Are you sure?
This action might not be possible to undo. Are you sure you want to continue?
Dissertation
zur Erlangung des akademischen Grades
Doktoringenieur
(Dr.Ing.)
von M.Sc. Mohamed Hassan Ahmed Mohamed
geb. am 15. July 1974 in Kairo,
¨
Agypten
genehmigt durch die Fakult¨ at f¨ ur Verfahrens und Systemtechnik
der OttovonGuerickeUniversit¨ at Magdeburg
Gutachter:
Prof. Dr.Ing. Dominique Th´evenin
Prof. Dr.Ing. Christian Oliver Paschereit
Prof. Dr.Ing. Zbigniew A. Styczynski
Promotionskolloquium am: 14.04.2011
I
I would like to dedicate this thesis
TO
MY PARENTS
TO
My wife DALIA and my children SAMA and AHMED
TO
My sisters SOMIA and ASMAA
AND TO
My brothers IBRAHIEM and AHMED
Acknowledgment
Firstly, I would like to express my thankfulness and gratitude to my country Egypt
for the ﬁnancial support during my research. Without that, I was not able to work and
search here in Germany.
I am greatly indebted to my supervisor Prof. Dr. Ing. Dominique Th´evenin. I
am extremely grateful for his support, invaluable guidance and for his continuing help
even before I came to Germany. I also wish to thank him for providing a wonderful
work atmosphere and facilities.
I am also grateful to Dr. Elem´er Pap for his valuable suggestions and help. I
am very grateful to Dr. G´abor Janiga for his great help. I acknowledge the eﬃcient
support of Matthias Lind concerning all experimental measurements.
I always feel lucky to be with so many excellent researchers. Thanks are due to all
colleagues of my institute, who were always quite helpful during my stay.
I would like to thank my colleague Dr. Hemdan Shalaby for his great help and
his valuable advices about the life in Magdeburg during starting time.
Finally, my sincere thanks go to my wife Dalia who oﬀered her invaluable support
to help me during this long education journey.
M. Mohamed
Magdeburg, Germany
1. October 2010
II
Abstract
Research and development activities in the ﬁeld of renewable energy have been
considerably increased in many countries recently, due to the worldwide energy crisis.
Wind energy is becoming particularly important. Although considerable progress
have already been achieved, the available technical design is not yet adequate to
develop reliable wind energy converters for conditions corresponding to low wind
speeds and urban areas. The Savonius turbine appears to be particularly promising
for such conditions, but suﬀers from a poor eﬃciency. The present study considers
improved designs in order to increase the output power of a classical Savonius turbine.
It aims at improving the output power of the Savonius turbine as well as its static
torque, which measures the selfstarting capability of the turbine. In order to achieve
both objectives, many designs have been investigated and optimized by placing in
an optimal manner an obstacle plate shielding the returning blade. The geometry
of the blade shape (skeleton line) has been optimized in presence of the obstacle
plate. Finally, frontal guiding plates have been considered and lead to a superior
performance of Savonius turbines. The optimization process is realized by coupling
an inhouse optimization library (OPAL, relying in the present case on Evolutionary
Algorithms) with an industrial ﬂow simulation code (ANSYSFluent). The target
function is the output power coeﬃcient. Compared to a standard Savonius turbine, a
relative increase of the power output coeﬃcient by 58% is ﬁnally obtained at design
point. The performance increases throughout the useful operating range. The static
torque is found to be positive at any angle, high enough to obtain selfstarting conditions.
Considering now ocean’s and sea’s energy, the Wells turbine is one of the technical
systems allowing an eﬃcient use of the power contained in waves with a relatively low
investment level. It consists of a selfrectifying air ﬂow turbine employed to convert
the pneumatic power of the air stream induced by an Oscillating Water Column into
mechanical energy. On the other hand, standard Wells turbines show several wellknown
disadvantages: a low tangential force, leading to a low power output from the turbine;
a high undesired axial force; usually a low aerodynamic eﬃciency and a limited range
of operation due to stall. In the present work an optimization process is employed
in order to increase the tangential force induced by a monoplane and twostage Wells
turbine using symmetric airfoil blades as well as by a twostage Wells turbine using
nonsymmetric airfoil blades. The automatic optimization procedure in this part of
the work is again carried out by coupling the inhouse optimization library OPAL with
IV
V
the industrial CFD code ANSYSFluent. This multiobjective optimization relying on
Evolutionary Algorithms takes into account both tangential force coeﬃcient and turbine
eﬃciency. Detailed comparisons are ﬁnally presented between the optimal designs and
the classical Wells turbine using symmetric airfoils, demonstrating the superiority of the
proposed solutions. The optimization of the airfoil shape lead to a considerably increased
power output (+12%) and simultaneously to an increase of eﬃciency throughout the
full operating range.
Zusammenfassung
Aufgrund der weltweiten Energiekrise wurden die Anstrengungen in Forschung und En
twicklung im Bereich der erneuerbaren Energien in den letzten Jahren in vielen L¨andern
erheblich erh¨oht. Dabei nimmt die Windenergie eine zunehmend wichtige Rolle ein.
Obwohl bereits erhebliche Fortschritte erzielt wurden, ist das zur Verf¨ ugung stehende
technische Design noch nicht ausreichend angepasst, um zuverl¨assige Windenergieanla
gen f¨ ur Bedingungen mit vergleichbar geringen Windgeschwindigkeiten und st¨adtischen
Gebieten zu entwickeln. Die SavoniusTurbine scheint besonders vielversprechend f¨ ur
solche Bedingungen zu sein, leidet aber unter einem schlechten Wirkungsgrad. Die vor
liegende Studie betrachtet verbesserte Designs, um die Ausgangsleistung einer klassis
chen Savonius Turbine zu erh¨ohen. Diese zielt sowohl auf die Verbesserung der Leistung
der SavoniusTurbine als auch auf die Steigerung des statischen Drehmoments, welches
f¨ ur die SelbstStartf¨ahigkeit der Turbine bestimmend ist. Um beide Ziele zu erreichen,
wurde eine Vielzahl von Entw¨ urfen untersucht und optimiert, wobei in optimaler Weise
ein Leitblech zur Abschirmung des r¨ uckkehrenden Turbinenblattes positioniert wurde.
Die Geometrie des Turbinenblattes (SkelettLinie) wurde in Anwesenheit des Leitblechs
optimiert. Dies f¨ uhrte schließlich zu einer gesteigerten Leistung der Savonius Turbine.
Die Optimierung wurde durch die Kopplung einer hauseigenen Optimierungsbibliothek
(OPAL im vorliegenden Fall auf Evolution¨ aren Algorithmen basierend) mit einem in
dustriellen Str¨omungssimulations Code (ANSYSFluent) realisiert. Hierbei ist der Aus
gangsleistungskoeﬃzient die Zielfunktion. Im Vergleich zu einer StandardSavonius Tur
bine ist eine relative Erh¨ohung des Ausgangsleistungskoeﬃzienten um 58% am Ausle
gungspunkt erreicht worden. Die Leistung steigt im gesamten Betriebsbereich. Das
statische Drehmoment erweist sich in jedem beliebigen Winkel positiv und ist hoch
genug, um die SelbstStartbedingung der Turbine zu erf¨ ullen.
Betrachtet man nun die Wasserkraft, so ist die WellsTurbine eines der technis
chen Systeme, welches eine eﬃziente Nutzung der Energie in Wellen unter Aufwendung
relativ geringer Investitionskosten erm¨oglicht. Diese besteht aus einer sich selbst richt
enden Luftstrom Turbine, welche die durch eine oszillierende Wassers¨ aule eingebrachte
pneumatische Energie des Luftstroms in mechanische Energie in Form von Rotation
umwandelt. Auf der anderen Seite zeigen Standard WellsTurbinen mehrere bekan
nte Nachteile: eine geringe Tangentialkraft, was zu einer geringen Leistung der Tur
bine f¨ uhrt, eine hohe unerw¨ unschte axiale Kraft, eine geringe aerodynamische Eﬃzienz
und einen begrenzten Betriebsbereich durch Str¨omungsabriss. In der vorliegenden Ar
beit wurde ein Optimierungsprozess eingesetzt, um die Tangentialkraft zu erh¨ohen,
VI
VII
welche durch eine zweistuﬁge WellsTurbine mit symmetrisch proﬁlierten Schaufeln,
sowie durch eine zweistuﬁge WellsTurbine mit nichtsymmetrisch proﬁlierten Schaufeln
induziert wird. Das automatische Optimierungsverfahren in diesem Teil der Arbeit wird
wiederum durch die Kopplung der hauseigenen Optimierungsbibliothek OPAL mit dem
industriellen CFDCode ANSYSFluent durchgef¨ uhrt. Diese Mehrzieloptimierung unter
Berufung auf Evolution¨ are Algorithmen ber¨ ucksichtigt sowohl den Tangentialkraftko
eﬃzienten als auch den Wirkungsgrad der Turbine. Abschließend werden detaillierte
Vergleiche zwischen den optimalen Designs und der klassischen WellsTurbine mit sym
metrischen Proﬁlen pr¨asentiert, welche die
¨
Uberlegenheit der vorgeschlagenen L¨osungen
veranschaulichen. Die Optimierung der Tragﬂ¨achenform f¨ uhrte zu einer beachtlich
gesteigerten Leistung (+12%) und gleichzeitig zu einer Erh¨ohung der Eﬃzienz ¨ uber
den gesamten Betriebsbereich.
Contents
Acknowledgment II
Abstract IV
Zusammenfassung VI
Index XII
Nomenclature XIV
List of Tables XVII
List of Figures XXIV
1 Introduction 2
1.1 Renewable energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.1 Some properties of renewable energy . . . . . . . . . . . . . . . . 4
1.1.2 A few numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.1.3 Future of renewable energy . . . . . . . . . . . . . . . . . . . . . . 6
1.2 Scope of the Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2.1 Conclusions and outline . . . . . . . . . . . . . . . . . . . . . . . 8
2 Basic concepts 10
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2 Wind energy conversion . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.1 Horizontal axis turbines . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.2 Vertical axis turbines . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2.3 Betz limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2.3.1 Betz’s momentum theory . . . . . . . . . . . . . . . . . 16
2.2.4 Principle operation of standard Savonius turbines . . . . . . . . . 19
2.2.5 Performance of a Savonius turbine . . . . . . . . . . . . . . . . . . 20
2.3 Wave energy conversion . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.3.1 Wave energy potential . . . . . . . . . . . . . . . . . . . . . . . . 22
2.3.2 Oscillating Water Column (OWC) principle . . . . . . . . . . . . 23
2.3.3 Wells turbines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
VIII
CONTENTS IX
2.3.4 Operation of Wells turbines . . . . . . . . . . . . . . . . . . . . . 25
2.3.5 Wells turbine performance . . . . . . . . . . . . . . . . . . . . . . 25
2.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3 Literature Review 28
3.1 Savonius turbine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.1.1 Experimental investigations of Savonius turbines . . . . . . . . . . 28
3.1.2 Numerical investigations of Savonius turbines . . . . . . . . . . . 31
3.1.3 Methods to improve Savonius turbine performance . . . . . . . . 33
3.1.3.1 Deﬂector plate . . . . . . . . . . . . . . . . . . . . . . . 33
3.1.3.2 Double and three steps Savonius rotor . . . . . . . . . . 33
3.1.3.3 Guide vanes . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.1.3.4 Twistedblade Savonius rotor . . . . . . . . . . . . . . . 34
3.1.3.5 Savonius rotor using a guidebox tunnel . . . . . . . . . 35
3.1.3.6 Modiﬁed Savonius rotor . . . . . . . . . . . . . . . . . . 36
3.1.4 Summary of Savonius turbine review . . . . . . . . . . . . . . . . 37
3.2 Wells Turbine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.2.1 Performance parameters of Wells turbine . . . . . . . . . . . . . 39
3.2.1.1 Solidity . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.2.1.2 Hubtotip ratio . . . . . . . . . . . . . . . . . . . . . . 40
3.2.1.3 Aspect ratio . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.2.1.4 Reynolds number . . . . . . . . . . . . . . . . . . . . . 41
3.2.1.5 Tip clearance . . . . . . . . . . . . . . . . . . . . . . . . 41
3.2.1.6 Inlet turbulence . . . . . . . . . . . . . . . . . . . . . . . 42
3.2.1.7 Blade shape . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.2.2 Unsteady ﬂow eﬀect on the performance of Wells turbine . . . . . 43
3.2.3 Methods to improve Wells turbine performance . . . . . . . . . . 44
3.2.3.1 Guide vanes installation . . . . . . . . . . . . . . . . . . 45
3.2.3.2 Selfpitchcontrolled blades . . . . . . . . . . . . . . . . 45
3.2.3.3 Lean blade (blade swept) . . . . . . . . . . . . . . . . . 47
3.2.3.4 End plate . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.2.4 Multiplane Wells turbine . . . . . . . . . . . . . . . . . . . . . . 48
3.2.4.1 Twostage Wells turbine (biplane turbine) . . . . . . . . 48
3.2.4.2 Twostage contrarotating Wells turbine . . . . . . . . . 49
3.2.5 Summary of Wells turbine review . . . . . . . . . . . . . . . . . . 52
3.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4 Optimization 54
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.2 Optimization uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.3 How can we achieve optimization? . . . . . . . . . . . . . . . . . . . . . . 55
4.4 Structure of optimization problems . . . . . . . . . . . . . . . . . . . . . 57
4.5 Types of mathematical programming . . . . . . . . . . . . . . . . . . . . 58
4.5.1 Linear programming (LP) . . . . . . . . . . . . . . . . . . . . . . 58
CONTENTS X
4.5.2 Integer programming (IP) . . . . . . . . . . . . . . . . . . . . . . 58
4.5.3 Quadratic programming (QP) . . . . . . . . . . . . . . . . . . . . 58
4.5.4 Nonlinear programming (NP) . . . . . . . . . . . . . . . . . . . . 59
4.5.5 Dynamic programming (DP) . . . . . . . . . . . . . . . . . . . . . 59
4.6 Requirements for optimization . . . . . . . . . . . . . . . . . . . . . . . 59
4.6.1 Deﬁning the system boundaries . . . . . . . . . . . . . . . . . . . 59
4.6.2 Performance criterion . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.6.3 Independent variables . . . . . . . . . . . . . . . . . . . . . . . . 60
4.6.4 System model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.7 Optimization methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.8 Evolutionary Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.8.1 Genetic Algorithm (GA) . . . . . . . . . . . . . . . . . . . . . . . 62
4.8.2 Advantage and disadvantage of Genetic Algorithm . . . . . . . . . 65
4.9 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
5 Numerical methods and algorithms 68
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.2 CFD & Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.3 Computational procedure . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5.3.1 Preprocess: geometry & grid generation . . . . . . . . . . . . . . 69
5.3.1.1 Savonius turbine: size of computational domain . . . . . 69
5.3.1.2 Savonius turbine mesh independence . . . . . . . . . . . 70
5.3.1.3 Wells turbine: size of computational domain . . . . . . . 70
5.3.1.4 Wells turbine mesh independence . . . . . . . . . . . . . 70
5.3.2 Numerical solution of the ﬂow ﬁeld . . . . . . . . . . . . . . . . . 72
5.3.2.1 Why Fluent? . . . . . . . . . . . . . . . . . . . . . . . . 72
5.3.2.2 Model validation and selection for Savonius turbine . . . 72
5.3.2.3 Model validation and selection for Wells turbine . . . . . 74
5.3.2.4 Realizable k − turbulence model . . . . . . . . . . . . . 75
5.3.3 PostProcessing: analysis of results . . . . . . . . . . . . . . . . . 77
5.3.3.1 Moment calculation for Savonius turbines . . . . . . . . 77
5.3.3.2 Forces calculation for Wells turbines . . . . . . . . . . . 77
5.4 CFD/Optimization coupling . . . . . . . . . . . . . . . . . . . . . . . . . 78
5.5 Optimization parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.5.1 Savonius turbine: single objective optimization . . . . . . . . . . . 80
5.5.2 Wells turbine: multiobjective concurrent optimization . . . . . . 81
5.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
6 Savonius turbine: singleobjective optimization 82
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
6.2 Optimal GWturbine: modiﬁed threeblade Savonius turbine without gap 82
6.2.1 Performance of the original GWturbine . . . . . . . . . . . . . . 82
6.2.2 Reducing the drag on the returning blade . . . . . . . . . . . . . 84
6.2.3 Inﬂuence of a rounded obstacle plate . . . . . . . . . . . . . . . . 86
CONTENTS XI
6.2.4 Optimization of the blade shape . . . . . . . . . . . . . . . . . . . 87
6.2.5 Conclusions on the GWturbine . . . . . . . . . . . . . . . . . . . 90
6.3 Optimal threeblade Savonius turbine . . . . . . . . . . . . . . . . . . . . 91
6.3.1 Inﬂuence of obstacle plate . . . . . . . . . . . . . . . . . . . . . . 91
6.3.1.1 Selfstarting capability . . . . . . . . . . . . . . . . . . . 92
6.3.1.2 Optimization of the obstacle position . . . . . . . . . . . 92
6.3.2 Optimal blade shape . . . . . . . . . . . . . . . . . . . . . . . . . 96
6.3.3 Optimal threeblade Savonius turbine with guiding plates . . . . 99
6.3.3.1 Selfstarting capability . . . . . . . . . . . . . . . . . . . 104
6.3.4 Conclusions on threeblade design . . . . . . . . . . . . . . . . . . 105
6.4 Optimal twoblade Savonius turbine . . . . . . . . . . . . . . . . . . . . . 105
6.4.1 Obstacle plate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
6.4.1.1 Selfstarting capability . . . . . . . . . . . . . . . . . . 106
6.4.1.2 Optimization . . . . . . . . . . . . . . . . . . . . . . . . 106
6.4.1.3 Oﬀ design performance . . . . . . . . . . . . . . . . . . 108
6.4.1.4 Practical realization . . . . . . . . . . . . . . . . . . . . 109
6.4.2 Optimal blade shape with obstacle plate . . . . . . . . . . . . . . 109
6.4.2.1 Optimization . . . . . . . . . . . . . . . . . . . . . . . . 109
6.4.2.2 Oﬀ design performance . . . . . . . . . . . . . . . . . . . 112
6.4.2.3 Selfstarting capability . . . . . . . . . . . . . . . . . . . 113
6.4.3 Optimal Savonius turbine with two guiding plates . . . . . . . . 113
6.4.3.1 Optimization . . . . . . . . . . . . . . . . . . . . . . . . 114
6.4.3.2 Oﬀ design performance . . . . . . . . . . . . . . . . . . 116
6.4.3.3 Proﬁled guiding plates . . . . . . . . . . . . . . . . . . . 117
6.5 Final optimization of Savonius turbine . . . . . . . . . . . . . . . . . . . 119
6.5.1 Oﬀ design performance . . . . . . . . . . . . . . . . . . . . . . . 124
6.5.2 Selfstarting capability . . . . . . . . . . . . . . . . . . . . . . . . 124
6.6 Preliminary experimental tests in windtunnel . . . . . . . . . . . . . . . 126
6.7 Conclusions on Savonius turbine . . . . . . . . . . . . . . . . . . . . . . . 127
7 Wells turbine: Concurrent optimization 130
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
7.2 Optimal monoplane Wells turbine . . . . . . . . . . . . . . . . . . . . . . 130
7.2.1 Optimization of airfoil shape . . . . . . . . . . . . . . . . . . . . . 132
7.2.2 Oﬀ design performance . . . . . . . . . . . . . . . . . . . . . . . 137
7.3 Optimal twostage Wells turbine with nonsymmetric airfoils . . . . . . . 137
7.3.1 Optimal airfoil shape without mutual interactions . . . . . . . . . 138
7.3.2 Optimal airfoil shape with mutual interactions . . . . . . . . . . . 140
7.3.2.1 Mutual interactions eﬀect . . . . . . . . . . . . . . . . . 140
7.3.2.2 Optimization . . . . . . . . . . . . . . . . . . . . . . . . 142
7.3.2.3 Oﬀ design performance . . . . . . . . . . . . . . . . . . . 145
7.4 Optimal twostage Wells turbine with symmetric airfoils . . . . . . . . . 148
7.4.1 Optimization of airfoil shape . . . . . . . . . . . . . . . . . . . . . 149
7.4.1.1 Oﬀ design performance . . . . . . . . . . . . . . . . . . 150
CONTENTS XII
7.5 Conclusions on Wells turbine . . . . . . . . . . . . . . . . . . . . . . . . . 152
8 Conclusions and Outlook 156
8.1 Suggestions for further research . . . . . . . . . . . . . . . . . . . . . . . 157
Bibliography 157
Curriculum Vitae 169
Related Publications 171
Nomenclature
Roman symbols
A Projected area of rotor (DH), m
2
B Rotor pitch of Wells turbine, m
b Blade span of Wells turbine, m
c Airfoil blade chord, m
C
A
Total pressure diﬀerence coeﬃcient
C
p
Power coeﬃcient (P/[1/2ρAU
3
])
C
m
Torque coeﬃcient (T/[ρR
2
HU
2
])
C
ms
Static torque coeﬃcient (T
s
/[ρR
2
HU
2
])
C
T
Tangential force coeﬃcient
D Turbine diameter of Savonius turbine (2R), m
D
t
Turbine diameter of Wells turbine, m
d Blade chord (2r) of Savonius turbine, m
∆p
0
Total pressure diﬀerence, Pa
F
D
Drag force, N
F
L
Lift force, N
F
T
Tangential force, N
F
X
Axial force, N
f
∗
Wave frequency, Hz
H Blade height, m
h Hub to tip ratio for Wells turbine
G Gap ratio
g
w
Gap width for the threeblade Savonius turbine, m
L
d
Deﬂector length, m
L
o
Obstacle length, m
N Rotational speed of rotor, rpm
P Output power (2πNT/60), W
Q Volumetric ﬂowrate, m
3
/s
XIV
CONTENTS XV
R Tip radius of Savonius turbine, m
r
t
Tip blade radius, m
r
h
Hub blade radius, m
r Blade radius of semicylindrical Savonius blade, m
s Blade solidity
T Output torque, Nm
T
s
Static torque quantifying selfstarting capability, Nm
t Airfoil max. half thickness, m
U Mean wind velocity in axial direction, m/s
u
t
Tip blade speed of Wells turbine, m/s
v
A
Axial air velocity, m/s
w Relative velocity, m/s
z Number of blades of Wells turbine
Greek symbols
α Angle of incidence, (
◦
)
β Obstacle angle, (
◦
)
η Aerodynamic eﬃcieny
γ Deﬂector angle, (
◦
)
γ
b
Setting blade angle, (
◦
)
φ Flow coeﬃcient
ρ Density, kg/m
3
θ Rotor angle, (
◦
)
ω Angular speed, 1/s
List of Tables
3.1 Summary of Savonius turbine main modiﬁcations . . . . . . . . . . . . . 37
3.2 Main modiﬁcations allowing to improve the performance of Wells turbines 52
5.1 Parameters of the Genetic Algorithm . . . . . . . . . . . . . . . . . . . . 81
6.1 Acceptable range for the input parameters (parameter space) . . . . . . . 93
6.2 Optimal conﬁgurations . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
6.3 Acceptable range for the input parameters (parameter space for blade
shape with the obstacle ) . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
6.4 Optimal blade shape with the obstacle . . . . . . . . . . . . . . . . . . . 98
6.5 Acceptable range for the input parameters (parameter space) . . . . . . . 102
6.6 Optimal conﬁguration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
6.7 Acceptable range for the input parameters . . . . . . . . . . . . . . . . . 107
6.8 Optimal conﬁgurations (obstacle position and angle) . . . . . . . . . . . 108
6.9 Acceptable range for the input parameters for the blade shape . . . . . . 111
6.10 Optimal conﬁguration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
6.11 Acceptable range for the input parameters (parameter space) . . . . . . . 116
6.12 Optimum parameters of guiding plates . . . . . . . . . . . . . . . . . . . 117
6.13 Acceptable range for the input parameters (parameter space) . . . . . . . 119
6.14 Acceptable range for the input parameters (parameter space) . . . . . . . 121
6.15 Optimum conﬁgurations . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
7.1 Parameter space for the moving points P
2
to P
12
for monoplane Wells
turbine. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
7.2 Optimum shape parameters for monoplane Wells turbine . . . . . . . . . 136
7.3 Polynomial coeﬃcients of optimal airfoil shape (best ﬁt) . . . . . . . . . 136
7.4 Parameter space for the moving points P
1
to P
16
for upper face and P
18
to P
33
for lower face . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
7.5 Optimum shape parameters for twostage Wells turbine with non
symmetric airfoils . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
7.6 Polynomial coeﬃcients of optimal airfoil shape (best ﬁt) for twostage
nonsymmetric airfoil Wells turbine considering mutual interaction be
tween the blades (upper and lower face) . . . . . . . . . . . . . . . . . . 147
7.7 Optimum shape parameters for twostage Wells turbine . . . . . . . . . 152
XVI
LIST OF TABLES XVII
7.8 Polynomial coeﬃcients of optimal airfoil shape (best ﬁt) for twostage
Wells turbine with symmetric airfoils . . . . . . . . . . . . . . . . . . . . 152
List of Figures
1.1 2008 worldwide renewable energy sources. Source: [12]. . . . . . . . . . . 3
1.2 2008 worldwide renewable energy sources [26]. . . . . . . . . . . . . . . . 7
2.1 Distribution of yearly annual mean values of the wind speed for 10 m
above ground worldwide [53]. . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Horizontal axis wind turbine. . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3 Flow velocities and aerodynamic forces at the airfoil crosssection of a
blade element [53]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.4 Flow conditions and drag force for vertical axis turbines [53]. . . . . . . . 15
2.5 Flow conditions due to the extraction of mechanical energy from a free
stream air ﬂow, according to the elementary momentum theory. . . . . . 17
2.6 Power coeﬃcients of wind rotors of diﬀerent designs [37]. . . . . . . . . . 19
2.7 Conventional Savonius rotor. . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.8 Schematic description and main parameters characterizing a Savonius rotor. 21
2.9 Approximate global distribution of wave power levels in kW/m of wave
front [133]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.10 The wave motion in the OWC device [73]. . . . . . . . . . . . . . . . . . 24
2.11 Wells turbine in state of upward and downward air ﬂow. . . . . . . . . . 25
2.12 Axial and tangential forces acting on a Wells turbine. . . . . . . . . . . . 26
2.13 Main geometrical parameters of a Wells turbine. . . . . . . . . . . . . . . 27
3.1 Flow in and around a nonrotating Savonius rotor (λ = 0); a: visualized
ﬂow ﬁeld; b: ﬂow inside the rotor; c: ﬂow model; d: surface pressure
distribution [29]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.2 Flow in and around a Savonius rotor in rotation (λ = 0.9); a: visualized
ﬂow ﬁeld; b: ﬂow inside the rotor; c: ﬂow model; d: surface pressure
distribution [29]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.3 Static torque coeﬃcient of a Savonius rotor (Re = 1.56 10
5
) [68]. . . . . . 32
3.4 Dynamic torque coeﬃcient of a Savonius rotor (Re = 1.56 10
5
, λ = 1) [68]. 32
3.5 Doublestep Savonius rotor [68]. . . . . . . . . . . . . . . . . . . . . . . . 34
3.6 Twistedblade Savonius Rotor [98]. . . . . . . . . . . . . . . . . . . . . . 35
3.7 Power coeﬃcient for twistedblade Savonius rotor [98]. . . . . . . . . . . 35
3.8 Savonius rotor using a guidebox tunnel [48]. . . . . . . . . . . . . . . . . 36
3.9 Power coeﬃcient of Savonius rotor using a guidebox tunnel [48]. . . . . . 37
XVIII
LIST OF FIGURES XIX
3.10 Modiﬁed Savonius Rotor [69]. . . . . . . . . . . . . . . . . . . . . . . . . 38
3.11 Static torque for the modiﬁed Savonius rotor (θ = 90
◦
, β = 45
◦
, e/d =
1/6) [69]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.12 Savonius turbine publication statistics in international journals and con
ferences. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.13 Axial and tangential forces acting on a Wells turbine. . . . . . . . . . . . 40
3.14 Solidity eﬀect on the eﬃciency [82, 92, 97]. . . . . . . . . . . . . . . . . . 41
3.15 Aspect ratio eﬀect on the eﬃciency [82, 92, 97]. . . . . . . . . . . . . . . 42
3.16 Eﬀect of unsteady ﬂow on Wells turbine performance [103]. . . . . . . . . 44
3.17 Eﬀect of solidity on the hysteretic behavior [103]. . . . . . . . . . . . . . 45
3.18 Illustration of the ﬂow structure in blade suction side [103]. . . . . . . . . 45
3.19 Installation of guide vanes [106]. . . . . . . . . . . . . . . . . . . . . . . . 46
3.20 Guide vanes eﬀect on the turbine performance [105]. . . . . . . . . . . . . 46
3.21 Principle of using selfpitchcontrolled blades [106]. . . . . . . . . . . . . 47
3.22 Swept blade (lean blade) [8]. . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.23 NACA 0015 and HSIM 152621231576 blades [8]. . . . . . . . . . . . . . 48
3.24 Rotor blade with end plate: (a) Middle type; (b) Forward type; (c) Back
ward type [124]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.25 Two stage (biplane) Wells turbine. . . . . . . . . . . . . . . . . . . . . . 50
3.26 Contrarotating Wells turbine [107]. . . . . . . . . . . . . . . . . . . . . . 50
3.27 Torque coeﬃcients for both LIMPET Rotors during intake and exhaust [24]. 51
3.28 Comparison of LIMPET and constant ﬂow model turbine eﬃciencies [24]. 51
3.29 Wells turbine publication statistics in international journals and confer
ences. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.1 Flowchart of a Genetic Algorithm. . . . . . . . . . . . . . . . . . . . . . . 63
4.2 Example of selection based on the roulettewheel for ten individuals. . . . 64
4.3 Example of crossover. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.1 Size of the computational domain and impact on the torque coeﬃcient . 70
5.2 Gridindependence study for the torque coeﬃcient . . . . . . . . . . . . 71
5.3 Size of the computational domain around monoplane and twostage Wells
turbines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.4 Gridindependence study for the tangential force coeﬃcient . . . . . . . . 72
5.5 Validation of computational model: a) torque coeﬃcient, b) power coef
ﬁcient, both compared to published experimental results for a twoblade
conventional Savonius turbine [38] . . . . . . . . . . . . . . . . . . . . . . 73
5.6 Validation of computational model: power coeﬃcient compared to exper
imental results for a threeblade Savonius turbine [48] . . . . . . . . . . . 74
5.7 Inﬂuence of the turbulence model on the tangential force coeﬃcient, com
pared to experimental results for a monoplane Wells turbine [107] . . . . 75
5.8 Inﬂuence of the number of revolutions on the instantaneous and on the
average power coeﬃcient C
p
computed by CFD for the optimum design
shown later. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
LIST OF FIGURES XX
5.9 Schematic description of optimizer (OPAL) and CFD code coupling. . . . 79
6.1 Schematic shape of the GWturbine . . . . . . . . . . . . . . . . . . . . . 83
6.2 Performance comparison between the GWturbine and the conventional,
twoblade Savonius rotor. . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
6.3 Instantaneous velocity vectors around the GWturbine . . . . . . . . . . 84
6.4 Schematic description of the GWturbine with open returning blade. . . . 84
6.5 Performance comparison between the GWturbine with or without open
returning blade. The performance of the conventional Savonius rotor
is also shown for comparison. Top: torque coeﬃcient. Bottom: power
coeﬃcient. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
6.6 Performance of open returning blade turbine for diﬀerent slit angles. Top:
torque coeﬃcient. Bottom: power coeﬃcient. . . . . . . . . . . . . . . . . 86
6.7 Schematic description of the counterrotating GWturbine with rounded
deﬂector. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
6.8 Performance comparison between the new concept in free ﬁeld or placed
behind a rounded deﬂector. . . . . . . . . . . . . . . . . . . . . . . . . . 88
6.9 Description of the blade shape with 3 discrete points connected by splines. 89
6.10 The two input parameters of the optimization and the power coeﬃcient. 89
6.11 The best (right) conﬁguration obtained during the optimization compared
to the original GWturbine (semicylindrical shape: left). . . . . . . . . . 89
6.12 Power coeﬃcient of the optimized conﬁguration compared to the GW
turbine as a function of λ. . . . . . . . . . . . . . . . . . . . . . . . . . . 90
6.13 Schematic description and main parameters characterizing a conventional
Savonius rotor with three blades. . . . . . . . . . . . . . . . . . . . . . . 91
6.14 Static torque coeﬃcient as a function of the rotor angle θ for three diﬀer
ent values of Y
1
choosing X
1
/R = −1.4 and X
2
/R = −1.76. . . . . . . . . 92
6.15 Schematic description of the geometry and free optimization parameters
X
1
, Y
1
, X
2
used to modify the position of the shielding obstacle. . . . . . 93
6.16 Input parameters of the optimization and power coeﬃcient represented
using parallel coordinates. The parameters of the optimal conﬁgurations
are connected with a thick red line. The power coeﬃcient of the conven
tional threeblade Savonius turbine is shown with a gray circle. . . . . . . 94
6.17 Performance of the optimized conﬁguration (red line) compared to the
conventional threeblade Savonius turbine without obstacle (black): a:
torque coeﬃcient; b: power coeﬃcient. The corresponding relative in
crease is shown with blue line. . . . . . . . . . . . . . . . . . . . . . . . . 95
6.18 Schematic description of the free optimization parameters
X
P1
, Y
P1
, X
P2
, Y
P2
, X
P3
and Y
P3
used to modify the blade shape. . . . . . 96
6.19 Input parameters of the optimization and power coeﬃcient represented
using parallel coordinates. The parameters of the optimal conﬁguration
are connected with a thick red line. The power coeﬃcient of the conven
tional threeblade turbine (semicylindrical shape) is also shown with a
black circle. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
LIST OF FIGURES XXI
6.20 Optimum conﬁguration obtained with the optimization procedure. . . . . 98
6.21 Performance of the optimized conﬁguration (red line) compared to the
conventional threeblade Savonius turbine (black line): a) torque coeﬃ
cient; b) power coeﬃcient. The corresponding relative increase compared
to the conventional threeblade conﬁguration is shown with blue line. . . 99
6.22 Schematic description of the free optimization parameters char
acterizing a threeblade Savonius rotor : a) plate parameters
(X
1
, Y
1
, X
2
, Y
2
, X
d1
, Y
d1
, X
d2
and Y
d2
); b) X
P1
, Y
P1
, X
P2
, Y
P2
, X
P3
and Y
P3
used to modify the blade shape; additionally, the gap width g
w
. . . . . . 100
6.23 Input parameters of the optimization and power coeﬃcient represented
using parallel coordinates. The parameters of the optimal conﬁguration
are connected with a thick red line. The power coeﬃcient of the conven
tional threeblade turbine (semicylindrical shape) is also shown with a
black circle. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
6.24 Optimum conﬁguration obtained with the optimization procedure. . . . . 102
6.25 Instantaneous velocity vectors magnitude (m/s) around the optimum con
ﬁguration (zoom) at the design point (λ = 0.7). . . . . . . . . . . . . . . 103
6.26 Performance of the optimized conﬁguration (red line) compared to the
conventional threeblade Savonius turbine (black line): Top: torque co
eﬃcient; Bottom: power coeﬃcient. The corresponding relative increase
compared to the standard conﬁguration is shown with blue line. . . . . . 104
6.27 Static torque coeﬃcient C
ms
as a function of the ﬁxed rotor angle θ for
the optimal design (ﬁlled red squares) compared to the classical three
blade Savonius turbine (blue plus). The experimental results of [48] are
also shown for comparison (empty black squares). . . . . . . . . . . . . . 105
6.28 Schematic description of the geometry and free optimization parameters
X
1
, Y
1
, X
2
used to modify the position of the shielding obstacle. . . . . . 106
6.29 Static torque coeﬃcient as a function of the rotor angle θ for three diﬀer
ent values of Y
1
choosing X
1
/R = −1.4 and X
2
/R = −1.76. . . . . . . . . 107
6.30 Input parameters of the optimization and power coeﬃcient represented
using parallel coordinates. The parameters of the optimal conﬁgurations
are connected with a thick red line. The power coeﬃcient of the conven
tional turbine is shown with a black circle. . . . . . . . . . . . . . . . . . 108
6.31 Instantaneous ﬂow structure when the advancing blade is in vertical po
sition for the optimal conﬁguration at λ = 0.7: Zoom on the vicinity of
the turbine (full CFD domain is much larger). . . . . . . . . . . . . . . . 109
6.32 Performance of the optimized conﬁguration (red line) compared to the
conventional twoblade Savonius turbine without obstacle (black line):
a: torque coeﬃcient; b: power coeﬃcient. The corresponding relative
increase is shown with blue stars. . . . . . . . . . . . . . . . . . . . . . . 110
6.33 Schematic description of the free optimization parameters
X
P1
, Y
P1
, X
P2
, Y
P2
, X
P3
and Y
P3
used to modify the blade shape. . . . . . 111
LIST OF FIGURES XXII
6.34 Input parameters of the optimization and power coeﬃcient represented
using parallel coordinates. The parameters of the optimal shape are con
nected with a thick red line. The power coeﬃcient of the conventional
turbine is shown with a gray circle. The power coeﬃcient of the conven
tional turbine (semicylindrical shape) with obstacle plate is also shown
with a black circle. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
6.35 Optimum conﬁguration (right) obtained with the optimization procedure
compared to the classical Savonius turbine (semicylindrical shape: left). 113
6.36 Performance of the optimized conﬁguration (red line) compared to the
conventional Savonius turbine with and without obstacle plate (blue and
black, respectively): a: torque coeﬃcient; b: power coeﬃcient. The
corresponding relative increase compared to the standard conﬁguration
is shown with green line. . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
6.37 Static torque coeﬃcient C
ms
as a function of the ﬁxed rotor angle θ for
the optimal conﬁguration compared to the standard Savonius turbine
with and without obstacle plate. For this last case, the experimental
results of [38] are also shown for comparison. . . . . . . . . . . . . . . . . 115
6.38 Schematic description of the geometry of Savonius turbine with frontal
guiding plates. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
6.39 Input parameters of the optimization and power coeﬃcient represented
using parallel coordinates. The parameters of the optimal design are
connected with a thick red line. The power coeﬃcient of the conventional
turbine is shown with a black circle. . . . . . . . . . . . . . . . . . . . . . 116
6.40 Optimum conﬁguration of guiding plates. . . . . . . . . . . . . . . . . . . 117
6.41 Performance of the optimized conﬁguration (green line) compared to the
conventional Savonius turbine (blue line): a: torque coeﬃcient; b: power
coeﬃcient. The corresponding relative increase compared to the standard
conﬁguration is shown with black line. . . . . . . . . . . . . . . . . . . . 118
6.42 Schematic description of the geometry of Savonius turbine with proﬁled
guiding plates. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
6.43 Input parameters of the optimization and power coeﬃcient represented
using parallel coordinates. The parameters of the optimal designs are
connected with a thick red line. The power coeﬃcient of the conventional
turbine is shown with a black circle. . . . . . . . . . . . . . . . . . . . . . 120
6.44 Optimum conﬁgurations of curved guiding plates. . . . . . . . . . . . . . 121
6.45 Schematic description of the geometry and free optimization parameters. 122
6.46 Input parameters of the optimization and power coeﬃcient represented
using parallel coordinates. The parameters of the optimal design are
connected with a thick red line. The power coeﬃcient of the conventional
turbine is shown with a black circle. . . . . . . . . . . . . . . . . . . . . . 123
6.47 Optimum design of Savonius turbine with guiding plates. . . . . . . . . . 124
LIST OF FIGURES XXIII
6.48 Instantaneous ﬂow ﬁelds around optimum conﬁgurations (zoom) at the
design point (λ = 0.7), static pressure (Pa), velocity and velocity vector
magnitudes (m/s); a) classical Savonius with optimal guiding plates, b)
optimal Savonius with optimal guiding plates. Note that the color scales
are identical to facilitate comparisons. . . . . . . . . . . . . . . . . . . . . 125
6.49 Performance of the optimized conﬁguration (red line) compared to the
conventional Savonius turbine (black line): a: torque coeﬃcient; b: power
coeﬃcient. The corresponding relative increase compared to the classical
conﬁguration is shown with blue line. . . . . . . . . . . . . . . . . . . . . 126
6.50 Static torque coeﬃcient C
ms
as a function of the ﬁxed rotor angle θ for
the optimal conﬁguration compared to the conventional Savonius turbine.
For this last case, the experimental results of [38] are also shown for
comparison. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
6.51 Model installation with open wind tunnel. . . . . . . . . . . . . . . . . . 127
6.52 Selected experimental results [64] a) conventional Savonius turbine; b)
Optimized design. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
6.53 Successive optimization steps for the twoblade Savonius turbine. . . . . 129
7.1 Standard airfoils NACA 0015 and NACA 0021. . . . . . . . . . . . . . . 131
7.2 Axial and tangential forces acting on a Wells turbine. . . . . . . . . . . . 131
7.3 Allowed parameter space for the moving points P
2
to P
12
. . . . . . . . . . 132
7.4 Objectives of the optimization; a) for all computed conﬁgurations b) for
the best conﬁgurations (i.e., zoom on the upperright part (red square)
of a). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
7.5 Input parameters of the optimization and objectives represented using
parallel coordinates. The parameters of the optimal shape are connected
with a thick red line. The standard design (NACA 0021) is shown with a
thick dashed blue line; a) Xcoordinates of the variable points (P
2
. . . P
12
);
b) Ycoordinates of the variable points (P
2
. . . P
12
). . . . . . . . . . . . . 135
7.6 Comparison between the original proﬁle NACA 0021 (solid line), the op
timal airfoil shape described by splines (black squares showing the posi
tion of the control points) and the corresponding polynomial ﬁt (Eq. 7.3,
dashed line). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
7.7 Performance of the spline optimal conﬁguration (red line), ﬁtting optimal
one (black cross) compared to the conventional Wells turbine relying on
the NACA 0021 proﬁle (green line). The corresponding relative increase
is shown with blue line; a) tangential force coeﬃcient; b) eﬃciency. . . . 138
7.8 Twostage Wells turbine based on nonsymmetric airfoils . . . . . . . . . 139
7.9 Comparison between the optimal shape of the airfoil and the original
proﬁle NACA 2421 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
7.10 Impact of mutual interaction between blades in the same plane on a)
tangential force (left) and b) axial force (right), as a function of the solidity.141
LIST OF FIGURES XXIV
7.11 Impact of solidity on the twostage Wells turbine performance considering
the tangential force coeﬃcient (a: left ﬁgure) and the turbine eﬃciency
(b: right ﬁgure). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
7.12 Projected shape of the turbine; a) Conventional turbine b) Suggestion of
isoperformance turbine. . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
7.13 Allowed parameter space for the moving points. . . . . . . . . . . . . . . 144
7.14 Objectives of the optimization: a) for all computed conﬁgurations; b) for
the best conﬁgurations, i.e., zoom on the upperright part (red square) of
(a). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
7.15 Input parameters of the optimization and objectives represented using
parallel coordinates. The parameters of the optimal shape are connected
with a thick red line. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
7.16 Comparison between the original proﬁle NACA 2421 (solid line) and the
optimal airfoil shape (dashed line), considering mutual interaction be
tween the blades. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
7.17 Performance of the optimal conﬁguration (red line), compared to the
nonsymmetric twostage Wells turbine relying on the NACA 2421 proﬁle
(black line). The corresponding relative increase is shown with blue line;
a) tangential force coeﬃcient; b) eﬃciency. . . . . . . . . . . . . . . . . . 148
7.18 Twostage Wells turbine with symmetric airfoils NACA 0021. . . . . . . . 149
7.19 Allowed parameter space for the moving points. . . . . . . . . . . . . . . 149
7.20 Objectives of the optimization: a) for all computed conﬁgurations; b) for
the best conﬁgurations, i.e., zoom on the upperright part (red circle) of
(a). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
7.21 Input parameters of the optimization and objectives represented using
parallel coordinates. The parameters of the optimal shape are connected
with a thick red line. The standard design (NACA 0021) is shown with
a thick dashed blue line. . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
7.22 Comparison between the original proﬁle NACA 0021 (solid line), the poly
nomial optimal airfoil shape (red line) and optimal shape by splines (black
square). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
7.23 Performance of the optimal conﬁguration (red line), compared to the
conventional twostage Wells turbine relying on the NACA 0021 proﬁle
(gray squares). The corresponding relative increase is shown with blue
line; a) tangential force coeﬃcient; b) eﬃciency. . . . . . . . . . . . . . . 153
7.24 Successive optimization steps for the monoplane and twostage Wells tur
bines. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
8.1 Power coeﬃcients of wind rotors of diﬀerent designs including newly de
veloped, optimal Savonius turbine. . . . . . . . . . . . . . . . . . . . . . 157
Chapter 1
Introduction
Life is nothing but a continuous process of energy conversion and transformation. The
accomplishments of civilization have largely been achieved through the increasingly ef
ﬁcient and extensive harnessing of various forms of energy to extend human capabilities
and ingenuity. Energy is similarly indispensable for continued human development and
economic growth. Providing adequate, aﬀordable energy is essential for eradicating
poverty, improving human welfare, and raising living standards worldwide. And with
out economic growth, it will be diﬃcult to address environmental challenges, especially
those associated with poverty. But energy production, conversion, and use always gener
ate undesirable byproducts and emissions at least in the form of dissipated heat. Energy
cannot be created or destroyed, but it can be converted from one form to another. Al
though it is common to discuss energy consumption, energy is actually transformed
rather than consumed. What is consumed is the ability of oil, gas, coal, biomass, or
wind to produce useful work. In this thesis, only conversion of selected renewable energy
sources has been investigated.
1.1 Renewable energy
Unlike fossil fuels, which are exhaustible, renewable energy sources regenerate and can
be sustained indeﬁnitely. The ﬁve renewable sources used most often are: Biomass
(including wood and wood waste, municipal solid waste, landﬁll gas, biogas, ethanol,
and biodiesel), Hydropower (including tidal, water potential and ocean wave energy),
Geothermal, Wind and Solar energy.
The use of renewable energy is not new. More than 150 years ago, wood, which is
one form of biomass, supplied up to 90% of our energy needs. Today, we are looking
again at renewable sources to ﬁnd new ways to use them to help meet our energy needs
(see Fig. 1.1).
In 2006, about 18% of global ﬁnal energy consumption came from renewable energies,
with 13% coming from traditional biomass, which is mainly used for heating, and 3%
from hydroelectricity. New renewables (small hydropower installations, modern biomass,
wind, solar, geothermal, and biofuels) accounted for another 2.4% and are growing very
2
CHAPTER 1. INTRODUCTION 3
Figure 1.1: 2008 worldwide renewable energy sources. Source: [12].
rapidly [11]. The share of renewables in electricity generation is around 18%, with 15%
of global electricity coming from hydroelectricity and 3.4% from new renewables.
The European policy concerning the use of electricity from renewable energy sources
aims at a 20% share of renewable energy in the European energy system. In summer
2010, the governments of the member states are to submit national plans to the European
Commission. Some countries have already developed national targets. In Denmark, the
governments longterm policy aims at achieving a 30% share of energy from renewable
energy sources in 2020 [79].
While most renewable energy projects and production is largescale, renewable tech
nologies are also suited to small oﬀgrid applications, sometimes in rural and remote
areas, where energy is often crucial in human development.
Some renewable energy technologies are criticized for being intermittent or unsightly,
yet the renewable energy market continues to grow. Climate change concerns, coupled
with high oil prices, peak oil, and increasing government support, are driving increas
ing renewable energy legislation, incentives and commercialization. New government
spending, regulation and policies helped the renewable energy industry weather the
2009 economic crisis better than many other sectors [65].
In the past, renewable energy has generally been more expensive to produce and
use than fossil fuels. Renewable resources are often located in remote areas, and it is
CHAPTER 1. INTRODUCTION 4
expensive to build power lines to the cities where the electricity they produce is needed.
The use of renewable sources is also limited by the fact that they are not always available;
cloudy days reduce solar power; calm days reduce wind power; and droughts reduce the
water available for hydropower.
1.1.1 Some properties of renewable energy
Renewable energy is in principle a clean, emission free power generation technology.
Some important properties of renewable energy conversion can be summarized as follows:
• Greenhouse eﬀect
First, renewable energy conversion produces no carbon dioxide (the main potential
greenhouse gas) during operation, and only minimal quantities during the manu
facture of its equipment and construction. By contrast, fossil fuels such as coal,
gas and oil are major emitters of carbon dioxide.
• Air pollution
Renewable energy also has a positive eﬀect on the quality of the air we breathe.
The combustion of fossil fuels produces sulphur dioxide and nitrogen oxide, both
serious sources of pollution. These gases are the main components of the ”acid
rain” eﬀect  killing forests, polluting water courses and corroding the stone facades
of buildings; not to mention the human health eﬀects.
• Water
Another consideration of renewable energy deployment concerns water. In an in
creasingly waterstressed world, renewable energy conversion uses virtually none
of this most precious of commodities in its operation. Most conventional tech
nologies, from mining and extraction to fuel processing and plant cooling measure
their water use in millions of liters per day. Other environmental eﬀects resulting
from the range of fuels currently used to generate electricity include the landscape
degradation and dangers of fossil fuel exploration and mining, the pollution caused
by accidental oil spills and the health risks associated with radiation produced by
the routine operation and waste management of the nuclear fuel cycle. Exploiting
renewable sources of energy, reduces these risks and hazards.
• Environmental impact
Renewable energy is arguably the cleanest electricity generation technology, but,
like any other industry, does have environmental impacts. The construction and
operation, often in rural areas, raises issues of visual impact, noise and the poten
tial eﬀects on local ecology and wildlife. Most of these issues are addressed during
consultation with local authorities. Since the early days of this relatively young
industry, signiﬁcant improvements have been made with regards to the siting of
wind farms and the design of turbines, for instance.
CHAPTER 1. INTRODUCTION 5
1.1.2 A few numbers
Renewable energy resources include solar energy, geothermal energy, energy from the
wind or waves, energy from tides and energy from biomass [42]. Only these relevant for
the present thesis are now discussed.
• Solar energy
Every year the Earth receives about 300,000,000,000,000,000,000,000 kJ of energy
from the sun. Some energy is absorbed by green plants and used to make food
by photosynthesis. So ultimately, the sun is the source of most energy resources
available to us, including fossil fuels. The two energy sources considered in the
present work (wind and waves) are indirectly a result of sun radiation: solar energy
drives processes in the atmosphere that cause the wind and waves [42].
• Wind energy
When the earth is irradiated by the sun the ground absorbs some of this radiation.
This heated ground warms the air above it. Hot air rises in what are called
convection currents. The uneven heating of the earth’s surface causes winds.
For example, if the sun’s rays fall on land and sea, the land heats up more quickly.
This results in the air above the land moving upwards more quickly than that over
the sea (hot air rises). As a result the colder air over the sea will rush in to ﬁll
the gap left by the rising air. It is processes like these that give rise to high and
low pressure areas, and thus to winds.
Wind energy is nonpolluting and is freely available in many areas. Wind turbines
are becoming more eﬃcient. The cost of the electricity they generate is falling.
Large balancing areas and aggregation beneﬁts of large areas help in reducing
the variability and forecast errors of wind power as well as in pooling more cost
eﬀective balancing resources [41]. There are already several power systems and
control areas coping with large amounts of wind power [115], like in Denmark,
Germany, Spain, Portugal and Ireland that have integrated 920 % of wind energy
(of yearly electricity demand).
However, the disadvantages of wind energy exist as well. To be eﬃcient, wind
turbines need to be linked together in wind farms, often with 20 turbines or
more. This looks unsightly, and can be noisy. The wind farms also need to
be sited reasonably close to populations so that the electricity generated can be
distributed. Another disadvantage is that winds are intermittent and do not blow
all the time [42].
In this thesis, one turbine used to convert wind energy is extensively optimized in
order to improve the output power.
• Wave energy
Waves are caused by the action of winds on the sea. Waves can be many meters
in height and contain a great deal of energy. This energy can be harnessed to
CHAPTER 1. INTRODUCTION 6
drive turbines that generate electricity. Wave energy collectors are of two main
types. The ﬁrst type directs waves into manmade channels, where the water
passes through a turbine that generates electricity. The second type uses the up
and down movement of a wave to push air.
For example, the Limpet shoreline wave energy concept has been commissioned in
December 2000 on the Island of Islay, oﬀ the west coast of Scotland. It is intended
to enable Islay to replace fossil fuels and become selfsuﬃcient through renewable
energy. The waves feed indirectly a pair of counterrotating Wells turbines, each of
which drives a 250 kW generator, giving a theoretical peak power of 500 kW [24].
Waves’ energy is nonpolluting, wave turbines are relatively quiet to operate and
do not aﬀect wildlife. However, some disadvantages exist: the turbines can be
unsightly; wave heights vary considerably, so they would not produce a constant
supply of energy. In this work, the Wells turbine used to convert wave energy into
mechanical energy is investigated and optimized to increase its eﬃciency.
1.1.3 Future of renewable energy
The worldwide energy demand is continuously growing and, according to the forecasts
of the International Energy Agency, it is expected to rise by approx. 50% until 2030.
Currently, over 80% of the primary energy demand is covered by fossil fuels. Although
their reserves will last for the next decades, they will not be able to cover the worldwide
energy consumption in the long run. In view of possible climatic changes due to the
increase in the atmospheric CO
2
content as well as the conceivable scarcity of fossil fuels,
it becomes clear that future energy supply can only be guaranteed through increased
use of renewable energy sources. With energy recovery through renewable sources like
sun, wind, water, tides, geothermal or biomass the global energy demand could be met
many times over; currently, however, it is still ineﬃcient and too expensive in many
cases to take over signiﬁcant parts of the energy supply.
Renewable energies have long since emerged from their much ridiculed niche existence
and established a ﬁrm place in the energy mix. Their further expansion is certain now
that the European Union has laid down ambitious and binding targets. These state
that by 2020 renewable energies are to account for as much as 20% of Europe’s energy
consumption. These targets focus attention not only on the electricity sector, but also on
the use of renewable energy sources in heat production and in the transport sector [67].
Due to the usual adaptation reactions on the markets, it is foreseeable that prices for
fossil fuels will rise, while signiﬁcantly reduced prices are expected for renewable ener
gies. Already today, wind, water and sun are economically competitive in some regions.
However, to solve energy and climate problems, it is not only necessary to economically
utilize renewable alternatives to fossil fuels, but also to optimize the whole value added
chain of energy, i.e., from development and conversion, transport and storage up to the
consumers’ utilization.
Innovation and increases in eﬃciency in conjunction with a general reduction of
energy consumption are urgently needed in all ﬁelds to reach the targets within the
CHAPTER 1. INTRODUCTION 7
given time since the world population is growing and striving for more prosperity [26]
(see Fig. 1.2).
Figure 1.2: 2008 worldwide renewable energy sources [26].
1.2 Scope of the Thesis
The optimization of renewable energy turbomachines is a completely new topic. While
gas turbine have been for instance considered extensively, turbomachines used for wind
and wave energy conversion are still at a very basic stage. The aim of this thesis is
to investigate and optimize two such devices. Therefore, the speciﬁc objectives of this
work are:
1. Savonius turbines (wind energy conversion)
• Study the impact of an obstacle shielding partially the returning blade(s) and
optimize the position and angle of this obstacle as well as the shape of the
blades under the eﬀect of this obstacle.
• Investigate the impact of a deﬂector installed simultaneously with the obsta
cle (frontal guiding plates) to redirect the ﬂow toward the advancing blade(s)
and optimize the position and angle of these guiding plates as well as the
shape of the blades under the eﬀect of these plates.
• Optimize the performance of the full turbine considering either two or three
three blades.
• At the end, a considerably improved design must be available.
CHAPTER 1. INTRODUCTION 8
2. Wells turbine (wave energy conversion)
• Study a modiﬁed Wells turbine consisting of nonsymmetric airfoils, taking
into account the mutual interaction between the blades.
• Optimize the blade shape of monoplane Wells turbine and obtain new, non
standard airfoils with a higher performance.
• Optimize the shape of the airfoil for a twostage, modiﬁed Wells turbine using
nonsymmetric airfoils.
• Obtain the optimal shape of the airfoil for a twostage Wells turbine using
symmetric airfoils.
• At the end, a considerably improved design must be available.
1.2.1 Conclusions and outline
This chapter provides a brief introduction to renewable energy conversion and its im
portance for human life. In chapter 2, wind and wave energy conversion are discussed
in details, followed by a discussion of the Savonius turbine and Wells turbine, which are
the main topic of this work. Chapter 3 illustrates the employed optimization technique.
A literature review of the performance of Savonius and Wells turbines is presented in
chapter 4 including previous attempts to improve the performance of both turbines.
The numerical methods are introduced in chapter 5, including the coupling between the
ﬂow solver and the optimizer (OPAL). All the results of the optimization are presented
in chapter 6 for Savonius turbine and chapter 7 for Wells turbine, comparing the new
designs with the classical ones. The thesis ends with conclusions and suggestions for
future work in chapter 8.
Chapter 2
Basic concepts
2.1 Introduction
The utilization of renewable energy sources is not at all new; in the history of mankind
renewable energies have for a long time been the primary possibility of generating en
ergy. This only changed with industrial revolution when lignite and hard coal became
increasingly important. Later on, also crude oil gained importance. Oﬀering the ad
vantages of easy transportation and processing, crude oil has become one of the prime
energy carriers applied today. As fossil energy carriers were increasingly used for en
ergy generation, at least by the industrialized countries, the application of renewable
energies decreased in absolute and relative terms; besides a few exceptions, renewable
energies are of secondary importance with regard to overall energy generation. Yet,
the utilization of fossil energy carriers involves a series of undesirable side eﬀects which
are less and less tolerated by industrialized societies increasingly sensitized to possible
environmental and climate eﬀects at the beginning of the 21
st
century. This is why the
search for environmental, climatefriendly and socially acceptable, alternatives suitable
to cover the energy demand has become increasingly important. Also with regard to the
considerable price increase for fossil fuel energy on the global energy markets in the last
few years, not only in Europe, high hopes and expectations are placed on the multiple
possibilities of utilizing renewable sources of energy. Considering this background, the
present chapter aims at presenting the physical and technical principles for using wind
and wave energy.
2.2 Wind energy conversion
Solar radiation induces the movement of the air masses within the atmosphere of the
earth. Of the total solar radiation incident on the outer layer of the atmosphere, ap
proximately 2.5% are utilized for the atmospheric movement. This leads to a theoretical
overall wind power of approximately 4.3 10
15
W. The energy contained in the moving
air masses, which for example can be converted into mechanical and electrical energy
by wind mills and turbines, is therefore a secondary form of solar energy. The aim of
10
CHAPTER 2. BASIC CONCEPTS 11
the following discourse is to show the main basic principles of the supply in wind energy
and to discuss its supply characteristics [53].
Wind power is growing at the rate of 30% annually, with a worldwide installed
capacity of 157,900 megawatts (MW) in 2009 [12]. It is widely used in Europe, Asia,
and the United States. The measured wind speeds can be analyzed and the annual
mean value can be calculated. If the yearly mean wind velocity is averaged over various
years, areas of similar wind speeds can be identiﬁed. Figure 2.1 shows these values on
a worldwide scale referring to 10 m above ground.
Figure 2.1: Distribution of yearly annual mean values of the wind speed for 10 m above
ground worldwide [53].
Wind energy converters harness the kinetic energy contained in ﬂowing air masses.
In the following, the fundamental physical principles of this type of energy conversion
are explained. Most modern wind energy converters are equipped with rotors to ex
tract wind power, and consist of one or several rotor blades. The extracted wind power
generates rotation and is thereby converted into mechanical power at the rotor shaft.
Mechanical power is taken up at the shaft in the form of a moment at a certain rotation
and is transfered to a machine (such as a generator). The entire wind power station
thus consists of a wind energy converter (rotor), a mechanical gear and a generator. It
is physically impossible to technically exploit the entire wind energy, as in this case air
ﬂow would come to a standstill; air would fail to enter the swept rotor area, and wind
power would no longer be available. There are two diﬀerent physical principles to extract
power from wind. The airfoil drag method is based on the wind drag force incident on
a windblown surface. The second principle, also referred to as aerodynamic or airfoil
lift principle, which is based on ﬂow deviation inside the rotor is at present predomi
nantly applied for wind energy conversion. Both principles are outlined throughout the
following sections to explain the main diﬀerences. The implementation of an innovative
aerodynamic control technique in wind turbines is a point under extensive investigation
CHAPTER 2. BASIC CONCEPTS 12
since the conventional wind turbine blade technology is reaching its limits. Almost all
the eﬀort of the wind turbine industry in the ﬁeld of aerodynamics is related to the
development of blades which oﬀer better performance, increased reliability and faster
control of larger wind turbines [78]. A discussion on the maximum achievable wind
power output by means of an ideal wind energy converter follows.
2.2.1 Horizontal axis turbines
These are the standard turbines used at present. According to the lift principle, wind is
deviated to generate peripheral force inside the rotor (Fig. 2.2). For highspeed propeller
type converters, rotor blades are mostly designed according to the wing theory.
Figure 2.2: Horizontal axis wind turbine.
The airfoil crosssection at radius r is set at a local blade pitch angle ϑ with respect
to the rotor plane of rotation (Fig. 2.3). The axial free stream velocity v
a
in the rotor
plane and the tangential speed u = rω at the radius of the blade crosssection combine
to form a resultant ﬂow velocity v
r
. Together with the airfoil chord line, it forms the
local aerodynamic angle of attack α. For the beneﬁt of those readers unfamiliar with
aerodynamics, the diﬀerence between the aerodynamic angle of attack α and the blade
pitch angle ϑ should be noted: the angle of attack is an aerodynamic parameter and the
blade pitch angle is a design parameter. The two angles are often confused, making it
more diﬃcult to understand the aerodynamic relationships. Linking the relationships of
ﬂuid mechanics for the momentum of the axial ﬂow and of the radial ﬂow components
of the rotating wake with the formulations for the aerodynamic forces at the blade
CHAPTER 2. BASIC CONCEPTS 13
element allows the ﬂow conditions at the blade element to be determined so that the
local aerodynamic lift and drag coeﬃcients can be calculated.
The calculation of the balance of forces includes not only the pure airfoil drag but
also other drag components which derive from the spatial ﬂow around the rotor blade.
In particular, the ﬂow around the blade tip, a result of the pressure diﬀerence between
the top and the underside of the blade, produces the socalled free tip vortices. The
resultant drag is called induced drag, a function of the local lift coeﬃcient and the
aspect ratio (slenderness) of the blades. The higher the aspect ratio, i.e., the more
slender the blades, the lower the induced drag. These blade tip losses are introduced
as additional drag components, as are the hub losses which are the result of vortices
in the wake of the ﬂow around the hub. They are derived from a complex vortex
model of the rotor ﬂow [37]. With several semiempirical approaches for these vortex
losses, the blade element theory provides the distribution of aerodynamic forces over the
length of the blade. This is usually divided into two components: one in the plane of
rotation of the rotor, the tangential force distribution; and one at right angles to it, the
thrust distribution. Integrating the tangential force distribution over the rotor radius
provides the driving torque of the rotor and, with the rotational speed of the rotor, the
rotor power or power coeﬃcient, respectively. Integrating the thrust distribution yields
the total rotor thrust for instance to the tower. Referred to the power rating of the air
stream, the simple momentum theory by Betz provides the ideal peak power coeﬃcient of
0.593 which is independent of the tip speed ratio. Taking into consideration the angular
momentum in the rotor wake shows that the power coeﬃcient becomes a function of
the tip speed ratio (see later section 2.2.3). It is only when the tip speed ratios become
inﬁnitely high that the power coeﬃcient approaches Betz’s ideal value. Introducing the
aerodynamic forces acting on the rotor blades, and particularly the aerodynamic drag,
further reduces the power coeﬃcient; in addition, the power coeﬃcient now exhibits an
optimum value at a certain tip speed ratio. The aerodynamic rotor theory based on the
momentum theory and on the blade element theory, yields the real rotor power curve
with good approximation. Nevertheless, it should be kept in mind that the momentum
theory as well as the blade element model include several simpliﬁcations, which limit
their validity to a discshaped wind energy converter. Sometimes, the momentum theory
is therefore called ”disc actuator theory” [37].
2.2.2 Vertical axis turbines
The oldest design of wind rotors features rotors with a vertical axis of rotation. At the
beginning, however, verticalaxis rotors could only be built as pure dragtype rotors (see
the idea of drag turbines in Fig. 2.4).
The Savonius rotor, which can be found as simple ventilator on some railroad car
riages or delivery vans, and the cup anemometer used to measure wind velocity are
wellknown examples of rotors with a vertical axis of rotation. It was only recently
that engineers succeeded in developing verticalaxis designs, which could also eﬀectively
utilize aerodynamic lift. The design proposed in 1925 by the French engineer Darrieus,
in particular, has been considered as a promising concept for modern wind turbines. As
CHAPTER 2. BASIC CONCEPTS 14
Figure 2.3: Flow velocities and aerodynamic forces at the airfoil crosssection of a blade
element [53].
is the case with horizontalaxis rotors, Darrieus rotors are preferably built with two or
three rotor blades. The speciﬁc advantages of vertical axis turbine concepts are that
their basically simple design includes the possibility of housing mechanical and electrical
components, gearbox and generator at ground level, and that there is no yaw system.
This is countered by disadvantages such as low tipspeed ratio, inability to selfstart
and not being able to control power output or speed by pitching the rotor blades. A
variation of the Darrieus rotor is the socalled Hrotor. Instead of curved rotor blades,
straight blades connected to the rotor shaft by struts are used. Attempts were made
particularly in the UK, in the US and in Germany to develop this design to commercial
maturity. Hrotors of a particularly simple structure, with the permanently excited gen
erator integrated directly into the rotor structure without intermediary gearbox, were
developed by a German manufacturer up until the beginning of the nineties but the
development was stopped then since there was no economic success in sight. Occasion
ally, the Savonius design is still used for small, simple wind rotors, especially for driving
small water pumps. It is not suitable for electricitygenerating wind turbines due to its
low tipspeed ratio and its comparatively low power coeﬃcient.
Altogether, it can be said that wind rotors with vertical axes and among these pri
marily the Darrieus rotor and Savonius rotor, might still have a large potential for de
velopment and optimization. Whether the basic advantages of these designs can prevail
over their disadvantages and whether it will become a serious rival to the horizontalaxis
rotors cannot be foreseen for the longterm. In any case, this will still require a relatively
long period of development. In half of this thesis, the development and optimization of
CHAPTER 2. BASIC CONCEPTS 15
U
Fwall
R
Figure 2.4: Flow conditions and drag force for vertical axis turbines [53].
a new concept based on a Savonius turbine is the target.
2.2.3 Betz limit
The main component of a wind turbine is the energy converter, which transforms the
kinetic energy contained in the moving air into mechanical energy. For an initial dis
cussions of basic principles, the exact nature of the energy converter is irrelevant. The
extraction of mechanical energy from a stream of moving air with the help of a disk
shaped, rotating wind energy converter follows its own basic rules. The credit for having
recognized this principle is owed to Albert Betz. Between 1922 and 1925, Betz published
writings in which he was able to show that, by applying elementary physical laws, the
mechanical energy extractable from an air stream passing through a given crosssectional
area is restricted to a certain ﬁxed proportion of the energy or power contained in the
air stream. Moreover, he found that optimal power extraction could only be realized
at a certain ratio between the ﬂow velocity of air in front of the energy converter and
the ﬂow velocity behind the converter. Although Betz’s ”momentum theory”, which
assumes an energy converter working without losses in a frictionless airﬂow, contains
simpliﬁcations, its results are still used for performing ﬁrst calculations in practical en
gineering. But its true signiﬁcance is founded in the fact that it provides a common
physical basis for the understanding and operation of wind energy converters of various
CHAPTER 2. BASIC CONCEPTS 16
designs. For this reason, the following pages will provide a summarized mathematical
derivation of the elementary momentum theory by Betz.
2.2.3.1 Betz’s momentum theory
The kinetic energy of an air mass m moving at a velocity U can be expressed as:
E =
mU
2
2
(2.1)
Considering a certain crosssectional area A, through which the air passes at velocity
U, the volume ﬂow rate Q (m/s
3
) ﬂowing through during a time unit, the socalled
volume ﬂow rate, is:
Q = AU (2.2)
and the mass ﬂow rate with the air density ρ is:
˙ m = ρAU (2.3)
The equations expressing the kinetic energy of the moving air and the mass ﬂow
yield the amount of energy passing through crosssection A per unit time. This energy
is physically identical to the power P in (W):
P =
ρAU
3
2
(2.4)
The question is how much mechanical energy can be extracted from the freestream
airﬂow by an energy converter? As mechanical energy can only be extracted at the cost
of the kinetic energy contained in the wind stream, this means that, with an unchanged
mass ﬂow, the ﬂow velocity behind the wind energy converter must decrease. Reduced
velocity, however, means at the same time a widening of the crosssection, as the same
mass ﬂow must pass through it. It is thus necessary to consider the conditions in front
of and behind the converter (Fig. 2.5). Here, U
1
is the undelayed freestream velocity,
the wind velocity before it reaches the converter, whereas U
2
is the ﬂow velocity behind
the converter. Neglecting any losses, the mechanical energy, which the diskshaped
converter extracts from the airﬂow corresponds to the power diﬀerence of the air stream
before and after the converter:
P =
ρAU
3
1
−ρAU
3
2
2
(2.5)
Maintaining the mass ﬂow (continuity equation) requires that:
ρA
1
U
1
= ρA
2
U
2
(2.6)
Thus,
P =
˙ m
2
(U
2
1
−U
2
2
) (2.7)
CHAPTER 2. BASIC CONCEPTS 17
U1
U2
U
Rotordisk
Figure 2.5: Flow conditions due to the extraction of mechanical energy from a freestream
air ﬂow, according to the elementary momentum theory.
From this equation it follows that, in purely formal terms, power would have to
be at its maximum when U
2
is zero, namely when the air is brought to a complete
standstill by the converter. However, this result does not make sense physically. If
the outﬂow velocity U
2
behind the converter is zero, then the inﬂow velocity before the
converter must also become zero, implying that there would be no more ﬂow through
the converter at all. As could be expected, a physically meaningful result consists in
a certain numerical ratio of U
2
/U
1
where the extractable power reaches its maximum.
This requires another equation expressing the mechanical power of the converter. Using
the law of conservation of momentum, the force which the air exerts on the converter
can be expressed as:
F = ˙ m(U
1
−U
2
) (2.8)
According to the principle of ”action equals reaction”, this force, the thrust, must be
counteracted by an equal force exerted by the converter on the airﬂow. The thrust,
so to speak, pushes the air mass at air velocity U
, present in the plane of ﬂow of the
converter. The power required for this is:
P = FU
= ˙ m(U
1
−U
2
)U
(2.9)
Thus, the mechanical power extracted from the air ﬂow can be derived from the energy
or power diﬀerence before and after the converter, on the one hand, and, on the other
hand, from the thrust and the ﬂow velocity. Equating these two expressions yields the
relationship for the ﬂow velocity U
:
˙ m
2
(U
2
1
−U
2
2
) = ˙ m(U
1
−U
2
)U
(2.10)
Thus, the ﬂow velocity in the converter plane is equal to the arithmetic mean of U
1
and
U
2
.
U
=
(U
1
+ U
2
)
2
(2.11)
CHAPTER 2. BASIC CONCEPTS 18
The mass ﬂow thus becomes:
˙ m = ρAU
=
ρA(U
1
+ U
2
)
2
(2.12)
The mechanical power output of the converter can be expressed as:
P =
ρA(U
2
1
−U
2
2
)(U
1
+ U
2
)
4
(2.13)
In order to provide a reference for this power output, it is compared with the power
of the freeair stream which ﬂows through the same crosssectional area A, without
mechanical power being extracted from it. This power was:
P
o
=
ρAU
3
1
2
(2.14)
The ratio between the mechanical power extracted by the converter and that of the
undisturbed air stream is called the ”power coeﬃcient” C
p
:
C
p
=
P
P
o
=
ρA(U
2
1
−U
2
2
)(U
1
+U
2
)
4
ρAU
3
1
2
(2.15)
After some rearrangement, the power coeﬃcient can be speciﬁed directly as a function
of the velocity ratio U
2
/U
1
:
C
p
=
P
P
o
=
1
2
¸
1 −
U
1
U
2
2
¸
¸
1 +
U
1
U
2
(2.16)
The power coeﬃcient, i.e., the ratio of the extractable mechanical power to the power
contained in the air stream, therefore, now only depends on the ratio of the air veloci
ties before and after the converter. If this interrelationship is diﬀerentiated to get the
maximum value of the power coeﬃcient it can be obtained that the power coeﬃcient
reaches a maximum at a certain velocity ratio with U
2
/U
1
= 1/3. The maximum ”ideal
power coeﬃcient” C
p
becomes
C
p(max)
=
16
27
= 0.593 (2.17)
Betz was the ﬁrst to derive this important value and it is, therefore, frequently
called the ”Betz factor”. Knowing that the maximum, ideal power coeﬃcient is reached
at U
2
/U
1
= 1/3, the ﬂow velocity U
in the rotor plane becomes:
U
=
2
3
U
1
(2.18)
It is worthwhile to recall that these basic relationships were derived for an ideal,
frictionless ﬂow (for the eﬀect of the friction, see for instance [77]), and that the result
was obviously derived without having a close look at the wind energy converter. In real
CHAPTER 2. BASIC CONCEPTS 19
Figure 2.6: Power coeﬃcients of wind rotors of diﬀerent designs [37].
cases, the power coeﬃcient will always be smaller than the ideal Betz value, as shown in
Fig. 2.6. The essential ﬁndings derived from the momentum theory can be summarized
in words as follows:
• The mechanical power which can be extracted from a freestream airﬂow by an
energy converter increases with the third power of the wind velocity.
• The power increases linearly with the crosssectional area of the converter tra
versed; it thus increases with the square of its diameter.
• Even with an ideal airﬂow and lossless conversion, the ratio of extractable me
chanical work to the power contained in the wind is limited to a maximum value
of 0.593. Hence, only less than 60% of the wind energy of a certain crosssection
can be converted into mechanical power.
• When the ideal power coeﬃcient achieves its maximum value (C
p
= 0.593), the
wind velocity in the plane of ﬂow of the converter amounts to two thirds of the
undisturbed wind velocity and is reduced to one third behind the converter.
2.2.4 Principle operation of standard Savonius turbines
In half of this thesis, optimizing of a Savonius wind turbine will be considered.
CHAPTER 2. BASIC CONCEPTS 20
The choice of a wind turbine is obviously based on its performance in connection
with the local wind conditions. To support the discussion, a comparison between the
characteristics of the main conventional wind turbines is shown in Fig. 2.6.
Figure 2.6 gives the power coeﬃcient C
p
, ratio of the aerodynamic power of the
turbine to the power of the incident wind, as a function of the speed ratio λ. This speed
ratio λ is also called velocity coeﬃcient and is equal to the ratio of the tip peripheral
speed to the wind velocity. The power coeﬃcient is directly linked to the global eﬃciency
of a wind machine. The curves in Fig. 2.6 show that the fast running horizontal axis
wind machines (two or threebladed airscrew) have incontestably the best eﬃciencies.
Consequently, theses machines are nowadays systematically chosen for the equipment of
largearea wind sites. On the other hand, the Savonius rotor, which is a slowrunning
vertical axis wind machine (λ 1.0) has a rather poor eﬃciency: C
p
0.15 to 0.2
at best [68]. Nevertheless, it can present some advantages for speciﬁc applications, in
particular due its simplicity, and resulting robustness and low cost. And there is room
for improvement!
S.J. Savonius initially developed the vertical axis Savonius rotor in the late 1920s.
The concept of the Savonius rotor is based on cutting a cylinder into two halves along
the central plane and then moving the two half cylinders sideways along the cutting
plane, so that the crosssection resembles the letter S [36] (Fig. 2.7).
Figure 2.7: Conventional Savonius rotor.
2.2.5 Performance of a Savonius turbine
Using the notations of Fig. 2.8, the velocity coeﬃcient is deﬁned as:
λ = ωR/U (2.19)
For a Savonius rotor of height H, a wind of incoming velocity U, the mechanical
power P and the mechanical torque on the axis of a Savonius turbine can respectively
be written as follows:
C
p
=
P
ρRHU
3
(2.20)
CHAPTER 2. BASIC CONCEPTS 21
Figure 2.8: Schematic description and main parameters characterizing a Savonius rotor.
and
C
m
=
T
ρR
2
HU
2
(2.21)
where C
p
and C
m
are respectively the power coeﬃcient and the torque coeﬃcient. In
the following sections, a rotor is called a conventional Savonius rotor if the geometrical
parameters a and e Fig. 2.8 are respectively equal to 0 and d/6. This reference conﬁgu
ration of the rotor has been extensively studied by many groups (see citations in [70] and
next chapter). This conventional design will be the starting point for the optimization
process described in later chapters.
2.3 Wave energy conversion
Let us switch to the second focus of this document: wave energy conversion with opti
mized Wells turbines. Very large energy ﬂuxes can be found in deep water sea waves.
The power in the wave is proportional to the square of the amplitude and to the period
of the motion. Therefore, the long period (≈10 s) and large amplitude (≈2 m) waves
have considerable interest for power generation, with energy ﬂuxes commonly averaging
between 50 and 70 kW per meter width of oncoming wave.
The possibility of generating electrical power from these deep water waves has been
realized for many years, and there are countless ideas for machines to extract the power.
The utilization of the sea wave was rarely considered on a practical scale prior to
1973. However, a great deal of research has been conducted since 1973, the year of
socalled oil crisis [5].
In recent years interest has revived, particularly in Japan, Britain and Scandinavia,
so the research and smallscale development has progressed to the stage of commercial
construction for power extraction. As with all renewable energy supplies, the scale of
operation has to be determined and present trends support moderate power generation
levels at about 1 MW from modular devices about 50 m wide across the wavefront.
Such devices should be economic to replace dieselgenerated electricity, especially on
islands [134].
CHAPTER 2. BASIC CONCEPTS 22
2.3.1 Wave energy potential
There is a large amount of ongoing work on wave energy due to a broad availability
(Fig. 2.9), which cannot be done justice in a brief overview. For ease of presentation,
the activities will be divided between the technologies suitable for deployment on the
shoreline, near the shore and oﬀshore [133].
Figure 2.9: Approximate global distribution of wave power levels in kW/m of wave
front [133].
• Shoreline devices
These devices are ﬁxed to or embedded in the shoreline itself, which has the
advantage of easier maintenance and/or installation. In addition these would
not require deep water moorings or long lengths of underwater electrical cable.
However, they would experience a less powerful wave regime.
• Near shore devices
The main prototype device for moderate water depths (i.e., < 20 m depth) is
the OSPREY developed by Wavegen. This is a 2 MW system, with provision for
addition of a 1.5 MW wind turbine. Since there could be environmental objections
to large farms of wind or wave energy devices close to the shore, this system aims
to maximize the amount of energy produced from a given amount of near shore
area.
• Oﬀshore devices
This class of device exploits the more powerful wave regimes available in deep water
(> 40 m depth) before energy dissipation mechanisms have had a signiﬁcant eﬀect.
CHAPTER 2. BASIC CONCEPTS 23
However, it is important to appreciate the many diﬃculties facing wave power de
velopments. It was summarized by Twidell and Weir [134] as follows:
1. Wave patterns are irregular in amplitude, phase and direction. It is diﬃcult to
design devices to extract power eﬃciently over this wide range of variables.
2. There is always probability of extreme gales or hurricanes producing waves of freak
intensity. The structure of the power devices must be able to withstand this.
3. Wave periods are commonly from 5 up to 10 sec (frequency
∼
=0.1 Hz). It is
extremely diﬃcult to couple this irregular slow motion to electrical generators
requiring about 500 times greater frequency.
Many attempts have been made to construct such devices and eﬃciently match
variable natural conditions, extract the wave power and withstand the listed diﬃculties.
Some of these devices being developed are placed underwater, anchored to the ocean
ﬂoor, while others ride on top of the waves. The following section gives an idea of
the most famous mechanism to convert the wave energy to mechanical energy, called
Oscillating Water Column mechanism (OWC).
2.3.2 Oscillating Water Column (OWC) principle
One of the many extraction systems proposed in recent years, the oscillating water
column device, provides the simplest and possibly the most reliable means of converting
slow irregular wave motion into high speed rotational movement required for electrical
power generation.
The device is essentially a caisson rested on sea bottom with a large submerged
opening at the front and a small nozzle at the ceiling (Fig. 2.10). An air turbine coupled
to an electric generator is connected to the nozzle. The water column within the lower
half of the caisson is caused to oscillate vertically by incident waves through the opening,
and it induces the compression and expansion of air mass within the upper half of the
caisson as shown in Fig. 2.10. The air motion generates a highvelocity ﬂow through the
nozzle, which activates the air turbine and generates electricity [73]. The wave power is
thus converted into air power in the air chamber of the wave powerextracting caisson.
OWC wave power extractors can be quite easily ﬁtted to a vertical breakwater,
because the latter is mostly built with a large concrete caisson rested on the foundation.
A combination of OWC wave power extractor and breakwater is also attractive from the
viewpoint of economical feasibility of wave power extraction, because the construction
cost of the total system can be jointly born by the accounts for power generation and
harbor protection.
2.3.3 Wells turbines
Serious research on waveenergy extraction methods began in several countries during
the 1970’s. Dr. Wells, a former professor of civil engineering at Queen’s university of
CHAPTER 2. BASIC CONCEPTS 24
Figure 2.10: The wave motion in the OWC device [73].
Belfast, proposed in 1976 a form of selfrectifying axial ﬂow air turbine as a device suit
able for wave energy conversion using the oscillating water column. In its simplest form
the air turbine rotor consists of several symmetrical airfoil blades positioned around a
hub. Because of its simple and eﬃcient operation, the Wells turbine has been widely
CHAPTER 2. BASIC CONCEPTS 25
applied for ocean waves energy absorption. Therefore, it has been subjected to a con
siderable amount of research and development in many counties. In the next chapter,
the state of the art of Wells turbines used as converter for wave energy will be proposed.
2.3.4 Operation of Wells turbines
Most research programs attempting to gain energy from waves depend on the OWC
as converter mechanism [7, 22, 25, 75, 137]. The water wave energy is converted to
pneumatic energy in the air, which passes periodically across a selfrectifying, axial air
ﬂow turbine. The ﬁrst prototype was constructed in UK in 1988. This device is located
on the Isle of Islay, one of the southern islands in the Inner Hebrides and depends on a
Wells turbine as ﬁnal converter [15, 16, 84, 138].
The turbine itself consists in a number of symmetric airfoils set around the hub radi
ally at 90
◦
stagger angle, with the chord plane normal to the axis of rotation (Fig. 2.11).
According to the standard airfoil concept, if the airfoil is set at an angle of attack α in a
ﬂuid ﬂow, it will generate a lift force F
L
normal to the free stream and a drag force F
D
in the direction of the free stream. These lift and drag forces can be combined to get the
tangential force F
T
and the axial force as shown in Fig. 2.12. For a symmetrical airfoil
as considered in [84, 86, 88, 90, 92, 97], the direction of tangential force, F
T
is the same
for both positive and negative values of angle of attack (±α) as shown in Fig. 2.12.
Oscillatingairflow
Airturbine
Figure 2.11: Wells turbine in state of upward and downward air ﬂow.
2.3.5 Wells turbine performance
If such symmetric airfoil blades are positioned around an axis of rotation, they will
rotate in the tangential force direction regardless of the direction of airﬂow, as shown in
Fig. 2.12. The force F
T
is responsible for the torque and consequently the blade power,
CHAPTER 2. BASIC CONCEPTS 26
while the axial force F
A
results in an axial thrust along the axis of the rotor, which has
to be absorbed by the bearings. This leads to a unidirectional device rotation for an
alternating airﬂow without the need for nonreturn valve. The tangential force F
T
and
the axial force F
A
shown in Fig. 2.12 can be obtained from:
FT
FA
R
ut
w
vA
Rotation
Flow
FT
FA
R
ut
w
vA
Rotation
Flow
Figure 2.12: Axial and tangential forces acting on a Wells turbine.
