Are you sure?
This action might not be possible to undo. Are you sure you want to continue?
DYNAMIC INFLUENCES OF WIND
POWER ON THE POWER SYSTEM
PhD t hes i s
Sec t i on of El ec t r i c Power Engi neer i ng
Ørsted•DTU
Mar ch 2003
ISBN : 8791184169
DYNAMIC INFLUENCES OF WIND
POWER ON THE POWER SYSTEM
By
Pedro Rosas
Thesis submitted to Ørsted Institute, Section of Electric Power Engineering
Technical University of Denmark
In partial fulfilment of the requirements for the degree of Doctor of Philosophy
Technical Report RISØ R1408
ISBN: 8791184169
Ørsted Institute, Risø National Laboratory & Brazilian Wind Energy Centre
Denmark, March 2003
iii
iv
DYNAMIC INFLUENCES OF WIND POWER ON THE POWER SYSTEM
Pedro André Carvalho Rosas
Risø National Laboratory, Wind Energy Department
&
Ørsted Institute – Section of Electric Power Engineering,
Technical University of Denmark
Abstract
The thesis first presents the basics influences of wind power on the power system
stability and quality by pointing out the main power quality issues of wind power in a
smallscale case and following, the expected largescale problems are introduced. Secondly,
a dynamic wind turbine model that supports power quality assessment of wind turbines is
presented. Thirdly, an aggregate wind farm model that support power quality and stability
analysis from large wind farms is presented. The aggregate wind farm model includes the
smoothing of the relative power fluctuation from a wind farm compared to a single wind
turbine. Finally, applications of the aggregate wind farm model to the power systems are
presented. The power quality and stability characteristics influenced by largescale wind
power are illustrated with three cases.
In this thesis, special emphasis has been given to appropriate models to represent the
wind acting on wind farms. The wind speed model to a single wind turbine includes
turbulence and tower shadow effects from the wind and the rotational sampling turbulence
due to the rotation of the blades. In a park scale, the wind speed model to the wind farm
includes the spatial coherence between different wind turbines. Here the wind speed model
is applied to a constant rotational speed wind turbine/farm, but the model is suitable to
variable speed wind turbine/farm as well.
The cases presented here illustrate the influences of the wind power on the power
system quality and stability. The flicker and frequency deviations are the main power
quality parameters presented. The power system stability concentrates on the voltage
stability and on the power system oscillations.
From the cases studied, voltage and the frequency variations were smaller than
expected from the largescale wind power integration due to the low spatial correlation of
the wind speed. The voltage quality analysed in a Brazilian power system and in the Nordel
power system from connecting large amount of wind power showed very small voltage
variations. The frequency variations analysed from the Nordel showed also small variations
in the frequency but it also showed that the wind turbines excites the power system in the
electromechanical modes.
Concerning the stability analysis, the study cases showed that largescale wind power
modifies the voltage stability of the power system and can cause power oscillations. It is
showed here that the reactive power from the wind farms is the key factor on the voltage
stability problem. During continuous operation, the distributed wind power variations did
not give any problems to the power system stability concerning the power oscillations.
v
DYNAMIC INFLUENCES OF WIND POWER ON THE POWER SYSTEM
Pedro André Carvalho Rosas
Risø National Laboratory, Wind Energy Department
&
Ørsted Institute – Section of Electric Power Engineering,
Technical University of Denmark
Resumé
Først i afhandlingen introduceres de vigtigste elkvalitetsproblemstillinger,
elsystemets stabilitet og elkvalitet, i relation til vindkraft, når vindmøller tilsluttes i mindre
og stor skala. Efterfølgende præsenteres en dynamisk vindmøllemodel, der er specielt egnet
til analyse af elkvalitet fra en vindmølle. Herefter introduceres en aggregeret
vindmølleparkmodel til brug ved analyse af elkvalitet og netstabilitet fra store
vindmølleparker. Den aggregerede vindmølleparkmodel indeholder den udglatning af de
relative effekfluktuationer fra en vindmøllepark sammenlignet med en enkelt vindmølle.
Endelig præsenteres nogle anvendelser af den aggregerede vindmølleparkmodel i
forskellige cases.