F
T
= F
L
sin(α) −F
D
cos(α) (2.22)
F
A
= F
L
cos(α) + F
D
sin(α) (2.23)
The running characteristics under steady ﬂow conditions are usually characterized
by the tangential force coeﬃcient C
T
, axial force coeﬃcient C
A
and eﬃciency η with
ﬂow coeﬃcient φ. The tangential force coeﬃcient C
T
and the axial force coeﬃcient C
A
are calculated as :
C
T
= F
T
/
(1/2)ρ
v
2
A
+ u
2
t
zbc
(2.24)
C
A
= ∆p
0
πr
2
t
/
(1/2)ρ
v
2
A
+ u
2
t
zbc
(2.25)
where u
t
= ωr
t
is the peripheral velocity, ω is the rotor angular velocity and r
t
is the tip
radius (Fig. 2.13). Furthermore, v
A
is the axial velocity normal to the plane of rotation,
CHAPTER 2. BASIC CONCEPTS 27
Rotorhub
onshaft
Forcesresolvedin
directionofrotation
Rotation
Sym. Airfoil
Blade
FA
FT
Chord(c)
s
p
a
n
(
b
)
r
t
rh
FA
Oscillating
flow
Figure 2.13: Main geometrical parameters of a Wells turbine.
z is the number of blades, b is the blade span, c is the blade chord (see Fig. 2.13) and
∆p
0
is the total pressure diﬀerence across the rotor.
The nondimensional variables expressing the performance of a Wells turbine are
function of the aerodynamic force coeﬃcients that are in turn function of several pa
rameters, as given by [84, 92]:
(C
T
, C
A
) = f(φ, s, h, AR, τ, τ
c
, T
u
, f
∗
, Re, blade profile shape) (2.26)
where, φ = v
A
/u
t
is the ﬂow coeﬃcient, s is the rotor solidity, h is the hub to tip ratio
(r
h
/r
t
), AR is the aspect ratio (b/c), c is blade chord, b is blade span, τ is the blade
thickness, τ
c
is the tip clearance ratio, T
u
is the turbulence level, f
∗
the frequency of
wave motion and the Reynolds number Re.
2.4 Conclusions
After having introduced all needed concepts and notations, its is now time to analyze
the state of the art for both systems considered in this thesis: Savonius turbine (wind
energy) and Wells turbine (wave energy).
Chapter 3
Literature Review
Due to the growing signiﬁcance of wind and wave energy conversion, a wealth of pub
lications can be found for the systems considered in this thesis. The present chapter is
again divided in two parts. The present state of Savonius turbines is ﬁrst considered,
before reviewing the literature on Wells turbines.
3.1 Savonius turbine
Savonius turbines show inherent drawbacks compared to conventional wind turbines:
mainly a low eﬃciency and poor starting characteristics. Therefore, many authors have
tried to identify the best principles of operation and to improve the characteristics of
Savonius turbines.
3.1.1 Experimental investigations of Savonius turbines
Some studies have been carried out in wind tunnels, using controlled conditions. Other
consider freespace experiments. Generally, the global performance of a rotor, identical
to or derived from the conventional Savonius rotor, is presented in such studies, but
without realizing any detailed, quantitative parametric study. Sometimes, some visual
izations of the ﬂow in and around the rotor are proposed, but with a poor description of
the physical phenomena. Such publications are of very limited scientiﬁc or technological
interest and are therefore not included here.
Beyond issues associated with intellectual properties, this lack of quantitative in
formation is sometimes associated to the complexity of the ﬂow in and around wind
turbines. In particular for Savonius turbines, the resulting ﬂow conditions are highly
unsteady. Furthermore, boundary layer separation is an essential aspect for the eﬃ
ciency of the system. As a consequence, detailed aerodynamic studies are rare and
often do not allow the prediction of the energetic behavior of the rotor. However, some
publications [10, 27, 31, 51] are of higher quality and give a precise description of the
aerodynamics of the conventional Savonius rotor, mainly obtained by pressure measure
ments on the paddles.
28
CHAPTER 3. LITERATURE REVIEW 29
Further articles describe an extensive experimental study in a windtunnel to evaluate
the importance of geometrical parameters on the Savonius rotor performance. Through
the corresponding rotor power output coeﬃcient versus the tip speed ratio, the inﬂuence
of each blade geometry parameter is investigated. Such parametric studies already
lead to geometrical conﬁgurations with a considerable increase in the rotor performance
eﬃciency [30, 39, 54, 128, 142].
The ﬂow in and around a Savonius rotor has also been studied using ﬂow visualiza
tion experiments in combination with the measured pressure distributions on the blade
surfaces. It is observed that the ﬂow separation regions on the blade surfaces are fairly
reduced by the rotation eﬀect (compare Fig. 3.1 and 3.2). Similarly, the ﬂow through
the overlap is weakened by the appearance of resisting ﬂow. The former contributes to
the torque production of the rotating rotor while the latter acts as a resistance. These
phenomena, together with the ﬂow stagnation region on the front side of the rotor,
contribute to the power producing mechanism of the Savonius rotor [4, 29].
Figure 3.1: Flow in and around a nonrotating Savonius rotor (λ = 0); a: visualized
ﬂow ﬁeld; b: ﬂow inside the rotor; c: ﬂow model; d: surface pressure distribution [29].
CHAPTER 3. LITERATURE REVIEW 30
Figure 3.2: Flow in and around a Savonius rotor in rotation (λ = 0.9); a: visualized
ﬂow ﬁeld; b: ﬂow inside the rotor; c: ﬂow model; d: surface pressure distribution [29].
Signiﬁcant features observed here are the downward movement of the separation
point (Fig. 3.2c) and the relative decrease in the pressure coeﬃcient on the convex side
of the advancing blade (Fig. 3.2d). These phenomena can be caused by the occurrence of
a Coandalike ﬂow pattern (Fig. 3.2a) on the convex side, which appears clearly at small
rotor angles of θ = 0
◦
to 45
◦
. The attached ﬂow on the convex side tends to separate at
large rotor angles (θ = 90
◦
to 135
◦
), which is due to the outward ﬂow motion at the tip of
the advancing blade. This ﬂow is induced by the pressure gradient distributed over the
concave side of the advancing blade. The injected ﬂow grows into a vortex circulating in
the rotating direction of the rotor, which increases in size downstream. It is considered
that the attached ﬂow patterns of the rotating rotor contribute to the rotating torque of
the rotor, as is expected from the pressure distributions (Fig. 3.2d). On the other hand,
a relative decrease in the stagnation torque is expected here in comparison with the non
rotating rotor (Fig. 3.1), since the relative velocity is decreased on the advancing blade
and is increased on the returning one. In addition, the stagnation point moves to the
CHAPTER 3. LITERATURE REVIEW 31
center of the blade due to the rotation eﬀect. It can be seen that the pressure coeﬃcients
are decreased overall by the eﬀect of circulation produced by the rotor rotation. Such
a circulation is a steady phenomenon. In comparison with the nonrotating rotor, the
ﬂow through the overlap is reduced here by the production of resisting ﬂow. This ﬂow is
expected to reduce the pressure recovery eﬀect on the back side of the returning blade,
which is supported by the measured pressure distributions near the overlap.
3.1.2 Numerical investigations of Savonius turbines
Numerical simulations have also been carried out on this kind of rotors. These studies
include static or dynamic modelling. Nevertheless, most results suﬀer from a too crude
description of the rotor. A few of these papers [23, 28, 49] used the discrete vortex
method to predict the ﬂow around a pair of coupled Savonius rotors. They suggested that
the reason why so few numerical studies had been successful was due to the complexity
of the ﬂow pattern about the rotor and to the separation of the ﬂow from the blade
surfaces.
Usually, wind sites are equipped today with fastrunning horizontal axis wind tur
bines of the airscrew type, associated to a high eﬃciency. Some articles point out that
the choice of a wind turbine must not be based only on high eﬃciency and proposed
a comparative criterion adapted to the comparison of a horizontal axis wind turbine
with a vertical axis wind turbine: the Lσ criterion. This criterion consists in compar
ing wind turbines which intercept the same front width of wind, by allocating them a
same reference value of the maximal mechanical stress on the blades or the paddles. On
the basis of this criterion, a quantitative comparison points to a clear advantage of the
Savonius rotors, because of their lower angular velocity, and provides some elements for
the improvement of their rotor [71]. The results show that the power per unit length
provided by the considered Savonius rotor is about four times as high as that provided
by a fastrunning two bladed airscrew. The American windmill and the Savonius rotor
have comparable values of their associated power per unit length. A favorable factor to
obtain a high power per unit length is consequently a low angular velocity. In terms
of the ease of setting up, and the potential for improvement in eﬃciency, the Savonius
rotor should be preferred to all other considered conﬁgurations following [71].
Flow ﬁelds around rotating Savonius rotors have been also simulated by solving the
twodimensional incompressible NavierStokes equations [95]. The results show a good
agreement with experimental performance for the following points: the torque grows
weaker in inverse proportion to the tip speed ratio; a gap between the rotor buckets is
eﬀective in increasing maximum power. These calculations have been realized using a
static approximation (the rotor is supposed to be ﬁxed whatever the wind direction) and
also a dynamic calculation; in this second case, the velocity coeﬃcient was equal to 1.0
(nominal working point in accordance with Fig. 2.6). These calculations are continued
until the residual values (variations in certain chosen parameters, e.g., velocity in the
wind direction), have all dropped below 10
−3
(criterion for convergence). From our
experience, this is a relatively weak and perhaps insuﬃcient criterion.
A separate study has been carried out to verify the model accuracy, comparing a
CHAPTER 3. LITERATURE REVIEW 32
static simulation and a dynamic one. To do so, a static simulation of the ﬂow around
the conventional Savonius rotor (e/d = 1/6; a = 0; no central shaft) has allowed to
determine the pressure distribution on the paddles. Then the static torque has been
calculated as a function of the wind velocity angle θ (Fig. 2.8). These numerical results
were compared to experimental data. The simulations give satisfactory results since
the diﬀerences between the experimental data and the numerical simulations are always
below 10%, except for angles θ around 0
◦
and 180
◦
where an instability of the torque is
observed (Fig. 3.3, [68]).
Figure 3.3: Static torque coeﬃcient of a Savonius rotor (Re = 1.56 10
5
) [68].
Figure 3.4: Dynamic torque coeﬃcient of a Savonius rotor (Re = 1.56 10
5
, λ = 1) [68].
In a second step, a dynamic calculation (rotating turbine) has been carried out for
the same value of the Reynolds number: Re= 1.56 10
5
, setting the velocity coeﬃcient
CHAPTER 3. LITERATURE REVIEW 33
equal to 1.0. The torque coeﬃcient C
m
has been evaluated by calculating the average
value of the torque on a whole revolution of the rotor. The results are compared to the
ones given with the static calculation. The diﬀerence between the two curves generally
does not exceed 2% whatever the angle θ (Fig. 3.4). This is clearly in contradiction
with the experimental studies discussed previously (see again Fig. 3.1 and 3.2), and
must therefore be considered with caution. The simulated ﬂows have been analyzed by
visualizing relative velocity, vorticity, pressure, etc. The behavior of the shed vortices
has been observed carefully, and it has been clariﬁed that the shed vortices have a large
eﬀect on the resulting ﬂow ﬁelds and on the global performance [49, 50, 70].
3.1.3 Methods to improve Savonius turbine performance
Several propositions can be found in the literature to improve the performance of con
ventional Savonius rotors.
3.1.3.1 Deﬂector plate
Attempts to improve the performance of a Sshaped Savonius rotor by using a deﬂecting
plate placed in front of the rotor have been documented. Using a deﬂecting plate placed
on the retreating side of the blade it has been observed that the power coeﬃcient can
be increased relatively by about 20% [27, 43].
3.1.3.2 Double and three steps Savonius rotor
When considering only the starting torque, it can be seen that, for some directions of
the wind velocity, the starting torque of the standard Savonius conﬁguration would be
so low that the rotor could not start alone. It is the reason why many authors have
chosen to use a doublestep (see Fig. 3.6) and threestep Savonius rotor, where the
upper and the lower paddle pairs are set at 90
◦
to each other (double step), respectively
at 120
◦
for the three step rotor. The doublestep and threestep rotors are said to be
slightly superior to the corresponding singlestep turbine (conventional Savonius rotor)
in selfstarting, but lower for both torque and power characteristics [38, 68].
3.1.3.3 Guide vanes
In order to decrease the torque variation of the standard Savonius rotor and to improve
its starting capability, a new type of Savonius turbine, using three stages with 120 degree
bucket phase shift between the adjacent stages with and without guide vanes, has also
been investigated [39].
The results indicate that the static torque coeﬃcient of the one stage turbine without
guide vanes is periodic with a cycle of 180 degrees (depending on the number of stages)
and that its variation is very large. The lower values of static torque are observed in
the ranges of θ = 140
◦
− 170
◦
and 320
◦
− 350
◦
. Guide vanes can be used to increase
the static torque and decrease its ﬂuctuation. But, unlike the static characteristic, the
dynamic eﬀect depends on the wind speed, so that the maximum values of C
m
and C
p
CHAPTER 3. LITERATURE REVIEW 34
Figure 3.5: Doublestep Savonius rotor [68].
of the threestage rotor are much smaller than those of the onestage rotor. The guide
vanes increase the dynamic torque coeﬃcient for small λ values (0 < λ < 0.3), but lead
to a decrease in the torque for larger values (λ > 0.3).
Recently, one study was published concerning the three bucket Savonius rotor [36].
Unfortunately, the authors rely on an erroneous formulation to compute the rotor power
and torque. They consider:
P
rotor
= (1/2ρAV
2
1
)u (3.1)
and
T = 60P
rotor
/2πN (3.2)
where V
1
is the testsection inlet air velocity and u is the tip peripheral blade velocity.
As a consequence the computed power is the power available in the incoming wind and
not the mechanical power of the turbine. Consequently, the results are not analyzed
further.
3.1.3.4 Twistedblade Savonius rotor
Another investigation aims at exploring the feasibility of a twistedbladed Savonius
rotor for power generation (Fig. 3.6). A twisted blade integrated within a threebladed
rotor system has been tested in a low speed windtunnel, and its performance has
been compared with conventional semicircular blades (corresponding to a twist angle
of 0
◦
). Performance analysis has been made on the basis of starting characteristics, static
torque and rotational speed. Experimental evidence shows the potential of the twisted
blade rotor in terms of smooth running, higher eﬃciency and selfstarting capability
as compared to that of the conventional rotor [94, 98]. Semicircular blades correspond
CHAPTER 3. LITERATURE REVIEW 35
to zero twist angle. By increasing this angle, the performance of the Savonius rotor is
increased in its performance as shown in Fig. 3.7. It is also shown that a larger twist
angle is preferable for a lower wind velocity in order to produce maximum power and
better starting characteristics. A twist angle α = 15
◦
gives optimum performance at
low airspeeds of U = 6.5 m/s in terms of starting acceleration and maximum noload
speed. Such blades shows a maximum of C
p
= 13.99 at tip speed ratio of λ = 0.65 (i.e.,
at U = 8.23 m/s), whereas the semicircular blade α = 0
◦
shows a C
p
= 11.04 at λ = 1.
Figure 3.6: Twistedblade Savonius Rotor [98].
Figure 3.7: Power coeﬃcient for twistedblade Savonius rotor [98].
3.1.3.5 Savonius rotor using a guidebox tunnel
This publication aims at improving and adjusting the output power of a Savonius rotor
under various wind power. A guidebox tunnel is employed as the appropriate device
to achieve this purpose. The guidebox tunnel is like a rectangular box used as wind
passage, in which a test rotor is included as shown in Fig. 3.8. The area ratio between
CHAPTER 3. LITERATURE REVIEW 36
the inlet and exit of the guidebox is variable in order to adjust the inlet mass ﬂow rate
or input power. At ﬁrst, experiments have been conducted to ﬁnd the conﬁguration
that would provide the best relative performance. The experiments measure the static
torque of the ﬁxed rotor at any phase angle and the dynamic torque under rotation.
Figure 3.9 shows that the maximum rotor rotational speed is obtained for a guidebox
area ratio between 0.3 and 0.7. The resulting value of the output power coeﬃcient of the
rotor with guidebox tunnel using an area ratio of 0.43 increases considerably (by a factor
1.5 using three blades; by a factor 1.23 using two blades) compared to the conventional
design without guidebox (Fig. 3.9). The optimum spacing ratio between the rotor tip
and the side walls of the guidebox tunnel is around 1.4. The resulting increase in
performance seems promising, but the guidebox of course increases considerably the
system complexity [48].
Figure 3.8: Savonius rotor using a guidebox tunnel [48].
3.1.3.6 Modiﬁed Savonius rotor
The conventional Savonius rotor is made from two vertical halfcylinders running around
a vertical axis. A modiﬁed rotor (Fig. 3.10) has also been proposed, which is just a modi
ﬁcation of the Savonius rotor, using now three geometrical parameters: the main overlap
e, the secondary overlap e
, and the angle β between the paddles. The characteristic
curve of such a rotor (values of the static torque coeﬃcient C
m
vs. wind angle) are
presented in Fig. 3.11.
The results are relatively encouraging, since the new rotor induces maximal values
of the static torque much higher than those obtained with the conventional rotor. Nev
ertheless, it also introduces low and negative values of the torque, with a large angular
variation. Overall, the mean value of the torque is increased: C
m
= 0.48, i.e., 60% more
than for the conventional rotor. Further studies are necessary to reﬁne the analysis.
CHAPTER 3. LITERATURE REVIEW 37
Figure 3.9: Power coeﬃcient of Savonius rotor using a guidebox tunnel [48].
Table 3.1: Summary of Savonius turbine main modiﬁcations
Design Gain Description Comments
Deﬂector Plate 20% Not veriﬁcated No details since 1992
MultiSteps Good selfstarting For same aspect ratio Lower C
p
by 30%
Guide Vanes Depends on wind speed Bad for large λ Good stability
Twistedblade 27% High cost Good selfstarting capability
GuideBox Tunnel 50% (3 blades) Complex design Not practical
Modiﬁed Savonius 60% Large vibrations Not practical
3.1.4 Summary of Savonius turbine review
All the research projects discussed in the present review have tried to understand and
improve the performance (torque and eﬃciency) of the Savonius turbine, considering
either the conventional geometry or slight modiﬁcations. All the improvements have
been tested manually, by means of a tedious trialanderror analysis. Such studies are
going on. All proposed modiﬁcations have been summarized and discussed in Table 3.1.
It can be ﬁnally concluded that:
• There is a renewed interest for Savonius turbines in recent years, as shown in
Fig. 3.12;
• Depending on the retained deﬁnition, their eﬃciency can be indeed considered as
quite high [71];
CHAPTER 3. LITERATURE REVIEW 38
Figure 3.10: Modiﬁed Savonius Rotor [69].
Figure 3.11: Static torque for the modiﬁed Savonius rotor (θ = 90
◦
, β = 45
◦
, e/d =
1/6) [69].
• Contradictory observations can be found in the literature, both qualitatively and
quantitatively. Therefore, the results presented in this review must all be consid
ered with great caution;
• Many small modiﬁcations and improvements have been proposed in the litera
ture to improve global eﬃciency or some speciﬁc characteristics (e.g., selfstarting
conditions) as summarized in Table 3.1 ;
• Nevertheless, a real optimization of the system has not been realized up to now
and would therefore be very interesting. This will be the subject considered in
later chapters of this thesis.
CHAPTER 3. LITERATURE REVIEW 39
1970 1980 1990 2000 2010
Years
0
4
8
12
16
N
u
m
b
e
r
o
f
p
a
p
e
r
s
O
i
l
C
r
i
s
i
s
Figure 3.12: Savonius turbine publication statistics in international journals and con
ferences.
3.2 Wells Turbine
Let us switch now to the second conﬁguration discussed in this work.
3.2.1 Performance parameters of Wells turbine
Many papers cited in the literature [9, 21, 63, 141] concern the principle of operation
and factors controlling the performance of Wells turbines both experimentally and the
oretically. The performance of Wells turbine includes the power output, the pressure
drop across the rotor, the turbine eﬃciency and the operating range of the turbine.
Here, we should note that the performance of the wave power converter depends on
the energy absorption eﬃciency of the OWC, which is closely related to the pressure
diﬀerence across the turbine, as well as to the turbine eﬃciency. An unidirectional airﬂow
test rig [82] constructed to investigate experimentally the eﬀect of the ﬂow coeﬃcient φ,
on both the pressure drop coeﬃcient and the eﬃciency of a monoplane turbine and in
same time a theoretical investigation was made, at h = 0.62 for two values of rotor
solidity (s = 0.5 and s = 0.75). The results indicated that, a linear relationship between
the pressure drop across the turbine and the ﬂow coeﬃcient φ, exists, for both low or
high rotor solidity. It also indicated that the blade eﬃciency for blade proﬁle NACA
0021 increases with the ﬂow coeﬃcient up to a certain value, and then decreases. This
is due to separation of the ﬂow around the turbine blades. In the next sections we will
discuss in details the diﬀerent parameters aﬀecting the performance of Wells turbines.
CHAPTER 3. LITERATURE REVIEW 40
3.2.1.1 Solidity
The solidity of the turbine, s = zc/πr
t
(1 + h) (see Fig. 3.13) is a measure of airﬂow
blockage within the turbine. It is also a measure of the mutual interaction between the
blades and is an important design variable that aﬀects the selfstarting of the turbine [82,
90, 97, 93, 118, 140]. In Fig. 3.14 the eﬃciency, η is normalized with respect to the
corresponding two dimensional single airfoil eﬃciency, η
o
. The results indicate that at
small values of solidity, its impact is small. To obtain selfstarting capability of the
turbine at hub to tip h = 0.6, s > 0.51 is needed in [89] but only s > 0.45 in [73]. The
reduction in eﬃciency for a Wells turbine at high solidity is due to increased kinetic
energy losses at the exit associated with swirl. In addition, there could be signiﬁcant
three dimensional eﬀects near the hub where, at high solidity, the blades are close to
each other and may interact with the boundary layer on the hub. These interactions
lead to endwall losses on the surface of the hub and also on the blade surfaces near the
hub. These eﬀects are compounded by the fact that the blades near the hub are always
at a larger incidence that at the tip. In this work, solidity is constant and equal to
0.67, which corresponds to a number of blades equal to 8. This solidity value has been
retained since it has been proved in a number of studies to be highly suitable for Wells
turbines (for instance [106, 110]).
Rotorhub
onshaft
Forcesresolvedin
directionofrotation
Rotation
Sym. Airfoil
Blade
FA
FT
Chord(c)
s
p
a
n
(
b
)
r
t
rh
FA
Oscillating
flow
Figure 3.13: Axial and tangential forces acting on a Wells turbine.
3.2.1.2 Hubtotip ratio
The eﬀect of the hub to tip ratio (h = d
h
/d
t
or r
h
/r
t
, see Fig. 3.13) was studied in
several papers [82, 90, 97]. The results indicated that the eﬃciency for diﬀerent proﬁles
decreases with increasing the hub to tip ratio. From our personal experience the eﬀect
of hub to tip ratio on the turbine eﬃciency is rather complex. This is due to the inter
action between the blades in hub region, which leads to higher incident angles locally.
Therefore, the stall occurs earlier in hub region than in the tip region. Admittedly based
on smallscale tests, publications suggest that values of h ≈ 0.6 are recommended for
optimal design.
CHAPTER 3. LITERATURE REVIEW 41
0.2 0.4 0.6 0.8 1.0
0
0.2
0.4
0.6
0.8
1.0
0
s
Figure 3.14: Solidity eﬀect on the eﬃciency [82, 92, 97].
3.2.1.3 Aspect ratio
The aspect ratio, AR = span(b)/chord(c), inﬂuences the turbine eﬃciency and ﬂow
ratio at which the turbine stalls, as can be seen in Fig. 3.15. The data shown here
are from experiments conducted at a ﬁxed hub to tip ratio and ﬁxed solidity where the
variation in the aspect ratio was obtained by varying the chord length. This means
that Re= wc/ν changed during the test. Therefore, there is a certain inﬂuence of the
Reynolds number on this data. These results, again limited to smallscale tests, suggest
that aspect ratio and Reynolds number play a dominating eﬀect on the eﬃciency. The
primary eﬀect of reducing the aspect ratio is to increase the eﬃciency by delaying stall,
associated with the positive eﬀect obtained on the blades due to increased mass ﬂow
through the tip [82, 97]. The eﬀect of aspect ratio is however not clear in this study due
to varying Reynolds number.
3.2.1.4 Reynolds number
A Wells turbine is very sensitive to the Reynolds number (Re= wc/ν) like all con
ventional turbomachines, due to the impact of Reynolds number on the aerodynamic
around the airfoils of the turbine and to its contribution to stall [97].
3.2.1.5 Tip clearance
The tip clearance ratio (τ = tip clearance/chord(c)) is a very important parameter for
the performance of turbomachines. Several groups studied the eﬀect of tip clearance
on stall and eﬃciency of the Wells turbine.The results indicated that decreasing the tip
clearance advances the stall point but increases the eﬃciency of the turbine, due to the
reduced leakage losses. On the other hand, if the turbine has a large tip clearance, it
CHAPTER 3. LITERATURE REVIEW 42
AR
NACA 00150.670.73x10
NACA 00150.510.653x10
5
5
h
Re
0.4 0.5
0.6 0.7 0.8 0.9 1.0
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.1
0.2
0.3
S
Figure 3.15: Aspect ratio eﬀect on the eﬃciency [82, 92, 97].
operates without stall. There is no signiﬁcant advantages to increase tip clearance above
2%. Therefore, the recommended tip clearance is τ ≤ 2% [45, 46, 82, 97].
3.2.1.6 Inlet turbulence
Turbomachines are sensitive to the inlet ﬂow conditions, such as distortion of velocity
proﬁle and turbulence levels. Increase in turbulence levels can alter boundary layer
development by advancing transition of the boundary layer and delaying stall. The
performance of turbomachines can be improved by increasing the turbulence level at
the inlet up to 3% [82, 97]. Experimental results indicate that the performance of
the Wells turbine improves with an increase in turbulence but a signiﬁcant increase in
turbulence levels is required to produce any appreciable improvement in performance.
As a whole the Wells turbine is less sensitive to inlet turbulence compared to most
conventional turbomachines.
3.2.1.7 Blade shape
The airfoil thickness is very important because it determines the aerodynamics around
the airfoil, the stall point and the turbine weight. The eﬀect of airfoil thickness on
the aerodynamic performance can not be separated from the eﬀect of the Reynolds
number, since they contribute together to separation on the airfoil. The published results
indicated that NACA 0021 airfoil proﬁles (21% thickness) lead to the best performance
for conventional Wells turbines [93]. Thick airfoil blades are advantages to improve self
starting characteristics of the turbine [46, 91, 129, 130, 135].
CHAPTER 3. LITERATURE REVIEW 43
3.2.2 Unsteady ﬂow eﬀect on the performance of Wells turbine
It is generally accepted that the airﬂow frequency in a wave energy device is so small
(f
∗
< 1 Hz) that dynamic eﬀects are negligible. All the results shown above are based
on quasisteady assumptions or experiments conducted in unidirectional airﬂow test
rigs. Some groups studied Wells turbine under oscillating ﬂow (unsteady considers).
The results indicated a hysteretic eﬀect on the force coeﬃcients (Fig. 3.16). The hys
teretic eﬀects are caused by asymmetry in the boundary layer development on the blade
surfaces and oscillating motion of the wake, the extent of which can be appreciable at
low Reynolds numbers [82, 85, 97] (Fig. 3.17).
Dynamic stall is a process resulting from a series of events, which involves a hysteresis
loop in the airfoil lift curve and account for a higher maximum lift force than that
achieved in the static curve, during a cycle of pitching motion. Helicopter, turbine and
windmill blades may all be aﬀected by this phenomenon, which increases forces and
moments applied to the blade and its root, and reduces fatigue life. But the loop of
hysteretic curve of the dynamic stall is diﬀerent from the one of Wells turbine. So, the
mechanism of the hysteretic characteristics of Wells turbine is still unclear [47, 57].
The ﬂow structure around the blade of Wells turbines has been checked by some
researchers. Figure 3.18 shows the illustration of the ﬂow structure obtained by nu
merical simulation [61, 62]. At high angles of attack, a separation vortex (2) appears
on the blade suction surface on the hub side and reduces blade circulation, because
of the excessive angle of attack near the hub. A strong downward ﬂow (4) is induced
by the separation vortex near the trailing edge. It brings about the clockwise vortical
wake ﬂow (3a), which enlarges the ﬂow separation on the adjacent blade suction surface
(3b) [58, 59, 104, 103].
The intensity of the vortical ﬂow varies in the accelerating and the decelerating
ﬂow process. In the accelerating ﬂow stroke, as the blade circulation increases, vortices
opposite to the blade circulation are shed from the trailing edge. The stronger vortices
are shed at a larger radius because the blade circulation increases more than at a smaller
radius.
Then, the clockwise trailing vortices are generated. Therefore, the clockwise vortical
ﬂow is intensiﬁed by these vortices. In the decelerating ﬂow process where the blade
circulation decreases, the shed vortices are in the same direction of the blade circulation.
They form counterclockwise trailing vortices, which suppress the vortical wake ﬂow.
Since the stronger vortical wake ﬂow enlarges the separation on the suction surface of
the adjacent blade, the performance in the accelerating ﬂow process becomes lower than
in the decelerating ﬂow process [103]. This eﬀect can be reduced by decreasing the
solidity (Fig. 3.18). This observation can be explained, since the vortical wake ﬂow
enlarging ﬂow separation becomes more distant from the blade suction surface for the
case of low solidity [103].
CHAPTER 3. LITERATURE REVIEW 44
Figure 3.16: Eﬀect of unsteady ﬂow on Wells turbine performance [103].
3.2.3 Methods to improve Wells turbine performance
Due to the drawbacks of Wells turbine, like low eﬃciency, bad starting capability at low
solidity and low output power, many researchers tried to suggest some ideas to improve
the performance, like installing guide vanes [87, 111, 117], or improving proﬁles [127],
as discussed in the next section.
CHAPTER 3. LITERATURE REVIEW 45
C
A
0 5 10
15
1.5
3
s=0.57
s=0.48
Figure 3.17: Eﬀect of solidity on the hysteretic behavior [103].
Figure 3.18: Illustration of the ﬂow structure in blade suction side [103].
3.2.3.1 Guide vanes installation
Many papers [13, 33, 105, 109, 110, 112, 116, 121, 122] have demonstrated the useful
ness of 2D and 3D (twisted) guide vanes. The eﬀect of guide vanes (Fig. 3.19) has
been investigated experimentally and theoretically by testing a model under steady ﬂow
conditions. It is found that the running and starting characteristics of the Wells turbine
with guide vanes are superior to those without guide vanes (Fig. 3.20).
The results indicated in particular that the three dimensional guide vanes (variable
angles along the vane span) providing a constant rotor blade angle of attack with radius
lead to the best characteristics and are therefore recommended.
3.2.3.2 Selfpitchcontrolled blades
Experimental investigations were performed by model testing of the rotor with ﬁxed
blades under steady ﬂow conditions [32, 44, 56, 60, 99, 100, 101, 106, 120]. The turbine
CHAPTER 3. LITERATURE REVIEW 46
Figure 3.19: Installation of guide vanes [106].
Figure 3.20: Guide vanes eﬀect on the turbine performance [105].
blade is set on the hub by a pivot located near the leading edge that enables it to oscillate
between two prescribed setting angles of ±γ
b
(Fig. 3.21). As an airfoil set at a certain
angle of incidence experiences a pitching moment M about the pivot, the turbine blades
CHAPTER 3. LITERATURE REVIEW 47
can ﬂip by themselves between +γ
b
and −γ
b
according to the ﬂow direction. Therefore,
higher torque and eﬃciency are obtained for high ﬂow coeﬃcients, while performance
decreases for lower ﬂow coeﬃcient. The results indicated that the operating range
improved when increasing setting angle. since the incidence angle (θ = α−γ
b
) decreased.
On the other hand, the torque improved only in the high operating range compared with
conventional Wells turbine (γ
b
= 0).
Wells
Turbine
AirChamber
M
Pivot
u
b
b
Figure 3.21: Principle of using selfpitchcontrolled blades [106].
3.2.3.3 Lean blade (blade swept)
Experimental research on diﬀerent types of rotor blades has been conducted recently
to improve the aerodynamic performance of the Wells turbine, using lean blade (swept
blade) as shown in Fig. 3.22 for two diﬀerent airfoils (NACA 0015 and HSIM15262123
1576, Fig. 3.23), which can operate with wider operating range [108] and acceptable
power output and eﬃciency. A numerical study indicated that the comparison between
standard NACA 0015 unswept blade turbine rotor and the swept one by 30
◦
is very
diﬃcult for small ﬂow coeﬃcients. For high ﬂow coeﬃcients, however, the swept one is
better in term of eﬃciency. However, the results are poor concerning power output. On
the other hand, the airfoil HSIM152621231576 is better than both unswept and swept
standard NACA 0015. Therefore, the rotor blade geometry has a remarkable inﬂuence
on the turbine performance. Some rotor geometries give a considerably wider range of
ﬂow rates for high eﬃciency and acceptable power output. Others have higher peak
eﬃciency but a narrower range of ﬂow rates [8, 102, 126, 136].
3.2.3.4 End plate
The eﬀect of end plate (Fig. 3.24) on the turbine characteristics has been investigated
experimentally for diﬀerent plate sizes (a/c) by model testing under steady ﬂow condi
tions and compared with the classical Wells turbine [119, 123, 124, 125], without end
plate. Experiments indicate that the best geometry corresponds to a/c = 0.033. The
eﬀectiveness of the end plate has been checked by using CFD to get the optimal position
of the plate.
CHAPTER 3. LITERATURE REVIEW 48
Figure 3.22: Swept blade (lean blade) [8].
Figure 3.23: NACA 0015 and HSIM 152621231576 blades [8].
The forward type case shows the highest tangential force coeﬃcient, while the back
ward type has the lowest value. On the other hand, the values of forward type and
middle type are almost the same concerning axial force coeﬃcient, C
A
. Globally, the
highest eﬃciency was for the forward type position, leading to wider operating range
and higher turbine performance.
3.2.4 Multiplane Wells turbine
In wave energy devices, where the available pressure drop is higher than a monoplane
could accommodate, multiplane turbine must be used. Several investigations studied
the performance of multiplane Wells turbine and are reviewed in this section [14, 55, 80].
3.2.4.1 Twostage Wells turbine (biplane turbine)
All the previous theoretical and experimental results for the ﬂow ﬁeld around a Wells
turbine rotor indicate that a considerable amount of exit kinetic energy is lost with
the swirl component of the ﬂow velocity, at least in the absence of guide vanes. This
kinetic energy can be partly recovered by using a second stage of blades. Twostage
Wells turbine have been already investigated experimentally and theoretically [34], but
considering only symmetric airfoils, as shown in Fig. 3.25.
CHAPTER 3. LITERATURE REVIEW 49
Figure 3.24: Rotor blade with end plate: (a) Middle type; (b) Forward type; (c) Backward
type [124].
3.2.4.2 Twostage contrarotating Wells turbine
The ﬁrst commercial wave power station (capacity of 500 kW), called LIMPET (Land
Installed Marine Pneumatic Energy Transmitter) in UK is constructed with a twostage
contrarotating Wells turbine [6]. In this conﬁguration, the two rotors are installed as
in Fig. 3.26.
Some researchers studied this conﬁguration and made a comparison with the con
ventional biplane Wells turbine [17, 18, 24, 83]. Smallscale experiments on the contra
rotating Wells turbine have been conducted using constant ﬂow windtunnels. The
turbine tip diameters for these tests were 0.2 and 0.59 m, respectively, with nominal
Mach numbers of 0.2 and 0.23 and Reynolds numbers of 3 10
5
and 6.5 10
5
, respectively.
The results indicated that a contrarotating Wells turbine is aerodynamically more eﬃ
cient than the biplane Wells turbine and operate without stalling over wider ﬂow rates
CHAPTER 3. LITERATURE REVIEW 50
Rotation
Flow
Flow
DownstreamRotor
UpstreamRotor
Chord pitch
d
Figure 3.25: Two stage (biplane) Wells turbine.
Figure 3.26: Contrarotating Wells turbine [107].
than the biplane turbine. They concluded also that the downstream rotor must rotate
at higher speed than the upstream to achieve some improvement. The performance
of the contrarotating Wells turbine installed in the LIMPET wave power station was
compared to the predicted performance from theoretical analysis and model tests [24].
Figure 3.27 gives the nondimensional turbine torque against ﬂow coeﬃcient for both
rotors during both the intake and exhaust strokes. During exhaust, the results indi
cate that the LIMPET turbine stalls at approximately the same ﬂow coeﬃcient as the
constantﬂow model, but that the torque at stall is reduced by approximately 25% for
the exhaust stroke and 46% for the intake stroke. In addition to comparing exhaust and
intake it is also interesting to consider the relative performance of either rotor shown in
Fig. 3.27. During exhaust there is no measurable diﬀerence in stall between the rotors.
However, on intake, the downstream rotor has a noticeably higher torque coeﬃcient at
stall condition.
Figure 3.28 shows a plot of the LIMPET turbine eﬃciency with ﬂow coeﬃcient
during the intake and exhaust, together with turbine eﬃciency derived from constant
ﬂow model tests. During exhaust, at low ﬂow coeﬃcients, the LIMPET turbine appears
to have a higher eﬃciency than in constant ﬂow model tests. This is possibly due to
CHAPTER 3. LITERATURE REVIEW 51
Figure 3.27: Torque coeﬃcients for both LIMPET Rotors during intake and exhaust [24].
a lower drag coeﬃcient at higher Reynolds numbers. However, measurement errors
are magniﬁed at low torques and ﬂow coeﬃcients, so the data in this region should be
viewed cautiously. However, the early onset of stall experienced by the LIMPET turbine
reduces the peak eﬃciency of the turbine to approximately 52%, whilst simultaneously
reducing its eﬀective operating range. During intake, a higher ﬂow coeﬃcient is required
for the turbine lift force to overcome the drag force due to relatively smaller torques
being generated for the same ﬂow coeﬃcients. Smaller turbine torques during intake
cause a smaller increase in turbine eﬃciency with ﬂow coeﬃcient, resulting in a peak
eﬃciency of only approximately 42%.
Figure 3.28: Comparison of LIMPET and constant ﬂow model turbine eﬃciencies [24].
This study concluded that the contrarotating Wells turbine ﬁnally has a lower ef
ﬁciency than a biplane or monoplane Wells turbine with guide vanes. In addition,
a contrarotating Wells turbine requires an additional generator (or a gearbox to re
CHAPTER 3. LITERATURE REVIEW 52
verse the direction of rotation of one rotor), making it more complex and expensive to
implement than biplane and monoplane Wells turbines. Consequently, unless the per
formance of contrarotating Wells turbines can be signiﬁcantly improved, they are not
recommended over other Wells turbine variants. Obviously, contradictory information
are found in the literature concerning contrarotating conﬁgurations highlighting the
need for further studies.
3.2.5 Summary of Wells turbine review
All the studies discussed in the present review have tried to understand and improve
the performance of the Wells turbine, considering either the conventional geometry or
slight modiﬁcations. All the improvements have been tested manually, by means of a
tedious trialanderror analysis. Such studies are going on. All proposed modiﬁcations
have been summarized and discussed in Table 3.2. It can be ﬁnally concluded that:
Table 3.2: Main modiﬁcations allowing to improve the performance of Wells turbines
Design Gain Description and Comments
Contrarotating [14] Improve eﬃciency by 5% Double shaft, complex
Pitch setting [106] Improve eﬃciency by 7% For positive small angles, complex
Guide vanes [33] Improve eﬃciency by 5% Smaller operating range
End plate [124] Improve eﬃciency by 5% Only for 0.2 ≤ φ ≤ 0.25
Multistage
∗
[81] Wider operating range Reduce eﬃciency by 10%
Multistage
∗∗
[74] Improve eﬃciency by 2% Small parameter space
∗
Symmetric airfoils
∗∗
NonSymmetric airfoils
1970 1980 1990 2000 2010
Years
0
4
8
12
16
N
u
m
b
e
r
o
f
p
a
p
e
r
s
O
i
l
c
r
i
s
i
s
Figure 3.29: Wells turbine publication statistics in international journals and confer
ences.
CHAPTER 3. LITERATURE REVIEW 53
• There is a renewed interest for Wells turbines in recent years, as shown in Fig. 3.29;
• Contradictory observations can be found in the literature, both qualitatively and
quantitatively. Therefore, the results presented in this review must all be consid
ered with great caution, in particular concerning the contrarotating design ;
• Many small modiﬁcations and improvements have been proposed in the literature
to improve global performance, as summarized in Table 3.2;
• Nevertheless, a real optimization of the system has not been realized up to now
and would therefore be very interesting. This will be the subject considered in
later chapters of this thesis.
3.3 Conclusions
In spite of many published studies, it still seems possible to increase considerably the
performance of both Savonius and Wells turbines by relying on modern computational
methods, as demonstrated in later chapters. The needed methodology for this purpose
is described in the next two chapters.
Chapter 4
Optimization
4.1 Introduction
Optimization pervades the ﬁelds of science, engineering, and business. In physics, many
diﬀerent optimal principles have been enunciated, describing natural phenomena in the
ﬁelds of optics or classical mechanics. Statistics treats various principles termed ”max
imum likelihood,” ”minimum loss,” and ”least squares,” and business makes use of
”maximum proﬁt,” ”minimum cost,” ”maximum use of resources,” ”minimum eﬀort,”
in its eﬀorts to increase proﬁts. A typical engineering problem can be posed as follows:
A process can be described by representative equations (or perhaps solely by experimen
tal data). You have a single performance criterion in mind such as maximum eﬃciency
or minimum cost. The goal of optimization is to ﬁnd the values of the variables in
the process that yield the best value of the performance criterion. A tradeoﬀ usually
exists. Typical problems in engineering process design or plant operation have many
(possibly an inﬁnite number) of solutions. Optimization is concerned with selecting the
best among the entire set by eﬃcient quantitative methods, thanks to computers and
associated software, which make the necessary computations feasible and cost eﬀective.
To obtain useful information using computers, however, requires (1) a critical analysis
of the process or design, (2) insight about what the appropriate performance objec
tives are (i.e., what is to be accomplished), (3) use of past experience, sometimes called
engineering judgment [35], and (4) suitable methods and algorithms.
Until recently, the denomination “optimization” was mostly used in the engineer
ing literature to describe a trialanderror, manual procedure (undoubtedly related to
optimization, but in a very minimalist sense), at the diﬀerence of a real, mathemati
cal optimization. This is now changing rapidly. In the present project, mathematical
optimization will be considered to obtain the optimal shape geometry. Hence, for us,
optimization means ”the design (or operation) of a system or process to make it as
good as possible in some deﬁned sense”. As a consequence, the best possible solution
constrained by appropriate conditions should be ideally found, and not simply a ”bet
ter” one [132]. Another deﬁnition is given by [2], which states that optimization is the
process of obtaining the ”best”, if it is possible to measure and change what is ”good”
or ”bad”. The deﬁnition in [96] is that the optimization theory is a body of mathe
54
CHAPTER 4. OPTIMIZATION 55
matical results and numerical methods for ﬁnding and identifying the best candidate
from a collection of alternatives without having to explicitly enumerate and evaluate all
possible alternatives.
In practice, one wishes the ”most” or ”maximum” (e.g., salary) or the ”least” or
”minimum” (e.g., expenses). Therefore, the word ”optimum” is taken to mean ”maxi
mum” or ”minimum” depending on the circumstances; ”optimum” is a technical term
which implies quantitative measurement and is a stronger word than ”best” which is
more appropriate for everyday use. Likewise, the word ”optimize”, which means to
achieve an optimum, is a stronger word than ”improve”. A computer is the perfect
tool for optimization as long as the idea or variable inﬂuencing the idea can be input
in electronic format. The process of optimization lies at the root of engineering, since
the classical function of the engineer is to design new, better, more eﬃcient, and less
expensive systems as well as to devise plans and procedures for the improved operation
of existing systems. The power of optimization methods to determine the best case
without actually testing all possible cases comes through the use of a modest level of
mathematics and at the cost of performing iterative numerical calculations using clearly
deﬁned logical procedures or algorithms implemented on computing machines.
4.2 Optimization uncertainty
For practical problems, optimization does not deliver a ”solution truth” because of the
uncertainty that exists in the mathematical and physical representation of the process
or the data used to model it. Engineers have to use their own judgment in applying
optimization techniques to problems that have considerable uncertainty associated with
them, both from the standpoint of accuracy and from the fact that operating parameters
are not always static. In some cases, it may be possible to carry out a ﬁrst analysis via
deterministic optimization and then add on stochastic features to the analysis to yield
quantitative predictions of the degree of uncertainty. Whenever the model of a process is
idealized and the input and parameter data only known approximately, the optimization
results must be treated judiciously. They can provide for instance upper limits on
expectations. Another way to evaluate the inﬂuence of uncertain parameters in optimal
design is to perform a sensitivity analysis. It is possible that the optimum value of a
process variable is unaﬀected by certain parameters (low sensitivity); therefore, having
precise values for these parameters will not be crucial to ﬁnding the true optimum [35].
Furthermore, optimization can only deliver accurate results when the quality of a speciﬁc
design can be measured accurately. For our problem, this issue will be considered in the
next chapter.
4.3 How can we achieve optimization?
Most reallife problems have several solutions and occasionally an inﬁnite number of
solutions may be possible. Assuming that the problem at hand admits more than one
solution, optimization can be achieved by ﬁnding the best solution of the problem in
CHAPTER 4. OPTIMIZATION 56
terms of some performance criterion. Several general approaches to optimization are
available, as follows:
• Analytical methods
• Graphical methods
• Experimental methods
• Numerical methods
Analytical methods are based on the classical techniques of diﬀerential calculus. In these
methods the maximum or minimum of a performance criterion is determined by ﬁnding
the values of parameters x
1
, x
2
, ..., x
n
that cause the derivatives of f(x
1
, x
2
, ..., x
n
) with
respect to x
1
, x
2
, ..., x
n
to assume zero values. The problem to be solved must obviously
be described in mathematical terms before the rules of calculus can be applied. The
method need not entail the use of a digital computer. However, it cannot be applied to
highly nonlinear problems or to problems where the number of independent parameters
exceedingly grows.
A graphical method can be used to plot the function to be maximized or minimized
if the number of variables does not exceed a few. If the function depends on only one
variable, say, x
1
, a plot of f(x
1
) versus x
1
will immediately reveal the maxima and/or
minima of the function. Similarly, if the function depends on only two variables, say, x
1
and x
2
, a set of contours can be constructed. A contour plot, like a topographical map of
a speciﬁc region, will reveal readily the peaks and valleys of the function. Unfortunately,
the graphical method is of limited usefulness since in most practical applications the
function to be optimized depends on many variables.
The optimum performance of a system can sometimes be achieved by direct experi
mentation. In this method, the system is set up and the process variables are adjusted
one by one and the performance criterion is measured in each case. This method may
lead to optimum or near optimum operating conditions. However, it can lead to unre
liable results since in certain systems, two or more variables interact with each other,
and must be adjusted simultaneously to yield the optimum performance criterion.
The most important general approach to optimization is based on numerical methods
(real optimization). In this approach, iterative numerical procedures are used to gen
erate a series of progressively improved solutions to the optimization problem, starting
with an initial estimate for the solution. The process is terminated when some conver
gence criterion is satisﬁed. For example, when changes in the independent variables or
the performance criterion from iteration to iteration become insigniﬁcant. Numerical
methods can be used to solve highly complex optimization problems of the type that
cannot be solved analytically. Furthermore, they can be readily programmed on digital
computers. The discipline encompassing the theory and practice of numerical optimiza
tion methods has come to be known as mathematical programming [2], covering:
• Linear programming
CHAPTER 4. OPTIMIZATION 57
• Integer programming
• Quadratic programming
• Nonlinear programming
• Dynamic programming
Each one of these branches of mathematical programming is concerned with a speciﬁc
class of optimization problems. The diﬀerences among them will be illustrated in the
nest sections.
4.4 Structure of optimization problems
Although the application problems discussed in the previous section originate from rad
ically diﬀerent sources and involve diﬀerent systems, at their root they have a remark
ably similar form. All ﬁve can be expressed as problems requiring the minimization
of a realvalued function f(x) of an Ncomponent vector argument x = (x
1
, x
2
, ..., x
N
)
whose values are restricted to satisfy a number of realvalued equations h
k
(x) = 0, a
set of inequalities g
j
(x) ≥ 0, and the variable bounds x
(U)
i
≥ x
i
≥ x
(L)
i
. In subsequent
discussions we will refer to the function f(x) as the objective function, to the equations
h
k
(x) = 0 as the equality constraints, and to the inequalities g
j
(x) ≥ 0 as the inequality
constraints. For our purposes, these problem functions will always be assumed to be
real valued, and their number will always be ﬁnite. Optimization then means:
Minimize or maximize f(x)
Subject to h
k
= 0 k = 1, ..., K
g
x
≥ 0 j = 1, ..., J
x
(U)
i
≥ x
i
≥ x
(L)
i
i = 1, ..., N
This general problem is called the constrained optimization problem. The problem
in which there are no constraints, that is, unconstrained optimization problem, would
correspond to
J = K = 0
and
x
(U)
i
= −x
(L)
i
= ∞ i = 1, ..., N
but it is almost never found in engineering.
CHAPTER 4. OPTIMIZATION 58
Optimization problems can be classiﬁed further based on the structure of the func
tions f, h
k
, and g
j
and on the dimensionality of x. Unconstrained problems in which
x is a onecomponent vector are called singlevariable problems and form the simplest.
Constrained problems in which the function h
k
and g
j
are all linear are called linearly
constrained problems. This subclass can further be subdivided into those with a linear
objective function f and those in which f is nonlinear. The category in which all problem
functions are linear in x includes problems with continuous variables, which are called
linear programs, and problems in integer variables, which are called integer programs.
Problems with nonlinear objective and linear constraints are sometimes called linearly
constrained nonlinear programs. This class can further be subdivided according to the
particular structure of the nonlinear objective function. If f(x) is quadratic, the problem
is a quadratic program; if it is a ratio of linear functions, it is called a fractional linear
program; and so on. Subdivision into these various classes is worthwhile because the
special structure of these problems can be sometimes eﬃciently exploited in devising
solution techniques [2]. Note that, in all this document, only direct optimization is con
sidered. Possibilities oﬀered by inverse design (based on solving the adjoint problem)
are not discussed.
4.5 Types of mathematical programming
Several branches of mathematical programming were enumerated in Section 4.3. Each
one of these branches consists of the theory and application of a collection of optimization
techniques that are suited to a speciﬁc class of optimization problems. The diﬀerences
among the various branches of mathematical programming are closely linked to the
structure of the optimization problem and to the mathematical nature of the objective
and constraint functions.
4.5.1 Linear programming (LP)
If the objective and constraint functions are linear and the variables are constrained to
be positive, optimization can be readily achieved by using some powerful LP algorithms.
4.5.2 Integer programming (IP)
In certain linear programming problems, at least some of the variables are required to
assume only integer values. This restriction renders the programming problem nonlinear
and more complex.
4.5.3 Quadratic programming (QP)
If the constraints are linear and the objective function is quadratic, such an optimization
problem is said to be a quadratic programming (QP) problem.
CHAPTER 4. OPTIMIZATION 59
4.5.4 Nonlinear programming (NP)
In nonlinear programming problems, the objective function and usually the constraint
functions are nonlinear. This is the most general branch of mathematical programming
and, in eﬀect, LP and QP can be considered as special cases of nonlinear programming.
The choice of an optimization algorithm depends on the mathematical behavior and
structure of the objective function. Rarely, the objective function is a well behaved
nonlinear function and all that is necessary is a general purpose, robust, and eﬃcient
algorithm. For many applications, however, specialized algorithms exist which are more
eﬃcient or more robust [19]. This is clearly the case for the problems considered in the
present thesis.
4.5.5 Dynamic programming (DP)
In some applications, a series of decisions must be made in sequence, where subsequent
decisions are inﬂuenced by earlier ones. In such applications, a number of optimizations
have to be performed in sequence and a general strategy may be required to achieve an
overall optimum solution.
4.6 Requirements for optimization
To apply the mathematical results and numerical techniques of optimization theory
to real engineering problems, it is necessary to clearly delineate the boundaries of the
engineering system to be optimized, to deﬁne the quantitative criterion on the basis of
which candidates will be ranked to determine the ”best”, to select the system variables
that will be used to characterize or identify candidates, and to deﬁne a model that
will express the manner in which the variables are related. This composite activity
constitutes the process of formulating the engineering optimization problem. Good
problem formulation is the key to the success of an optimization study [96].
4.6.1 Deﬁning the system boundaries
Before undertaking any optimization study, it is important to clearly deﬁne the bound
aries of the system under investigation. They serve to isolate the system from its sur
roundings, because, for purpose of analysis, all interactions between the system and its
surroundings are assumed to be frozen. Nonetheless, since interactions always exist, the
act of deﬁning the system boundaries is the ﬁrst step in the process of approximating
the real system. In many situations it may turn out that the initial choice of boundary
is too restrictive. To fully analyze a given engineering system, it may be necessary to
expand the system boundaries and to include other subsystems that strongly aﬀect the
operation of the system under study.
CHAPTER 4. OPTIMIZATION 60
4.6.2 Performance criterion
Given that we have selected the system of interest and have deﬁned its boundaries,
we next need to select a criterion on the basis of which the performance or design of
the system can be evaluated, so that the best design or set of operating conditions
can be identiﬁed. There is a considerable choice in the precise deﬁnition of such a
criterion: minimum production time, maximum production rate, minimum energy uti
lization, maximum torque, minimum weight, and so on.
4.6.3 Independent variables
The third key element in formulating a problem for optimization is the selection of the
independent variables that are adequate to characterize the possible candidate designs or
operating conditions of the system. There are several factors to be considered in selecting
the independent variables. First, it is necessary to distinguish between variables whose
values are amenable to change and variables whose values are ﬁxed by external factors,
lying outside the boundaries selected for the system in question. Furthermore, it is
important to diﬀerentiate between system parameters that can be treated as ﬁxed and
those that are subject to ﬂuctuations inﬂuenced by external and uncontrollable factors.
Clearly, variations in these key system parameters must be taken into account in the
formulation of the problem. Second, it is important to include in the formulation all
the important variables that inﬂuence the operation of the system or aﬀect the design
deﬁnition. For instance, if in the design of a gas storage system we include the height,
diameter, and wall thickness of a cylindrical tank as independent variables but exclude
the possibility of using a compressor to raise the storage pressure, we may well obtain a
very poor design. For the selected ﬁxed pressure, we would certainly ﬁnd the leastcost
tank dimensions. However, by including the storage pressure as an independent variable
and adding the compressor cost to our performance criteria, we could obtain a design
with a much lower overall cost because of a reduction in the required tank volume.
Thus, the independent variables must be selected so that all important alternatives are
included in the formulation. Finally, another consideration in the selection of variables
is the level of detail to which the system is considered. While it is important to treat
all key independent variables, it is equally important not to obscure the problem by the
inclusion of a large number of ﬁne details of subordinate importance. A good rule is
to include only those variables that have a signiﬁcant impact on the composite system
performance criterion.
4.6.4 System model
Once the performance criterion and the independent variables have been selected, the
next step in problem formulation is to assemble the mathematical and physical models
that describe the manner in which the problem variables are related and the way in
which the performance criterion is inﬂuenced by the independent variables. In principle,
optimization studies may sometimes be performed by experimenting directly with the
CHAPTER 4. OPTIMIZATION 61
real system. Thus, the independent variables of the system or process may be set to
selected values, the system operated under those conditions, and the system performance
index evaluated using the observed performance. The optimization methodology would
then be used to predict improved choices of the independent variable values and the
experiments continued in this fashion. In practice, most optimization studies are carried
out with the help of a representation of the real system, called a model. Models are
typically used in engineering design because they oﬀer the cheapest and fastest way of
studying the eﬀects of changes in key design variables on system performance. For ﬂow
optimization the NavierStokes equations constitute the natural model.
4.7 Optimization methods
In this work we consider methods that iteratively produce estimates of x
∗
(optimum
solution), that set of design variables that causes f(x) to take on its optimum value.
The methods that have been devised for the solution of this problem can be classiﬁed
into three broad categories [96] based on the type of information that must be supplied
by the user:
• Directsearch methods, which use only function values;
• Gradient methods, which require estimates of the ﬁrst derivative of f;
• Secondorder methods, which require estimates of the ﬁrst and second derivatives
of f.
No single method can be expected to uniformly solve all problems with equal ef
ﬁciency. For instance, in some applications available computer storage is limited; in
others, function evaluations are very time consuming; in still others, high accuracy in
the ﬁnal solution is desired. In many applications it is either impossible or very time con
suming to obtain the derivatives. For our problem, it is almost impossible and extremely
time consuming to estimate the function derivatives. Furthermore, function evaluations
rely again on computerbased simulations, and are therefore noisy and of limited ac
curacy. As a consequence, only Genetic Algorithms (a stochastic direct methods) are
discussed further.
4.8 Evolutionary Algorithms
Evolutionary Algorithms (EA) are computer programs that attempt to solve complex
problems by mimicking the processes of Darwinian evolution (e.g., [76]). In an EA
a number of artiﬁcial creatures search over the space of the problem. They compete
continually with each other to discover optimal areas of the search space. It is hoped
that over time the most successful of these creatures will evolve to describe the optimal
solution. The artiﬁcial creatures in EAs, known as individuals, are typically represented
by ﬁxed length strings or vectors. Each individual encodes a single possible solution to
CHAPTER 4. OPTIMIZATION 62
the problem under consideration. For example, in order to construct an EA to search
the conformation space of a molecule, each angle of rotation around a ﬂexible bond
could be encoded as a real number. Concatenating these numbers gives a string which
can be used within an EA. Thus, each individual would encode a speciﬁc set of torsion
angles. EAs manipulate pools or populations of individuals. The EA is started with
an initial population of size N comprising random or quasirandom individuals. Every
individual is then assigned a ﬁtness value. To generate a ﬁtness score the individual
is decoded to produce a possible solution to the problem. The value of this solution is
then calculated using the ﬁtness function. Population members with high ﬁtness scores
therefore represent better solutions to the problem than individuals with lower ﬁtness
scores. Following this initial phase the main iterative cycle of the algorithm begins.
Using mutation (perturbation) and recombination operators, the N individuals in the
current population produce n children according to a survival probability ratio. The n
children are assigned ﬁtness scores. A new population of N individuals is then selected
from the N individuals in the current population and the n children. This new popu
lation becomes the current population and the iterative cycle is repeated. The survival
of the ﬁttest is employed and individuals compete against each other. The selection is
applied either when choosing individuals to become parent of children or when choos
ing individuals to form a new population. There have been three main independent
implementation instances of EAs: Genetic Algorithms (GAs), ﬁrst developed by Hol
land (1975) and thoroughly reviewed by Goldberg (1989); evolution strategies (ESs),
developed in Germany by Rechenberg (1973) and Schwefel (1981); and evolutionary
programming (EP), originally developed by L. J. Fogel et al.(1966) and subsequently
reﬁned by D. B. Fogel (1995). Each of these three algorithms has been proved capable of
yielding approximately optimal solutions given complex, multimodal, nondiﬀerential,
and discontinuous search spaces. In the present project, only GA have been employed
and therefore described now.
4.8.1 Genetic Algorithm (GA)
This is the most popular type of EA. One seeks the solution of a problem in the form of
strings of numbers, by applying operators such as recombination and mutation [114]. A
path through the components of the GA is shown as a ﬂowchart in Fig. 4.1. Each block
in this ”big picture” overview is discussed in detail in what follows.
The canonical GA encodes the problem within binary string individuals. Nowadays,
real or doubleprecision representations are mostly used for engineering problems. Evo
lutionary pressure is applied in the steps of the iterations, where the stochastic technique
of roulette wheel parent selection is used to pick parents for the new population. The
concept is as follow:
1. A population of N random or quasirandom individuals is initialized
2. Fitness scores are assigned to each individual
CHAPTER 4. OPTIMIZATION 63
Definecostfunction,cost,variables
SelectGA parameters
Generateinitialpopulation
Decodechromosomes
Findcostforeachchromosome
Selectmates
Mating
CrossoverandMutation
ConvergenceCheck
Done
Figure 4.1: Flowchart of a Genetic Algorithm.
3. Using roulette wheel parent selection N/2 pairs of parents are chosen from the
current population to form a new population.
4. With probability Pc (crossover), children are formed by performing crossover on
the N/2 pairs of parents. The children replace the parents in the new population.
5. With probability Pm (mutation), mutation is performed on the new population.
6. The new population becomes the current population.
7. If the termination conditions are satisﬁed exit, otherwise go to step 3.
In Fig. 4.2, selection based on roulette is expained using a trivial example with a
population of ten individuals. Each individual is assigned a sector of a roulette wheel
that is proportional to its ﬁtness and the wheel is spun to select a parent.
While selection is random and any individual has the capacity to become a parent,
selection is clearly biassed towards ﬁtter individuals. Parents are not required to be
unique and, in each iteration, ﬁt individuals may produce many oﬀsprings. From a
CHAPTER 4. OPTIMIZATION 64
Figure 4.2: Example of selection based on the roulettewheel for ten individuals.
population of size N, N/2 pairs of parents are chosen. These parents initiate a new
population. With probability Pc each pair is recombined using the crossover operator
to produce a pair of children. This cut and splice operator is illustrated in Fig. 4.3.
Crossoverpoint
Parents
Children
Figure 4.3: Example of crossover.
A cross point is selected at random. Each child is identical to one parent before
the cross point and identical to the other after the cross point. The child individuals
then replace their parents in the new population. Following crossover, mutation is
applied to all or some individuals in the new population. The probabilities of mutation
and crossover, Pm and Pc are parameters of the algorithm and must be set by the
user. With probability Pm, each bit on every string is inverted or modiﬁed. The new
population then becomes the current population and the cycle is repeated until some
termination criteria are satisﬁed. The algorithm typically runs for some ﬁxed number
of iterations, or until convergence is detected within the population.
CHAPTER 4. OPTIMIZATION 65
Many GAs applied to real world problems bear only a passing resemblance to the
canonical GA, and GAs are best viewed as a paradigm for evolutionary search, rather
than a speciﬁc algorithm. The binary encoding is often inappropriate for many problems
and may be extended to nonbinary representations. Successful GAs have used integer
string individuals or even more general representations such as tree and matrix struc
tures. Specialized crossover operators have been devised to handle unusual encodings.
In order to increase program eﬀectiveness hybrid GAs mix problemspeciﬁc operations
with crossover and mutation. Selection pressure is deﬁned as the relative probability
that the ﬁttest individual in the population will be chosen as a parent relative to an
individual of average ﬁtness. Too high a selection pressure and a GA will rapidly con
verge to a suboptimal solution. While encouraging search, a low selection pressure can
result in a GA taking an inordinate time to converge. In order to control selection pres
sure within a GA ﬁtness values are often rescaled when applying roulette wheel parent
selection. One problem with a canonical GA is that there is no guarantee that good
individuals will survive from one iteration to the next. Not all algorithms produce an
entirely new population at each iteration. An elitist strategy involves copying the best
individuals unchanged from the current population to the new population, as done in
the present work. In a steadystate GA, each iteration involves the application of one
crossover or mutation operator and only one or two new individuals are added to the
population, usually replacing the worst individuals. As a whole, GAs have proved to be
the most popular EAs. They provide an eﬃcient a simple framework for attempting to
solve complex search problems and have been widely applied.
4.8.2 Advantage and disadvantage of Genetic Algorithm
Some of the advantages of a GA include that it
• Optimizes with continuous or discrete variables,
• Does not require any derivative information,
• Simultaneously searches from a wide sampling of the cost surface,
• Deals with a large number of variables,
• Is well suited for parallel computers,
• Optimizes variables with extremely complex cost surfaces (they can jump out of
a local minimum),
• Provides a list of optimum variables, not just a single solution,
• Works with numerically generated data, experimental data, or analytical functions.
These advantages produce stunning results when traditional optimization approaches
fail miserably. Of course, the GA is not the best way to solve every problem. For
instance, the traditional methods have been tuned to quickly ﬁnd the solution of a well
CHAPTER 4. OPTIMIZATION 66
behaved convex analytical function of only a few variables. For such cases the calculus
based methods by far outperform the GA, quickly ﬁnding the minimum while the GA
is still analyzing the costs of the initial population. However, many realistic problems
do not fall into this category.
4.9 Conclusions
For the optimization problem considered in this work:
• Function evaluations rely on a complex computation with a limited accuracy,
• Concurrent objectives will be considered,
• Local minima are often expected,
• Results for non optimal conﬁgurations may be nevertheless interesting,
• Parallel computers are available.
For all these reasons, GA appears to be ideally suited and will always be used in what
follows. Now, it is essential to compute the cost function as eﬃciently and accurately
as possible. This is the subject of the next chapter.
Chapter 5
Numerical methods and algorithms
5.1 Introduction
Accurate computer simulations of ﬂuid ﬂows involve a wide range of issues, from grid
generation to turbulence modelling to the applicability of various simpliﬁed forms of the
NavierStokes equations. Many of these issues are not addressed at all in this thesis,
like acoustics or reacting ﬂows [3]. Instead, we focus on selected numerical issues,
with emphasis on ﬁnitevolume solutions of the NavierStokes equations, coupled with
optimization to improve wind and wave energy turbines. We present in this chapter a
foundation for developing, analyzing, and understanding such methods.
5.2 CFD & Optimization
Computational Fluid Dynamics (CFD) deals with the numerical analysis of complex
ﬂows. Despite impressive progress in recent years, CFD remains an imperfect tool in
the comparatively mature discipline of ﬂuid dynamics, partly because electronic digital
computers have been in widespread use for only thirty years or so. The NavierStokes
equations, which govern the motion of a Newtonian viscous ﬂuid were formulated well
over a century ago. The most straightforward method of attacking any ﬂuid dynamics
problem is to solve these equations for appropriate boundary conditions. Analytical
solutions are few and trivial and, even with today’s supercomputers, numerically exact
solution of the complete equations for the threedimensional, timedependent motion
of turbulent ﬂow is prohibitively expensive except for basic research studies in simple
conﬁgurations at low Reynolds numbers. Therefore, the straightforward approach is
still impracticable for engineering purposes. Moreover, CFD will be considered for
optimization in this work.
Optimization methods allowing to identify a constrained, best possible solution have
been known for a long time, but have not permeated all engineering disciplines yet. Con
cerning more speciﬁcally ﬂuid dynamics, the ﬁrst applications of optimization are found
for aeronautical problems, in particular to improve wing proﬁle and ﬂight properties
(typically, reduce drag). This is a problem with a high addedvalue and involves only
68
CHAPTER 5. NUMERICAL METHODS AND ALGORITHMS 69
the basic equations of ﬂuid dynamics (Euler or NavierStokes equations, depending on
the investigated properties). This explains why most available books and articles dealing
with optimization relying on evaluations obtained by Computational Fluid Dynamics
concern such situations. Even then, the number of such books and review articles re
mains quite limited [132]. In our group, a considerable experience is available concerning
such mathematical optimization relying on evaluations based on Computational Fluid
Dynamics. We therefore employ our own optimization library, OPAL (for OPtimization
ALgorithms), containing many diﬀerent optimization techniques. Diﬀerent CFD solvers
(inhouse codes, ANSYSFluent, ANSYSCFX) have been coupled in the past with this
optimizer. It has already been employed successfully to improve a variety of applications
like for instance heat exchangers [40] or burners [52].
5.3 Computational procedure
The optimization procedure can only deliver the right solution, if all evaluations relying
on Computational Fluid Dynamics indeed lead to an accurate ﬂow description. From
the literature it is known that an accurate CFD simulation of the ﬂow around a Savonius
turbine or Wells turbine is a particularly challenging task, mainly due to its highly time
dependent nature and to the fact that ﬂow separation plays an important role for the
eﬃciency of the system. It is therefore necessary to check the full CFD procedure with
great care. Afterwards, the resulting methodology must be validated.
5.3.1 Preprocess: geometry & grid generation
Note ﬁrst that both problems considered in this thesis are indeed twodimensional in
space, allowing an easier representation and discretization. The ﬂow is timedependent
for Savonius turbine, while a steady solution is suﬃcient for Wells turbine.
5.3.1.1 Savonius turbine: size of computational domain
The appropriate size of the computational domain has ﬁrst been investigated. It must
be indeed checked that this size does not impact the results of CFD. A computational
domain of increasing dimensions (square domain of size 2L × 2L, suitably normalized
by the rotor radius R, see Fig. 5.1) has been considered in the CFD computation. It
is easy to notice from Fig. 5.1 that the three smaller domains are associated with a
large variation of the torque coeﬃcient. On the other hand, the remaining three (larger
domains) lead to a nearly constant value, with a relative variation of the output quantity
below 1.1%. This demonstrates that the computational domain should extend at least
over 20 times the rotor radius in each direction. In a smaller domain, the boundary
conditions inﬂuence the ﬂow results in an inappropriate manner. Finally, the domain
marked in Fig. 5.1 has been retained for all further Savonius computations in this work.
CHAPTER 5. NUMERICAL METHODS AND ALGORITHMS 70
R
2L
2
L
Turbine
Computational
domain
Employeddomain
0 10 20 30 40
Relative size of the computational
domain (L/R)
0
0.4
0.8
1.2
1.6
T
o
r
q
u
e
c
o
e
f
f
i
c
i
e
n
t
(
C
m
)
Figure 5.1: Size of the computational domain and impact on the torque coeﬃcient
5.3.1.2 Savonius turbine mesh independence
Several diﬀerent twodimensional grids of increasing density and quality, composed of
5 400 up to 120 000 cells, have been tested for a conventional twoblade Savonius turbine
with obstacle and deﬂector plates, using a representative example of the target solutions.
It is easy to notice from Fig. 5.2 that the ﬁve coarsest grids are associated with a large
variation of the torque coeﬃcient. On the other hand, all remaining grids employing
more than 80 000 cells lead to a relative variation of the output quantity below 1.3%.
Since the cost of a CFD evaluation obviously increases rapidly with the number of grid
cells, the intermediate grid range between 85 000 and 100 000 cells has been retained for
all further results shown in the present work concerning Savonius turbines.
5.3.1.3 Wells turbine: size of computational domain
The mutual interaction between the blades constrains the size of the computational
domain (Fig. 5.3) since only a single blade is considered. The appropriate size of the
computational domain has been selected in the spanwise direction by using constant
solidity and periodic boundary condition for both sides of the domain. In the axial
direction recommendation from the literature have been implemented (s = 0.67, see
section 3.2.1.1).
5.3.1.4 Wells turbine mesh independence
Corresponding results are shown in Fig. 5.4. Several diﬀerent twodimensional grids of
increasing density and quality, composed of 12 200 up to 108 000 cells, have been tested
for the baseline, nonsymmetric blade conﬁguration NACA 2421. All other parameters
CHAPTER 5. NUMERICAL METHODS AND ALGORITHMS 71
0 40000 80000 120000
Number of cells
0
0.1
0.2
0.3
0.4
T
o
r
q
u
e
c
o
e
f
f
i
c
i
e
n
t
(
C
m
)
l=1
X
d1
/R= 0.597
Y
d1
/R= 1.191
X
d2
/R= 0.516
Figure 5.2: Gridindependence study for the torque coeﬃcient
Figure 5.3: Size of the computational domain around monoplane and twostage Wells
turbines
of the CFD are unchanged. It is easy to notice that the six coarsest grids are associated
with a large variation of the objective functions (here, the tangential force coeﬃcient is
represented). On the other hand, all remaining grids employing more than 53 000 cells
lead to a variation of the target variables smaller than 1.5%. Therefore, the intermediate
CHAPTER 5. NUMERICAL METHODS AND ALGORITHMS 72
grid range between 55 000 and 80 000 cells is retained for all further results shown in the
present work for Wells turbine, depending on the type (one or two stages).
Figure 5.4: Gridindependence study for the tangential force coeﬃcient
5.3.2 Numerical solution of the ﬂow ﬁeld
5.3.2.1 Why Fluent?
From all CFD codes available in our group, ANSYSFluent has ﬁnally been retained.
Fluent is the world’s leading commercial supplier of Computational Fluid Dynamics
software and services. Fluent enables engineers to simulate ﬂuid ﬂow, ﬂuid machines,
heat and mass transfer, and a host of related phenomena involving turbulent, reacting,
and multiphase ﬂows. Many researchers have used Fluent for simulating Savonius
turbines [1, 69, 70] and Wells turbine [9, 103, 130, 131] in the past. Therefore, Fluent
has a capability to predict the performance of such turbines. Other commercial tools
do not show any noticeable advantage. Opensource CFD codes like OpenFOAM are
deﬁnitely cheaper but of limited numerical eﬃciency when considering a moving mesh
or using parallel computers [139]. Fluent is already coupled to our own optimization
library and has been ﬁnally selected as CFD solver for this study.
5.3.2.2 Model validation and selection for Savonius turbine
All ﬂow simulations presented in this work rely on the software ANSYSFluent version
6.3. The unsteady ReynoldsAveraged NavierStokes equations are solved using the
SIMPLE (SemiImplicit Method for Pressurelinked Equations) algorithm for pressure
velocity coupling. The ﬂow variables and all turbulent quantities are discretized in a
ﬁnitevolume formulation using a secondorder upwind scheme. The unsteady ﬂow is
solved by using the Sliding Mesh Model (SMM).
The full numerical procedure and in particular the employed turbulence model have
been validated by comparison with published experimental results for conventional Savo
nius turbines (twoblade and threeblade Savonius rotor of [38] and [48], respectively).
CHAPTER 5. NUMERICAL METHODS AND ALGORITHMS 73
The inﬂuence of the turbulence model is shown in Fig. 5.5 and Fig. 5.6. These results
demonstrate the excellent agreement obtained between CFD and experiments for the
target function, C
p
, when using the Realizable k − ε turbulence model. The employed
computational procedure thus appears suitable to predict the performance of the tur
bine in the investigated range of operation and is now kept for all further simulations
of Savonius turbines.
0.2 0.4 0.6 0.8 1 1.2 1.4
Speed ratio (l)
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
T
o
r
q
u
e
c
o
e
f
f
i
c
i
e
n
t
(
C
m
)
Exp. T. Hayashi et al (2005)
Realizable ke model
SSTmodel
Standard ke model
RSM model
0.2 0.4 0.6 0.8 1 1.2 1.4
Speed ratio (l)
0
0.05
0.1
0.15
0.2
0.25
0.3
P
o
w
e
r
c
o
e
f
f
i
c
i
e
n
t
(
C
p
)
(a)
(b)
Figure 5.5: Validation of computational model: a) torque coeﬃcient, b) power coeﬃcient,
both compared to published experimental results for a twoblade conventional Savonius
turbine [38]
CHAPTER 5. NUMERICAL METHODS AND ALGORITHMS 74
0.2 0.4 0.6 0.8 1 1.2
Speed ratio(l)
0
0.05
0.1
0.15
0.2
P
o
w
e
r
C
o
e
f
f
i
c
i
e
n
t
(
C
p
)
Exp. (K. Irabu & J. Roy, 2007)
Realizable ke model
SST model
Standard ke model
RSM model
Figure 5.6: Validation of computational model: power coeﬃcient compared to experi
mental results for a threeblade Savonius turbine [48]
5.3.2.3 Model validation and selection for Wells turbine
The full numerical model and in particular the employed turbulence model have been
again validated by comparison with published experimental results for a standard, mono
plane Wells turbine at a ﬂow Reynolds number Re= 2.4 10
5
using the chord as char
acteristic length. Published studies usually consider a range Re= 1 10
5
to Re= 5 10
5
,
since this corresponds to realistic conditions for employing a Wells turbine. Solidity is
assumed constant and equal to s = 0.67 following [110], which corresponds to a number
of blades equal to 8. The inﬂuence of the turbulence model is shown in Fig. 5.7. These
results demonstrate again the excellent agreement obtained between CFD and experi
ments for this standard conﬁguration, in particular when using the Realizable kε turbu
lence model. This model thus appears suitable to predict the performance of the turbine
in the later investigated range of operation (ﬂow coeﬃcient varying from φ = 0.08 to
0.25) and is now kept for all further simulations. For both turbines, ReynoldsStress
Model (RSM) leads to a considerably longer computing time but surprisingly to a worse
agreement than the kε models. This is probably due to the low turbulence level and
to a larger inﬂuence of the inﬂow turbulence boundary conditions, which are not prop
erly characterized in the experiments. Inlet boundary conditions for the RSM model
have been implemented using diﬀerent possibilities, prescribing either k and or the
turbulence intensity together with a length scale. Nevertheless, it has been impossible
to obtain a better agreement. Therefore, the RSM model appears to be unappropriate
for such conﬁgurations, associated with a low but unknown inﬂow turbulence level.
CHAPTER 5. NUMERICAL METHODS AND ALGORITHMS 75
0 0.05 0.1 0.15 0.2 0.25
Flow coefficient
0.1
0
0.1
0.2
0.3
0.4
T
a
n
g
e
n
t
i
a
l
f
o
r
c
e
c
o
e
f
f
i
c
i
e
n
t
s=0.67
a=8
°
Exp. T.Setoguchi et al. (1998)
Standard ke model
RNG ke model
Realizable ke model
sst kw model
RSM model
Figure 5.7: Inﬂuence of the turbulence model on the tangential force coeﬃcient, compared
to experimental results for a monoplane Wells turbine [107]
5.3.2.4 Realizable k − turbulence model
As explained in the previous sections, the realizable k turbulence model developed
by Shih et al. [113] has always been retained. This model contains a new transport
equation for the turbulent dissipation rate. Also, a critical coeﬃcient of the model, C
µ
,
is expressed as a function of mean ﬂow and turbulence properties, rather than assumed
to be constant as in the standard model. This allows the model to satisfy additional
mathematical constraints on the normal stresses, consistent with the physics of tur
bulence (realizability). The concept of a variable is also consistent with experimental
observations in boundary layers. The Realizable k model usually provides improved
results for swirling ﬂows and ﬂows involving separation when compared to the standard
k model.
• Transport equations:
∂
∂t
(ρk) +
∂
∂x
j
(ρ k u
j
) =
∂
∂x
j
¸
µ +
µ
t
σ
k
∂k
∂x
j
+P
k
+ P
b
−ρ −Y
M
+ S
k
(5.1)
∂
∂t
(ρ)+
∂
∂x
j
(ρ u
j
) =
∂
∂x
j
¸
µ +
µ
t
σ
∂
∂x
j
+ρ C
1
S −ρ C
2
2
k +
√
ν
+C
1
k
C
3
P
b
+S
(5.2)
where
C
1
= max
¸
0.43,
E
E + 5
, E = S
k
, S =
2S
ij
S
ij
(5.3)
CHAPTER 5. NUMERICAL METHODS AND ALGORITHMS 76
where S is the modulus of the mean rate of strain tensor.
In these equations, P
k
represents the generation of turbulence kinetic energy due
to the mean velocity gradients, calculated as follows:
P
k
= µ
t
S
2
(5.4)
P
b
is the generation of the turbulence kinetic energy due to buoyancy, negligible
for our applications:
P
b
= βg
i
µ
t
Pr
t
∂T
∂x
i
(5.5)
where Pr
t
is the turbulent Prandtl number for energy and g
i
is the component of
the gravitational vector in the i
th
direction. The default value of Pr
t
is 0.85.
The coﬃcient of thermal expansion, β is deﬁned as:
β = −
1
ρ
(
∂ρ
∂T
)
P
(5.6)
• Modelling turbulent viscosity
µ
t
= ρ C
µ
k
2
(5.7)
While C
µ
is constant in the standard k model, in the Realizable k model this
coeﬃcient is calculated as follows:
C
µ
=
1
A
0
+ A
s
kU
∗
(5.8)
U
∗
=
S
ij
S
ij
+
˜
Ω
ij
˜
Ω
ij
(5.9)
˜
Ω = Ω −2
ijk
ω
k
(5.10)
and
Ω = Ω
ij
−
ijk
ω
k
(5.11)
where Ω
ij
is the mean rate of rotation viewed in a rotating reference frame with
the angular velocity ω
k
. The model constants A
0
and A
s
are given by:
A
0
= 4.04, A
s
=
√
6 cos φ (5.12)
where
φ =
1
3
arccos(
√
6 W) (5.13)
W =
S
ij
S
jk
S
ki
˜
S
(5.14)
CHAPTER 5. NUMERICAL METHODS AND ALGORITHMS 77
˜
S =
S
ij
S
ij
(5.15)
S
ij
=
1
2
∂u
j
∂x
i
+
∂u
i
∂x
j
(5.16)
• Model Constants
C
1
= 1.44, C
2
= 1.9, σ
k
= 1.0, σ
= 1.3 (5.17)
5.3.3 PostProcessing: analysis of results
5.3.3.1 Moment calculation for Savonius turbines
The unsteady ﬂow is solved by using the Sliding Mesh Model. Since many diﬀerent con
ﬁgurations must be evaluated during the optimization, the computing time associated
with one single CFD computation must be kept acceptable. Three complete revolutions
are always computed, using an appropriate, constant value of the timestep; the ﬁrst
revolution is only used to initiate the correct ﬂow solution, while the ﬂow properties
(in particular the power coeﬃcient C
p
and the torque coeﬃcient C
m
) are obtained by
averaging the results during the last two revolutions. This combination (1 revolution
for initializing the ﬂow + 2 revolutions to compute the target function) has been kept
throughout. The moment coeﬃcient C
m
and the power coeﬃcient C
p
are calculated
according to Eqs. (2.20) and (2.21), respectively.
We have checked separately the inﬂuence of the number of revolutions on the com
puted turbine performance for the optimal design (Fig. 5.8), by continuing the CFD
simulation. After 10 revolutions, the average power coeﬃcient reaches a constant value.
The absolute diﬀerence in C
p
between this value and the one obtained after only three
revolutions equals 0.024. This is an estimation of the uncertainty associated with the
considered optimization process. This inherent uncertainty is very small compared to
the range of C
p
explored during the optimization and is thus deemed acceptable. It
amounts to only 6% of the pressure coeﬃcient associated with the optimal design. The
inﬂuence of the number of revolutions on the estimation of C
p
by CFD has been also
investigated systematically in a separate project [139], conﬁrming the present ﬁndings.
Only three revolutions have been thus computed for each Savonius design in order to
reduce the needed computational time.
5.3.3.2 Forces calculation for Wells turbines
When an airfoil is set at an angle of incidence α in a ﬂuid ﬂow, it will generate a lift force
F
L
normal to the free stream and a drag force F
D
in the direction of the free stream.
These lift and drag forces can then be combined to get the tangential force F
T
and
the axial force F
A
(Fig. 2.12). The corresponding force coeﬃcients are the tangential
force coeﬃcient C
T
and the axial force coeﬃcient C
A
, respectively which are calculated
according to Eqs. (2.24) and (2.25). Together with the turbine eﬃciency, C
T
will be the
objective function for the optimization.
CHAPTER 5. NUMERICAL METHODS AND ALGORITHMS 78
0 15 30 45 60 75
Time
0
0.4
0.8
1.2
P
o
w
e
r
c
o
e
f
f
i
c
i
e
n
t
(
C
p
)
0 20 40 60 80 100
No. of revolutions
l =0.7
Instantaneous power coeff.
Average Power Coeff.
Employednumberofrevolutions
Figure 5.8: Inﬂuence of the number of revolutions on the instantaneous and on the
average power coeﬃcient C
p
computed by CFD for the optimum design shown later.
5.4 CFD/Optimization coupling
A fully automatic optimization ﬁnally takes place, using OPAL (decisionmaker for the
conﬁgurations to investigate), the commercial tool Gambit for geometry and grid gener
ation (including quality check) and the industrial CFD code ANSYSFluent to compute
the ﬂow ﬁeld around the turbines. As a result of the CFD computation the objective
function(s) is determined, and stored in a result ﬁle. The procedure is automated using
journal scripts (to restart Gambit, Fluent) and a master program written in C (Algo
rithm 5.5.1), calling all codes in the right sequence as shown in Fig. 5.9. By checking
the values stored in the result ﬁle, OPAL is able to decide how to modify the input
parameters before starting a new iteration. The fully coupled optimization procedure is
a complex task, which has been described in detail in previous publications [40, 52, 132].
Algorithm 5.4.1
/∗ Block 1 – Generate input ﬁle ∗/
begin
sprintf( ﬁlename, ”input.dat”);
if ( !( f input := fopen(path, ”w”) ) )
Open the inputﬁle for the simulation.
then
printf( ”Cannot write ﬁle %s !”, path );
exit; Error opening the inputﬁle.
ﬁ
comment: Create the header of the input ﬁle.
fprintf( f input, ”$DesignVariable1 = %le”,
CHAPTER 5. NUMERICAL METHODS AND ALGORITHMS 79
Figure 5.9: Schematic description of optimizer (OPAL) and CFD code coupling.
DesignV ariable1);
fprintf( f input, ”$DesignVariable2 = %le”,
DesignV ariable2);
fclose( f input ); Close the inputﬁle.
end
/* Block 2 – Perform the simulation */
begin system( ”rm f old results.dat” ); Remove the previous result.
comment: Call the simulation tool.
system( ”ﬂuent 2ddp g i journal ﬁle.jou” );
end
/* Block 3 – Import the result(s) of the simulation */
begin
if ( !(f result := fopen(”Result.dat”, ”r”) ) )
Open the result ﬁle.
then
ErrorVariable := 1; Error reading the result ﬁle.
continue;
if
fscanf( f result,”%le”, &Drag ); Read the objective value.
fclose( f result ); Close the result ﬁle.
end
Many types of ﬁles have to be prepared in order to start optimization process, as
follows:
• Fluent journal ﬁle (ﬂuent.jou)
CHAPTER 5. NUMERICAL METHODS AND ALGORITHMS 80
In this ﬁle, all the steps needed for CFD have been coded by text interface as
script ﬁle. For example, a statement like ”de bc vi 3 y y n 10 n 0.9902 n 
0.13917 y n 0.49 n 4.9” means ”deﬁne boundaryconditions of the velocity inlet
(zone number 3), the velocity speciﬁcation method is magnitude and direction
with absolute reference frame and constant value. The inlet velocity value equal
10 m/s with constant Xcomponent of the ﬂow direction equal 0.9902 and constant
Ycomponent equal −0.13917. The Inlet turbulent speciﬁcation method is k and
ε with constant values equal 0.49 and 4.9, respectively”. This procedure is then
repeated for every step in ﬂuent sequence.
• Gambit journal ﬁle (gambit.jou)
By the same method, a ﬁle is also constructed for geometry and grid generation.
However, in our cases, this ﬁle has been built from two subﬁles (head and tail).
Since this work considers shape optimization, the tail ﬁle describes the body and
contains all geometry steps. The values of all parameters describing the geometry
are placed in the head ﬁle at the beginning of each optimization iteration.
• Optimization parameters ﬁle (*.tcl)
In this ﬁle, all the optimization parameters have been implemented, containing
number of generations, population size, mutation rate, crossover probability and
so on (Table 5.1). Beside these parameters, all constrains of each optimization
parameter have been deﬁned in this ﬁle.
Further, ﬁles have been constructed during the optimization, as:
• Input ﬁle (*.in)
This ﬁle contains the values of the optimization parameters for every generation.
• Fluent output ﬁle (ﬂuent.out)
with this ﬁle, we can follow the convergence of the CFD solution for every conﬁg
uration during evaluation.
• Output ﬁle (Result.dat)
The output ﬁle includes all the objective results for every tested conﬁguration.
Results are stored in this ﬁle after every generation.
5.5 Optimization parameters
5.5.1 Savonius turbine: single objective optimization
The design variables considered for the optimization will be described in the next chap
ter, since many diﬀerent cases have been optimized in this work. For this purpose,
diﬀerent parameters are considered in each conﬁguration, which, together, are suﬃcient
to ﬁx clearly the geometry of this speciﬁc case. The objective function considers only
CHAPTER 5. NUMERICAL METHODS AND ALGORITHMS 81
one output of the simulation (single objective optimization), that should be maximized
as far as possible: the output power coeﬃcient C
p
. The parameters of the GA are listed
in Table 5.1.
Table 5.1: Parameters of the Genetic Algorithm
Parameter Value
Population size of the ﬁrst generation, N 20 to 30
Number of generations N
g
Survival probability 50%
Average probability 33.3%
Crossover probability 16.7%
Mutation probability 100%
Mutation magnitude 30%
a
(i.e. ±15%)
a
This value is multiplied by 0.8 at each generation. For example the mutation
magnitude is only 4% (±2%) after 10 generations. Mutation magnitude must be
decreased during the optimization process to stabilize the population.
The number of generations N
g
will be varied, depending on the number of free
parameters, since there is a very strong relation between the number of parameters,
the population size and the eﬃciency of the Genetic Algorithm (computing time and
solution quality) [20].
5.5.2 Wells turbine: multiobjective concurrent optimization
The central goal when designing an improved Wells turbine is to achieve high eﬃciency
and high power output (i.e., high tangential force coeﬃcient). The objective function
hence contains simultaneously two outputs of the simulation (multiobjective concurrent
optimization), that should both be maximized as far as possible: the tangential force
coeﬃcient C
T
; and the turbine eﬃciency η , which is inversely proportional to the axial
force coeﬃcient, and is deﬁned for a negligible density change as:
η =
F
T
u
t
∆p
0
Q
(5.18)
with Q the volumetric ﬂowrate through the turbine.
The design variables again depend on the speciﬁc problem considered and will be
listed in Chapter 7. The parameters of the GA are given in Table 5.1.
5.6 Conclusions
All the tools required for the optimization have been now developed and validated. The
ﬁnal analysis can be started, beginning with the Savonius turbine.
Chapter 6
Savonius turbine: singleobjective
optimization
6.1 Introduction
As already discussed in section 3.1.3, both twoblade and threeblade Savonius turbines
have been proposed and constructed. The threeblade conﬁguration should be in par
ticular advantageous to obtain good selfstarting conditions. At the beginning of this
project, a small company contacted us with a modiﬁed, threeblade design without gap,
called in what follows GWturbine. We will start by considering and optimizing this
speciﬁc conﬁguration, before comparing with the standard (but optimized) threeblade
Savonius turbine. At the end of the chapter, the best overall solution, involving indeed
only two blades, will be fully optimized, yielding very promising results.
6.2 Optimal GWturbine: modiﬁed threeblade
Savonius turbine without gap
6.2.1 Performance of the original GWturbine
A modiﬁed design, a threeblade rotor without passage in between has been proposed by
a small company, in an eﬀort to improve the performance compared to the conventional
system (Fig. 6.1).
Three issues must be speciﬁcally investigated in this case:
1. is it possible to improve the performance by changing the shape of the blade (not
being semicylindrical any more)?
2. is it possible to improve the performance by employing a deﬂector nose in front of
the turbine?
3. is it possible to improve the performance by using mobile parts for the returning
blade, thus reducing drag by ”opening” the returning blade?
82
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 83
Figure 6.1: Schematic shape of the GWturbine
We begin by considering the newly proposed design compared to the conventional
turbine.
0.2 0.4 0.6 0.8 1 1.2 1.4
Speed ratio (l)
0.1
0.05
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
T
o
r
q
u
e
c
o
e
f
f
i
c
i
e
n
t
(
C
m
)
Conventional Savoniusturbine(twoblades)
Threebladewithout passage
0.2 0.4 0.6 0.8 1 1.2 1.4
Speed ratio (l)
0.1
0.05
0
0.05
0.1
0.15
0.2
0.25
0.3
P
o
w
e
r
c
o
e
f
f
i
c
i
e
n
t
(
C
p
)
Conventional Savonius turbine (twoblades)
Threeblade without passage
Negative Torque
Figure 6.2: Performance comparison between the GWturbine and the conventional,
twoblade Savonius rotor.
Figure 6.2 shows the comparison between the GWturbine (denoted “three blades
without passage”) and the conventional Savonius turbine performance. Unfortunately,
it appears clearly from these ﬁrst comparisons that the new design systematically leads
to a poorer performance, both from the point of view of power coeﬃcient and of torque
coeﬃcient. For very low values of λ, the diﬀerence is small, but the loss of performance
becomes considerable for increasing λ. The peak value of C
p
is only 0.16 compared to
0.18 for the conventional design. An analysis of the ﬂow can readily identify the reason
for this loss of performance. Increasing the number of blades increases the reverse
moment as well. Closing the passage between the blades leads to a reduction of the air
ﬂow entering the system and increases the global drag on the returning blade (Fig. 6.3).
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 84
Figure 6.3: Instantaneous velocity vectors around the GWturbine
6.2.2 Reducing the drag on the returning blade
In order to improve the performance of the GWturbine, it was proposed to reduce the
drag of the returning blade by employing mobile blade sections. In closed position, these
mobile parts return back to the standard blade shape. When open, they should lead to
a considerable reduction of the drag on the returning blade. This procedure is described
schematically in Fig. 6.4.
Wind
Justclosedblade
Justopenedblade
Closedblade
(advancingblade)
Figure 6.4: Schematic description of the GWturbine with open returning blade.
The new design with an open returning blade has been investigated numerically for
diﬀerent values of the speed ratio λ. For these computations, the slits are considered
to be inclined by 30
◦
(constant value) compared to the local blade direction. The
resulting performance has been compared with that of the baseline GWturbine, as
shown in Fig. 6.4. The results show a considerable improvement of the performance
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 85
when comparing with the baseline GWturbine. This improvement is visible both for
the torque and for the power coeﬃcient. The gain is directly a consequence of the drag
reduction on the returning blade, due to air passing between the slits of this returning
blade, as seen in Fig. 6.5. The increase of the power coeﬃcient reaches 0.0843 at
λ = 0.8, which means a relative improvement of performance by almost 36% under such
conditions. The GWturbine with open returning blade is also systematically better
than the conventional Savonius turbine. For example, at λ = 0.7 the relative increase
in performance is 25.9%.
0.2 0.4 0.6 0.8 1 1.2 1.4
Speed ratio (l)
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
T
o
r
q
u
e
c
o
e
f
f
i
c
i
e
n
t
(
C
m
)
Conventional Savoniusturbine
Threebladeswithout passage(modifieddesign)
Threebladeswithopenreturningblade
%Relativeincreaseof opendesignvs. modifieddesign
%Relativeincreaseof opendesignvs. conventional turbine
0
30
60
90
120
150
180
210
%
R
e
l
a
t
i
v
e
i
n
c
r
e
a
s
e
0.2 0.4 0.6 0.8 1 1.2 1.4
Speed ratio (l)
0.1
0.05
0
0.05
0.1
0.15
0.2
0.25
0.3
P
o
w
e
r
c
o
e
f
f
i
c
i
e
n
t
(
C
p
)
Conventional Savonius turbine
Three blades without passage (modified design)
Three blades with open returning blade
% increase of open design vs. modified design
% increase of open design vs. Savonius turbine
0
30
60
90
120
150
180
210
%
R
e
l
a
t
i
v
e
i
n
c
r
e
a
s
e
Negative Torque
Figure 6.5: Performance comparison between the GWturbine with or without open re
turning blade. The performance of the conventional Savonius rotor is also shown for
comparison. Top: torque coeﬃcient. Bottom: power coeﬃcient.
Due to the eﬀectiveness of this design, the eﬀect of the slits opening angle on the
performance has been investigated. Therefore, nine diﬀerent angles (from 10
◦
to 90
◦
)
have been studied, as shown in Fig. 6.6. From the results for both torque and power
coeﬃcients, the best performance for this design is when the slit angle lies between 30
◦
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 86
0.2 0.4 0.6 0.8 1 1.2 1.4
Speed ratio (l)
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
T
o
r
q
u
e
C
o
e
f
f
i
c
i
e
n
t
(
C
m
)
open with 10°
open with 20°
open with 30°
open with 40°
open with 50°
open with 60°
open with 70°
open with 80°
open with 90°
0.2 0.4 0.6 0.8 1 1.2 1.4
Speed ratio(l)
0.2
0.15
0.1
0.05
0
0.05
0.1
0.15
0.2
0.25
0.3
P
o
w
e
r
c
o
e
f
f
i
c
i
e
n
t
(
C
p
)
open with 10°
open with 20°
open with 30°
open with 40°
open with 50°
open with 60°
open with 70°
open with 80°
open with 90°
Figure 6.6: Performance of open returning blade turbine for diﬀerent slit angles. Top:
torque coeﬃcient. Bottom: power coeﬃcient.
and 40
◦
considering at eﬀective operating range between λ = 0.6 to λ = 1.2, correspond
ing to practical applications.
6.2.3 Inﬂuence of a rounded obstacle plate
In the original concept, a rounded deﬂector structure is placed in front of two counter
rotating GWturbines. It is therefore important to check the inﬂuence of this deﬂect
ing surface on the global system performance. The corresponding geometry is shown
schematically in Fig. 6.7.
Figure 6.8 shows a comparison between an isolated GWturbine and the turbine
placed behind the rounded deﬂector. Here again, it appears unfortunately that the
performance is systematically reduced by the deﬂector. A GWturbine placed directly
in the free ﬂow leads systematically to a higher performance, both in terms of torque
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 87
Rotation
0
.8
5
m
5cm
R
o
t
a
t
i
o
n
5
c
m
Wind 0
.8
5
m
F
l
o
w
F
l
o
w
Figure 6.7: Schematic description of the counterrotating GWturbine with rounded de
ﬂector.
coeﬃcient and of power coeﬃcient. When analyzing the resulting ﬂow ﬁeld, the decrease
in performance can be related to the excessive size and large radius of the planned
deﬂector. As a consequence, the air ﬂow is directed away from the blades instead of
entering the system.
6.2.4 Optimization of the blade shape
Both for the conventional Savonius turbine and for the new design, semicylindrical
blades are considered as a starting condition. Even if such blades are traditionally
employed, it has never been demonstrated that such a blade shape leads to an optimal
performance. Therefore, an optimization of this shape geometry reconstructed by splines
based on three discrete points (Fig. 6.9) has been carried out, moving only one point
(p
1
).
The mathematical optimization procedure described previously (Genetic Algorithm
relying on automated evaluations through CFD) is employed to ﬁnd the optimal blade
shape. This is done for a speed ratio λ = 0.7, considering an incident wind velocity
U = 10 m/s, following the literature. This value of λ is retained, since it is known
from the literature that it corresponds to the peak power coeﬃcient of the conventional
turbine (nominal conditions).
Two degrees of freedom are left simultaneously to the OPAL optimizer: X
1
and Y
1
deﬁne the coordinate of the center point of the blades (Fig. 6.9).
In this study, a relatively large domain has been deﬁned for the optimization in the
parameter space. The limits of this domain for the two parameters are (0.32 : 0.675) for
(X
1
/R) and (−0.294 : 0.294) for (Y
1
/R), where R is the radius of the original design,
kept constant during the optimization.
Finally, the optimization process thus involves simultaneously two parameters (or
degrees of freedom): X
1
and Y
1
. For each geometrical conﬁguration one single objective
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 88
0.2 0.4 0.6 0.8 1 1.2 1.4
Speed ratio (l)
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5
T
o
r
q
u
e
c
o
e
f
f
i
c
i
e
n
t
(
C
m
)
Conventional Savonius turbine
Three blades without passage
Three blades with rounded obstacle
0.2 0.4 0.6 0.8 1 1.2 1.4
Speed ratio (l)
0.2
0.15
0.1
0.05
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
P
o
w
e
r
c
o
e
f
f
i
c
i
e
n
t
(
C
p
)
Conventional Savonius turbine
Three blades without passage
Three blades with rounded obstacle
Negative Torque
Negative Torque
Figure 6.8: Performance comparison between the new concept in free ﬁeld or placed
behind a rounded deﬂector.
(power output coeﬃcient) is determined by CFD evaluations, and should be maximized
by the optimization procedure.
The results presented in Fig. 6.10 indicate that the considered objective is indeed
considerably inﬂuenced by the two free parameters, X
1
and Y
1
. As a whole, 140 diﬀerent
geometrical settings have been evaluated by CFD.
An optimal conﬁguration can readily be identiﬁed for λ = 0.7. This optimum point
diﬀers noticeably from the original design and corresponds to the point of coordinate
X
1
/R = 0.6315 and Y
1
/R = −0.0521 as shown in Fig. 6.10. This optimal condition
leads to a power coeﬃcient C
p
= 0.1638 and a torque coeﬃcient C
m
= 0.2339.
When compared with the GWturbine (semicylindrical blade shape, Fig. 6.11), the
optimal point found by the optimization procedure corresponds simultaneously to an
increase of the power coeﬃcient by 0.01487 and of the torque coeﬃcient by 0.02126 at
λ = 0.7. For the power coeﬃcient, this means a relative increase of the performance
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 89
P2
P3
P (X ,Y ) 1 1 1
Y1
r
Variablepoint
X1
Figure 6.9: Description of the blade shape with 3 discrete points connected by splines.
Figure 6.10: The two input parameters of the optimization and the power coeﬃcient.
SemicylindricalShape
BestShape
Figure 6.11: The best (right) conﬁguration obtained during the optimization compared
to the original GWturbine (semicylindrical shape: left).
by 7.1% compared to the original GWturbine. Since such turbines must operate also
outside of the design conditions, it is now important to check how this gain will change
as a function of λ. Therefore, the performance of the obtained optimal conﬁguration
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 90
has been computed for the full range of useful λvalues, as shown in Fig. 6.12. The
results of the new design with semicylindrical blades are also shown for comparison.
Figure 6.12 demonstrates that the improvement of power output coeﬃcient is observed
at all conditions for intermediate values of λ (in particular between λ = 0.7 and λ=1.1),
compared to the semicylindrical design. The highest gain in eﬃciency with the new
blade shape is obtained around λ = 1 and is roughly equal to 15%. For very low
(λ < 0.6) and very high (λ > 1.2) values of λ, the modiﬁed shape is less eﬃcient than
the semicylindrical one. Therefore, operating at such conditions should be avoided.
0.2 0.4 0.6 0.8 1 1.2 1.4
Speed ratio (l)
0.05
0
0.05
0.1
0.15
0.2
0.25
P
o
w
e
r
c
o
e
f
f
i
c
i
e
n
t
(
C
p
)
Three blades without passage
Optimal blade shape
% Relative increase
60
50
40
30
20
10
0
10
20
%
R
e
l
a
t
i
v
e
i
n
c
r
e
a
s
e
%
Figure 6.12: Power coeﬃcient of the optimized conﬁguration compared to the GW
turbine as a function of λ.
6.2.5 Conclusions on the GWturbine
We can summarize all these results as:
• The standard GWdesign (threeblade without gap) is less favorable than the
conventional Savonius rotor.
• The rounded deﬂector is much too large in the original design.
• Opening the returning blade leads indeed to a considerable increase of perfor
mance. Diﬀerent opening angles between 10
◦
to 90
◦
have been tested, best values
are found between 30
◦
and 40
◦
.
• Similar modiﬁcations could possibly lead to even better results for classical Savo
nius turbines. It is therefore interesting to examine now such conﬁgurations, start
ing with the conventional, threeblade turbine.
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 91
6.3 Optimal threeblade Savonius turbine
The conventional, threeblade conﬁguration with gap (Fig. 6.13) of the Savonius rotor
has been extensively studied in the past [48]. The corresponding values of C
p
and
C
m
have been determined numerically and sometimes experimentally as a function of
the speed ratio λ. This has already been used to validate extensively our numerical
procedure by comparison with published results, (see section 5.3.2.2). Published studies
have demonstrated that threeblade turbines show considerable drawbacks compared to
classical Savonius turbines (twoblade), in particular a lower eﬃciency. Nevertheless, we
will try now to improve the performance of this design through optimization.
Wind
R
ThreebladeSavoniusTurbine
gw
120
Figure 6.13: Schematic description and main parameters characterizing a conventional
Savonius rotor with three blades.
6.3.1 Inﬂuence of obstacle plate
Since one of the major advantages of the Savonius turbine is its simplicity and cor
responding compactness, robustness and low cost, a modiﬁcation introducing a high
complexity should probably not be retained. Considering the results of the previous
studies and of section 3.1.3, some simple guiding or deﬂecting plate could lead to the
best eﬃciency improvement at the lowest possible cost and complexity. Therefore, we
will now investigate numerically the eﬀect of an obstacle shielding partly the returning
blade of a Savonius turbine. This part builds on top of a previous investigation [72]
considering Savonius turbines with two as well as with three blades. Adding a shielding
obstacle should in principle reduce the reverse moment, and as a consequence the total
moment of the turbine will be increased, since the total moment is the moment diﬀer
ence between the advancing and the returning blades. Related ideas have already been
proposed by other groups in the past [48]. But, at the diﬀerence of the previous studies,
we are not looking here for a better solution, but directly for the best possible one.
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 92
6.3.1.1 Selfstarting capability
One important issue associated with wind energy is the selfstarting capability of the
system. For decentral, lowcost applications as considered here, it is essential to obtain
a selfstarting system. To investigate this issue, the static torque exerted on a turbine at
a ﬁxed angle has been computed by CFD as a function of this angle θ. Figure 6.14 shows
the obtained static torque coeﬃcient C
m
s
obtained for three diﬀerent positions of the
obstacle plate as a function of θ. The experimental results of [48] for the conventional
threeblade turbines are also shown for comparison in Fig. 6.14. Due to periodicity, the
results are only plotted for θ between 0 and 120
◦
. These computations demonstrate
that the obstacle plate has a considerable and mostly positive eﬀect on the static torque
coeﬃcient. The obstacle plate improves the selfstarting capacity for part of the θrange.
A selfstarting capability (C
m
s
> 0) is always obtained in principle at any angle, which
is a major advantage.
0 30 60 90 120
Rotation angle(q)
0
0.1
0.2
0.3
0.4
0.5
S
t
a
t
i
c
T
o
r
q
u
e
c
o
e
f
f
i
c
i
e
n
t
(
C
m
s
)
Exp. K. Irabu & J. Roy (2007)
(no obstacle)
Three blade without obstacle
Y1/R=0.0 , b=73
o
Y1/R=0.24 , b=68.5
o
Y1/R=0.47 , b=60.5
o
Figure 6.14: Static torque coeﬃcient as a function of the rotor angle θ for three diﬀerent
values of Y
1
choosing X
1
/R = −1.4 and X
2
/R = −1.76.
6.3.1.2 Optimization of the obstacle position
The position and the angle of the shielding obstacle will now be optimized. The free
design variables considered for the optimization will describe the obstacle position. For
this purpose, three parameters are considered (X
1
, Y
1
and X
2
) which, together with
a ﬁxed value for Y
2
are suﬃcient to ﬁx clearly the geometry of the shielding obstacle
(Fig. 6.15). The objective function considers only one output of the simulation, that
should be maximized as far as possible: the output power coeﬃcient C
p
. The mathemat
ical optimization procedure described previously can be employed to ﬁnd the optimal
position of the obstacle. This is done again for a speed ratio λ = 0.7, considering a ﬁxed
incident wind velocity U = 10 m/s.
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 93
Wind
X >R 1
X >R 2
R
Obstacle
Y
1
Advancing
Blade
Returning
Blade
Y
2
x
y
(0,0)
Figure 6.15: Schematic description of the geometry and free optimization parameters
X
1
, Y
1
, X
2
used to modify the position of the shielding obstacle.
As explained previously, three degrees of freedom are left simultaneously to the
OPAL optimizer: X
1
and Y
1
deﬁne the upper tip of the shielding obstacle; the value
X
2
is then suﬃcient to deﬁne the position of the lower tip, since Y
2
is taken constant,
with Y
2
/R = −1.177 (Fig. 6.15). With these three factors, the position of the obstacle
is perfectly determined, and the angle β can be deduced as well. In this manner, a
truly optimal solution can be obtained considering a large set of possible parameters.
Of course, when choosing the parameter space, it must be checked that the shielding
obstacle cannot come into direct contact with the rotor ( X
1
> R and X
2
> R), so
the acceptable range for the input parameters is shown in Table 6.1. The corresponding
positions of the obstacle all lead to conﬁgurations that shield partially the returning
blade.
Table 6.1: Acceptable range for the input parameters (parameter space)
Parameter minimum maximum
X
1
/R −1.88 −1.017
Y
1
/R −0.88 0.0
X
2
/R −1.88 −1.017
In Fig. 6.16 a parallel coordinate representation has been chosen since it is the
most popular way to analyze output data from optimization involving several degrees
of freedom. Such ﬁgures might be at ﬁrst diﬃcult to understand. Each thin connecting
line represents all numerical parameters associated with one speciﬁc conﬁguration, where
each parameter is associated with its own vertical axis. The ﬁrst three columns therefore
show the value of the three free parameters, X
1
/R, X
2
/R and Y
1
/R; the last column on
the right corresponds to the value of the objective function, here the power coeﬃcient
C
p
. The scale of the ﬁrst parameter (X1/R) is for instance bounded between −1.88
(minimum) and −1.017 (maximum). By following a single line, the reader can therefore
determine quantitatively the values of all parameters associated with one conﬁguration.
The optimal solution is shown with a thick red line. The results presented in Fig. 6.16
indicate that the considered objective is indeed considerably inﬂuenced by the three free
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 94
X1/R Y1/R X2/R Powercoeff.
1.88
1.017 0.0 1.017 0.212
0.88
1.88
0.133

Optimalconfiguration
0.153
Without
obstacle
Figure 6.16: Input parameters of the optimization and power coeﬃcient represented
using parallel coordinates. The parameters of the optimal conﬁgurations are connected
with a thick red line. The power coeﬃcient of the conventional threeblade Savonius
turbine is shown with a gray circle.
parameters, X
1
, Y
1
and X
2
. As a whole, 210 diﬀerent geometrical settings have been
evaluated by CFD, requesting 22 days of computing time on a standard PC for the
threeblade Savonius turbine. Note that the userwaiting time could be considerably
reduced by carrying out the requested CFD in parallel on a PC cluster [132]. Such a
parallel procedure, already implemented in OPAL, has not been used in the present case
but could reduce the needed time by more than an order of magnitude, as demonstrated
in other studies.
The optimal conﬁguration (highest point in the right column in Fig. 6.16, all corre
sponding parameters being connected by a thick red line) can now readily be identiﬁed
for λ = 0.7. The optimum obstacle position, optimum angle β and corresponding opti
mal power coeﬃcient C
p
are listed in Table 6.2.
Table 6.2: Optimal conﬁgurations
Design parameter value β power coeﬀ. C
p
Threeblade Savonius turbine X
1
/R −1.05632
Y
1
/R −0.36912 (80.52
◦
) 0.2120
X
2
/R −1.38162
When compared with the threeblade Savonius turbines (without shielding obsta
cle), the optimal point found by the optimization procedure corresponds to an absolute
increase of the power coeﬃcient by 0.058 at λ = 0.7. This means a relative increase of
the performance (measured by the power output coeﬃcient) by 27.5%, compared to the
conventional threeblade Savonius design without any obstacle.
It is now important to check how this gain would change as a function of λ, since such
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 95
0.2 0.4 0.6 0.8 1 1.2
Speed ratio (l)
0
0.1
0.2
0.3
0.4
0.5
T
o
r
q
u
e
C
o
e
f
f
i
c
i
e
n
t
(
C
m
)
Three blade Savonius
without obstacle
Three blade with obstacle
(optimum configuration)
% Relative increase
0
20
40
60
80
100
R
e
l
a
t
i
v
e
i
n
c
e
a
s
e
(
%
)
0.2 0.4 0.6 0.8 1 1.2
Speed ratio(l)
0
0.05
0.1
0.15
0.2
0.25
P
o
w
e
r
C
o
e
f
f
i
c
i
e
n
t
(
C
p
)
0
20
40
60
80
100
R
e
l
a
t
i
v
e
i
n
c
e
a
s
e
(
%
)
(a)
(b)
%
%
Figure 6.17: Performance of the optimized conﬁguration (red line) compared to the
conventional threeblade Savonius turbine without obstacle (black): a: torque coeﬃcient;
b: power coeﬃcient. The corresponding relative increase is shown with blue line.
a turbine must be able to work also for oﬀdesign conditions. Therefore, the performance
of the optimal conﬁgurations have been ﬁnally computed for the full range of useful λ
values, as shown in Fig. 6.17. This ﬁgure demonstrates that the improvement of both
torque coeﬃcient and power output coeﬃcient is observed throughout for all values of
λ, compared to the conventional threeblade Savonius turbine without obstacle. The
absolute gain for C
p
increases even slightly with λ at ﬁrst, the relative increase being
highest for the largest values of λ considered in the present study.
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 96
6.3.2 Optimal blade shape
In the previous section, the eﬃciency of the conventional threeblade Savonius turbine
has been increased by placing in an optimal manner an obstacle plate shielding the
returning blade. The present study now aims at improving further the output power
of the threeblade Savonius turbine. In order to achieve this objective, the geometry of
the blade shape (skeleton line) is now optimized in presence of the obstacle plate.
Optimization is used here to ﬁnd the best blade shape while taking into account
the obstacle shielding the returning blade in the optimum position. The free design
variables considered for the optimization describe the blade skeleton line for a constant
blade thickness of 2 mm. For this purpose, six parameters are considered (X
P1
, Y
P1
,
X
P2
, Y
P2
, X
P3
and Y
P3
, see Fig. 6.18). The obstacle plate is kept ﬁxed in the opti
mal position identiﬁed in the previous section (X
1
/R = −1.05632, Y
1
/R = −0.36912,
X
2
/R = −1.38162 and Y
2
/R = −1.1770, which leads to an angle β = 80.52
◦
, see
Fig. 6.15). Now, the blade shape of the Savonius turbine will be optimized in order
to ﬁnd the best possible ﬂow conditions. The six shape parameters are suﬃcient to
determine uniquely the geometry of the system, since the obstacle position is ﬁxed.
P4
P (X ,Y ) 1 P1 P1
P5
r
r
Variablepoints
P (X ,Y ) 2 P2 P2
P (X ,Y ) 3 P3 P3
X
Y
(P0)
Fixedpoints
Blade
center
Turbineshaft
gw
Figure 6.18: Schematic description of the free optimization parameters
X
P1
, Y
P1
, X
P2
, Y
P2
, X
P3
and Y
P3
used to modify the blade shape.
Knowing all 5 points, the full proﬁle of the blade is reconstructed using standard
splines (Nonuniform rational Bsplines, NURBS). The order of a NURBS curve deﬁnes
the number of nearby control points that inﬂuence any given point on the curve. The
curve is represented mathematically by a polynomial of degree one less than the order
of the curve; this means that the spline order is 5 in our case and the degree of the
polynomial is 4. The objective function contains one single output of the simulation,
that should be maximized as far as possible: the power coeﬃcient C
p
. The parameter
space considered in the optimization has been deﬁned as documented in Table 6.3. These
domains are selected to prevent any domain overlap along the Y direction and to keep
realistic blade shapes. The reference point of the parameter space is point P
0
, which is
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 97
the center of the original, semicylindrical shape with radius r as shown in Fig. 6.18.
During the calculations, a circular turbine shaft is included with a ﬁxed radius R
sh
computed from R
sh
/R = 0.03.
Table 6.3: Acceptable range for the input parameters (parameter space for blade shape
with the obstacle )
Parameter Minimum allowed Maximum allowed
X
P1
/r 0.53 1.47
Y
P1
/r −0.24 0.24
X
P2
/r 0.24 1.2
Y
P2
/r −0.94 −0.24
X
P3
/r 0.24 1.2
Y
P3
/r 0.24 0.94
The results presented in Fig. 6.19 indicate that the considered objective is again
considerably inﬂuenced by the six free parameters, X
P1
, Y
P1
, X
P2
, Y
P2
, X
P3
and Y
P3
,
and thus by the blade shape. As a whole, 210 diﬀerent geometrical settings have been
evaluated by CFD, requesting one and a half month of total computing time on a
standard PC.
0.53 0.24 0.24
0.94 0.24
0.24 0.0
0.24 0.94 1.2 0.24
1.2 0.24 1.47
X /r 1 Y /r 1 X /r 2 Y /r 2 X /r 3 Y /r 3 Cp
Optimalconfiguration
Threeblade
Savoniuswithout
obstacle
0.153
0.233
Figure 6.19: Input parameters of the optimization and power coeﬃcient represented
using parallel coordinates. The parameters of the optimal conﬁguration are connected
with a thick red line. The power coeﬃcient of the conventional threeblade turbine (semi
cylindrical shape) is also shown with a black circle.
The optimal conﬁguration (highest point in the right column in Fig. 6.19, all corre
sponding parameters being connected by a thick red line) can now readily be identiﬁed
for λ = 0.7. The optimum point positions and corresponding optimal power coeﬃcient
C
p
are listed in Table 6.4.
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 98
Table 6.4: Optimal blade shape with the obstacle
X
P1
/r Y
P1
/r X
P2
/r Y
P2
/r X
P3
/r Y
P3
/r Optimal power coeﬀ. C
p
0.7519 0.0307 0.4345 −0.5456 0.6594 0.5464 0.233
At λ = 0.7 the optimal point found by the optimization procedure corresponds to
an absolute increase of the power coeﬃcient by 0.08 compared with the conventional
threeblade Savonius turbine, respectively by 0.021 compared with the conventional
Savonius rotor (semicylindrical blade shape) with obstacle plate. As a whole, this
means a relative increase of the performance (measured by the power output coeﬃcient)
by 34.3% for the optimum shape with obstacle plate, compared to the conventional
threeblade Savonius design without obstacle.
Note that this new shape (Fig. 6.20) is only optimum in combination with the em
ployed obstacle plate. Indeed, the modiﬁed shape exploits best the ﬂow redirection in
duced by the obstacle. As a consequence, the reverse moment is reduced by the shielding
obstacle; simultaneously, the modiﬁed shape leads to an increase of the positive moment
of the advancing blade.
Obstacle
Wind
Figure 6.20: Optimum conﬁguration obtained with the optimization procedure.
It is now important to check how this gain would change as a function of λ. Therefore,
the performance of the optimal conﬁguration has been ﬁnally computed for the full range
of useful λvalues, as shown in Fig. 6.21. This ﬁgure demonstrates that the improvement
of both torque coeﬃcient and power output coeﬃcient is observed throughout for all
values of λ, compared to the conventional threeblade Savonius turbine. The absolute
gain for C
p
and C
m
is even higher for lower λvalues. The relative performance increase
compared to the conventional threeblade Savonius conﬁguration is always higher than
30% in the eﬀective operating range.
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 99
0.2 0.4 0.6 0.8 1 1.2 1.4
Speed ratio (l)
0
0.1
0.2
0.3
0.4
0.5
0.6
T
o
r
q
u
e
C
o
e
f
f
i
c
i
e
n
t
(
C
m
) Three blade Savonius
without obstacle
optimal Configuration
0.2 0.4 0.6 0.8 1 1.2 1.4
Speed ratio(l)
0
0.05
0.1
0.15
0.2
0.25
P
o
w
e
r
C
o
e
f
f
i
c
i
e
n
t
(
C
p
)
0
20
40
60
80
100
R
e
l
a
t
i
v
e
i
n
c
e
a
s
e
(
%
)
Three blade Savonius
without obstacle
Optimal configuration
% Relative increase
(a)
(b)
%
Figure 6.21: Performance of the optimized conﬁguration (red line) compared to the
conventional threeblade Savonius turbine (black line): a) torque coeﬃcient; b) power
coeﬃcient. The corresponding relative increase compared to the conventional threeblade
conﬁguration is shown with blue line.
6.3.3 Optimal threeblade Savonius turbine with guiding
plates
After improving the eﬃciency of the threeblade Savonius turbine by placing appropri
ately an obstacle shielding the returning blade, the present study aims at investigating
and improving further the output power of the threeblade Savonius turbine and im
proving the static torque, which measures the selfstarting capability of the turbine. For
this purpose, a modiﬁed design is considered, involving simultaneously an obstacle plate
shielding the returning blade and a ﬂow deﬂector (frontal guiding plates). Four geomet
rical properties are optimized simultaneously: 1) the position of an obstacle shielding
the returning blade; 2) the position of a deﬂector guiding the wind toward the advancing
blade; 3) the blade skeleton line and 4) the gap width.
The optimization process thus relies on free design variables that describe the posi
tion and angles of the plates, the blade shape (skeleton line) as well as the gap width s
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 100
Wind
X >R 1
X >R 2
R
Obstacle
Y
1
Advancing
Blade
Returning
Blade
Y
2
x
y
(0,0)
Deflector
Xd1
Xd2
Y
d
1
Y
d
2
(a)
P4
P (X ,Y ) 1 P1 P1
P5
r
r
Variablepoints
P (X ,Y ) 2 P2 P2
P (X ,Y ) 3 P3 P3
X
Y
(P0)
Fixedpoints
Blade
center
Turbineshaft
(b)
gw
Figure 6.22: Schematic description of the free optimization parameters characterizing a
threeblade Savonius rotor : a) plate parameters (X
1
, Y
1
, X
2
, Y
2
, X
d1
, Y
d1
, X
d2
and Y
d2
);
b) X
P1
, Y
P1
, X
P2
, Y
P2
, X
P3
and Y
P3
used to modify the blade shape; additionally, the gap
width g
w
.
(Fig. 6.22). At the end, ﬁfteen free parameters are thus considered (X
1
, Y
1
, X
2
, Y
2
, X
d1
,
Y
d1
, X
d2
, Y
d2
, X
P1
, Y
P1
, X
P2
, Y
P2
, X
P3
, Y
P3
and g
w
). The objective function considers
again only one output of the simulation, that should be maximized: the output power
coeﬃcient C
p
.
The mathematical optimization procedure described previously can now be em
ployed. This is done as usual for a constant speed ratio λ = 0.7, considering a ﬁxed
incident wind velocity U = 10 m/s. The parameter space considered in the optimization
has been deﬁned as documented in Table 6.5. These domains are selected to prevent any
domain overlap along the Y direction, to keep realistic blade shapes and to cover a wide
region for positioning the guiding plates. The reference point of the parameter space for
the blade skeleton line is point P
0
, which is the center of the original, semicylindrical
shape with radius r as shown in Fig. 6.22. The reference point for the remaining space
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 101
parameters (guiding plates and gap width) is the global center of turbine rotation. Dur
ing the calculations, a circular turbine shaft is included with a radius R
sh
computed
from R
sh
/R = 0.03.
0.53 0.24 0.24 0.94 0.24 0.24 1.2 1.1 1.88 1.76 1.88 0.7 1.88 1.88 0.03 0.024
1.2 0.24 1.1 0.24 1.1 0.94 0 1.65 0 2.6 1.1 0 1.1 0.7 0.18 0.38
X /r p1 X /r p2 X /r p3
Y /r p1 Y /r p2 Y /r p3
X /R d1 X /R d2 X /R 1 X /R 2
Y /R d1 Y /R d2 Y /R 1 X /R 2
s/R
CP
Bladeshape Guidingplatespositions
Gapwidth
0.153
Threeblade
Savoniusturbine
withoutguiding
plates
0.363
Optimum
configuration
Optimum
configuration
Figure 6.23: Input parameters of the optimization and power coeﬃcient represented
using parallel coordinates. The parameters of the optimal conﬁguration are connected
with a thick red line. The power coeﬃcient of the conventional threeblade turbine (semi
cylindrical shape) is also shown with a black circle.
The results presented in Fig. 6.23 indicate that the considered objective is indeed
considerably inﬂuenced by the ﬁfteen free parameters. As a whole, 240 diﬀerent geomet
rical settings have been evaluated by CFD, requesting 47 days of total computing time
on a standard PC. Relying on parallel computers and possibly carrying out each CFD
evaluation again in parallel [132] is clearly necessary when considering threedimensional
problems. Fortunately, this is quite straightforward to implement, so that researchers
having access to parallel clusters can solve corresponding problems within an acceptable
lapse of time.
The optimal conﬁguration (highest point in the right column in Fig. 6.23, all corre
sponding parameters being connected by a thick red line) can now readily be identiﬁed
for λ = 0.7. The corresponding geometry is shown in Fig. 6.24. The optimum parameter
values are listed in Table 6.6.
One instantaneous picture of the velocity ﬁeld is shown as an example in Fig. 6.25,
demonstrating that the employed grid captures all important ﬂow features in the vicinity
of the rotor and guiding plates. This is of course a dynamic process, diﬃcult to illustrate
in a static ﬁgure.
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 102
Table 6.5: Acceptable range for the input parameters (parameter space)
Parameter Minimum allowed Maximum allowed
Blade shape
X
P1
/r 0.53 1.2
Y
P1
/r −0.24 0.24
X
P2
/r 0.24 1.1
Y
P2
/r −0.94 −0.24
X
P3
/r 0.24 1.1
Y
P3
/r 0.24 0.94
Guiding plates
X
1d
/R −1.2 0.0
Y
d1
/R 1.1 1.65
X
2d
/R −1.88 0.0
Y
d2
/R 1.76 2.6
X
1
/R −1.88 −1.1
Y
1
/R −0.7 0.0
X
2
/R −1.88 −1.1
Y
2
/R −1.88 −0.7
Gap width
g
w
/R 0.03 0.18
Advancing
Blade
Deflector
Obstacle
Optimumdesign
Wind
U
Returningblade
=77.58 °
=81.13 °
Figure 6.24: Optimum conﬁguration obtained with the optimization procedure.
At λ = 0.7 the optimal point found by the optimization procedure corresponds to
an absolute increase of the power coeﬃcient by 0.207 compared with the conventional
threeblade Savonius turbine (semicylindrical blade shape). As a whole, this means a
relative increase of the performance (measured by the power output coeﬃcient) by 57%
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 103
Table 6.6: Optimal conﬁguration
Part Parameter Value Angle
Blade shape X
P1
/r 0.6077 
Y
P1
/r 0.1338
X
P2
/r 0.2735
Y
P2
/r 0.7136
X
P3
/r 0.7065
Y
P3
/r 0.5901
Guiding plates X
d1
/R 0.3089 γ = 81.13
◦
Y
d1
/R 1.436
X
d2
/R 0.4591
Y
d2
/R 2.388
X
1
/R 1.3638 β = 77.58
◦
Y
1
/R 0.1075
X
2
/R 1.691
Y
2
/R 1.5935
Gap width g
w
/R 0.0988 
Figure 6.25: Instantaneous velocity vectors magnitude (m/s) around the optimum con
ﬁguration (zoom) at the design point (λ = 0.7).
for the optimum design.
The performance of the optimal conﬁguration has been ﬁnally computed for the full
range of useful λvalues, as shown in Fig. 6.26. This ﬁgure demonstrates that the im
provement of both torque coeﬃcient and power output coeﬃcient is observed throughout
for all values of λ, compared to the conventional threeblade Savonius turbine. The rel
ative performance increase compared to the standard Savonius conﬁguration is always
higher than 50% in the usual operating range (0.6 ≤ λ ≤ 1), demonstrating again the
interest of the optimized conﬁguration.
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 104
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Speed ratio (l)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
T
o
r
q
u
e
C
o
e
f
f
i
c
i
e
n
t
(
C
m
)
Optimum design
Conventional threeblade
Savonius
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Speed ratio(l)
0
0.1
0.2
0.3
0.4
0.5
P
o
w
e
r
C
o
e
f
f
i
c
i
e
n
t
(
C
p
)
0
20
40
60
80
100
R
e
l
a
t
i
v
e
i
n
c
e
a
s
e
(
%
)
Conventional three blade
Savonius
Optimum design
% Relative increase
Figure 6.26: Performance of the optimized conﬁguration (red line) compared to the con
ventional threeblade Savonius turbine (black line): Top: torque coeﬃcient; Bottom:
power coeﬃcient. The corresponding relative increase compared to the standard conﬁg
uration is shown with blue line.
6.3.3.1 Selfstarting capability
For decentralized, lowcost windenergy applications, it is essential to obtain a self
starting system. To investigate this issue, the static torque exerted on the turbine at
a ﬁxed angle has been computed by CFD as a function of this angle θ. Figure 6.27
shows the obtained static torque coeﬃcient C
ms
for the optimal design compared to
the classical threeblade turbine. The experimental results of [48] for a conventional
threeblade turbine are also shown for comparison in Fig. 6.27. Due to periodicity, the
results are only plotted for θ between 0 and 120
◦
. Compared to the classical turbine,
these computations demonstrate that the modiﬁcations have a considerable and positive
eﬀect on the static torque coeﬃcient, except in a small range (90
◦
≤ θ ≤ 100
◦
). There,
the static torque coeﬃcient is less than the classical one, but remains strictly positive.
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 105
Averaging over all angle positions, C
ms
is increased by 0.091 for the optimum design.
0 30 60 90 120
Rotation angle(q)
0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
S
t
a
t
i
c
T
o
r
q
u
e
c
o
e
f
f
i
c
i
e
n
t
(
C
m
s
)
Our CFDresults:
Conventional Threeblade
Savonius Turbine
Optimumdesign
Exp. K. Irabu&J. Roy(2007)
Negative Torque (no selfstarting)
Figure 6.27: Static torque coeﬃcient C
ms
as a function of the ﬁxed rotor angle θ for the
optimal design (ﬁlled red squares) compared to the classical threeblade Savonius turbine
(blue plus). The experimental results of [48] are also shown for comparison (empty black
squares).
6.3.4 Conclusions on threeblade design
It is possible to increase the performance of the conventional threeblade design using
optimization by modifying the blade shape and gap width while placing suitable obstacle
and deﬂector plates. However, it is known from the literature that the twoblade design
is usual better in terms of power coeﬃcient. Therefore, in a last step, the conventional
twoblade design will be optimized along the same lines.
6.4 Optimal twoblade Savonius turbine
In this section, new designs will be step by step investigated and optimized to improve
the performance of the conventional twoblade Savonius turbine.
6.4.1 Obstacle plate
To achieve better performance, the position of an obstacle shielding the returning blade
of the Savonius turbine and possibly leading to a better ﬂow orientation toward the
advancing blade is ﬁrst optimized (Fig. 6.28).
Adding a shielding obstacle should in principle reduce the reverse moment, and as a
consequence the total moment of the turbine will be increased.
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 106
Wind
X >R 1
X >R 2
R
Obstacle
Y
1
Advancing
Blade
Returning
Blade
y
x
(0,0)
Y
2
Figure 6.28: Schematic description of the geometry and free optimization parameters
X
1
, Y
1
, X
2
used to modify the position of the shielding obstacle.
6.4.1.1 Selfstarting capability
It is essential to obtain a selfstarting system. To investigate this issue, the static torque
exerted on a turbine at a ﬁxed angle has been computed by CFD as a function of this
angle θ. Figure 6.29 shows the obtained static torque coeﬃcient C
m
s
obtained for three
diﬀerent positions of the obstacle plate as a function of θ. The experimental results
of [38] for the conventional turbine are also shown for comparison in Fig. 6.29. Due to
periodicity, the results are only plotted for θ between 0 and 180
◦
. These computations
demonstrate that the obstacle plate has a considerable and positive eﬀect on the static
torque coeﬃcient for the classical conﬁguration. The conventional Savonius turbine
shows a very large variation of the static torque coeﬃcient as a function of θ, with
negative values around θ = 140
◦
−170
◦
(no selfstarting). For all investigated positions
involving an obstacle, the negative torque region completely disappears, with a minimum
value of C
m
s
higher than 0.07. Apart from that, the evolution as a function of θ is similar
to that of the conventional turbine. As a whole, employing an obstacle plate improves
noticeably the selfstarting properties for the classical conﬁguration. A selfstarting
capability (C
m
s
> 0) is always obtained in principle at any angle, which is a major
advantage.
6.4.1.2 Optimization
Three degrees of freedom are left simultaneously to the OPAL optimizer: X
1
and Y
1
deﬁne the upper tip of the shielding obstacle; the value X
2
is then suﬃcient to deﬁne
the position of the lower tip, since Y
2
is taken constant, with Y
2
/R = −1.177 (Fig. 6.28).
With these three factors, the position of the obstacle plate is perfectly determined, and
the angle β can be deduced as well.
In a previous study [72], a small range of variation had been deﬁned for the parameter
space. Here, a much larger accessible domain has been prescribed, as documented in
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 107
0 30 60 90 120 150 180
Rotation angle(q)
0.2
0
0.2
0.4
0.6
0.8
1
S
t
a
t
i
c
T
o
r
q
u
e
C
o
e
f
f
i
c
i
e
n
t
(
C
m
s
)
Our CFD Results:
Savonius without obstacle
Y1/R=0.0 , b=73
o
Y1/R=0.24 , b=68.5
o
Y1/R=0.47, b=60.5
o
Exp. T. Hayashi et al (2005)
(no obstacle)
Figure 6.29: Static torque coeﬃcient as a function of the rotor angle θ for three diﬀerent
values of Y
1
choosing X
1
/R = −1.4 and X
2
/R = −1.76.
Table 6.7. In this manner, a truly optimal solution can be obtained considering a large
set of possible parameters. Of course, when choosing the parameter space, it must
be checked that the shielding obstacle cannot come into direct contact with the rotor
( X
1
> R and X
2
> R). The corresponding positions of the obstacle all lead to
conﬁgurations that shield partially the returning blade.
Table 6.7: Acceptable range for the input parameters
Parameter minimum maximum
X
1
/R −1.88 −1.017
Y
1
/R −0.88 0.0
X
2
/R −1.88 −1.017
The results presented in Fig. 6.30 indicate that the considered objective is consider
ably inﬂuenced by the three free parameters X
1
, Y
1
and X
2
. As a whole, 210 diﬀerent
geometrical settings have been evaluated by CFD, requesting 24 days of computing
time on a standard PC. The optimal conﬁgurations (highest point in the right column
in Fig. 6.30, all corresponding parameters being connected by a thick red line) can
now be identiﬁed for λ = 0.7. The optimum obstacle position, optimum angle β and
corresponding optimal power coeﬃcient C
p
are listed in Table 6.8.
When compared with the standard Savonius turbines (without shielding obstacle),
the optimal point found by the optimization procedure corresponds to an absolute in
crease of the power coeﬃcient by 0.068 at λ = 0.7. This means a relative increase of
the performance (measured by the power output coeﬃcient) by 27.3% for the twoblade
Savonius turbine, compared to the conventional Savonius design without any obstacle.
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 108
X1/R
Y1/R X2/R Powercoeff.
1.88
0.88
1.88 0.133
1.017
0.0
1.017 0.2503
Optimalconfiguration
Without
obstacle
0.182
Figure 6.30: Input parameters of the optimization and power coeﬃcient represented
using parallel coordinates. The parameters of the optimal conﬁgurations are connected
with a thick red line. The power coeﬃcient of the conventional turbine is shown with a
black circle.
Table 6.8: Optimal conﬁgurations (obstacle position and angle)
Design parameter value β power coeﬀ. C
p
Twoblade Savonius turbine X
1
/R −1.23830
Y
1
/R −0.45390 (100.83
◦
) 0.2503
X
2
/R −1.09993
Note that we have been at ﬁrst surprised to obtain an optimal geometry leading
to β > 90
◦
for the twoblade turbine. After a thorough ﬂow analysis, exempliﬁed
in Fig. 6.31, it is possible to understand ﬁnally that the ﬂow direction induced by
the obstacle at β = 100.83
◦
is indeed optimal for the advancing blade in the relative
reference frame. Furthermore, for most conﬁgurations associated with β < 90
◦
, the
ﬂow behind the shielding obstacle points partly toward the returning blade and thus
increases the reverse moment; this eﬀect is reduced for the optimal conﬁguration. The
optimal geometry of the shielding obstacle is of course highly dependent on the speciﬁc
rotor conﬁguration.
6.4.1.3 Oﬀ design performance
The performance of the optimal conﬁguration has been ﬁnally computed for the full
range of useful λvalues, as shown in Figure 6.32 . This ﬁgure demonstrate that the im
provement of both torque coeﬃcient and power output coeﬃcient is observed throughout
for all values of λ, compared to the conventional Savonius turbine without obstacle. The
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 109
Figure 6.31: Instantaneous ﬂow structure when the advancing blade is in vertical position
for the optimal conﬁguration at λ = 0.7: Zoom on the vicinity of the turbine (full CFD
domain is much larger).
absolute gain for C
p
increases even slightly with λ at ﬁrst, the relative increase being
highest for the largest values of λ considered in the present study.
6.4.1.4 Practical realization
From the technical point of view, many existing systems already rely on a tail vane
for optimal alignment into the wind direction. A similar technical solution would be
used for the Savonius turbine using the obstacle. In this manner the orientation of
the system can be simply, eﬃciently and automatically controlled. As a whole, the
optimized conﬁguration is only slightly more complex, more expensive and heavier than
the original system. Therefore, the improved power and torque coeﬃcients should easily
compensate these drawbacks within a short time after installation.
6.4.2 Optimal blade shape with obstacle plate
In the last section, the eﬃciency of the classical Savonius turbine has been increased
by placing in an optimal manner an obstacle plate shielding the returning blade. The
study now aims at improving further the output power of the Savonius turbine as well
as the static torque, which measures the selfstarting capability of the turbine. In order
to achieve both objectives, the geometry of the blade shape is now optimized in presence
of the obstacle plate. Six free parameters are considered in this optimization process.
6.4.2.1 Optimization
The free design variables considered for the optimization describe the blade skeleton line
for a constant blade thickness of 2 mm. For this purpose, six parameters are considered
(X
P1
, Y
P1
, X
P2
, Y
P2
, X
P3
and Y
P3
, see Fig. 6.33). The obstacle plate is kept ﬁxed in the
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 110
0.2 0.4 0.6 0.8 1 1.2 1.4
Speed ratio (l)
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
T
o
r
q
u
e
c
o
e
f
f
i
c
i
e
n
t
(
C
m
)
0.2 0.4 0.6 0.8 1 1.2 1.4
Speed ratio (l)
0
0.05
0.1
0.15
0.2
0.25
0.3
P
o
w
e
r
c
o
e
f
f
i
c
i
e
n
t
(
C
p
)
Savoniuswithout obstacle
Savoniuswithobstacle
(optimal configuration)
%Relativeincrease
0
20
40
60
80
100
R
e
l
a
t
i
v
e
i
n
c
e
a
s
e
(
%
)
(a)
(b)
%
Figure 6.32: Performance of the optimized conﬁguration (red line) compared to the con
ventional twoblade Savonius turbine without obstacle (black line): a: torque coeﬃcient;
b: power coeﬃcient. The corresponding relative increase is shown with blue stars.
optimal position identiﬁed in Section 6.4.1 (X
1
/R = −1.2383, Y
1
/R = −0.4539, X
2
/R =
−1.0999 and Y
2
/R = −1.1770, which leads to an angle β = 100.8
◦
, see Fig. 6.28). The
six shape parameters are suﬃcient to determine uniquely the geometry of the system,
since the obstacle position is ﬁxed. The objective function considers only one output of
the simulation, that should be maximized as far as possible: the power coeﬃcient C
p
.
The parameter space considered in the optimization has been deﬁned as documented
in Table 6.9. These domains are selected to prevent any domain overlap along the Y 
direction and to keep realistic blade shapes. The reference point of the parameter space
is point P
0
, which is the center of the original, semicylindrical shape with radius r as
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 111
P4
P (X ,Y ) 1 P1 P1
P5
r
r
Variablepoints
P (X ,Y ) 2 P2 P2
P (X ,Y ) 3 P3 P3
X
Y
(P0) Fixedpoints
Blade
center
Figure 6.33: Schematic description of the free optimization parameters
X
P1
, Y
P1
, X
P2
, Y
P2
, X
P3
and Y
P3
used to modify the blade shape.
shown in Fig. 6.33. During the calculations, a circular turbine shaft is included with a
ﬁxed radius R
sh
computed from R
sh
/R = 0.03.
Table 6.9: Acceptable range for the input parameters for the blade shape
Parameter Minimum allowed Maximum allowed
X
P1
/r 0.53 1.47
Y
P1
/r −0.24 0.24
X
P2
/r 0.24 1.2
Y
P2
/r −0.94 −0.24
X
P3
/r 0.24 1.2
Y
P3
/r 0.24 0.94
The results presented in Fig. 6.34 indicate that the considered objective is consider
ably inﬂuenced by the six free parameters, X
P1
, Y
P1
, X
P2
, Y
P2
, X
P3
and Y
P3
, and thus
by the blade shape. As a whole, 210 diﬀerent geometrical settings have been evaluated
by CFD, requesting one and a half month of total computing time on a standard PC.
The optimal conﬁguration (highest point in the right column in Fig. 6.34, all cor
responding parameters being connected by a thick red line) can now be identiﬁed for
λ = 0.7. The optimum point positions and corresponding optimal power coeﬃcient C
p
are listed in Table 6.10.
Table 6.10: Optimal conﬁguration
X
P1
/r Y
P1
/r X
P2
/r Y
P2
/r X
P3
/r Y
P3
/r Optimal power coeﬀ. C
p
0.6909 0.0386 0.3940 −0.6067 0.6389 0.6357 0.298
At λ = 0.7 the optimal point found by the optimization procedure corresponds to
an absolute increase of the power coeﬃcient by 0.116 compared with the conventional
Savonius turbine, respectively by 0.0475 compared with the conventional Savonius rotor
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 112
0.53
0.24 0.24 0.94 0.24 0.24 0.1
0.3
0.94 1.2 0.24 1.2 0.24 1.47
X1/r Y1/r X2/r Y2/r X3/r
Y3/r Cp
Optimalconfiguration
0.298
0.182
Savoniuswithout
obstacle
ClassicalSavonius
withobstacle
0.250
Figure 6.34: Input parameters of the optimization and power coeﬃcient represented
using parallel coordinates. The parameters of the optimal shape are connected with a
thick red line. The power coeﬃcient of the conventional turbine is shown with a gray
circle. The power coeﬃcient of the conventional turbine (semicylindrical shape) with
obstacle plate is also shown with a black circle.
(semicylindrical blade shape) with obstacle plate. As a whole, this means a relative
increase of the performance (measured by the power output coeﬃcient) by 38.9% for
the optimum shape with obstacle plate, compared to the conventional Savonius design
without obstacle.
Note that this new shape (Fig. 6.35) is only optimum in combination with the em
ployed obstacle plate. Indeed, the modiﬁed shape exploits best the ﬂow redirection in
duced by the obstacle. As a consequence, the reverse moment is reduced by the shielding
obstacle; simultaneously, the modiﬁed shape leads to an increase of the positive moment
of the advancing blade.
6.4.2.2 Oﬀ design performance
The performance of the optimal conﬁguration has been ﬁnally computed for the full
range of useful λvalues, as shown in Fig. 6.36. This ﬁgure demonstrates that the im
provement of both torque coeﬃcient and power output coeﬃcient is observed throughout
for all values of λ, compared to the classical Savonius turbine. The absolute gain for C
p
and C
m
is even higher for lower λvalues. The relative performance increase compared to
the standard Savonius conﬁguration is always higher than 30% in the eﬀective operating
range.
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 113
ClassicalSavonius
turbine
Optimalshape
Advancingblade
Returningblade
obstacle
Wind Wind
Returningblade
Figure 6.35: Optimum conﬁguration (right) obtained with the optimization procedure
compared to the classical Savonius turbine (semicylindrical shape: left).
6.4.2.3 Selfstarting capability
The static torque exerted on the turbine at a ﬁxed angle has been computed by CFD as
a function of this angle θ. Figure 6.37 shows the obtained static torque coeﬃcient C
ms
obtained for the optimal conﬁguration compared to the classical turbine with obstacle,
as a function of θ. The experimental results of [38] for a conventional turbine are
also shown for comparison in Fig. 6.37. Compared to the standard Savonius without
obstacle plate, these computations demonstrate that the modiﬁed blade shape has a
considerable and positive eﬀect on the static torque coeﬃcient. In the present, optimal
conﬁguration the negative torque region completely disappears, with a minimum value
of C
ms
higher than 0.2. Apart from that, the evolution as a function of θ is similar to
that obtained with the conventional shape involving an obstacle plate. A selfstarting
capability (C
ms
> 0) is always obtained at any angle in both cases, which is a major
advantage.
6.4.3 Optimal Savonius turbine with two guiding plates
From the summary of the most important modiﬁcation proposals listed in Chapter 3,
the highest advantage of the Savonius turbine is its robustness. Modiﬁcations should not
involve an exceedingly complex or expensive design. Therefore, simple guiding plates
seem to be the best compromise between the increase of eﬃciency and the increase of
cost and complexity. In the present section we will thus investigate numerically the
eﬀect of two combined guiding plates: a deﬂector plate is employed to obtain the best
possible ﬂow conditions for the advancing blade, while an obstacle plate shields partly
the returning blade (Fig. 6.38)
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 114
0.2 0.4 0.6 0.8 1 1.2 1.4
Speed ratio (l)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
T
o
r
q
u
e
c
o
e
f
f
i
c
i
e
n
t
(
C
m
)
0.2 0.4 0.6 0.8 1 1.2 1.4
Speed ratio (l)
0
0.04
0.08
0.12
0.16
0.2
0.24
0.28
0.32
P
o
w
e
r
c
o
e
f
f
i
c
i
e
n
t
(
C
p
)
0
20
40
60
80
100
R
e
l
a
t
i
v
e
i
n
c
e
a
s
e
(
%
)
Optimal shape
(noncylindrical blade)
Savonius with obstacle
Savonius without obstacle
%Relative increase
(a)
(b)
%
Figure 6.36: Performance of the optimized conﬁguration (red line) compared to the con
ventional Savonius turbine with and without obstacle plate (blue and black, respectively):
a: torque coeﬃcient; b: power coeﬃcient. The corresponding relative increase compared
to the standard conﬁguration is shown with green line.
6.4.3.1 Optimization
The optimization work will be carried out for obstacle and deﬂector simultaneously,
with eight free space parameters (X
1
, Y
1
, X
2
, Y
2
, X
d1
, Y
d1
, X
d2
and Y
d2
), which together
are suﬃcient to obtain clearly the position, length and angles of these guiding plates
(obstacle and deﬂector) as shown in see Fig. 6.38. The parameter spaces considered in
the optimization have been deﬁned as documented in Table 6.11.
The results in Fig. 6.39 indicate that the considered objective is considerably inﬂu
enced by the eight free parameters, X
d1
, Y
d1
, X
d2
, Y
d2
, X
1
, Y
1
, X
2
and Y
2
, and thus by
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 115
0 30 60 90 120 150 180
Rotor angle (q)
0.2
0
0.2
0.4
0.6
0.8
1
1.2
S
t
a
t
i
c
T
o
r
q
u
e
C
o
e
f
f
i
c
i
e
n
t
(
C
m
s
)
Our CFD results:
without obstacle
Classical Savonius
with obstacle
Optimal shape
Exp. T. Hayashi et al (2005)
(no obstacle)
Negative torque (no selfstarting)
Figure 6.37: Static torque coeﬃcient C
ms
as a function of the ﬁxed rotor angle θ for
the optimal conﬁguration compared to the standard Savonius turbine with and without
obstacle plate. For this last case, the experimental results of [38] are also shown for
comparison.
Wind
X >R 1
X >R 2
R
Obstacle
Y
1
Advancing
Blade
Returning
Blade
Y
2
Deflector
Xd1
Xd2
Y
d
1
Y
d
2
y
X
(0,0)
Figure 6.38: Schematic description of the geometry of Savonius turbine with frontal
guiding plates.
the positions, angles and lengths of the guiding plates. As a whole, 210 diﬀerent geo
metrical settings have been evaluated by CFD, requesting 40 days of total computing
time on a standard PC.
The optimum point positions are listed in Table 6.12. These optimum parameters
lead to the corresponding angles of 92.35
◦
and 82.15
◦
for obstacle and deﬂector, respec
tively and lead also to lengths of the guiding plates L
o
/d = 1.094 and L
d
/d = 0.782, as
shown in Fig. 6.40.
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 116
Table 6.11: Acceptable range for the input parameters (parameter space)
Parameter Minimum Maximum
X
d1
/d −1.04 0
Y
d1
/d 0.9 1.87
X
d2
/d −1.66 0
Y
d2
/d 1.3 1.87
X
1
/d −1.66 −0.9
Y
1
/d −0.78 0
X
2
/d −1.66 −0.9
Y
2
/d −1.66 −0.53
X /d d1 Y /d d1 X /d d2
1.04 0.9 1.66
0.0 1.87
0.0
1.3
Y /d d2
1.87
0.18
Cp
0.36
X /d 1 Y /d 1 X /d 2
0.9 0 0.9
Y /d 2
0.53
1.66 1.66 0.78 1.66
0.182
Optimalconf.
Conventional
Savonius
0.349
Figure 6.39: Input parameters of the optimization and power coeﬃcient represented
using parallel coordinates. The parameters of the optimal design are connected with a
thick red line. The power coeﬃcient of the conventional turbine is shown with a black
circle.
At λ = 0.7 the optimal points found by the optimization procedure corresponds
to an absolute increase of the power coeﬃcient by 0.167 compared with the conven
tional Savonius turbine. This means a relative increase of the performance (measured
by the power output coeﬃcient) by 47.85% for the optimum design compared to the
conventional Savonius design.
6.4.3.2 Oﬀ design performance
The performance of the optimal conﬁguration has been next computed for the full range
of useful λvalues, as shown in Fig. 6.41. The improvement of both torque coeﬃcient
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 117
Table 6.12: Optimum parameters of guiding plates
X
d1
/d 0.066981
Y
d1
/d 1.509671
X
d2
/d 0.17377
Y
d2
/d 2.28434
X
1
/d 1.134209
Y
1
/d 0.286166
X
2
/d 1.0892
Y
2
/d 1.37918
L
d
opt.
opt.
L
o
Wind
Figure 6.40: Optimum conﬁguration of guiding plates.
and power output coeﬃcient is observed throughout for all values of λ. The absolute
gain for C
p
and C
m
is even higher for lower λvalues. The relative performance increase
compared to the standard Savonius conﬁguration is always higher than 35% in the
eﬀective operating range.
6.4.3.3 Proﬁled guiding plates
There is no reason to believe that ﬂat guiding plates are the best choice. Therefore
we started an optimization for nonﬂat guiding plates. Optimization is repeated for
obstacle and deﬂector simultaneously, with fourteen free space parameters describing
the optimum shape of both guiding plates. These fourteen parameters (X
d1
, X
d2
Y
d2
,
X
d3
, Y
d3
, X
d4
, Y
d4
, X
1
X
2
, Y
2
, X
3
, Y
3
, X
4
and Y
4
, while Y
1
and Y
d1
are constant and
taken from the previous section, Table 6.12). Fourteen parameters are suﬃcient to
deﬁne clearly the position and shapes of these guiding plates, as shown in Fig. 6.42.
The parameter space considered in the optimization has been deﬁned as documented in
Table 6.13.
The results are shown in Fig. 6.43. More than 200 diﬀerent geometrical settings have
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 118
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Speed ratio (l)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
T
o
r
q
u
e
c
o
e
f
f
i
c
i
e
n
t
(
C
m
)
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Speed ratio (l)
0
0.1
0.2
0.3
0.4
0.5
0.6
P
o
w
e
r
c
o
e
f
f
i
c
i
e
n
t
(
C
p
)
Optimal conf.
(Savoniuswithguidingplates)
Conventional Savoniusturbine
%Relativeincrease
0
20
40
60
80
100
R
e
l
a
t
i
v
e
i
n
c
r
e
a
s
e
(
%
)
(a)
(b)
Figure 6.41: Performance of the optimized conﬁguration (green line) compared to the
conventional Savonius turbine (blue line): a: torque coeﬃcient; b: power coeﬃcient.
The corresponding relative increase compared to the standard conﬁguration is shown
with black line.
been evaluated by CFD, requesting 40 days of total computing time on a standard PC.
However, by comparison with the results of the ﬂat guiding plates (see Section 6.4.3.1),
the diﬀerence in power coeﬃcient is only minimal (around 0.01). Therefore, from
the manufacturing point of view, the slight improvement in the performance does not
compensate the additional complexity, cost and weight of the proﬁled guiding plates
(Fig. 6.44). As a consequence, only ﬂat guiding plates are considered in the ﬁnal opti
mization step.
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 119
Wind
X >R 4
X >R 1
R
Y
4
Advancing
Blade
Returning
Blade
Y
1
Xd4
Xd2
Y
d
4
Y
d
2
y
X
(0,0)
Xd1
Xd3
Y
d
1
Y
d
3
Y
3
Y
2
X >R 2
X >R 3
Figure 6.42: Schematic description of the geometry of Savonius turbine with proﬁled
guiding plates.
Table 6.13: Acceptable range for the input parameters (parameter space)
Parameter Minimum Maximum
Deﬂector
X
d1
/d −1.5 0
X
d2
/d −1.5 0
Y
d2
/d 2 2.4
X
d3
/d −1.5 0
Y
d3
/d 1.5 2
X
d4
/d −1.5 0
Y
d4
/d 1.15 1.5
Obstacle
X
1
/d −1.5 −0.8
X
2
/d −1.5 −1.15
Y
2
/d −1.5 −1
X
3
/d −1.5 −1.15
Y
3
/d −1 −0.5
X
4
/d −1.5 −1.15
Y
4
/d −0.5 0
6.5 Final optimization of Savonius turbine
The optimization steps of increasing complexity described in the previous sections seem
very promising. Having now full conﬁdence in the process, a last step involving all
characteristic geometrical parameters is now attempted.
The eﬀect of blade shape and guiding plates positions will be incorporated simul
taneously during the optimization. In this aggressive optimization work, we will opti
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 120
2 0.18 1.5 1.5 1.5 1.5 1.5 1.15 1.5 1.5 1.5 1.5 1 1.5 0.5
2.4 0.37
0 0
0 2
0
1.5
0.8
1.15
1
1.15
0.5
1.15
0 X /d d1 Y /d d2
X /d d3
Y /d d3 Cp
X /d 1 Y /d 2
X /d 3
Y /d 3 X /d d2 X /d d4
Y /d d4 X /d 2
Y /d 4
X /d 4
0.182
Classical
Savonius
Optimalconf.(Savoniuswithcurvedguidingplates)
0.361
Figure 6.43: Input parameters of the optimization and power coeﬃcient represented
using parallel coordinates. The parameters of the optimal designs are connected with a
thick red line. The power coeﬃcient of the conventional turbine is shown with a black
circle.
mize sixteen parameters simultaneously; eight parameters for both guiding plates (X
1
,
Y
1
, X
2
, Y
2
, X
d1
, Y
d1
, X
d2
and Y
d2
, see Fig. 6.45a), considered ﬂat for the reasons ex
plained in the previous section. The blade shape is described by ﬁve points, two ﬁxed
points (P
4
and P
5
) and three movable points (P
1
, P
2
, P
3
). Every point has two coor
dinates (X
Pi
, Y
Pi
) this means we have another six free parameters (X
P1
, Y
P1
, X
P2
, Y
P2
,
X
P3
and Y
P3
Fig. 6.45b). Knowing all 5 points, the full proﬁle is reconstructed us
ing standard splines. Another two parameters will taken into consideration for internal
spaces of the turbine (a and e see Fig. 6.45c). The objective function considers only one
output of the simulation that should be maximized as far as possible: the output power
coeﬃcient C
p
. The employed optimization parameters have been listed in chapter 5.
The parameter space considered in the optimization has been deﬁned as documented
in Table 6.14.
Optimization results are shown in Fig. 6.46. 310 diﬀerent geometrical settings have
been evaluated by CFD, requesting two and a half months of total computing time on
a standard PC for this ambitious analysis.
The optimal conﬁguration (highest point in the right column in Fig. 6.46, all corre
sponding parameters being connected by a thick red line) can now readily be identiﬁed
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 121
Optimaldeflector
shape
Optimalobstacle
shape
Figure 6.44: Optimum conﬁgurations of curved guiding plates.
Table 6.14: Acceptable range for the input parameters (parameter space)
Parameter Minimum Maximum
Blade shape
X
P1
/r 0.53 1.47
Y
P1
/r −0.24 0.24
X
P2
/r 0.24 1.2
Y
P2
/r −0.94 −0.24
X
P3
/r 0.24 1.2
Y
P3
/r 0.24 0.94
Guiding plates
X
d1
/d −1.04 0
Y
d1
/d 0.9 1.87
X
d2
/d −1.66 0
Y
d2
/d 1.3 1.87
X
1
/d −1.66 −0.9
Y
1
/d −0.78 0
X
2
/d −1.66 −0.9
Y
2
/d −1.66 −0.53
Internal spaces
a/d −0.09 0.18
e/d 0.05 0.28
for λ = 0.7. The optimum parameters are listed in Table 6.15, leading to the correspond
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 122
Wind
X >R 1
X >R 2
R
Obstacle
Y
1
Advancing
Blade
Returning
Blade
Y
2
Deflector
Xd1
Xd2
Y
d
1
Y
d
2
P4
P (X ,Y ) 1 P1 P1
P5
r
r
Variablepoints
P (X ,Y ) 2 P2 P2
P (X ,Y ) 3 P3 P3
X
Y
(P0) Fixedpoints
Blade
center
Turbineshaft
(b)
a
e
(c)
(a)
Figure 6.45: Schematic description of the geometry and free optimization parameters.
ing angles of 90.41
◦
and 94.13
◦
for obstacle and deﬂector, respectively. The lengths of
the guiding plates are L
o
/d = 0.7597 and L
d
/d = 1.048, as shown in Fig. 6.47.
Note that this new shape (Fig. 6.47) is only optimum in combination with the em
ployed guiding plates. Indeed, the modiﬁed shape exploits best the ﬂow redirection
induced by the guiding plates. As a consequence, the reverse moment is reduced by
the shielding obstacle and the ﬂow redirected by the deﬂector to the advancing blade;
simultaneously, the modiﬁed shape leads to an increase of the positive moment of the
advancing blade.
One instantaneous picture of the ﬂow ﬁeld (pressure, velocity magnitude and veloc
ity vectors) is shown as an example in Fig. 6.48 for the ﬁnally optimized design and for
the classical semicylindrical turbine with guiding plates, demonstrating that the em
ployed CFD captures all important ﬂow features in the vicinity of the rotor and guiding
plates. This is of course a dynamic process, diﬃcult to illustrate in a static ﬁgure. The
pressure diﬀerence obtained for the optimal design is much larger, explaining the better
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 123
1.04
0.9
1.66
1.3
0.16
1.66
0.78
1.66 0.09
0.05
0.01
Xd1/d
Yd1/d
Xd2/d
0.0
1.87
0.0
Yd2/d
1.87
Cp
0.45
X /d 1
Y1/d
X2/d
0.9
0
0.9
Y2/d
0.53 0.94
1.2
0.24
1.2
0.24
1.47
X1/r
Y1/r
X2/r
Y2/r
X3/r
Y3/r
e/d
0.28
a/d
0.18
0.53
0.24
0.24
0.94
0.24
0.24
Classical
Savonius
0.182
Optimalconf.(modifiedSavoniuswithguidingplates)
0.436
Figure 6.46: Input parameters of the optimization and power coeﬃcient represented
using parallel coordinates. The parameters of the optimal design are connected with a
thick red line. The power coeﬃcient of the conventional turbine is shown with a black
circle.
Table 6.15: Optimum conﬁgurations
Blade shape
X
P1
/r 0.54822
Y
P1
/r 0.19762
X
P2
/r 0.34849
Y
P2
/r −0.37885
X
P3
/r 0.54593
Y
P3
/r 0.770346
Guiding plates
X
d1
/d −0.41882
Y
d1
/d 1.24505
X
d2
/d −0.3433
Y
d2
/d 2.29074
X
1
/d −1.2828
Y
1
/d −0.4037
X
2
/d −1.27654
Y
2
/d −1.16339
Internal spaces
a/d −0.00635
e/d 0.18286
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 124
opt.
L
d
Wind
L
o
opt.
Figure 6.47: Optimum design of Savonius turbine with guiding plates.
performance.
6.5.1 Oﬀ design performance
It is now important to check how this gain would change as a function of λ, since such a
turbine must be able to work also for oﬀdesign conditions. Therefore, the performance
of the optimal conﬁguration has been ﬁnally computed for the full range of useful λ
values, as shown in Fig. 6.49. This ﬁgure demonstrates that the improvement of both
torque coeﬃcient and power output coeﬃcient is observed throughout for all values of
λ, compared to the conventional Savonius turbine. The absolute gain for C
p
and C
m
is even higher for lower λvalues. The relative performance increase compared to the
standard Savonius conﬁguration is always higher than 40% in the eﬀective operating
range.
6.5.2 Selfstarting capability
To investigate the selfstarting capability of the system, the static torque exerted on
a turbine at a ﬁxed angle has been computed by CFD as a function of this angle θ.
Figure 6.50 shows the obtained static torque coeﬃcient C
ms
obtained for the optimal
conﬁguration compared to the classical turbine, as a function of θ. The experimental
results of [38] for a conventional turbine are again shown for comparison. Compared
to the classical Savonius, these computations demonstrate that the new design has a
considerable and positive eﬀect on the static torque coeﬃcient. For the optimal con
ﬁguration the negative torque region completely disappears, with a minimum value of
C
ms
of about 0.2. A selfstarting capability (C
ms
> 0) is always obtained at any angle
in both cases, which is an essential property.
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 125
Pressuredistribution
Velocityvectordistribution
Velocitydistribution
(a) (b)
Figure 6.48: Instantaneous ﬂow ﬁelds around optimum conﬁgurations (zoom) at the de
sign point (λ = 0.7), static pressure (Pa), velocity and velocity vector magnitudes (m/s);
a) classical Savonius with optimal guiding plates, b) optimal Savonius with optimal guid
ing plates. Note that the color scales are identical to facilitate comparisons.
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 126
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Speed ratio (l)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
T
o
r
q
u
e
c
o
e
f
f
i
c
i
e
n
t
(
C
m
)
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Speed ratio (l)
0
0.1
0.2
0.3
0.4
0.5
0.6
P
o
w
e
r
c
o
e
f
f
i
c
i
e
n
t
(
C
p
)
Optimal conf.
(modifieddesign)
Conventional Savoniusturbine
0
20
40
60
80
100
R
e
l
a
t
i
v
e
i
n
c
r
e
a
s
e
(
%
)
Optimal conf.
(modified design)
Conventional Savonius turbine
%Relative increase
(a) (b)
Figure 6.49: Performance of the optimized conﬁguration (red line) compared to the
conventional Savonius turbine (black line): a: torque coeﬃcient; b: power coeﬃcient.
The corresponding relative increase compared to the classical conﬁguration is shown with
blue line.
0 30 60 90 120 150 180
Rotor angle (q)
0.2
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
S
t
a
t
i
c
T
o
r
q
u
e
C
o
e
f
f
i
c
i
e
n
t
(
C
m
s
)
Our CFD results:
Optimal modified Savonius
with guiding plates
Savonius without guiding plates
Exp. T. Hayashi et al (2005)
(no obstacle)
Negative torque (no selfstarting)
Figure 6.50: Static torque coeﬃcient C
ms
as a function of the ﬁxed rotor angle θ for the
optimal conﬁguration compared to the conventional Savonius turbine. For this last case,
the experimental results of [38] are also shown for comparison.
6.6 Preliminary experimental tests in windtunnel
It is now important to check the CFD results experimentally. Therefore, we have con
structed two small models for Savonius turbine, one model for the conventional two
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 127
blade Savonius and an other for the optimized shape with shielding obstacle. The main
objective of this work is to check that the optimized design is superior to the original
design.
InstallationofSavoniusturbine
atthetestsection
Figure 6.51: Model installation with open wind tunnel.
The wind tunnel operates with open or closed test section. The maximum ﬂow rate
for open test section is 40 m/s (Fig. 6.51). When closed, velocity up to nearly 60 m/s
is possible. The setting of the load is carried out by a brake handle, ﬁnetuned by a
screw. To get the torque / speed curve, the signal of the torque sensor was transmitted
to a computer, which also is used to control the wind speed of the wind tunnel. The
software used for both tasks is LabView.
Five test series were run for each rotor design. Before each measurement the oﬀ
set of the torque sensor was set to zero. The model has been submitted to increasing
wind speed, selfstarting, until a nominal wind speed of 20 m/s was achieved. Once
the rotor speed reaches a constant value, the load was progressively increased. Then,
output torque and rotation speed of the rotor have been recorded at every load to get
performance curves for the two designs.
First results (Fig. 6.52) indeed show increased power coeﬃcient and torque coeﬃcient
for the optimized design.
Technical diﬃculties (limited accuracy of the torque meter, vibrations of the set
up, rapid wear of the employed brake) have prevented up to now a more thorough
comparison.
6.7 Conclusions on Savonius turbine
The conventional Savonius turbine is a promising concept for smallscale windenergy
systems, but suﬀers from a poor eﬃciency. Therefore, the major objective of the present
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 128
0,00
0,05
0,10
0,15
0,20
0,25
0,30
0,35
0,40
0,4 0,6 0,8 1,0 1,2 1,4 1,6
?
c
m
cp1 cp2 cp3 cp4 cp5 Trend, gemittelt
0,00
0,05
0,10
0,15
0,20
0,25
0,4 0,6 0,8 1,0 1,2 1,4 1,6
?
c
p
cp1 cp2 cp3 cp4 cp5 Trend, gemittelt
0,00
0,05
0,10
0,15
0,20
0,25
0,30
0,35
0,40
0,4 0,6 0,8 1,0 1,2 1,4 1,6
?
c
m
cp1 cp2 cp3 cp4 cp5 Trend, gemittelt
0,00
0,05
0,10
0,15
0,20
0,25
0,4 0,6 0,8 1,0 1,2 1,4 1,6
?
c
p
cp1 cp2 cp3 cp4 cp5 Trend, gemittelt
C
p
C
p
C
m
C
p
C
m
Meanvalue
Meanvalue
Meanvalue
a)ConventionalSavoniusturbine b)OptimizedSavoniusturbine
Figure 6.52: Selected experimental results [64] a) conventional Savonius turbine; b)
Optimized design.
study was to identify an optimal design, leading to higher values of the power coeﬃcient
and of the static torque, thus obtaining a higher eﬃciency and better selfstarting capa
bility. For this purpose, many designs have been successively introduced and optimized
in this chapter.
• Threeblade Savonius turbine
After some preliminary steps, all the geometrical parameters are simultaneously
taken into consideration during optimization. Therefore, the blade shape, position
and angles of the guiding plates and gap width have been optimized in a fully au
tomatic manner, in order to obtain the best possible performance, as measured by
the power coeﬃcient C
p
. The optimization relies on evolutionary algorithms, while
all geometrical conﬁgurations are evaluated by CFD. This optimization procedure
is able to identify considerably better conﬁgurations than the conventional three
blade Savonius turbine. The best one leads in particular to a relative increase of
the power output coeﬃcient by 57% at λ = 0.7. A performance gain of at least
25% is found for the full operating range of the conventional design. At the same
time, the operating range is extended up to λ = 1.5. A peak power coeﬃcient of
C
p
0.39 is obtained for λ = 0.9. This positive eﬀect is also observed for the
torque coeﬃcient. The optimal design still ensures selfstarting capability for all
rotating angles.
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 129
• Twostage Savonius turbine
Figure 6.53: Successive optimization steps for the twoblade Savonius turbine.
Here again, preliminary steps have been followed by an optimization involving all
important parameters. This optimization procedure is able to identify a tremen
dously better conﬁguration than the conventional Savonius turbine, leading in
particular to a relative increase of the power output coeﬃcient by 58% at λ = 0.7.
A performance gain of at least 35% is found for the full conventional operating
range (0.3 ≤ λ ≤ 1.4). The optimal design shows a much larger range and still
delivers power at λ = 1.7. This positive eﬀect is also observed for the torque
coeﬃcient. The optimal design furthermore leads to selfstarting capability at
any angle, at the diﬀerence of the conventional Savonius. Therefore, this opti
mal conﬁguration appears indeed to be very promising for lowpower wind energy
generation in urban areas.
The successive steps of the optimization are documented in Fig. 6.53. At λ = 0.7, the
power coeﬃcient has been increased from C
p
0.18 to C
p
0.44 thanks to optimization.
For the best design, a peak C
p
of 0.48 is obtained for λ = 0.9, approaching Betz’ limit.
As a whole, and supporting previous ﬁndings from the literature, the twoblade
design appears superior to the threeblade conﬁguration in terms of low weight, cost,
as well as eﬃciency and operating range. Selfstating capabilities are similar, with only
a very slight advantage for the threeblade design. Therefore, the optimized twoblade
conﬁguration should indeed be very useful to supplement wind energy conversion.
Chapter 7
Wells turbine: Concurrent
optimization
7.1 Introduction
Wells turbine is a selfrectifying air ﬂow turbine employed to convert the pneumatic
power of the air stream induced by an Oscillating Water Column into mechanical energy.
Standard Wells turbines have several wellknown disadvantages: a very low tangential
force, leading to a low power output from the turbine; a high undesired axial force;
usually a low aerodynamic eﬃciency and a limited range of operation due to stall. All the
theoretical and experimental investigations listed in the section 3.2.3 only considered the
performance of Wells turbines using standard symmetric airfoils of type NACA 00XX.
As an illustration, Fig. 7.1 shows NACA 0015 and NACA 0021. Most investigations
pertaining to Wells turbines have considered NACA 0012, NACA 0015, NACA 0018
and NACA 0021 (e.g., [73, 81, 91, 93]). The formula for the shape of a NACA 00XX
foil, with “XX” being replaced by the percentage of maximum thickness to chord length
c, is
y =
ct
0.2
¸
0.2969
x
c
−0.126
x
c
−0.3517
x
c
2
+ 0.2843
x
c
3
−0.1015
x
c
4
(7.1)
where x is the position along the chord from 0 to c, y is the halfthickness at a given
value of x (centerline to external surface), and t is the maximum halfthickness as a
fraction of the chord (so that 100 t gives a half of the last two digits in the NACA
4digit denomination). Both monoplane and twoplane Wells turbines are considered
in the literature and show diﬀerent advantage and drawbacks. As a consequence, it is
interesting to optimize both designs separately, starting with the simple conﬁguration.