I afhandlingen er der lagt specielt vægt på at opstille relevante og anvendelige
modeller af vinden i en vindmøllepark. Modellen for vindhastigheden for enkelt vindmølle
inkluderer turbulens og tårnskygge herunder især roterende sampling af turbulensen som
fremkommer pga. vingernes rotation. I parkskala modellen er der taget hensyn til den
rummelige koherens mellem de forskellige vindmøller. Vindhastighedsmodellen er i
afhandlingen anvendt på vindmøller med konstant omløbstal, men den kan også
umiddelbart anvendes på møller med variabelt omløbstal.
De cases, der præsenteres, illustrerer indflydelsen af vindmølleparker på elsystemets
stabilitet og på elkvaliteten. Flicker og frekvensafvigelser er de vigtigste
elkvalitetsparameter, der benyttes til vurderingen af indflydelsen. Stabiliteten af elsystemet
vurderes vha. spændingsstabilitet og effekt og frekvenssvingninger.
Resultaterne fra de forskellige cases viser at spændings og frekvensvariationerne er
mindre end man kunne forvente på grund af den lille rumlige korrelation af
vindhastighederne ved de forskellige vindmøller. Undersøgelsen af spændingskvaliteten i
den brasilianske case og og i Nordelcasen viser at der kan forventes små
spændingsvariationer selv når der tilsluttes meget vindkraft. Undersøgelsen af
frekvensvariationerne i Nordelcasen viser også kun små variationer i frekvensen, men den
viser også at der kan være tilfælde, hvor vindmøllerne anslår elektromekaniske
egensvingninger i elsystemet.
Med hensyn til stabilitets analyserne viser de forskellige cases at storskala
integration af vindkraft kan ændre grænsen for spændingstabilitet og kan forårsage
effektfluktuationer. Det fremgår endvidere af analyserne at den reaktive effekt der
forbruges i parkerne spiller en nøglerolle med hensyn til spændingstabiliteten.
Effektfluktuationerne under normal drift af vindmølleparkerne i Nordelsystemet i de
undersøgte tilfælde viser at vindfluktuationerne ikke giver anledning til noget problem med
stabiliteten af elsystemet.
vi
Acknowledgment
This work has been carried out at the Risø National Laboratory and Technical
University of Denmark and supported financially by the CAPES (Coordenação de
Aperfeiçoamento de Pessoal de Nível Superior), Brazil, through a doctoral scholarship.
This work has been also economically supported by Risø National Laboratory, which I am
thankful for. The work is a fruit of an international cooperation between the Brazilian Wind
Energy Centre, Risø National Laboratory and Technical University of Denmark.
First, I would like to tank my wife Alexsandra Rosas for the long support in my
dreams and for understanding and handling long lone periods while I was working to finish
this phase of our life.
I would like also to tank my friends and supervisors Poul Sørensen and Henrik
Bindner, Risø, and Dr. Arne Hejde, Technical University of Denmark, for the supervision,
guidance, helpful discussions and valuable time. I would like also to express my gratitude
to Prof. Jan RønneHansen (In memoriam), Technical University of Denmark, who strongly
supported and helped me and to Prof. Everaldo Feitosa, Brazilian Wind Energy Centre,
who also strongly supported and initiated the project.
I am also grateful to EFISintef where I did my external research in special to John
Olav Tande, Kjetil Uhlen and Trond Toftevaag for the great opportunity, technical
discussions and support in Trondheim, Norway.
In addition, I would like to thank the staff and management at the Wind Energy and
Atmospheric Department from Risø National Laboratory and at the Section of Electric
Power Engineering, Ørsted Institute at DTU for the general assistance in different ways.
Last, but not least, I would like to thank all my friends, family, in special to my
mother (Elaine Carvalho), my brother (Gustavo Rosas) and relatives in special to Ms. Maria
de Lourdes, Mr. Claudino Araújo and Alexandre Rosas. You have contributed to this thesis
more than you can imagine...