7.2 Optimal monoplane Wells turbine
Reference investigations indicated that NACA 0021 airfoil proﬁles (21% thickness) lead
to the best performance for conventional monoplane Wells turbines [93]. There is nev
130
CHAPTER 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 131
c
t
y
x
0 0.2 0.4 0.6 0.8 1
0.15
0.1
0.05
0
0.05
0.1
0.15
NACA 0015
NACA 0021
0.15c
0.1c
0.05c
0.05c
0.1c
0.15c
0
0 0.2c 0.4c 0.6c 0.8c c
Figure 7.1: Standard airfoils NACA 0015 and NACA 0021.
ertheless no proof that NACA proﬁles, as deﬁned by Eq. 7.1, automatically lead to the
best possible performance. An alternative geometry might be much better, in particular
for such very speciﬁc applications. As a consequence, the present section now concen
trates on the optimization of a symmetric airfoil shape, leading to the best possible
performance of a Wells turbine (i.e., maximal tangential force coeﬃcient and eﬃciency).
Due to the complexity of the underlying optimization procedure, this ﬁrst study consid
ers only monoplane Wells turbines (the original design) and a constant turbine solidity
(s = 0.67, as proposed by [110]), while taking into account the mutual interaction eﬀect
between the blades.
Rotorhub
onshaft
Forcesresolvedin
directionofrotation
Rotation
Sym. Airfoil
Blade
FA
FT
Chord(c)
s
p
a
n
(
b
)
r
t
rh
FA
Oscillating
flow
Figure 7.2: Axial and tangential forces acting on a Wells turbine.
In the present section, the free design variables considered for the optimization will
be the shape of the blade using a constant solidity (s = zc/[πr
t
(1 + h)]) where h is the
ratio between hub radius r
h
and tip radius r
t
(Fig. 7.2). The objective function contains
simultaneously two outputs of the simulation, that should both be maximized as far
as possible: the tangential force coeﬃcient C
T
; and the turbine eﬃciency η, which is
inversely proportional to the axial force coeﬃcient, and is deﬁned for a negligible density
change as:
η =
F
T
u
t
∆p
0
Q
(7.2)
CHAPTER 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 132
with Q the volumetric ﬂowrate through the turbine. For conﬁgurations involving con
current objectives, Evolutionary Algorithms are particulary robust and have therefore
been used in the present study. The employed optimization parameters are listed in
Table 5.1.
7.2.1 Optimization of airfoil shape
After having checked the accuracy of an individual evaluation relying on CFD as de
scribed in section 5.3.2.3, it is now possible to start the optimization procedure. As
explained previously, only symmetric blades are considered in what follows, based on
the proﬁle NACA 0021 for a ﬁrst guess.
To illustrate the optimization, a ﬁxed angle of incidence α = 8
◦
(ﬂow coeﬃcient
φ = 0.14) is considered. Twentytwo free parameters are varied simultaneously by the
OPAL optimizer, explaining the diﬃculty of the process. In the present case the outer
boundary of the airfoil (or airfoil shape) is constructed with thirteen points; two ﬁxed
points (P
1
and P
13
) and eleven variable points (P
2
, P
3
, P
4
, P
5
, P
6
, P
7
, P
8
, P
9
, P
10
,
P
11
and P
12
) as shown in Fig. 7.3. Knowing the exact position of these 13 points,
the full proﬁle is ﬁnally reconstructed for one face of the airfoil using standard splines
(Nonuniform rational Bsplines, NURBS). The order of a NURBS curve deﬁnes the
number of nearby control points that inﬂuence any given point on the curve. The curve
is represented mathematically by a polynomial of degree one less than the order of the
curve; this means that the spline order is 13 in our case. Then, the obtained face is
mirrored to obtain the full symmetric airfoil. Every point P
2
to P
12
has two coordinates
(X
pi
, Y
pi
), where i = 2 . . . 12. The parameter space considered in the optimization has
been deﬁned as documented in Table 7.1 and illustrated in Fig. 7.3. The corresponding
parameter spaces have been selected to cover all usual NACA 00XX, while avoiding
collisions between reference points and keeping acceptable geometries. The reference
point is point P
1
(0,0), origin of the cartesian coordinate system.
Figure 7.3: Allowed parameter space for the moving points P
2
to P
12
.
CHAPTER 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 133
Table 7.1: Parameter space for the moving points P
2
to P
12
for monoplane Wells turbine.
Point Parameter Minimum Maximum Point Parameter Minimum Maximum
P
2
X
P2
/c 0.01 0.015 P
8
X
P8
/c 0.45 0.55
Y
P2
/c 0.023 0.043 Y
P8
/c 0.043 0.143
P
3
X
P3
/c 0.015 0.035 P
9
X
P9
/c 0.55 0.65
Y
P3
/c 0.025 0.066 Y
P9
/c 0.039 0.11
P
4
X
P4
/c 0.05 0.1 P
10
X
P10
/c 0.65 0.75
Y
P4
/c 0.034 0.11 Y
P10
/c 0.034 0.094
P
5
X
P5
/c 0.1 0.2 P
11
X
P11
/c 0.75 0.85
Y
P5
/c 0.043 0.143 Y
P11
/c 0.026 0.066
P
6
X
P6
/c 0.2 0.3 P
12
X
P12
/c 0.85 0.95
Y
P6
/c 0.054 0.154 Y
P12
/c 0.015 0.035
P
7
X
P7
/c 0.35 0.45
Y
P7
/c 0.052 0.152
As a whole, the optimization process thus involves twenty two parameters (or degrees
of freedom) X
pi
and Y
pi
with i = 2 . . . 12 and two objectives (eﬃciency and tangential
force coeﬃcient) that should be simultaneously maximized in a concurrent manner.
The results presented in Figs. 7.4(a) and 7.4(b) indicate that the two considered
objectives are indeed considerably inﬂuenced by the airfoil shape. Figure 7.4(a) shows
all evaluation results. As a whole, 615 diﬀerent conﬁgurations have been ﬁnally tested by
the optimizer, leading to 15 days of total computing time on a standard PC. In Fig. 7.4(a)
the performance of the standard airfoil NACA 0021 (tangential force coeﬃcient and
eﬃciency (C
T
, η) = (0.1163; 0.5109)) is also plotted for comparison. Globally, the two
considered objectives are not fully concurrent but increase simultaneously, which is not a
complete surprise since the tangential force appears on the numerator in Eq. 7.2 deﬁning
the eﬃciency.
When considering now only the best conﬁgurations of Fig. 7.4(a), located in the
upperright corner (marked by a red square in dashed line), a more complex picture ap
pears, as documented in Fig. 7.4(b). For the last percent of performance improvement,
the two objectives (tangential force coeﬃcient and eﬃciency) become indeed slightly
concurrent and cannot be optimized simultaneously. Two optimal conditions are ﬁ
nally found: Point A, (C
T
A
, η
A
) =(0.1325;0.5187) (highest tangential force); and Point
B,(C
T
B
, η
B
) =(0.1281;0.5197) (highest eﬃciency). By analyzing in detail the resulting
geometries and considering daily engineering purposes, the increase in tangential force
coeﬃcient (higher power output) appears to be more signiﬁcant and valuable than the
very slightly increased eﬃciency. Therefore, the most interesting point is globally Point
A with (C
T
A
, η
A
) ≈ (0.1325; 0.519).
The results of the optimization process can be usefully visualized in a diﬀerent man
ner using parallel coordinates (Fig. 7.5). Here again, the performance of the standard
airfoil NACA 0021 is also plotted for comparison, close to the middle of the parameter
space (thick dashed blue line). Figures 7.5(a) and 7.5(b) indicate by parallel coordinates
CHAPTER 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 134
0 0.04 0.08 0.12 0.16
Tangential force coefficient (C
T
)
0.1
0.2
0.3
0.4
0.5
0.6
E
f
f
i
c
i
e
n
c
y
Optimization output
Monoplane Wells turbine
NACA 0021
0.1275 0.129 0.1305 0.132 0.1335
Tangential force coefficient (C
T
)
0.495
0.5
0.505
0.51
0.515
0.52
0.525
E
f
f
i
c
i
e
n
c
y
Optimization output
Optimum tang. force coeff.
Optimum efficiency
(A)
(B)
(a)
(b)
Figure 7.4: Objectives of the optimization; a) for all computed conﬁgurations b) for the
best conﬁgurations (i.e., zoom on the upperright part (red square) of a).
the X and Y coordinates of the eleven moving points (P
2
. . . P
12
), together with the two
objectives. This ﬁgure demonstrates that very diﬀerent shapes have been evaluated on
the way toward the optimal solution. The optimum conﬁguration (Point A) is indicated
by a thick red line.
It can be seen that the optimal airfoil shape leads only to a slightly higher eﬃciency
(+0.78%) compared to the standard airfoil (NACA 0021). However, the tangential force
coeﬃcient C
T
is at the same time increased by 0.0162. This means a relative increase
of 12.2% for the present ﬂow coeﬃcient, equal to 0.14.
The geometrical parameters corresponding to the optimal shape are listed in Ta
ble 7.2. The resulting shape of the optimal airfoil in comparison with the standard
CHAPTER 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 135
Figure 7.5: Input parameters of the optimization and objectives represented using parallel
coordinates. The parameters of the optimal shape are connected with a thick red line. The
standard design (NACA 0021) is shown with a thick dashed blue line; a) Xcoordinates
of the variable points (P
2
. . . P
12
); b) Ycoordinates of the variable points (P
2
. . . P
12
).
NACA 0021 is shown in Fig. 7.6. Knowing all points P
1
to P
13
, the full proﬁle is again
reconstructed using standard splines of order 13. Nevertheless, a simple polynomial de
scription of this proﬁle would be helpful for practical purposes. An excellent ﬁt (average
CHAPTER 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 136
Table 7.2: Optimum shape parameters for monoplane Wells turbine
Point Parameter Value Point Parameter Value
P
2
X
P2
/c 0.0124947 P
8
X
P8
/c 0.469922
Y
P2
/c 0.033155 Y
P8
/c 0.0666932
P
3
X
P3
/c 0.0269924 P
9
X
P9
/c 0.56625
Y
P3
/c 0.045006 Y
P9
/c 0.0493063
P
4
X
P4
/c 0.0678583 P
10
X
P10
/c 0.723109
Y
P4
/c 0.0608499 Y
P10
/c 0.037401
P
5
X
P5
/c 0.155326 P
11
X
P11
/c 0.798623
Y
P5
/c 0.078402 Y
P11
/c 0.0315091
P
6
X
P6
/c 0.248998 P
12
X
P12
/c 0.8891079
Y
P6
/c 0.0877814 Y
P12
/c 0.0214618
P
7
X
P7
/c 0.405838
Y
P7
/c 0.0751153
residual error of 0.38%) has been obtained with following polynomial description:
Y
c
= A
X
c
5
+ B
X
c
4
+ D
X
c
3
+ E
X
c
2
+H
X
c
+ K (7.3)
with the constants A to H listed in Table 7.3. Furthermore, this polynomial removes
any possible oscillations of the proﬁle shape induced by the spline description.
Table 7.3: Polynomial coeﬃcients of optimal airfoil shape (best ﬁt)
A B D E H K
1.6958588 −5.4277515 6.560073 −3.737973 0.898334 0.01409362
0.12c
0.08c
0.04c
0
0 0.2c
0.4c 0.6c
0.8c c
Optimum shape
Polynomial fit of optimum shape
NACA 0021
Figure 7.6: Comparison between the original proﬁle NACA 0021 (solid line), the optimal
airfoil shape described by splines (black squares showing the position of the control points)
and the corresponding polynomial ﬁt (Eq. 7.3, dashed line).
CHAPTER 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 137
7.2.2 Oﬀ design performance
It is important to check how the gain induced by the new airfoil shape would change
as a function of the ﬂow coeﬃcient φ, since such a turbine must be able to work also
for oﬀdesign conditions. Therefore, the performance of the optimal shape has been
ﬁnally computed for the full range of useful φvalues, as shown in Fig. 7.7. These results
demonstrate that the improvement of tangential force coeﬃcient is observed throughout
for all values of φ, compared to the conventional turbine based on standard airfoils
NACA 0021. The absolute gain for C
T
increases even slightly with φ. The relative
increase is higher than 8.8% throughout the useful operating range, with an average
gain of 11.3% (Fig. 7.7a). At the same time the eﬃciency of the optimized shape
is always higher than for the conventional design, the diﬀerence being lower for large
ﬂow coeﬃcients. The corresponding gain varies between 0.2% and up to 3.2%, with
an average increase of 1% (Fig. 7.7b). No signiﬁcant diﬀerence is observed in Fig. 7.7
between the performance of the exact proﬁle described by splines and the associated
polynomial ﬁt (Eq. 7.3).
7.3 Optimal twostage Wells turbine with non
symmetric airfoils
In this section we investigate extensively the potential of nonsymmetric airfoil blades
to improve the tangential force and eﬃciency associated to a twostage Wells turbine.
Since these two stages are mirrored, the system stays globally symmetric, as requested.
Nonsymmetric blades could perhaps allow to increase considerably the power output
from the turbine and the global eﬃciency of the system.
All published results concerning the ﬂow ﬁeld around a Wells turbine rotor indicate
that a considerable amount of exit kinetic energy is lost with the swirl component
of the ﬂow velocity, at least in the absence of guide vanes (see Section 3.2.3). This
kinetic energy can be partly recovered by using a second stage of blades. Twostage
Wells turbine have been already investigated experimentally and theoretically [21, 61,
66, 80, 107], but considering only symmetric airfoils. The present work concentrates
on a modiﬁed, twostage Wells turbine constructed from nonsymmetric airfoils based
initially on NACA 2421. The two stages are mirrored to keep the global symmetry of
the turbine as shown in Fig. 7.8.
The performance of the twostage rotor is modiﬁed by mutual aerodynamic interfer
ences due to the proximity of the two planes. The upstream rotor aﬀects the performance
of the downstream rotor by producing a deﬂection of the air stream. Therefore, the gap
between the two rotors is an important parameter to control performance. Using nu
merical optimization, the solidity and the shape of the nonsymmetric airfoils will be
optimized in this section.
CHAPTER 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 138
0.05 0.1 0.15 0.2 0.25 0.3
Flow coefficient
0
0.1
0.2
0.3
0.4
0.5
T
a
n
g
e
n
t
i
a
l
f
o
r
c
e
c
o
e
f
f
i
c
i
e
n
t
(
C
T
)
0
10
20
30
40
50
%
R
e
l
a
t
i
v
e
i
n
c
r
e
a
s
e
NACA0021
Spline optimumshape
Fitting optimumshape
%Relative increase
%
0.05 0.1 0.15 0.2 0.25 0.3
Flow coefficient
0.4
0.42
0.44
0.46
0.48
0.5
0.52
0.54
E
f
f
i
c
i
e
n
c
y
NACA 0021
Spline optimum shape
Fitting optimum shape
(a)
(b)
Figure 7.7: Performance of the spline optimal conﬁguration (red line), ﬁtting optimal
one (black cross) compared to the conventional Wells turbine relying on the NACA 0021
proﬁle (green line). The corresponding relative increase is shown with blue line; a)
tangential force coeﬃcient; b) eﬃciency.
7.3.1 Optimal airfoil shape without mutual interactions
In order to assess the potential of nonsymmetric airfoils, a ﬁrst optimization is carried
out without mutual interactions between the blades. To illustrate the process, a ﬁxed
angle of incidence α = 12
◦
(ﬂow coeﬃcient φ = 0.21) is considered. Two degrees of
freedom are left to the OPAL optimizer: the thickness factors y
1
for the upper side of
the airfoil and y
2
for the lower side (see Fig. 7.8). These two factors are constrained
by the user between 0.2 and 1.6, so that considerable variations are still allowed (both
CHAPTER 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 139
Figure 7.8: Twostage Wells turbine based on nonsymmetric airfoils
for increasing and decreasing thickness). All blades of both rotors are always modiﬁed
simultaneously and the two rotors are still mirrored to preserve the global symmetry of
the turbine. As a whole, the optimization process thus involves only two parameters
y
1
and y
2
and two objectives (eﬃciency and tangential force coeﬃcient) that should be
simultaneously maximized in a concurrent manner.
The optimal point found in this ﬁrst study, corresponds to the scaling factors y
1
=
1.22966 (23% thickening) for the upper face and y
2
= 1.1795 (18% thickening) for the
lower face. The resulting shape of the optimal airfoil in comparison with the standard
one is shown in Fig. 7.9.
Figure 7.9: Comparison between the optimal shape of the airfoil and the original proﬁle
NACA 2421
When compared with the baseline case (NACA 2421) the optimal point corresponds
simultaneously to a relative improvement of the eﬃciency by 2.1% and of the tangential
force coeﬃcient by almost 6%, demonstrating the interest of nonsymmetric airfoils.
However, mutual interactions could modify these results and must now be taken into
CHAPTER 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 140
account.
7.3.2 Optimal airfoil shape with mutual interactions
7.3.2.1 Mutual interactions eﬀect
The present section describes the optimization of the airfoil shape considering mutual
interactions between the blades. A Wells turbine consisting of two mirrored airfoil stages
is considered again. Mirroring is needed to keep the global symmetry of the system,
as requested when using power from an Oscillating Water Column. But, in order to
increase system eﬃciency, nonsymmetric blades are considered. The inﬂuence of the
mutual interactions between the blades building one plane (perpendicular to the main
ﬂow direction) are considered extensively in what follows.
The results of the previous section indicated that nonsymmetric airfoils are better
than symmetric ones. However, these results have been obtained neglecting any inter
actions between the blades. The validity of these ﬁrst results is therefore questionable,
since the performance of the twostage Wells turbine is aﬀected by mutual aerodynamic
interaction between the blades in a single plane. This aerodynamic interaction results
from the wakes produced by the preceeding blades and is a function of the angle of
incidence of the air ﬂow as well as of the solidity of the blades.
In the past, the inﬂuence of the mutual interaction between blades has been studied
using windtunnel measurements or simple theoretical correlations relying on potential
ﬂow analysis and singularity theory for ﬂat plate aerofoils in cascade [86]. According
to such studies, the correlation factor m describing the ratio between the lift coeﬃcient
considering the interaction and the coeﬃcient of an isolated ﬂat plate is given by:
m =
2B
πc
tan
πc
2B
(7.4)
where c is the blade chord length and B is the pitch between the blades.
In this section, the interaction eﬀect of the modiﬁed Wells turbine is investigated
numerically in a systematic manner. The axial and tangential force coeﬃcients normal
ized by the corresponding coeﬃcients obtained for an isolated blade (single airfoil) have
been computed for a constant gap ratio G = 1 (Fig. 7.10). The results demonstrate that
both the normalized tangential and axial force coeﬃcients increase exponentially with
the solidity s due to the growing inﬂuence of wake eﬀects. Increasing the tangential
force is a positive aspect, but an increasing axial force is of course a major drawback.
The eﬃciency of the modiﬁed Wells turbine decreases very rapidly when increasing
the solidity due to high losses, as shown in Fig. 7.11b. As the same time, the tangential
force coeﬃcient increases with the solidity (Fig. 7.11a).
This decrease in eﬃciency can be explained by the fact that the losses near the
turbine hub are considerably higher than near the turbine tip, because the ﬂow passage
at the tip is much wider than the ﬂow passage near the hub for a standard rectangular
blade shape as shown in Fig. 7.12a. Therefore, we suggest to replace the rectangular
shape of the blades by a trapezoidal shape as shown in Fig. 7.12b. This trapezoidal
CHAPTER 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 141
0.4 0.5 0.6 0.7 0.8
Solidity
1
2
3
4
5
C
x
/
C
x
0
G=1
a=8 deg.
a=10 deg.
a=12 deg.
0.4 0.5 0.6 0.7 0.8
Solidity
1
1.2
1.4
1.6
1.8
2
2.2
2.4
C
T
/
C
T
0
G=1
a=8 deg.
a=10 deg.
a=12 deg.
(b)
(a)
Figure 7.10: Impact of mutual interaction between blades in the same plane on a) tan
gential force (left) and b) axial force (right), as a function of the solidity.
0.4 0.5 0.6 0.7 0.8
Solidity
0
0.2
0.4
0.6
0.8
C
T
G=1.0
a=8 deg.
a=10 deg.
a=12 deg.
0.4 0.5 0.6 0.7 0.8
Solidity
0.3
0.4
0.5
0.6
0.7
E
f
f
i
c
i
e
n
c
y
G=1.0
a=8 deg.
a=10 deg.
a=12 deg.
(a)
(b)
Figure 7.11: Impact of solidity on the twostage Wells turbine performance considering
the tangential force coeﬃcient (a: left ﬁgure) and the turbine eﬃciency (b: right ﬁgure).
shape leads to the same passage at both hub and tip, and consequently such blades
should give globally a better eﬃciency along the blade span (isoperformance blade).
The best conditions should be obtained when the ratio between hub radius and tip
radius equals the ratio between hub chord and tip chord:
R
h
R
t
=
c
h
c
t
(7.5)
Checking the performance of this modiﬁed design would unfortunately require three
dimensional CFD which are beyond reach when coupled with an optimizer. Therefore,
this issue is left for future studies.
CHAPTER 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 142
Tipchord(c ) t
Tipradius(R ) t
Hubradius (R ) h
Hubchord(c ) h
IsoPerformance
blade
Re e ctangularblad
Tipradius(R ) t
Hubradius (R ) h
(a)
(b)
Figure 7.12: Projected shape of the turbine; a) Conventional turbine b) Suggestion of
isoperformance turbine.
7.3.2.2 Optimization
Starting from NACA 2421, the present section concentrates on the optimization of a
nonsymmetric airfoil shape, leading to the best possible performance (as usual, maxi
mal tangential force coeﬃcient and eﬃciency) of a twostage Wells turbine. A constant
turbine solidity (s = 0.67, as proposed by [110]) and a constant gap between the rotors
G = d/c = 1.0 are again considered, while taking into account the mutual interac
tion eﬀect between the blades (see previous section). In the present case the outer
boundary of the airfoil is constructed with thirty four points; two ﬁxed points (P
0
and P
17
) and thirty two variable points (sixteen variable points for every face of the
airfoil, P
1
, P
2
, P
3
, P
4
, P
5
, P
6
, P
7
, P
8
, P
9
, P
10
, P
11
, P
12
, P
13
, P
14
, P
15
and P
16
for upper face,
P
18
, P
19
, P
20
, P
21
, P
22
, P
23
, P
24
, P
25
, P
26
, P
27
, P
28
, P
29
, P
30
, P
31
, P
32
and P
33
for lower face)
as shown in Fig. 7.13. Knowing the exact position of these 34 points, the full proﬁle is
reconstructed for each face of the airfoil using standard splines (NURBS). The param
eter space considered in the optimization has been deﬁned as documented in Table 7.4
and illustrated in Fig. 7.13. The corresponding parameter spaces have been selected
to cover all usual NACA airfoils, while avoiding collisions between reference points and
keeping acceptable geometries. The reference point is point P
0
(0,0), origin of the carte
sian coordinate system.
CHAPTER 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 143
Table 7.4: Parameter space for the moving points P
1
to P
16
for upper face and P
18
to
P
33
for lower face
Point Parameter Minimum Maximum Point Parameter Minimum Maximum
P
1
X
P1
/c 0.006 0.018 P
18
X
P18
/c 0.006 0.018
Y
P1
/c 0.03 0.06 Y
P18
/c 0.03 0.016
P
2
X
P2
/c 0.018 0.032 P
19
X
P19
/c 0.018 0.032
Y
P2
/c 0.03 0.07 Y
P19
/c 0.04 0.025
P
3
X
P3
/c 0.032 0.06 P
20
X
P20
/c 0.032 0.06
Y
P3
/c 0.05 0.07 Y
P20
/c 0.055 0.035
P
4
X
P4
/c 0.06 0.08 P
21
X
P21
/c 0.06 0.08
Y
P4
/c 0.07 0.13 Y
P21
/c 0.06 0.045
P
5
X
P5
/c 0.08 0.13 P
22
X
P22
/c 0.08 0.13
Y
P5
/c 0.08 0.14 Y
P22
/c 0.076 0.055
P
6
X
P6
/c 0.13 0.17 P
23
X
P23
/c 0.13 0.17
Y
P6
/c 0.1 0.15 Y
P23
/c 0.08 0.06
P
7
X
P7
/c 0.17 0.23 P
24
X
P24
/c 0.17 0.23
Y
P7
/c 0.11 0.16 Y
P24
/c 0.08 0.06
P
8
X
P8
/c 0.23 0.27 P
25
X
P25
/c 0.23 0.27
Y
P8
/c 0.11 0.17 Y
P25
/c 0.08 0.06
P
9
X
P9
/c 0.27 0.35 P
26
X
P26
/c 0.27 0.35
Y
P9
/c 0.12 0.18 Y
P26
/c 0.075 0.06
P
10
X
P10
/c 0.35 0.45 P
27
X
P27
/c 0.35 0.45
Y
P10
/c 0.12 0.17 Y
P27
/c 0.07 0.055
P
11
X
P11
/c 0.45 0.55 P
28
X
P28
/c 0.45 0.55
Y
P11
/c 0.11 0.16 Y
P28
/c 0.06 0.045
P
12
X
P12
/c 0.55 0.65 P
29
X
P29
/c 0.55 0.65
Y
P12
/c 0.1 0.15 Y
P29
/c 0.055 0.035
P
13
X
P13
/c 0.65 0.75 P
30
X
P30
/c 0.65 0.75
Y
P13
/c 0.08 0.13 Y
P30
/c 0.04 0.025
P
14
X
P14
/c 0.75 0.85 P
31
X
P31
/c 0.75 0.85
Y
P14
/c 0.06 0.09 Y
P31
/c 0.03 0.016
P
15
X
P15
/c 0.85 0.93 P32 X
P32
/c 0.85 0.93
Y
P15
/c 0.03 0.06 Y
P32
/c 0.014 0.01
P
16
X
P16
/c 0.93 0.98 P
33
X
P33
/c 0.93 0.98
Y
P16
/c 0.01 0.04 Y
P33
/c 0.009 0.006
The results are presented in Fig. 7.14. As a whole, 615 diﬀerent conﬁgurations have
been tested by the optimizer, leading to 18 days of total computing time on a stan
dard PC. In Fig. 7.14a the performance of the standard airfoil NACA 2421 (tangential
force coeﬃcient and eﬃciency (C
T
, η )=(0.2074;0.4453)) is also plotted for comparison.
Globally, the two considered objectives are not fully concurrent but increase simulta
neously. When considering now only the best conﬁgurations of Fig. 7.14, located in
the upperright corner (marked by a red square), a more complex picture appears, as
CHAPTER 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 144
P18
P19
P20
P21
P22
P23
P24P25
P26
P27
P28
P29
P30
P31
P32
P33
P17
P0
C C C C C
C
C
C
C
C
Lowerface
P0
P17
P1
P16
P15
P14
P13
P12
P11
P10
P17
P9 P8 P7
P6
P5
P4
P3
P2
C C C C
C
C
C
C
C
UpperFace
0 0.2 0.4 0.6 0.8 1
0
0.04
0.08
0.12
0.16
c
0 0.2 0.4 0.6 0.8 1
0.1
0.08
0.06
0.04
0.02
0
c
Figure 7.13: Allowed parameter space for the moving points.
in previous cases, with slightly concurrent objectives. Two optimal conditions are ﬁ
nally found: Point A, (C
T
, η)= (0.2247;0.4482) (highest tangential force); and Point B,
(C
T
, η)= (0.2212;0.4506) (highest eﬃciency). As usual, the increase in tangential force
coeﬃcient (higher power output) appears to be more signiﬁcant and valuable than the
very slightly increased eﬃciency.
Therefore, the most interesting point is globally Point A with (C
T
, η)(0.225;0.448).
The results of the optimization process can be also visualized using parallel coordinates
(Fig. 7.15). Figure 7.15a indicates by parallel coordinates the X and Y coordinates of
the eleven moving points of upper face (P
1
. . . P
16
), together with the two objectives,
while Fig. 7.15b shows by parallel coordinates the X and Y coordinates of the eleven
moving points of lower face (P
18
. . . P
33
), with the same two objectives.
This ﬁgure demonstrates that very diﬀerent shapes have been evaluated on the way
toward the optimal solution. The optimum conﬁguration (Point A) is indicated by a
thick red line. It can be seen that the optimal airfoil shape leads only to a slightly
higher eﬃciency (+0.5%) compared to the standard airfoil (NACA 2421). However,
the tangential force coeﬃcient C
T
is at the same time increased by 0.017, i.e., a relative
increase of more than 7.7% for the present ﬂow coeﬃcient, equal to 0.14. The geometrical
parameters corresponding to the optimal shape are listed in Table 7.5.
The resulting shape of the optimal airfoil in comparison with the standard NACA
2421 is shown in Fig. 7.16. Knowing all points P
0
to P
33
, the full proﬁle is again
CHAPTER 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 145
0 0.05 0.1 0.15 0.2 0.25
Tangential force coefficient
0
0.1
0.2
0.3
0.4
0.5
E
f
f
i
c
i
e
n
c
y
Optimization output
NACA 2421
0.21 0.213 0.216 0.219 0.222 0.225
Tangential force coefficient
0.43
0.435
0.44
0.445
0.45
0.455
E
f
f
i
c
i
e
n
c
y
Optimization output
Max. tangential force coeff.
Max. Efficiency
(a)
(b)
Point(B)
Point(A)
Figure 7.14: Objectives of the optimization: a) for all computed conﬁgurations; b) for
the best conﬁgurations, i.e., zoom on the upperright part (red square) of (a).
reconstructed using standard splines. Nevertheless, a simple polynomial description of
these proﬁles is helpful for practical purposes. An excellent ﬁt (average residual error
of 0.3%) has been obtained using Eq. 7.3
with the constants A to H listed in Table 7.6 for the upper face and the lower face.
7.3.2.3 Oﬀ design performance
The results shown in Fig. 7.17 demonstrate that the improvement of tangential force
coeﬃcient is observed throughout for all values of φ, compared to the nonsymmetric
turbine based on standard airfoils NACA 2421, with an average increase of 5.5%. At
the same time the eﬃciency of the optimized shape is slightly higher than the standard
airfoil design. The corresponding gain varies between 0.2% and 0.7%, with an average
CHAPTER 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 146
Optimalshapeofupperface
Optimalshapeoflowerface
CT Eff.
0.451
0.2251
CT Eff.
0.451 0.2251
0.083
0.02
X /c 1
Y /c 1
.......
X /c 16
Y /c 16
X /c 17
Y /c 17
.......
X /c 32
Y /c 32
0.207
0.445
0.207
0.445
NACA 2421
Figure 7.15: Input parameters of the optimization and objectives represented using par
allel coordinates. The parameters of the optimal shape are connected with a thick red
line.
0 0.2 0.4 0.6 0.8 1
0.1
0.05
0
0.05
0.1
0.15 c
c
c
c
c
c c c c c
Optimalconfiguration
NACA 2421
Figure 7.16: Comparison between the original proﬁle NACA 2421 (solid line) and the
optimal airfoil shape (dashed line), considering mutual interaction between the blades.
CHAPTER 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 147
Table 7.5: Optimum shape parameters for twostage Wells turbine with nonsymmetric
airfoils
Point Parameter Value Point Parameter Value Point Parameter Value
P
1
X
P1
/c 0.01129 P
12
X
P12
/c 0.5748 P
24
X
P24
/c 0.18961
Y
P1
/c 0.0405 Y
P12
/c 0.125 Y
P24
/c 0.0653
P
2
X
P2
/c 0.02179 P
13
X
P13
/c 0.68696 P
25
X
P25
/c 0.2485
Y
P2
/c 0.0527 Y
P13
/c 0.09814 Y
P25
/c 0.06995
P
3
X
P3
/c 0.0377 P
14
X
P14
/c 0.7556 P
26
X
P26
/c 0.3229
Y
P3
/c 0.06611 Y
P14
/c 0.07551 Y
P26
/c 0.06117
P
4
X
P4
/c 0.06938 P
15
X
P15
/c 0.8759 P
27
X
P27
/c 0.3846
Y
P4
/c 0.0928 Y
P15
/c 0.0386 Y
P27
/c 0.0602
P
5
X
P5
/c 0.1169 P
16
X
P16
/c 0.9365 P
28
X
P28
/c 0.5078
Y
P5
/c 0.1229 Y
P16
/c 0.0149 Y
P28
/c 0.0584
P
6
X
P6
/c 0.1372 P
18
X
P18
/c 0.0096 P
29
X
P29
/c 0.559
Y
P6
/c 0.1302 Y
P18
/c 0.02274 Y
P29
/c 0.0464
P
7
X
P7
/c 0.1795 P
19
X
P19
/c 0.0196 P
30
X
P30
/c 0.6882
Y
P7
/c 0.1345 Y
P19
/c 0.03159 Y
P30
/c 0.0327
P
8
X
P8
/c 0.247 P
20
X
P20
/c 0.0577 P
31
X
P31
/c 0.8019
Y
P8
/c 0.1351 Y
P20
/c 0.0453 Y
P31
/c 0.0219
P
9
X
P9
/c 0.2783 P
21
X
P21
/c 0.0689 P
32
X
P32
/c 0.862
Y
P9
/c 0.1417 Y
P21
/c 0.0514 Y
P32
/c 0.0121
P
10
X
P10
/c 0.3936 P
22
X
P22
/c 0.1069 P
33
X
P33
/c 0.9508
Y
P10
/c 0.1353 Y
P22
/c 0.0633 Y
P33
/c 0.0073
P
11
X
P11
/c 0.5205 P
23
X
P23
/c 0.1478
Y
P11
/c 0.1334 Y
P23
/c 0.0709
Table 7.6: Polynomial coeﬃcients of optimal airfoil shape (best ﬁt) for twostage non
symmetric airfoil Wells turbine considering mutual interaction between the blades (up
per and lower face)
Face A B D E H
Upper face −1.31571 3.19739 −2.9773 1.06296 0.02429
Lower face 0.88881 −2.18127 1.8927 −0.58341 −0.014714
increase of 0.3% (Fig. 7.17b).
This optimization study considers the largest number of free optimization parameters
(64) ever considered in our group. By analyzing the results we observed that the resulting
Pareto frontier was relatively irregular and poorly populated.
The presented results should therefore be considered with caution. It is probably
better to keep a lower number of free parameters in association with concurrent opti
mization problems. For this reason, the next section considers only symmetric proﬁles,
that might be described with much fewer control points while keeping a larger parameter
space.
CHAPTER 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 148
0.08 0.12 0.16 0.2 0.24 0.28
Flow coefficient
0.4
0.41
0.42
0.43
0.44
0.45
0.46
E
f
f
i
c
i
e
n
c
y
Optimum airfoil
NACA 2421
0.08 0.12 0.16 0.2 0.24 0.28
Flow coefficient
0
0.2
0.4
0.6
0.8
T
a
n
g
e
n
t
i
a
l
f
o
r
c
e
c
o
e
f
f
i
c
i
e
n
t
0
10
20
30
40
50
%
R
e
l
a
t
i
v
e
i
n
c
r
e
a
s
e
Optimum airfoil
NACA 2421
% Relative increase
(a)
(b)
Figure 7.17: Performance of the optimal conﬁguration (red line), compared to the non
symmetric twostage Wells turbine relying on the NACA 2421 proﬁle (black line). The
corresponding relative increase is shown with blue line; a) tangential force coeﬃcient; b)
eﬃciency.
7.4 Optimal twostage Wells turbine with symmet
ric airfoils
The present section ﬁnally concentrates on the optimization of a symmetric airfoil shape,
leading to the best possible performance (i.e., maximal tangential force coeﬃcient and
eﬃciency) for a twostage Wells turbine. A constant turbine solidity (s = 0.67)) and a
constant gap between the rotors G = d/c = 1.0 (Fig. 7.18) are again considered, while
taking into account the mutual interaction eﬀect between the blades.
CHAPTER 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 149
Flow
Flow
Rotation d
Figure 7.18: Twostage Wells turbine with symmetric airfoils NACA 0021.
7.4.1 Optimization of airfoil shape
The blade shape is now constructed with thirteen points, two ﬁxed points (P
1
and P
13
)
and eleven variable points (P
2
, P
3
, P
4
, P
5
, P
6
, P
7
, P
8
, P
9
, P
10
, P
11
and P
12
) as shown in
Fig. 7.19. Knowing the exact position of these 13 points, the full proﬁle is ﬁnally re
constructed for one face of the airfoil using standard splines. Figure 7.20a shows all
Figure 7.19: Allowed parameter space for the moving points.
evaluation results. As a whole, 615 diﬀerent conﬁgurations have been ﬁnally tested
by the optimizer, leading to 17 days of total computing time on a standard PC. In
Fig. 7.20a the performance of the standard airfoil NACA 0021 (tangential force co
eﬃcient and eﬃciency (C
T
, η)=(0.2104;0.4351)) is also plotted for comparison. Glob
ally, the two considered objectives are not fully concurrent but increase simultaneously.
When considering now only the best conﬁgurations of Fig. 7.20, located in the upper
right corner (marked by a red circle), a more complex picture appears, as documented in
CHAPTER 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 150
Figure 7.20b. For the last percent of performance improvement, the two objectives (tan
gential force coeﬃcient and eﬃciency) become again slightly concurrent and cannot be
optimized simultaneously. Two optimal conditions are ﬁnally found: Point A, (C
T
, η)=
(0.2489;0.4430) (highest tangential force); and Point B, (C
T
, η)= (0.2384;0.4450) (high
est eﬃciency). Here again, the increase in tangential force coeﬃcient (higher power
output) appears to be more signiﬁcant and valuable than the very slightly increased
eﬃciency.
0 0.05 0.1 0.15 0.2 0.25
Tangential force coefficient (C
T
)
0
0.1
0.2
0.3
0.4
0.5
E
f
f
i
c
i
e
n
c
y
Optimization output
NACA 0021
0.235 0.24 0.245 0.25
Tangential force coefficient (C
T
)
0.434
0.436
0.438
0.44
0.442
0.444
0.446
E
f
f
i
c
i
e
n
c
y
Optimization output
Max. tang. force coeff.
Max. efficiency
Point(A)
Point(B)
Figure 7.20: Objectives of the optimization: a) for all computed conﬁgurations; b) for
the best conﬁgurations, i.e., zoom on the upperright part (red circle) of (a).
Therefore, the most interesting point is Point A with (C
T
, η)≈(0.249;0.443). The
results of the optimization process can be also visualized using parallel coordinates
(Fig. 7.21). Here again, the performance of the standard airfoil NACA 0021 is also
plotted for comparison (thick dashed blue line).
The optimum conﬁguration (Point A) is indicated by a thick red line. It can be seen
that the optimal airfoil shape leads only to a slightly higher eﬃciency (+0.8%) compared
to the standard airfoil (NACA 0021). However, the tangential force coeﬃcient C
T
is at
the same time increased by 0.0385, i.e., a relative increase of more than 15% for the
present ﬂow coeﬃcient, equal to 0.14. The geometrical parameters corresponding to the
optimal shape are listed in Table 7.7.
The resulting shape of the optimal airfoil in comparison with the standard NACA
0021 is shown in Fig. 7.22. Knowing all points P1 to P13, the full proﬁle is again
reconstructed using standard splines of order 13. A polynomial description of this proﬁle
would be helpful for practical purposes. An excellent ﬁt (average residual error of 0.16%)
has been obtained using again Eq. 7.3 with the constants A to H listed in Table 7.8.
7.4.1.1 Oﬀ design performance
The results shown in Fig. 7.23 demonstrate that the improvement of tangential force
coeﬃcient is observed throughout for all values of φ, compared to the conventional
CHAPTER 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 151
X /c 2 X /c 3
X /c 12
X /c 11 X /c 10 X /c 9 X /c 8 X /c 7 X /c 6 X /c 4 X /c 5 Cp Effic.
0.445 0.249 0.02 0.05 0.1 0.2 0.35 0.45 0.55 0.65 0.75 0.85 0.97
0.006 0.02
0.05 0.1 0.2 0.35
0.55 0.65 0.75 0.85 0.017 0.45 0.05
Y /c 2 Y /c 3
Y /c 12 Y /c 11 Y /c 10 Y /c 9 Y /c 8 Y /c 7 Y /c 6 Y /c 4 Y /c 5
0.053 0.076 0.12 0.153 0.164
0.16
0.13 0.12 0.076 0.036
0.16
0.013 0.015 0.024 0.034 0.041 0.04 0.029 0.02 0.02 0.01 0.04
Optimumconfiguration
NACA 0021
Figure 7.21: Input parameters of the optimization and objectives represented using par
allel coordinates. The parameters of the optimal shape are connected with a thick red
line. The standard design (NACA 0021) is shown with a thick dashed blue line.
Optimum shape
NACA 0021
Polynomial fit of the optimum shape
0
0.2c 0.4c 0.6c 0.8c
c
0.04c
0.08c
0.12c
0
Figure 7.22: Comparison between the original proﬁle NACA 0021 (solid line), the poly
nomial optimal airfoil shape (red line) and optimal shape by splines (black square).
turbine based on standard airfoils NACA 0021. The absolute gain for C
T
increases even
slightly with φ. The relative increase is higher than 10% throughout the useful operating
CHAPTER 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 152
Table 7.7: Optimum shape parameters for twostage Wells turbine
Point Parameter Value Point Parameter Value
P
2
X
P2
/c 0.01188 P
8
X
P8
/c 0.53128
Y
P2
/c 0.03242 Y
P8
/c 0.090718
P
3
X
P3
/c 0.03056 P
9
X
P9
/c 0.58674
Y
P3
/c 0.0458 Y
P9
/c 0.08169
P
4
X
P4
/c 0.0746 P
10
X
P10
/c 0.70306
Y
P4
/c 0.0636 Y
P10
/c 0.065626
P
5
X
P5
/c 0.11839 P
11
X
P11
/c 0.78848
Y
P5
/c 0.07726 Y
P11
/c 0.047259
P
6
X
P6
/c 0.29647 P
12
X
P12
/c 0.89755
Y
P6
/c 0.111736 Y
P12
/c 0.027875
P
7
X
P7
/c 0.3825
Y
P7
/c 0.11015
Table 7.8: Polynomial coeﬃcients of optimal airfoil shape (best ﬁt) for twostage Wells
turbine with symmetric airfoils
A B D E H K
0.5233101 −2.1617185 3.357793 −2.588445 0.8563828 0.01314256
range, with an average gain of 12% (Fig. 7.23a). At the same time the eﬃciency of
the optimized shape is higher than the conventional design for φ < 0.18; the diﬀerence
disappears for large ﬂow coeﬃcients due to stall. The corresponding gain varies between
0.5% and 2%, with an average increase of 0.6% (Fig. 7.23b).
7.5 Conclusions on Wells turbine
As discussed previously, standard Wells turbines have several wellknown disadvantages:
low tangential force, high (undesired) axial force; low aerodynamic eﬃciency and lim
ited range of operation. In the present work we have shown the potential of CFDbased
optimization to improve the tangential force induced by monoplane and twostage Wells
turbines. Two concurrent objectives (eﬃciency and tangential force coeﬃcient) have
been maximized in a concurrent manner. The optimization relied on Genetic Algo
rithms, all geometrical conﬁgurations being evaluated in an automatic manner by CFD,
taking into account the inﬂuence of the mutual interaction between the blades.
• Monoplane Wells turbine
In this case, only symmetric airfoils can be considered. Due to the importance
of the airfoil shape, a mathematical optimization procedure has been carried out
considering simultaneously up to twentytwo free parameters.
This optimization procedure is able to identify a considerably better conﬁguration
CHAPTER 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 153
0.08 0.12 0.16 0.2 0.24 0.28
Flow coefficient
0
0.2
0.4
0.6
0.8
1
T
a
n
g
e
n
t
i
o
n
a
l
f
o
r
c
e
c
o
e
f
f
i
c
i
e
n
t
(
C
T
)
0
20
40
60
80
100
%
R
e
l
a
t
i
v
e
i
n
c
r
e
a
s
e
Optimum design
NACA 0021
Relative increase
0.08 0.12 0.16 0.2 0.24 0.28
Flow coefficient
0.38
0.4
0.42
0.44
0.46
E
f
f
i
c
i
e
n
c
y
Optimum design
NACA 0021
%
(a)
(b)
Figure 7.23: Performance of the optimal conﬁguration (red line), compared to the con
ventional twostage Wells turbine relying on the NACA 0021 proﬁle (gray squares). The
corresponding relative increase is shown with blue line; a) tangential force coeﬃcient; b)
eﬃciency.
than the standard design relying on NACA 0021. A relative increase of the tan
gential force coeﬃcient exceeding 8.8% (as a mean, 11.3%) is obtained for the full
operating range. At the same time, the eﬃciency improves also by at least 0.2%
and up to 3.2% (as a mean, 1%).
• Twostage Wells turbine with nonsymmetric airfoils
Then, the aerodynamic performance of a modiﬁed twostage Wells turbine consist
CHAPTER 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 154
0.3
0.35
0.4
0.45
0.5
0.55
0.6
0.65
0.7
0.75
0.8
0.85
P
e
a
k
t
a
n
g
e
n
t
i
a
l
f
o
r
c
e
c
o
e
f
f
i
c
i
e
n
t
Optimizationdirection
W
e
l
l
s
t
u
r
b
i
n
e
(
N
A
C
A
0
0
2
1
)
O
p
t
i
m
i
z
e
d
W
e
l
l
s
t
u
r
b
i
n
e
a)MonoplaneWellsturbine
(lowpotentialwave)
0.4
0.425
0.45
0.475
0.5
0.525
P
e
a
k
e
f
f
i
c
i
e
n
c
y
N
o
n

s
y
m
.
W
e
l
l
s
t
u
r
b
i
n
e
(
N
A
C
A
2
4
2
1
)
N
o
n

s
y
m
.
O
p
t
i
m
i
z
e
d
W
e
l
l
s
t
u
r
b
i
n
e
S
y
m
.
W
e
l
l
s
t
u
r
b
i
n
e
(
N
A
C
A
0
0
2
1
)
S
y
m
.
O
p
t
i
m
i
z
e
d
W
e
l
l
s
t
u
r
b
i
n
e
b)TwostageWellsturbine
(highpotentialwaves)
0.3
0.35
0.4
0.45
0.5
0.55
0.6
0.65
0.7
0.75
0.8
0.85
P
e
a
k
t
a
n
g
e
n
t
i
a
l
f
o
r
c
e
c
o
e
f
f
i
c
i
e
n
t
Optimizationdirection
0.4
0.425
0.45
0.475
0.5
0.525
P
e
a
k
e
f
f
i
c
i
e
n
c
y
Optimizationdirection
W
e
l
l
s
t
u
r
b
i
n
e
(
N
A
C
A
0
0
2
1
)
O
p
t
i
m
i
z
e
d
W
e
l
l
s
t
u
r
b
i
n
e
N
o
n

s
y
m
.
W
e
l
l
s
t
u
r
b
i
n
e
(
N
A
C
A
2
4
2
1
)
N
o
n

s
y
m
.
O
p
t
i
m
i
z
e
d
W
e
l
l
s
t
u
r
b
i
n
e
S
y
m
.
W
e
l
l
s
t
u
r
b
i
n
e
(
N
A
C
A
0
0
2
1
)
S
y
m
.
O
p
t
i
m
i
z
e
d
W
e
l
l
s
t
u
r
b
i
n
e
Optimizationdirection
f»0.25
f»0.25
Figure 7.24: Successive optimization steps for the monoplane and twostage Wells tur
bines.
ing of nonsymmetric airfoils has been optimized. The two stages are mirrored, so
that the system globally stays symmetric. Mutual interaction eﬀects between the
blades are taken in account. An aggressive mathematical optimization procedure
has been carried out considering simultaneously sixtyfour free parameters. This
optimization procedure is able to identify a better conﬁguration than the standard
design (NACA 2421). It can be seen that the optimal airfoil shape leads only to
a slightly higher eﬃciency compared to the standard airfoil (NACA 2421). The
relative increase of the tangential force coeﬃcient is however 6.2% at φ = 0.14,
with an average increase of 5.5% along the operating range. The gain in eﬃciency
varies between 0.2% and 0.7%, with an average increase of 0.3%.
• Twostage Wells turbine with symmetric airfoils
Finally, twostage Wells turbines consisting of classical, symmetric airfoils have
been considered. Optimization has been carried out considering twentytwo free
parameters. This optimization procedure is again able to ﬁnd a considerably better
CHAPTER 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 155
conﬁguration than the standard design (NACA 0021). The absolute gain for C
T
increases even slightly with the ﬂow coeﬃcient. The relative increase is higher
than 10% throughout the useful operating range, with an average gain of 12%. At
the same time the eﬃciency of the optimized shape is higher than the conventional
design for a ﬂow coeﬃcient φ ≤ 0.18. The corresponding gain varies between 0.5%
and 2%, with an average increase of 0.6%.
As a whole, and supporting previous ﬁndings from the literature, the monoplane
design appears clearly superior to the twostage design in terms of eﬃciency. At φ =
0.14, the eﬃciency of the optimized monoplane turbine exceeds η 52%, compared to
η 45% for optimized twostage design (Fig. 7.24). The tangential force coeﬃcient of
the twostage design is approximately twice the monoplane one, keeping in mind that
the twostage design needs higher potential waves to operate (see section 2.3.1). The
twostage Wells turbine consisting of symmetric airfoils is found better than the two
stage design consisting of nonsymmetric airfoils in terms of tangential force coeﬃcient
with only minor diﬀerence in eﬃciency (Fig. 7.24).
Finally, we can conclude that these optimized Wells turbine conﬁgurations should
help indeed to improve wave energy conversion.
Chapter 8
Conclusions and Outlook
In this thesis, two important systems allowing energy generation from renewable sources
(Savonius turbine: wind; Wells turbine: sea waves) have been optimized by Genetic
Algorithms, evaluating the performance by Computational Fluid Dynamics. After many
new development, check, and validation steps, the optimization process can be carried
out in a fully automatic manner allowing to explore eﬃciently a variety of conﬁgurations.
Concerning the process, the main ﬁndings are as follows:
• An optimization requires typically one month of computing time on a single PC.
While acceptable for research purposes, this is too long for industrial needs, high
lighting the need for parallel computing. Fortunately, GAs are ideally suited for
parallelization. Furthermore, it must be kept in mind that only twodimensional
geometries have been considered here. For 3D cases, parallel optimization is a
clear must, even for academic research.
• It is possible to optimize considering a large number of design parameters. Up to
64 such free parameters have been employed in this work. This is, however, the
maximum number that can be reasonably considered with the present algorithm.
For concurrent optimization, GA works better for only 10 to 30 parameters and
improvements are needed before considering, say, 100 variables.
• Concurrent optimization involving several objectives is also possible but leads to
additional issues. In particular, it is usually necessary to involve additional criteria
in the ﬁnal decision process, when Pareto fronts are encountered.
Concerning engineering results:
• The performance of the original Savonius turbine could be tremendously increased,
with an extraordinary peak C
p
of about 0.48 compared to 0.18 for the original
system (Fig. 8.1). A patent has been submitted for this conﬁguration.
• In a similar but somewhat less impressive manner, the tangential force coeﬃcient
and eﬃciency of the original Wells turbine could be improved, typically by 12%
for the tangential force coeﬃcient and 1% for eﬃciency. Though limited, this gain
156
CHAPTER 8. CONCLUSIONS AND OUTLOOK 157
is already highly interesting for practical purposes and discussions with possible
industrial partners are running.
Betzlimit(Cp=0.593)
OptimalSavonius
P
o
w
e
r
c
o
e
f
f
i
c
i
e
n
t
Speedratio
Onebladerotor
Twobladerotor Threebladerotor
Darrieusrotor
Savoniusrotor
Dutchwindmill
Americanwindturbine
0
0
2 4 6 8 10 12 14 16 18
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Figure 8.1: Power coeﬃcients of wind rotors of diﬀerent designs including newly devel
oped, optimal Savonius turbine.
8.1 Suggestions for further research
The present study introduced optimized designs with superior performance for Savonius
and Wells turbines. However, a full experimental veriﬁcation of the ﬁndings is needed:
• Wind tunnel or urban wind measurements should be carried out for the optimal
Savonius turbine.
• Experimental measurements should be carried out as well for the optimal Wells
turbine considering an oscillating ﬂow.
Concerning methodology:
• Fully parallel optimization involving parallel CFD should be implemented as a
future standard.
• Improvements in the algorithm are needed to consider many parameters in a con
current optimization.
• The performance of GA could be improved by coupling with other alternatives like
particle swarm or surface responce techniques, allowing to reduce the numbers of
(very costly) evaluations.
Bibliography
[1] B. D. Altan and M. At´ ylgan. The use of a curtain design to increase the perfor
mance level of Savonius wind rotors. Renewable Energy, 35(4):821–829, 2008.
[2] A. Antoniou and L. WuSheng. Practical Optimization, Algorithms and Engi
neering Applications. Springer Science+Business Media, LLC, New York, USA,
2007.
[3] V. Bellucci, B. Schuermans, D. Nowak, P. Flohr, and C. O. Paschereit. Ther
moacoustic modeling of a gas turbine combustor equipped with acoustic dampers.
Journal of Turbomachinery, 127(2), 2005.
[4] D. Benghrib, A. Ahram, and L. Bchir. Description of the vortex formation on a
Savonius rotor in a water channel. Comptes Rendus de l’Academie de Sciences 
Serie IIb: Mecanique, Physique, Chimie, Astronomie, 326(8):495–500, 1998.
[5] R. Bhattacharyya and M. E. McCormick. Wave Energy Conversion. Elsevier
Ocean Engineering Series, 6(7):1–187, 2003.
[6] C. B. Boake, T. J. T. Whittaker, M. Folley, and H. Ellen. Overview and initial
operational experience of the LIMPET wave energy plant. Proceedings of the
International Oﬀshore and Polar Engineering Conference, Kitakyushu, Japan, 586
594, 2002.
[7] P. Boccotti. Comparison between a UOWC and a conventional OWC. Ocean
Engineering, 34(56):799–805, 2007.
[8] A. BritoMelo, L.M.C. Gato, and A.J.N.A. Sarmento. Analysis of Wells turbine
design parameters by numerical simulation of the OWC performance. Ocean En
gineering, 29(1):1463–1477, 2002.
[9] S. M. Camporeale, M. Torresi, G. Pascazio, and B. Fortunato. A 3D unsteady
analysis of a Wells turbine in a seawave energy conversion device. American
Society of Mechanical Engineers, International Gas Turbine Institute, Turbo Expo,
Atlanta, Georgia, USA, GT200338715, 2003.
[10] A. Chauvin and D. Benghrib. Drag and lift coeﬃcients evolution of a Savonius
rotor. Experiments in Fluids, 8(12):118–120, 1989.
158
BIBLIOGRAPHY 159
[11] REN21 Steering Committee. Renewables global status report 2007. RENE21
Renewable Energy policy Network for the 21st century, pages 1–54, 2007.
[12] REN21 Steering Committee. Renewables global status report 2009 update.
RENE21 Renewable Energy policy Network for the 21st century, pages 1–32, 2009.
[13] R. Curran and L. M. C. Gato. Aerodynamic losses incurred by the Wells turbine.
Proceedings of the International Oﬀshore and Polar Engineering Conference, Los
Angeles, California, USA, (1):5865, 1996.
[14] R. Curran and L. M. C. Gato. The energy conversion performance of several types
of Wells turbine designs. Proceedings of the Institution of Mechanical Engineers,
Part A: Journal of Power and Energy, 211(2):133–145, 1997.
[15] R. Curran, R. S. Raghunathan, T. P. Stewart, and T. J. T. Whittaker. Match
ing a Wells turbine to the Islay oscillating water column wave power converter.
Proceedings of the International Conference on Oﬀshore Mechanics and Arctic
Engineering  OMAE 1, Copenhagen, Denmark, Part A, 147154, 1995.
[16] R. Curran, T. P. Stewart, and T. J. T. Whittaker. Design synthesis of oscillating
water column wave energy converters: Performance matching. Proceedings of
the Institution of Mechanical Engineers, Part A: Journal of Power and Energy,
211(6):489–505, 1997.
[17] R. Curran, T. J. T. Whittaker, S. Raghunathan, and W. C. Beattie. Performance
prediction of contrarotating Wells turbines for wave energy converter design.
Journal of Energy Engineering, 124(2):35–52, 1998.
[18] R. Curran, T. J. T. Whittaker, S. Raghunathan, and W. C. Beattie. Performance
prediction of contrarotating Wells turbines for wave energy converter design. Jour
nal of Energy Engineering, 124(2):35–52, 1998.
[19] K. DaeHo, L. MinJeong, J. WonHyouk, and U. Krewer. Nonisothermal dy
namic modelling and optimization of a direct methanol fuel cell. Journal of Power
Sources, 180(1):71–83, 2008.
[20] R. A. Van den Braembussche. Numerical optimization for advanced turbomachin
ery design. In D. Th´evenin and G. Janiga, editors, Optimization and Computa
tional Fluid Dynamics, Berlin, Heidelberg, 2008. SpringerVerlag.
[21] T. S. Dhanasekaran and M. Govardhan. Computational analysis of performance
and ﬂow investigation on Wells turbine for wave energy conversion. Journal of
Renewable Energy, 30(14):2129–2147, 2005.
[22] A. F. D. O. Falca and R. J. A. Rodrigues. Stochastic modelling of OWC wave
power plant performance. Applied Ocean Research, 24(2):59–71, 2002.
BIBLIOGRAPHY 160
[23] M. S. U. K. Fernando and V. J. Modi. A numerical analysis of the unsteady ﬂow
past bluﬀ bodies. Computational Mechanics, 6(1):11–34, 1990.
[24] M. Folley, R. Curran, and T. Whittaker. Comparison of LIMPET contrarotating
Wells turbine with theoretical and model test predictions. Ocean Engineering,
33(89):1056–1069, 2006.
[25] M. Folley and T. Whittaker. The eﬀect of plenum chamber volume and air turbine
hysteresis on the optimal performance of oscillating water columns. Proceedings
of the International Conference on Oﬀshore Mechanics and Arctic Engineering,
Halkidiki, Greece, (2):493498, 2005.
[26] EUJapan Centre for Industrial Cooperation Seminar Report 2008. The future of
energy: Shell energy scenarios to 2050. Climate Change, Environment and Energy
seminar series, (4):1–59, 2008.
[27] N. Fujisawa. On the torque mechanism of Savonius rotors. Wind Engineering and
Industrial Aerodynamics, 40(3):277–292, 1992.
[28] N. Fujisawa. Velocity measurements and numerical calculations of ﬂow ﬁelds in
and around Savonius rotors. Wind Engineering and Industrial Aerodynamics,
59(1):39–50, 1996.
[29] N. Fujisawa and F. Gotoh. Visualization study of the ﬂow in and around a Savonius
rotor. Experiments in Fluids, 12(6):407–412, 1992.
[30] N. Fujisawa and F. Gotoh. Experimental study on the aerodynamic perfor
mance of a Savonius rotor. Solar Energy Engineering, Transactions of the ASME,
116(3):148–152, 1994.
[31] N. Fujisawa, Y. Taguchi, T. Satoh, T. Uemura, K. Nagaya, and S. Ikai. Inves
tigation of ﬂow ﬁeld inside a Savonius rotor by image processing technique with
conditional sampling. Nippon Kikai Gakkai Ronbunshu, B Hen/Transactions of
the Japan Society of Mechanical Engineers, Part B, 59(567):3519–3523, 1993.
[32] L. M. C. Gato and A. F. D. O. Falcao. Aerodynamics of the Wells turbine:
control by swinging rotorblades. International Journal of Mechanical Sciences,
31(6):425–434, 1989.
[33] L. M. C. Gato and A. F. d. O. Falcao. Performance of the Wells turbine with a dou
ble row of guide vanes. JSME International Journal, Series 2: Fluids Engineering,
Heat Transfer, Power, Combustion, Thermophysical Properties, 33(2):265–271,
1990.
[34] L. M. C. Gato, A. F. d. O. Falcao, N. H. C. Pereira, and R. M. R. J. Pereira. Design
of wells turbine for OWC wave power plant. Proceedings of the First International
Oﬀshore and Polar Engineering Conference, Colorado, USA, (1):380384, 1991.
BIBLIOGRAPHY 161
[35] E. D. Glandt, M. T. Klein, and T. E. Edgar. Optimization of chemical processes.
McGrawHill, New York, USA, 2001.
[36] R. Gupta, A. Biswas, and K. K. Sharma. Comparative study of a threebucket
Savonius rotor with a combined threebucket Savoniusthreebladed Darrieus ro
tor. Renewable Energy, 33(9):1974–1981, 2008.
[37] E. Hau. Renewable Energy, Fundamental, Technology, Applications and Eco
nomics. SpringerVerlag, Berlin, Heidelberg, 2006.
[38] T. Hayashi, Y. Li, and Y. Hara. Wind tunnel tests on a diﬀerent phase three
stage Savonius rotor. JSME International Journal, Series B: Fluids and Thermal
Engineering, 48(1):9–16, 2005.
[39] T. Hayashi, Y. Li, and Y. Hara. Wind tunnel tests on a diﬀerent phase three
stage Savonius rotor. JSME International Journal, Series B: Fluids and Thermal
Engineering, 48(1):9–16, 2005.
[40] R. Hilbert, G. Janiga, R. Baron, and D. Th´evenin. Multiobjective shape opti
mization of the ﬂow around a heat exchanger using parallel genetic algorithms.
International Journal of Heat and Mass Transfer, 49:2567–2577, 2006.
[41] H. Holttinen, H. P. Meibom, A. G. Orths, B. Lange, M. O’Malley, J. O. Tande,
A. Estanqueiro, E. Gomez, L. S¨oder, G. Strbac, J. C. Smith, and F. V. Hulle.
Impacts of large amounts of wind power on design and operation of power systems,
results of IEA collaboration. Proceeding of 14th Kassel Symposium on Energy
Systems Technology, Kassel, Germany, 724, 2009.
[42] http://www.scienceonline.co.uk/energy/renewable energy.html. Renewable en
ergy resources. last accessed April 2010.
[43] M. D. Huda, M. A. Selim, A. K. M. S. Islam, and M. Q. Islam. Performance of
an Sshaped Savonius rotor with a deﬂecting plate. RERIC International Energy
Journal, 14(1):25–32, 1992.
[44] M. Inoue, K. Kaneko, T. Setoguchi, and H. Hamakawa. Air turbine with selfpitch
controlled blades for wave power generator (estimation of performance by model
testing). JSME International Journal, Series B: Fluids and Thermal Engineering,
32(1):19–24, 1989.
[45] M. Inoue, K. Kaneko, T. Setoguchi, and S. Raghunathan. Starting and running
characteristics of biplane Wells turbine. Proceedings of the 5th International
Symposium on Oﬀshore Mechanics and Arctic Engineering, Tokyo,Japan, 604
609, 1986.
[46] M. Inoue, K. Kaneko, T. Setoguchi, and S. Raghunathan. The fundamental char
acteristics and future of Wells turbine for wave power generator. Journal of Science
of Machines, 39:275–280, 1987.
BIBLIOGRAPHY 162
[47] M. Inoue, K. Kenji, T. Setoguchi, and F. Koura. Hysteretic characteris
tics of Wells turbine in reciprocating ﬂow. Nippon Kikai Gakkai Ronbun
shu, BHen/Transactions of the Japan Society of Mechanical Engineers,Part B.,
53(496):3699–3704, 1997.
[48] K. Irabu and J. N. Roy. Characteristics of wind power on Savonius rotor using a
guidebox tunnel. Experimental Thermal and Fluid Science, 32(2):580–586, 2007.
[49] K. Ishimatsu, T. Shinohara, K. Kage, and T. Okubayashi. Numerical simulation
for Savonius rotors (eﬀects of shed vortices on running performance). Nippon
Kikai Gakkai Ronbunshu, B Hen/Transactions of the Japan Society of Mechanical
Engineers, Part B, 61(581):12–17, 1995.
[50] K. Ishimatsu, T. Shinohara, and F. Takuma. Numerical simulation for Savo
nius rotors (running performance and ﬂow ﬁelds). Nippon Kikai Gakkai Ronbun
shu, B Hen/Transactions of the Japan Society of Mechanical Engineers, Part B,
60(569):154–160, 1994.
[51] S. A. K. M. Islam, Q. M. Islam, A. C. Mandal, and M. M. Razzaque. Aerody
namic characteristics of a stationary Savonius rotor. RERIC International Energy
Journal, 15(2):125–136, 1993.
[52] G. Janiga and D. Th´evenin. Reducing the CO emissions in a laminar burner
using diﬀerent numerical optimization methods. Journal of Power and Energy,
221(5):647–655, 2007.
[53] M. Kaltschmitt, W. Streicher, and A. Wiese. Renewable Energy, Technology and
Environment Economics. SpringerVerlag, Berlin, Heidelberg, 2007.
[54] M. A. Kamoji, S. B. Kedare, and S. V. Prabhu. Experimental investigations on
modiﬁed Savonius rotor. Collection of Technical Papers  AIAA Applied Aerody
namics Conference, 1:768–780, 2007.
[55] K. Kaneko, T. Setoguchi, H. Hamakawa, and M. Inoue. Biplane axial turbine for
wave power generator. International Journal of Oﬀshore and Polar Engineering,
1(2):122–128, 1991.
[56] K. Kaneko, T. Setoguchi, and S. Raghunathan. Selfrectifying turbines for wave
energy conversion. Proceedings of the First International Oﬀshore and Polar En
gineering Conference, Colorado, USA, (1):385392, 1991.
[57] T. Kim, Y. Lee, I. Park, T. Setoguchi, and C. Kang. Numerical analysis for
unsteady ﬂow characteristics of the Wells turbine. Proceedings of 12th the Inter
national Oﬀshore and Polar Engineering Conference, Kitakyushu, Japan, 694699,
2002.
BIBLIOGRAPHY 163
[58] T. Kim, M. Takao, T. Setoguchi, K. Kaneko, and M. Inoue. Performance com
parison of turbines for wave power conversion. International Journal of Thermal
Sciences, 40(7):681–689, 2001.
[59] T. H. Kim, T. Setoguchi, Y. Kinoue, K. Kaneko, and M. Inoue. Hysteretic char
acteristics of Wells turbine for wave power conversion. Proceedings of the Inter
national Oﬀshore and Polar Engineering Conference, 2002.
[60] T. H. Kim, T. Setoguchi, M. Takao, K. Kaneko, and S. Santhakumar. Study of
turbine with selfpitchcontrolled blades for wave energy conversion. International
Journal of Thermal Sciences, 41(1), 2002.
[61] Y. Kinoue, T. H. Kim, T. Setoguchi, M. Mohammad, K. Kaneko, and M. Inoue.
Hysteretic characteristics of monoplane and biplane Wells turbine for wave power
conversion. Energy Conversion and Management, 45(910):1617–1629, 2004.
[62] Y. Kinoue, T. Setoguchi, T. Kim, K. Kaneko, and M. Inoue. Mechanism of
hysteretic characteristics of Wells turbine. Nippon Kikai Gakkai Ronbunshu,
B Hen/Transactions of the Japan Society of Mechanical Engineers, Part B,
69(679):610–616, 2003.
[63] H. Lee, J. Kim, Y. Lee, T. Setoguchi, and C. Kang. Numerical analysis of ﬂow
characteristics in a Wells turbine for wave power conversion. Proceedings of the
10th International Oﬀshore and Polar Engineering Conference, Seattle, USA, 427
433, 2000.
[64] M. Lind. Experimental Investigation of a modiﬁed Savonius Turbine. Studienarbeit
Thesis, LSSS01/10, Univ. of Magdeburg, 2010.
[65] J. Makower, R. Pernick, and C. Wilder. Clean energy trends 2009. Clean Edge,
122, 2009.
[66] M. Mamun, Y. Kinoue, T. Setoguchi, K. Kaneko, and Y. Lee. Hysteretic charac
teristics of biplane Wells turbine in a deep stall condition. Proceedings of the 15th
International Oﬀshore and Polar Engineering Conference, Seoul, Korea, (1):523
529, 2005.
[67] M. V. Mark and W. D¨ urrschmidt. Electricity from renewable energy sources.
Federal Ministry for the Environment, Nature Conservation and Nuclear Safety
(BMU), Berlin  Germany:1–44, 2009.
[68] J. Menet. A doublestep Savonius rotor for local production of electricity: A design
study. Renewable Energy, 29(11):1843–1862, 2004.
[69] J. Menet. Aerodynamic behaviour of a new type of slowrunning VAWT. Wind
Energy Proceedings of the Euromech Colloquium, Springer,(43):235–240, 2007.
BIBLIOGRAPHY 164
[70] J. Menet and B. Nachida. Increase in the Savonius rotors eﬃciency via a para
metric investigation. European Wind Energy Conference EWEA, Aerodynamics
and Aeroacoustics, 2004.
[71] J. Menet, L. Vald´es, and B. M´enart. A comparative calculation of the wind tur
bines capacities on the basis of the L−σ criterion. Renewable Energy, 22(4):491–
506, 2001.
[72] M. H. Mohamed, G. Janiga, E. Pap, and D Th´evenin. Optimal performance of a
Savonius turbine using an obstacle shielding the returning blade. Ninth Interna
tional Congress of Fluid Dynamics and Propulsion ASMEICFDP9, ICFDP9EG
249, Alexandria, Egypt, 2008.
[73] M.H. Mohamed. Aerodynamic performance of the Wells turbine used in the sea
wave conversion. Master’s Thesis, Helwan University, Cairo, Egypt, 2003.
[74] M.H Mohamed, G. Janiga, and D. Th´evenin. Performance optimization of a
modiﬁed Wells turbine using nonsymmetric airfoil blades. American Society of
Mechanical Engineers, International Gas Turbine Institute, Turbo Expo, Berlin,
Germany, GT200850815, (6):24812488, 2008.
[75] I. G. Morrison and C. A. Greated. Oscillating water column modeling. Proceedings
of the Coastal Engineering Conference, (1):502–511, 1993.
[76] C. O. Paschereit, B. Schuermans, and D. B¨ uche. Combustion process optimization
using evolutionary algorithm. American Society of Mechanical Engineers, Interna
tional Gas Turbine Institute, Turbo Expo, Atlanta, Georgia, USA, GT200338393,
(2):281291, 2003.
[77] G. Pechlivanoglou, S. Fuehr, C. N. Nayeri, and C. O. Paschereit. The eﬀect
of distributed roughness on the power performance of wind turbines. American
Society of Mechanical Engineers, International Gas Turbine Institute, Turbo Expo,
Glasgow, UK, GT201023258, 2010.
[78] G. Pechlivanoglou, J. Wagner, C. N. Nayeri, and C. O. Paschereit. Active aerody
namic control of wind turbine blades with high deﬂection ﬂexible ﬂaps. 48th AIAA
Aerospace Sciences Meeting, Orlando, Florida, USA, (AIAA 2010644), 2010.
[79] M. Powalko, A. G. Orths, and P. B. Eriksen. Grid operation supported by wind
farms: The Danish experience. Proceeding of 14th Kassel Symposium Energy
Systems Technology, Kassel, Germany, 6371, 2009.
[80] S. Raghunathan. Prediction of performance of biplane Wells turbine. Proceedings
of the Third International Oﬀshore and Polar Engineering Conference, Singapore,
(1):167171, 1993.
BIBLIOGRAPHY 165
[81] S. Raghunathan. A methodology for Wells turbine design for waves energy conver
sion. Proceedings  IMechE: A, Journal of Power and Energy, 209(A3):221–232,
1995.
[82] S. Raghunathan. The Wells air turbine for wave energy conversion. Progress in
Aerospace Sciences, 31(4):335–386, 1995.
[83] S. Raghunathan and W. C. Beattie. Aerodynamic performance of contrarotating
Wells turbine for wave energy conversion. Proceedings of the Institution of Me
chanical Engineers, Part A: Journal of Power and Energy, 210(6):431–447, 1996.
[84] S. Raghunathan, R. Curran, and T. J. T. Whittaker. Performance of the Islay
Wells air turbine. Proceedings of the Institution of Mechanical Engineers, Part A:
Journal of Power and Energy, 209(1):55–62, 1995.
[85] S. Raghunathan, T. Setoguchi, and K. Kaneko. The eﬀect of inlet conditions on the
performance of Wells turbine. USA, Proceedings of the ASME Fluids Engineering
Conference, 87FE2, 1987.
[86] S. Raghunathan, T. Setoguchi, and K. Kaneko. Some techniques to improve the
operation range of the Wells turbine for wave power generator. JSME Interna
tional Journal, Series B: Fluids and Thermal Engineering, 32(1):71–77, 1989.
[87] S. Raghunathan, T. Setoguchi, and K. Kaneko. Some techniques to improve the
operation range of the Wells turbine for wave power generator. JSME Interna
tional Journal, Series B: Fluids and Thermal Engineering, 32(1):71–77, 1989.
[88] S. Raghunathan, T. Setoguchi, and K. Kaneko. Aerodynamics of monoplane Wells
turbine a review. Proceedings of the First International Oﬀshore and Polar En
gineering Conference, Colorado, USA, (1):370379, 1991.
[89] S. Raghunathan and C. P. Tan. Performance of the Wells turbine at starting.
AIAA Journal (Journal of Energy), 6(6):430–431, 1982.
[90] S. Raghunathan and C. P. Tan. Aerodynamic performance of a Wells air turbine.
Journal of Energy, 7(3):226–230, 1983.
[91] S. Raghunathan and C. P. Tan. The eﬀect of blade proﬁle on the performance of
Wells turbine. International Journal of Heat and Fluid Flow, 6:17–22, 1985.
[92] S. Raghunathan, C. P. Tan, and O. O. Ombaka. Performance of the Wells self
rectifying air turbine. Aeronautical Journal, 89(890):369–379, 1985.
[93] S. Raghunathan, C. P. Tan, and N. A. J. Wells. Wind tunnel tests on airfoil in
tandem cascade. AIAA Journal, 19(11):1490–1492, 1981.
[94] M. J. Rajkumar and U. K. Saha. Valveaided twisted Savonius rotor. Wind
Engineering, 30(3):243–254, 2006.
BIBLIOGRAPHY 166
[95] M. J. Rajkumar, U. K. Saha, and D. Maity. Simulation of ﬂow around and behind
a Savonius rotor. International Energy Journal, 6(2):83–90, 2005.
[96] A. Ravindran, K. M. Ragsdell, and G. V. Reklaitis. Engineering optimization:
methods and applications. John Wiley & Sons, Inc., New Jersey, USA, 2006.
[97] T. Setoguchi S. Raghunathan and K. Kaneko. Aerodynamic of monoplane Wells
turbine  A review. International Journal of Oﬀshore and Polar Engineering,
4(1):68–75, 1994.
[98] U. K. Saha and M. J. Rajkumar. On the performance analysis of Savonius rotor
with twisted blades. Renewable Energy, 31(11):1776–1788, 2006.
[99] T. Setoguchi, K. Kaneko, M. Takao, and M. Inoue. Air turbine with selfpitch
controlled blades for wave energy conversion. Proceedings of the 6 th International
Oﬀshore and Polar Engineering Conference, Los Angeles, California, USA, (1):70
74, 1996.
[100] T. Setoguchi, T. Kim, K. Kaneko, M. Takao, Y. Lee, and M. Inoue. Air turbine
with staggered blades for wave power conversion. International Journal of Oﬀshore
and Polar Engineering, 13(4):316–320, 2003.
[101] T. Setoguchi, T. H. Kim, K. Kaneko, M. Takao, Y. W. Lee, and M. Inoue. Air
turbine with staggered blades for wave power conversion. Proceedings of the
International Oﬀshore and Polar Engineering Conference, Kitakyushu, Japan, 662
667, 2002.
[102] T. Setoguchi, T. W. Kim, M. Takao, A. Thakker, and S. Raghunathan. The
eﬀect of rotor geometry on the performance of a Wells turbine for wave energy
conversion. International Journal of Ambient Energy, 25(3):137–150, 2004.
[103] T. Setoguchi, Y. Kinoue, T. H. Kim, K. Kaneko, and M. Inoue. Hysteretic
characteristics of Wells turbine for wave power conversion. Renewable Energy,
28(13):2113–2127, 2003.
[104] T. Setoguchi, Y. Kinoue, M. Mohammad, K. Kaneko, and M. Takao. Unsteady
ﬂow phenomena of Wells turbine in deep stall condition. Proceedings of the Inter
national Oﬀshore and Polar Engineering Conference, Toulon, France, (1):266271,
2004.
[105] T. Setoguchi, S. Santhakumar, M. Takao, T. H. Kim, and K. Kaneko. Eﬀect of
guide vane shape on the performance of a wells turbine. Renewable Energy, 1:1–15,
2001.
[106] T. Setoguchi, S. Santhakumar, M. Takao, T. H. Kim, and K. Kaneko. A modiﬁed
Wells turbine for wave energy conversion. Renewable Energy, 28(1):79–91, 2003.
BIBLIOGRAPHY 167
[107] T. Setoguchi and M. Takao. Current status of self rectifying air turbines for wave
energy conversion. Energy Conversion and Management, 47(1516):2382–2396,
2006.
[108] T. Setoguchi, M. Takao, K. Itakura, M. Mohammad, K. Kaneko, and A. Thakker.
Eﬀect of rotor geometry on the performance of Wells turbine. Proceedings of the
International Oﬀshore and Polar Engineering Conference, Honolulu, Hawaii, USA,
(1):374381, 2003.
[109] T. Setoguchi, M. Takao, K. Kaneko, and M. Inoue. Eﬀect of guide vanes on
the performance of a Wells turbine for wave energy conversion. Proceedings of
the International Oﬀshore and Polar Engineering Conference, Honolulu, Hawaii,
(1):8388, 1997.
[110] T. Setoguchi, M. Takao, K. Kaneko, and M. Inoue. Eﬀect of guide vanes on the
performance of a Wells turbine for wave energy conversion. International Journal
of Oﬀshore and Polar Engineering, 8(2):155–160, 1998.
[111] T. Setoguchi, M. Takao, K. Kaneko, and M. Inoue. Eﬀect of guide vanes on the
performance of a Wells turbine for wave energy conversion. International Journal
of Oﬀshore and Polar Engineering, 8(2):155–160, 1998.
[112] T. Setoguchi, M. Takao, E. Kinoue, K. Kaneko, and M. Inoue. Comparative
study of performances of turbines for wave power conversion. Proceedings of the
International Oﬀshore and Polar Engineering Conference, Seattle, USA, (4):351
357, 2000.
[113] T. Shih, W. W. Liou, A. Shabbir, Z. Yang, and J. Zhu. A new k eddy vis
cosity model for high Reynolds number turbulent ﬂows. Computers and Fluids,
24(3):227–238, 1995.
[114] S. N. Sivanandam and S. N. Deepa. Introduction to Genetic Algorithms. Springer
Verlag, Berlin, Heidelberg, 2008.
[115] L. S¨oder, L. Hofmann, A. G. Orths, H. Holttinen, Y. H. Wan, and A. Tuohy. Ex
perience from wind integration in some high penetration areas. IEEE Transactions
on Energy Conversion, 22(2):4–12, 2007.
[116] M. Suzuki. Design method of guide vane for Wells turbine. Journal of Thermal
Science, 15(2):126–131, 2006.
[117] M. Suzuki and C. Arakawa. Guide vanes eﬀect of Wells turbine for wave power
generator. Proceedings of the Ninth International Oﬀshore and Polar Engineering
Conference, Brest, France, (1):162168, 1999.
[118] M. Suzuki, C. Arakawa, and T. Tagori. Fundamental studies on Wells turbine for
wave power generator (1st report, the eﬀect of solidity, and selfstarting). Bulletin
of the JSME, 27(231):1925–1931, 1984.
BIBLIOGRAPHY 168
[119] M. Takao, Y. Kinoue, T. Setoguchi, K. Kaneko, and S. Nagata. Improvement of
Wells turbine performance by means of end plate. Nihon Kikai Gakkai Ronbun
shu, B Hen/Transactions of the Japan Society of Mechanical Engineers, Part B,
72(10):2381–2385, 2006.
[120] M. Takao, T. Setoguchi, K. Kaneko, and M. Inoue. Air turbine with selfpitch
controlled blades for wave energy conversion. International Journal of Oﬀshore
and Polar Engineering, 7(4):308–312, 1997.
[121] M. Takao, T. Setoguchi, K. Kaneko, and M. Inoue. Eﬀect of guide vanes
on the performance of a Wells turbine. Nippon Kikai Gakkai Ronbunshu, B
Hen/Transactions of the Japan Society of Mechanical Engineers, Part B, 63:577–
582, 1997.
[122] M. Takao, T. Setoguchi, T. H. Kim, K. Kaneko, and M. Inoue. The performance
of Wells turbine with 3D guide vanes. Proceedings of the International Oﬀshore
and Polar Engineering Conference, Seattle, USA, (4):381386, 2000.
[123] M. Takao, T. Setoguchi, Y. Kinoue, and K. Kaneko. Eﬀect of end plates on the
performance of a Wells turbine for wave energy conversion. Journal of Thermal
Science, 15(4):319–323, 2006.
[124] M. Takao, T. Setoguchi, Y. Kinoue, and K. Kaneko. Wells turbine with end plates
for wave energy conversion. Ocean Engineering, 34(1112):1790–1795, 2007.
[125] M. Takao, T. Setoguchi, Y. Kinoue, K. Kaneko, and S. Nagata. Improvement
of Wells turbine performance by means of end plate. Proceedings of the Inter
national Oﬀshore and Polar Engineering Conference, San Francisco, California,
USA, (1):480484, 2006.
[126] M. Takao, A. Thakker, R. Abdulhadi, and T. Setoguchi. Eﬀect of blade proﬁle
on the performance of largescale Wells turbine. Proceedings of the International
Oﬀshore and Polar Engineering Conference, Toulon, France, (1):272276, 2004.
[127] M. Takao, A. Thakker, R. Abdulhadi, and T. Setoguchi. Eﬀect of blade proﬁle on
the performance of largescale Wells turbine. International Journal of Sustainable
Energy, 25(1):53–61, 2006.
[128] S. Tan, Y. Shimizu, and K. Kikuyama. Experimental studies on a Savonius rotor
with casing. Nippon Kikai Gakkai Ronbunshu, B Hen/Transactions of the Japan
Society of Mechanical Engineers, Part B, 63(611):2356–2363, 1997.
[129] A. Thakker, P. Frawley, and E. S. Bajeet. CA9: Analysis of a stallresistant aerofoil
in view of wave power conversion. Proceedings of the International Oﬀshore and
Polar Engineering Conference, Stavanger, Norway, (1):614619, 2001.
BIBLIOGRAPHY 169
[130] A. Thakker, P. Frawley, and E. S. Bajeet. Numerical analysis of Wells turbine per
formance using a 3D NavierStokes explicit solver. Proceedings of the International
Oﬀshore and Polar Engineering Conference, Stavanger, Norway, (1):604608, 2001.
[131] A. Thakker, P. Frawley, and E. S. Bajeet. Performance simulation on CA9 Wells
turbine rotor using a 3D NavierStokes explicit solver. Proceedings of the fourth
European Conference on Turbomachinery, 2001.
[132] D. Th´evenin and G. Janiga. Optimization and Computational Fluid Dynamics.
SpringerVerlag, Berlin, Heidelberg, 2008.
[133] T Thorpe. An overview of wave energy technologies: status, performance and
costs. waveenergy.net. Retrieved 20081013.
[134] J. W. Twidell and A. D. Weir. Renewable Energy Resources. ELBS/E. & F.N.
Spon, London, UK, 1987.
[135] M. Webster and L. M. C. Gato. Eﬀect of rotor blade shape on the performance of
the Wells turbine. Proceedings of the International Oﬀshore and Polar Engineering
Conference, Brest, France, (1):169173, 1999.
[136] M. Webster and L. M. C. Gato. The eﬀect of rotor blade sweep on the performance
of the Wells turbine. International Journal of Oﬀshore and Polar Engineering,
9(3):233–239, 1999.
[137] P. White, M. Webster, A. Thakker, and M. O’Dowd. Utilization of CFD to study
an oscillating water column. Proceedings of the International Oﬀshore and Polar
Engineering Conference, Honolulu, Hawaii, USA, (1):7076, 1997.
[138] T. J. T. Whittaker and S. McIlwaine. Shoreline wave power experience with
the Islay prototype. Proceedings of the First International Oﬀshore and Polar
Engineering Conference, Colorado, USA, (1):393397, 1991.
[139] H. Yu. Numerical investigation of a rotating system using OpenFOAM. Master’s
Thesis, LSSM02/09, Univ. of Magdeburg, 2009.
[140] Z. Yu. Selfstarting process simulation of Wells turbine in regular and irregular
oscillating air ﬂows. China Ocean Engineering, 4(4):469–476, 1990.
[141] Z. Yu and H. Maeda. Optimal matching design method of the Wells turbine with
the air chamber in an oscillating air ﬂow. Journal of Hydrodynamics, 3(2), 1991.
[142] J. Zarinchang and A. Sabzevari. Variable geometry Savonius windmill: modeling
and wind tunnel testing. Modelling, simulation and control. B, 56(1):57–64, 1994.
BIBLIOGRAPHY 170
********* Curriculum Vitae *********
Name: Mohamed Hassan Ahmed Mohamed
Date of Birth: Juli 15, 1974; in Cairo, Egypt
Nationality: Egyptian
Status: Married, two Children
Address: Coswiger Str. 8,
D39114 Magdeburg, Germany
Email: mohamed.mohamed@st.ovgu.de
moh75202@yahoo.de
Education:
1989  1992 Secondary School Certiﬁcate, Cairo, Egypt.
1992  1997 B.Sc. in Mechanical Power Engineering, Faculty of Engineering
Mattria Helwan University, Cairo, Egypt.
1998  1999 Military service.
1999  2001 Postgraduate courses for M.Sc. Faculty of EngineeringMattria,
Helwan University, Cairo, Egypt.
2001  2003 M.Sc. in Mechanical Power Engineering, Helwan University,
Cairo, Egypt.
2003  2005 Postgraduate courses for Ph.D. Faculty of EngineeringMattria,
Helwan University, Cairo, Egypt.
Aug. 2007  till Now Works towards Ph.D. degree at Institut f¨ ur Str¨omungstechnik
& Thermodynamik, Lehrstuhl Str¨omungsmechanik & Str¨omungstechnik
(ISUT/LSS), OttovonGuerickeUniversity Magdeburg, Germany.
Work Experience:
1999  2003 Demonstrator in Dept. of Mechanical Power Engineering, Faculty of
EngineeringMattria, Helwan University, Cairo, Egypt.
2003  2007 Assistant lecturer in Dept. of Mechanical Power Engineering, Faculty
of EngineeringMattria, Helwan University, Cairo, Egypt.
Aug. 2007  till Now PhD student and Assistant lecturer in Turbomachines, Chair of Fluid
Dynamics and Technical Flows, OttovonGuericke University
Magdeburg, Germany.
Magdeburg, 4. October 2010
M. H. Mohamed
BIBLIOGRAPHY 172
——————————————————–
Related Publications
The presented thesis is partly based on the following publication in international
reviewed journals and conferences:
1. Mohamed, M.H., Janiga, G., Pap, E. and Th´evenin, D., MultiObjective Op
timization of the Airfoil Shape of Wells Turbine used for Wave Energy
Conversion. Energy, 36(1):438446, 2011.
2. Mohamed, M.H., Janiga, G., Pap, E. and Th´evenin, D., Optimal Blade
Shape of a Modiﬁed Savonius Turbine Using an Obstacle Shielding the
Returning Blade. Energy Conversion and Management, 52(1):236242, 2011.
3. Mohamed, M.H., Janiga, G., Pap, E. and Th´evenin, D., Optimization of
Savonius turbines using an obstacle shielding the returning blade. Re
newable Energy, 35(11):26182626, 2010.
4. Mohamed, M.H., and Th´evenin, D.: Performance optimization of a Savo
nius Turbine Considering Diﬀerent Shapes for Frontal Guiding Plates.
In: 10th Int. Congress of Fluid Dynamics ASMEICFD10, EG3026, Cairo, Egypt,
2010.
5. Mohamed, M.H., Janiga, G., Pap, E. and Th´evenin, D.: Optimal perfor
mance of a modiﬁed threeblade Savonius turbine using frontal guiding
plates. In: ASME Turbo Expo Conference, Glasgow, Scotland, GT201022538,
2010.
6. Mohamed, M.H., Janiga, G. and Th´evenin, D.: Airfoil shape optimization
of a twostage Wells turbine. In: microCAD10 International Scientiﬁc Confer
ence, (Bikfalvi, P., Ed.), Miskolc, Hungary, ISBN 9789636619107, 5156, 2010.
7. Mohamed, M.H., Janiga, G. and Th´evenin, D.: Airfoil shape optimization
of a twostage Wells turbine. In: proc. MagdeburgMiskolc Interuniversity
Cooperation (50 Year Cooperation), (Bikfalvi, P., Ed.), Miskolc, Hungary, ISBN
9789636619244, 99104, 2010.
8. Mohamed, M.H., Janiga, G., Pap, E. and Th´evenin, D.: Optimal perfor
mance of a Savonius turbine using guiding plates. In: Conference on Mod
elling Fluid Flow (the 14th International Conference on Fluid Flow Technologies),
(Vad, J., Ed.), Budapest, Hungary, ISBN 9789634209874, 871878, 2009.
9. Mohamed, M.H., Janiga, G., Pap, E. and Th´evenin, D.: Optimal shape of
a modiﬁed Wells turbine considering mutual interaction between the
blades. In: First International Conference of Energy Engineering ICEE1, Aswan,
Egypt, 2008.
BIBLIOGRAPHY 173
10. Mohamed, M.H., Janiga, G., Pap, E. and Th´evenin, D.: Optimal perfor
mance of a Savonius turbine using an obstacle shielding the returning
blade. In: 9th Int. Congress of Fluid Dynamics and Propulsion ASMEICFDP9,
Alexandria, Egypt, 249/1249/9, 2008.
11. Mohamed, M.H., Janiga, G. and Th´evenin, D.: Performance optimization
of a modiﬁed Wells turbine using nonsymmetric airfoil blades. In:
ASME Turbo Expo Conference, Berlin, Germany, GT200850815, 24812488, 2008.
I
I would like to dedicate this thesis
TO
MY PARENTS
TO
My wife DALIA and my children SAMA and AHMED
TO
My sisters SOMIA and ASMAA
AND TO
My brothers IBRAHIEM and AHMED
Acknowledgment
Firstly, I would like to express my thankfulness and gratitude to my country Egypt for the ﬁnancial support during my research. Without that, I was not able to work and search here in Germany.
I am greatly indebted to my supervisor Prof. Dr. Ing. Dominique Th´venin. I e am extremely grateful for his support, invaluable guidance and for his continuing help even before I came to Germany. I also wish to thank him for providing a wonderful work atmosphere and facilities.
I am also grateful to Dr. Elem´r Pap for his valuable suggestions and help. I e am very grateful to Dr. G´bor Janiga for his great help. I acknowledge the eﬃcient a support of Matthias Lind concerning all experimental measurements.
I always feel lucky to be with so many excellent researchers. Thanks are due to all colleagues of my institute, who were always quite helpful during my stay.
I would like to thank my colleague Dr. Hemdan Shalaby for his great help and his valuable advices about the life in Magdeburg during starting time.
Finally, my sincere thanks go to my wife Dalia who oﬀered her invaluable support to help me during this long education journey. M. Mohamed Magdeburg, Germany 1. October 2010
II
.
Wind energy is becoming particularly important. standard Wells turbines show several wellknown disadvantages: a low tangential force. frontal guiding plates have been considered and lead to a superior performance of Savonius turbines.Abstract Research and development activities in the ﬁeld of renewable energy have been considerably increased in many countries recently. The target function is the output power coeﬃcient. high enough to obtain selfstarting conditions. The static torque is found to be positive at any angle. leading to a low power output from the turbine. a high undesired axial force. which measures the selfstarting capability of the turbine. The performance increases throughout the useful operating range. the Wells turbine is one of the technical systems allowing an eﬃcient use of the power contained in waves with a relatively low investment level. The Savonius turbine appears to be particularly promising for such conditions. Compared to a standard Savonius turbine. a relative increase of the power output coeﬃcient by 58% is ﬁnally obtained at design point. It consists of a selfrectifying air ﬂow turbine employed to convert the pneumatic power of the air stream induced by an Oscillating Water Column into mechanical energy. The present study considers improved designs in order to increase the output power of a classical Savonius turbine. It aims at improving the output power of the Savonius turbine as well as its static torque. In the present work an optimization process is employed in order to increase the tangential force induced by a monoplane and twostage Wells turbine using symmetric airfoil blades as well as by a twostage Wells turbine using nonsymmetric airfoil blades. the available technical design is not yet adequate to develop reliable wind energy converters for conditions corresponding to low wind speeds and urban areas. due to the worldwide energy crisis. The geometry of the blade shape (skeleton line) has been optimized in presence of the obstacle plate. On the other hand. The optimization process is realized by coupling an inhouse optimization library (OPAL. In order to achieve both objectives. many designs have been investigated and optimized by placing in an optimal manner an obstacle plate shielding the returning blade. The automatic optimization procedure in this part of the work is again carried out by coupling the inhouse optimization library OPAL with IV . but suﬀers from a poor eﬃciency. Finally. usually a low aerodynamic eﬃciency and a limited range of operation due to stall. Although considerable progress have already been achieved. Considering now ocean’s and sea’s energy. relying in the present case on Evolutionary Algorithms) with an industrial ﬂow simulation code (ANSYSFluent).
The optimization of the airfoil shape lead to a considerably increased power output (+12%) and simultaneously to an increase of eﬃciency throughout the full operating range.V the industrial CFD code ANSYSFluent. This multiobjective optimization relying on Evolutionary Algorithms takes into account both tangential force coeﬃcient and turbine eﬃciency. demonstrating the superiority of the proposed solutions. Detailed comparisons are ﬁnally presented between the optimal designs and the classical Wells turbine using symmetric airfoils. .
welche die durch eine oszillierende Wassers¨ule eingebrachte a pneumatische Energie des Luftstroms in mechanische Energie in Form von Rotation umwandelt. Hierbei ist der Auso gangsleistungskoeﬃzient die Zielfunktion. um zuverl¨ssige Windenergieanlaa gen f¨r Bedingungen mit vergleichbar geringen Windgeschwindigkeiten und st¨dtischen u a Gebieten zu entwickeln. o VI . so ist die WellsTurbine eines der technischen Systeme. um die Ausgangsleistung einer klassischen Savonius Turbine zu erh¨hen. o Obwohl bereits erhebliche Fortschritte erzielt wurden. eine geringe aerodynamische Eﬃzienz u u und einen begrenzten Betriebsbereich durch Str¨mungsabriss. u a wurde eine Vielzahl von Entw¨rfen untersucht und optimiert. welches eine eﬃziente Nutzung der Energie in Wellen unter Aufwendung relativ geringer Investitionskosten erm¨glicht. leidet aber unter einem schlechten Wirkungsgrad. Dabei nimmt die Windenergie eine zunehmend wichtige Rolle ein. Im Vergleich zu einer StandardSavonius Turbine ist eine relative Erh¨hung des Ausgangsleistungskoeﬃzienten um 58% am Ausleo gungspunkt erreicht worden. Die vorliegende Studie betrachtet verbesserte Designs. Auf der anderen Seite zeigen Standard WellsTurbinen mehrere bekannte Nachteile: eine geringe Tangentialkraft. was zu einer geringen Leistung der Turbine f¨hrt. welches f¨r die SelbstStartf¨higkeit der Turbine bestimmend ist. Das statische Drehmoment erweist sich in jedem beliebigen Winkel positiv und ist hoch genug. eine hohe unerw¨nschte axiale Kraft. wobei in optimaler Weise u ein Leitblech zur Abschirmung des r¨ckkehrenden Turbinenblattes positioniert wurde. u Die Optimierung wurde durch die Kopplung einer hauseigenen Optimierungsbibliothek (OPAL im vorliegenden Fall auf Evolution¨ren Algorithmen basierend) mit einem ina dustriellen Str¨mungssimulations Code (ANSYSFluent) realisiert. Um beide Ziele zu erreichen. Diese zielt sowohl auf die Verbesserung der Leistung o der SavoniusTurbine als auch auf die Steigerung des statischen Drehmoments. Die SavoniusTurbine scheint besonders vielversprechend f¨r u solche Bedingungen zu sein. um die SelbstStartbedingung der Turbine zu erf¨llen. u Die Geometrie des Turbinenblattes (SkelettLinie) wurde in Anwesenheit des Leitblechs optimiert. Dies f¨hrte schließlich zu einer gesteigerten Leistung der Savonius Turbine. Die Leistung steigt im gesamten Betriebsbereich. u Betrachtet man nun die Wasserkraft.Zusammenfassung Aufgrund der weltweiten Energiekrise wurden die Anstrengungen in Forschung und Entwicklung im Bereich der erneuerbaren Energien in den letzten Jahren in vielen L¨ndern a erheblich erh¨ht. In der vorliegenden Aro beit wurde ein Optimierungsprozess eingesetzt. ist das zur Verf¨gung stehende u technische Design noch nicht ausreichend angepasst. Diese besteht aus einer sich selbst richto enden Luftstrom Turbine. um die Tangentialkraft zu erh¨hen.
. Abschließend werden detaillierte Vergleiche zwischen den optimalen Designs und der klassischen WellsTurbine mit sym¨ metrischen Proﬁlen pr¨sentiert.VII welche durch eine zweistuﬁge WellsTurbine mit symmetrisch proﬁlierten Schaufeln. Die Optimierung der Tragﬂ¨chenform f¨hrte zu einer beachtlich a u gesteigerten Leistung (+12%) und gleichzeitig zu einer Erh¨hung der Eﬃzienz uber o ¨ den gesamten Betriebsbereich. Das automatische Optimierungsverfahren in diesem Teil der Arbeit wird wiederum durch die Kopplung der hauseigenen Optimierungsbibliothek OPAL mit dem industriellen CFDCode ANSYSFluent durchgef¨hrt. sowie durch eine zweistuﬁge WellsTurbine mit nichtsymmetrisch proﬁlierten Schaufeln induziert wird. welche die Uberlegenheit der vorgeschlagenen L¨sungen a o veranschaulichen. Diese Mehrzieloptimierung unter u Berufung auf Evolution¨re Algorithmen ber¨cksichtigt sowohl den Tangentialkraftkoa u eﬃzienten als auch den Wirkungsgrad der Turbine.
. . . . 2. . . . . 2. . .2 A few numbers .2. .1. . . . . . . . . . .2 Oscillating Water Column (OWC) principle . .3 Wave energy conversion . . . . . . . . .1 Some properties of renewable energy 1. . .3 Future of renewable energy . . . . . . . 2.1. . . . . . . . . . . . . . . . . . . . .1 Betz’s momentum theory . 1.1 Renewable energy . . . . . VIII . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 2 4 5 6 7 8 10 10 10 12 13 15 16 19 20 21 22 23 23 2 Basic concepts 2. . . . . . . . . . . . . . . . . . 2. . . . . . .3. . . . . . . . . 2. . . . . 1.2. . . . . . . . . . . .1 Wave energy potential . . . . . .2. . 1. . .Contents Acknowledgment Abstract Zusammenfassung Index Nomenclature List of Tables List of Figures 1 Introduction 1. . . . . . . . . . . . . . . . . . . . . . . . . 2. . . . . . . . . . . .2. . . . . . . . . . . . . . .2. . 2. . . . . . . .3. . . . 2. . . . . . . . .2 Scope of the Thesis . . . . . .2 Vertical axis turbines . . .1 Introduction . . . . .2. . . . . . . . . . . . . . . . . .4 Principle operation of standard Savonius turbines 2. . . . . . . . . . . . . .3. . . . . . . . . . . . . . . 2. . . . . . . . . 2. . . . . . . . . . . . . . . . . . 1. .3. . . . . . . .1 Horizontal axis turbines . . . . .3 Betz limit . . . . . . . . . . . . . . . . . . .3 Wells turbines . . . . . . . . . II IV VI XII XIV XVII XXIV . . . .2. . . . . . . .5 Performance of a Savonius turbine . . . . . .1 Conclusions and outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1. . . . .2 Wind energy conversion . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . .1. . . . . . .2 Hubtotip ratio . .2. . . . .3. . . . 3.3. . . . .4 Multiplane Wells turbine . . . . . . 3. . 3. . . . . . 4. . . . . . . . .2.2 Unsteady ﬂow eﬀect on the performance of Wells turbine 3. . . . . . . . . . . . . . . .1 Experimental investigations of Savonius turbines .2. . . . .1. . . . . . . . . . . . . . . . . . . . . 3. . . . . . . . . . . . . . . . . . . . . . 3. . . . . . . . . . . . .2. . . . . . . .2. . . . . . . . . . .2. . . . . . . . .3 How can we achieve optimization? . . . . . . . 3. . .2. . . . .CONTENTS IX 2. . . . .3 Aspect ratio . .1. . . . . . . . . . . . .1. . . . . . . . . . . .2. . . . . .3 Conclusions . . . . . . . . . . . . . . .3 Guide vanes . . . .6 Modiﬁed Savonius rotor . . . . . . 3. .1 Introduction . . .2. . . . . . . . . . . . . . . .2.1. . . . . .1. .3. . . . . . . . . . . . . . . . . .4 Reynolds number . . . . . . . 25 Conclusions . . . . . . . . . . . . . . .4 Summary of Savonius turbine review . . .1. . . . .3. . . . . .3. . 4.3 Methods to improve Savonius turbine performance . .5. . . . . . . . . . . . . . . . . . . . . . . . .3 Methods to improve Wells turbine performance . . . . .3. . .2 Double and three steps Savonius rotor .4 Twistedblade Savonius rotor . 3. .5 Types of mathematical programming 4. . . . . . . . . . . . . . . . 3. . . .4 2. . . . . .5 Wells turbine performance .2 Selfpitchcontrolled blades . . . . . . .1. . . . 4. . . . .5 Summary of Wells turbine review . . . . . . . . .2. .2. . 3. . . 3.1 Solidity . . . . . .1. . . . . . . . . . . . . . . . . . . . . .1. . .1 Guide vanes installation .2. . . . . . .2 Twostage contrarotating Wells turbine . . . .1 Savonius turbine . . 25 2. . . . 3. . . . . 3. . . . . .1. . . 3. . . . 3. .4 Operation of Wells turbines . . . . . . . . . . . . . 3. . . . . . . . .1 Deﬂector plate . . . . . . . . . . . . . . .3 Lean blade (blade swept) . . . . . . . . . . . . . . . . .1.4 Structure of optimization problems . 3. 3. .3. . .4. . . . . . . .2. . . . . . . . . . 3. . . . . . . . . . 3. . . 3. . . . . . . . . . . . . . .2. . . . . . . . . .7 Blade shape . . . . 3. . 3. . . . . .2. . . . . 3. . . . . . 28 28 28 31 33 33 33 33 34 35 36 37 39 39 40 40 41 41 41 42 42 43 44 45 45 47 47 48 48 49 52 53 54 54 55 55 57 58 58 3 Literature Review 3. . . . . . . . . . . . . .1 Twostage Wells turbine (biplane turbine) . .5 Savonius rotor using a guidebox tunnel .2 Optimization uncertainty . . . . . . . . . . 3. . . . . . . . . . . . . . . . . . . . . . . . . . 3. . . . . . . . . . . . . . .2. . . . . . . . . . . . . .4. . . . . . 27 . . . . .1 Linear programming (LP) . . . .1 Performance parameters of Wells turbine . . 3. . . . . . . . . . . . . . . . . . . . . . . . .2.1. . . . .1. . . . . . . . . . . . . . .3. . . .1. . . . .3. . . . . . . . . . . . . . . . . . .3. . . . . 3. . 3. . . .4 End plate . . . . .1. . . . . . . . . . . . . . . . . .2 Numerical investigations of Savonius turbines . . . . . . . . .5 Tip clearance . .6 Inlet turbulence . . . . . . . . . 4 Optimization 4.3. . . . . 4. . . .2 Wells Turbine .3. . . . . . . . . . .1. . . . . .
. . . . . . . .3. . . . . .1 Introduction . . . . . . . . . . . . . . . . . . . . . . . .2 Reducing the drag on the returning blade . . . . . .1 Deﬁning the system boundaries . . . . . . .5. . . .3 Model validation and selection for Wells turbine . . . .3. . . 4.1 Introduction . . . .4 Realizable k − turbulence model . . . . 4. . . . . .2 Numerical solution of the ﬂow ﬁeld . . . . . . . . . . . . . . . . . . . . . . . . .3. 5. . . . 5.2. . . . . . X 58 58 59 59 59 59 60 60 60 61 61 62 65 66 68 68 68 69 69 69 70 70 70 72 72 72 74 75 77 77 77 78 80 80 81 81 82 82 82 82 84 86 4. . Optimization methods . . . . . . . . .3 PostProcessing: analysis of results . . . . . . . .6. . . . 6. . . .3. . . . . . .1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5. . . 5. . . . . . . 4. . .5. . . . . . . . . . . .5 Dynamic programming (DP) . . . . .2 Performance criterion . . . . . . . . . .3. . . . . 5. . . . 5. . . . . . . . . . . . . . . . . .3. .2 Savonius turbine mesh independence . . . . . . . Evolutionary Algorithms . . . . . . .2. .1 Savonius turbine: single objective optimization . . . . . . .7 4. . . . . . . . . . . 4. 4. . . . . . . . . . . . . . . . . . . . . .5. .1 Preprocess: geometry & grid generation . . . . . . .5. . . . . .3. . . . . . . . . . . .3. . . . . . 6. . . . . . . . . 6 Savonius turbine: singleobjective optimization 6. . . . 5. . . . . . . . 4. . . . . . . . . . . . . . . . 5. . .2. . . . . . . . . . . . . . . .3 Computational procedure . . . . . . . . . . . . . 5. . . . .3 Quadratic programming (QP) . . . .4 Wells turbine mesh independence . . .6. . . . .2. .3. . . . . . . .5 Optimization parameters . . . . 5. . . . . . . .2 CFD & Optimization . . . . . . . . . . .2 Optimal GWturbine: modiﬁed threeblade Savonius turbine without gap 6. 5. . 5. . . . . . . . . . . .1 Savonius turbine: size of computational domain . . . . . . . . . . . . . . .5. . . . . . . . . . . . . . 6. . . . . . . . . . . . . . . . . . .3. . . .3 Independent variables . . . . . .1 Performance of the original GWturbine . . .2. . .3. . . .1 Why Fluent? .3. . . . .6. . . . . . . . 5. 5.2 Forces calculation for Wells turbines . . . . . . 4. . . . . . . . . . .CONTENTS 4. . . .2 Integer programming (IP) . . .4 Nonlinear programming (NP) . . . . . .6 4.1. . . . .4 System model . . . 5. .8. . .2 Wells turbine: multiobjective concurrent optimization . . .2. . 4. .1 Genetic Algorithm (GA) . . . . . . . . . . . . . .4 CFD/Optimization coupling . . .8. . . . 5. . . . . . . 5. .2 Advantage and disadvantage of Genetic Conclusions . . . . . . . . . . . . . . . . . . . . . . . .3 Inﬂuence of a rounded obstacle plate .1. .3. .1 Moment calculation for Savonius turbines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .9 5 Numerical methods and algorithms 5. . . .2. . . . . . . . . . . . . . . . . Requirements for optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . Algorithm . . . . . . . . . .5. . . . . . . . . 4. . . . . . . . . .3 Wells turbine: size of computational domain . . . .6 Conclusions . . .1. . . . . 5.6. .8 4. . . . .2 Model validation and selection for Savonius turbine 5. . . . . . .3. . . 5.3. . . . . . . . . . . .
2 Selfstarting capability . .2. . . .2. . . . . . . XI 87 90 91 91 92 92 96 99 104 105 105 105 106 106 108 109 109 109 112 113 113 114 116 117 119 124 124 126 127 6. 148 . . . . . . . . . 7. . . . . . . . . . . . . . . . . . .3. . . . . . . . . . 6. . . . . .3 Selfstarting capability . .2 Optimal monoplane Wells turbine . . . . . . . . . . .1. . . . . . . . . . . . . . . . . .4. .3. . 137 . . . . . 6.1 Mutual interactions eﬀect . . . . . . . . . . . . . . . . . .1 Introduction . . . . . . . . . . . 7.1 Selfstarting capability . . . 137 . .1 Optimization of airfoil shape . . . . . 150 . . 7.3. . . . . . . .4. . .3. . . Optimal threeblade Savonius turbine .3. . . . . . . . .1. . . . . . .4 6. . . . . . . . . . . . . . . . . . . . . . .2. . . . . .4. . Final optimization of Savonius turbine . . . . . . . . . .3 Optimal Savonius turbine with two guiding plates . . . . . . . . . . . . . . . . . . Optimal twoblade Savonius turbine . . . .1 Optimization .4. . . . . 6. . . . . 6. . . . . . . . . . . . . . . . . 6. . . . . . . 6.3 Optimal threeblade Savonius turbine with guiding plates 6. . .2 Optimal blade shape . . . 6. .3. . .2 Optimization . . . .1. . . . .1. . . . . . . . 7. 7. . . . . . . . . . . . . . . . . . 7. . .3 Oﬀ design performance . . . . . . . . . 7. . . . . . .5. . . . . . . . . . . . . . . .1 Oﬀ design performance . . . . . . . . . .1 Selfstarting capability . . . .2 Oﬀ design performance . . . . . . . . . . . . . . . . . 7.5 6. . . . . . .4. . . . . .1 Inﬂuence of obstacle plate . . . . . . . . . . . . .5 Conclusions on the GWturbine . . . . . . . . 130 . . .4.4. . . . . . . . .2. . . . . 6. .1 Optimization . . . 140 .3. 6. . . . . . . . .4. . 7. . . . Preliminary experimental tests in windtunnel .3. . . .4. . . . . . . . .3. .6 6. . . . . . .4.4 Practical realization . . . . . . . . . . . .2. . . . . . . . . . . . . . 145 . . .4 Conclusions on threeblade design .1.7 7 Wells turbine: Concurrent optimization 7. . . .4. . . . . . . .4. . . . . . . . 6. . . . . . . 6. .3. . . . . . . . . . . .3. . . . . . . .1 Obstacle plate . . .2. . . . . . . . Conclusions on Savonius turbine . . . . . . . . . . . . . . . . . . . . . . . .CONTENTS 6. . 6.3. . . . . . 6. . . . . . . . . . . . . .5. . . . .1 Selfstarting capability . .4.3.1. . . . . . 142 . . . .2.3 Optimal twostage Wells turbine with nonsymmetric airfoils 7. . . . . . . . . . . . . . . . . 149 . . . . . . . 7. . . . . . . . . . .2. . . . . . . 6. . . . . . . .3. . 6. . . . . . . . . .3. . . . . . . . . . . . . . . . .2 Optimal airfoil shape with mutual interactions . . . . . . . .2 Oﬀ design performance . . 6. . . . . .3 6. . . . . .4 Optimization of the blade shape . . . . . . . . 130 . . . . . . 130 . . . . . . . . . .4. . .1 Oﬀ design performance . . . . . . . . . .2 Optimization of the obstacle position . . . . . . . 6. . . . . . . . . . . . . .1 Optimal airfoil shape without mutual interactions .4. . . . .1. .2. . . 6. . 6. .3. . . . . .2 Oﬀ design performance .3 Oﬀ design performance . . 132 . . . . . . . . 7. . . . . . . . . . . . . . 6.3 Proﬁled guiding plates .2 Optimal blade shape with obstacle plate . . . .1 Optimization of airfoil shape . . 140 . . . . . . . . 6. . . .4 Optimal twostage Wells turbine with symmetric airfoils .2 Optimization . . . 138 . 6. . .2. . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . .1 Suggestions for further research . . . . . . . 157 Bibliography Curriculum Vitae Related Publications 157 169 171 . . . . . . . . . . . . .5 XII Conclusions on Wells turbine . 152 8 Conclusions and Outlook 156 8.CONTENTS 7. . . . .
.
m2 Rotor pitch of Wells turbine. m Airfoil blade chord. N Lift force. Hz Blade height. N Axial force. N Tangential force.Nomenclature Roman symbols A B b c CA Cp Cm Cms CT D Dt d ∆p0 FD FL FT FX f∗ H h G gw Ld Lo N P Q Projected area of rotor (DH). m Total pressure diﬀerence coeﬃcient Power coeﬃcient (P/[1/2ρAU 3 ]) Torque coeﬃcient (T /[ρR2 HU 2 ]) Static torque coeﬃcient (Ts /[ρR2 HU 2 ]) Tangential force coeﬃcient Turbine diameter of Savonius turbine (2R). m Turbine diameter of Wells turbine. Pa Drag force. m Blade span of Wells turbine. m3 /s XIV . rpm Output power (2πN T /60). W Volumetric ﬂowrate. m Blade chord (2r) of Savonius turbine. m Total pressure diﬀerence. m Hub to tip ratio for Wells turbine Gap ratio Gap width for the threeblade Savonius turbine. N Wave frequency. m Rotational speed of rotor. m Deﬂector length. m Obstacle length.
(◦ ) Aerodynamic eﬃcieny Deﬂector angle. m Blade solidity Output torque. kg/m3 Rotor angle. m Hub blade radius. (◦ ) Setting blade angle. half thickness. (◦ ) Obstacle angle. m Mean wind velocity in axial direction. m/s Relative velocity. m/s Number of blades of Wells turbine Angle of incidence. (◦ ) Flow coeﬃcient Density. Nm Airfoil max. Nm Static torque quantifying selfstarting capability.CONTENTS XV R rt rh r s T Ts t U ut vA w z Greek symbols α β η γ γb φ ρ θ ω Tip radius of Savonius turbine. m/s Axial air velocity. 1/s . m Blade radius of semicylindrical Savonius blade. m/s Tip blade speed of Wells turbine. (◦ ) Angular speed. m Tip blade radius.
. . . . . . . . . . . . . . . . Optimal conﬁguration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Polynomial coeﬃcients of optimal airfoil shape (best ﬁt) . . . Optimum shape parameters for twostage Wells turbine with nonsymmetric airfoils . . . . . . . . . . . . . . . . .3 6. . .3 7. .14 6. . Acceptable range for the input parameters (parameter space) . . . . . . . 81 Acceptable range for the input parameters (parameter space) . . . . . . . . . . . . . Optimal conﬁgurations . . . .2 5. . Optimum shape parameters for twostage Wells turbine . . . . . . . . . . . Optimum parameters of guiding plates . . . . . . . . .10 6.1 7. . . . .1 6. . . . . . . . . .1 3. .1 6. . . . . . . . . . Polynomial coeﬃcients of optimal airfoil shape (best ﬁt) for twostage nonsymmetric airfoil Wells turbine considering mutual interaction between the blades (upper and lower face) . . . .6 Summary of Savonius turbine main modiﬁcations . . .List of Tables 3. . . . . . . Optimal conﬁgurations (obstacle position and angle) . . . . . .2 6. . Acceptable range for the input parameters (parameter space for blade shape with the obstacle ) . . . . . . . . . . . . . . .7 6. . . . . . . .13 6. . .2 7. . .8 6. . . . . . . 37 Main modiﬁcations allowing to improve the performance of Wells turbines 52 Parameters of the Genetic Algorithm . . . . . . . Acceptable range for the input parameters (parameter space) . . . . . . . . . . . . . . . Acceptable range for the input parameters (parameter space) . . . . . . . . Optimal blade shape with the obstacle . Optimum shape parameters for monoplane Wells turbine . . . . . . . . .7 147 152 XVI . .12 6. . . . . . . .11 6.5 6. . . . . . . . . . . . . Parameter space for the moving points P1 to P16 for upper face and P18 to P33 for lower face . . . . . . . .4 7. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Optimum conﬁgurations . . . . . Parameter space for the moving points P2 to P12 for monoplane Wells turbine. . . . . . . .5 7. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acceptable range for the input parameters (parameter space) . . . . . . . . . . Acceptable range for the input parameters . . . . . .4 6. . . . . . .15 7. . . . . . . . . . . . . . . . . . . .6 6. . . . . Acceptable range for the input parameters for the blade shape . . . . . . . . . . . . . . . . . . . . Optimal conﬁguration . . . . . . . . . . . . . .9 6. . . . . . . . 93 94 97 98 102 103 107 108 111 111 116 117 119 121 123 133 136 136 143 147 7. . .
. . 152 .8 XVII Polynomial coeﬃcients of optimal airfoil shape (best ﬁt) for twostage Wells turbine with symmetric airfoils . . . . . .LIST OF TABLES 7. . . . . . . . . . . . .
. .5 3. . . . . . . . . . . . . . Twistedblade Savonius Rotor [98]. . . .9).2 2. . . . . . . . according to the elementary momentum theory. .5 2. . λ = 1) [68]. . . . . c: ﬂow model. Flow in and around a nonrotating Savonius rotor (λ = 0). . . . . XVIII 3 7 11 12 14 15 17 19 20 21 22 24 25 26 27 29 3. . . . . .9 30 32 32 34 35 35 36 37 . . . Static torque coeﬃcient of a Savonius rotor (Re = 1. . . . . . . d: surface pressure distribution [29]. . Power coeﬃcients of wind rotors of diﬀerent designs [37]. . . . . . .8 3. . . .12 2. . . . . . . . . . . . . . . . . . . . . Power coeﬃcient of Savonius rotor using a guidebox tunnel [48]. . . . . .10 2. . . . . . . . . . . . . . . . . . . . . . . . . .4 3. . . . . . . . . . . . .7 2. 2008 worldwide renewable energy sources [26]. . . Distribution of yearly annual mean values of the wind speed for 10 m above ground worldwide [53]. . . c: ﬂow model.3 2. . .2 2. . . . . . . b: ﬂow inside the rotor. . . . . .56 105 ) [68]. . . . . . . . . . . . . . . . . . Conventional Savonius rotor. . . . . . . . . .56 105 . . . Approximate global distribution of wave power levels in kW/m of wave front [133]. Savonius rotor using a guidebox tunnel [48]. . .List of Figures 1. . . . . . . . . . . .2 3. . . Flow conditions and drag force for vertical axis turbines [53]. . . . . . . Flow conditions due to the extraction of mechanical energy from a freestream air ﬂow. . . . Flow in and around a Savonius rotor in rotation (λ = 0. . . . . . . . . . . . . . . a: visualized ﬂow ﬁeld. . . . . . . .9 2. . d: surface pressure distribution [29]. . . Source: [12]. a: visualized ﬂow ﬁeld. . . Horizontal axis wind turbine. .6 2. . . . . . . . . . . . .8 2. . . .1 1. . b: ﬂow inside the rotor.1 2. Flow velocities and aerodynamic forces at the airfoil crosssection of a blade element [53]. . . . . . . . . . Doublestep Savonius rotor [68]. . . . . . . . . .11 2. . . . . . . . . . . . . . . . . . .1 2008 worldwide renewable energy sources. . Power coeﬃcient for twistedblade Savonius rotor [98]. . . . . Dynamic torque coeﬃcient of a Savonius rotor (Re = 1. . Wells turbine in state of upward and downward air ﬂow. . . . . . . . . . . . . . . . . . . . .13 3. . . . . . . . .7 3. . . . . . . Axial and tangential forces acting on a Wells turbine. . . The wave motion in the OWC device [73]. Schematic description and main parameters characterizing a Savonius rotor. . . . . . . . . . .4 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 3. .6 3. . . . Main geometrical parameters of a Wells turbine. . . . . . . . . . . . . . . . . . . . . . . .
12 Savonius turbine publication statistics in international journals and conferences. . .1 5. . . . . . . . . . . . . . . . . . . . . . 3. . . . . .18 Illustration of the ﬂow structure in blade suction side [103]. . . . . . . .LIST OF FIGURES XIX 38 38 39 40 41 42 44 45 45 46 46 47 48 48 49 50 50 51 51 52 3. . . . 3.22 Swept blade (lean blade) [8]. . . . . . Gridindependence study for the torque coeﬃcient . .16 Eﬀect of unsteady ﬂow on Wells turbine performance [103]. . . . . . . . . . . . . . . Inﬂuence of the number of revolutions on the instantaneous and on the average power coeﬃcient Cp computed by CFD for the optimum design shown later. . . . . . . . . . . . . . . . . 3. . . .21 Principle of using selfpitchcontrolled blades [106]. . 3. .5 Flowchart of a Genetic Algorithm. . . . . . . both compared to published experimental results for a twoblade conventional Savonius turbine [38] . . . . . . . . 64 Example of crossover. . . . 3. . . . . . Inﬂuence of the turbulence model on the tangential force coeﬃcient. . . . . . . . . . . . 92. 97]. . . . . . . .3 5. . . . . . . .6 5. . . . . . .25 Two stage (biplane) Wells turbine. . . . (b) Forward type. . . . . . . . . . . . . . . . . . . . . Validation of computational model: power coeﬃcient compared to experimental results for a threeblade Savonius turbine [48] . . . . . . . . . . . . . . . . . . . . . . . . 3. . . . . . . . . . . . compared to experimental results for a monoplane Wells turbine [107] . . . . . . . . . . . . . 3. . . . . . . . . . . . . . . . .27 Torque coeﬃcients for both LIMPET Rotors during intake and exhaust [24]. 3. . . . 64 Size of the computational domain and impact on the torque coeﬃcient . . . . . . 92. . . . . . 3. . . . . . . . . . . . . . .11 Static torque for the modiﬁed Savonius rotor (θ = 90◦ . . . . . . . . . .10 Modiﬁed Savonius Rotor [69]. .8 78 . . 3. . 70 71 71 72 73 74 75 5. 3. . . . . . . . . . . 3. . . . . . . . . . . b) power coefﬁcient. . . . . . . . . . .15 Aspect ratio eﬀect on the eﬃciency [82. . . . . . . . . . . . . . . . . . . . . . . . . . .3 5. . . 3. . . . . . . . . . . . . . . . 97]. . . . . . . . . . . . . . . . . . . . . . . . . 63 Example of selection based on the roulettewheel for ten individuals. . 3. 4.7 5. . . . .1 4. . . . . . Gridindependence study for the tangential force coeﬃcient . . . . . . . . . . . . . . . . . . . . (c) Backward type [124]. . . . . . . . . . . . . . . . . . . . . . . . . .24 Rotor blade with end plate: (a) Middle type. . . . . . . .19 Installation of guide vanes [106]. . .26 Contrarotating Wells turbine [107]. . . . .28 Comparison of LIMPET and constant ﬂow model turbine eﬃciencies [24]. . . . . . . . . . 3. . . . 3. . . . . Validation of computational model: a) torque coeﬃcient. . . . .17 Eﬀect of solidity on the hysteretic behavior [103]. .2 5. .23 NACA 0015 and HSIM 152621231576 blades [8]. . . . . . . . . . . . . . . . . . . . . . . . .14 Solidity eﬀect on the eﬃciency [82. . . . .4 5.13 Axial and tangential forces acting on a Wells turbine. . . . Size of the computational domain around monoplane and twostage Wells turbines . . . . .2 4. . . 3. . . . . . . .29 Wells turbine publication statistics in international journals and conferences. . . β = 45◦ . . . . . . e/d = 1/6) [69]. . . . . . . . . . . . . . . . . . . . . . . . . . 3. 3.20 Guide vanes eﬀect on the turbine performance [105]. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . The power coeﬃcient of the conventional threeblade turbine (semicylindrical shape) is also shown with a black circle. . . . . . . . . . . . . . . . . .15 6. . . . . . . . . . . . . Schematic description of the geometry and free optimization parameters X1 . Top: torque coeﬃcient. . . . Input parameters of the optimization and power coeﬃcient represented using parallel coordinates. . . . Top: torque coeﬃcient. . . . . . .14 6. . . . . .17 95 96 6. . . . . . . . . . . . .12 6. . . . . . . . . The performance of the conventional Savonius rotor is also shown for comparison. . X2 used to modify the position of the shielding obstacle. . . . . . . . . . . .4 and X2 /R = −1. . .LIST OF FIGURES 5. . . .18 6. . Schematic description and main parameters characterizing a conventional Savonius rotor with three blades. YP 2 . . YP 1 . . . . . . XP 2 . . . . . .10 6. . . . . . . . . . . . . Description of the blade shape with 3 discrete points connected by splines. . . . .5 XX Schematic description of optimizer (OPAL) and CFD code coupling. .7 6. . .16 94 6. . . . . . . . XP 3 and YP 3 used to modify the blade shape. The parameters of the optimal conﬁgurations are connected with a thick red line. . . . . . . . . . . . . . . . Schematic description of the GWturbine with open returning blade. . . . . . . .76. . . . 79 Schematic shape of the GWturbine . . .6 6. . . twoblade Savonius rotor. . . Schematic description of the counterrotating GWturbine with rounded deﬂector. . . . Performance of open returning blade turbine for diﬀerent slit angles. . . . The parameters of the optimal conﬁguration are connected with a thick red line. .1 6. The power coeﬃcient of the conventional threeblade Savonius turbine is shown with a gray circle. . . . . . . . . . . . . . . .19 97 . . . . . . Y1 . . . . . . Performance comparison between the GWturbine and the conventional. . . . . . . . . . . . . The two input parameters of the optimization and the power coeﬃcient. . b: power coeﬃcient. The corresponding relative increase is shown with blue line. . . . . . . . . . . . . . . . . . . Bottom: power coeﬃcient. .3 6. . . . Instantaneous velocity vectors around the GWturbine . . . . . . . . . . . .4 6. . . . Performance comparison between the GWturbine with or without open returning blade. . .9 6. .13 6. . . . . . . . . .2 6. . . . . . . . . . . . . . . . . . . . . . . . . . . . .9 6. Power coeﬃcient of the optimized conﬁguration compared to the GWturbine as a function of λ. . . 83 83 84 84 85 86 87 88 89 89 89 90 91 92 93 6. The best (right) conﬁguration obtained during the optimization compared to the original GWturbine (semicylindrical shape: left). . . . . . . Bottom: power coeﬃcient.8 6. Input parameters of the optimization and power coeﬃcient represented using parallel coordinates. . . . . . . Performance of the optimized conﬁguration (red line) compared to the conventional threeblade Savonius turbine without obstacle (black): a: torque coeﬃcient. . . . Schematic description of the free optimization parameters XP 1 . . . . . . . . . . . . Performance comparison between the new concept in free ﬁeld or placed behind a rounded deﬂector. Static torque coeﬃcient as a function of the rotor angle θ for three diﬀerent values of Y1 choosing X1 /R = −1. .11 6. . .
. Yd1 . 6. . The corresponding relative increase is shown with blue stars. . . . . YP 2 . . . . 6. . . . .27 Static torque coeﬃcient Cms as a function of the ﬁxed rotor angle θ for the optimal design (ﬁlled red squares) compared to the classical threeblade Savonius turbine (blue plus). . .LIST OF FIGURES 6. . YP 1 . . . . . XP 2 . . . . .4 and X2 /R = −1. . Bottom: power coeﬃcient. . 6. . . . . . .32 Performance of the optimized conﬁguration (red line) compared to the conventional twoblade Savonius turbine without obstacle (black line): a: torque coeﬃcient. YP 1 . .20 Optimum conﬁguration obtained with the optimization procedure. . . Y2 . . . The corresponding relative increase compared to the standard conﬁguration is shown with blue line. . . . . . . 6. . additionally.7). The parameters of the optimal conﬁguration are connected with a thick red line. . . . . . XP 3 and YP 3 used to modify the blade shape. .24 Optimum conﬁguration obtained with the optimization procedure. . . . . . . 6. . . . . . . The corresponding relative increase compared to the conventional threeblade conﬁguration is shown with blue line. . . 6. . . . . . . .7: Zoom on the vicinity of the turbine (full CFD domain is much larger). . . . . Xd1 . .23 Input parameters of the optimization and power coeﬃcient represented using parallel coordinates. . . . . 6. . . The experimental results of [48] are also shown for comparison (empty black squares). . Y1 . YP 2 . . . . . . .25 Instantaneous velocity vectors magnitude (m/s) around the optimum conﬁguration (zoom) at the design point (λ = 0. .28 Schematic description of the geometry and free optimization parameters X1 .26 Performance of the optimized conﬁguration (red line) compared to the conventional threeblade Savonius turbine (black line): Top: torque coeﬃcient. . . . . . . . . . . . . The power coeﬃcient of the conventional turbine is shown with a black circle. . The power coeﬃcient of the conventional threeblade turbine (semicylindrical shape) is also shown with a black circle.33 Schematic description of the free optimization parameters XP 1 . . Xd2 and Yd2 ). X2 . XP 3 and YP 3 used to modify the blade shape. . . 6. b: power coeﬃcient. . . .30 Input parameters of the optimization and power coeﬃcient represented using parallel coordinates. Y1 . . . . .76. . 6. . . . . . . 6. . . b) power coeﬃcient. . . . XXI 98 99 100 101 102 103 104 105 106 107 108 109 110 111 . . . . . . b) XP 1 .31 Instantaneous ﬂow structure when the advancing blade is in vertical position for the optimal conﬁguration at λ = 0. . . X2 used to modify the position of the shielding obstacle. . . . .21 Performance of the optimized conﬁguration (red line) compared to the conventional threeblade Savonius turbine (black line): a) torque coeﬃcient. . . . 6. 6. . . .29 Static torque coeﬃcient as a function of the rotor angle θ for three diﬀerent values of Y1 choosing X1 /R = −1. 6. . . . . . . The parameters of the optimal conﬁgurations are connected with a thick red line. . .22 Schematic description of the free optimization parameters characterizing a threeblade Savonius rotor : a) plate parameters (X1 . XP 2 . the gap width gw . . . .
. . . . . . . . . . . . . . . . . The parameters of the optimal design are connected with a thick red line. . . . . . . . . .35 Optimum conﬁguration (right) obtained with the optimization procedure compared to the classical Savonius turbine (semicylindrical shape: left). . . . . .36 Performance of the optimized conﬁguration (red line) compared to the conventional Savonius turbine with and without obstacle plate (blue and black. .44 Optimum conﬁgurations of curved guiding plates. . . . . . . . . . . . . . . . .45 Schematic description of the geometry and free optimization parameters. . . . . The parameters of the optimal designs are connected with a thick red line. . .39 Input parameters of the optimization and power coeﬃcient represented using parallel coordinates. The parameters of the optimal design are connected with a thick red line. . . . . . . . . . . . . . . . . . .34 Input parameters of the optimization and power coeﬃcient represented using parallel coordinates. 6. . . . b: power coeﬃcient. . . . . . . . The parameters of the optimal shape are connected with a thick red line. 6. . .38 Schematic description of the geometry of Savonius turbine with frontal guiding plates. . . . . 6. .43 Input parameters of the optimization and power coeﬃcient represented using parallel coordinates. . . . . . The corresponding relative increase compared to the standard conﬁguration is shown with black line.LIST OF FIGURES XXII 6. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . the experimental results of [38] are also shown for comparison. The power coeﬃcient of the conventional turbine is shown with a black circle. . . . . . 6. . . 6. . . . . . . respectively): a: torque coeﬃcient. . . . . . . .37 Static torque coeﬃcient Cms as a function of the ﬁxed rotor angle θ for the optimal conﬁguration compared to the standard Savonius turbine with and without obstacle plate. . . 6. . .40 Optimum conﬁguration of guiding plates. . . . . . . . . . 6. . The power coeﬃcient of the conventional turbine (semicylindrical shape) with obstacle plate is also shown with a black circle. . The corresponding relative increase compared to the standard conﬁguration is shown with green line. . The power coeﬃcient of the conventional turbine is shown with a black circle. . . . . . . . . . . . . . b: power coeﬃcient. . . . . . . . . .41 Performance of the optimized conﬁguration (green line) compared to the conventional Savonius turbine (blue line): a: torque coeﬃcient. . . . 112 113 114 115 115 116 117 118 119 120 121 122 123 124 . . . . For this last case. . . . 6. . . . . . . 6. . . . . .46 Input parameters of the optimization and power coeﬃcient represented using parallel coordinates. 6. . . . . . . . . . The power coeﬃcient of the conventional turbine is shown with a gray circle. . .42 Schematic description of the geometry of Savonius turbine with proﬁled guiding plates. . . . . . . . . . . . . . . . . . . The power coeﬃcient of the conventional turbine is shown with a black circle. . . 6. 6. . . . . . . . . . .47 Optimum design of Savonius turbine with guiding plates. 6. . . . . . . . . . . . .
. . . . . The parameters of the optimal shape are connected with a thick red line.e. . . . . . . . For this last case. .7 Performance of the spline optimal conﬁguration (red line). . . . . 6. . . . a) Xcoordinates of the variable points (P2 . . The corresponding relative increase is shown with blue line. .7). the optimal airfoil shape described by splines (black squares showing the position of the control points) and the corresponding polynomial ﬁt (Eq. . . . . .10 Impact of mutual interaction between blades in the same plane on a) tangential force (left) and b) axial force (right). . . . . . . 132 Objectives of the optimization. . . . . . . as a function of the solidity. . . . . . . . 139 7.3 7. . . . . P12 ). . 139 7. . .. . . . . . . . b: power coeﬃcient. . . . . . . . . . . zoom on the upperright part (red square) of a). . . . . . . .9 Comparison between the optimal shape of the airfoil and the original proﬁle NACA 2421 . . . . .4 125 126 126 127 128 129 Standard airfoils NACA 0015 and NACA 0021. . . 6. . . . . b) eﬃciency. . . . 7. 135 7.6 Comparison between the original proﬁle NACA 0021 (solid line). . . . . . .3. 6. a) for all computed conﬁgurations b) for the best conﬁgurations (i. . . . . . . . . . . . . b) Optimized design. . b) Ycoordinates of the variable points (P2 . . . . . . static pressure (Pa). . . . . . . . . . . . . . 7. a) classical Savonius with optimal guiding plates. . .5 Input parameters of the optimization and objectives represented using parallel coordinates. . . . . . . . . . . . 131 Allowed parameter space for the moving points P2 to P12 . . . . . .50 Static torque coeﬃcient Cms as a function of the ﬁxed rotor angle θ for the optimal conﬁguration compared to the conventional Savonius turbine. . . . . . . . 138 7. the experimental results of [38] are also shown for comparison. . . . . 6. . . . The corresponding relative increase compared to the classical conﬁguration is shown with blue line. . . . . . . . . . . . . . . . . . .53 Successive optimization steps for the twoblade Savonius turbine. . . . . . . .48 Instantaneous ﬂow ﬁelds around optimum conﬁgurations (zoom) at the design point (λ = 0. . . .2 7. . .8 Twostage Wells turbine based on nonsymmetric airfoils . . .52 Selected experimental results [64] a) conventional Savonius turbine. . . . . . . ﬁtting optimal one (black cross) compared to the conventional Wells turbine relying on the NACA 0021 proﬁle (green line). . . . . . . . . . . . . . . . . . . . . The standard design (NACA 0021) is shown with a thick dashed blue line. . . . . . . . . . . a) tangential force coeﬃcient.141 . . . . . . . . . . . . . . . . 131 Axial and tangential forces acting on a Wells turbine.49 Performance of the optimized conﬁguration (red line) compared to the conventional Savonius turbine (black line): a: torque coeﬃcient. . . 136 7. .51 Model installation with open wind tunnel. . b) optimal Savonius with optimal guiding plates. . .1 7. . velocity and velocity vector magnitudes (m/s). . . 134 7. . . Note that the color scales are identical to facilitate comparisons. . .LIST OF FIGURES XXIII 6. . . . . dashed line). . . P12 ). . . . . . . . . . . . . . . . . . . . . . . . . . 6. .
. 8. . . . . . . . . . . . . . . . . . . . . . . 7. . . . . . The standard design (NACA 0021) is shown with a thick dashed blue line. . . . . . . zoom on the upperright part (red circle) of (a). . . compared to the conventional twostage Wells turbine relying on the NACA 0021 proﬁle (gray squares). . . . . . . . . . . . . . . . . . . . .22 Comparison between the original proﬁle NACA 0021 (solid line). . . . . . . . . b) eﬃciency. . . . . . . . 7. . . . .19 Allowed parameter space for the moving points. . . . . . a) tangential force coeﬃcient. . . . . . . . . . . . . 7. . . . . .17 Performance of the optimal conﬁguration (red line). . . . . . . . . . . .e. . . .14 Objectives of the optimization: a) for all computed conﬁgurations. 7. . . .21 Input parameters of the optimization and objectives represented using parallel coordinates. . . . . . i. . . . . . . . . . . i. . . . . . . . . . . . The parameters of the optimal shape are connected with a thick red line.23 Performance of the optimal conﬁguration (red line). . . . . . . . . . . . 7. . . . . . . . . . . The corresponding relative increase is shown with blue line. the polynomial optimal airfoil shape (red line) and optimal shape by splines (black square). . . . . . . . . zoom on the upperright part (red square) of (a). . . . . . . . . 7. . . . 7. 7. . . . . . .LIST OF FIGURES XXIV 7. . . . . . . . . . . . . . . . . . . . . . . . . . The parameters of the optimal shape are connected with a thick red line. . . . . . . . . . b) for the best conﬁgurations. . . . . . . . . . . . considering mutual interaction between the blades. . . . . . . . . . . . . . . . The corresponding relative increase is shown with blue line. . .13 Allowed parameter space for the moving points. . . . . . . . . . .e. . . . . . . a) Conventional turbine b) Suggestion of isoperformance turbine. . . . . . . . . . . . b) for the best conﬁgurations. . . .1 141 142 144 145 146 146 148 149 149 150 151 151 153 154 Power coeﬃcients of wind rotors of diﬀerent designs including newly developed. . . . . . . . .12 Projected shape of the turbine. . . compared to the nonsymmetric twostage Wells turbine relying on the NACA 2421 proﬁle (black line). .. . . . . 7.18 Twostage Wells turbine with symmetric airfoils NACA 0021. . . . . . . . . . . . . . . 7. .20 Objectives of the optimization: a) for all computed conﬁgurations.15 Input parameters of the optimization and objectives represented using parallel coordinates. . optimal Savonius turbine. . . . . . . . . .. . . . . b) eﬃciency. . . . . . . . . . . . . a) tangential force coeﬃcient. . . 7. . . . . . . . . . . . . . .11 Impact of solidity on the twostage Wells turbine performance considering the tangential force coeﬃcient (a: left ﬁgure) and the turbine eﬃciency (b: right ﬁgure). . . . . . . . . . . 7. . . . 7.16 Comparison between the original proﬁle NACA 2421 (solid line) and the optimal airfoil shape (dashed line). . . . . .24 Successive optimization steps for the monoplane and twostage Wells turbines. . . . . . 157 . . . . . . . . . . . .
.
or wind to produce useful work. In 2006. New renewables (small hydropower installations. What is consumed is the ability of oil. coal. and 3% from hydroelectricity. In this thesis. Geothermal. and raising living standards worldwide. wood. renewable energy sources regenerate and can be sustained indeﬁnitely. improving human welfare. The use of renewable energy is not new. municipal solid waste.Chapter 1 Introduction Life is nothing but a continuous process of energy conversion and transformation. Energy cannot be created or destroyed. ethanol. 1. and use always generate undesirable byproducts and emissions at least in the form of dissipated heat. water potential and ocean wave energy). But energy production. and biodiesel). but it can be converted from one form to another. aﬀordable energy is essential for eradicating poverty. Hydropower (including tidal. especially those associated with poverty. 1. Although it is common to discuss energy consumption. which is one form of biomass.1 Renewable energy Unlike fossil fuels. supplied up to 90% of our energy needs. gas. wind. energy is actually transformed rather than consumed. landﬁll gas. biomass. solar. which are exhaustible. we are looking again at renewable sources to ﬁnd new ways to use them to help meet our energy needs (see Fig. about 18% of global ﬁnal energy consumption came from renewable energies. conversion. it will be diﬃcult to address environmental challenges. only conversion of selected renewable energy sources has been investigated. Today.4% and are growing very 2 . More than 150 years ago. Wind and Solar energy. Energy is similarly indispensable for continued human development and economic growth. Providing adequate. geothermal. biogas. and biofuels) accounted for another 2. The ﬁve renewable sources used most often are: Biomass (including wood and wood waste. modern biomass. with 13% coming from traditional biomass.1). which is mainly used for heating. And without economic growth. The accomplishments of civilization have largely been achieved through the increasingly efﬁcient and extensive harnessing of various forms of energy to extend human capabilities and ingenuity.
coupled with high oil prices. regulation and policies helped the renewable energy industry weather the 2009 economic crisis better than many other sectors [65]. New government spending. In Denmark. yet the renewable energy market continues to grow. In summer 2010. The share of renewables in electricity generation is around 18%. and increasing government support. peak oil. INTRODUCTION 3 Figure 1. renewable technologies are also suited to small oﬀgrid applications. Renewable resources are often located in remote areas. with 15% of global electricity coming from hydroelectricity and 3. Source: [12]. rapidly [11]. Some renewable energy technologies are criticized for being intermittent or unsightly. where energy is often crucial in human development. incentives and commercialization. the governments of the member states are to submit national plans to the European Commission.1: 2008 worldwide renewable energy sources.CHAPTER 1. renewable energy has generally been more expensive to produce and use than fossil fuels. The European policy concerning the use of electricity from renewable energy sources aims at a 20% share of renewable energy in the European energy system.4% from new renewables. the governments longterm policy aims at achieving a 30% share of energy from renewable energy sources in 2020 [79]. In the past. and it is . While most renewable energy projects and production is largescale. Some countries have already developed national targets. sometimes in rural and remote areas. are driving increasing renewable energy legislation. Climate change concerns.
The combustion of fossil fuels produces sulphur dioxide and nitrogen oxide. . These gases are the main components of the ”acid rain” eﬀect . 1. cloudy days reduce solar power. fossil fuels such as coal. polluting water courses and corroding the stone facades of buildings. Exploiting renewable sources of energy. calm days reduce wind power. reduces these risks and hazards. The use of renewable sources is also limited by the fact that they are not always available. Most of these issues are addressed during consultation with local authorities. renewable energy conversion uses virtually none of this most precious of commodities in its operation. The construction and operation. • Water Another consideration of renewable energy deployment concerns water. signiﬁcant improvements have been made with regards to the siting of wind farms and the design of turbines. does have environmental impacts. for instance. from mining and extraction to fuel processing and plant cooling measure their water use in millions of liters per day. and droughts reduce the water available for hydropower. Some important properties of renewable energy conversion can be summarized as follows: • Greenhouse eﬀect First. gas and oil are major emitters of carbon dioxide. often in rural areas. By contrast. renewable energy conversion produces no carbon dioxide (the main potential greenhouse gas) during operation. Most conventional technologies. and only minimal quantities during the manufacture of its equipment and construction.killing forests. not to mention the human health eﬀects. both serious sources of pollution. like any other industry. emission free power generation technology. INTRODUCTION 4 expensive to build power lines to the cities where the electricity they produce is needed. but. Since the early days of this relatively young industry. • Air pollution Renewable energy also has a positive eﬀect on the quality of the air we breathe. the pollution caused by accidental oil spills and the health risks associated with radiation produced by the routine operation and waste management of the nuclear fuel cycle. raises issues of visual impact.1 Some properties of renewable energy Renewable energy is in principle a clean.CHAPTER 1. In an increasingly waterstressed world. • Environmental impact Renewable energy is arguably the cleanest electricity generation technology. Other environmental eﬀects resulting from the range of fuels currently used to generate electricity include the landscape degradation and dangers of fossil fuel exploration and mining. noise and the potential eﬀects on local ecology and wildlife.1.
2 A few numbers Renewable energy resources include solar energy. energy from tides and energy from biomass [42]. like in Denmark. Wind turbines are becoming more eﬃcient. The wind farms also need to be sited reasonably close to populations so that the electricity generated can be distributed.1. Spain. geothermal energy.000. Wind energy is nonpolluting and is freely available in many areas. This looks unsightly. Large balancing areas and aggregation beneﬁts of large areas help in reducing the variability and forecast errors of wind power as well as in pooling more cost eﬀective balancing resources [41]. There are already several power systems and control areas coping with large amounts of wind power [115]. • Wind energy When the earth is irradiated by the sun the ground absorbs some of this radiation.CHAPTER 1. Waves can be many meters in height and contain a great deal of energy. In this thesis. • Solar energy Every year the Earth receives about 300. Germany. The two energy sources considered in the present work (wind and waves) are indirectly a result of sun radiation: solar energy drives processes in the atmosphere that cause the wind and waves [42]. energy from the wind or waves. This results in the air above the land moving upwards more quickly than that over the sea (hot air rises). This energy can be harnessed to . This heated ground warms the air above it. It is processes like these that give rise to high and low pressure areas. The uneven heating of the earth’s surface causes winds.000. To be eﬃcient.000. the land heats up more quickly. For example. Hot air rises in what are called convection currents. wind turbines need to be linked together in wind farms.000. if the sun’s rays fall on land and sea. So ultimately. and thus to winds. including fossil fuels. However. The cost of the electricity they generate is falling. As a result the colder air over the sea will rush in to ﬁll the gap left by the rising air. Another disadvantage is that winds are intermittent and do not blow all the time [42].000. • Wave energy Waves are caused by the action of winds on the sea. Portugal and Ireland that have integrated 920 % of wind energy (of yearly electricity demand).000 kJ of energy from the sun. the disadvantages of wind energy exist as well. Some energy is absorbed by green plants and used to make food by photosynthesis. often with 20 turbines or more. the sun is the source of most energy resources available to us. one turbine used to convert wind energy is extensively optimized in order to improve the output power. Only these relevant for the present thesis are now discussed.000. and can be noisy. INTRODUCTION 5 1.
to solve energy and climate problems. the Wells turbine used to convert wave energy into mechanical energy is investigated and optimized to increase its eﬃciency. 1. Waves’ energy is nonpolluting. it is still ineﬃcient and too expensive in many cases to take over signiﬁcant parts of the energy supply. Already today. wind. For example. from development and conversion. according to the forecasts of the International Energy Agency. Although their reserves will last for the next decades. water. so they would not produce a constant supply of energy. INTRODUCTION 6 drive turbines that generate electricity. Renewable energies have long since emerged from their much ridiculed niche existence and established a ﬁrm place in the energy mix. over 80% of the primary energy demand is covered by fossil fuels. they will not be able to cover the worldwide energy consumption in the long run. Wave energy collectors are of two main types. but also on the use of renewable energy sources in heat production and in the transport sector [67]. it is not only necessary to economically utilize renewable alternatives to fossil fuels. wave heights vary considerably. tides. the Limpet shoreline wave energy concept has been commissioned in December 2000 on the Island of Islay. The second type uses the up and down movement of a wave to push air. The ﬁrst type directs waves into manmade channels. geothermal or biomass the global energy demand could be met many times over. water and sun are economically competitive in some regions. however. These state that by 2020 renewable energies are to account for as much as 20% of Europe’s energy consumption. The waves feed indirectly a pair of counterrotating Wells turbines. In this work. while signiﬁcantly reduced prices are expected for renewable energies. In view of possible climatic changes due to the increase in the atmospheric CO2 content as well as the conceivable scarcity of fossil fuels. These targets focus attention not only on the electricity sector. 50% until 2030. wind. it becomes clear that future energy supply can only be guaranteed through increased use of renewable energy sources. oﬀ the west coast of Scotland. some disadvantages exist: the turbines can be unsightly. it is foreseeable that prices for fossil fuels will rise. where the water passes through a turbine that generates electricity. currently. Innovation and increases in eﬃciency in conjunction with a general reduction of energy consumption are urgently needed in all ﬁelds to reach the targets within the . With energy recovery through renewable sources like sun. each of which drives a 250 kW generator. However.e.3 Future of renewable energy The worldwide energy demand is continuously growing and. Due to the usual adaptation reactions on the markets. Their further expansion is certain now that the European Union has laid down ambitious and binding targets.1. However. giving a theoretical peak power of 500 kW [24].CHAPTER 1. i. transport and storage up to the consumers’ utilization. It is intended to enable Islay to replace fossil fuels and become selfsuﬃcient through renewable energy. it is expected to rise by approx. Currently. but also to optimize the whole value added chain of energy. wave turbines are relatively quiet to operate and do not aﬀect wildlife..
Figure 1. a considerably improved design must be available.CHAPTER 1. • At the end. INTRODUCTION 7 given time since the world population is growing and striving for more prosperity [26] (see Fig. . • Investigate the impact of a deﬂector installed simultaneously with the obstacle (frontal guiding plates) to redirect the ﬂow toward the advancing blade(s) and optimize the position and angle of these guiding plates as well as the shape of the blades under the eﬀect of these plates. Savonius turbines (wind energy conversion) • Study the impact of an obstacle shielding partially the returning blade(s) and optimize the position and angle of this obstacle as well as the shape of the blades under the eﬀect of this obstacle. the speciﬁc objectives of this work are: 1.2: 2008 worldwide renewable energy sources [26]. 1.2 Scope of the Thesis The optimization of renewable energy turbomachines is a completely new topic. Therefore. turbomachines used for wind and wave energy conversion are still at a very basic stage. The aim of this thesis is to investigate and optimize two such devices. While gas turbine have been for instance considered extensively.2). 1. • Optimize the performance of the full turbine considering either two or three three blades.
• Optimize the blade shape of monoplane Wells turbine and obtain new. followed by a discussion of the Savonius turbine and Wells turbine. .2. INTRODUCTION 2. taking into account the mutual interaction between the blades. The thesis ends with conclusions and suggestions for future work in chapter 8. 1. a considerably improved design must be available. Chapter 3 illustrates the employed optimization technique. including the coupling between the ﬂow solver and the optimizer (OPAL). The numerical methods are introduced in chapter 5. Wells turbine (wave energy conversion) 8 • Study a modiﬁed Wells turbine consisting of nonsymmetric airfoils. • Optimize the shape of the airfoil for a twostage. In chapter 2. modiﬁed Wells turbine using nonsymmetric airfoils. nonstandard airfoils with a higher performance. which are the main topic of this work.1 Conclusions and outline This chapter provides a brief introduction to renewable energy conversion and its importance for human life. comparing the new designs with the classical ones. • Obtain the optimal shape of the airfoil for a twostage Wells turbine using symmetric airfoils. • At the end. A literature review of the performance of Savonius and Wells turbines is presented in chapter 4 including previous attempts to improve the performance of both turbines. wind and wave energy conversion are discussed in details.CHAPTER 1. All the results of the optimization are presented in chapter 6 for Savonius turbine and chapter 7 for Wells turbine.
.
1 Introduction The utilization of renewable energy sources is not at all new. is therefore a secondary form of solar energy. This leads to a theoretical overall wind power of approximately 4. This only changed with industrial revolution when lignite and hard coal became increasingly important. 2. Oﬀering the advantages of easy transportation and processing.Chapter 2 Basic concepts 2. Later on.2 Wind energy conversion Solar radiation induces the movement of the air masses within the atmosphere of the earth. at least by the industrialized countries.3 1015 W. Considering this background. high hopes and expectations are placed on the multiple possibilities of utilizing renewable sources of energy. climatefriendly and socially acceptable. in the history of mankind renewable energies have for a long time been the primary possibility of generating energy. Yet. The aim of 10 . The energy contained in the moving air masses. crude oil has become one of the prime energy carriers applied today.5% are utilized for the atmospheric movement. Of the total solar radiation incident on the outer layer of the atmosphere. approximately 2. also crude oil gained importance. renewable energies are of secondary importance with regard to overall energy generation. the utilization of fossil energy carriers involves a series of undesirable side eﬀects which are less and less tolerated by industrialized societies increasingly sensitized to possible environmental and climate eﬀects at the beginning of the 21st century. This is why the search for environmental. the application of renewable energies decreased in absolute and relative terms. alternatives suitable to cover the energy demand has become increasingly important. not only in Europe. Also with regard to the considerable price increase for fossil fuel energy on the global energy markets in the last few years. which for example can be converted into mechanical and electrical energy by wind mills and turbines. As fossil energy carriers were increasingly used for energy generation. the present chapter aims at presenting the physical and technical principles for using wind and wave energy. besides a few exceptions.
Wind energy converters harness the kinetic energy contained in ﬂowing air masses. the fundamental physical principles of this type of energy conversion are explained. areas of similar wind speeds can be identiﬁed. There are two diﬀerent physical principles to extract power from wind. as in this case air ﬂow would come to a standstill.1: Distribution of yearly annual mean values of the wind speed for 10 m above ground worldwide [53]. BASIC CONCEPTS 11 the following discourse is to show the main basic principles of the supply in wind energy and to discuss its supply characteristics [53]. The airfoil drag method is based on the wind drag force incident on a windblown surface. The extracted wind power generates rotation and is thereby converted into mechanical power at the rotor shaft. with a worldwide installed capacity of 157. and consist of one or several rotor blades. Mechanical power is taken up at the shaft in the form of a moment at a certain rotation and is transfered to a machine (such as a generator). and the United States. Figure 2. a mechanical gear and a generator. The measured wind speeds can be analyzed and the annual mean value can be calculated. Asia. The implementation of an innovative aerodynamic control technique in wind turbines is a point under extensive investigation .CHAPTER 2. Both principles are outlined throughout the following sections to explain the main diﬀerences. If the yearly mean wind velocity is averaged over various years. and wind power would no longer be available. air would fail to enter the swept rotor area. Figure 2.1 shows these values on a worldwide scale referring to 10 m above ground. The second principle. which is based on ﬂow deviation inside the rotor is at present predominantly applied for wind energy conversion. Most modern wind energy converters are equipped with rotors to extract wind power. It is widely used in Europe.900 megawatts (MW) in 2009 [12]. also referred to as aerodynamic or airfoil lift principle. In the following. The entire wind power station thus consists of a wind energy converter (rotor). It is physically impossible to technically exploit the entire wind energy. Wind power is growing at the rate of 30% annually.
The airfoil crosssection at radius r is set at a local blade pitch angle ϑ with respect to the rotor plane of rotation (Fig. BASIC CONCEPTS 12 since the conventional wind turbine blade technology is reaching its limits. 2. For the beneﬁt of those readers unfamiliar with aerodynamics. the diﬀerence between the aerodynamic angle of attack α and the blade pitch angle ϑ should be noted: the angle of attack is an aerodynamic parameter and the blade pitch angle is a design parameter. A discussion on the maximum achievable wind power output by means of an ideal wind energy converter follows. According to the lift principle.1 Horizontal axis turbines These are the standard turbines used at present. The axial free stream velocity va in the rotor plane and the tangential speed u = rω at the radius of the blade crosssection combine to form a resultant ﬂow velocity vr . it forms the local aerodynamic angle of attack α. 2. The two angles are often confused.2). 2. Figure 2. increased reliability and faster control of larger wind turbines [78].CHAPTER 2.2. Almost all the eﬀort of the wind turbine industry in the ﬁeld of aerodynamics is related to the development of blades which oﬀer better performance. rotor blades are mostly designed according to the wing theory. For highspeed propellertype converters. Linking the relationships of ﬂuid mechanics for the momentum of the axial ﬂow and of the radial ﬂow components of the rotating wake with the formulations for the aerodynamic forces at the blade . making it more diﬃcult to understand the aerodynamic relationships.2: Horizontal axis wind turbine. wind is deviated to generate peripheral force inside the rotor (Fig. Together with the airfoil chord line.3).
i. in addition. and particularly the aerodynamic drag. 2. it should be kept in mind that the momentum theory as well as the blade element model include several simpliﬁcations. The design proposed in 1925 by the French engineer Darrieus. verticalaxis rotors could only be built as pure dragtype rotors (see the idea of drag turbines in Fig. It is only when the tip speed ratios become inﬁnitely high that the power coeﬃcient approaches Betz’s ideal value.CHAPTER 2.4). BASIC CONCEPTS 13 element allows the ﬂow conditions at the blade element to be determined so that the local aerodynamic lift and drag coeﬃcients can be calculated. The aerodynamic rotor theory based on the momentum theory and on the blade element theory.2. As . Referred to the power rating of the air stream. This is usually divided into two components: one in the plane of rotation of the rotor. the rotor power or power coeﬃcient. the thrust distribution. 2. In particular. At the beginning. the lower the induced drag. The higher the aspect ratio. the ﬂow around the blade tip. Taking into consideration the angular momentum in the rotor wake shows that the power coeﬃcient becomes a function of the tip speed ratio (see later section 2. a result of the pressure diﬀerence between the top and the underside of the blade. has been considered as a promising concept for modern wind turbines. Integrating the tangential force distribution over the rotor radius provides the driving torque of the rotor and. These blade tip losses are introduced as additional drag components. produces the socalled free tip vortices. a function of the local lift coeﬃcient and the aspect ratio (slenderness) of the blades. which can be found as simple ventilator on some railroad carriages or delivery vans. Introducing the aerodynamic forces acting on the rotor blades.3).2. which limit their validity to a discshaped wind energy converter. The calculation of the balance of forces includes not only the pure airfoil drag but also other drag components which derive from the spatial ﬂow around the rotor blade.e. the power coeﬃcient now exhibits an optimum value at a certain tip speed ratio. Sometimes. Integrating the thrust distribution yields the total rotor thrust for instance to the tower.2 Vertical axis turbines The oldest design of wind rotors features rotors with a vertical axis of rotation. further reduces the power coeﬃcient. however. They are derived from a complex vortex model of the rotor ﬂow [37]. and the cup anemometer used to measure wind velocity are wellknown examples of rotors with a vertical axis of rotation. which could also eﬀectively utilize aerodynamic lift. and one at right angles to it. the blade element theory provides the distribution of aerodynamic forces over the length of the blade. the tangential force distribution. as are the hub losses which are the result of vortices in the wake of the ﬂow around the hub. the more slender the blades. yields the real rotor power curve with good approximation. with the rotational speed of the rotor. With several semiempirical approaches for these vortex losses. the momentum theory is therefore called ”disc actuator theory” [37]. The Savonius rotor. respectively. It was only recently that engineers succeeded in developing verticalaxis designs. in particular. the simple momentum theory by Betz provides the ideal peak power coeﬃcient of 0..593 which is independent of the tip speed ratio. The resultant drag is called induced drag. Nevertheless.
gearbox and generator at ground level. and that there is no yaw system. straight blades connected to the rotor shaft by struts are used.3: Flow velocities and aerodynamic forces at the airfoil crosssection of a blade element [53]. It is not suitable for electricitygenerating wind turbines due to its low tipspeed ratio and its comparatively low power coeﬃcient. Hrotors of a particularly simple structure. this will still require a relatively long period of development. The speciﬁc advantages of vertical axis turbine concepts are that their basically simple design includes the possibility of housing mechanical and electrical components. Altogether.CHAPTER 2. This is countered by disadvantages such as low tipspeed ratio. were developed by a German manufacturer up until the beginning of the nineties but the development was stopped then since there was no economic success in sight. inability to selfstart and not being able to control power output or speed by pitching the rotor blades. Attempts were made particularly in the UK. simple wind rotors. Occasionally. is the case with horizontalaxis rotors. Whether the basic advantages of these designs can prevail over their disadvantages and whether it will become a serious rival to the horizontalaxis rotors cannot be foreseen for the longterm. the Savonius design is still used for small. especially for driving small water pumps. Darrieus rotors are preferably built with two or three rotor blades. the development and optimization of . BASIC CONCEPTS 14 Figure 2. A variation of the Darrieus rotor is the socalled Hrotor. Instead of curved rotor blades. might still have a large potential for development and optimization. with the permanently excited generator integrated directly into the rotor structure without intermediary gearbox. In any case. in the US and in Germany to develop this design to commercial maturity. In half of this thesis. it can be said that wind rotors with vertical axes and among these primarily the Darrieus rotor and Savonius rotor.
Between 1922 and 1925. which transforms the kinetic energy contained in the moving air into mechanical energy. Moreover. he found that optimal power extraction could only be realized at a certain ratio between the ﬂow velocity of air in front of the energy converter and the ﬂow velocity behind the converter.2. Betz published writings in which he was able to show that. which assumes an energy converter working without losses in a frictionless airﬂow. Although Betz’s ”momentum theory”. BASIC CONCEPTS 15 U Fwall R Figure 2. 2. contains simpliﬁcations. the exact nature of the energy converter is irrelevant.4: Flow conditions and drag force for vertical axis turbines [53]. its results are still used for performing ﬁrst calculations in practical engineering. But its true signiﬁcance is founded in the fact that it provides a common physical basis for the understanding and operation of wind energy converters of various . rotating wind energy converter follows its own basic rules. a new concept based on a Savonius turbine is the target. The credit for having recognized this principle is owed to Albert Betz.CHAPTER 2. For an initial discussions of basic principles.3 Betz limit The main component of a wind turbine is the energy converter. The extraction of mechanical energy from a stream of moving air with the help of a diskshaped. by applying elementary physical laws. the mechanical energy extractable from an air stream passing through a given crosssectional area is restricted to a certain ﬁxed proportion of the energy or power contained in the air stream.
P = m 2 ˙ 2 (U − U2 ) 2 1 (2. the following pages will provide a summarized mathematical derivation of the elementary momentum theory by Betz. with an unchanged mass ﬂow. the mechanical energy.CHAPTER 2. which the diskshaped converter extracts from the airﬂow corresponds to the power diﬀerence of the air stream before and after the converter: P = 3 3 ρAU1 − ρAU2 2 (2. the wind velocity before it reaches the converter. the socalled volume ﬂow rate.3. is: Q = AU and the mass ﬂow rate with the air density ρ is: m = ρAU ˙ (2. Reduced velocity. This energy is physically identical to the power P in (W): ρAU 3 P = (2.2. For this reason. this means that. Neglecting any losses. U1 is the undelayed freestream velocity. means at the same time a widening of the crosssection. 2. whereas U2 is the ﬂow velocity behind the converter.2) The equations expressing the kinetic energy of the moving air and the mass ﬂow yield the amount of energy passing through crosssection A per unit time. as the same mass ﬂow must pass through it.4) 2 The question is how much mechanical energy can be extracted from the freestream airﬂow by an energy converter? As mechanical energy can only be extracted at the cost of the kinetic energy contained in the wind stream. the ﬂow velocity behind the wind energy converter must decrease.7) .1) Considering a certain crosssectional area A. the volume ﬂow rate Q (m/s3 ) ﬂowing through during a time unit. BASIC CONCEPTS 16 designs. Here.3) (2. however. through which the air passes at velocity U.5).1 Betz’s momentum theory The kinetic energy of an air mass m moving at a velocity U can be expressed as: E= mU 2 2 (2. It is thus necessary to consider the conditions in front of and behind the converter (Fig.5) Maintaining the mass ﬂow (continuity equation) requires that: ρA1 U1 = ρA2 U2 Thus. 2.6) (2.
9) Thus.CHAPTER 2. namely when the air is brought to a complete standstill by the converter. the force which the air exerts on the converter can be expressed as: F = m(U1 − U2 ) ˙ (2. then the inﬂow velocity before the converter must also become zero. on the one hand. As could be expected. If the outﬂow velocity U2 behind the converter is zero. according to the elementary momentum theory. from the thrust and the ﬂow velocity. and. in purely formal terms. Equating these two expressions yields the relationship for the ﬂow velocity U : m 2 ˙ 2 ˙ (2. power would have to be at its maximum when U2 is zero. From this equation it follows that. the mechanical power extracted from the air ﬂow can be derived from the energy or power diﬀerence before and after the converter. U = (U1 + U2 ) 2 (2. so to speak.5: Flow conditions due to the extraction of mechanical energy from a freestream air ﬂow. Using the law of conservation of momentum.8) According to the principle of ”action equals reaction”. must be counteracted by an equal force exerted by the converter on the airﬂow. this force.11) . this result does not make sense physically. a physically meaningful result consists in a certain numerical ratio of U2 /U1 where the extractable power reaches its maximum. on the other hand. present in the plane of ﬂow of the converter. This requires another equation expressing the mechanical power of the converter. The thrust. pushes the air mass at air velocity U . BASIC CONCEPTS Rotor disk 17 U U1 U2 Figure 2. the thrust. the ﬂow velocity in the converter plane is equal to the arithmetic mean of U1 and U2 .10) (U − U2 ) = m(U1 − U2 )U 2 1 Thus. The power required for this is: P = F U = m(U1 − U2 )U ˙ (2. implying that there would be no more ﬂow through the converter at all. However.
. the ratio of the extractable mechanical power to the power contained in the air stream. without mechanical power being extracted from it.16) The power coeﬃcient.e. therefore. Knowing that the maximum. the ﬂow velocity U in the rotor plane becomes: U = 2 3 U1 (2.18) It is worthwhile to recall that these basic relationships were derived for an ideal. now only depends on the ratio of the air velocities before and after the converter. see for instance [77]). frictionless ﬂow (for the eﬀect of the friction. i.CHAPTER 2.593 27 (2.15) After some rearrangement. and that the result was obviously derived without having a close look at the wind energy converter.14) 2 The ratio between the mechanical power extracted by the converter and that of the undisturbed air stream is called the ”power coeﬃcient” Cp : P Cp = = Po 2 2 ρA(U1 −U2 )(U1 +U2 ) 4 3 ρAU1 2 (2. the power coeﬃcient can be speciﬁed directly as a function of the velocity ratio U2 /U1 : P 1 Cp = = 1− Po 2 U1 U2 2 1+ U1 U2 (2. frequently called the ”Betz factor”.17) Betz was the ﬁrst to derive this important value and it is.12) The mechanical power output of the converter can be expressed as: 2 2 ρA(U1 − U2 )(U1 + U2 ) (2. This power was: P = 3 ρAU1 Po = (2. it is compared with the power of the freeair stream which ﬂows through the same crosssectional area A. In real . ideal power coeﬃcient is reached at U2 /U1 = 1/3. therefore.13) 4 In order to provide a reference for this power output. BASIC CONCEPTS The mass ﬂow thus becomes: m = ρAU = ˙ ρA(U1 + U2 ) 2 18 (2. The maximum ”ideal power coeﬃcient” Cp becomes Cp(max) = 16 = 0. If this interrelationship is diﬀerentiated to get the maximum value of the power coeﬃcient it can be obtained that the power coeﬃcient reaches a maximum at a certain velocity ratio with U2 /U1 = 1/3.
• Even with an ideal airﬂow and lossless conversion.593. • When the ideal power coeﬃcient achieves its maximum value (Cp = 0. it thus increases with the square of its diameter. • The power increases linearly with the crosssectional area of the converter traversed. BASIC CONCEPTS 19 Figure 2.593).6: Power coeﬃcients of wind rotors of diﬀerent designs [37]. 2. The essential ﬁndings derived from the momentum theory can be summarized in words as follows: • The mechanical power which can be extracted from a freestream airﬂow by an energy converter increases with the third power of the wind velocity. as shown in Fig. the power coeﬃcient will always be smaller than the ideal Betz value. Hence. optimizing of a Savonius wind turbine will be considered. cases. the wind velocity in the plane of ﬂow of the converter amounts to two thirds of the undisturbed wind velocity and is reduced to one third behind the converter. . the ratio of extractable mechanical work to the power contained in the wind is limited to a maximum value of 0. only less than 60% of the wind energy of a certain crosssection can be converted into mechanical power.2.4 Principle operation of standard Savonius turbines In half of this thesis.6.CHAPTER 2. 2.
ratio of the aerodynamic power of the turbine to the power of the incident wind. 2. 2.6 gives the power coeﬃcient Cp . as a function of the speed ratio λ.7: Conventional Savonius rotor. the mechanical power P and the mechanical torque on the axis of a Savonius turbine can respectively be written as follows: P (2.6 show that the fast running horizontal axis wind machines (two. Consequently.7). 2.6. a wind of incoming velocity U . The curves in Fig. Figure 2. Figure 2.CHAPTER 2.8. it can present some advantages for speciﬁc applications.2. Savonius initially developed the vertical axis Savonius rotor in the late 1920s. This speed ratio λ is also called velocity coeﬃcient and is equal to the ratio of the tip peripheral speed to the wind velocity. And there is room for improvement! S.15 to 0.19) Using the notations of Fig. 2. theses machines are nowadays systematically chosen for the equipment of largearea wind sites. To support the discussion. On the other hand.20) Cp = ρRHU 3 .0) has a rather poor eﬃciency: Cp 0. Nevertheless. The power coeﬃcient is directly linked to the global eﬃciency of a wind machine. which is a slowrunning vertical axis wind machine (λ 1. the Savonius rotor. BASIC CONCEPTS 20 The choice of a wind turbine is obviously based on its performance in connection with the local wind conditions.J. the velocity coeﬃcient is deﬁned as: For a Savonius rotor of height H.5 Performance of a Savonius turbine λ = ωR/U (2. 2.or threebladed airscrew) have incontestably the best eﬃciencies. so that the crosssection resembles the letter S [36] (Fig. a comparison between the characteristics of the main conventional wind turbines is shown in Fig. and resulting robustness and low cost. in particular due its simplicity. The concept of the Savonius rotor is based on cutting a cylinder into two halves along the central plane and then moving the two half cylinders sideways along the cutting plane.2 at best [68].
with energy ﬂuxes commonly averaging between 50 and 70 kW per meter width of oncoming wave. Such devices should be economic to replace dieselgenerated electricity. a rotor is called a conventional Savonius rotor if the geometrical parameters a and e Fig. especially on islands [134]. and there are countless ideas for machines to extract the power. As with all renewable energy supplies.21) where Cp and Cm are respectively the power coeﬃcient and the torque coeﬃcient. 2. so the research and smallscale development has progressed to the stage of commercial construction for power extraction. . Very large energy ﬂuxes can be found in deep water sea waves. In the following sections. the year of socalled oil crisis [5]. 2. BASIC CONCEPTS 21 Figure 2. the long period (≈10 s) and large amplitude (≈2 m) waves have considerable interest for power generation.8 are respectively equal to 0 and d/6. This reference conﬁguration of the rotor has been extensively studied by many groups (see citations in [70] and next chapter). However. the scale of operation has to be determined and present trends support moderate power generation levels at about 1 MW from modular devices about 50 m wide across the wavefront.8: Schematic description and main parameters characterizing a Savonius rotor. Britain and Scandinavia. The utilization of the sea wave was rarely considered on a practical scale prior to 1973. and Cm = T ρR2 HU 2 (2. The possibility of generating electrical power from these deep water waves has been realized for many years.CHAPTER 2. Therefore. This conventional design will be the starting point for the optimization process described in later chapters. a great deal of research has been conducted since 1973.3 Wave energy conversion Let us switch to the second focus of this document: wave energy conversion with optimized Wells turbines. In recent years interest has revived. The power in the wave is proportional to the square of the amplitude and to the period of the motion. particularly in Japan.
they would experience a less powerful wave regime. This is a 2 MW system. Figure 2. the activities will be divided between the technologies suitable for deployment on the shoreline. 2.e.3. near the shore and oﬀshore [133].. this system aims to maximize the amount of energy produced from a given amount of near shore area. BASIC CONCEPTS 22 2.9). which has the advantage of easier maintenance and/or installation. • Oﬀshore devices This class of device exploits the more powerful wave regimes available in deep water (> 40 m depth) before energy dissipation mechanisms have had a signiﬁcant eﬀect. • Near shore devices The main prototype device for moderate water depths (i. • Shoreline devices These devices are ﬁxed to or embedded in the shoreline itself. .5 MW wind turbine.1 Wave energy potential There is a large amount of ongoing work on wave energy due to a broad availability (Fig. In addition these would not require deep water moorings or long lengths of underwater electrical cable. which cannot be done justice in a brief overview. < 20 m depth) is the OSPREY developed by Wavegen.CHAPTER 2. However. Since there could be environmental objections to large farms of wind or wave energy devices close to the shore. with provision for addition of a 1.9: Approximate global distribution of wave power levels in kW/m of wave front [133]. For ease of presentation.