30 March 2003
Pedro Rosas,
vii
8/152
Table of Contents
Abstract ____________________________________________________________ v
Resumé____________________________________________________________ vi
Acknowledgment ____________________________________________________vii
1 Introduction ____________________________________________________17
1.1 Motivation _________________________________________________17
1.2 Literature Review ___________________________________________18
1.3 Wind Power Basics __________________________________________20
1.4 Thesis Outline ______________________________________________21
2 Wind Power Integration ___________________________________________23
2.1 SmallScale Integration of Wind Power _________________________23
2.1.1 Steady – state operation____________________________________24
2.1.2 Dynamic Operation _______________________________________26
2.2 LargeScale Integration of Wind Energy ________________________29
2.2.1 Voltage Stability Problem__________________________________31
2.2.1.1 Analysis of Voltage Stability _____________________________33
2.2.2 Frequency Control Problem_________________________________34
2.2.2.1 Analysis of Power System Oscillations _____________________35
2.3 Remarks on Wind Energy Integration __________________________36
3 Wind Speed Model _______________________________________________39
3.1 Model Description ___________________________________________41
3.1.1 Equivalent Wind Speed Model ______________________________43
3.1.2 Description of the Equivalent Wind Model_____________________44
3.1.3 Deterministic Part of the Wind ______________________________47
3.1.3.1 Tower Shadow ________________________________________47
3.1.3.2 Wind Shear ___________________________________________50
3.1.4 Implementation of the Deterministic Component ________________53
3.1.5 Stochastic Part – Turbulence ________________________________53
3.1.5.1 Power Spectral Density of Turbulence ______________________54
3.1.5.2 Coherence of the Wind __________________________________55
3.1.5.3 Power Spectral Density of a Rotating Blade – Rotational Turbulence
_____________________________________________________55
3.1.6 Implementation of the Stochastic Component___________________56
3.2 Validation of the Equivalent Wind Speed Model __________________58
3.3 Equivalent Wind Speed Model Remarks ________________________61
4 Wind Turbine Model______________________________________________63
9/152
4.1 Simulation Tool _____________________________________________ 63
4.2 Wind Turbine Model ________________________________________ 64
4.2.1 Aeroelastic Components ___________________________________ 65
4.2.1.1 Aerodynamic Rotor ____________________________________ 65
4.2.1.2 Drive Train ___________________________________________ 67
4.2.2 Electrical Components ____________________________________ 70
4.2.2.1 Electrical generator_____________________________________ 70
4.3 Verification of the Complete Wind Turbine Model________________ 71
4.4 Dynamic Wind Turbine Model Remarks ________________________ 76
5 Aggregate Wind Farm Model ______________________________________ 77
5.1 Aggregate Wind Speed Model _________________________________ 77
5.2 Coherence of Turbulence in a Park Scale________________________ 78
5.3 Aggregate Turbulence _______________________________________ 79
5.4 Simulation of Wind Speeds ___________________________________ 80
5.5 Aggregate Wind Turbine Machine _____________________________ 84
5.6 Results and Discussions ______________________________________ 84
5.6.1 Case Description _________________________________________ 84
5.6.2 Results_________________________________________________ 86
5.6.2.1 Wind Speed Simulator __________________________________ 86
5.6.2.2 Wind Farm Power Production ____________________________ 87
5.6.2.3 Extension to Large Wind Farms and Different Random Seeds ___ 94
5.7 Aggregate Wind Farm Model Remarks _________________________ 97
6 Large Scale Integration – Case Analysis _____________________________ 99
6.1 Case Study 1: Voltage Stability in a Modern Power System ________ 99
6.1.1 Power System Characteristics _______________________________ 99
6.1.2 Wind Power Representation _______________________________ 102
6.1.3 Wind Power Impacts on Voltage Stability ____________________ 102
6.2 Case Study 2: Voltage Stability and Quality in a Brazilian Power
System __________________________________________________________ 106
6.2.1 Power System Characteristics ______________________________ 106
6.2.2 Wind Power Representation _______________________________ 108
6.2.3 Wind Power Impacts on the Voltage Stability _________________ 109
6.2.4 Wind Power Impact on the Voltage Quality___________________ 111
6.2.4.1 Light Load Condition __________________________________ 113
6.2.4.2 Heavy Load Condition _________________________________ 115
6.3 Case Study 3: Power System Interactions – NORDEL ____________ 116
6.3.1 Power System Description ________________________________ 117
6.3.1.1 Reduced Nordel Model_________________________________ 118
6.3.1.2 Nordel Characteristics _________________________________ 120
6.3.2 Wind Power Projects and Representation_____________________ 123
6.3.2.1 Aggregate Wind Farm Model ____________________________ 123
10/152
6.3.3 Wind Power Impacts on the Power System Voltage and Frequency
Regulation ______________________________________________________126
6.3.3.1 Frequency controllers __________________________________126
6.3.3.2 Voltage Quality_______________________________________131
6.4 Case Analysis Remarks ______________________________________133
6.4.1 Remarks on Case 1 ______________________________________134
6.4.2 Remarks on Case 2 ______________________________________134
6.4.3 Remarks on Case 3 ______________________________________135
6.4.3.1 Frequency control _____________________________________135
6.4.3.2 Voltage controllers ____________________________________136