CHAPTER 2. because the construction cost of the total system can be jointly born by the accounts for power generation and harbor protection.3. 3. Some of these devices being developed are placed underwater. 2.3 Wells turbines Serious research on waveenergy extraction methods began in several countries during the 1970’s. Dr.10. It was summarized by Twidell and Weir [134] as follows: 1. anchored to the ocean ﬂoor. The water column within the lower half of the caisson is caused to oscillate vertically by incident waves through the opening. The following section gives an idea of the most famous mechanism to convert the wave energy to mechanical energy. provides the simplest and possibly the most reliable means of converting slow irregular wave motion into high speed rotational movement required for electrical power generation. and it induces the compression and expansion of air mass within the upper half of the caisson as shown in Fig. 2. the oscillating water column device. it is important to appreciate the many diﬃculties facing wave power developments. Wave periods are commonly from 5 up to 10 sec (frequency ∼ =0. 2. Wells. The structure of the power devices must be able to withstand this.1 Hz). while others ride on top of the waves. An air turbine coupled to an electric generator is connected to the nozzle. There is always probability of extreme gales or hurricanes producing waves of freak intensity. A combination of OWC wave power extractor and breakwater is also attractive from the viewpoint of economical feasibility of wave power extraction. a former professor of civil engineering at Queen’s university of . called Oscillating Water Column mechanism (OWC). Many attempts have been made to construct such devices and eﬃciently match variable natural conditions. The air motion generates a highvelocity ﬂow through the nozzle.10). 2. which activates the air turbine and generates electricity [73]. extract the wave power and withstand the listed diﬃculties. Wave patterns are irregular in amplitude.3. It is extremely diﬃcult to couple this irregular slow motion to electrical generators requiring about 500 times greater frequency. because the latter is mostly built with a large concrete caisson rested on the foundation. The wave power is thus converted into air power in the air chamber of the wave powerextracting caisson.2 Oscillating Water Column (OWC) principle One of the many extraction systems proposed in recent years. phase and direction. It is diﬃcult to design devices to extract power eﬃciently over this wide range of variables. OWC wave power extractors can be quite easily ﬁtted to a vertical breakwater. The device is essentially a caisson rested on sea bottom with a large submerged opening at the front and a small nozzle at the ceiling (Fig. BASIC CONCEPTS 23 However. 2.
the Wells turbine has been widely .CHAPTER 2. BASIC CONCEPTS 24 Figure 2. Belfast.10: The wave motion in the OWC device [73]. Because of its simple and eﬃcient operation. In its simplest form the air turbine rotor consists of several symmetrical airfoil blades positioned around a hub. proposed in 1976 a form of selfrectifying axial ﬂow air turbine as a device suitable for wave energy conversion using the oscillating water column.
This device is located on the Isle of Islay.12. 2. 2. 25.12. According to the standard airfoil concept. 86. FT is the same for both positive and negative values of angle of attack (±α) as shown in Fig. 75. 90. 97]. if the airfoil is set at an angle of attack α in a ﬂuid ﬂow. In the next chapter.CHAPTER 2. The water wave energy is converted to pneumatic energy in the air. The ﬁrst prototype was constructed in UK in 1988. the direction of tangential force.11). 138]. they will rotate in the tangential force direction regardless of the direction of airﬂow. as shown in Fig.3. it will generate a lift force FL normal to the free stream and a drag force FD in the direction of the free stream. 2. 2. 2. 92. one of the southern islands in the Inner Hebrides and depends on a Wells turbine as ﬁnal converter [15. Oscillating airflow Air turbine Figure 2. 22. 2. Therefore. 137]. BASIC CONCEPTS 25 applied for ocean waves energy absorption. 84. For a symmetrical airfoil as considered in [84.5 Wells turbine performance If such symmetric airfoil blades are positioned around an axis of rotation. These lift and drag forces can be combined to get the tangential force FT and the axial force as shown in Fig.3. which passes periodically across a selfrectifying. .12. the state of the art of Wells turbines used as converter for wave energy will be proposed. axial air ﬂow turbine. it has been subjected to a considerable amount of research and development in many counties.4 Operation of Wells turbines Most research programs attempting to gain energy from waves depend on the OWC as converter mechanism [7. 88. The turbine itself consists in a number of symmetric airfoils set around the hub radially at 90◦ stagger angle. The force FT is responsible for the torque and consequently the blade power. with the chord plane normal to the axis of rotation (Fig. 16.11: Wells turbine in state of upward and downward air ﬂow.
The tangential force coeﬃcient CT and the axial force coeﬃcient CA are calculated as : 2 CT = FT / (1/2)ρ vA + u2 zbc (2. ω is the rotor angular velocity and rt is the tip radius (Fig. 2. 2. FT = FL sin(α) − FD cos(α) FA = FL cos(α) + FD sin(α) (2. This leads to a unidirectional device rotation for an alternating airﬂow without the need for nonreturn valve.CHAPTER 2.12 can be obtained from: FT FA R Rotation w vA ut Flow Flow ut vA w Rotation FA R FT Figure 2.25) where ut = ωrt is the peripheral velocity. which has to be absorbed by the bearings. . axial force coeﬃcient CA and eﬃciency η with ﬂow coeﬃcient φ. BASIC CONCEPTS 26 while the axial force FA results in an axial thrust along the axis of the rotor. The tangential force FT and the axial force FA shown in Fig.12: Axial and tangential forces acting on a Wells turbine.13).22) (2. Furthermore.23) The running characteristics under steady ﬂow conditions are usually characterized by the tangential force coeﬃcient CT .24) t 2 2 CA = ∆p0 πrt / (1/2)ρ vA + u2 zbc t (2. vA is the axial velocity normal to the plane of rotation.
as given by [84. blade prof ile shape) (2. Airfoil Blade sp an (b ) rt Rotation Chord (c) FA FT Oscillating flow FA Forces resolved in direction of rotation Figure 2. The nondimensional variables expressing the performance of a Wells turbine are function of the aerodynamic force coeﬃcients that are in turn function of several parameters. CA ) = f (φ. h is the hub to tip ratio (rh /rt ). 2. τ. z is the number of blades. Tu . 2. its is now time to analyze the state of the art for both systems considered in this thesis: Savonius turbine (wind energy) and Wells turbine (wave energy). Re.4 Conclusions After having introduced all needed concepts and notations. AR is the aspect ratio (b/c). h. f ∗ . Tu is the turbulence level.13: Main geometrical parameters of a Wells turbine.26) where. τc is the tip clearance ratio. c is the blade chord (see Fig. f ∗ the frequency of wave motion and the Reynolds number Re. s. φ = vA /ut is the ﬂow coeﬃcient. b is the blade span.13) and ∆p0 is the total pressure diﬀerence across the rotor. AR.CHAPTER 2. c is blade chord. 92]: (CT . b is blade span. BASIC CONCEPTS Rotor hub on shaft 27 rh Sym. . τc . τ is the blade thickness. s is the rotor solidity.
1 Experimental investigations of Savonius turbines Some studies have been carried out in wind tunnels. this lack of quantitative information is sometimes associated to the complexity of the ﬂow in and around wind turbines. 28 . The present state of Savonius turbines is ﬁrst considered.1 Savonius turbine Savonius turbines show inherent drawbacks compared to conventional wind turbines: mainly a low eﬃciency and poor starting characteristics. In particular for Savonius turbines. 51] are of higher quality and give a precise description of the aerodynamics of the conventional Savonius rotor. Such publications are of very limited scientiﬁc or technological interest and are therefore not included here. before reviewing the literature on Wells turbines. As a consequence. some visualizations of the ﬂow in and around the rotor are proposed. Therefore. some publications [10. the resulting ﬂow conditions are highly unsteady. 27. but with a poor description of the physical phenomena. However. identical to or derived from the conventional Savonius rotor.Chapter 3 Literature Review Due to the growing signiﬁcance of wind and wave energy conversion. 3. mainly obtained by pressure measurements on the paddles. many authors have tried to identify the best principles of operation and to improve the characteristics of Savonius turbines. a wealth of publications can be found for the systems considered in this thesis. Sometimes. is presented in such studies. boundary layer separation is an essential aspect for the eﬃciency of the system. using controlled conditions. The present chapter is again divided in two parts. but without realizing any detailed. detailed aerodynamic studies are rare and often do not allow the prediction of the energetic behavior of the rotor. Beyond issues associated with intellectual properties. 31. the global performance of a rotor. quantitative parametric study. Furthermore.1. 3. Generally. Other consider freespace experiments.
b: ﬂow inside the rotor. Through the corresponding rotor power output coeﬃcient versus the tip speed ratio. Such parametric studies already lead to geometrical conﬁgurations with a considerable increase in the rotor performance eﬃciency [30. These phenomena.1: Flow in and around a nonrotating Savonius rotor (λ = 0). The former contributes to the torque production of the rotating rotor while the latter acts as a resistance. d: surface pressure distribution [29]. 142].2). contribute to the power producing mechanism of the Savonius rotor [4. LITERATURE REVIEW 29 Further articles describe an extensive experimental study in a windtunnel to evaluate the importance of geometrical parameters on the Savonius rotor performance. 39.1 and 3. the inﬂuence of each blade geometry parameter is investigated. . together with the ﬂow stagnation region on the front side of the rotor. a: visualized ﬂow ﬁeld. Similarly. 54. the ﬂow through the overlap is weakened by the appearance of resisting ﬂow. 128. Figure 3. It is observed that the ﬂow separation regions on the blade surfaces are fairly reduced by the rotation eﬀect (compare Fig.CHAPTER 3. The ﬂow in and around a Savonius rotor has also been studied using ﬂow visualization experiments in combination with the measured pressure distributions on the blade surfaces. c: ﬂow model. 29]. 3.
On the other hand. 3. The attached ﬂow on the convex side tends to separate at large rotor angles (θ = 90◦ to 135◦ ).1). a: visualized ﬂow ﬁeld.2a) on the convex side.2d). the stagnation point moves to the . 3. which increases in size downstream. Signiﬁcant features observed here are the downward movement of the separation point (Fig. c: ﬂow model. b: ﬂow inside the rotor.CHAPTER 3.2: Flow in and around a Savonius rotor in rotation (λ = 0. In addition. which is due to the outward ﬂow motion at the tip of the advancing blade. since the relative velocity is decreased on the advancing blade and is increased on the returning one. as is expected from the pressure distributions (Fig. a relative decrease in the stagnation torque is expected here in comparison with the nonrotating rotor (Fig. d: surface pressure distribution [29]. The injected ﬂow grows into a vortex circulating in the rotating direction of the rotor. LITERATURE REVIEW 30 Figure 3.2d). 3. 3. It is considered that the attached ﬂow patterns of the rotating rotor contribute to the rotating torque of the rotor.9).2c) and the relative decrease in the pressure coeﬃcient on the convex side of the advancing blade (Fig. These phenomena can be caused by the occurrence of a Coandalike ﬂow pattern (Fig. 3. This ﬂow is induced by the pressure gradient distributed over the concave side of the advancing blade. which appears clearly at small rotor angles of θ = 0◦ to 45◦ .
This ﬂow is expected to reduce the pressure recovery eﬀect on the back side of the returning blade. and the potential for improvement in eﬃciency. The results show that the power per unit length provided by the considered Savonius rotor is about four times as high as that provided by a fastrunning two bladed airscrew. It can be seen that the pressure coeﬃcients are decreased overall by the eﬀect of circulation produced by the rotor rotation.. e. a gap between the rotor buckets is eﬀective in increasing maximum power. In terms of the ease of setting up.1. This criterion consists in comparing wind turbines which intercept the same front width of wind. The American windmill and the Savonius rotor have comparable values of their associated power per unit length. These calculations are continued until the residual values (variations in certain chosen parameters. by allocating them a same reference value of the maximal mechanical stress on the blades or the paddles. From our experience. A separate study has been carried out to verify the model accuracy. Such a circulation is a steady phenomenon. LITERATURE REVIEW 31 center of the blade due to the rotation eﬀect. Some articles point out that the choice of a wind turbine must not be based only on high eﬃciency and proposed a comparative criterion adapted to the comparison of a horizontal axis wind turbine with a vertical axis wind turbine: the Lσ criterion. These studies include static or dynamic modelling. have all dropped below 10−3 (criterion for convergence). the velocity coeﬃcient was equal to 1. a quantitative comparison points to a clear advantage of the Savonius rotors. in this second case.6). which is supported by the measured pressure distributions near the overlap. A favorable factor to obtain a high power per unit length is consequently a low angular velocity. They suggested that the reason why so few numerical studies had been successful was due to the complexity of the ﬂow pattern about the rotor and to the separation of the ﬂow from the blade surfaces. Usually. A few of these papers [23. 28. this is a relatively weak and perhaps insuﬃcient criterion. 2. 49] used the discrete vortex method to predict the ﬂow around a pair of coupled Savonius rotors.2 Numerical investigations of Savonius turbines Numerical simulations have also been carried out on this kind of rotors. Nevertheless. and provides some elements for the improvement of their rotor [71]. The results show a good agreement with experimental performance for the following points: the torque grows weaker in inverse proportion to the tip speed ratio. These calculations have been realized using a static approximation (the rotor is supposed to be ﬁxed whatever the wind direction) and also a dynamic calculation. comparing a . wind sites are equipped today with fastrunning horizontal axis wind turbines of the airscrew type.0 (nominal working point in accordance with Fig. most results suﬀer from a too crude description of the rotor. because of their lower angular velocity. 3.g. On the basis of this criterion. velocity in the wind direction). the ﬂow through the overlap is reduced here by the production of resisting ﬂow. the Savonius rotor should be preferred to all other considered conﬁgurations following [71]. In comparison with the nonrotating rotor. Flow ﬁelds around rotating Savonius rotors have been also simulated by solving the twodimensional incompressible NavierStokes equations [95].CHAPTER 3. associated to a high eﬃciency.
These numerical results were compared to experimental data. 3.8). Figure 3. except for angles θ around 0◦ and 180◦ where an instability of the torque is observed (Fig.56 105 . 2. To do so. λ = 1) [68]. no central shaft) has allowed to determine the pressure distribution on the paddles. LITERATURE REVIEW 32 static simulation and a dynamic one.3. a static simulation of the ﬂow around the conventional Savonius rotor (e/d = 1/6. a = 0. a dynamic calculation (rotating turbine) has been carried out for the same value of the Reynolds number: Re= 1. In a second step. The simulations give satisfactory results since the diﬀerences between the experimental data and the numerical simulations are always below 10%.56 105 ) [68].CHAPTER 3. [68]).56 105 . Then the static torque has been calculated as a function of the wind velocity angle θ (Fig.3: Static torque coeﬃcient of a Savonius rotor (Re = 1.4: Dynamic torque coeﬃcient of a Savonius rotor (Re = 1. setting the velocity coeﬃcient . Figure 3.
Guide vanes can be used to increase the static torque and decrease its ﬂuctuation. so that the maximum values of Cm and Cp . 70].3. The results indicate that the static torque coeﬃcient of the one stage turbine without guide vanes is periodic with a cycle of 180 degrees (depending on the number of stages) and that its variation is very large.1. vorticity. 3.2 Double and three steps Savonius rotor When considering only the starting torque.4). The behavior of the shed vortices has been observed carefully. the dynamic eﬀect depends on the wind speed.1. 3. The results are compared to the ones given with the static calculation.3 Guide vanes In order to decrease the torque variation of the standard Savonius rotor and to improve its starting capability. but lower for both torque and power characteristics [38. 50. it can be seen that. 3. pressure. has also been investigated [39]. 3. 43]. for some directions of the wind velocity. The doublestep and threestep rotors are said to be slightly superior to the corresponding singlestep turbine (conventional Savonius rotor) in selfstarting.1. unlike the static characteristic. The simulated ﬂows have been analyzed by visualizing relative velocity. 3. 3. The torque coeﬃcient Cm has been evaluated by calculating the average value of the torque on a whole revolution of the rotor. and it has been clariﬁed that the shed vortices have a large eﬀect on the resulting ﬂow ﬁelds and on the global performance [49. The lower values of static torque are observed in the ranges of θ = 140◦ − 170◦ and 320◦ − 350◦ . LITERATURE REVIEW 33 equal to 1.3. the starting torque of the standard Savonius conﬁguration would be so low that the rotor could not start alone.6) and threestep Savonius rotor.CHAPTER 3. using three stages with 120 degree bucket phase shift between the adjacent stages with and without guide vanes. respectively at 120◦ for the three step rotor. 3.2).1 Deﬂector plate Attempts to improve the performance of a Sshaped Savonius rotor by using a deﬂecting plate placed in front of the rotor have been documented.3.3 Methods to improve Savonius turbine performance Several propositions can be found in the literature to improve the performance of conventional Savonius rotors. and must therefore be considered with caution.1. a new type of Savonius turbine.1 and 3. where the upper and the lower paddle pairs are set at 90◦ to each other (double step). 68]. This is clearly in contradiction with the experimental studies discussed previously (see again Fig. Using a deﬂecting plate placed on the retreating side of the blade it has been observed that the power coeﬃcient can be increased relatively by about 20% [27. But. The diﬀerence between the two curves generally does not exceed 2% whatever the angle θ (Fig.0. etc. It is the reason why many authors have chosen to use a doublestep (see Fig.
2) where V1 is the testsection inlet air velocity and u is the tip peripheral blade velocity.3).CHAPTER 3. and its performance has been compared with conventional semicircular blades (corresponding to a twist angle of 0◦ ).1. Experimental evidence shows the potential of the twisted blade rotor in terms of smooth running. but lead to a decrease in the torque for larger values (λ > 0. A twisted blade integrated within a threebladed rotor system has been tested in a low speed windtunnel. Semicircular blades correspond . the results are not analyzed further. As a consequence the computed power is the power available in the incoming wind and not the mechanical power of the turbine. 98]. Consequently. 3. static torque and rotational speed. Unfortunately. one study was published concerning the three bucket Savonius rotor [36].5: Doublestep Savonius rotor [68].1) and T = 60Protor /2πN (3. Recently. 3.3.3). The guide vanes increase the dynamic torque coeﬃcient for small λ values (0 < λ < 0. of the threestage rotor are much smaller than those of the onestage rotor. higher eﬃciency and selfstarting capability as compared to that of the conventional rotor [94.4 Twistedblade Savonius rotor Another investigation aims at exploring the feasibility of a twistedbladed Savonius rotor for power generation (Fig. the authors rely on an erroneous formulation to compute the rotor power and torque. Performance analysis has been made on the basis of starting characteristics. LITERATURE REVIEW 34 Figure 3.6). They consider: Protor = (1/2ρAV12 )u (3.
Such blades shows a maximum of Cp = 13.65 (i. The guidebox tunnel is like a rectangular box used as wind passage. It is also shown that a larger twist angle is preferable for a lower wind velocity in order to produce maximum power and better starting characteristics.99 at tip speed ratio of λ = 0. in which a test rotor is included as shown in Fig. the performance of the Savonius rotor is increased in its performance as shown in Fig. Figure 3. 3. A twist angle α = 15◦ gives optimum performance at low airspeeds of U = 6. Figure 3.e. whereas the semicircular blade α = 0◦ shows a Cp = 11.5 Savonius rotor using a guidebox tunnel This publication aims at improving and adjusting the output power of a Savonius rotor under various wind power. LITERATURE REVIEW 35 to zero twist angle..8.23 m/s).1.6: Twistedblade Savonius Rotor [98]. By increasing this angle. 3. 3.04 at λ = 1.3.CHAPTER 3. at U = 8. A guidebox tunnel is employed as the appropriate device to achieve this purpose.5 m/s in terms of starting acceleration and maximum noload speed.7: Power coeﬃcient for twistedblade Savonius rotor [98]. The area ratio between .7.
Overall.7. since the new rotor induces maximal values of the static torque much higher than those obtained with the conventional rotor. Further studies are necessary to reﬁne the analysis. The resulting value of the output power coeﬃcient of the rotor with guidebox tunnel using an area ratio of 0.43 increases considerably (by a factor 1.3. 3.1. The resulting increase in performance seems promising. Nevertheless.8: Savonius rotor using a guidebox tunnel [48]. The results are relatively encouraging..CHAPTER 3. Figure 3. i.5 using three blades.9 shows that the maximum rotor rotational speed is obtained for a guidebox area ratio between 0. which is just a modiﬁcation of the Savonius rotor. wind angle) are presented in Fig. The experiments measure the static torque of the ﬁxed rotor at any phase angle and the dynamic torque under rotation. LITERATURE REVIEW 36 the inlet and exit of the guidebox is variable in order to adjust the inlet mass ﬂow rate or input power. it also introduces low and negative values of the torque. At ﬁrst.48. 3. 3.10) has also been proposed. 3.4. the secondary overlap e . and the angle β between the paddles. The characteristic curve of such a rotor (values of the static torque coeﬃcient Cm vs.6 Modiﬁed Savonius rotor The conventional Savonius rotor is made from two vertical halfcylinders running around a vertical axis.3 and 0. but the guidebox of course increases considerably the system complexity [48]. experiments have been conducted to ﬁnd the conﬁguration that would provide the best relative performance. by a factor 1. .23 using two blades) compared to the conventional design without guidebox (Fig. A modiﬁed rotor (Fig. 60% more than for the conventional rotor. using now three geometrical parameters: the main overlap e. The optimum spacing ratio between the rotor tip and the side walls of the guidebox tunnel is around 1. the mean value of the torque is increased: Cm = 0. with a large angular variation.e.11. Figure 3.9).
12.1: Summary of Savonius Design Gain Deﬂector Plate 20% MultiSteps Good selfstarting Guide Vanes Depends on wind speed Twistedblade 27% GuideBox Tunnel 50% (3 blades) Modiﬁed Savonius 60% turbine main modiﬁcations Description Comments Not veriﬁcated No details since 1992 For same aspect ratio Lower Cp by 30% Bad for large λ Good stability High cost Good selfstarting capability Complex design Not practical Large vibrations Not practical 3. Table 3. by means of a tedious trialanderror analysis. All the improvements have been tested manually. LITERATURE REVIEW 37 Figure 3. All proposed modiﬁcations have been summarized and discussed in Table 3. Such studies are going on.CHAPTER 3. 3.9: Power coeﬃcient of Savonius rotor using a guidebox tunnel [48]. their eﬃciency can be indeed considered as quite high [71]. as shown in Fig. It can be ﬁnally concluded that: • There is a renewed interest for Savonius turbines in recent years.1. .1. • Depending on the retained deﬁnition. considering either the conventional geometry or slight modiﬁcations.4 Summary of Savonius turbine review All the research projects discussed in the present review have tried to understand and improve the performance (torque and eﬃciency) of the Savonius turbine.
• Many small modiﬁcations and improvements have been proposed in the literature to improve global eﬃciency or some speciﬁc characteristics (e. • Contradictory observations can be found in the literature. LITERATURE REVIEW 38 Figure 3.g. • Nevertheless.1 . the results presented in this review must all be considered with great caution.CHAPTER 3. Therefore. both qualitatively and quantitatively.11: Static torque for the modiﬁed Savonius rotor (θ = 90◦ . a real optimization of the system has not been realized up to now and would therefore be very interesting. .10: Modiﬁed Savonius Rotor [69]. Figure 3. β = 45◦ . selfstarting conditions) as summarized in Table 3.. e/d = 1/6) [69]. This will be the subject considered in later chapters of this thesis.
2.5 and s = 0.62 for two values of rotor solidity (s = 0. In the next sections we will discuss in details the diﬀerent parameters aﬀecting the performance of Wells turbines. we should note that the performance of the wave power converter depends on the energy absorption eﬃciency of the OWC. . the turbine eﬃciency and the operating range of the turbine. for both low or high rotor solidity.CHAPTER 3. exists. An unidirectional airﬂow test rig [82] constructed to investigate experimentally the eﬀect of the ﬂow coeﬃcient φ. and then decreases. 63. It also indicated that the blade eﬃciency for blade proﬁle NACA 0021 increases with the ﬂow coeﬃcient up to a certain value. as well as to the turbine eﬃciency.75). The performance of Wells turbine includes the power output. at h = 0. 141] concern the principle of operation and factors controlling the performance of Wells turbines both experimentally and theoretically.1 Performance parameters of Wells turbine Many papers cited in the literature [9. a linear relationship between the pressure drop across the turbine and the ﬂow coeﬃcient φ. LITERATURE REVIEW 16 39 Number of papers 12 8 4 0 1970 Oil Crisis 1980 1990 2000 2010 Years Figure 3. which is closely related to the pressure diﬀerence across the turbine. The results indicated that.12: Savonius turbine publication statistics in international journals and conferences. 3. This is due to separation of the ﬂow around the turbine blades. 21. on both the pressure drop coeﬃcient and the eﬃciency of a monoplane turbine and in same time a theoretical investigation was made. the pressure drop across the rotor. 3.2 Wells Turbine Let us switch now to the second conﬁguration discussed in this work. Here.
45 in [73]. These eﬀects are compounded by the fact that the blades near the hub are always at a larger incidence that at the tip.6. These interactions lead to endwall losses on the surface of the hub and also on the blade surfaces near the hub. Admittedly based on smallscale tests. 3. . The reduction in eﬃciency for a Wells turbine at high solidity is due to increased kinetic energy losses at the exit associated with swirl. From our personal experience the eﬀect of hub to tip ratio on the turbine eﬃciency is rather complex. 110]). s > 0. Therefore. In Fig.6 are recommended for optimal design. the stall occurs earlier in hub region than in the tip region. 140]. η is normalized with respect to the corresponding two dimensional single airfoil eﬃciency.67. solidity is constant and equal to 0. To obtain selfstarting capability of the turbine at hub to tip h = 0.1. In this work. which leads to higher incident angles locally. 3. The results indicated that the eﬃciency for diﬀerent proﬁles decreases with increasing the hub to tip ratio. see Fig. It is also a measure of the mutual interaction between the blades and is an important design variable that aﬀects the selfstarting of the turbine [82. 118. s = zc/πrt (1 + h) (see Fig. which corresponds to a number of blades equal to 8. 97]. the blades are close to each other and may interact with the boundary layer on the hub.13) was studied in several papers [82. 93.1 Solidity 40 The solidity of the turbine.2 Hubtotip ratio The eﬀect of the hub to tip ratio (h = dh /dt or rh /rt . 3.2. publications suggest that values of h ≈ 0.1. Airfoil Blade sp an (b ) rt Rotation Chord (c) FA FT Oscillating flow FA Forces resolved in direction of rotation Figure 3. LITERATURE REVIEW 3.2. at high solidity. This is due to the interaction between the blades in hub region. 90. The results indicate that at small values of solidity. 90.13: Axial and tangential forces acting on a Wells turbine. 3. ηo .14 the eﬃciency.51 is needed in [89] but only s > 0. 97.CHAPTER 3. Rotor hub on shaft rh Sym.13) is a measure of airﬂow blockage within the turbine. its impact is small. there could be signiﬁcant three dimensional eﬀects near the hub where. In addition. This solidity value has been retained since it has been proved in a number of studies to be highly suitable for Wells turbines (for instance [106.
97]. again limited to smallscale tests.1. 92. 3.1. Several groups studied the eﬀect of tip clearance on stall and eﬃciency of the Wells turbine.2.8 0. The data shown here are from experiments conducted at a ﬁxed hub to tip ratio and ﬁxed solidity where the variation in the aspect ratio was obtained by varying the chord length. 3.The results indicated that decreasing the tip clearance advances the stall point but increases the eﬃciency of the turbine. The eﬀect of aspect ratio is however not clear in this study due to varying Reynolds number.6 0. On the other hand. AR = span(b)/chord(c). 3. it .5 Tip clearance The tip clearance ratio (τ = tip clearance/chord(c)) is a very important parameter for the performance of turbomachines.15. 97].4 0.4 Reynolds number A Wells turbine is very sensitive to the Reynolds number (Re= wc/ν) like all conventional turbomachines.2.0 41 0.14: Solidity eﬀect on the eﬃciency [82.1. due to the reduced leakage losses.2.6 0. as can be seen in Fig. due to the impact of Reynolds number on the aerodynamic around the airfoils of the turbine and to its contribution to stall [97]. inﬂuences the turbine eﬃciency and ﬂow ratio at which the turbine stalls. associated with the positive eﬀect obtained on the blades due to increased mass ﬂow through the tip [82.CHAPTER 3. there is a certain inﬂuence of the Reynolds number on this data. Therefore. suggest that aspect ratio and Reynolds number play a dominating eﬀect on the eﬃciency.3 Aspect ratio The aspect ratio.4 0.8 s 1. The primary eﬀect of reducing the aspect ratio is to increase the eﬃciency by delaying stall. LITERATURE REVIEW 1.2 0 0 0. if the turbine has a large tip clearance. This means that Re= wc/ν changed during the test. These results. 3.0 Figure 3.2 0.
65 Re 5 3x105 3x10 42 0.2 0. 82.1 AR Figure 3. The published results indicated that NACA 0021 airfoil proﬁles (21% thickness) lead to the best performance for conventional Wells turbines [93]. such as distortion of velocity proﬁle and turbulence levels.7 0.15: Aspect ratio eﬀect on the eﬃciency [82. since they contribute together to separation on the airfoil.1 0 0.2.1.6 Inlet turbulence Turbomachines are sensitive to the inlet ﬂow conditions. LITERATURE REVIEW NACA 0015 NACA 0015 S 0. operates without stall.5 0. 129. 91.7 0.9 1.6 0. The eﬀect of airfoil thickness on the aerodynamic performance can not be separated from the eﬀect of the Reynolds number. As a whole the Wells turbine is less sensitive to inlet turbulence compared to most conventional turbomachines. 46.7 Blade shape The airfoil thickness is very important because it determines the aerodynamics around the airfoil. 135]. 3.5 0. 97].4 0.2 0. the stall point and the turbine weight.0 0. 92.67 0.3 0.3 0.1. Therefore. 130.51 h 0. Increase in turbulence levels can alter boundary layer development by advancing transition of the boundary layer and delaying stall. the recommended tip clearance is τ ≤ 2% [45.CHAPTER 3. 97]. Thick airfoil blades are advantages to improve self starting characteristics of the turbine [46. Experimental results indicate that the performance of the Wells turbine improves with an increase in turbulence but a signiﬁcant increase in turbulence levels is required to produce any appreciable improvement in performance. .4 0.6 0.7 0. The performance of turbomachines can be improved by increasing the turbulence level at the inlet up to 3% [82.8 0. There is no signiﬁcant advantages to increase tip clearance above 2%.2. 3. 97].
The intensity of the vortical ﬂow varies in the accelerating and the decelerating ﬂow process. 59. . Helicopter. 3. the shed vortices are in the same direction of the blade circulation.2 Unsteady ﬂow eﬀect on the performance of Wells turbine It is generally accepted that the airﬂow frequency in a wave energy device is so small (f ∗ < 1 Hz) that dynamic eﬀects are negligible. The ﬂow structure around the blade of Wells turbines has been checked by some researchers. This observation can be explained. Then. The stronger vortices are shed at a larger radius because the blade circulation increases more than at a smaller radius. 97] (Fig. the clockwise vortical ﬂow is intensiﬁed by these vortices. Dynamic stall is a process resulting from a series of events. A strong downward ﬂow (4) is induced by the separation vortex near the trailing edge. Since the stronger vortical wake ﬂow enlarges the separation on the suction surface of the adjacent blade. In the accelerating ﬂow stroke. a separation vortex (2) appears on the blade suction surface on the hub side and reduces blade circulation.18 shows the illustration of the ﬂow structure obtained by numerical simulation [61. the clockwise trailing vortices are generated. since the vortical wake ﬂow enlarging ﬂow separation becomes more distant from the blade suction surface for the case of low solidity [103]. during a cycle of pitching motion. the performance in the accelerating ﬂow process becomes lower than in the decelerating ﬂow process [103]. which suppress the vortical wake ﬂow.18). All the results shown above are based on quasisteady assumptions or experiments conducted in unidirectional airﬂow test rigs. In the decelerating ﬂow process where the blade circulation decreases. 103]. which increases forces and moments applied to the blade and its root. They form counterclockwise trailing vortices. This eﬀect can be reduced by decreasing the solidity (Fig. Some groups studied Wells turbine under oscillating ﬂow (unsteady considers). 3. the extent of which can be appreciable at low Reynolds numbers [82. and reduces fatigue life. which involves a hysteresis loop in the airfoil lift curve and account for a higher maximum lift force than that achieved in the static curve. Therefore.16). The results indicated a hysteretic eﬀect on the force coeﬃcients (Fig. LITERATURE REVIEW 43 3.2. 85. which enlarges the ﬂow separation on the adjacent blade suction surface (3b) [58. vortices opposite to the blade circulation are shed from the trailing edge. 104. as the blade circulation increases.17). turbine and windmill blades may all be aﬀected by this phenomenon. the mechanism of the hysteretic characteristics of Wells turbine is still unclear [47. 3. 57]. So. At high angles of attack. But the loop of hysteretic curve of the dynamic stall is diﬀerent from the one of Wells turbine. because of the excessive angle of attack near the hub. 62]. It brings about the clockwise vortical wake ﬂow (3a). Figure 3.CHAPTER 3. The hysteretic eﬀects are caused by asymmetry in the boundary layer development on the blade surfaces and oscillating motion of the wake.
like installing guide vanes [87.16: Eﬀect of unsteady ﬂow on Wells turbine performance [103]. like low eﬃciency. .CHAPTER 3.3 Methods to improve Wells turbine performance Due to the drawbacks of Wells turbine. or improving proﬁles [127]. 3.2. bad starting capability at low solidity and low output power. 111. as discussed in the next section. LITERATURE REVIEW 44 Figure 3. 117]. many researchers tried to suggest some ideas to improve the performance.
48 1. 3.1 Guide vanes installation Many papers [13. 56.3. 120]. 33.CHAPTER 3. 106. 116. 121.18: Illustration of the ﬂow structure in blade suction side [103].3. The results indicated in particular that the three dimensional guide vanes (variable angles along the vane span) providing a constant rotor blade angle of attack with radius lead to the best characteristics and are therefore recommended. 3. 99. 60. Figure 3.2. 110.20). 100. The eﬀect of guide vanes (Fig. 112. 105.5 0 5 10 15 Figure 3.19) has been investigated experimentally and theoretically by testing a model under steady ﬂow conditions. 122] have demonstrated the usefulness of 2D and 3D (twisted) guide vanes.2. 3.57 s=0. 44.2 Selfpitchcontrolled blades Experimental investigations were performed by model testing of the rotor with ﬁxed blades under steady ﬂow conditions [32. It is found that the running and starting characteristics of the Wells turbine with guide vanes are superior to those without guide vanes (Fig. 109. The turbine . LITERATURE REVIEW 3 45 CA s=0. 3.17: Eﬀect of solidity on the hysteretic behavior [103]. 101.
LITERATURE REVIEW 46 Figure 3.20: Guide vanes eﬀect on the turbine performance [105]. Figure 3. 3.21).CHAPTER 3. blade is set on the hub by a pivot located near the leading edge that enables it to oscillate between two prescribed setting angles of ±γb (Fig. the turbine blades .19: Installation of guide vanes [106]. As an airfoil set at a certain angle of incidence experiences a pitching moment M about the pivot.
033. 3.2. 126. For high ﬂow coeﬃcients. . On the other hand.22 for two diﬀerent airfoils (NACA 0015 and HSIM152621231576.2. 3. Pivot Wells Turbine u Air Chamber M b b Figure 3. the results are poor concerning power output. the torque improved only in the high operating range compared with conventional Wells turbine (γb = 0).23). The results indicated that the operating range improved when increasing setting angle. since the incidence angle (θ = α−γb ) decreased. Therefore. However. LITERATURE REVIEW 47 can ﬂip by themselves between +γb and −γb according to the ﬂow direction. using lean blade (swept blade) as shown in Fig. Therefore. The eﬀectiveness of the end plate has been checked by using CFD to get the optimal position of the plate. 3. higher torque and eﬃciency are obtained for high ﬂow coeﬃcients. 3.4 End plate The eﬀect of end plate (Fig. 3.24) on the turbine characteristics has been investigated experimentally for diﬀerent plate sizes (a/c) by model testing under steady ﬂow conditions and compared with the classical Wells turbine [119.3 Lean blade (blade swept) Experimental research on diﬀerent types of rotor blades has been conducted recently to improve the aerodynamic performance of the Wells turbine. 125]. A numerical study indicated that the comparison between standard NACA 0015 unswept blade turbine rotor and the swept one by 30◦ is very diﬃcult for small ﬂow coeﬃcients.3.CHAPTER 3. however. while performance decreases for lower ﬂow coeﬃcient.3. Experiments indicate that the best geometry corresponds to a/c = 0. Others have higher peak eﬃciency but a narrower range of ﬂow rates [8. the airfoil HSIM152621231576 is better than both unswept and swept standard NACA 0015. 136]. 124. Fig. 123. without end plate. which can operate with wider operating range [108] and acceptable power output and eﬃciency. the swept one is better in term of eﬃciency. the rotor blade geometry has a remarkable inﬂuence on the turbine performance.21: Principle of using selfpitchcontrolled blades [106]. On the other hand. 102. Some rotor geometries give a considerably wider range of ﬂow rates for high eﬃciency and acceptable power output.
3.4 Multiplane Wells turbine In wave energy devices. the values of forward type and middle type are almost the same concerning axial force coeﬃcient. Globally.2.4. the highest eﬃciency was for the forward type position. This kinetic energy can be partly recovered by using a second stage of blades. where the available pressure drop is higher than a monoplane could accommodate.25. Twostage Wells turbine have been already investigated experimentally and theoretically [34]. The forward type case shows the highest tangential force coeﬃcient. multiplane turbine must be used. Several investigations studied the performance of multiplane Wells turbine and are reviewed in this section [14. CA . 3.22: Swept blade (lean blade) [8]. 55. leading to wider operating range and higher turbine performance. . at least in the absence of guide vanes.CHAPTER 3. On the other hand. but considering only symmetric airfoils. LITERATURE REVIEW 48 Figure 3.2. 3. while the backward type has the lowest value.23: NACA 0015 and HSIM 152621231576 blades [8]. Figure 3.1 Twostage Wells turbine (biplane turbine) All the previous theoretical and experimental results for the ﬂow ﬁeld around a Wells turbine rotor indicate that a considerable amount of exit kinetic energy is lost with the swirl component of the ﬂow velocity. as shown in Fig. 80].
In this conﬁguration.23 and Reynolds numbers of 3 105 and 6. called LIMPET (Land Installed Marine Pneumatic Energy Transmitter) in UK is constructed with a twostage contrarotating Wells turbine [6].2 and 0. respectively. 3. with nominal Mach numbers of 0. LITERATURE REVIEW 49 Figure 3. respectively.24: Rotor blade with end plate: (a) Middle type. (c) Backward type [124].2 and 0. 83]. Some researchers studied this conﬁguration and made a comparison with the conventional biplane Wells turbine [17.CHAPTER 3.59 m. The turbine tip diameters for these tests were 0. 3.2 Twostage contrarotating Wells turbine The ﬁrst commercial wave power station (capacity of 500 kW). 18. 24. the two rotors are installed as in Fig.5 105 . The results indicated that a contrarotating Wells turbine is aerodynamically more eﬃcient than the biplane Wells turbine and operate without stalling over wider ﬂow rates .2.26. (b) Forward type. Smallscale experiments on the contrarotating Wells turbine have been conducted using constant ﬂow windtunnels.4.
CHAPTER 3. LITERATURE REVIEW
Flow
Downstream Rotor
50
pitch
Chord
Upstream Rotor
Flow
Figure 3.25: Two stage (biplane) Wells turbine.
Figure 3.26: Contrarotating Wells turbine [107].
than the biplane turbine. They concluded also that the downstream rotor must rotate at higher speed than the upstream to achieve some improvement. The performance of the contrarotating Wells turbine installed in the LIMPET wave power station was compared to the predicted performance from theoretical analysis and model tests [24]. Figure 3.27 gives the nondimensional turbine torque against ﬂow coeﬃcient for both rotors during both the intake and exhaust strokes. During exhaust, the results indicate that the LIMPET turbine stalls at approximately the same ﬂow coeﬃcient as the constantﬂow model, but that the torque at stall is reduced by approximately 25% for the exhaust stroke and 46% for the intake stroke. In addition to comparing exhaust and intake it is also interesting to consider the relative performance of either rotor shown in Fig. 3.27. During exhaust there is no measurable diﬀerence in stall between the rotors. However, on intake, the downstream rotor has a noticeably higher torque coeﬃcient at stall condition. Figure 3.28 shows a plot of the LIMPET turbine eﬃciency with ﬂow coeﬃcient during the intake and exhaust, together with turbine eﬃciency derived from constant ﬂow model tests. During exhaust, at low ﬂow coeﬃcients, the LIMPET turbine appears to have a higher eﬃciency than in constant ﬂow model tests. This is possibly due to
d
Rotation
CHAPTER 3. LITERATURE REVIEW
51
Figure 3.27: Torque coeﬃcients for both LIMPET Rotors during intake and exhaust [24].
a lower drag coeﬃcient at higher Reynolds numbers. However, measurement errors are magniﬁed at low torques and ﬂow coeﬃcients, so the data in this region should be viewed cautiously. However, the early onset of stall experienced by the LIMPET turbine reduces the peak eﬃciency of the turbine to approximately 52%, whilst simultaneously reducing its eﬀective operating range. During intake, a higher ﬂow coeﬃcient is required for the turbine lift force to overcome the drag force due to relatively smaller torques being generated for the same ﬂow coeﬃcients. Smaller turbine torques during intake cause a smaller increase in turbine eﬃciency with ﬂow coeﬃcient, resulting in a peak eﬃciency of only approximately 42%.
Figure 3.28: Comparison of LIMPET and constant ﬂow model turbine eﬃciencies [24]. This study concluded that the contrarotating Wells turbine ﬁnally has a lower efﬁciency than a biplane or monoplane Wells turbine with guide vanes. In addition, a contrarotating Wells turbine requires an additional generator (or a gearbox to re
CHAPTER 3. LITERATURE REVIEW
52
verse the direction of rotation of one rotor), making it more complex and expensive to implement than biplane and monoplane Wells turbines. Consequently, unless the performance of contrarotating Wells turbines can be signiﬁcantly improved, they are not recommended over other Wells turbine variants. Obviously, contradictory information are found in the literature concerning contrarotating conﬁgurations highlighting the need for further studies.
3.2.5
Summary of Wells turbine review
All the studies discussed in the present review have tried to understand and improve the performance of the Wells turbine, considering either the conventional geometry or slight modiﬁcations. All the improvements have been tested manually, by means of a tedious trialanderror analysis. Such studies are going on. All proposed modiﬁcations have been summarized and discussed in Table 3.2. It can be ﬁnally concluded that: Table 3.2: Main modiﬁcations allowing to improve Design Gain Contrarotating [14] Improve eﬃciency by 5% Pitch setting [106] Improve eﬃciency by 7% Guide vanes [33] Improve eﬃciency by 5% End plate [124] Improve eﬃciency by 5% ∗ Multistage [81] Wider operating range ∗∗ Improve eﬃciency by 2% Multistage [74] ∗ ∗∗ Symmetric airfoils NonSymmetric airfoils
16
the performance of Wells turbines Description and Comments Double shaft, complex For positive small angles, complex Smaller operating range Only for 0.2 ≤ φ ≤ 0.25 Reduce eﬃciency by 10% Small parameter space
Number of papers
12
8
4
0 1970 1980 1990 2000 2010
Oil crisis
Years
Figure 3.29: Wells turbine publication statistics in international journals and conferences.
CHAPTER 3. LITERATURE REVIEW
53
• There is a renewed interest for Wells turbines in recent years, as shown in Fig. 3.29; • Contradictory observations can be found in the literature, both qualitatively and quantitatively. Therefore, the results presented in this review must all be considered with great caution, in particular concerning the contrarotating design ; • Many small modiﬁcations and improvements have been proposed in the literature to improve global performance, as summarized in Table 3.2; • Nevertheless, a real optimization of the system has not been realized up to now and would therefore be very interesting. This will be the subject considered in later chapters of this thesis.
3.3
Conclusions
In spite of many published studies, it still seems possible to increase considerably the performance of both Savonius and Wells turbines by relying on modern computational methods, as demonstrated in later chapters. The needed methodology for this purpose is described in the next two chapters.
Chapter 4 Optimization
4.1 Introduction
Optimization pervades the ﬁelds of science, engineering, and business. In physics, many diﬀerent optimal principles have been enunciated, describing natural phenomena in the ﬁelds of optics or classical mechanics. Statistics treats various principles termed ”maximum likelihood,” ”minimum loss,” and ”least squares,” and business makes use of ”maximum proﬁt,” ”minimum cost,” ”maximum use of resources,” ”minimum eﬀort,” in its eﬀorts to increase proﬁts. A typical engineering problem can be posed as follows: A process can be described by representative equations (or perhaps solely by experimental data). You have a single performance criterion in mind such as maximum eﬃciency or minimum cost. The goal of optimization is to ﬁnd the values of the variables in the process that yield the best value of the performance criterion. A tradeoﬀ usually exists. Typical problems in engineering process design or plant operation have many (possibly an inﬁnite number) of solutions. Optimization is concerned with selecting the best among the entire set by eﬃcient quantitative methods, thanks to computers and associated software, which make the necessary computations feasible and cost eﬀective. To obtain useful information using computers, however, requires (1) a critical analysis of the process or design, (2) insight about what the appropriate performance objectives are (i.e., what is to be accomplished), (3) use of past experience, sometimes called engineering judgment [35], and (4) suitable methods and algorithms. Until recently, the denomination “optimization” was mostly used in the engineering literature to describe a trialanderror, manual procedure (undoubtedly related to optimization, but in a very minimalist sense), at the diﬀerence of a real, mathematical optimization. This is now changing rapidly. In the present project, mathematical optimization will be considered to obtain the optimal shape geometry. Hence, for us, optimization means ”the design (or operation) of a system or process to make it as good as possible in some deﬁned sense”. As a consequence, the best possible solution constrained by appropriate conditions should be ideally found, and not simply a ”better” one [132]. Another deﬁnition is given by [2], which states that optimization is the process of obtaining the ”best”, if it is possible to measure and change what is ”good” or ”bad”. The deﬁnition in [96] is that the optimization theory is a body of mathe54
CHAPTER 4. OPTIMIZATION
55
matical results and numerical methods for ﬁnding and identifying the best candidate from a collection of alternatives without having to explicitly enumerate and evaluate all possible alternatives. In practice, one wishes the ”most” or ”maximum” (e.g., salary) or the ”least” or ”minimum” (e.g., expenses). Therefore, the word ”optimum” is taken to mean ”maximum” or ”minimum” depending on the circumstances; ”optimum” is a technical term which implies quantitative measurement and is a stronger word than ”best” which is more appropriate for everyday use. Likewise, the word ”optimize”, which means to achieve an optimum, is a stronger word than ”improve”. A computer is the perfect tool for optimization as long as the idea or variable inﬂuencing the idea can be input in electronic format. The process of optimization lies at the root of engineering, since the classical function of the engineer is to design new, better, more eﬃcient, and less expensive systems as well as to devise plans and procedures for the improved operation of existing systems. The power of optimization methods to determine the best case without actually testing all possible cases comes through the use of a modest level of mathematics and at the cost of performing iterative numerical calculations using clearly deﬁned logical procedures or algorithms implemented on computing machines.
4.2
Optimization uncertainty
For practical problems, optimization does not deliver a ”solution truth” because of the uncertainty that exists in the mathematical and physical representation of the process or the data used to model it. Engineers have to use their own judgment in applying optimization techniques to problems that have considerable uncertainty associated with them, both from the standpoint of accuracy and from the fact that operating parameters are not always static. In some cases, it may be possible to carry out a ﬁrst analysis via deterministic optimization and then add on stochastic features to the analysis to yield quantitative predictions of the degree of uncertainty. Whenever the model of a process is idealized and the input and parameter data only known approximately, the optimization results must be treated judiciously. They can provide for instance upper limits on expectations. Another way to evaluate the inﬂuence of uncertain parameters in optimal design is to perform a sensitivity analysis. It is possible that the optimum value of a process variable is unaﬀected by certain parameters (low sensitivity); therefore, having precise values for these parameters will not be crucial to ﬁnding the true optimum [35]. Furthermore, optimization can only deliver accurate results when the quality of a speciﬁc design can be measured accurately. For our problem, this issue will be considered in the next chapter.
4.3
How can we achieve optimization?
Most reallife problems have several solutions and occasionally an inﬁnite number of solutions may be possible. Assuming that the problem at hand admits more than one solution, optimization can be achieved by ﬁnding the best solution of the problem in
. iterative numerical procedures are used to generate a series of progressively improved solutions to the optimization problem. However. The most important general approach to optimization is based on numerical methods (real optimization). . Unfortunately.. OPTIMIZATION 56 terms of some performance criterion.. A graphical method can be used to plot the function to be maximized or minimized if the number of variables does not exceed a few. the system is set up and the process variables are adjusted one by one and the performance criterion is measured in each case. If the function depends on only one variable. x2 . as follows: • Analytical methods • Graphical methods • Experimental methods • Numerical methods Analytical methods are based on the classical techniques of diﬀerential calculus. The discipline encompassing the theory and practice of numerical optimization methods has come to be known as mathematical programming [2]. . xn to assume zero values. will reveal readily the peaks and valleys of the function. the graphical method is of limited usefulness since in most practical applications the function to be optimized depends on many variables.. x1 and x2 . if the function depends on only two variables. it can lead to unreliable results since in certain systems. a plot of f (x1 ) versus x1 will immediately reveal the maxima and/or minima of the function. In this method. they can be readily programmed on digital computers. like a topographical map of a speciﬁc region.. Furthermore. The problem to be solved must obviously be described in mathematical terms before the rules of calculus can be applied. it cannot be applied to highly nonlinear problems or to problems where the number of independent parameters exceedingly grows. when changes in the independent variables or the performance criterion from iteration to iteration become insigniﬁcant. and must be adjusted simultaneously to yield the optimum performance criterion. In this approach. xn ) with respect to x1 . two or more variables interact with each other. A contour plot. Several general approaches to optimization are available. a set of contours can be constructed... xn that cause the derivatives of f(x1 . In these methods the maximum or minimum of a performance criterion is determined by ﬁnding the values of parameters x1 . However. For example. Numerical methods can be used to solve highly complex optimization problems of the type that cannot be solved analytically. The method need not entail the use of a digital computer. say. The process is terminated when some convergence criterion is satisﬁed. x2 . x2 . say. covering: • Linear programming . The optimum performance of a system can sometimes be achieved by direct experimentation.CHAPTER 4. This method may lead to optimum or near optimum operating conditions.. starting with an initial estimate for the solution. . x1 .. Similarly.
.. xN ) whose values are restricted to satisfy a number of realvalued equations hk (x) = 0.4 Structure of optimization problems Although the application problems discussed in the previous section originate from radically diﬀerent sources and involve diﬀerent systems. . a (U ) (L) set of inequalities gj (x) ≥ 0. and the variable bounds xi ≥ xi ≥ xi . and their number will always be ﬁnite. J i = 1. . .. 4. at their root they have a remarkably similar form.. that is. .. to the equations hk (x) = 0 as the equality constraints. unconstrained optimization problem.. Optimization then means: Minimize or maximize Subject to f(x) hk = 0 gx ≥ 0 (U ) (L) xi ≥ xi ≥ xi k = 1. and to the inequalities gj (x) ≥ 0 as the inequality constraints... these problem functions will always be assumed to be real valued. The diﬀerences among them will be illustrated in the nest sections.. N but it is almost never found in engineering. x2 . . OPTIMIZATION • Integer programming • Quadratic programming • Nonlinear programming • Dynamic programming 57 Each one of these branches of mathematical programming is concerned with a speciﬁc class of optimization problems. K j = 1. For our purposes.. . All ﬁve can be expressed as problems requiring the minimization of a realvalued function f(x) of an Ncomponent vector argument x = (x1 ... N This general problem is called the constrained optimization problem.CHAPTER 4. In subsequent discussions we will refer to the function f(x) as the objective function. The problem in which there are no constraints... would correspond to J =K=0 and xi (U ) = −xi (L) =∞ i = 1..
3 Quadratic programming (QP) If the constraints are linear and the objective function is quadratic.5. Problems with nonlinear objective and linear constraints are sometimes called linearly constrained nonlinear programs. Constrained problems in which the function hk and gj are all linear are called linearly constrained problems. at least some of the variables are required to assume only integer values. . only direct optimization is considered. if it is a ratio of linear functions.5. such an optimization problem is said to be a quadratic programming (QP) problem.5. hk .CHAPTER 4.2 Integer programming (IP) In certain linear programming problems. optimization can be readily achieved by using some powerful LP algorithms. 4. Unconstrained problems in which x is a onecomponent vector are called singlevariable problems and form the simplest. This subclass can further be subdivided into those with a linear objective function f and those in which f is nonlinear. 4. 4. and so on. 4. and problems in integer variables. Note that. which are called linear programs.5 Types of mathematical programming Several branches of mathematical programming were enumerated in Section 4. This class can further be subdivided according to the particular structure of the nonlinear objective function. and gj and on the dimensionality of x. If f(x) is quadratic.1 Linear programming (LP) If the objective and constraint functions are linear and the variables are constrained to be positive. The diﬀerences among the various branches of mathematical programming are closely linked to the structure of the optimization problem and to the mathematical nature of the objective and constraint functions. it is called a fractional linear program. Subdivision into these various classes is worthwhile because the special structure of these problems can be sometimes eﬃciently exploited in devising solution techniques [2]. The category in which all problem functions are linear in x includes problems with continuous variables. OPTIMIZATION 58 Optimization problems can be classiﬁed further based on the structure of the functions f. Possibilities oﬀered by inverse design (based on solving the adjoint problem) are not discussed.3. in all this document. This restriction renders the programming problem nonlinear and more complex. the problem is a quadratic program. which are called integer programs. Each one of these branches consists of the theory and application of a collection of optimization techniques that are suited to a speciﬁc class of optimization problems.
to deﬁne the quantitative criterion on the basis of which candidates will be ranked to determine the ”best”. the objective function and usually the constraint functions are nonlinear.1 Deﬁning the system boundaries Before undertaking any optimization study. This composite activity constitutes the process of formulating the engineering optimization problem. Good problem formulation is the key to the success of an optimization study [96]. This is the most general branch of mathematical programming and.CHAPTER 4. and to deﬁne a model that will express the manner in which the variables are related. robust. OPTIMIZATION 59 4. it is necessary to clearly delineate the boundaries of the engineering system to be optimized. The choice of an optimization algorithm depends on the mathematical behavior and structure of the objective function. where subsequent decisions are inﬂuenced by earlier ones. and eﬃcient algorithm. it is important to clearly deﬁne the boundaries of the system under investigation. it may be necessary to expand the system boundaries and to include other subsystems that strongly aﬀect the operation of the system under study.5. This is clearly the case for the problems considered in the present thesis.6. however.4 Nonlinear programming (NP) In nonlinear programming problems. Rarely.6 Requirements for optimization To apply the mathematical results and numerical techniques of optimization theory to real engineering problems. the act of deﬁning the system boundaries is the ﬁrst step in the process of approximating the real system. 4. in eﬀect. for purpose of analysis. a series of decisions must be made in sequence. a number of optimizations have to be performed in sequence and a general strategy may be required to achieve an overall optimum solution.5. In many situations it may turn out that the initial choice of boundary is too restrictive. They serve to isolate the system from its surroundings. all interactions between the system and its surroundings are assumed to be frozen. 4. LP and QP can be considered as special cases of nonlinear programming. For many applications. 4. specialized algorithms exist which are more eﬃcient or more robust [19]. since interactions always exist. To fully analyze a given engineering system. . to select the system variables that will be used to characterize or identify candidates. Nonetheless. because. the objective function is a well behaved nonlinear function and all that is necessary is a general purpose.5 Dynamic programming (DP) In some applications. In such applications.
Clearly. optimization studies may sometimes be performed by experimenting directly with the . Second. maximum production rate. OPTIMIZATION 60 4. and wall thickness of a cylindrical tank as independent variables but exclude the possibility of using a compressor to raise the storage pressure.2 Performance criterion Given that we have selected the system of interest and have deﬁned its boundaries.4 System model Once the performance criterion and the independent variables have been selected. we next need to select a criterion on the basis of which the performance or design of the system can be evaluated. 4. variations in these key system parameters must be taken into account in the formulation of the problem. There is a considerable choice in the precise deﬁnition of such a criterion: minimum production time. There are several factors to be considered in selecting the independent variables.6. For instance. minimum weight. another consideration in the selection of variables is the level of detail to which the system is considered. if in the design of a gas storage system we include the height. and so on. it is equally important not to obscure the problem by the inclusion of a large number of ﬁne details of subordinate importance. the next step in problem formulation is to assemble the mathematical and physical models that describe the manner in which the problem variables are related and the way in which the performance criterion is inﬂuenced by the independent variables.6.CHAPTER 4. In principle. A good rule is to include only those variables that have a signiﬁcant impact on the composite system performance criterion.6. we may well obtain a very poor design. maximum torque. Finally.3 Independent variables The third key element in formulating a problem for optimization is the selection of the independent variables that are adequate to characterize the possible candidate designs or operating conditions of the system. lying outside the boundaries selected for the system in question. While it is important to treat all key independent variables. it is important to include in the formulation all the important variables that inﬂuence the operation of the system or aﬀect the design deﬁnition. However. First. we could obtain a design with a much lower overall cost because of a reduction in the required tank volume. so that the best design or set of operating conditions can be identiﬁed. it is important to diﬀerentiate between system parameters that can be treated as ﬁxed and those that are subject to ﬂuctuations inﬂuenced by external and uncontrollable factors. we would certainly ﬁnd the leastcost tank dimensions. 4. For the selected ﬁxed pressure. diameter. by including the storage pressure as an independent variable and adding the compressor cost to our performance criteria. the independent variables must be selected so that all important alternatives are included in the formulation. Furthermore. it is necessary to distinguish between variables whose values are amenable to change and variables whose values are ﬁxed by external factors. Thus. minimum energy utilization.
Furthermore. that set of design variables that causes f(x) to take on its optimum value. called a model. • Gradient methods. Models are typically used in engineering design because they oﬀer the cheapest and fastest way of studying the eﬀects of changes in key design variables on system performance. The artiﬁcial creatures in EAs.8 Evolutionary Algorithms Evolutionary Algorithms (EA) are computer programs that attempt to solve complex problems by mimicking the processes of Darwinian evolution (e. It is hoped that over time the most successful of these creatures will evolve to describe the optimal solution. which require estimates of the ﬁrst and second derivatives of f. the independent variables of the system or process may be set to selected values. OPTIMIZATION 61 real system. known as individuals. Each individual encodes a single possible solution to . As a consequence. In many applications it is either impossible or very time consuming to obtain the derivatives. which require estimates of the ﬁrst derivative of f. For instance. The methods that have been devised for the solution of this problem can be classiﬁed into three broad categories [96] based on the type of information that must be supplied by the user: • Directsearch methods. which use only function values. function evaluations rely again on computerbased simulations. 4.CHAPTER 4. The optimization methodology would then be used to predict improved choices of the independent variable values and the experiments continued in this fashion. They compete continually with each other to discover optimal areas of the search space. In practice. it is almost impossible and extremely time consuming to estimate the function derivatives. are typically represented by ﬁxed length strings or vectors. in some applications available computer storage is limited. In an EA a number of artiﬁcial creatures search over the space of the problem. • Secondorder methods.g. For ﬂow optimization the NavierStokes equations constitute the natural model. the system operated under those conditions. [76]). only Genetic Algorithms (a stochastic direct methods) are discussed further. in still others.7 Optimization methods In this work we consider methods that iteratively produce estimates of x∗ (optimum solution). For our problem.. No single method can be expected to uniformly solve all problems with equal efﬁciency. in others. Thus. and are therefore noisy and of limited accuracy. and the system performance index evaluated using the observed performance. high accuracy in the ﬁnal solution is desired. function evaluations are very time consuming. 4. most optimization studies are carried out with the help of a representation of the real system.
nondiﬀerential. Fitness scores are assigned to each individual .8.1 Genetic Algorithm (GA) This is the most popular type of EA. Nowadays. ﬁrst developed by Holland (1975) and thoroughly reviewed by Goldberg (1989). where the stochastic technique of roulette wheel parent selection is used to pick parents for the new population. Fogel et al. One seeks the solution of a problem in the form of strings of numbers. 4. developed in Germany by Rechenberg (1973) and Schwefel (1981). and evolutionary programming (EP). Each of these three algorithms has been proved capable of yielding approximately optimal solutions given complex. Every individual is then assigned a ﬁtness value.(1966) and subsequently reﬁned by D. Concatenating these numbers gives a string which can be used within an EA. OPTIMIZATION 62 the problem under consideration. and discontinuous search spaces. each individual would encode a speciﬁc set of torsion angles. Following this initial phase the main iterative cycle of the algorithm begins. only GA have been employed and therefore described now.1. evolution strategies (ESs). by applying operators such as recombination and mutation [114]. 4. This new population becomes the current population and the iterative cycle is repeated. The concept is as follow: 1. In the present project. A path through the components of the GA is shown as a ﬂowchart in Fig. Using mutation (perturbation) and recombination operators. The n children are assigned ﬁtness scores. B. EAs manipulate pools or populations of individuals. A population of N random or quasirandom individuals is initialized 2. J. The survival of the ﬁttest is employed and individuals compete against each other. For example. each angle of rotation around a ﬂexible bond could be encoded as a real number. Evolutionary pressure is applied in the steps of the iterations. The selection is applied either when choosing individuals to become parent of children or when choosing individuals to form a new population. Each block in this ”big picture” overview is discussed in detail in what follows.CHAPTER 4. originally developed by L. in order to construct an EA to search the conformation space of a molecule. There have been three main independent implementation instances of EAs: Genetic Algorithms (GAs). To generate a ﬁtness score the individual is decoded to produce a possible solution to the problem. The canonical GA encodes the problem within binary string individuals. real or doubleprecision representations are mostly used for engineering problems. Population members with high ﬁtness scores therefore represent better solutions to the problem than individuals with lower ﬁtness scores. A new population of N individuals is then selected from the N individuals in the current population and the n children. The value of this solution is then calculated using the ﬁtness function. Thus. The EA is started with an initial population of size N comprising random or quasirandom individuals. multimodal. the N individuals in the current population produce n children according to a survival probability ratio. Fogel (1995).
In Fig. The children replace the parents in the new population. otherwise go to step 3. mutation is performed on the new population. Parents are not required to be unique and. If the termination conditions are satisﬁed exit. cost. ﬁt individuals may produce many oﬀsprings. 5.2. With probability Pc (crossover). Each individual is assigned a sector of a roulette wheel that is proportional to its ﬁtness and the wheel is spun to select a parent. 4. OPTIMIZATION Define cost function. children are formed by performing crossover on the N/2 pairs of parents. While selection is random and any individual has the capacity to become a parent. Using roulette wheel parent selection N /2 pairs of parents are chosen from the current population to form a new population. 4. The new population becomes the current population. From a . selection based on roulette is expained using a trivial example with a population of ten individuals.1: Flowchart of a Genetic Algorithm. With probability Pm (mutation). 3.CHAPTER 4. 7. variables Select GA parameters 63 Generate initial population Decode chromosomes Find cost for each chromosome Select mates Mating Cross over and Mutation Convergence Check Done Figure 4. 6. selection is clearly biassed towards ﬁtter individuals. in each iteration.
With probability Pc each pair is recombined using the crossover operator to produce a pair of children. Pm and Pc are parameters of the algorithm and must be set by the user. mutation is applied to all or some individuals in the new population. Parents Crossover point Children Figure 4. The new population then becomes the current population and the cycle is repeated until some termination criteria are satisﬁed. N/2 pairs of parents are chosen.3: Example of crossover. . population of size N.CHAPTER 4. These parents initiate a new population. 4. Following crossover. The algorithm typically runs for some ﬁxed number of iterations. A cross point is selected at random. OPTIMIZATION 64 Figure 4. With probability Pm.2: Example of selection based on the roulettewheel for ten individuals. The probabilities of mutation and crossover. each bit on every string is inverted or modiﬁed. or until convergence is detected within the population. Each child is identical to one parent before the cross point and identical to the other after the cross point.3. The child individuals then replace their parents in the new population. This cut and splice operator is illustrated in Fig.
In order to control selection pressure within a GA ﬁtness values are often rescaled when applying roulette wheel parent selection. One problem with a canonical GA is that there is no guarantee that good individuals will survive from one iteration to the next. each iteration involves the application of one crossover or mutation operator and only one or two new individuals are added to the population. a low selection pressure can result in a GA taking an inordinate time to converge. • Simultaneously searches from a wide sampling of the cost surface.2 Advantage and disadvantage of Genetic Algorithm Some of the advantages of a GA include that it • Optimizes with continuous or discrete variables. • Does not require any derivative information. In order to increase program eﬀectiveness hybrid GAs mix problemspeciﬁc operations with crossover and mutation. as done in the present work.8. and GAs are best viewed as a paradigm for evolutionary search. Successful GAs have used integer string individuals or even more general representations such as tree and matrix structures. Too high a selection pressure and a GA will rapidly converge to a suboptimal solution. Of course. An elitist strategy involves copying the best individuals unchanged from the current population to the new population. Selection pressure is deﬁned as the relative probability that the ﬁttest individual in the population will be chosen as a parent relative to an individual of average ﬁtness. Not all algorithms produce an entirely new population at each iteration. usually replacing the worst individuals. or analytical functions. • Optimizes variables with extremely complex cost surfaces (they can jump out of a local minimum). rather than a speciﬁc algorithm. • Works with numerically generated data. • Provides a list of optimum variables. The binary encoding is often inappropriate for many problems and may be extended to nonbinary representations.CHAPTER 4. For instance. the traditional methods have been tuned to quickly ﬁnd the solution of a well . OPTIMIZATION 65 Many GAs applied to real world problems bear only a passing resemblance to the canonical GA. experimental data. the GA is not the best way to solve every problem. These advantages produce stunning results when traditional optimization approaches fail miserably. GAs have proved to be the most popular EAs. not just a single solution. • Deals with a large number of variables. 4. As a whole. While encouraging search. Specialized crossover operators have been devised to handle unusual encodings. • Is well suited for parallel computers. In a steadystate GA. They provide an eﬃcient a simple framework for attempting to solve complex search problems and have been widely applied.
. • Concurrent objectives will be considered. quickly ﬁnding the minimum while the GA is still analyzing the costs of the initial population. it is essential to compute the cost function as eﬃciently and accurately as possible. • Local minima are often expected. For all these reasons.9 Conclusions For the optimization problem considered in this work: • Function evaluations rely on a complex computation with a limited accuracy. many realistic problems do not fall into this category.CHAPTER 4. 4. This is the subject of the next chapter. • Parallel computers are available. However. For such cases the calculusbased methods by far outperform the GA. Now. OPTIMIZATION 66 behaved convex analytical function of only a few variables. GA appears to be ideally suited and will always be used in what follows. • Results for non optimal conﬁgurations may be nevertheless interesting.
.
Chapter 5 Numerical methods and algorithms
5.1 Introduction
Accurate computer simulations of ﬂuid ﬂows involve a wide range of issues, from grid generation to turbulence modelling to the applicability of various simpliﬁed forms of the NavierStokes equations. Many of these issues are not addressed at all in this thesis, like acoustics or reacting ﬂows [3]. Instead, we focus on selected numerical issues, with emphasis on ﬁnitevolume solutions of the NavierStokes equations, coupled with optimization to improve wind and wave energy turbines. We present in this chapter a foundation for developing, analyzing, and understanding such methods.
5.2
CFD & Optimization
Computational Fluid Dynamics (CFD) deals with the numerical analysis of complex ﬂows. Despite impressive progress in recent years, CFD remains an imperfect tool in the comparatively mature discipline of ﬂuid dynamics, partly because electronic digital computers have been in widespread use for only thirty years or so. The NavierStokes equations, which govern the motion of a Newtonian viscous ﬂuid were formulated well over a century ago. The most straightforward method of attacking any ﬂuid dynamics problem is to solve these equations for appropriate boundary conditions. Analytical solutions are few and trivial and, even with today’s supercomputers, numerically exact solution of the complete equations for the threedimensional, timedependent motion of turbulent ﬂow is prohibitively expensive except for basic research studies in simple conﬁgurations at low Reynolds numbers. Therefore, the straightforward approach is still impracticable for engineering purposes. Moreover, CFD will be considered for optimization in this work. Optimization methods allowing to identify a constrained, best possible solution have been known for a long time, but have not permeated all engineering disciplines yet. Concerning more speciﬁcally ﬂuid dynamics, the ﬁrst applications of optimization are found for aeronautical problems, in particular to improve wing proﬁle and ﬂight properties (typically, reduce drag). This is a problem with a high addedvalue and involves only 68
CHAPTER 5. NUMERICAL METHODS AND ALGORITHMS
69
the basic equations of ﬂuid dynamics (Euler or NavierStokes equations, depending on the investigated properties). This explains why most available books and articles dealing with optimization relying on evaluations obtained by Computational Fluid Dynamics concern such situations. Even then, the number of such books and review articles remains quite limited [132]. In our group, a considerable experience is available concerning such mathematical optimization relying on evaluations based on Computational Fluid Dynamics. We therefore employ our own optimization library, OPAL (for OPtimization ALgorithms), containing many diﬀerent optimization techniques. Diﬀerent CFD solvers (inhouse codes, ANSYSFluent, ANSYSCFX) have been coupled in the past with this optimizer. It has already been employed successfully to improve a variety of applications like for instance heat exchangers [40] or burners [52].
5.3
Computational procedure
The optimization procedure can only deliver the right solution, if all evaluations relying on Computational Fluid Dynamics indeed lead to an accurate ﬂow description. From the literature it is known that an accurate CFD simulation of the ﬂow around a Savonius turbine or Wells turbine is a particularly challenging task, mainly due to its highly timedependent nature and to the fact that ﬂow separation plays an important role for the eﬃciency of the system. It is therefore necessary to check the full CFD procedure with great care. Afterwards, the resulting methodology must be validated.
5.3.1
Preprocess: geometry & grid generation
Note ﬁrst that both problems considered in this thesis are indeed twodimensional in space, allowing an easier representation and discretization. The ﬂow is timedependent for Savonius turbine, while a steady solution is suﬃcient for Wells turbine. 5.3.1.1 Savonius turbine: size of computational domain
The appropriate size of the computational domain has ﬁrst been investigated. It must be indeed checked that this size does not impact the results of CFD. A computational domain of increasing dimensions (square domain of size 2L × 2L, suitably normalized by the rotor radius R, see Fig. 5.1) has been considered in the CFD computation. It is easy to notice from Fig. 5.1 that the three smaller domains are associated with a large variation of the torque coeﬃcient. On the other hand, the remaining three (larger domains) lead to a nearly constant value, with a relative variation of the output quantity below 1.1%. This demonstrates that the computational domain should extend at least over 20 times the rotor radius in each direction. In a smaller domain, the boundary conditions inﬂuence the ﬂow results in an inappropriate manner. Finally, the domain marked in Fig. 5.1 has been retained for all further Savonius computations in this work.
CHAPTER 5. NUMERICAL METHODS AND ALGORITHMS
Computational domain 2L
70
1.6
Torque coefficient (Cm)
1.2
R
0.8
Turbine
0.4
Employed domain
0 0 10 20 30 40
Relative size of the computational domain (L/R)
Figure 5.1: Size of the computational domain and impact on the torque coeﬃcient
5.3.1.2
Savonius turbine mesh independence
Several diﬀerent twodimensional grids of increasing density and quality, composed of 5 400 up to 120 000 cells, have been tested for a conventional twoblade Savonius turbine with obstacle and deﬂector plates, using a representative example of the target solutions. It is easy to notice from Fig. 5.2 that the ﬁve coarsest grids are associated with a large variation of the torque coeﬃcient. On the other hand, all remaining grids employing more than 80 000 cells lead to a relative variation of the output quantity below 1.3%. Since the cost of a CFD evaluation obviously increases rapidly with the number of grid cells, the intermediate grid range between 85 000 and 100 000 cells has been retained for all further results shown in the present work concerning Savonius turbines. 5.3.1.3 Wells turbine: size of computational domain
The mutual interaction between the blades constrains the size of the computational domain (Fig. 5.3) since only a single blade is considered. The appropriate size of the computational domain has been selected in the spanwise direction by using constant solidity and periodic boundary condition for both sides of the domain. In the axial direction recommendation from the literature have been implemented (s = 0.67, see section 3.2.1.1). 5.3.1.4 Wells turbine mesh independence
Corresponding results are shown in Fig. 5.4. Several diﬀerent twodimensional grids of increasing density and quality, composed of 12 200 up to 108 000 cells, have been tested for the baseline, nonsymmetric blade conﬁguration NACA 2421. All other parameters
2L
On the other hand.597 Yd1/R= 1. It is easy to notice that the six coarsest grids are associated with a large variation of the objective functions (here. all remaining grids employing more than 53 000 cells lead to a variation of the target variables smaller than 1. the intermediate .5%.0.0.191 Xd2/R=. Therefore.516 0.1 0 0 40000 80000 120000 Number of cells Figure 5.2: Gridindependence study for the torque coeﬃcient Figure 5. NUMERICAL METHODS AND ALGORITHMS 0.CHAPTER 5. the tangential force coeﬃcient is represented).3: Size of the computational domain around monoplane and twostage Wells turbines of the CFD are unchanged.2 0.4 71 Torque coefficient (Cm) 0.3 l=1 Xd1/R=.
reacting. The full numerical procedure and in particular the employed turbulence model have been validated by comparison with published experimental results for conventional Savonius turbines (twoblade and threeblade Savonius rotor of [38] and [48].CHAPTER 5. . 130. ﬂuid machines. NUMERICAL METHODS AND ALGORITHMS 72 grid range between 55 000 and 80 000 cells is retained for all further results shown in the present work for Wells turbine. Many researchers have used Fluent for simulating Savonius turbines [1.1 Numerical solution of the ﬂow ﬁeld Why Fluent? From all CFD codes available in our group.2. 5. The unsteady ReynoldsAveraged NavierStokes equations are solved using the SIMPLE (SemiImplicit Method for Pressurelinked Equations) algorithm for pressurevelocity coupling. and multiphase ﬂows.2 Model validation and selection for Savonius turbine All ﬂow simulations presented in this work rely on the software ANSYSFluent version 6.3. Fluent is the world’s leading commercial supplier of Computational Fluid Dynamics software and services. Fluent is already coupled to our own optimization library and has been ﬁnally selected as CFD solver for this study. Fluent has a capability to predict the performance of such turbines.3. 70] and Wells turbine [9. and a host of related phenomena involving turbulent. Figure 5. respectively).3. 103. heat and mass transfer. The ﬂow variables and all turbulent quantities are discretized in a ﬁnitevolume formulation using a secondorder upwind scheme.2 5.2. Opensource CFD codes like OpenFOAM are deﬁnitely cheaper but of limited numerical eﬃciency when considering a moving mesh or using parallel computers [139]. Therefore. ANSYSFluent has ﬁnally been retained. 131] in the past. 69. Fluent enables engineers to simulate ﬂuid ﬂow.4: Gridindependence study for the tangential force coeﬃcient 5. Other commercial tools do not show any noticeable advantage.3. The unsteady ﬂow is solved by using the Sliding Mesh Model (SMM). depending on the type (one or two stages).
1 0.15 0. 5.8 1 1. (a) 0.1 0. NUMERICAL METHODS AND ALGORITHMS 73 The inﬂuence of the turbulence model is shown in Fig. The employed computational procedure thus appears suitable to predict the performance of the turbine in the investigated range of operation and is now kept for all further simulations of Savonius turbines.2 1.2 0.05 0 0.2 0. T.2 1. These results demonstrate the excellent agreement obtained between CFD and experiments for the target function. Cp . Hayashi et al (2005) Realizable ke model SSTmodel Standard ke model RSM model Torque coefficient (Cm) 0.4 0. b) power coeﬃcient.8 1 1.3 0.2 0.5 0.6.25 0.4 0.4 Speed ratio (l) (b) 0.5: Validation of computational model: a) torque coeﬃcient.45 Exp.6 0. both compared to published experimental results for a twoblade conventional Savonius turbine [38] .4 Power coefficient (Cp) Speed ratio (l) Figure 5.3 0.CHAPTER 5.25 0. 5.5 and Fig.35 0.05 0 0.2 0.15 0.4 0. when using the Realizable k − ε turbulence model.6 0.
associated with a low but unknown inﬂow turbulence level.08 to 0.4 0. Therefore. in particular when using the Realizable kε turbulence model. 2007) Realizable ke model SST model Standard ke model RSM model 0.05 0 0.1 Exp.3.3 Model validation and selection for Wells turbine The full numerical model and in particular the employed turbulence model have been again validated by comparison with published experimental results for a standard.4 105 using the chord as characteristic length. Irabu & J. Roy.7. prescribing either k and or the turbulence intensity together with a length scale.CHAPTER 5. the RSM model appears to be unappropriate for such conﬁgurations.25) and is now kept for all further simulations.15 0. The inﬂuence of the turbulence model is shown in Fig. (K. This model thus appears suitable to predict the performance of the turbine in the later investigated range of operation (ﬂow coeﬃcient varying from φ = 0. it has been impossible to obtain a better agreement. Inlet boundary conditions for the RSM model have been implemented using diﬀerent possibilities.67 following [110]. These results demonstrate again the excellent agreement obtained between CFD and experiments for this standard conﬁguration. ReynoldsStress Model (RSM) leads to a considerably longer computing time but surprisingly to a worse agreement than the kε models. Published studies usually consider a range Re= 1 105 to Re= 5 105 . which corresponds to a number of blades equal to 8.6 0.2 0.2. .2 74 Power Coefficient (Cp) 0.8 1 1. monoplane Wells turbine at a ﬂow Reynolds number Re= 2. NUMERICAL METHODS AND ALGORITHMS 0.2 Speed ratio(l) Figure 5. Nevertheless. since this corresponds to realistic conditions for employing a Wells turbine. Solidity is assumed constant and equal to s = 0. For both turbines. This is probably due to the low turbulence level and to a larger inﬂuence of the inﬂow turbulence boundary conditions.6: Validation of computational model: power coeﬃcient compared to experimental results for a threeblade Savonius turbine [48] 5. 5. which are not properly characterized in the experiments.
model usually provides improved results for swirling ﬂows and ﬂows involving separation when compared to the standard k.2.67 a=8° 0.4 Realizable k − turbulence model As explained in the previous sections.2 0. (1998) Standard ke model RNG ke model Realizable ke model sst kw model RSM model 0.Setoguchi et al.2) E . This allows the model to satisfy additional mathematical constraints on the normal stresses. a critical coeﬃcient of the model.3 Exp.3) .CHAPTER 5.25 Flow coefficient Figure 5. µt σk ∂k + Pk + Pb − ρ − YM + Sk (5. the realizable k.3. E+5 k E=S . • Transport equations: ∂ ∂ ∂ (ρk) + (ρ k uj ) = ∂t ∂xj ∂xj ∂ ∂ ∂ (ρ )+ (ρ uj ) = ∂t ∂xj ∂xj where C1 = max 0.43.turbulence model developed by Shih et al. compared to experimental results for a monoplane Wells turbine [107] 5.7: Inﬂuence of the turbulence model on the tangential force coeﬃcient.model.1 0 0. Also.4 75 Tangential force coefficient s=0.1 0. NUMERICAL METHODS AND ALGORITHMS 0. [113] has always been retained. S= 2Sij Sij (5.2 0.05 0. rather than assumed to be constant as in the standard model. The Realizable k. The concept of a variable is also consistent with experimental observations in boundary layers. is expressed as a function of mean ﬂow and turbulence properties. T. Cµ .15 0. consistent with the physics of turbulence (realizability).1) ∂xj µ+ µ+ µt σ 2 ∂ √ +C1 C3 Pb +S +ρ C1 S −ρ C2 ∂xj k k+ ν (5.1 0 0. This model contains a new transport equation for the turbulent dissipation rate.
12) where φ= √ 1 arccos( 6 W ) 3 Sij Sjk Ski W = ˜ S (5.5) Pb = βgi P rt ∂xi where Prt is the turbulent Prandtl number for energy and gi is the component of the gravitational vector in the ith direction.85.10) (5. negligible for our applications: µt ∂T (5. The coﬃcient of thermal expansion. The default value of Prt is 0. 76 In these equations.04. As = 6 cos φ (5. NUMERICAL METHODS AND ALGORITHMS where S is the modulus of the mean rate of strain tensor.6) While Cµ is constant in the standard k.13) (5. calculated as follows: P k = µt S 2 (5.7) (5.model this coeﬃcient is calculated as follows: Cµ = U∗ = 1 A0 + As (5. in the Realizable k. Pk represents the generation of turbulence kinetic energy due to the mean velocity gradients.4) Pb is the generation of the turbulence kinetic energy due to buoyancy. β is deﬁned as: 1 ∂ρ β = − ( )P ρ ∂T • Modelling turbulent viscosity µt = ρ C µ k2 (5. The model constants A0 and As are given by: √ A0 = 4.CHAPTER 5.8) (5.14) .model.11) kU ∗ ˜ ˜ Sij Sij + Ωij Ωij ωk ωk ˜ Ω=Ω−2 and Ω = Ωij − ijk ijk where Ωij is the mean rate of rotation viewed in a rotating reference frame with the angular velocity ωk .9) (5.