7 Conclusions____________________________________________________137
8 Reference List __________________________________________________139
9 Annexes _______________________________________________________147
9.1 Electrical Components Model in SIMPOW _____________________147
9.1.1 Electrical generator ______________________________________148
9.1.2 Stepup Transformer _____________________________________149
9.1.3 Reactive Power Compensation. _____________________________150
9.1.4 Lines and Cables ________________________________________150
9.1.5 Slack bus ______________________________________________150
11/152
List of Figures
Figure 1.1 Basic components of a wind turbine unity. .........................................................20
Figure 1.2 Power smoothing effect from wind farms. ..........................................................21
Figure 2.1 Basic components of a wind farm. ......................................................................24
Figure 2.2 Single line equivalent for a wind turbine connection. .........................................25
Figure 2.3 Voltage fluctuations corresponding to flicker emission unity [34]. ....................27
Figure 2.4 Measured power spectra of the electrical power from a 225kW pitch regulated
wind turbine. .................................................................................................................27
Figure 2.5 Basic Power System Structure.............................................................................29
Figure 2.6 Single line equivalent of a Power System. ..........................................................30
Figure 2.7 Simplified transmission line equivalent diagram. ...............................................31
Figure 2.8 Power transfer to a node as function of the voltage (“nose curve”). ..................33
Figure 3.1 Illustration of the wind on the rotor area of a wind turbine [43]. ........................39
Figure 3.2 Power produced by a 500kW stall regulated wind turbine..................................40
Figure 3.3 General overview of wind turbine models...........................................................41
Figure 3.4 Wind speed measured on a section of a rotating blade [45]. ...............................41
Figure 3.5 PSD of measured electrical power output of a 500kW stall regulated wind
turbine. ..........................................................................................................................42
Figure 3.6 Reference axis used in the wind turbine. .............................................................43
Figure 3.7 Equivalent Wind Speed Model principle.............................................................44
Figure 3.8 The tower shadow effects on the horizontal wind (top view). ............................48
Figure 3.9 Wind speed field interference by the tower shadow............................................49
Figure 3.10 Normalised torque influenced by the tower shadow. ........................................49
Figure 3.11 Wind shear for different sites (ground is reference for height). ........................51
Figure 3.12 Reference axis and angles used in the wind turbine. .........................................52
Figure 3.13. Normalised torque influenced by wind shear (site with a medium z
0
). ............52
Figure 3.14 Implementation of the deterministic model in Simulink/MATLAB. ................53
Figure 3.15. Kaimal PSD of Turbulence...............................................................................54
Figure 3.16 Normalised admittance function to 0p...............................................................57
Figure 3.17 Normalised admittance function to 3p...............................................................57
Figure 3.18 Implementation of the stochastic model in Simulink/MATLAB. .....................58
12/152
Figure 3.19 Normalised simulated deterministic wind component compared to a DBP2
model. ........................................................................................................................... 59
Figure 3.20 PSD comparisons of the stochastic model. ....................................................... 60
Figure 3.21 Measured and the simulated equivalent wind speeds on rotating blade section.
...................................................................................................................................... 61
Figure 4.1 Interaction between each components of a wind turbine unity........................... 64
Figure 4.2 Power coefficients to compute the dynamic power coefficient. ......................... 66
Figure 4.3 Example of drivetrain components. ................................................................... 67
Figure 4.4 Dynamic representation of the drive train model................................................ 68
Figure 4.5 Active power as function of speed and voltage terminals for an asynchronous
generator. ...................................................................................................................... 70
Figure 4.6 Reactive power as a function of the speed and voltage for an asynchronous
generator (1p.u. = rated reactive power at 1pu volts). .................................................. 71
Figure 4.7 C
p
(λ) static characteristic of the wind turbine Nortank 500kW.......................... 72
Figure 4.8 Measured wind speed. ......................................................................................... 73
Figure 4.9 Time series of simulated and measured power to the Nortank 500kW. ............. 73
Figure 4.10 Verification of the standard deviation of the dynamic wind turbine model. .... 74
Figure 4.11. Verification of the flicker P
st
to different frequencies. .................................... 75
Figure 5.1 Spatial disposition of the two wind turbines (α
xy
= 90° means lateral disposition).