CHAPTER 5. conﬁrming the present ﬁndings.3 5. σk = 1. The inﬂuence of the number of revolutions on the estimation of Cp by CFD has been also investigated systematically in a separate project [139].3.44.24) and (2. constant value of the timestep. (2. After 10 revolutions. Three complete revolutions are always computed. The absolute diﬀerence in Cp between this value and the one obtained after only three revolutions equals 0. the ﬁrst revolution is only used to initiate the correct ﬂow solution. (2. CT will be the objective function for the optimization.3. NUMERICAL METHODS AND ALGORITHMS ˜ S= 1 2 Sij Sij ∂uj ∂ui + ∂xi ∂xj 77 (5.17) 5. Since many diﬀerent conﬁgurations must be evaluated during the optimization. .21).3. This is an estimation of the uncertainty associated with the considered optimization process.8). 5.25). We have checked separately the inﬂuence of the number of revolutions on the computed turbine performance for the optimal design (Fig. while the ﬂow properties (in particular the power coeﬃcient Cp and the torque coeﬃcient Cm ) are obtained by averaging the results during the last two revolutions.3. the average power coeﬃcient reaches a constant value. it will generate a lift force FL normal to the free stream and a drag force FD in the direction of the free stream.1 PostProcessing: analysis of results Moment calculation for Savonius turbines The unsteady ﬂow is solved by using the Sliding Mesh Model. respectively which are calculated according to Eqs. 5.3 (5. These lift and drag forces can then be combined to get the tangential force FT and the axial force FA (Fig. by continuing the CFD simulation.2 Forces calculation for Wells turbines When an airfoil is set at an angle of incidence α in a ﬂuid ﬂow.15) Sij = • Model Constants C1 = 1. respectively. σ = 1. the computing time associated with one single CFD computation must be kept acceptable. (5. using an appropriate. It amounts to only 6% of the pressure coeﬃcient associated with the optimal design.16) C2 = 1.20) and (2. Only three revolutions have been thus computed for each Savonius design in order to reduce the needed computational time. The corresponding force coeﬃcients are the tangential force coeﬃcient CT and the axial force coeﬃcient CA .12). This combination (1 revolution for initializing the ﬂow + 2 revolutions to compute the target function) has been kept throughout.024.9. 2. This inherent uncertainty is very small compared to the range of Cp explored during the optimization and is thus deemed acceptable. The moment coeﬃcient Cm and the power coeﬃcient Cp are calculated according to Eqs.3. Together with the turbine eﬃciency.0.
OPAL is able to decide how to modify the input parameters before starting a new iteration. By checking the values stored in the result ﬁle.7 Instantaneous power coeff. and stored in a result ﬁle.8 Employed number of revolutions 0. of revolutions 0 1. 0. then printf( ”Cannot write ﬁle %s !”. ”w”) ) ) Open the inputﬁle for the simulation.4 0 0 15 30 45 60 75 Time Figure 5. Average Power Coeff. if ( !( f input := fopen(path.dat”).2 20 40 60 80 100 78 Power coefficient (Cp) l =0. Error opening the inputﬁle. ”input.1 /∗ Block 1 – Generate input ﬁle ∗/ begin sprintf( ﬁlename. The procedure is automated using journal scripts (to restart Gambit. the commercial tool Gambit for geometry and grid generation (including quality check) and the industrial CFD code ANSYSFluent to compute the ﬂow ﬁeld around the turbines.5.8: Inﬂuence of the number of revolutions on the instantaneous and on the average power coeﬃcient Cp computed by CFD for the optimum design shown later.CHAPTER 5. ﬁ comment: Create the header of the input ﬁle. 52. exit. Fluent) and a master program written in C (Algorithm 5. 5.4 CFD/Optimization coupling A fully automatic optimization ﬁnally takes place. calling all codes in the right sequence as shown in Fig. 132].9.4. NUMERICAL METHODS AND ALGORITHMS No. path ). As a result of the CFD computation the objective function(s) is determined. ”$DesignVariable1 = %le”. fprintf( f input. The fully coupled optimization procedure is a complex task. which has been described in detail in previous publications [40. using OPAL (decisionmaker for the conﬁgurations to investigate). Algorithm 5. 5.1). .
DesignV ariable2).jou” ). comment: Call the simulation tool. fprintf( f input. Close the result ﬁle.9: Schematic description of optimizer (OPAL) and CFD code coupling. NUMERICAL METHODS AND ALGORITHMS 79 Figure 5. Remove the previous result. fclose( f input ). then ErrorVariable := 1.dat”. DesignV ariable1). &Drag ). if Read the objective value. fscanf( f result.jou) .CHAPTER 5. fclose( f result ).dat” ). end /* Block 3 – Import the result(s) of the simulation */ begin if ( !(f result := fopen(”Result. Error reading the result ﬁle.”%le”. end /* Block 2 – Perform the simulation */ begin system( ”rm f old results. continue. ”r”) ) ) Open the result ﬁle. system( ”ﬂuent 2ddp g i journal ﬁle. Close the inputﬁle. end Many types of ﬁles have to be prepared in order to start optimization process. as follows: • Fluent journal ﬁle (ﬂuent. ”$DesignVariable2 = %le”.
the velocity speciﬁcation method is magnitude and direction with absolute reference frame and constant value. crossover probability and so on (Table 5. ﬁles have been constructed during the optimization. Since this work considers shape optimization. we can follow the convergence of the CFD solution for every conﬁguration during evaluation. The objective function considers only . this ﬁle has been built from two subﬁles (head and tail). which. This procedure is then repeated for every step in ﬂuent sequence. The values of all parameters describing the geometry are placed in the head ﬁle at the beginning of each optimization iteration. NUMERICAL METHODS AND ALGORITHMS 80 In this ﬁle. since many diﬀerent cases have been optimized in this work.5.5 5.1). • Optimization parameters ﬁle (*. containing number of generations. respectively”.out) with this ﬁle. For example. together. all the optimization parameters have been implemented.jou) By the same method.tcl) In this ﬁle.49 n 4. in our cases.9. mutation rate. as: • Input ﬁle (*.9902 and constant Ycomponent equal −0. • Output ﬁle (Result.dat) The output ﬁle includes all the objective results for every tested conﬁguration. diﬀerent parameters are considered in each conﬁguration. the tail ﬁle describes the body and contains all geometry steps. • Fluent output ﬁle (ﬂuent.9” means ”deﬁne boundaryconditions of the velocity inlet (zone number 3).49 and 4. Further. • Gambit journal ﬁle (gambit. all constrains of each optimization parameter have been deﬁned in this ﬁle. a statement like ”de bc vi 3 y y n 10 n 0. a ﬁle is also constructed for geometry and grid generation.1 Optimization parameters Savonius turbine: single objective optimization The design variables considered for the optimization will be described in the next chapter. Results are stored in this ﬁle after every generation. The inlet velocity value equal 10 m/s with constant Xcomponent of the ﬂow direction equal 0. all the steps needed for CFD have been coded by text interface as script ﬁle.13917 y n 0. However. population size. The Inlet turbulent speciﬁcation method is k and ε with constant values equal 0.in) This ﬁle contains the values of the optimization parameters for every generation.CHAPTER 5. are suﬃcient to ﬁx clearly the geometry of this speciﬁc case. For this purpose. 5. Beside these parameters.9902 n 0.13917.
The parameters of the GA are listed in Table 5.5. Mutation magnitude must be decreased during the optimization process to stabilize the population.8 at each generation. the population size and the eﬃciency of the Genetic Algorithm (computing time and solution quality) [20]. since there is a very strong relation between the number of parameters. The ﬁnal analysis can be started. ±15%) a This value is multiplied by 0.6 Conclusions All the tools required for the optimization have been now developed and validated.e. The objective function hence contains simultaneously two outputs of the simulation (multiobjective concurrent optimization).. and is deﬁned for a negligible density change as: η= F T ut ∆p0 Q (5. and the turbine eﬃciency η . 5.18) with Q the volumetric ﬂowrate through the turbine.7% Mutation probability 100% a Mutation magnitude 30% (i.CHAPTER 5. Table 5. . that should be maximized as far as possible: the output power coeﬃcient Cp .1. that should both be maximized as far as possible: the tangential force coeﬃcient CT .1. 5. depending on the number of free parameters.e. which is inversely proportional to the axial force coeﬃcient.1: Parameters of the Genetic Algorithm Parameter Value Population size of the ﬁrst generation. The parameters of the GA are given in Table 5. beginning with the Savonius turbine. NUMERICAL METHODS AND ALGORITHMS 81 one output of the simulation (single objective optimization). high tangential force coeﬃcient). The design variables again depend on the speciﬁc problem considered and will be listed in Chapter 7.2 Wells turbine: multiobjective concurrent optimization The central goal when designing an improved Wells turbine is to achieve high eﬃciency and high power output (i. For example the mutation magnitude is only 4% (±2%) after 10 generations.3% Crossover probability 16. The number of generations Ng will be varied. N 20 to 30 Number of generations Ng Survival probability 50% Average probability 33.
At the beginning of this project. called in what follows GWturbine.2 6. a threeblade rotor without passage in between has been proposed by a small company. At the end of the chapter. We will start by considering and optimizing this speciﬁc conﬁguration. yielding very promising results. The threeblade conﬁguration should be in particular advantageous to obtain good selfstarting conditions. is it possible to improve the performance by changing the shape of the blade (not being semicylindrical any more)? 2. Three issues must be speciﬁcally investigated in this case: 1. a small company contacted us with a modiﬁed. before comparing with the standard (but optimized) threeblade Savonius turbine. 6.Chapter 6 Savonius turbine: singleobjective optimization 6.2.1 Introduction As already discussed in section 3. thus reducing drag by ”opening” the returning blade? 82 . is it possible to improve the performance by using mobile parts for the returning blade. involving indeed only two blades. 6.1. is it possible to improve the performance by employing a deﬂector nose in front of the turbine? 3. threeblade design without gap. both twoblade and threeblade Savonius turbines have been proposed and constructed. will be fully optimized. in an eﬀort to improve the performance compared to the conventional system (Fig. the best overall solution.1).1 Optimal GWturbine: modiﬁed threeblade Savonius turbine without gap Performance of the original GWturbine A modiﬁed design.3.
1: Schematic shape of the GWturbine We begin by considering the newly proposed design compared to the conventional turbine. Increasing the number of blades increases the reverse moment as well. Closing the passage between the blades leads to a reduction of the airﬂow entering the system and increases the global drag on the returning blade (Fig. An analysis of the ﬂow can readily identify the reason for this loss of performance.6 0.1 0. it appears clearly from these ﬁrst comparisons that the new design systematically leads to a poorer performance. The peak value of Cp is only 0.8 1 1.05 0.15 0.3 0.18 for the conventional design.2 0. . Unfortunately.2 0.15 0.25 0.3). For very low values of λ.3 0.2 0.2 0.CHAPTER 6.25 0.05 0 0. the diﬀerence is small.05 0.6 0.4 Negative Torque Conventional Savonius turbine (two blades) Threeblade without passage Torque coefficient (Cm) Power coefficient (Cp) 0. but the loss of performance becomes considerable for increasing λ. 0.4 Conventional Savonius turbine (two blades) Threeblade without passage 0. Figure 6.1 0.4 0.2 1.16 compared to 0.2: Performance comparison between the GWturbine and the conventional. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 83 Figure 6. both from the point of view of power coeﬃcient and of torque coeﬃcient.4 Speed ratio (l) Speed ratio (l) Figure 6. twoblade Savonius rotor.8 1 1.45 0.4 0.1 0.05 0 0.1 0.2 shows the comparison between the GWturbine (denoted “three blades without passage”) and the conventional Savonius turbine performance.2 1.35 0. 6.
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION
84
Figure 6.3: Instantaneous velocity vectors around the GWturbine
6.2.2
Reducing the drag on the returning blade
In order to improve the performance of the GWturbine, it was proposed to reduce the drag of the returning blade by employing mobile blade sections. In closed position, these mobile parts return back to the standard blade shape. When open, they should lead to a considerable reduction of the drag on the returning blade. This procedure is described schematically in Fig. 6.4.
Closed blade (advancing blade)
Wind
Just closed blade
Just opened blade
Figure 6.4: Schematic description of the GWturbine with open returning blade. The new design with an open returning blade has been investigated numerically for diﬀerent values of the speed ratio λ. For these computations, the slits are considered to be inclined by 30◦ (constant value) compared to the local blade direction. The resulting performance has been compared with that of the baseline GWturbine, as shown in Fig. 6.4. The results show a considerable improvement of the performance
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION
85
when comparing with the baseline GWturbine. This improvement is visible both for the torque and for the power coeﬃcient. The gain is directly a consequence of the drag reduction on the returning blade, due to air passing between the slits of this returning blade, as seen in Fig. 6.5. The increase of the power coeﬃcient reaches 0.0843 at λ = 0.8, which means a relative improvement of performance by almost 36% under such conditions. The GWturbine with open returning blade is also systematically better than the conventional Savonius turbine. For example, at λ = 0.7 the relative increase in performance is 25.9%.
Conventional Savonius turbine Three blades without passage (modified design) Three blades with open returning blade %Relative increase of open design vs. modified design %Relative increase of open design vs. conventional turbine
0.6
210 180 150 120
Torque coefficient (Cm)
0.5 0.4 0.3 0.2
90 0.1 0 0.1 0.2 0.2 0.4 0.6 0.8 1 1.2 60
Negative Torque
30 0 1.4
Speed ratio (l)
0.3
Conventional Savonius turbine Three blades without passage (modified design) Three blades with open returning blade % increase of open design vs. modified design % increase of open design vs. Savonius turbine
210 180 150 120
Power coefficient (Cp)
0.25 0.2 0.15 0.1
90 0.05 0 0.05 0.1 0.2 0.4 0.6 0.8 1 1.2 60 30 0 1.4
Speed ratio (l)
Figure 6.5: Performance comparison between the GWturbine with or without open returning blade. The performance of the conventional Savonius rotor is also shown for comparison. Top: torque coeﬃcient. Bottom: power coeﬃcient. Due to the eﬀectiveness of this design, the eﬀect of the slits opening angle on the performance has been investigated. Therefore, nine diﬀerent angles (from 10◦ to 90◦ ) have been studied, as shown in Fig. 6.6. From the results for both torque and power coeﬃcients, the best performance for this design is when the slit angle lies between 30◦
% Relative increase
% Relative increase
CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION
0.6
86
Torque Coefficient (Cm)
0.5 0.4 0.3 0.2 0.1 0 0.1 0.2
open with 10° open with 20° open with 30° open with 40° open with 50° open with 60° open with 70° open with 80° open with 90°
0.2
0.4
0.6
0.8
1
1.2
1.4
Speed ratio (l)
0.3 0.25
Power coefficient (Cp)
0.2 0.15 0.1 0.05 0 0.05 0.1 0.15 0.2 0.2 0.4 0.6 0.8 1 1.2 1.4
open with 10° open with 20° open with 30° open with 40° open with 50° open with 60° open with 70° open with 80° open with 90°
Speed ratio(l)
Figure 6.6: Performance of open returning blade turbine for diﬀerent slit angles. Top: torque coeﬃcient. Bottom: power coeﬃcient.
and 40◦ considering at eﬀective operating range between λ = 0.6 to λ = 1.2, corresponding to practical applications.
6.2.3
Inﬂuence of a rounded obstacle plate
In the original concept, a rounded deﬂector structure is placed in front of two counterrotating GWturbines. It is therefore important to check the inﬂuence of this deﬂecting surface on the global system performance. The corresponding geometry is shown schematically in Fig. 6.7. Figure 6.8 shows a comparison between an isolated GWturbine and the turbine placed behind the rounded deﬂector. Here again, it appears unfortunately that the performance is systematically reduced by the deﬂector. A GWturbine placed directly in the free ﬂow leads systematically to a higher performance, both in terms of torque
As a consequence.8 5m Figure 6. This value of λ is retained. moving only one point (p1 ). When analyzing the resulting ﬂow ﬁeld. The mathematical optimization procedure described previously (Genetic Algorithm relying on automated evaluations through CFD) is employed to ﬁnd the optimal blade shape.9). Two degrees of freedom are left simultaneously to the OPAL optimizer: X1 and Y1 deﬁne the coordinate of the center point of the blades (Fig. coeﬃcient and of power coeﬃcient. 6.7.2.675) for (X1 /R) and (−0.32 : 0. The limits of this domain for the two parameters are (0. since it is known from the literature that it corresponds to the peak power coeﬃcient of the conventional turbine (nominal conditions). This is done for a speed ratio λ = 0.9) has been carried out.8 0. an optimization of this shape geometry reconstructed by splines based on three discrete points (Fig. 6.CHAPTER 6. it has never been demonstrated that such a blade shape leads to an optimal performance. kept constant during the optimization. the optimization process thus involves simultaneously two parameters (or degrees of freedom): X1 and Y1 . the air ﬂow is directed away from the blades instead of entering the system.4 Optimization of the blade shape Both for the conventional Savonius turbine and for the new design. Finally. where R is the radius of the original design. Therefore. semicylindrical blades are considered as a starting condition.294) for (Y1 /R). 6.294 : 0. the decrease in performance can be related to the excessive size and large radius of the planned deﬂector. Even if such blades are traditionally employed. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 87 w Flo Rotation 5 cm 5m Wind 0. In this study.7: Schematic description of the counterrotating GWturbine with rounded deﬂector. For each geometrical conﬁguration one single objective Rotation 5 cm Flo w . following the literature. considering an incident wind velocity U = 10 m/s. a relatively large domain has been deﬁned for the optimization in the parameter space.
SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 0. (power output coeﬃcient) is determined by CFD evaluations.8 1 1. The results presented in Fig.4 Speed ratio (l) Figure 6.8: Performance comparison between the new concept in free ﬁeld or placed behind a rounded deﬂector.2 0.1 0 0.35 0. This optimum point diﬀers noticeably from the original design and corresponds to the point of coordinate X1 /R = 0. An optimal conﬁguration can readily be identiﬁed for λ = 0. When compared with the GWturbine (semicylindrical blade shape.1 0.11).2 0.15 0.8 1 1. 6.25 0.CHAPTER 6.2 1.2339.10 indicate that the considered objective is indeed considerably inﬂuenced by the two free parameters. X1 and Y1 .3 Power coefficient (Cp) 0.7.5 88 Torque coefficient (Cm) 0.0521 as shown in Fig.1 0.1 0.6315 and Y1 /R = −0.1638 and a torque coeﬃcient Cm = 0. 140 diﬀerent geometrical settings have been evaluated by CFD.4 0.02126 at λ = 0. As a whole. 6. this means a relative increase of the performance .05 0.2 0. 6. Fig. and should be maximized by the optimization procedure. the optimal point found by the optimization procedure corresponds simultaneously to an increase of the power coeﬃcient by 0.2 0.4 Speed ratio (l) 0.6 0.2 0.01487 and of the torque coeﬃcient by 0.3 0.2 1.10.15 0.05 0 0. For the power coeﬃcient.6 0.2 0.4 Conventional Savonius turbine Three blades without passage Three blades with rounded obstacle Negative Torque 0. This optimal condition leads to a power coeﬃcient Cp = 0.4 Conventional Savonius turbine Three blades without passage Three blades with rounded obstacle Negative Torque 0.7.
it is now important to check how this gain will change as a function of λ.Y1) Variable point r P2 Figure 6.9: Description of the blade shape with 3 discrete points connected by splines. by 7.11: The best (right) conﬁguration obtained during the optimization compared to the original GWturbine (semicylindrical shape: left).10: The two input parameters of the optimization and the power coeﬃcient. the performance of the obtained optimal conﬁguration . Therefore. Semicylindrical Shape Best Shape Figure 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION P3 89 Y1 X1 P1(X1. Since such turbines must operate also outside of the design conditions.1% compared to the original GWturbine.CHAPTER 6. Figure 6.
12: Power coeﬃcient of the optimized conﬁguration compared to the GWturbine as a function of λ. Three blades without passage Optimal blade shape % Relative increase 0. threeblade turbine.6) and very high (λ > 1. • The rounded deﬂector is much too large in the original design.4 20 30 Speed ratio (l) Figure 6.8 1 1.2 0. %Relative increase 0.12 demonstrates that the improvement of power output coeﬃcient is observed at all conditions for intermediate values of λ (in particular between λ = 0. the modiﬁed shape is less eﬃcient than the semicylindrical one. best values are found between 30◦ and 40◦ .2 60 1. Figure 6.CHAPTER 6.7 and λ=1.1). It is therefore interesting to examine now such conﬁgurations.1 0.2. • Opening the returning blade leads indeed to a considerable increase of performance.5 Conclusions on the GWturbine We can summarize all these results as: • The standard GWdesign (threeblade without gap) is less favorable than the conventional Savonius rotor. 6.2 .6 0. operating at such conditions should be avoided. For very low (λ < 0. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 90 has been computed for the full range of useful λvalues.05 0.15 10 0.2) values of λ.05 40 0 50 0.25 20% 10 0 Power coefficient (Cp) 0. compared to the semicylindrical design. • Similar modiﬁcations could possibly lead to even better results for classical Savonius turbines. starting with the conventional.12. The highest gain in eﬃciency with the new blade shape is obtained around λ = 1 and is roughly equal to 15%. as shown in Fig. The results of the new design with semicylindrical blades are also shown for comparison. Diﬀerent opening angles between 10◦ to 90◦ have been tested. 6. Therefore.4 0.
But. Published studies have demonstrated that threeblade turbines show considerable drawbacks compared to classical Savonius turbines (twoblade).1 Inﬂuence of obstacle plate Since one of the major advantages of the Savonius turbine is its simplicity and corresponding compactness. Adding a shielding obstacle should in principle reduce the reverse moment. at the diﬀerence of the previous studies. we are not looking here for a better solution. This part builds on top of a previous investigation [72] considering Savonius turbines with two as well as with three blades. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 91 6. The corresponding values of Cp and Cm have been determined numerically and sometimes experimentally as a function of the speed ratio λ. (see section 5. and as a consequence the total moment of the turbine will be increased. Related ideas have already been proposed by other groups in the past [48]. . 120 Wind gw R Threeblade Savonius Turbine Figure 6. but directly for the best possible one. Therefore. threeblade conﬁguration with gap (Fig. we will try now to improve the performance of this design through optimization. we will now investigate numerically the eﬀect of an obstacle shielding partly the returning blade of a Savonius turbine. since the total moment is the moment diﬀerence between the advancing and the returning blades. a modiﬁcation introducing a high complexity should probably not be retained. some simple guiding or deﬂecting plate could lead to the best eﬃciency improvement at the lowest possible cost and complexity.2.1.3. in particular a lower eﬃciency. 6.CHAPTER 6.13: Schematic description and main parameters characterizing a conventional Savonius rotor with three blades.3. 6.3. Nevertheless. robustness and low cost.3 Optimal threeblade Savonius turbine The conventional.13) of the Savonius rotor has been extensively studied in the past [48]. This has already been used to validate extensively our numerical procedure by comparison with published results. Considering the results of the previous studies and of section 3.2).
. Figure 6. The obstacle plate improves the selfstarting capacity for part of the θrange.15). Y1 and X2 ) which. it is essential to obtain a selfstarting system.1.14 shows the obtained static torque coeﬃcient Cms obtained for three diﬀerent positions of the obstacle plate as a function of θ.3.14: Static torque coeﬃcient as a function of the rotor angle θ for three diﬀerent values of Y1 choosing X1 /R = −1. For decentral. three parameters are considered (X1 . Roy (2007) (no obstacle) Three blade without obstacle Y1/R=0. which is a major advantage. These computations demonstrate that the obstacle plate has a considerable and mostly positive eﬀect on the static torque coeﬃcient.7. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 6.24 .4 Y1/R=0.14. that should be maximized as far as possible: the output power coeﬃcient Cp .5 0. Irabu & J.47 . b=68. The experimental results of [48] for the conventional threeblade turbines are also shown for comparison in Fig.CHAPTER 6. For this purpose. considering a ﬁxed incident wind velocity U = 10 m/s.5 Exp. This is done again for a speed ratio λ = 0.1.3.76. b=73 o o o Static Torque coefficient (Cms) 0. b=60. together with a ﬁxed value for Y2 are suﬃcient to ﬁx clearly the geometry of the shielding obstacle (Fig. Due to periodicity.2 Optimization of the obstacle position The position and the angle of the shielding obstacle will now be optimized. The mathematical optimization procedure described previously can be employed to ﬁnd the optimal position of the obstacle.5 Y1/R=0. the results are only plotted for θ between 0 and 120◦ .2 0.4 and X2 /R = −1.3 0. The objective function considers only one output of the simulation. The free design variables considered for the optimization will describe the obstacle position.1 0 0 30 60 90 120 Rotation angle(q) Figure 6.0 . To investigate this issue. K. 0. A selfstarting capability (Cms > 0) is always obtained in principle at any angle.1 Selfstarting capability 92 One important issue associated with wind energy is the selfstarting capability of the system. lowcost applications as considered here. the static torque exerted on a turbine at a ﬁxed angle has been computed by CFD as a function of this angle θ. 6. 6. 6.
X2 used to modify the position of the shielding obstacle. it must be checked that the shielding obstacle cannot come into direct contact with the rotor ( X1 > R and X2 > R). The corresponding positions of the obstacle all lead to conﬁgurations that shield partially the returning blade.88 0.CHAPTER 6. 6.16 indicate that the considered objective is indeed considerably inﬂuenced by the three free .1. so the acceptable range for the input parameters is shown in Table 6.0 X2 /R −1. Of course. Y1 .88 −1.15). By following a single line. The optimal solution is shown with a thick red line.88 −1. X1 /R. Each thin connecting line represents all numerical parameters associated with one speciﬁc conﬁguration. a truly optimal solution can be obtained considering a large set of possible parameters. 6.0) Obstacle x R Y2 X2>R Returning Blade Figure 6.177 (Fig. The ﬁrst three columns therefore show the value of the three free parameters. here the power coeﬃcient Cp . With these three factors.15: Schematic description of the geometry and free optimization parameters X1 . As explained previously.1: Acceptable range for the input parameters (parameter space) Parameter minimum maximum X1 /R −1. 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION y Advancing Blade X1>R 93 Wind Y1 (0. since Y2 is taken constant. In this manner.017 In Fig. the reader can therefore determine quantitatively the values of all parameters associated with one conﬁguration. the last column on the right corresponds to the value of the objective function. the position of the obstacle is perfectly determined.16 a parallel coordinate representation has been chosen since it is the most popular way to analyze output data from optimization involving several degrees of freedom. with Y2 /R = −1. when choosing the parameter space. The scale of the ﬁrst parameter (X1/R) is for instance bounded between −1. where each parameter is associated with its own vertical axis. Table 6. three degrees of freedom are left simultaneously to the OPAL optimizer: X1 and Y1 deﬁne the upper tip of the shielding obstacle. Such ﬁgures might be at ﬁrst diﬃcult to understand.88 (minimum) and −1.017 Y1 /R −0. The results presented in Fig. and the angle β can be deduced as well.017 (maximum). the value X2 is then suﬃcient to deﬁne the position of the lower tip. X2 /R and Y1 /R.
since such . optimum angle β and corresponding optimal power coeﬃcient Cp are listed in Table 6. Cp −1. The parameters of the optimal conﬁgurations are connected with a thick red line. X1 . This means a relative increase of the performance (measured by the power output coeﬃcient) by 27. The optimal conﬁguration (highest point in the right column in Fig.52◦ ) 0. As a whole. Figure 6. has not been used in the present case but could reduce the needed time by more than an order of magnitude.212 94 0.7.5%.2.36912 (80.05632 −0.2120 −1. all corresponding parameters being connected by a thick red line) can now readily be identiﬁed for λ = 0.CHAPTER 6.88 X2/R 0. Table 6.017 0.153 Without obstacle 1. Note that the userwaiting time could be considerably reduced by carrying out the requested CFD in parallel on a PC cluster [132]. compared to the conventional threeblade Savonius design without any obstacle. 6. The optimum obstacle position.017 0. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION Optimal configuration 1.88 Y1/R 1.16: Input parameters of the optimization and power coeﬃcient represented using parallel coordinates. requesting 22 days of computing time on a standard PC for the threeblade Savonius turbine.88 X1/R 0.7. 210 diﬀerent geometrical settings have been evaluated by CFD. Such a parallel procedure. It is now important to check how this gain would change as a function of λ. The power coeﬃcient of the conventional threeblade Savonius turbine is shown with a gray circle. the optimal point found by the optimization procedure corresponds to an absolute increase of the power coeﬃcient by 0. already implemented in OPAL.38162 When compared with the threeblade Savonius turbines (without shielding obstacle).2: Optimal Design parameter Threeblade Savonius turbine X1 /R Y1 /R X2 /R conﬁgurations value β power coeﬀ.0 1.16. as demonstrated in other studies. Y1 and X2 .058 at λ = 0. parameters.133 Power coeff.
5 Three blade Savonius without obstacle Three blade with obstacle (optimum configuration) % Relative increase 95 100 % Torque Coefficient (Cm) 0.25 100 % Power Coefficient (Cp) 0.1 20 0 0.4 80 0.2 Speed ratio(l) Figure 6.2 80 0.4 0. a turbine must be able to work also for oﬀdesign conditions.3 60 0.2 40 0.05 20 0 0.8 1 0 1.1 40 0.2 0. The corresponding relative increase is shown with blue line.2 Speed ratio (l) (b) 0.CHAPTER 6.17: Performance of the optimized conﬁguration (red line) compared to the conventional threeblade Savonius turbine without obstacle (black): a: torque coeﬃcient.2 0. This ﬁgure demonstrates that the improvement of both torque coeﬃcient and power output coeﬃcient is observed throughout for all values of λ. Therefore.6 0. the relative increase being highest for the largest values of λ considered in the present study.17.4 0. compared to the conventional threeblade Savonius turbine without obstacle.6 0.15 60 0. b: power coeﬃcient. The absolute gain for Cp increases even slightly with λ at ﬁrst. 6. as shown in Fig. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION (a) 0. Relative incease (%) Relative incease (%) . the performance of the optimal conﬁgurations have been ﬁnally computed for the full range of useful λvalues.8 1 0 1.
The reference point of the parameter space is point P0 . These domains are selected to prevent any domain overlap along the Y direction and to keep realistic blade shapes. which is . The order of a NURBS curve deﬁnes the number of nearby control points that inﬂuence any given point on the curve. six parameters are considered (XP 1 .2 Optimal blade shape In the previous section. YP 1 . YP 1 . that should be maximized as far as possible: the power coeﬃcient Cp . In order to achieve this objective. this means that the spline order is 5 in our case and the degree of the polynomial is 4. 6. the geometry of the blade shape (skeleton line) is now optimized in presence of the obstacle plate. since the obstacle position is ﬁxed. X2 /R = −1.3. XP 3 and YP 3 used to modify the blade shape.18: Schematic description of the free optimization XP 1 . The present study now aims at improving further the output power of the threeblade Savonius turbine.52◦ . The curve is represented mathematically by a polynomial of degree one less than the order of the curve. YP 2 . see Fig.38162 and Y2 /R = −1. The six shape parameters are suﬃcient to determine uniquely the geometry of the system. Optimization is used here to ﬁnd the best blade shape while taking into account the obstacle shielding the returning blade in the optimum position. Y1 /R = −0.05632. which leads to an angle β = 80.36912. the blade shape of the Savonius turbine will be optimized in order to ﬁnd the best possible ﬂow conditions.CHAPTER 6. 6. Now.YP2) P5 Turbine shaft Figure 6. The objective function contains one single output of the simulation. XP 3 and YP 3 . YP 2 . P4 P3(XP3.YP3) Y Fixed points Blade center (P0) r X P1(XP1. parameters Knowing all 5 points. For this purpose.YP1) Variable points r gw P2(XP2.3. NURBS). see Fig. The obstacle plate is kept ﬁxed in the optimal position identiﬁed in the previous section (X1 /R = −1.1770. the full proﬁle of the blade is reconstructed using standard splines (Nonuniform rational Bsplines.15). The free design variables considered for the optimization describe the blade skeleton line for a constant blade thickness of 2 mm. XP 2 . SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 96 6. the eﬃciency of the conventional threeblade Savonius turbine has been increased by placing in an optimal manner an obstacle plate shielding the returning blade. The parameter space considered in the optimization has been deﬁned as documented in Table 6. XP 2 .18).
24 1. YP 1 .03.153 Threeblade Savonius without obstacle 0.94 The results presented in Fig. XP 1 .24 XP 2 /r 0. a circular turbine shaft is included with a ﬁxed radius Rsh computed from Rsh /R = 0. all corresponding parameters being connected by a thick red line) can now readily be identiﬁed for λ = 0.2 Y3/r 0. XP 3 and YP 3 . YP 2 . and thus by the blade shape.94 −0.24 0.47 YP 1 /r −0.24 0. XP 2 .24 0.94 0. requesting one and a half month of total computing time on a standard PC.24 X2/r 1. The power coeﬃcient of the conventional threeblade turbine (semicylindrical shape) is also shown with a black circle. Optimal configuration X1/r 1. 6.18. 210 diﬀerent geometrical settings have been evaluated by CFD. The parameters of the optimal conﬁguration are connected with a thick red line.0 Figure 6.24 0.53 1. The optimum point positions and corresponding optimal power coeﬃcient Cp are listed in Table 6.24 X3/r 1. Table 6.CHAPTER 6. .4.2 YP 2 /r −0.94 Cp 0.19 indicate that the considered objective is again considerably inﬂuenced by the six free parameters. As a whole. 6.2 YP 3 /r 0.53 0. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 97 the center of the original. During the calculations.24 0.24 0.19: Input parameters of the optimization and power coeﬃcient represented using parallel coordinates.24 XP 3 /r 0.233 0. The optimal conﬁguration (highest point in the right column in Fig. semicylindrical shape with radius r as shown in Fig.24 1.19.47 Y1/r 0. 6.2 Y2/r 0.7.24 0.3: Acceptable range for the input parameters (parameter space for blade shape with the obstacle ) Parameter Minimum allowed Maximum allowed XP 1 /r 0.
It is now important to check how this gain would change as a function of λ.4345 −0. This ﬁgure demonstrates that the improvement of both torque coeﬃcient and power output coeﬃcient is observed throughout for all values of λ.3% for the optimum shape with obstacle plate. Note that this new shape (Fig.08 compared with the conventional threeblade Savonius turbine. as shown in Fig.021 compared with the conventional Savonius rotor (semicylindrical blade shape) with obstacle plate. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION Table 6. the modiﬁed shape exploits best the ﬂow redirection induced by the obstacle.6594 0. compared to the conventional threeblade Savonius design without obstacle. the reverse moment is reduced by the shielding obstacle. the modiﬁed shape leads to an increase of the positive moment of the advancing blade. The relative performance increase compared to the conventional threeblade Savonius conﬁguration is always higher than 30% in the eﬀective operating range.0307 0.CHAPTER 6.20: Optimum conﬁguration obtained with the optimization procedure. The absolute gain for Cp and Cm is even higher for lower λvalues. Wind Obstacle Figure 6. Indeed. respectively by 0. this means a relative increase of the performance (measured by the power output coeﬃcient) by 34. 6. As a whole.20) is only optimum in combination with the employed obstacle plate. . compared to the conventional threeblade Savonius turbine.7519 At λ = 0. Therefore.7 the optimal point found by the optimization procedure corresponds to an absolute increase of the power coeﬃcient by 0. simultaneously. 6. Cp 0.4: Optimal blade shape with the obstacle YP 1 /r XP 2 /r YP 2 /r XP 3 /r YP 3 /r Optimal power coeﬀ. the performance of the optimal conﬁguration has been ﬁnally computed for the full range of useful λvalues.5456 0.233 98 XP 1 /r 0. As a consequence.21.5464 0.
SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION (a) 0. 6. involving simultaneously an obstacle plate shielding the returning blade and a ﬂow deﬂector (frontal guiding plates). Four geometrical properties are optimized simultaneously: 1) the position of an obstacle shielding the returning blade.15 60 0.2 0.2 0.2 0.CHAPTER 6. The optimization process thus relies on free design variables that describe the position and angles of the plates.21: Performance of the optimized conﬁguration (red line) compared to the conventional threeblade Savonius turbine (black line): a) torque coeﬃcient. b) power coeﬃcient.25 100 % Power Coefficient (Cp) 0.3.6 0. 2) the position of a deﬂector guiding the wind toward the advancing blade.3 Optimal threeblade Savonius turbine with guiding plates After improving the eﬃciency of the threeblade Savonius turbine by placing appropriately an obstacle shielding the returning blade.05 20 0 0.4 0.8 1 1.6 99 Torque Coefficient (Cm) 0. the present study aims at investigating and improving further the output power of the threeblade Savonius turbine and improving the static torque. the blade shape (skeleton line) as well as the gap width s Relative incease (%) .4 0.4 Speed ratio(l) Figure 6. 3) the blade skeleton line and 4) the gap width.1 0 0.5 Three blade Savonius without obstacle optimal Configuration 0.6 0.4 0.1 Three blade Savonius without obstacle Optimal configuration % Relative increase 40 0.3 0. The corresponding relative increase compared to the conventional threeblade conﬁguration is shown with blue line.2 1.2 80 0.2 0 1. which measures the selfstarting capability of the turbine.4 Speed ratio (l) (b) 0. a modiﬁed design is considered.8 1 1. For this purpose.
The mathematical optimization procedure described previously can now be employed. At the end. Yd1 . Y2 . semicylindrical shape with radius r as shown in Fig. These domains are selected to prevent any domain overlap along the Y direction. 6. YP 1 .7. Y2 . XP 1 .YP3) Y Fixed points Blade center (P0) r X P1(XP1. b) XP 1 .22). Xd1 .YP1) Variable points r gw P2(XP2. Y1 . considering a ﬁxed incident wind velocity U = 10 m/s. the gap width gw . This is done as usual for a constant speed ratio λ = 0. YP 2 .22. that should be maximized: the output power coeﬃcient Cp . The objective function considers again only one output of the simulation. which is the center of the original.YP2) P5 Turbine shaft Figure 6. XP 2 . Yd1 . YP 3 and gw ). ﬁfteen free parameters are thus considered (X1 . to keep realistic blade shapes and to cover a wide region for positioning the guiding plates. Y1 . The parameter space considered in the optimization has been deﬁned as documented in Table 6. Yd2 . Xd1 . XP 2 . YP 2 .5. XP 3 and YP 3 used to modify the blade shape. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 100 (a) Xd2 Xd1 y Deflector Yd2 X1>R Wind Yd1 Advancing Blade x Y1 (0. YP 1 . additionally.CHAPTER 6. The reference point for the remaining space . 6. (Fig.0) Obstacle R Y2 X2>R Returning Blade (b) P4 P3(XP3. Xd2 . X2 . The reference point of the parameter space for the blade skeleton line is point P0 . Xd2 and Yd2 ). X2 .22: Schematic description of the free optimization parameters characterizing a threeblade Savonius rotor : a) plate parameters (X1 . XP 3 .
7. . all corresponding parameters being connected by a thick red line) can now readily be identiﬁed for λ = 0. 240 diﬀerent geometrical settings have been evaluated by CFD.03.65 0 2.88 0.024 Figure 6.24 0.24 0.24 1.53 0.153 Threeblade Savonius turbine without guiding plates 0. The optimum parameter values are listed in Table 6.24.24 0.23 indicate that the considered objective is indeed considerably inﬂuenced by the ﬁfteen free parameters.88 1. diﬃcult to illustrate in a static ﬁgure.1 0. The power coeﬃcient of the conventional threeblade turbine (semicylindrical shape) is also shown with a black circle. demonstrating that the employed grid captures all important ﬂow features in the vicinity of the rotor and guiding plates. a circular turbine shaft is included with a radius Rsh computed from Rsh /R = 0.38 1. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 101 parameters (guiding plates and gap width) is the global center of turbine rotation.2 1.1 1.24 1.1 0.7 0. 6. The parameters of the optimal conﬁguration are connected with a thick red line. 6. 6.88 1. Fortunately.94 0.03 0.6 1. The results presented in Fig.363 Optimum configuration 0. The optimal conﬁguration (highest point in the right column in Fig. this is quite straightforward to implement. The corresponding geometry is shown in Fig.76 1.CHAPTER 6.23: Input parameters of the optimization and power coeﬃcient represented using parallel coordinates. requesting 47 days of total computing time on a standard PC.94 0 1.6.18 0. so that researchers having access to parallel clusters can solve corresponding problems within an acceptable lapse of time.25.88 0. 6. During the calculations.24 Xp2/r Guiding plates positions Gap width s/R X2/R Xd1/R Xd2/R X1/R Xp3/r CP X2/R Y1/R Yd2/R Yp2/r Yd1/R Yp3/r 0.1 0 1. As a whole. Optimum configuration Blade shape Xp1/r Yp1/r 1.7 1. One instantaneous picture of the velocity ﬁeld is shown as an example in Fig.1 0. This is of course a dynamic process.2 0. Relying on parallel computers and possibly carrying out each CFD evaluation again in parallel [132] is clearly necessary when considering threedimensional problems.23.
88 0.1 1.0 Yd1 /R 1.1 Y2 /R −1.24 XP 3 /r 0.65 X2d /R −1.53 1.18 U Deflector =81.24 0. this means a relative increase of the performance (measured by the power output coeﬃcient) by 57% . At λ = 0.88 −1.24 0.2 0.7 the optimal point found by the optimization procedure corresponds to an absolute increase of the power coeﬃcient by 0.7 0.76 2.7 Gap width gw /R 0.94 Guiding plates X1d /R −1.1 Y1 /R −0.207 compared with the conventional threeblade Savonius turbine (semicylindrical blade shape). SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 102 Table 6. As a whole.0 X2 /R −1.2 YP 1 /r −0.94 −0.1 YP 2 /r −0.CHAPTER 6.88 −1.58 ° Returning blade Optimum design Figure 6.6 X1 /R −1.24 1.13 ° Advancing Blade Wind Obstacle =77.24 XP 2 /r 0.88 −0.0 Yd2 /R 1.24: Optimum conﬁguration obtained with the optimization procedure.1 YP 3 /r 0.24 1.03 0.5: Acceptable range for the input parameters (parameter space) Parameter Minimum allowed Maximum allowed Blade shape XP 1 /r 0.
This ﬁgure demonstrates that the improvement of both torque coeﬃcient and power output coeﬃcient is observed throughout for all values of λ. demonstrating again the interest of the optimized conﬁguration.7). compared to the conventional threeblade Savonius turbine.691 1. .7136 0. 6.388 1. for the optimum design.25: Instantaneous velocity vectors magnitude (m/s) around the optimum conﬁguration (zoom) at the design point (λ = 0. The relative performance increase compared to the standard Savonius conﬁguration is always higher than 50% in the usual operating range (0.58◦ 0.3638 β = 77.13◦ 1.5935 0.CHAPTER 6.2735 0.3089 γ = 81.6: Optimal Part Parameter Blade shape XP 1 /r YP 1 /r XP 2 /r YP 2 /r XP 3 /r YP 3 /r Guiding plates Xd1 /R Yd1 /R Xd2 /R Yd2 /R X1 /R Y1 /R X2 /R Y2 /R Gap width gw /R conﬁguration Value Angle 0.6077 0.6 ≤ λ ≤ 1).1075 1. The performance of the optimal conﬁguration has been ﬁnally computed for the full range of useful λvalues.1338 0. as shown in Fig.5901 0. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 103 Table 6.436 0.7065 0.4591 2.0988  Figure 6.26.
6 0. the static torque coeﬃcient is less than the classical one. To investigate this issue.2 40 0. 6.5 100 Power Coefficient (Cp) 0.6 Speed ratio(l) Figure 6. the static torque exerted on the turbine at a ﬁxed angle has been computed by CFD as a function of this angle θ. the results are only plotted for θ between 0 and 120◦ . There.8 1 1.2 1.CHAPTER 6.6 0. Compared to the classical turbine.4 0.6 0.2 0. The corresponding relative increase compared to the standard conﬁguration is shown with blue line. Figure 6. Relative incease (%) .3.3.3 0.3 60 0.4 0 1.27.2 1.4 80 0.2 0.7 Optimum design Conventional threeblade Savonius Torque Coefficient (Cm) 0.4 0. The experimental results of [48] for a conventional threeblade turbine are also shown for comparison in Fig.2 0.5 0.6 Speed ratio (l) Conventional three blade Savonius Optimum design % Relative increase 0. but remains strictly positive. except in a small range (90◦ ≤ θ ≤ 100◦ ).1 0 0. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 104 0.4 0.26: Performance of the optimized conﬁguration (red line) compared to the conventional threeblade Savonius turbine (black line): Top: torque coeﬃcient.1 20 0 0.4 1. Bottom: power coeﬃcient.8 1 1. Due to periodicity. 6.27 shows the obtained static torque coeﬃcient Cms for the optimal design compared to the classical threeblade turbine.1 Selfstarting capability For decentralized. lowcost windenergy applications. it is essential to obtain a selfstarting system. these computations demonstrate that the modiﬁcations have a considerable and positive eﬀect on the static torque coeﬃcient.
6 Static Torque coefficient (Cms) Exp. Cms is increased by 0. Therefore.7 0. Adding a shielding obstacle should in principle reduce the reverse moment.27: Static torque coeﬃcient Cms as a function of the ﬁxed rotor angle θ for the optimal design (ﬁlled red squares) compared to the classical threeblade Savonius turbine (blue plus).2 0. the conventional twoblade design will be optimized along the same lines. 0. Irabu&J.28). Roy(2007) Our CFD results: Conventional Threeblade Savonius Turbine Optimum design 0. in a last step. and as a consequence the total moment of the turbine will be increased.5 0.4 Conclusions on threeblade design It is possible to increase the performance of the conventional threeblade design using optimization by modifying the blade shape and gap width while placing suitable obstacle and deﬂector plates.3. The experimental results of [48] are also shown for comparison (empty black squares).3 0.1 0 Negative Torque (no selfstarting) 0.1 Obstacle plate To achieve better performance.4 Optimal twoblade Savonius turbine In this section.CHAPTER 6. it is known from the literature that the twoblade design is usual better in terms of power coeﬃcient. . new designs will be step by step investigated and optimized to improve the performance of the conventional twoblade Savonius turbine.4. 6. 6. K. 6.091 for the optimum design. the position of an obstacle shielding the returning blade of the Savonius turbine and possibly leading to a better ﬂow orientation toward the advancing blade is ﬁrst optimized (Fig. 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 105 Averaging over all angle positions.1 0 30 60 Rotation angle(q) 90 120 Figure 6.4 0. However.
In a previous study [72]. 6.1 Selfstarting capability It is essential to obtain a selfstarting system. A selfstarting capability (Cms > 0) is always obtained in principle at any angle. since Y2 is taken constant. as documented in . For all investigated positions involving an obstacle. the value X2 is then suﬃcient to deﬁne the position of the lower tip.177 (Fig. and the angle β can be deduced as well.28).29 shows the obtained static torque coeﬃcient Cms obtained for three diﬀerent positions of the obstacle plate as a function of θ. with negative values around θ = 140◦ − 170◦ (no selfstarting). the position of the obstacle plate is perfectly determined. 6. Figure 6. the evolution as a function of θ is similar to that of the conventional turbine. The experimental results of [38] for the conventional turbine are also shown for comparison in Fig.4. which is a major advantage. a much larger accessible domain has been prescribed.28: Schematic description of the geometry and free optimization parameters X1 . SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 106 y Advancing Blade Wind Y1 X1>R x (0.0) R Obstacle Y2 X2>R Returning Blade Figure 6.07. 6. With these three factors. Here.29. Y1 . the results are only plotted for θ between 0 and 180◦ . 6. Apart from that. a small range of variation had been deﬁned for the parameter space. To investigate this issue.CHAPTER 6. with a minimum value of Cms higher than 0. employing an obstacle plate improves noticeably the selfstarting properties for the classical conﬁguration.2 Optimization Three degrees of freedom are left simultaneously to the OPAL optimizer: X1 and Y1 deﬁne the upper tip of the shielding obstacle. X2 used to modify the position of the shielding obstacle.1. with Y2 /R = −1. As a whole. The conventional Savonius turbine shows a very large variation of the static torque coeﬃcient as a function of θ. the static torque exerted on a turbine at a ﬁxed angle has been computed by CFD as a function of this angle θ.4. Due to periodicity.1. the negative torque region completely disappears. These computations demonstrate that the obstacle plate has a considerable and positive eﬀect on the static torque coeﬃcient for the classical conﬁguration.
068 at λ = 0. As a whole.4 and X2 /R = −1.5 o Y1/R=0. the optimal point found by the optimization procedure corresponds to an absolute increase of the power coeﬃcient by 0. Table 6. This means a relative increase of the performance (measured by the power output coeﬃcient) by 27.017 Y1 /R −0.7.7. T. Y1 and X2 .88 −1.017 The results presented in Fig. Of course. optimum angle β and corresponding optimal power coeﬃcient Cp are listed in Table 6.76.30 indicate that the considered objective is considerably inﬂuenced by the three free parameters X1 .3% for the twoblade Savonius turbine. When compared with the standard Savonius turbines (without shielding obstacle).7.8.88 −1.2 0 0. it must be checked that the shielding obstacle cannot come into direct contact with the rotor ( X1 > R and X2 > R).7: Acceptable range for the input parameters Parameter minimum maximum X1 /R −1. b=68.6 0.29: Static torque coeﬃcient as a function of the rotor angle θ for three diﬀerent values of Y1 choosing X1 /R = −1. Table 6. a truly optimal solution can be obtained considering a large set of possible parameters.5 Static Torque Coefficient (Cms) 1 0.0 . b=60. . The corresponding positions of the obstacle all lead to conﬁgurations that shield partially the returning blade. 6.0 X2 /R −1. 210 diﬀerent geometrical settings have been evaluated by CFD.8 0. In this manner. when choosing the parameter space. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 107 Exp. 6.CHAPTER 6. requesting 24 days of computing time on a standard PC. compared to the conventional Savonius design without any obstacle. Hayashi et al (2005) (no obstacle) Our CFD Results: Savonius without obstacle o Y1/R=0.4 0.47. The optimum obstacle position. all corresponding parameters being connected by a thick red line) can now be identiﬁed for λ = 0. The optimal conﬁgurations (highest point in the right column in Fig.30.2 0 30 60 90 120 150 180 Rotation angle(q) Figure 6.24 .88 0. b=73 o Y1/R=0.
09993 Note that we have been at ﬁrst surprised to obtain an optimal geometry leading to β > 90◦ for the twoblade turbine. as shown in Figure 6. for most conﬁgurations associated with β < 90◦ .88 X1/R 0.3 Oﬀ design performance The performance of the optimal conﬁguration has been ﬁnally computed for the full range of useful λvalues.88 Y1/R 1.23830 Y1 /R −0.1. Cp Twoblade Savonius turbine X1 /R −1.182 Without obstacle 1.2503 X2 /R −1. The . exempliﬁed in Fig.4.31.83◦ is indeed optimal for the advancing blade in the relative reference frame.30: Input parameters of the optimization and power coeﬃcient represented using parallel coordinates. the ﬂow behind the shielding obstacle points partly toward the returning blade and thus increases the reverse moment. The parameters of the optimal conﬁgurations are connected with a thick red line.133 Power coeff. compared to the conventional Savonius turbine without obstacle.88 X2/R 0. Figure 6. Table 6. 6. 6.8: Optimal conﬁgurations (obstacle position and angle) Design parameter value β power coeﬀ.0 1. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 108 Optimal configuration 1. this eﬀect is reduced for the optimal conﬁguration.83◦ ) 0.2503 0. After a thorough ﬂow analysis. This ﬁgure demonstrate that the improvement of both torque coeﬃcient and power output coeﬃcient is observed throughout for all values of λ. The optimal geometry of the shielding obstacle is of course highly dependent on the speciﬁc rotor conﬁguration. The power coeﬃcient of the conventional turbine is shown with a black circle. it is possible to understand ﬁnally that the ﬂow direction induced by the obstacle at β = 100.017 0.CHAPTER 6.45390 (100. Furthermore.32 .017 0.
which measures the selfstarting capability of the turbine. more expensive and heavier than the original system. The study now aims at improving further the output power of the Savonius turbine as well as the static torque. the eﬃciency of the classical Savonius turbine has been increased by placing in an optimal manner an obstacle plate shielding the returning blade. Six free parameters are considered in this optimization process.4. see Fig. the improved power and torque coeﬃcients should easily compensate these drawbacks within a short time after installation. eﬃciently and automatically controlled. In order to achieve both objectives. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 109 Figure 6. the geometry of the blade shape is now optimized in presence of the obstacle plate.4. 6.2. YP 2 .4. For this purpose.2 Optimal blade shape with obstacle plate In the last section. absolute gain for Cp increases even slightly with λ at ﬁrst. XP 2 . six parameters are considered (XP 1 . As a whole. YP 1 . the relative increase being highest for the largest values of λ considered in the present study.1.7: Zoom on the vicinity of the turbine (full CFD domain is much larger). In this manner the orientation of the system can be simply. many existing systems already rely on a tail vane for optimal alignment into the wind direction. XP 3 and YP 3 .1 Optimization The free design variables considered for the optimization describe the blade skeleton line for a constant blade thickness of 2 mm. A similar technical solution would be used for the Savonius turbine using the obstacle. 6.31: Instantaneous ﬂow structure when the advancing blade is in vertical position for the optimal conﬁguration at λ = 0. Therefore.4 Practical realization From the technical point of view.33). 6. 6. the optimized conﬁguration is only slightly more complex.CHAPTER 6. The obstacle plate is kept ﬁxed in the .
9. b: power coeﬃcient.2 1. optimal position identiﬁed in Section 6.8◦ . see Fig. The parameter space considered in the optimization has been deﬁned as documented in Table 6.35 0. since the obstacle position is ﬁxed.8 1 1.32: Performance of the optimized conﬁguration (red line) compared to the conventional twoblade Savonius turbine without obstacle (black line): a: torque coeﬃcient.1 0.45 0.2 60 0.5 0. These domains are selected to prevent any domain overlap along the Y direction and to keep realistic blade shapes.2383. X2 /R = −1.4539. The reference point of the parameter space is point P0 .05 0 0.1 20 0. The six shape parameters are suﬃcient to determine uniquely the geometry of the system.6 Torque coefficient (Cm) 0.2 0.4 0. semicylindrical shape with radius r as Relative incease (%) 0. The corresponding relative increase is shown with blue stars.4 Speed ratio (l) (b) 0. 6.05 0 0.2 0.25 0.4 0.28).25 .1 (X1 /R = −1. The objective function considers only one output of the simulation.3 100 % Power coefficient (Cp) 80 0.CHAPTER 6. Y1 /R = −0.2 0 1.4 0.8 1 1.15 40 0.2 0.3 0.0999 and Y2 /R = −1. which is the center of the original.4 Speed ratio (l) Figure 6. that should be maximized as far as possible: the power coeﬃcient Cp . SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 110 Savonius without obstacle Savonius with obstacle (optimal configuration) %Relative increase (a) 0.15 0. which leads to an angle β = 100.6 0.4.1770.
10: Optimal conﬁguration YP 2 /r XP 3 /r YP 3 /r Optimal power coeﬀ. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 111 P4 P3(XP3.0386 0.3940 At λ = 0. 6.94 −0.24 0. requesting one and a half month of total computing time on a standard PC.53 1.9: Acceptable range for the input parameters for the blade shape Parameter Minimum allowed Maximum allowed XP 1 /r 0.2 YP 3 /r 0. As a whole. The optimum point positions and corresponding optimal power coeﬃcient Cp are listed in Table 6.24 0.33: Schematic description of the free optimization XP 1 .6067 0. all corresponding parameters being connected by a thick red line) can now be identiﬁed for λ = 0.CHAPTER 6. YP 1 .6357 0. Table 6. XP 3 and YP 3 .116 compared with the conventional Savonius turbine.94 The results presented in Fig. The optimal conﬁguration (highest point in the right column in Fig.YP3) Y Fixed points Blade center (P0) r X P1(XP1.YP1) Variable points r P2(XP2. and thus by the blade shape.47 YP 1 /r −0.34. Cp −0.298 XP 1 /r 0.6389 0.YP2) P5 Figure 6. parameters shown in Fig.7.24 XP 2 /r 0. respectively by 0. XP 2 .24 XP 3 /r 0. YP 2 . 6.24 1. During the calculations.03. XP 1 . YP 2 . Table 6.7 the optimal point found by the optimization procedure corresponds to an absolute increase of the power coeﬃcient by 0.0475 compared with the conventional Savonius rotor .10. 6.34 indicate that the considered objective is considerably inﬂuenced by the six free parameters. a circular turbine shaft is included with a ﬁxed radius Rsh computed from Rsh /R = 0. 210 diﬀerent geometrical settings have been evaluated by CFD. XP 3 and YP 3 used to modify the blade shape. XP 2 . YP 1 .6909 YP 1 /r XP 2 /r 0.2 YP 2 /r −0.33.24 1.
4.182 Savonius without obstacle 0.24 0.9% for the optimum shape with obstacle plate. As a consequence. Indeed. compared to the classical Savonius turbine. the modiﬁed shape exploits best the ﬂow redirection induced by the obstacle. (semicylindrical blade shape) with obstacle plate.CHAPTER 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 112 Optimal configuration X1/r 1.24 0. this means a relative increase of the performance (measured by the power output coeﬃcient) by 38. the modiﬁed shape leads to an increase of the positive moment of the advancing blade. simultaneously. As a whole. The parameters of the optimal shape are connected with a thick red line. Note that this new shape (Fig. the reverse moment is reduced by the shielding obstacle.1 Figure 6.250 Classical Savonius with obstacle 0. as shown in Fig.36.24 X3/r 1. The relative performance increase compared to the standard Savonius conﬁguration is always higher than 30% in the eﬀective operating range.298 0. The absolute gain for Cp and Cm is even higher for lower λvalues.94 Cp 0.2 Y3/r 0.34: Input parameters of the optimization and power coeﬃcient represented using parallel coordinates.53 0. The power coeﬃcient of the conventional turbine is shown with a gray circle. 6.24 0.94 0.24 X2/r 1. 6.2. 6.2 Oﬀ design performance The performance of the optimal conﬁguration has been ﬁnally computed for the full range of useful λvalues.35) is only optimum in combination with the employed obstacle plate. .2 Y2/r 0.3 0. The power coeﬃcient of the conventional turbine (semicylindrical shape) with obstacle plate is also shown with a black circle.24 0.47 Y1/r 0. This ﬁgure demonstrates that the improvement of both torque coeﬃcient and power output coeﬃcient is observed throughout for all values of λ. compared to the conventional Savonius design without obstacle.
In the present section we will thus investigate numerically the eﬀect of two combined guiding plates: a deﬂector plate is employed to obtain the best possible ﬂow conditions for the advancing blade.2. 6. 6.35: Optimum conﬁguration (right) obtained with the optimization procedure compared to the classical Savonius turbine (semicylindrical shape: left). A selfstarting capability (Cms > 0) is always obtained at any angle in both cases. which is a major advantage.3 Selfstarting capability The static torque exerted on the turbine at a ﬁxed angle has been computed by CFD as a function of this angle θ. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 113 Advancing blade Wind Wind Returning blade Returning blade Classical Savonius turbine obstacle Optimal shape Figure 6. optimal conﬁguration the negative torque region completely disappears.38) . these computations demonstrate that the modiﬁed blade shape has a considerable and positive eﬀect on the static torque coeﬃcient. simple guiding plates seem to be the best compromise between the increase of eﬃciency and the increase of cost and complexity.37.2.4. Apart from that. as a function of θ. with a minimum value of Cms higher than 0. the evolution as a function of θ is similar to that obtained with the conventional shape involving an obstacle plate.CHAPTER 6. 6.3 Optimal Savonius turbine with two guiding plates From the summary of the most important modiﬁcation proposals listed in Chapter 3. 6. Therefore.37 shows the obtained static torque coeﬃcient Cms obtained for the optimal conﬁguration compared to the classical turbine with obstacle. Modiﬁcations should not involve an exceedingly complex or expensive design. Figure 6. In the present. the highest advantage of the Savonius turbine is its robustness. The experimental results of [38] for a conventional turbine are also shown for comparison in Fig. while an obstacle plate shields partly the returning blade (Fig. Compared to the standard Savonius without obstacle plate.4.
1 0 0.CHAPTER 6. length and angles of these guiding plates (obstacle and deﬂector) as shown in see Fig.8 1 1.2 0.32 100 % Power coefficient (Cp) 0. 6.4 0.4.7 Torque coefficient (Cm) 0.2 1.6 0.2 0 1.4 0. b: power coeﬃcient.1 Optimization The optimization work will be carried out for obstacle and deﬂector simultaneously.12 0.36: Performance of the optimized conﬁguration (red line) compared to the conventional Savonius turbine with and without obstacle plate (blue and black. Xd1 .28 80 0. respectively): a: torque coeﬃcient.04 0 0. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 114 Optimal shape (noncylindrical blade) Savonius with obstacle Savonius without obstacle %Relative increase (a) 0. The parameter spaces considered in the optimization have been deﬁned as documented in Table 6.39 indicate that the considered objective is considerably inﬂuenced by the eight free parameters.08 20 0. The results in Fig.4 40 60 Speed ratio (l) Figure 6. Xd2 . which together are suﬃcient to obtain clearly the position. Y1 . The corresponding relative increase compared to the standard conﬁguration is shown with green line. Y2 . Yd1 . Xd1 .38.4 0.3 0. Xd2 and Yd2 ). X2 .4 Speed ratio (l) (b) 0. Yd2 . Y1 .5 0. 6.6 0.3.11.2 0.24 0.2 0.16 0. 6. Yd1 . with eight free space parameters (X1 .6 0.2 0. and thus by Relative incease (%) . X1 .8 1 1. X2 and Y2 .
as shown in Fig.CHAPTER 6.15◦ for obstacle and deﬂector. angles and lengths of the guiding plates. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 115 Exp. y Xd2 Xd1 Deflector Yd2 Wind Y1 Obstacle Yd1 Advancing Blade X1>R X (0.0) Y2 R X2>R Returning Blade Figure 6. the experimental results of [38] are also shown for comparison. .8 0. 210 diﬀerent geometrical settings have been evaluated by CFD. The optimum point positions are listed in Table 6. Hayashi et al (2005) (no obstacle) Static Torque Coefficient (Cms) 1. For this last case.2 1 0.4 0. requesting 40 days of total computing time on a standard PC. the positions.6 0.38: Schematic description of the geometry of Savonius turbine with frontal guiding plates.37: Static torque coeﬃcient Cms as a function of the ﬁxed rotor angle θ for the optimal conﬁguration compared to the standard Savonius turbine with and without obstacle plate. These optimum parameters lead to the corresponding angles of 92. 6. T.782.094 and Ld /d = 0.2 0 Our CFD results: without obstacle Classical Savonius with obstacle Optimal shape Negative torque (no selfstarting) 0.40.2 0 30 60 90 120 150 180 Rotor angle (q) Figure 6.35◦ and 82. As a whole.12. respectively and lead also to lengths of the guiding plates Lo /d = 1.
9 1.9 Y2/d 0.87 X1 /d −1.167 compared with the conventional Savonius turbine.0 Yd1/d 1. 6.182 1. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 116 Table 6.04 0 Yd1 /d 0.66 −0.3. 6.78 0 X2 /d −1.87 Xd2 /d −1.9 Y1 /d −0.04 0.53 Cp 0.9 1.41.66 0 Yd2 /d 1.7 the optimal points found by the optimization procedure corresponds to an absolute increase of the power coeﬃcient by 0.9 Y1/d 0 X2/d 0.78 1.66 0. The improvement of both torque coeﬃcient .0 Yd2/d 1.9 Y2 /d −1.36 0.66 −0. At λ = 0.18 Figure 6. The parameters of the optimal design are connected with a thick red line. The power coeﬃcient of the conventional turbine is shown with a black circle.3 1.4.53 Optimal conf.39: Input parameters of the optimization and power coeﬃcient represented using parallel coordinates.66 1.87 Xd2/d 0.3 1.87 X1/d 0. as shown in Fig.85% for the optimum design compared to the conventional Savonius design.66 −0.CHAPTER 6. This means a relative increase of the performance (measured by the power output coeﬃcient) by 47.11: Acceptable range for the input parameters (parameter space) Parameter Minimum Maximum Xd1 /d −1.2 Oﬀ design performance The performance of the optimal conﬁguration has been next computed for the full range of useful λvalues.66 0.349 Conventional Savonius 0. Xd1/d 0.66 1.
6. Table 6. X4 and Y4 . Fourteen parameters are suﬃcient to deﬁne clearly the position and shapes of these guiding plates. as shown in Fig.28434 X1 /d 1. and power output coeﬃcient is observed throughout for all values of λ.3.43. Y2 .134209 Y1 /d 0. Wind Lo opt. Figure 6. 6. Optimization is repeated for obstacle and deﬂector simultaneously.066981 Yd1 /d 1. 6. Xd2 Yd2 . X1 X2 . X3 .40: Optimum conﬁguration of guiding plates. The absolute gain for Cp and Cm is even higher for lower λvalues.CHAPTER 6. while Y1 and Yd1 are constant and taken from the previous section. Xd3 .3 Proﬁled guiding plates There is no reason to believe that ﬂat guiding plates are the best choice. Therefore we started an optimization for nonﬂat guiding plates.12). The results are shown in Fig. More than 200 diﬀerent geometrical settings have . with fourteen free space parameters describing the optimum shape of both guiding plates.12: Optimum parameters of guiding plates Xd1 /d 0.42.509671 Xd2 /d 0. Xd4 .286166 X2 /d 1.17377 Yd2 /d 2.13. The relative performance increase compared to the standard Savonius conﬁguration is always higher than 35% in the eﬀective operating range.37918 Ld opt. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 117 Table 6. Yd4 . Yd3 . These fourteen parameters (Xd1 . The parameter space considered in the optimization has been deﬁned as documented in Table 6.4. Y3 .0892 Y2 /d 1.
7 Optimal conf.CHAPTER 6.8 20 (b) 100 Power coefficient (Cp) 0 Speed ratio (l) Figure 6.5 0. However.2 1.8 1 1. As a consequence.3.41: Performance of the optimized conﬁguration (green line) compared to the conventional Savonius turbine (blue line): a: torque coeﬃcient. by comparison with the results of the ﬂat guiding plates (see Section 6.8 1 1.6 0.6 1. been evaluated by CFD.4 0.3 0. Relative increase (%) 80 .4 0.6 1.4 60 0. 6. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 118 (a) 0. b: power coeﬃcient.1 0 0.5 0. only ﬂat guiding plates are considered in the ﬁnal optimization step.2 0. requesting 40 days of total computing time on a standard PC.4 1.6 0.2 0.4 0.2 1. cost and weight of the proﬁled guiding plates (Fig.44).1 0 0. Therefore.01).4 1.8 0. the slight improvement in the performance does not compensate the additional complexity.8 Speed ratio (l) 0.4. from the manufacturing point of view.3 40 0.1).6 0.2 0. (Savonius with guiding plates) Conventional Savonius turbine %Relative increase Torque coefficient (Cm) 0. the diﬀerence in power coeﬃcient is only minimal (around 0.6 0.2 0. The corresponding relative increase compared to the standard conﬁguration is shown with black line.
42: Schematic description of the geometry of Savonius turbine with proﬁled guiding plates.5 −1.15 1.5 0 Yd4 /d 1.5 −1.15 Y4 /d −0.5 −0.5 −1. In this aggressive optimization work.13: Acceptable range for the input parameters (parameter space) Parameter Minimum Maximum Deﬂector Xd1 /d −1.5 0 Yd2 /d 2 2.5 0 Xd2 /d −1.4 Xd3 /d −1. a last step involving all characteristic geometrical parameters is now attempted.5 −1 X3 /d −1.5 Obstacle X1 /d −1.5 X4 /d −1. Table 6.0) Y2 Y3 Yd4 R Y1 X3>R X2>R X1>R Returning Blade Figure 6.15 Y3 /d −1 −0. Having now full conﬁdence in the process. we will opti .5 2 Xd4 /d −1.8 X2 /d −1. The eﬀect of blade shape and guiding plates positions will be incorporated simultaneously during the optimization.5 0 Yd3 /d 1.5 0 6.5 Final optimization of Savonius turbine The optimization steps of increasing complexity described in the previous sections seem very promising.CHAPTER 6.15 Y2 /d −1. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 119 Xd1 Xd2 Xd3 Xd4 y Yd1 Yd3 Wind Y4 Yd2 Advancing Blade X4>R X (0.
15 0. The objective function considers only one output of the simulation that should be maximized as far as possible: the output power coeﬃcient Cp . 6. requesting two and a half months of total computing time on a standard PC for this ambitious analysis.5 1 1.5 1.18 Figure 6. all corresponding parameters being connected by a thick red line) can now readily be identiﬁed .5 0 Cp 0 2. YP 2 . 6.5 1.5 1. The parameters of the optimal designs are connected with a thick red line. eight parameters for both guiding plates (X1 .5 1. The optimal conﬁguration (highest point in the right column in Fig. X2 .45c).46. 6. mize sixteen parameters simultaneously. Xd2 and Yd2 . XP 3 and YP 3 Fig.5 0. two ﬁxed points (P4 and P5 ) and three movable points (P1 .4 2 1. YP i ) this means we have another six free parameters (XP 1 . The power coeﬃcient of the conventional turbine is shown with a black circle.46. considered ﬂat for the reasons explained in the previous section.182 1.43: Input parameters of the optimization and power coeﬃcient represented using parallel coordinates. see Fig.5 1. Xd1 . P2 . Every point has two coordinates (XP i . XP 2 . 6. the full proﬁle is reconstructed using standard splines.15 1. Yd1 .15 1. The employed optimization parameters have been listed in chapter 5.14.8 1 0. Y2 .361 Classical Savonius 0.15 1. Optimization results are shown in Fig.CHAPTER 6. Y1 .5 1. The parameter space considered in the optimization has been deﬁned as documented in Table 6.5 0.45a). YP 1 . ( Savonius with curved guiding plates ) Xd2/d Xd1/d Xd3/d Yd2/d Xd4/d Yd3/d X1/d Yd4/d Y2/d X2/d Y3/d X3/d Y4/d X4/d 0 0 0 0. 6. Another two parameters will taken into consideration for internal spaces of the turbine (a and e see Fig. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 120 Optimal conf.37 0. 310 diﬀerent geometrical settings have been evaluated by CFD.5 2 1. The blade shape is described by ﬁve points. P3 ).5 1. Knowing all 5 points.45b).5 1.
04 0 Yd1 /d 0.09 0.24 XP 3 /r 0.78 0 X2 /d −1.2 YP 3 /r 0.87 X1 /d −1.87 Xd2 /d −1.24 0.94 Guiding plates Xd1 /d −1.24 1.9 1. Table 6.28 for λ = 0.7.94 −0.3 1.18 e/d 0.66 −0.44: Optimum conﬁgurations of curved guiding plates.9 Y1 /d −0.2 YP 2 /r −0.05 0.24 0.47 YP 1 /r −0.24 1. leading to the correspond .9 Y2 /d −1.CHAPTER 6.14: Acceptable range for the input parameters (parameter space) Parameter Minimum Maximum Blade shape XP 1 /r 0. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 121 Optimal deflector shape Optimal obstacle shape Figure 6.24 XP 2 /r 0.66 −0. The optimum parameters are listed in Table 6.53 1.66 −0.66 0 Yd2 /d 1.15.53 Internal spaces a/d −0.
As a consequence. This is of course a dynamic process. 6. the modiﬁed shape leads to an increase of the positive moment of the advancing blade.45: Schematic description of the geometry and free optimization parameters.YP2) P5 a Turbine shaft Figure 6. demonstrating that the employed CFD captures all important ﬂow features in the vicinity of the rotor and guiding plates. Indeed.YP1) Variable points r (c) P2(XP2. velocity magnitude and velocity vectors) is shown as an example in Fig. The lengths of the guiding plates are Lo /d = 0.47) is only optimum in combination with the employed guiding plates. simultaneously. explaining the better e . Note that this new shape (Fig. diﬃcult to illustrate in a static ﬁgure. One instantaneous picture of the ﬂow ﬁeld (pressure.47.048. 6.YP3) Y Returning Blade Fixed points Blade center (P0) r X P1(XP1.48 for the ﬁnally optimized design and for the classical semicylindrical turbine with guiding plates. ing angles of 90. 6. the modiﬁed shape exploits best the ﬂow redirection induced by the guiding plates. the reverse moment is reduced by the shielding obstacle and the ﬂow redirected by the deﬂector to the advancing blade. as shown in Fig.13◦ for obstacle and deﬂector. The pressure diﬀerence obtained for the optimal design is much larger. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 122 (a) Xd2 Xd1 Deflector Yd2 Advancing Blade X1>R Wind Y1 Obstacle Y2 Yd1 R X2>R (b) P4 P3(XP3. respectively.7597 and Ld /d = 1.CHAPTER 6.41◦ and 94.
54593 0.24 1.66 0.770346 −0.18 0.436 Classical Savonius 0.0 0.94 1.41882 1.27654 −1.2 0.24 0.9 1.16 0.09 0.54822 0.18286 .01 Figure 6.46: Input parameters of the optimization and power coeﬃcient represented using parallel coordinates.3433 2.04 0.19762 0. The parameters of the optimal design are connected with a thick red line. The power coeﬃcient of the conventional turbine is shown with a black circle.24 0.37885 0.29074 −1.9 1.24505 −0.53 0.05 0.24 0.00635 0.66 0.0 0.3 1.87 0.94 0.28 Cp a/d X3/r X2/r X1/d X2/d X1/r Xd2/d Xd1/d 0.47 0.182 0.4037 −1.24 0.2 0.CHAPTER 6.53 0. (modified Savonius with guiding plates ) Y3/r Y2/r Yd1/d Yd2/d Y1/r Y1/d Y2/d e/d 1.45 1.2828 −0.34849 −0. Table 6.66 0.78 1.16339 −0.9 1.87 0 0.24 1.15: Optimum Blade shape XP 1 /r YP 1 /r XP 2 /r YP 2 /r XP 3 /r YP 3 /r Guiding plates Xd1 /d Yd1 /d Xd2 /d Yd2 /d X1 /d Y1 /d X2 /d Y2 /d Internal spaces a/d e/d conﬁgurations 0. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 123 Optimal conf.
since such a turbine must be able to work also for oﬀdesign conditions. The absolute gain for Cp and Cm is even higher for lower λvalues.2. 6. The experimental results of [38] for a conventional turbine are again shown for comparison. Compared to the classical Savonius. which is an essential property. as a function of θ. .49. 6. Figure 6. A selfstarting capability (Cms > 0) is always obtained at any angle in both cases. the performance of the optimal conﬁguration has been ﬁnally computed for the full range of useful λvalues. 6. the static torque exerted on a turbine at a ﬁxed angle has been computed by CFD as a function of this angle θ. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 124 Ld opt. Figure 6.50 shows the obtained static torque coeﬃcient Cms obtained for the optimal conﬁguration compared to the classical turbine. Therefore.1 Oﬀ design performance It is now important to check how this gain would change as a function of λ. For the optimal conﬁguration the negative torque region completely disappears.47: Optimum design of Savonius turbine with guiding plates. Wind Lo opt. performance.5. with a minimum value of Cms of about 0. compared to the conventional Savonius turbine. This ﬁgure demonstrates that the improvement of both torque coeﬃcient and power output coeﬃcient is observed throughout for all values of λ.CHAPTER 6.2 Selfstarting capability To investigate the selfstarting capability of the system. The relative performance increase compared to the standard Savonius conﬁguration is always higher than 40% in the eﬀective operating range. these computations demonstrate that the new design has a considerable and positive eﬀect on the static torque coeﬃcient.5. as shown in Fig.
b) optimal Savonius with optimal guiding plates. .CHAPTER 6.48: Instantaneous ﬂow ﬁelds around optimum conﬁgurations (zoom) at the design point (λ = 0. static pressure (Pa). SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 125 (a) (b) Pressure distribution Velocity vector distribution Velocity distribution Figure 6. a) classical Savonius with optimal guiding plates.7). velocity and velocity vector magnitudes (m/s). Note that the color scales are identical to facilitate comparisons.
T.2 0 30 60 90 120 150 180 Rotor angle (q) Figure 6. (modified design) Conventional Savonius turbine % Relative increase (a) 0.6 Preliminary experimental tests in windtunnel It is now important to check the CFD results experimentally.2 1.6 1.1 0 0.4 1.5 0.1 0 0. we have constructed two small models for Savonius turbine.8 20 0 Speed ratio (l) Speed ratio (l) Figure 6. 6.2 0.2 1 0.6 1.2 0.2 0 Negative torque (no selfstarting) 0.4 0. (modified design) Conventional Savonius turbine 0.6 0.8 1 1. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 126 Optimal conf.7 Optimal conf.6 0. one model for the conventional two Relative increase (%) 80 .50: Static torque coeﬃcient Cms as a function of the ﬁxed rotor angle θ for the optimal conﬁguration compared to the conventional Savonius turbine.4 1.4 0.3 0.8 1 1. For this last case.5 0. Hayashi et al (2005) (no obstacle) Our CFD results: Optimal modified Savonius with guiding plates Savonius without guiding plates Static Torque Coefficient (Cms) 1.4 1. the experimental results of [38] are also shown for comparison. b: power coeﬃcient. Exp.6 0.3 40 0.6 1.4 (b) 100 Torque coefficient (Cm) Power coefficient (Cp) 0.4 0.8 60 0.6 0.2 1.8 0.4 0. The corresponding relative increase compared to the classical conﬁguration is shown with blue line.CHAPTER 6.8 0.2 0. Therefore.49: Performance of the optimized conﬁguration (red line) compared to the conventional Savonius turbine (black line): a: torque coeﬃcient.2 0.6 0.
until a nominal wind speed of 20 m/s was achieved.51). First results (Fig.51: Model installation with open wind tunnel. Technical diﬃculties (limited accuracy of the torque meter. The main objective of this work is to check that the optimized design is superior to the original design. the load was progressively increased. Before each measurement the oﬀ set of the torque sensor was set to zero. rapid wear of the employed brake) have prevented up to now a more thorough comparison. ﬁnetuned by a screw. The wind tunnel operates with open or closed test section. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 127 blade Savonius and an other for the optimized shape with shielding obstacle. To get the torque / speed curve. but suﬀers from a poor eﬃciency. selfstarting. Five test series were run for each rotor design. the signal of the torque sensor was transmitted to a computer. 6.CHAPTER 6. The software used for both tasks is LabView. velocity up to nearly 60 m/s is possible. 6. Once the rotor speed reaches a constant value. Installation of Savonius turbine at the test section Figure 6. The setting of the load is carried out by a brake handle. which also is used to control the wind speed of the wind tunnel.52) indeed show increased power coeﬃcient and torque coeﬃcient for the optimized design. The maximum ﬂow rate for open test section is 40 m/s (Fig. output torque and rotation speed of the rotor have been recorded at every load to get performance curves for the two designs. When closed.7 Conclusions on Savonius turbine The conventional Savonius turbine is a promising concept for smallscale windenergy systems. Therefore. the major objective of the present . vibrations of the setup. The model has been submitted to increasing wind speed. 6. Then.
in order to obtain the best possible performance.6 0. This optimization procedure is able to identify considerably better conﬁgurations than the conventional threeblade Savonius turbine.25 cp2 cp3 cp4 cp5 Trend. The best one leads in particular to a relative increase of the power output coeﬃcient by 57% at λ = 0.15 0. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 128 cp1 0. Therefore.4 0. as measured by the power coeﬃcient Cp .4 1. thus obtaining a higher eﬃciency and better selfstarting capability.4 1. A performance gain of at least 25% is found for the full operating range of the conventional design.8 1. gemittelt Mean value cp1 0.2 1.35 0.9.6 a) Conventional Savonius turbine b) Optimized Savonius turbine Figure 6. all the geometrical parameters are simultaneously taken into consideration during optimization. At the same time.05 0.05 0.15 0.39 is obtained for λ = 0.2 1. For this purpose.25 cm Cm cp2 cp3 cp4 c p5 Trend.20 0. position and angles of the guiding plates and gap width have been optimized in a fully automatic manner.30 0. This positive eﬀect is also observed for the torque coeﬃcient.15 0. • Threeblade Savonius turbine After some preliminary steps.15 0.10 0. leading to higher values of the power coeﬃcient and of the static torque.7.0 ? 1.6 0.6 0.6 0. gemittelt cp1 0.40 0.4 0.8 1.05 0.6 0.52: Selected experimental results [64] a) conventional Savonius turbine.35 0.20 0.10 0.25 cp2 cp3 cp4 cp5 Trend.4 0. the blade shape.00 0.CHAPTER 6.2 1. b) Optimized design.0 ? 1.00 0. The optimal design still ensures selfstarting capability for all rotating angles.2 1.5. gemittelt Mean value Cm cm 0.4 1. the operating range is extended up to λ = 1.00 0.0 ? 1.8 1.8 1.4 1.4 0.25 cp2 cp3 cp4 cp5 Trend.10 0. A peak power coeﬃcient of Cp 0.40 0.20 0.05 0. while all geometrical conﬁgurations are evaluated by CFD.20 Cp cp 0. many designs have been successively introduced and optimized in this chapter. The optimization relies on evolutionary algorithms.10 cp Cp 0. gemittelt Mean value 0.0 ? 1.00 0. .6 0. study was to identify an optimal design.30 0.6 cp1 0.
as well as eﬃciency and operating range.48 is obtained for λ = 0. 6.4).7. For the best design. This optimization procedure is able to identify a tremendously better conﬁguration than the conventional Savonius turbine.7.3 ≤ λ ≤ 1.53. the twoblade design appears superior to the threeblade conﬁguration in terms of low weight. cost. The optimal design shows a much larger range and still delivers power at λ = 1. at the diﬀerence of the conventional Savonius.CHAPTER 6. .53: Successive optimization steps for the twoblade Savonius turbine.44 thanks to optimization. Therefore. this optimal conﬁguration appears indeed to be very promising for lowpower wind energy generation in urban areas. the power coeﬃcient has been increased from Cp 0. A performance gain of at least 35% is found for the full conventional operating range (0. The optimal design furthermore leads to selfstarting capability at any angle. with only a very slight advantage for the threeblade design. This positive eﬀect is also observed for the torque coeﬃcient. At λ = 0.7. The successive steps of the optimization are documented in Fig. the optimized twoblade conﬁguration should indeed be very useful to supplement wind energy conversion. leading in particular to a relative increase of the power output coeﬃcient by 58% at λ = 0. Therefore. Here again. a peak Cp of 0.9. preliminary steps have been followed by an optimization involving all important parameters. Selfstating capabilities are similar. SAVONIUS TURBINE: SINGLEOBJECTIVE OPTIMIZATION 129 • Twostage Savonius turbine Figure 6. and supporting previous ﬁndings from the literature. approaching Betz’ limit.18 to Cp 0. As a whole.
91. and t is the maximum halfthickness as a fraction of the chord (so that 100 t gives a half of the last two digits in the NACA 4digit denomination).Chapter 7 Wells turbine: Concurrent optimization 7. The formula for the shape of a NACA 00XX foil. it is interesting to optimize both designs separately. As an illustration.1 Introduction Wells turbine is a selfrectifying air ﬂow turbine employed to convert the pneumatic power of the air stream induced by an Oscillating Water Column into mechanical energy. There is nev130 . starting with the simple conﬁguration.1015 (7. Both monoplane and twoplane Wells turbines are considered in the literature and show diﬀerent advantage and drawbacks. 7.3517 + 0.g. is x x 2 x 3 x 4 ct x 0.2 Optimal monoplane Wells turbine Reference investigations indicated that NACA 0021 airfoil proﬁles (21% thickness) lead to the best performance for conventional monoplane Wells turbines [93]. [73. As a consequence..1) 0. usually a low aerodynamic eﬃciency and a limited range of operation due to stall. Most investigations pertaining to Wells turbines have considered NACA 0012. NACA 0015. with “XX” being replaced by the percentage of maximum thickness to chord length c. y= 7. Standard Wells turbines have several wellknown disadvantages: a very low tangential force. NACA 0018 and NACA 0021 (e.126 − 0. 81.2969 − 0. Fig. 93]). All the theoretical and experimental investigations listed in the section 3.2 c c c c c where x is the position along the chord from 0 to c. a high undesired axial force. leading to a low power output from the turbine.2.2843 − 0.3 only considered the performance of Wells turbines using standard symmetric airfoils of type NACA 00XX.1 shows NACA 0015 and NACA 0021. y is the halfthickness at a given value of x (centerline to external surface).
maximal tangential force coeﬃcient and eﬃciency). the free design variables considered for the optimization will be the shape of the blade using a constant solidity (s = zc/[πrt (1 + h)]) where h is the ratio between hub radius rh and tip radius rt (Fig.05c 0.6 0. An alternative geometry might be much better.1.15 0.2). as proposed by [110]). and is deﬁned for a negligible density change as: F T ut (7.1: Standard airfoils NACA 0015 and NACA 0021. 7.8 0.1 0.1 0. as deﬁned by Eq.6c 0.05 0. The objective function contains simultaneously two outputs of the simulation.e. this ﬁrst study considers only monoplane Wells turbines (the original design) and a constant turbine solidity (s = 0. In the present section.8c x 1 c Figure 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 0.1c 0.05 0. Rotor hub on shaft rh Sym. while taking into account the mutual interaction eﬀect between the blades. Airfoil Blade sp an (b ) rt Rotation Chord (c) FA FT Oscillating flow FA Forces resolved in direction of rotation Figure 7. the present section now concentrates on the optimization of a symmetric airfoil shape.2 0. leading to the best possible performance of a Wells turbine (i. As a consequence. ertheless no proof that NACA proﬁles.4 0.67. 7.2) η= ∆p0 Q .2: Axial and tangential forces acting on a Wells turbine.CHAPTER 7. which is inversely proportional to the axial force coeﬃcient. and the turbine eﬃciency η.15c 0. in particular for such very speciﬁc applications.2c 131 NACA 0015 NACA 0021 y c t 0.. Due to the complexity of the underlying optimization procedure. automatically lead to the best possible performance. that should both be maximized as far as possible: the tangential force coeﬃcient CT .15c 0 0 0.15 0.05c 0 0 0.1c 0.4c 0.
Every point P2 to P12 has two coordinates (Xpi . P4 . The curve is represented mathematically by a polynomial of degree one less than the order of the curve. NURBS). As explained previously. P9 . The parameter space considered in the optimization has been deﬁned as documented in Table 7. The corresponding parameter spaces have been selected to cover all usual NACA 00XX. while avoiding collisions between reference points and keeping acceptable geometries. Then. based on the proﬁle NACA 0021 for a ﬁrst guess. 7. P7 . 7. In the present case the outer boundary of the airfoil (or airfoil shape) is constructed with thirteen points.3. To illustrate the optimization. this means that the spline order is 13 in our case. P11 and P12 ) as shown in Fig.2.1. 12. a ﬁxed angle of incidence α = 8◦ (ﬂow coeﬃcient φ = 0.CHAPTER 7.3. origin of the cartesian coordinate system.0).1 and illustrated in Fig.14) is considered. . Knowing the exact position of these 13 points. only symmetric blades are considered in what follows. 7. it is now possible to start the optimization procedure. P6 . the obtained face is mirrored to obtain the full symmetric airfoil. The employed optimization parameters are listed in Table 5.3. The order of a NURBS curve deﬁnes the number of nearby control points that inﬂuence any given point on the curve.2. For conﬁgurations involving concurrent objectives. . Evolutionary Algorithms are particulary robust and have therefore been used in the present study. The reference point is point P1 (0. WELLS TURBINE: CONCURRENT OPTIMIZATION 132 with Q the volumetric ﬂowrate through the turbine.3. Ypi ).1 Optimization of airfoil shape After having checked the accuracy of an individual evaluation relying on CFD as described in section 5. where i = 2 .3: Allowed parameter space for the moving points P2 to P12 . P8 . Twentytwo free parameters are varied simultaneously by the OPAL optimizer. the full proﬁle is ﬁnally reconstructed for one face of the airfoil using standard splines (Nonuniform rational Bsplines. two ﬁxed points (P1 and P13 ) and eleven variable points (P2 . . P3 . P5 . P10 . Figure 7. explaining the diﬃculty of the process.
052 0.4(a) the performance of the standard airfoil NACA 0021 (tangential force coeﬃcient and eﬃciency (CT .11 YP 10 /c 0.035 P9 XP 9 /c 0. which is not a complete surprise since the tangential force appears on the numerator in Eq.75 0.CHAPTER 7.65 YP 3 /c 0. (CTA . the increase in tangential force coeﬃcient (higher power output) appears to be more signiﬁcant and valuable than the very slightly increased eﬃciency.154 YP 12 /c 0.85 0.55 YP 2 /c 0. Therefore.45 YP 7 /c 0. the most interesting point is globally Point A with (CTA . Two optimal conditions are ﬁnally found: Point A. ηA ) =(0. . located in the upperright corner (marked by a red square in dashed line).143 P3 XP 3 /c 0.85 YP 5 /c 0.4(a).519).2 0. the performance of the standard airfoil NACA 0021 is also plotted for comparison.015 0.066 YP 9 /c 0. the optimization process thus involves twenty two parameters (or degrees of freedom) Xpi and Ypi with i = 2 . When considering now only the best conﬁgurations of Fig.043 0. 7.1 YP 4 /c 0.1: Parameter space for the moving points P2 to P12 for monoplane Wells turbine.1 0.034 0. 7.026 0.4(a) shows all evaluation results.1281. 615 diﬀerent conﬁgurations have been ﬁnally tested by the optimizer. Globally.2 P11 XP 11 /c 0.066 P6 XP 6 /c 0. WELLS TURBINE: CONCURRENT OPTIMIZATION 133 Table 7.75 P4 XP 4 /c 0.65 0.152 As a whole.1325.3 P12 XP 12 /c 0.5187) (highest tangential force).4(b). The results presented in Figs. as documented in Fig.043 YP 8 /c 0.5197) (highest eﬃciency). 0. ηB ) =(0.025 0. the two considered objectives are not fully concurrent but increase simultaneously. Figure 7. As a whole. Point Parameter Minimum Maximum Point Parameter Minimum Maximum P2 XP 2 /c 0. 0. Figures 7. ηA ) ≈ (0.11 P10 XP 10 /c 0. 12 and two objectives (eﬃciency and tangential force coeﬃcient) that should be simultaneously maximized in a concurrent manner. the two objectives (tangential force coeﬃcient and eﬃciency) become indeed slightly concurrent and cannot be optimized simultaneously. and Point B.0. η) = (0.5). 7.(CTB .1163.05 0.4(a) and 7. Here again. close to the middle of the parameter space (thick dashed blue line).01 0.015 P8 XP 8 /c 0.5(b) indicate by parallel coordinates .015 0.1325. By analyzing in detail the resulting geometries and considering daily engineering purposes.2 deﬁning the eﬃciency.143 YP 11 /c 0. .034 0. In Fig.035 P7 XP 7 /c 0.95 YP 6 /c 0.5(a) and 7.45 0.5109)) is also plotted for comparison.023 0.039 0.043 0.094 P5 XP 5 /c 0. leading to 15 days of total computing time on a standard PC. 7. For the last percent of performance improvement. 7. a more complex picture appears.55 0. The results of the optimization process can be usefully visualized in a diﬀerent manner using parallel coordinates (Fig.054 0. 7.35 0.4(b) indicate that the two considered objectives are indeed considerably inﬂuenced by the airfoil shape.0.
a) for all computed conﬁgurations b) for the best conﬁgurations (i.16 Tangential force coefficient (CT) (b) 0. P12 ).1275 0.4 0. .2% for the present ﬂow coeﬃcient. the X and Y coordinates of the eleven moving points (P2 . The geometrical parameters corresponding to the optimal shape are listed in Table 7.51 0.e.52 0. WELLS TURBINE: CONCURRENT OPTIMIZATION (a) 0.. the tangential force coeﬃcient CT is at the same time increased by 0.1305 (A) Efficiency Optimization output Optimum tang.12 0. Optimum efficiency 0.495 0.5 Efficiency 0.129 0.3 Optimization output Monoplane Wells turbine NACA 0021 0. zoom on the upperright part (red square) of a).4: Objectives of the optimization.1335 Tangential force coefficient (CT) Figure 7.515 0.1 0 0. This ﬁgure demonstrates that very diﬀerent shapes have been evaluated on the way toward the optimal solution.132 0.5 0.2.CHAPTER 7.78%) compared to the standard airfoil (NACA 0021).08 0. The optimum conﬁguration (Point A) is indicated by a thick red line. It can be seen that the optimal airfoil shape leads only to a slightly higher eﬃciency (+0.2 0.04 0.505 0.14. However. together with the two objectives.0162. The resulting shape of the optimal airfoil in comparison with the standard .6 134 0. .525 (B) 0. This means a relative increase of 12. equal to 0. force coeff.
a) Xcoordinates of the variable points (P2 . 7. The standard design (NACA 0021) is shown with a thick dashed blue line. An excellent ﬁt (average .5: Input parameters of the optimization and objectives represented using parallel coordinates. b) Ycoordinates of the variable points (P2 . . P12 ). a simple polynomial description of this proﬁle would be helpful for practical purposes. Nevertheless. .6. . NACA 0021 is shown in Fig. the full proﬁle is again reconstructed using standard splines of order 13. P12 ). Knowing all points P1 to P13 . WELLS TURBINE: CONCURRENT OPTIMIZATION 135 Figure 7.CHAPTER 7. The parameters of the optimal shape are connected with a thick red line. .
737973 0.248998 P12 XP 12 /c 0.0214618 136 residual error of 0.898334 0.155326 P11 XP 11 /c 0.3. .033155 YP 8 /c 0. the optimal airfoil shape described by splines (black squares showing the position of the control points) and the corresponding polynomial ﬁt (Eq.405838 0.0666932 0. 7.38%) has been obtained with following polynomial description: Y =A c X c 5 +B X c 4 +D X c 3 +E X c 2 +H X c +K (7.01409362 Optimum shape Polynomial fit of optimum shape NACA 0021 0.0877814 YP 12 /c 0.078402 YP 11 /c 0.2: Optimum Point Parameter P2 XP 2 /c YP 2 /c P3 XP 3 /c YP 3 /c P4 XP 4 /c YP 4 /c P5 XP 5 /c YP 5 /c P6 XP 6 /c YP 6 /c P7 XP 7 /c YP 7 /c shape parameters for monoplane Value Point Parameter 0.4c 0.8891079 0.56625 0.6: Comparison between the original proﬁle NACA 0021 (solid line). Furthermore.469922 0.0678583 P10 0.CHAPTER 7.04c 0 0 0.2c 0. Table 7.560073 −3.6c 0.0269924 P9 XP 9 /c 0.6958588 −5.3.8c c Figure 7. this polynomial removes any possible oscillations of the proﬁle shape induced by the spline description.0608499 YP 10 /c 0.12c 0.798623 0.037401 0.0315091 0.4277515 6. WELLS TURBINE: CONCURRENT OPTIMIZATION Table 7.08c 0.0751153 Wells turbine Value 0. dashed line).3) with the constants A to H listed in Table 7.723109 0.3: Polynomial coeﬃcients of optimal airfoil shape (best ﬁt) A B D E H K 1.0493063 0.0124947 P8 XP 8 /c 0.045006 YP 9 /c XP 10 /c 0.
Twostage Wells turbine have been already investigated experimentally and theoretically [21. All published results concerning the ﬂow ﬁeld around a Wells turbine rotor indicate that a considerable amount of exit kinetic energy is lost with the swirl component of the ﬂow velocity. Using numerical optimization. 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 137 7. 7.3% (Fig. the gap between the two rotors is an important parameter to control performance. twostage Wells turbine constructed from nonsymmetric airfoils based initially on NACA 2421.2%. at least in the absence of guide vanes (see Section 3. as shown in Fig. Therefore. The relative increase is higher than 8. 7.7.3 Optimal twostage Wells turbine with nonsymmetric airfoils In this section we investigate extensively the potential of nonsymmetric airfoil blades to improve the tangential force and eﬃciency associated to a twostage Wells turbine.7b). The corresponding gain varies between 0. 80. Nonsymmetric blades could perhaps allow to increase considerably the power output from the turbine and the global eﬃciency of the system.2 Oﬀ design performance It is important to check how the gain induced by the new airfoil shape would change as a function of the ﬂow coeﬃcient φ. with an average increase of 1% (Fig.8. since such a turbine must be able to work also for oﬀdesign conditions. The absolute gain for CT increases even slightly with φ. .2% and up to 3.CHAPTER 7. 7.7 between the performance of the exact proﬁle described by splines and the associated polynomial ﬁt (Eq. The two stages are mirrored to keep the global symmetry of the turbine as shown in Fig.2. 7. 61. 107]. Therefore. At the same time the eﬃciency of the optimized shape is always higher than for the conventional design. the solidity and the shape of the nonsymmetric airfoils will be optimized in this section. but considering only symmetric airfoils.8% throughout the useful operating range.2. These results demonstrate that the improvement of tangential force coeﬃcient is observed throughout for all values of φ. as requested.7a). The performance of the twostage rotor is modiﬁed by mutual aerodynamic interferences due to the proximity of the two planes. the diﬀerence being lower for large ﬂow coeﬃcients. This kinetic energy can be partly recovered by using a second stage of blades. the system stays globally symmetric. The upstream rotor aﬀects the performance of the downstream rotor by producing a deﬂection of the air stream. 66. 7. Since these two stages are mirrored. with an average gain of 11.3). compared to the conventional turbine based on standard airfoils NACA 0021. No signiﬁcant diﬀerence is observed in Fig. The present work concentrates on a modiﬁed.3). the performance of the optimal shape has been ﬁnally computed for the full range of useful φvalues. 7.
a) tangential force coeﬃcient. Two degrees of freedom are left to the OPAL optimizer: the thickness factors y1 for the upper side of the airfoil and y2 for the lower side (see Fig.48 0.6.42 0.15 0. b) eﬃciency.2 0. The corresponding relative increase is shown with blue line.7: Performance of the spline optimal conﬁguration (red line).25 0 0. 7.44 0.5 Flow coefficient Efficiency 0. To illustrate the process.2 0.4 40 0.3.05 0.21) is considered.8). a ﬁxed angle of incidence α = 12◦ (ﬂow coeﬃcient φ = 0.1 Optimal airfoil shape without mutual interactions In order to assess the potential of nonsymmetric airfoils.15 0.25 0.46 0.1 10 0 0. These two factors are constrained by the user between 0.54 0. 7. so that considerable variations are still allowed (both % Relative increase .5 NACA 0021 Spline optimum shape Fitting optimum shape %Relative increase 50 % 0.4 0.CHAPTER 7. WELLS TURBINE: CONCURRENT OPTIMIZATION (a) 138 Tangential force coefficient (CT) 0.2 20 0.3 NACA 0021 Spline optimum shape Fitting optimum shape Flow coefficient Figure 7.1 0.05 0.52 0.1 0.3 30 0. ﬁtting optimal one (black cross) compared to the conventional Wells turbine relying on the NACA 0021 proﬁle (green line).3 (b) 0. a ﬁrst optimization is carried out without mutual interactions between the blades.2 and 1.
CHAPTER 7. the optimization process thus involves only two parameters y1 and y2 and two objectives (eﬃciency and tangential force coeﬃcient) that should be simultaneously maximized in a concurrent manner.9. All blades of both rotors are always modiﬁed simultaneously and the two rotors are still mirrored to preserve the global symmetry of the turbine. mutual interactions could modify these results and must now be taken into . Figure 7.8: Twostage Wells turbine based on nonsymmetric airfoils for increasing and decreasing thickness). demonstrating the interest of nonsymmetric airfoils.1795 (18% thickening) for the lower face. The resulting shape of the optimal airfoil in comparison with the standard one is shown in Fig. The optimal point found in this ﬁrst study. 7. WELLS TURBINE: CONCURRENT OPTIMIZATION 139 Figure 7.1% and of the tangential force coeﬃcient by almost 6%. corresponds to the scaling factors y1 = 1. As a whole. However.9: Comparison between the optimal shape of the airfoil and the original proﬁle NACA 2421 When compared with the baseline case (NACA 2421) the optimal point corresponds simultaneously to a relative improvement of the eﬃciency by 2.22966 (23% thickening) for the upper face and y2 = 1.
In this section.12b.1 Optimal airfoil shape with mutual interactions Mutual interactions eﬀect The present section describes the optimization of the airfoil shape considering mutual interactions between the blades. The eﬃciency of the modiﬁed Wells turbine decreases very rapidly when increasing the solidity due to high losses.11a).3. But. the inﬂuence of the mutual interaction between blades has been studied using windtunnel measurements or simple theoretical correlations relying on potential ﬂow analysis and singularity theory for ﬂat plate aerofoils in cascade [86]. The results of the previous section indicated that nonsymmetric airfoils are better than symmetric ones. The axial and tangential force coeﬃcients normalized by the corresponding coeﬃcients obtained for an isolated blade (single airfoil) have been computed for a constant gap ratio G = 1 (Fig. According to such studies. The validity of these ﬁrst results is therefore questionable. The results demonstrate that both the normalized tangential and axial force coeﬃcients increase exponentially with the solidity s due to the growing inﬂuence of wake eﬀects. WELLS TURBINE: CONCURRENT OPTIMIZATION account. these results have been obtained neglecting any interactions between the blades. 7. 7. This aerodynamic interaction results from the wakes produced by the preceeding blades and is a function of the angle of incidence of the air ﬂow as well as of the solidity of the blades. 140 7. Increasing the tangential force is a positive aspect. In the past.12a.CHAPTER 7. This decrease in eﬃciency can be explained by the fact that the losses near the turbine hub are considerably higher than near the turbine tip. the interaction eﬀect of the modiﬁed Wells turbine is investigated numerically in a systematic manner. as requested when using power from an Oscillating Water Column. The inﬂuence of the mutual interactions between the blades building one plane (perpendicular to the main ﬂow direction) are considered extensively in what follows.3. 7. the correlation factor m describing the ratio between the lift coeﬃcient considering the interaction and the coeﬃcient of an isolated ﬂat plate is given by: m= 2B πc tan πc 2B (7. we suggest to replace the rectangular shape of the blades by a trapezoidal shape as shown in Fig. in order to increase system eﬃciency. However. 7. since the performance of the twostage Wells turbine is aﬀected by mutual aerodynamic interaction between the blades in a single plane. 7. but an increasing axial force is of course a major drawback.4) where c is the blade chord length and B is the pitch between the blades. A Wells turbine consisting of two mirrored airfoil stages is considered again. Mirroring is needed to keep the global symmetry of the system. This trapezoidal . nonsymmetric blades are considered.2. the tangential force coeﬃcient increases with the solidity (Fig.11b. because the ﬂow passage at the tip is much wider than the ﬂow passage near the hub for a standard rectangular blade shape as shown in Fig. Therefore.10). As the same time.2 7. as shown in Fig.
2 2 G=1 a=8 deg. Therefore.2 1 0. a=10 deg. a=10 deg.6 0.5 0.4 1. shape leads to the same passage at both hub and tip. (b) 0.10: Impact of mutual interaction between blades in the same plane on a) tangential force (left) and b) axial force (right).5 0. as a function of the solidity.4 2.7 0. The best conditions should be obtained when the ratio between hub radius and tip radius equals the ratio between hub chord and tip chord: Rh Rt = ch ct (7.2 0. a=12 deg.7 0.7 0. WELLS TURBINE: CONCURRENT OPTIMIZATION (a) 2. 141 (b) 5 G=1 a=8 deg. this issue is left for future studies.6 0. and consequently such blades should give globally a better eﬃciency along the blade span (isoperformance blade).6 0.8 Cx/Cx0 CT/CT0 3 2 1 0.6 1. 4 1.6 Efficiency CT 0.8 Solidity Solidity Figure 7.5 G=1.4 0. a=12 deg.3 0.5) Checking the performance of this modiﬁed design would unfortunately require threedimensional CFD which are beyond reach when coupled with an optimizer. .0 0. a=12 deg.8 Solidity Solidity Figure 7. a=12 deg.4 0.8 1.4 0. a=10 deg.11: Impact of solidity on the twostage Wells turbine performance considering the tangential force coeﬃcient (a: left ﬁgure) and the turbine eﬃciency (b: right ﬁgure).4 0. 0 0. (a) 0.5 0.4 a=8 deg.CHAPTER 7.5 0.7 0.7 0. a=10 deg.6 0.0 a=8 deg.8 0.6 0.8 G=1.4 0.
P19 . P9 . P2 . The corresponding parameter spaces have been selected to cover all usual NACA airfoils.4 and illustrated in Fig. P22 . 7. P25 . maximal tangential force coeﬃcient and eﬃciency) of a twostage Wells turbine. P20 . P26 . a) Conventional turbine b) Suggestion of isoperformance turbine.67.12: Projected shape of the turbine. The parameter space considered in the optimization has been deﬁned as documented in Table 7.13.3. P23 . two ﬁxed points (P0 and P17 ) and thirty two variable points (sixteen variable points for every face of the airfoil. P21 . while avoiding collisions between reference points and keeping acceptable geometries. P5 . P24 . 7. P14 . Knowing the exact position of these 34 points. P18 . P31 . P6 . 7. P8 .2. leading to the best possible performance (as usual. as proposed by [110]) and a constant gap between the rotors G = d/c = 1.0). P10 . P29 . while taking into account the mutual interaction eﬀect between the blades (see previous section).13. In the present case the outer boundary of the airfoil is constructed with thirty four points. WELLS TURBINE: CONCURRENT OPTIMIZATION (a) Rectangular blade 142 Hub radius (Rh) Tip radius(Rt) (b) Tip chord (ct) IsoPerformance blade Tip radius(Rt) Hub radius (Rh) Hub chord(ch) Figure 7. P30 . . P7 . P28 . P3 . P1 .2 Optimization Starting from NACA 2421. P27 . P4 .0 are again considered. A constant turbine solidity (s = 0. P12 . P11 .CHAPTER 7. the full proﬁle is reconstructed for each face of the airfoil using standard splines (NURBS). P32 and P33 for lower face) as shown in Fig. the present section concentrates on the optimization of a nonsymmetric airfoil shape. The reference point is point P0 (0. P15 and P16 for upper face. origin of the cartesian coordinate system. P13 .
055 P28 XP 28 /c 0.03 P16 XP 16 /c 0.03 P3 XP 3 /c 0.055 0.55 YP 12 /c 0. as .15 0.13 YP 22 /c 0.4453)) is also plotted for comparison.07 0.14 0.55 0.08 0. WELLS TURBINE: CONCURRENT OPTIMIZATION Table 7.045 P29 XP 29 /c 0.07 0.27 YP 9 /c 0.93 0.006 The results are presented in Fig.85 0.12 P10 XP 10 /c 0.08 0.1 P7 XP 7 /c 0.018 0.06 YP 20 /c 0.65 0.35 YP 26 /c 0.CHAPTER 7.08 0. a more complex picture appears.55 YP 28 /c 0.11 P8 XP 8 /c 0. 7.018 YP 2 /c 0.03 0. When considering now only the best conﬁgurations of Fig.17 YP 7 /c 0.075 0.27 0. In Fig.016 P19 XP 19 /c 0.06 0.35 0.018 0.93 0.07 0.045 P22 XP 22 /c 0.018 YP 18 /c 0.025 P20 XP 20 /c 0.08 0. 615 diﬀerent conﬁgurations have been tested by the optimizer.18 0. located in the upperright corner (marked by a red square).65 YP 13 /c 0.12 P11 XP 11 /c 0.15 0.11 P12 XP 12 /c 0.0.23 0.14.45 YP 27 /c 0.032 0.01 P33 XP 33 /c 0.85 YP 31 /c 0.08 P14 XP 14 /c 0.04 0.45 0.75 YP 14 /c 0.08 0.035 P30 XP 30 /c 0.1 P13 XP 13 /c 0.11 P9 XP 9 /c 0.85 YP 15 /c 0.09 0.93 YP 16 /c 0.032 0.75 0.06 0.14.16 0.35 YP 10 /c 0.93 YP 32 /c 0.04 Point Parameter Minimum Maximum P18 XP 18 /c 0.98 0.07 P5 XP 5 /c 0.01 143 the moving points P1 to P16 for upper face and P18 to Maximum 0.23 YP 8 /c 0.75 0.032 YP 19 /c 0.009 0.17 YP 23 /c 0.006 0.45 0.13 0.05 P4 XP 4 /c 0.27 YP 25 /c 0.06 P15 XP 15 /c 0.03 0.27 0. As a whole. Globally.13 0.75 YP 30 /c 0.06 0.016 P 32 XP 32 /c 0.06 P24 XP 24 /c 0.65 YP 29 /c 0.06 0.14a the performance of the standard airfoil NACA 2421 (tangential force coeﬃcient and eﬃciency (CT . 7.65 0.06 P25 XP 25 /c 0.35 0.17 0.45 YP 11 /c 0.025 P31 XP 31 /c 0.08 YP 21 /c 0.032 YP 3 /c 0.13 0.17 0.035 P21 XP 21 /c 0.23 0.08 P6 XP 6 /c 0. leading to 18 days of total computing time on a standard PC.055 P23 XP 23 /c 0.17 0.98 YP 33 /c 0.08 YP 5 /c 0. the two considered objectives are not fully concurrent but increase simultaneously.06 YP 4 /c 0.006 YP 1 /c 0. η )=(0.06 P26 XP 26 /c 0.16 0.13 0.04 0.2074. 7.23 YP 24 /c 0.85 0.076 0.4: Parameter space for P33 for lower face Point Parameter Minimum P1 XP 1 /c 0.13 YP 6 /c 0.06 0.03 P2 XP 2 /c 0.17 0.055 0.06 0.06 P27 XP 27 /c 0.014 0.55 0.
the most interesting point is globally Point A with (CT .8 C Upper Face 0 0 0. Two optimal conditions are ﬁnally found: Point A..4 C 0.4C 0.2212.2 C P9 P10 P11 P12 P13 P14 P15 P16 P17 C c 1 P5 0. It can be seen that the optimal airfoil shape leads only to a slightly higher eﬃciency (+0. a relative increase of more than 7. i. The resulting shape of the optimal airfoil in comparison with the standard NACA 2421 is shown in Fig.0. This ﬁgure demonstrates that very diﬀerent shapes have been evaluated on the way toward the optimal solution. However.04C 0.14.08C 0. with the same two objectives. while Fig.4482) (highest tangential force). the increase in tangential force coeﬃcient (higher power output) appears to be more signiﬁcant and valuable than the very slightly increased eﬃciency. The geometrical parameters corresponding to the optimal shape are listed in Table 7.02C 0.5%) compared to the standard airfoil (NACA 2421). η) (0.13: Allowed parameter space for the moving points.4506) (highest eﬃciency). in previous cases.04 C 0 144 P6 P7 P8 P4 P3 P2 P1 P0 0 0.16. 7. As usual. Therefore.16 C 0.2 C 0.0.15a indicates by parallel coordinates the X and Y coordinates of the eleven moving points of upper face (P1 . .15b shows by parallel coordinates the X and Y coordinates of the eleven moving points of lower face (P18 . WELLS TURBINE: CONCURRENT OPTIMIZATION 0.e. 7. equal to 0. P16 ).6 C 0. with slightly concurrent objectives. Knowing all points P0 to P33 .15).06C 0. the full proﬁle is again .8 C c 1 C P31 P30 P29 P23 P24P25 P26 P27 P28 P32 P33 P17 Lower face Figure 7.12 C 0.017. The optimum conﬁguration (Point A) is indicated by a thick red line. 7.5. (CT . . η)= (0. together with the two objectives. (CT .1C P0 P18 P19 P20 P21 P22 0. The results of the optimization process can be also visualized using parallel coordinates (Fig. .2247.0.CHAPTER 7. P33 ). η)= (0. the tangential force coeﬃcient CT is at the same time increased by 0. Figure 7. . and Point B.7% for the present ﬂow coeﬃcient.08 C 0.225.448).6 C 0.
14: Objectives of the optimization: a) for all computed conﬁgurations.3 Oﬀ design performance The results shown in Fig.6 for the upper face and the lower face. a simple polynomial description of these proﬁles is helpful for practical purposes. 7. The corresponding gain varies between 0. An excellent ﬁt (average residual error of 0.05 0.216 0.225 Tangential force coefficient Figure 7.3 0.222 0. compared to the nonsymmetric turbine based on standard airfoils NACA 2421. Max. WELLS TURBINE: CONCURRENT OPTIMIZATION 0.2.45 Efficiency 0.5%.3 with the constants A to H listed in Table 7.21 0..4 Efficiency 0.1 0.e. i. Nevertheless. b) for the best conﬁgurations.7%.219 0. tangential force coeff. 7.CHAPTER 7.43 0.2% and 0.17 demonstrate that the improvement of tangential force coeﬃcient is observed throughout for all values of φ. 7. with an average .435 Optimization output Max. reconstructed using standard splines.15 0.2 0.2 Optimization output NACA 2421 0. with an average increase of 5.3.25 Tangential force coefficient (b) 0.5 145 (a) 0. At the same time the eﬃciency of the optimized shape is slightly higher than the standard airfoil design. Efficiency 0.213 0.455 Point (B) Point (A) 0.44 0.445 0. zoom on the upperright part (red square) of (a).3%) has been obtained using Eq.1 0 0 0.
083 Figure 7..6 c 0..207 0...445 0.1 c 0. Optimal shape of upper face NACA 2421 Y16/c X16/c CT Eff.. Optimal shape of lower face Y32/c CT Eff..16: Comparison between the original proﬁle NACA 2421 (solid line) and the optimal airfoil shape (dashed line). The parameters of the optimal shape are connected with a thick red line.207 Y17/c X17/c .4 c 0. X32/c 0.8 c c 1 Figure 7.15 c 0.2251 0.CHAPTER 7. .05 c 0.451 0....05 c 0 0. considering mutual interaction between the blades.. WELLS TURBINE: CONCURRENT OPTIMIZATION Y1/c X1/c 146 .445 0. 0.451 0.15: Input parameters of the optimization and objectives represented using parallel coordinates..02 0.2251 0. Optimal configuration NACA 2421 0.2 c 0..1 c 0 0.
the next section considers only symmetric proﬁles.1353 P11 XP 11 /c 0.1229 YP 16 /c 0.1351 YP 20 /c 0.9508 YP 22 /c 0.8019 YP 8 /c 0.88881 −2.0405 YP 12 /c 0.3846 P4 XP 4 /c 0.0149 YP 28 /c 0.0196 P30 XP 30 /c 0.2485 YP 2 /c 0. that might be described with much fewer control points while keeping a larger parameter space.07551 YP 26 /c 0.1417 YP 21 /c 0.6: Polynomial coeﬃcients of optimal airfoil shape (best ﬁt) for twostage nonsymmetric airfoil Wells turbine considering mutual interaction between the blades (upper and lower face) Face A B D E H Upper face −1.6882 YP 7 /c 0.0514 YP 32 /c 0.02274 YP 29 /c 0.06938 P15 YP 4 /c 0.0073 YP 10 /c 0.1372 P18 XP 18 /c 0.03159 YP 30 /c 0.0327 P8 XP 8 /c 0. This optimization study considers the largest number of free optimization parameters (64) ever considered in our group.0584 P6 XP 6 /c 0.0653 P2 XP 2 /c 0.3229 YP 3 /c 0.1795 P19 XP 19 /c 0. WELLS TURBINE: CONCURRENT OPTIMIZATION 147 Table 7.31571 3.1334 YP 23 /c 0.01129 P12 XP 12 /c 0.0527 YP 13 /c 0.1069 P33 XP 33 /c 0.9773 1.06296 0.3936 P22 XP 22 /c 0.0219 P9 XP 9 /c 0.0386 YP 27 /c 0.0709 Table 7.0577 P31 XP 31 /c 0. By analyzing the results we observed that the resulting Pareto frontier was relatively irregular and poorly populated.5078 YP 5 /c 0.1169 P16 XP 16 /c 0. 7.7556 P26 XP 26 /c 0.06611 YP 14 /c 0.1302 YP 18 /c 0.125 YP 24 /c 0.17b).014714 increase of 0.19739 −2.0453 YP 31 /c 0. It is probably better to keep a lower number of free parameters in association with concurrent optimization problems.862 YP 9 /c 0. .68696 P25 XP 25 /c 0.0464 P7 XP 7 /c 0.06117 XP 15 /c 0.02429 Lower face 0.8927 −0.559 YP 6 /c 0.3% (Fig.9365 P28 XP 28 /c 0.06995 P3 XP 3 /c 0.CHAPTER 7.5205 P23 XP 23 /c 0.0377 P14 XP 14 /c 0.18127 1.1478 YP 11 /c 0.247 P20 XP 20 /c 0.8759 P27 XP 27 /c 0.09814 YP 25 /c 0.0121 P10 XP 10 /c 0.18961 YP 1 /c 0.0602 P5 XP 5 /c 0.5748 P24 XP 24 /c 0.58341 −0.02179 P13 XP 13 /c 0.0633 YP 33 /c 0.5: Optimum shape parameters for twostage Wells turbine with nonsymmetric airfoils Point Parameter Value Point Parameter Value Point Parameter Value P1 XP 1 /c 0. The presented results should therefore be considered with caution. For this reason.0689 P32 XP 32 /c 0.0096 P29 XP 29 /c 0.1345 YP 19 /c 0.2783 P21 XP 21 /c 0.0928 YP 15 /c 0.
17: Performance of the optimal conﬁguration (red line).28 Flow coefficient Figure 7. % Relative increase Optimum airfoil NACA 2421 .16 0.4 0.42 Optimum airfoil NACA 2421 0.12 0. maximal tangential force coeﬃcient and eﬃciency) for a twostage Wells turbine.24 0.e. The corresponding relative increase is shown with blue line.CHAPTER 7. 7. compared to the nonsymmetric twostage Wells turbine relying on the NACA 2421 proﬁle (black line).46 0. a) tangential force coeﬃcient.12 0.4 20 0.2 0.08 0.43 0.16 0.8 50 148 Tangential force coefficient 40 0.08 0.2 0.67)) and a constant gap between the rotors G = d/c = 1.41 0.4 Optimal twostage Wells turbine with symmetric airfoils The present section ﬁnally concentrates on the optimization of a symmetric airfoil shape. while taking into account the mutual interaction eﬀect between the blades.28 0 Flow coefficient (b) 0.. WELLS TURBINE: CONCURRENT OPTIMIZATION (a) 0.45 Efficiency 0. leading to the best possible performance (i.2 10 0 0. A constant turbine solidity (s = 0.24 0.18) are again considered.0 (Fig.44 0.6 % Relative increase 30 0. 7. b) eﬃciency.
P11 and P12 ) as shown in Fig. two ﬁxed points (P1 and P13 ) and eleven variable points (P2 . located in the upperright corner (marked by a red circle).4. Knowing the exact position of these 13 points. P4 . In Fig. the full proﬁle is ﬁnally reconstructed for one face of the airfoil using standard splines. P9 . evaluation results. P3 .1 Optimization of airfoil shape The blade shape is now constructed with thirteen points.4351)) is also plotted for comparison.CHAPTER 7. 7.0. leading to 17 days of total computing time on a standard PC. Globally.19. 7.18: Twostage Wells turbine with symmetric airfoils NACA 0021. P8 . When considering now only the best conﬁgurations of Fig. P10 . as documented in . P7 . WELLS TURBINE: CONCURRENT OPTIMIZATION 149 Flow Rotation d Flow Figure 7. 7. P6 . 615 diﬀerent conﬁgurations have been ﬁnally tested by the optimizer. the two considered objectives are not fully concurrent but increase simultaneously.19: Allowed parameter space for the moving points.2104. η)=(0. a more complex picture appears.20. P5 . As a whole.20a the performance of the standard airfoil NACA 0021 (tangential force coeﬃcient and eﬃciency (CT . Figure 7. 7.20a shows all Figure 7.
7.05 0.e. and Point B.4 0. Therefore. 7.1 Optimization output NACA 0021 0 Efficiency Tangential force coefficient (CT) Tangential force coefficient (CT) Figure 7. It can be seen that the optimal airfoil shape leads only to a slightly higher eﬃciency (+0.20: Objectives of the optimization: a) for all computed conﬁgurations. η)= (0.444 0.44 0.22.434 0 0.2384. compared to the conventional . 7.4. Knowing all points P1 to P13.24 0.249. 7. Here again. tang. i.5 0.25 Optimization output Max.3 with the constants A to H listed in Table 7.0.442 0.245 0.. a relative increase of more than 15% for the present ﬂow coeﬃcient.3 0. An excellent ﬁt (average residual error of 0.20b. A polynomial description of this proﬁle would be helpful for practical purposes.0385. 7. 0.0. η)= (0.CHAPTER 7. (CT .1 Oﬀ design performance The results shown in Fig. efficiency Efficiency 0.2489.8. the full proﬁle is again reconstructed using standard splines of order 13.0.443). the most interesting point is Point A with (CT . Max.446 Point (B) Point (A) 0. b) for the best conﬁgurations.7.8%) compared to the standard airfoil (NACA 0021).25 0. The geometrical parameters corresponding to the optimal shape are listed in Table 7. The optimum conﬁguration (Point A) is indicated by a thick red line. the performance of the standard airfoil NACA 0021 is also plotted for comparison (thick dashed blue line). i. force coeff.436 0.. the tangential force coeﬃcient CT is at the same time increased by 0.235 0.15 0. η)≈(0. The resulting shape of the optimal airfoil in comparison with the standard NACA 0021 is shown in Fig.4430) (highest tangential force).438 0. However. Two optimal conditions are ﬁnally found: Point A.1. The results of the optimization process can be also visualized using parallel coordinates (Fig. the two objectives (tangential force coeﬃcient and eﬃciency) become again slightly concurrent and cannot be optimized simultaneously. zoom on the upperright part (red circle) of (a). (CT . WELLS TURBINE: CONCURRENT OPTIMIZATION 150 Figure 7. For the last percent of performance improvement.e.2 0.21).2 0.16%) has been obtained using again Eq.14.1 0. the increase in tangential force coeﬃcient (higher power output) appears to be more signiﬁcant and valuable than the very slightly increased eﬃciency. Here again.23 demonstrate that the improvement of tangential force coeﬃcient is observed throughout for all values of φ.4450) (highest eﬃciency). equal to 0.
CHAPTER 7. The absolute gain for CT increases even slightly with φ.22: Comparison between the original proﬁle NACA 0021 (solid line).01 0.249 0. WELLS TURBINE: CONCURRENT OPTIMIZATION Optimum configuration NACA 0021 151 Y2/c 0.2c 0. The standard design (NACA 0021) is shown with a thick dashed blue line.015 0.1 X5/c 0.445 X3/c 0. 0.4c 0.55 0.12 Y11/c 0.1 0.55 X9/c 0.164 X6/c 0.08c 0.034 0.053 X2/c 0. the polynomial optimal airfoil shape (red line) and optimal shape by splines (black square).2 X7/c 0.65 0.45 X8/c 0.75 X11/c 0. turbine based on standard airfoils NACA 0021.036 Cp Effic.6c 0.013 0.8c c Figure 7.12c 0.02 0.05 0.13 Y10/c 0.153 0.85 X12/c 0.017 0.02 Y3/c 0.16 Y9/c 0.02 0.12 Y5/c Y6/c 0.97 0.041 0.35 Y7/c 0.029 0.024 0.75 0.04 0.02 0.16 Y8/c 0. The parameters of the optimal shape are connected with a thick red line.21: Input parameters of the optimization and objectives represented using parallel coordinates.45 0. The relative increase is higher than 10% throughout the useful operating .35 0.2 0.65 X10/c 0.006 0.85 0. Optimum shape NACA 0021 Polynomial fit of the optimum shape 0.05 0.04 Figure 7.05 X4/c 0.076 Y4/c 0.04c 0 0 0.076 Y12/c 0.
3825 YP 7 /c 0.01188 P8 XP 8 /c YP 2 /c 0.18.03242 YP 8 /c P3 XP 3 /c 0.23b).03056 P9 XP 9 /c YP 3 /c 0.29647 P12 XP 12 /c YP 6 /c 0.89755 0.07726 YP 11 /c P6 XP 6 /c 0. the diﬀerence disappears for large ﬂow coeﬃcients due to stall.5233101 −2. This optimization procedure is able to identify a considerably better conﬁguration .027875 152 Table 7.01314256 range.7: Optimum shape parameters for twostage Point Parameter Value Point Parameter P2 XP 2 /c 0.58674 0. high (undesired) axial force.11839 P11 XP 11 /c YP 5 /c 0. 7. At the same time the eﬃciency of the optimized shape is higher than the conventional design for φ < 0. standard Wells turbines have several wellknown disadvantages: low tangential force.0636 YP 10 /c P5 XP 5 /c 0.357793 −2. with an average increase of 0.CHAPTER 7. only symmetric airfoils can be considered. In the present work we have shown the potential of CFDbased optimization to improve the tangential force induced by monoplane and twostage Wells turbines.78848 0. 7. • Monoplane Wells turbine In this case. WELLS TURBINE: CONCURRENT OPTIMIZATION Table 7. a mathematical optimization procedure has been carried out considering simultaneously up to twentytwo free parameters.1617185 3. Due to the importance of the airfoil shape.047259 0. with an average gain of 12% (Fig.53128 0.0746 YP 4 /c 0.70306 0.111736 YP 12 /c P7 XP 7 /c 0.8: Polynomial coeﬃcients of optimal airfoil shape (best ﬁt) for twostage Wells turbine with symmetric airfoils A B D E H K 0. 7.8563828 0.065626 0.23a). The optimization relied on Genetic Algorithms.11015 Wells turbine Value 0. low aerodynamic eﬃciency and limited range of operation.5 Conclusions on Wells turbine As discussed previously. all geometrical conﬁgurations being evaluated in an automatic manner by CFD.5% and 2%. The corresponding gain varies between 0.08169 0. taking into account the inﬂuence of the mutual interaction between the blades.090718 0.0458 YP 9 /c P10 XP 10 /c P4 XP 4 /c 0.588445 0.6% (Fig. Two concurrent objectives (eﬃciency and tangential force coeﬃcient) have been maximized in a concurrent manner.
08 0.2 0.2 0.12 0.16 0. a) tangential force coeﬃcient. • Twostage Wells turbine with nonsymmetric airfoils Then. The corresponding relative increase is shown with blue line.46 0.2 20 0 0. b) eﬃciency.28 Flow coefficient (b) 0.08 0. WELLS TURBINE: CONCURRENT OPTIMIZATION (a) 153 Tangentional force coefficient (CT) 1 Optimum design NACA 0021 Relative increase 100 % 0.28 Flow coefficient Figure 7. At the same time. A relative increase of the tangential force coeﬃcient exceeding 8. 1%). the eﬃciency improves also by at least 0.3%) is obtained for the full operating range.2% and up to 3.44 Efficiency 0. the aerodynamic performance of a modiﬁed twostage Wells turbine consist % Relative increase 0.42 0.CHAPTER 7.12 0.16 0.8% (as a mean. than the standard design relying on NACA 0021.24 0 0.38 0. compared to the conventional twostage Wells turbine relying on the NACA 0021 proﬁle (gray squares).2% (as a mean.6 60 0.24 0.23: Performance of the optimal conﬁguration (red line).4 40 0. 11.4 Optimum design NACA 0021 0.8 80 .
425 0.8 0.25 Optimized Wells turbine 0.45 0. The relative increase of the tangential force coeﬃcient is however 6. Wells turbine (NACA 0021) .75 0.3 0.4 0. Optimized Wells turbine Wells turbine (NACA 0021) 0.65 0.24: Successive optimization steps for the monoplane and twostage Wells turbines. An aggressive mathematical optimization procedure has been carried out considering simultaneously sixtyfour free parameters.65 0. It can be seen that the optimal airfoil shape leads only to a slightly higher eﬃciency compared to the standard airfoil (NACA 2421).525 0.75 0.45 0. Optimized Wells turbine 0. The gain in eﬃciency varies between 0.5 0.7 0.475 0.5 0. Wells turbine (NACA 2421) Peak tangential force coefficient Peak tangential force coefficient Nonsym.3 Optimization direction Optimization direction a) Monoplane Wells turbine (low potential wave) b) Twostage Wells turbine (high potential waves) Figure 7.5 0.5 0. Optimization has been carried out considering twentytwo free parameters.7 0.45 0.14.425 0.2% and 0.6 0.55 0.35 0.475 0. This optimization procedure is again able to ﬁnd a considerably better Sym.4 Optimization direction Optimization direction Nonsym.35 0.6 0. The two stages are mirrored. so that the system globally stays symmetric. Optimized Wells turbine 0.525 Sym.CHAPTER 7.55 0.25 Sym.2% at φ = 0. WELLS TURBINE: CONCURRENT OPTIMIZATION Nonsym. with an average increase of 5. Mutual interaction eﬀects between the blades are taken in account. Wells turbine (NACA 0021) f»0.45 0.3%. ing of nonsymmetric airfoils has been optimized.85 Sym.7%.4 0.5% along the operating range. symmetric airfoils have been considered. This optimization procedure is able to identify a better conﬁguration than the standard design (NACA 2421). Wells turbine (NACA 2421) Nonsym.4 0. Optimized Wells turbine 154 Wells turbine (NACA 0021) Optimized Wells turbine Peak efficiency Peak efficiency 0.85 f»0. twostage Wells turbines consisting of classical. with an average increase of 0. • Twostage Wells turbine with symmetric airfoils Finally.8 0.
At φ = 0. 7. The corresponding gain varies between 0.18. with an average gain of 12%. keeping in mind that the twostage design needs higher potential waves to operate (see section 2. The tangential force coeﬃcient of the twostage design is approximately twice the monoplane one. The relative increase is higher than 10% throughout the useful operating range. and supporting previous ﬁndings from the literature. The twostage Wells turbine consisting of symmetric airfoils is found better than the twostage design consisting of nonsymmetric airfoils in terms of tangential force coeﬃcient with only minor diﬀerence in eﬃciency (Fig. 7. Finally. we can conclude that these optimized Wells turbine conﬁgurations should help indeed to improve wave energy conversion. with an average increase of 0. WELLS TURBINE: CONCURRENT OPTIMIZATION 155 conﬁguration than the standard design (NACA 0021).14. the monoplane design appears clearly superior to the twostage design in terms of eﬃciency.1). .24).CHAPTER 7.6%. As a whole. The absolute gain for CT increases even slightly with the ﬂow coeﬃcient.3. the eﬃciency of the optimized monoplane turbine exceeds η 52%.24). compared to η 45% for optimized twostage design (Fig.5% and 2%. At the same time the eﬃciency of the optimized shape is higher than the conventional design for a ﬂow coeﬃcient φ ≤ 0.
when Pareto fronts are encountered. this gain 156 . Concerning engineering results: • The performance of the original Savonius turbine could be tremendously increased. GA works better for only 10 to 30 parameters and improvements are needed before considering. this is too long for industrial needs. the maximum number that can be reasonably considered with the present algorithm.48 compared to 0. the optimization process can be carried out in a fully automatic manner allowing to explore eﬃciently a variety of conﬁgurations. GAs are ideally suited for parallelization. Though limited. Fortunately. the main ﬁndings are as follows: • An optimization requires typically one month of computing time on a single PC. For concurrent optimization. the tangential force coeﬃcient and eﬃciency of the original Wells turbine could be improved.18 for the original system (Fig. however. • Concurrent optimization involving several objectives is also possible but leads to additional issues. typically by 12% for the tangential force coeﬃcient and 1% for eﬃciency.Chapter 8 Conclusions and Outlook In this thesis. say. Concerning the process. highlighting the need for parallel computing. parallel optimization is a clear must. This is. In particular. Furthermore. and validation steps. For 3D cases. A patent has been submitted for this conﬁguration. While acceptable for research purposes. two important systems allowing energy generation from renewable sources (Savonius turbine: wind. • It is possible to optimize considering a large number of design parameters. 100 variables. with an extraordinary peak Cp of about 0. • In a similar but somewhat less impressive manner. After many new development. 8. it must be kept in mind that only twodimensional geometries have been considered here. it is usually necessary to involve additional criteria in the ﬁnal decision process. check. Wells turbine: sea waves) have been optimized by Genetic Algorithms. even for academic research. evaluating the performance by Computational Fluid Dynamics.1). Up to 64 such free parameters have been employed in this work.
optimal Savonius turbine.3 0. CONCLUSIONS AND OUTLOOK 157 is already highly interesting for practical purposes and discussions with possible industrial partners are running.2 0. • Improvements in the algorithm are needed to consider many parameters in a concurrent optimization. • The performance of GA could be improved by coupling with other alternatives like particle swarm or surface responce techniques.1: Power coeﬃcients of wind rotors of diﬀerent designs including newly developed. .6 Power coefficient 0. 0.1 Suggestions for further research The present study introduced optimized designs with superior performance for Savonius and Wells turbines.593) 0.5 0.4 0. a full experimental veriﬁcation of the ﬁndings is needed: • Wind tunnel or urban wind measurements should be carried out for the optimal Savonius turbine.1 0 Optimal Savonius Threeblade rotor Twoblade rotor Oneblade rotor Darrieus rotor Dutch windmill American wind turbine Savonius rotor 0 2 4 6 8 10 12 14 16 18 Speed ratio Figure 8. • Experimental measurements should be carried out as well for the optimal Wells turbine considering an oscillating ﬂow. 8. Concerning methodology: • Fully parallel optimization involving parallel CFD should be implemented as a future standard. allowing to reduce the numbers of (very costly) evaluations.CHAPTER 8.7 Betz limit (Cp=0. However.
T. Boake. P. L.M. Physique. [6] C. Bchir. Georgia. and H. 586594. Ocean Engineering. [7] P. Ahram. and L. Journal of Turbomachinery. 35(4):821–829. At´lgan. M. International Gas Turbine Institute. Practical Optimization. BritoMelo.N. 326(8):495–500. WuSheng. Benghrib. 127(2). [9] S. G. 34(56):799–805. LLC. Folley.C. Bellucci. Drag and lift coeﬃcients evolution of a Savonius rotor. Experiments in Fluids. Analysis of Wells turbine design parameters by numerical simulation of the OWC performance.J. A 3D unsteady analysis of a Wells turbine in a seawave energy conversion device. D. USA. Benghrib. Torresi. Turbo Expo. Nowak. Thermoacoustic modeling of a gas turbine combustor equipped with acoustic dampers. Gato. Comparison between a UOWC and a conventional OWC. Flohr. 2003. The use of a curtain design to increase the perfory mance level of Savonius wind rotors. E. Boccotti. A.Bibliography [1] B. 6(7):1–187. Springer Science+Business Media. Bhattacharyya and M. Sarmento. 2003. [3] V. Pascazio. 1998. Paschereit. USA. Fortunato. 2005. New York. Ocean Engineering. and C. Astronomie. and A. and B. 2007. 2002. Schuermans. Japan. Wave Energy Conversion.A. Overview and initial operational experience of the LIMPET wave energy plant. 8(12):118–120. M. McCormick. Whittaker. American Society of Mechanical Engineers. 2008. B. 2002. O. Camporeale. Elsevier Ocean Engineering Series. [2] A. Ellen. Atlanta. Algorithms and Engineering Applications. 158 . Proceedings of the International Oﬀshore and Polar Engineering Conference. M. T. 29(1):1463–1477. GT200338715. J. Kitakyushu. D. Altan and M. Chauvin and D. Description of the vortex formation on a Savonius rotor in a water channel. 1989. Comptes Rendus de l’Academie de Sciences Serie IIb: Mecanique. [4] D. [8] A. [10] A. 2007. Renewable Energy. Antoniou and L. B. Chimie. [5] R.
2002. 2008. and T. 2005. Beattie. and W. Whittaker. Gato. [15] R. Performance prediction of contrarotating Wells turbines for wave energy converter design. [16] R. [19] K. C. Raghunathan. [14] R. 24(2):59–71. . R. pages 1–54. Krewer. Stewart. USA. and U. Proceedings of the International Conference on Oﬀshore Mechanics and Arctic Engineering . 1997. Optimization and Computae tional Fluid Dynamics. J.OMAE 1. 147154.BIBLIOGRAPHY 159 [11] REN21 Steering Committee. 2007. T. Journal of Power Sources. (1):5865. Whittaker. Part A: Journal of Power and Energy. Janiga. Van den Braembussche. Stewart. T. Proceedings of the Institution of Mechanical Engineers. [18] R. A. [12] REN21 Steering Committee. Aerodynamic losses incurred by the Wells turbine. Part A. WonHyouk. Rodrigues. Computational analysis of performance and ﬂow investigation on Wells turbine for wave energy conversion. and W. Whittaker. Curran and L. Curran. 124(2):35–52. T. RENE21 Renewable Energy policy Network for the 21st century. Stochastic modelling of OWC wave power plant performance. 2008. D. Copenhagen. J. S. J. Denmark. Raghunathan. Berlin. SpringerVerlag. Applied Ocean Research. Journal of Renewable Energy. M. Whittaker. T. M. T. 1997. 211(2):133–145. A. T. Th´venin and G. 1995. T. Los Angeles. Curran. [22] A. C. Heidelberg. J. 1998. [17] R. California. P. Falca and R. 124(2):35–52. J. RENE21 Renewable Energy policy Network for the 21st century. 180(1):71–83. Performance prediction of contrarotating Wells turbines for wave energy converter design. In D. Curran. Journal of Energy Engineering. Part A: Journal of Power and Energy. The energy conversion performance of several types of Wells turbine designs. Numerical optimization for advanced turbomachinery design. MinJeong. Curran. Dhanasekaran and M. [13] R. S. [20] R. 1998. Journal of Energy Engineering. 211(6):489–505. Proceedings of the International Oﬀshore and Polar Engineering Conference. DaeHo. S. Beattie. C. T. Matching a Wells turbine to the Islay oscillating water column wave power converter. S. Design synthesis of oscillating water column wave energy converters: Performance matching. 30(14):2129–2147. and T. Raghunathan. Renewables global status report 2009 update. Gato. 1996. J. F. C. Curran and L. editors. pages 1–32. O. Nonisothermal dynamic modelling and optimization of a direct methanol fuel cell. [21] T. P. Renewables global status report 2007. Proceedings of the Institution of Mechanical Engineers. Govardhan. L. 2009.
M. R. Wind Engineering and Industrial Aerodynamics. C. (1):380384. Whittaker. JSME International Journal. Transactions of the ASME. B Hen/Transactions of the Japan Society of Mechanical Engineers. F. 1990. H. 1993. Solar Energy Engineering. O. T. [34] L. Design of wells turbine for OWC wave power plant. 1992. Comparison of LIMPET contrarotating Wells turbine with theoretical and model test predictions. The eﬀect of plenum chamber volume and air turbine hysteresis on the optimal performance of oscillating water columns. d. 31(6):425–434. Uemura. Experimental study on the aerodynamic performance of a Savonius rotor. 59(567):3519–3523. 1989. Gato and A. C. [30] N. 33(89):1056–1069. Investigation of ﬂow ﬁeld inside a Savonius rotor by image processing technique with conditional sampling. Gato and A. Part B. [31] N. 1990. Performance of the Wells turbine with a double row of guide vanes. Fujisawa. 2006. Halkidiki. K. Power. Fernando and V. 1996. Environment and Energy seminar series. Proceedings of the International Conference on Oﬀshore Mechanics and Arctic Engineering. Gotoh. C. 2005. K. Proceedings of the First International Oﬀshore and Polar Engineering Conference. On the torque mechanism of Savonius rotors. R. Wind Engineering and Industrial Aerodynamics. Series 2: Fluids Engineering.BIBLIOGRAPHY 160 [23] M. M. T. Falcao. Colorado. Pereira. Y. 6(1):11–34. M. Modi. A. 1994. Ikai. Visualization study of the ﬂow in and around a Savonius rotor. 40(3):277–292. Greece. 33(2):265–271. 2008. D. USA. Taguchi. Nippon Kikai Gakkai Ronbunshu. Fujisawa. M. Fujisawa and F. [27] N. J. A numerical analysis of the unsteady ﬂow past bluﬀ bodies. Falcao. Falcao. Experiments in Fluids. 1991. F. Nagaya. and S. U. [28] N. Gato. F. N. . O. (2):493498. Velocity measurements and numerical calculations of ﬂow ﬁelds in and around Savonius rotors. S. [26] EUJapan Centre for Industrial Cooperation Seminar Report 2008. Curran. Fujisawa and F. C. 12(6):407–412. Gotoh. d. [24] M. Fujisawa. Whittaker. O. [25] M. Satoh. Aerodynamics of the Wells turbine: control by swinging rotorblades. 1992. Combustion. [29] N. International Journal of Mechanical Sciences. [33] L. J. and T. [32] L. Folley and T. Computational Mechanics. Heat Transfer. Pereira. Thermophysical Properties. Ocean Engineering. (4):1–59. 59(1):39–50. and R. The future of energy: Shell energy scenarios to 2050. 116(3):148–152. Climate Change. Folley.
C. Kaneko.scienceonline. . Estanqueiro. M. o Impacts of large amounts of wind power on design and operation of power systems. 724.html. RERIC International Energy Journal. Selim. 32(1):19–24. Setoguchi. International Journal of Heat and Mass Transfer. Germany. Raghunathan. 2001. Series B: Fluids and Thermal Engineering. USA. [40] R. The fundamental characteristics and future of Wells turbine for wave power generator. Strbac. and M. [41] H. S. Comparative study of a threebucket Savonius rotor with a combined threebucket Savoniusthreebladed Darrieus rotor. and D. Orths. Hara. K. results of IEA collaboration. [38] T. Huda. 2006. Inoue. 1992. Inoue. O’Malley. K. 1986. JSME International Journal. Optimization of chemical processes. K. Islam. Hara. B. Janiga. Klein. Sharma. Renewable en [43] M. Islam. Glandt. Gomez. A. R. Renewable Energy. Proceeding of 14th Kassel Symposium on Energy Systems Technology. Wind tunnel tests on a diﬀerent phase threestage Savonius rotor. Li. Hau. P. and T. S¨der. Meibom. J. A. [42] http://www. and F. ergy resources. Hayashi. Heidelberg. Holttinen. Series B: Fluids and Thermal Engineering.uk/energy/renewable energy. 2009. Tande. Biswas. 2006. 1987. K. Setoguchi. O. Edgar. Inoue. E. K. Fundamental. A. Q. Kaneko. Series B: Fluids and Thermal Engineering. and H. Tokyo. [37] E. Renewable Energy. 1989. Y. G. A. E. Hamakawa. 48(1):9–16. H. M. T. A. M. 33(9):1974–1981. Air turbine with selfpitchcontrolled blades for wave power generator (estimation of performance by model testing). and K. Berlin. 48(1):9–16. Hulle. [46] M. G. T. 2008. Technology. and Y. 49:2567–2577. Hayashi. Gupta. Smith. D.co. and Y. Setoguchi. 39:275–280. Y. Lange. Kaneko. SpringerVerlag. Wind tunnel tests on a diﬀerent phase threestage Savonius rotor. Proceedings of the 5th International Symposium on Oﬀshore Mechanics and Arctic Engineering. Li. Performance of an Sshaped Savonius rotor with a deﬂecting plate. M. Kassel.Japan. G. JSME International Journal. J. 604609. D. Journal of Science of Machines. last accessed April 2010. JSME International Journal. T. 2005. 2005. 14(1):25–32. [44] M. [36] R. Raghunathan. Multiobjective shape optie mization of the ﬂow around a heat exchanger using parallel genetic algorithms. Applications and Economics. [45] M. V. and S. T.BIBLIOGRAPHY 161 [35] E. L. Starting and running characteristics of biplane Wells turbine. Th´venin. Baron. New York. Hilbert. McGrawHill. [39] T. and S.
[55] K. 221(5):647–655. [56] K. T. and M. Irabu and J. Colorado. Th´venin. Kitakyushu. Collection of Technical Papers . Raghunathan.BIBLIOGRAPHY 162 [47] M. Japan. M. Hysteretic characteristics of Wells turbine in reciprocating ﬂow. [51] S. [54] M. 1994. Experimental Thermal and Fluid Science. and S. M. A. Q. T. Y. S. [57] T. (1):385392. Technology and Environment Economics.Part B. Inoue. and C. 1(2):122–128. Selfrectifying turbines for wave energy conversion. Kenji. Numerical analysis for unsteady ﬂow characteristics of the Wells turbine. 15(2):125–136. Islam. B. C. Koura. Numerical simulation for Savonius rotors (running performance and ﬂow ﬁelds). International Journal of Oﬀshore and Polar Engineering. Kaneko. and F. I. Setoguchi. [50] K. Kedare. Janiga and D. . Kamoji. 1991. 2007. Experimental investigations on modiﬁed Savonius rotor. A. Nippon Kikai Gakkai Ronbunshu. H. Renewable Energy. Prabhu. 1993. Kaneko. Streicher. Heidelberg. T. 1997. and S. Ishimatsu. Kang.AIAA Applied Aerodynamics Conference. [48] K. Lee. T. 2002. Part B. 694699. 2007. 61(581):12–17. Proceedings of the First International Oﬀshore and Polar Engineering Conference. Kage. Setoguchi. 2007. B Hen/Transactions of the Japan Society of Mechanical Engineers. Kim. and A. Mandal. Park. 2007. Biplane axial turbine for wave power generator. Nippon Kikai Gakkai Ronbunshu. Okubayashi. Islam. Razzaque. N. 60(569):154–160. Inoue. 53(496):3699–3704. Characteristics of wind power on Savonius rotor using a guidebox tunnel. Hamakawa. [53] M. A. Setoguchi. Part B. V. Kaltschmitt. [49] K. 1:768–780. Proceedings of 12th the International Oﬀshore and Polar Engineering Conference. Reducing the CO emissions in a laminar burner e using diﬀerent numerical optimization methods. 32(2):580–586. Wiese. Journal of Power and Energy. Nippon Kikai Gakkai Ronbunshu. K. Berlin. BHen/Transactions of the Japan Society of Mechanical Engineers. and F. K. Aerodynamic characteristics of a stationary Savonius rotor. Ishimatsu. and T. and M. Roy. USA. T. 1991. K. Setoguchi. B Hen/Transactions of the Japan Society of Mechanical Engineers. [52] G. RERIC International Energy Journal. M. Takuma. T. Numerical simulation for Savonius rotors (eﬀects of shed vortices on running performance). Shinohara. SpringerVerlag. 1995. W. Shinohara..
A doublestep Savonius rotor for local production of electricity: A design study. International Journal of Thermal Sciences. T. of Magdeburg. Inoue. Kaneko. Proceedings of the 10th International Oﬀshore and Polar Engineering Conference. Lind. Kim. Kinoue. . [68] J. Part B. [65] J. Proceedings of the 15th International Oﬀshore and Polar Engineering Conference. 2004. 2009. Kim. Kim. Menet. Y. 2005. Kaneko. Setoguchi. [62] Y. K. V. USA. Wind Energy Proceedings of the Euromech Colloquium. Inoue. Seoul. Kinoue. [59] T. T. Kaneko. [66] M. Setoguchi. Kaneko. T. Takao. Numerical analysis of ﬂow characteristics in a Wells turbine for wave power conversion. International Journal of Thermal Sciences. Lee. T. 2004. Kaneko. T. T. [61] Y. Univ. K. and M. Nature Conservation and Nuclear Safety (BMU). Setoguchi. Setoguchi. 2003. Proceedings of the International Oﬀshore and Polar Engineering Conference. Santhakumar. Mechanism of hysteretic characteristics of Wells turbine. B Hen/Transactions of the Japan Society of Mechanical Engineers. 2000. Study of turbine with selfpitchcontrolled blades for wave energy conversion. Mamun. Electricity from renewable energy sources. 2009. [63] H. Experimental Investigation of a modiﬁed Savonius Turbine. Kim. Performance comparison of turbines for wave power conversion. H. Berlin . Inoue. Pernick. K. 427433. Setoguchi. H. Setoguchi. M. R. LSSS01/10. and C.BIBLIOGRAPHY 163 [58] T. Clean energy trends 2009. Kang. 2010. and M. Makower. M. 41(1). [67] M. K. Kinoue. Mark and W. Setoguchi. Hysteretic characteristics of monoplane and biplane Wells turbine for wave power conversion. [64] M. Springer. and Y. T. J. Korea. [60] T. H. K. T. Hysteretic characteristics of Wells turbine for wave power conversion. Hysteretic characteristics of biplane Wells turbine in a deep stall condition. Menet. Mohammad. 69(679):610–616. [69] J. 40(7):681–689. Kim. Kinoue. 122. and C. Renewable Energy. Inoue. (1):523529. M. Aerodynamic behaviour of a new type of slowrunning VAWT. Lee. Clean Edge. Nippon Kikai Gakkai Ronbunshu. Kaneko. Energy Conversion and Management. and S. 29(11):1843–1862. T. 45(910):1617–1629. Seattle. 2001. Studienarbeit Thesis. 2002.(43):235–240. u Federal Ministry for the Environment. Y. 2007. Wilder. Kim. Y.Germany:1–44. and M. Lee. and M. D¨rrschmidt. 2002. Takao. K.
[77] G. [74] M. Th´venin. Renewable Energy. J. GT201023258. International Gas Turbine Institute.H Mohamed. 2003. Wagner. O. Proceedings of the Coastal Engineering Conference. and D Th´venin. USA. Berlin. Mohamed. GT200850815. [76] C. A. G. G. Nayeri. USA. Pechlivanoglou. O. A. Kassel. Turbo Expo. H. [71] J. 2008. Pechlivanoglou. 6371. Raghunathan. American Society of Mechanical Engineers. E.BIBLIOGRAPHY 164 [70] J. and P. C. GT200338393. Cairo. L. Grid operation supported by wind farms: The Danish experience. Florida. [72] M. ICFDP9EG249. . Aerodynamics and Aeroacoustics. Singapore. Proceedings of the Third International Oﬀshore and Polar Engineering Conference. 48th AIAA Aerospace Sciences Meeting. International Gas Turbine Institute. UK. M´nart. Proceeding of 14th Kassel Symposium Energy Systems Technology. Nayeri. B¨che. [75] I. Aerodynamic performance of the Wells turbine used in the sea wave conversion. Vald´s. G. Janiga. Active aerodynamic control of wind turbine blades with high deﬂection ﬂexible ﬂaps. 2010. Germany. N. Paschereit. Paschereit. Powalko. O. Increase in the Savonius rotors eﬃciency via a parametric investigation. Atlanta. Helwan University. Menet and B. and C. (1):502–511. Schuermans. Fuehr. N. Glasgow. [80] S. Prediction of performance of biplane Wells turbine. Ninth International Congress of Fluid Dynamics and Propulsion ASMEICFDP9. 1993. Pap. C. Georgia. G. Eriksen. 1993. and D. Oscillating water column modeling. Germany. International Gas Turbine Institute. Turbo Expo. and C. Orths. A comparative calculation of the wind ture e bines capacities on the basis of the L − σ criterion. [78] G. B. 2001. European Wind Energy Conference EWEA. 2010. [73] M. The eﬀect of distributed roughness on the power performance of wind turbines. Nachida. American Society of Mechanical Engineers. Menet. and B. 2004. Greated. Master’s Thesis. Paschereit. Mohamed. Egypt. Janiga. and D. Morrison and C. 2008. S. Combustion process optimization u using evolutionary algorithm. (6):24812488. Performance optimization of a e modiﬁed Wells turbine using nonsymmetric airfoil blades. (1):167171. Orlando. Egypt. Turbo Expo. B. American Society of Mechanical Engineers. Alexandria. (AIAA 2010644). 2003. 22(4):491– 506. [79] M. 2009. (2):281291.H. Optimal performance of a e Savonius turbine using an obstacle shielding the returning blade.
Raghunathan. Proceedings . . Kaneko. Aerodynamic performance of contrarotating Wells turbine for wave energy conversion. Proceedings of the Institution of Mechanical Engineers. Raghunathan. [86] S. Colorado. Beattie. Performance of the Islay Wells air turbine. T. Tan. and T. [92] S. 1985. 1985. Series B: Fluids and Thermal Engineering. [84] S. K. Tan. 6(6):430–431. 1995. O. Some techniques to improve the operation range of the Wells turbine for wave power generator. C. Aerodynamics of monoplane Wells turbine. P. Raghunathan. Raghunathan. Wells. 1995. Part A: Journal of Power and Energy. Tan. 2006. 7(3):226–230. P. Part A: Journal of Power and Energy. AIAA Journal (Journal of Energy). The eﬀect of inlet conditions on the performance of Wells turbine. Kaneko. Ombaka. T. Proceedings of the ASME Fluids Engineering Conference. J. USA. Valveaided twisted Savonius rotor. 209(1):55–62. 87FE2.IMechE: A. Wind Engineering. Raghunathan. International Journal of Heat and Fluid Flow. C. 1982. Proceedings of the First International Oﬀshore and Polar Engineering Conference. A methodology for Wells turbine design for waves energy conversion. and K. 6:17–22. Proceedings of the Institution of Mechanical Engineers. Raghunathan. Setoguchi. J. 210(6):431–447. Setoguchi. P. 209(A3):221–232. Tan. Tan. AIAA Journal. [91] S.a review. 1996. Raghunathan and W. T. Aerodynamic performance of a Wells air turbine. 1995. Performance of the Wells selfrectifying air turbine. [89] S. Journal of Energy. 1987. P. Wind tunnel tests on airfoil in tandem cascade. Kaneko. 32(1):71–77. [83] S. 1983. (1):370379. Performance of the Wells turbine at starting. R. The eﬀect of blade proﬁle on the performance of Wells turbine. 89(890):369–379. P. [85] S. [94] M. Kaneko. 1989. 1981. and K. Progress in Aerospace Sciences. Raghunathan and C. Curran. Journal of Power and Energy. Whittaker. T. [82] S. Raghunathan. A. Saha. [87] S. 1989. C. and K.BIBLIOGRAPHY 165 [81] S. Setoguchi. [93] S. 1991. Aeronautical Journal. Some techniques to improve the operation range of the Wells turbine for wave power generator. Raghunathan. 19(11):1490–1492. USA. [90] S. T. JSME International Journal. Raghunathan. and O. Setoguchi. 31(4):335–386. Raghunathan and C. 30(3):243–254. J. Raghunathan and C. The Wells air turbine for wave energy conversion. and K. JSME International Journal. 32(1):71–77. Series B: Fluids and Thermal Engineering. Rajkumar and U. [88] S. and N.
K. Setoguchi. Y. 662667. Kaneko. Engineering optimization: methods and applications. T. Raghunathan. Reklaitis. Y. Inoue. 1994. K. Mohammad. Kaneko. and M. 2001. 2005. Kim. [102] T. Rajkumar. Kaneko. Kaneko. 2003. Takao. (1):266271. Kaneko. and K. T. M. K. 13(4):316–320. U. The eﬀect of rotor geometry on the performance of a Wells turbine for wave energy conversion. Takao. and M. 2002. Kaneko. Air turbine with selfpitchcontrolled blades for wave energy conversion. S. Takao. Inc. [99] T. Maity. V. Unsteady ﬂow phenomena of Wells turbine in deep stall condition. Renewable Energy. Air turbine with staggered blades for wave power conversion. . International Energy Journal. Japan. 25(3):137–150. Thakker. Inoue. Rajkumar. J. M. John Wiley & Sons. M. K. Toulon. California. T. [106] T. Proceedings of the International Oﬀshore and Polar Engineering Conference. Ravindran. USA. M. Hysteretic characteristics of Wells turbine for wave power conversion. K.A review. K. New Jersey. and M. H. Santhakumar. T. Renewable Energy. Air turbine with staggered blades for wave power conversion. Takao. Simulation of ﬂow around and behind a Savonius rotor. [103] T. [97] T. Kitakyushu. Kinoue. Setoguchi. [96] A. 2004. 2003. and D. Renewable Energy. Setoguchi. [100] T. 1996. [104] T. M. Inoue. Aerodynamic of monoplane Wells turbine . H. [105] T. Kim. Lee. Kim. H. Setoguchi S.BIBLIOGRAPHY 166 [95] M. Kim. Setoguchi. A modiﬁed Wells turbine for wave energy conversion. Setoguchi. On the performance analysis of Savonius rotor with twisted blades. (1):7074. M. Santhakumar. International Journal of Oﬀshore and Polar Engineering. Kinoue. and M. Inoue. 4(1):68–75. and M. A. Y. International Journal of Ambient Energy. [101] T. Kaneko. 31(11):1776–1788.. International Journal of Oﬀshore and Polar Engineering. Saha. France. M. 28(13):2113–2127. Takao. J. Kim. and K. [98] U. 2006. Los Angeles. Setoguchi. W. W. Y. and G. 6(2):83–90. Kaneko. K. 2003. Eﬀect of guide vane shape on the performance of a wells turbine. USA. T. Proceedings of the 6 th International Oﬀshore and Polar Engineering Conference. Raghunathan and K. Renewable Energy. Setoguchi. 2006. 2004. Takao. Takao. Saha and M. M. 28(1):79–91. S. Setoguchi. 1:1–15. T. Ragsdell. K. Kim. Lee. and S. Proceedings of the International Oﬀshore and Polar Engineering Conference. H.
M. Inoue. Brest. Takao. Proceedings of the International Oﬀshore and Polar Engineering Conference. Setoguchi. Thakker. [118] M. K. Mohammad. Eﬀect of rotor geometry on the performance of Wells turbine. Takao. A. Suzuki. Arakawa. 47(1516):2382–2396. Takao. (4):351357. Kinoue. SpringerVerlag. France. Guide vanes eﬀect of Wells turbine for wave power generator. Deepa. Inoue. Current status of self rectifying air turbines for wave energy conversion. M. Setoguchi and M. the eﬀect of solidity. K. Journal of Thermal Science. 1984. Design method of guide vane for Wells turbine. 1999. Orths. IEEE Transactions on Energy Conversion. 8(2):155–160. (1):8388. [108] T. Hofmann. K. Introduction to Genetic Algorithms. 1998. Comparative study of performances of turbines for wave power conversion. 2008. M. Setoguchi. Exo perience from wind integration in some high penetration areas. W. Takao. Berlin. [110] T. K. M. Shabbir. L. H. 1995. [112] T. M. Yang. [109] T. [114] S. N. Seattle. and A. K. S¨der. and M. and selfstarting). [113] T. Kaneko. W. 2000. [117] M. G. C. Hawaii. 27(231):1925–1931. and M. H. A. 2003. 2006. Wan. 2006. (1):162168. Setoguchi. Proceedings of the International Oﬀshore and Polar Engineering Conference. Tuohy. Kaneko. and A. Sivanandam and S. 1998. A new k. 8(2):155–160. USA. International Journal of Oﬀshore and Polar Engineering. Takao. Energy Conversion and Management. International Journal of Oﬀshore and Polar Engineering. and J.eddy viscosity model for high Reynolds number turbulent ﬂows. Inoue. Bulletin of the JSME. 2007. (1):374381. Takao. Holttinen. Eﬀect of guide vanes on the performance of a Wells turbine for wave energy conversion.BIBLIOGRAPHY 167 [107] T. Shih. Suzuki. [116] M. Eﬀect of guide vanes on the performance of a Wells turbine for wave energy conversion. Fundamental studies on Wells turbine for wave power generator (1st report. and T. Eﬀect of guide vanes on the performance of a Wells turbine for wave energy conversion. Liou. Hawaii. [115] L. Kaneko. 1997. Proceedings of the International Oﬀshore and Polar Engineering Conference. M. Tagori. Setoguchi. Proceedings of the Ninth International Oﬀshore and Polar Engineering Conference. Z. Itakura. Suzuki and C. Computers and Fluids. Zhu. Inoue. 24(3):227–238. Kaneko. [111] T. 22(2):4–12. Setoguchi. and M. E. N. Kaneko. USA. Arakawa. Honolulu. 15(2):126–131. and M. . Honolulu. Heidelberg. Y. K.
and M. International Journal of Sustainable Energy. Y. 25(1):53–61. 2000. 2001. [126] M. Takao. Kikuyama. Nippon Kikai Gakkai Ronbunshu. Kaneko. Kinoue. USA. Part B. Bajeet. 15(4):319–323. T. [128] S. Takao. Y. Inoue. Inoue. and M. K. 2004. Wells turbine with end plates for wave energy conversion. [125] M. Kaneko. P. T. A. [129] A. . Takao. Abdulhadi. and T. Setoguchi. 63:577– 582. Kinoue. Kaneko. and K. and M. Part B. CA9: Analysis of a stallresistant aerofoil in view of wave power conversion. Nippon Kikai Gakkai Ronbunshu. Eﬀect of blade proﬁle on the performance of largescale Wells turbine. 2006. California. San Francisco. Experimental studies on a Savonius rotor with casing. Frawley. [127] M. Journal of Thermal Science. Ocean Engineering. Thakker.BIBLIOGRAPHY 168 [119] M. Setoguchi. Tan. Air turbine with selfpitch controlled blades for wave energy conversion. K. Takao. Takao. Toulon. T. [124] M. K. Y. Takao. (4):381386. [121] M. Y. R. 1997. T. T. and T. Takao. Shimizu. Kaneko. (1):614619. Setoguchi. Kaneko. T. Abdulhadi. 2006. Thakker. France. and S. T. A. (1):272276. and K. Kaneko. B Hen/Transactions of the Japan Society of Mechanical Engineers. Eﬀect of guide vanes on the performance of a Wells turbine. and S. Proceedings of the International Oﬀshore and Polar Engineering Conference. 1997. 72(10):2381–2385. Norway. 63(611):2356–2363. Kaneko. B Hen/Transactions of the Japan Society of Mechanical Engineers. S. 7(4):308–312. Setoguchi. Seattle. Proceedings of the International Oﬀshore and Polar Engineering Conference. [120] M. Stavanger. and E. Part B. R. Proceedings of the International Oﬀshore and Polar Engineering Conference. K. Nagata. Nagata. 34(1112):1790–1795. Setoguchi. Takao. Takao. Setoguchi. Nihon Kikai Gakkai Ronbunshu. Setoguchi. [122] M. T. The performance of Wells turbine with 3D guide vanes. Kinoue. 1997. Eﬀect of blade proﬁle on the performance of largescale Wells turbine. 2006. Improvement of Wells turbine performance by means of end plate. (1):480484. International Journal of Oﬀshore and Polar Engineering. Inoue. B Hen/Transactions of the Japan Society of Mechanical Engineers. Kinoue. Kim. 2007. Improvement of Wells turbine performance by means of end plate. Thakker. [123] M. USA. Y. Proceedings of the International Oﬀshore and Polar Engineering Conference. 2006. K. and K. H. Eﬀect of end plates on the performance of a Wells turbine for wave energy conversion. Setoguchi. Setoguchi.
W. Retrieved 20081013. M. 1994. [142] J. Numerical analysis of Wells turbine performance using a 3D NavierStokes explicit solver. [135] M. Spon. (1):7076. Gato. [131] A. S. Selfstarting process simulation of Wells turbine in regular and irregular oscillating air ﬂows. Frawley. 2001. Heidelberg. 9(3):233–239. e SpringerVerlag. Renewable Energy Resources. . 1999. and E. Eﬀect of rotor blade shape on the performance of the Wells turbine. Hawaii. [141] Z. McIlwaine. waveenergy. [133] T Thorpe. Weir. 1999. Maeda. Thakker. [139] H. 1991. UK. 1991. 3(2). Univ. Shoreline wave power experience with the Islay prototype. Journal of Hydrodynamics. 2008. Webster and L. USA. 1987. Modelling. Optimization and Computational Fluid Dynamics. 2009. Yu. 1997. Berlin. ELBS/E.N. Bajeet. Janiga. A. Performance simulation on CA9 Wells turbine rotor using a 3D NavierStokes explicit solver. (1):169173. 1990. Proceedings of the International Oﬀshore and Polar Engineering Conference. Thakker. Twidell and A. Webster. (1):393397. USA. [136] M. Bajeet. Proceedings of the fourth European Conference on Turbomachinery. M. Optimal matching design method of the Wells turbine with the air chamber in an oscillating air ﬂow. Brest. 4(4):469–476. S. of Magdeburg. Variable geometry Savonius windmill: modeling and wind tunnel testing. O’Dowd. Numerical investigation of a rotating system using OpenFOAM.BIBLIOGRAPHY 169 [130] A. LSSM02/09. [134] J. and M. White. P. Webster and L. 56(1):57–64. China Ocean Engineering. Sabzevari. Stavanger. [140] Z. J. & F.net. simulation and control. Norway. M. C. Zarinchang and A. Thakker. Master’s Thesis. P. Yu. [138] T. [137] P. Utilization of CFD to study an oscillating water column. C. Frawley. Proceedings of the International Oﬀshore and Polar Engineering Conference. The eﬀect of rotor blade sweep on the performance of the Wells turbine. (1):604608. Proceedings of the International Oﬀshore and Polar Engineering Conference. Th´venin and G. France. London. Colorado. [132] D. B. D. T. Whittaker and S. International Journal of Oﬀshore and Polar Engineering. Gato. Honolulu. performance and costs. and E. Proceedings of the First International Oﬀshore and Polar Engineering Conference. An overview of wave energy technologies: status. Yu and H. 2001.
2005 Aug.2007 Aug. 8.BIBLIOGRAPHY ********* Curriculum Vitae ********* Name: Date of Birth: Nationality: Status: Address: Email: Mohamed Hassan Ahmed Mohamed Juli 15. D39114 Magdeburg. 4. Military service.ovgu. Egypt.1992 1992 . Postgraduate courses for Ph. B.Sc. Mohamed . Faculty of Engineering Mattria Helwan University. Chair of Fluid Dynamics and Technical Flows. in Mechanical Power Engineering. OttovonGuericke University Magdeburg.Sc. Egypt. Egypt.1999 1999 . Assistant lecturer in Dept. of Mechanical Power Engineering. Faculty of EngineeringMattria. Cairo. Work Experience: 1999 . in Mechanical Power Engineering. Germany. Germany mohamed. Helwan University. Lehrstuhl Str¨mungsmechanik & Str¨mungstechnik o o (ISUT/LSS). Cairo. Postgraduate courses for M. degree at Institut f¨r Str¨mungstechnik u o & Thermodynamik. Cairo. Helwan University. 1974. Egypt Egyptian Married. of Mechanical Power Engineering. Cairo. Germany. Egypt.till Now Secondary School Certiﬁcate.Sc. 2007 . Helwan University.2001 2001 . Egypt. Helwan University.2003 2003 . Faculty of EngineeringMattria.till Now Demonstrator in Dept. Works towards Ph. H. Cairo. Faculty of EngineeringMattria.mohamed@st. PhD student and Assistant lecturer in Turbomachines.D. 2007 . October 2010 M. Cairo.D.de 170 Education: 1989 . Faculty of EngineeringMattria. Cairo. Egypt. two Children Coswiger Str. Helwan University. OttovonGuerickeUniversity Magdeburg. M. in Cairo. Magdeburg.2003 2003 . Egypt.1997 1998 .de moh75202@yahoo.
.
. Optimal Blade e Shape of a Modiﬁed Savonius Turbine Using an Obstacle Shielding the Returning Blade.: Optimal shape of e a modiﬁed Wells turbine considering mutual interaction between the blades.. 2011. D. Mohamed. Janiga.. Pap. In: ASME Turbo Expo Conference. J. Optimization of e Savonius turbines using an obstacle shielding the returning blade. Ed. ISBN 9789634209874. 2009. M. ISBN 9789636619244. and Th´venin. P. G. P. Janiga. . Miskolc. M. Janiga. and Th´venin. Glasgow.). G.. M. E. E. Mohamed. M. M. Pap.. Mohamed. 3. In: microCAD10 International Scientiﬁc Conference. Pap. EG3026. Energy. Hungary..H. Aswan. 9. Egypt. Scotland. 2011.: Optimal perfore mance of a modiﬁed threeblade Savonius turbine using frontal guiding plates..H. Budapest. Miskolc. Pap.BIBLIOGRAPHY 172 ——————————————————– Related Publications The presented thesis is partly based on the following publication in international reviewed journals and conferences: 1. 52(1):236242. G. 35(11):26182626.). Mohamed. Janiga. Janiga. and Th´venin.H. Pap. Mohamed. (Bikfalvi. E. and Th´venin. In: 10th Int.: Optimal perfore mance of a Savonius turbine using guiding plates. 36(1):438446. Congress of Fluid Dynamics ASMEICFD10. Mohamed. 4. and Th´venin. D.. In: proc. G. (Vad. Ed. 2010.H. D. Hungary. Hungary. D. 6.H.). Mohamed. In: First International Conference of Energy Engineering ICEE1. 2010.. Janiga. G. D. E. Ed. 2010. Janiga. E. GT201022538. and Th´venin. and Th´venin. 99104. G. In: Conference on Modelling Fluid Flow (the 14th International Conference on Fluid Flow Technologies). and Th´venin.. 2010.H.: Airfoil shape optimization e of a twostage Wells turbine. Cairo. 5156.: Performance optimization of a Savoe nius Turbine Considering Diﬀerent Shapes for Frontal Guiding Plates.. Energy Conversion and Management.. 5..H. and Th´venin. Egypt.. 2008.. Janiga. G. Mohamed. 7..H.. D. ISBN 9789636619107. 871878.: Airfoil shape optimization e of a twostage Wells turbine.. Mohamed. Pap. (Bikfalvi. 2010... 8. D. Renewable Energy. MultiObjective Ope timization of the Airfoil Shape of Wells Turbine used for Wave Energy Conversion. MagdeburgMiskolc Interuniversity Cooperation (50 Year Cooperation). M. 2. D. M.. M. E. D.H. M. G.
Mohamed. Mohamed. and Th´venin.BIBLIOGRAPHY 173 10. 11. 24812488. Janiga. D. Egypt. GT200850815. Congress of Fluid Dynamics and Propulsion ASMEICFDP9. Alexandria. .: Performance optimization e of a modiﬁed Wells turbine using nonsymmetric airfoil blades.: Optimal perfore mance of a Savonius turbine using an obstacle shielding the returning blade. Pap.H. In: 9th Int.H. Germany. 249/1249/9. E. 2008. Berlin. M.. 2008. Janiga. D. In: ASME Turbo Expo Conference. G.. G.. and Th´venin. M.
This action might not be possible to undo. Are you sure you want to continue?