...................................................................................................................................... 78
Figure 5.2 Coherence factor for different distances between two points. ............................ 79
Figure 5.3 Structure of the AWFWS generator. ................................................................... 81
Figure 5.4 Distribution of random constants. ....................................................................... 84
Figure 5.5. Static power curve of the wind turbine. ............................................................. 85
Figure 5.6 Wind farm layout. ............................................................................................... 86
Figure 5.7 Comparisons of the wind speed simulator. ......................................................... 87
Figure 5.8 Power characteristics evolution with the wind speed (10% turbulence intensity).
...................................................................................................................................... 88
Figure 5.9 Power characteristics evolution with the wind speed (20% turbulence intensity).
...................................................................................................................................... 89
Figure 5.10. Non linear effects on the power variations. ..................................................... 89
Figure 5.11 Power simulated 13 m/s and 10% turbulence intensity. ................................... 90
Figure 5.12 Power spectral comparisons at 13 m/s and turbulence intensity 10%. ............. 90
Figure 5.13 Power simulated at 16 m/s and turbulence intensity 20%................................. 91
Figure 5.14 Power spectral comparisons at 16 m/s and turbulence intensity 20%. ............. 92
Figure 5.15 Flicker coefficients comparisons (computed according to [34])....................... 93
13/152
Figure 5.16 Evolution of flicker coefficients at 13m/s 20% turbulence intensity. ...............93
Figure 5.17 Evolution of flicker coefficients at 16m/s 10% turbulence intensity. ...............94
Figure 5.18 Influences of different sizes of aggregate wind farms to the power
characteristics at 13m/s, 20 % turbulence intensity ......................................................95
Figure 5.19 Influences of different sizes of aggregate wind farms to power characteristics at
16m/s, 20 % turbulence intensity..................................................................................96
Figure 6.1 Case Studies.........................................................................................................99
Figure 6.2 Diagram of the Power System used in analysis (loads in MW and MVAr)......100
Figure 6.3 Loadability curve to bus 3 without wind turbines. ............................................100
Figure 6.4 Loadability curve to bus 3 without wind turbines. ............................................102
Figure 6.5 Maximum wind power integration concerning voltage stability. ......................103
Figure 6.6 Maximum wind power integration concerning voltage stability (load factor
unity). ..........................................................................................................................104
Figure 6.7. Loadability curve to bus 3 with wind turbines using power electronics. .........105
Figure 6.8 Evolution of the power production from the wind turbines (with electronic
power converter). ........................................................................................................106
Figure 6.9 Brazilian interconnected system power system [72] .........................................107
Figure 6.10 Brazilian network studied [72]. .......................................................................108
Figure 6.11 Layout of a single wind farm applied to the Brazilian power system studied.109
Figure 6.12 Network topology of Brazilian power system studied. ...................................110
Figure 6.13 Maximum wind power to MOSSORO bus (light load condition)...................111
Figure 6.14 Wind power influences on the voltage to different wind speeds. ....................112
Figure 6.15 General wind power influences on MOSSORO bus. ......................................113
Figure 6.16. Voltage at MOSSORO and power flux from the wind farm in light load
condition (mean wind speed 10m/s). ..........................................................................114
Figure 6.17 Statistics wind speed, active power, reactive power and voltage variations at
MOSSORO (light load)...............................................................................................114
Figure 6.18 Voltage at MOSSORO and power flux from the wind farm in heavy load
condition (mean wind speed 10m/s). ..........................................................................115
Figure 6.19 Statistics wind speed, active power, reactive power and voltage variations on
MOSSORO (Heavy load). ..........................................................................................116
Figure 6.20 High voltage Nordic power network [75]........................................................117
Figure 6.21 Reduced model to the Nordic power system [38]. ..........................................119
Figure 6.22 Relevant eigenvalues of the reduced model to the Nordel. .............................121
Figure 6.23 Modal analysis of the eigenvalue 0.34962 + 0.55164 Hz –Nordic Power
System. ........................................................................................................................122
14/153
Figure 6.24 Aggregate wind farm power simulation.......................................................... 124
Figure 6.25 Aggregate wind farm power variation. ........................................................... 125
Figure 6.26 Aggregate wind farm power characteristics (at 20% turbulence intensity). ... 125
Figure 6.27 Power variations and power balance in the Nordel case studied. ................... 127
Figure 6.28 Wind power, frequency and standard deviation of power (mean wind speed at
12m/s to Finland)........................................................................................................ 128
Figure 6.29 Power spectral distribution of the power and speed of selected machines (mean
wind speed at 12m/s to Finland)................................................................................. 129
Figure 6.30 Power spectral distribution of power and speed to selected machines (AWF
modified to lower rotational speed (3p~0.5Hz)). ....................................................... 130
Figure 6.31 Wind power, frequency and standard deviation of power (average wind speed
16m/s to Finland)........................................................................................................ 131
Figure 6.32 Wind power and voltage deviations simulated in Nordel system (mean wind
speed 12m/s to Finland).............................................................................................. 132
Figure 6.33 Power spectral distribution of voltage and voltage deviation to selected
machines in the Nordel (Finland AWF at 12m/s)....................................................... 132
Figure 6.34 Power spectra distribution of voltage and voltage deviation simulated in Nordel
System (mean wind speed 12m/s to Finland AWF low frequency (3p=0.5Hz)). ...... 133
Figure 9.1. Electrical generator model parameters............................................................. 148
Figure 9.2. Electrical generator equivalent......................................................................... 148
Figure 9.3 Structure of the transformer model. .................................................................. 149
Figure 9.4 Electrical transformer model............................................................................. 149
Figure 9.5 Electrical Transmission structure model. .......................................................... 150
Figure 9.6 Electrical transmission model. .......................................................................... 150
Figure 9.7 Slack bus structure. ........................................................................................... 151
Figure 9.8 Slack bus model (positive sequence). ............................................................... 151
15/153
List of Tables
Table 2.1 Main steady state parameters defined in IEC 6140021 [1]..................................25
Table 2.2 Main power system influences from the wind energy integration........................36
Table 3.1 Typical values of surface roughness length z
0
for various types of terrain [53]...50
Table 3.2 Parameters used in the simulation for tower shadow............................................59
Table 3.3. Parameters of the wind turbine used in the measurement comparisons. .............60
Table 4.1 Basic characteristics of the Nortank 500kW wind turbine modelled....................72
Table 5.1 Basic characteristics of the 660kW wind turbine modelled..................................85
Table 5.2 Representative power characteristic values of all simulations..............................88
Table 6.1 Bus 3 loadability limits keeping the reactive power constant.............................103
Table 6.2 Bus 3 loadability limits keeping the active power constant................................104
Table 6.3 Relevant loads in the Brazilian power system studied........................................108
Table 6.4 Loadability limits to MOSSORO........................................................................110
Table 6.5 Loadability limits to MOSSORO with wind power............................................111
Table 6.6. Requirements on frequency response in the Nordel power system. ..................118
Table 6.7 Nordel reduced machines connection nodes. ......................................................120
Table 6.8 Wind power plans simulated (Figure 6.21 identifies the buses’ name). .............123
Table 6.9 Basic characteristics of the 1.83MW wind turbine. ............................................124
Table 6.10 Aggregate wind farms average wind speeds.....................................................126
16/152
Chapter 1
1 Introduction
Wind energy is said to be one of the most prominent sources of electrical energy in
years to come. The increasing concerns to environmental issues demand the search for
more sustainable electrical sources. Wind turbines along with solar energy and fuel cells are
possible solutions for the environmentalfriendly energy production. In this report, the
focus is on the wind power as it is said to hit large integration in the near future. This
technology has already reached a penetration level in some areas, which raises some
technical problems concerning grid integration.
Wind power has to overcome some technical as well as economical barriers if it
should produce a substantial part of the electricity. In this report, some of the technical
aspects are treated, particularly those regarding the power system quality and stability.
1.1 Motivation
It is possible to state that the significant impact of wind power started in the
beginning of 80s very much related to the mid 70s oil crises. During the period, a simple
and robust wind turbine concept emerged and became very popular pulling the wind power
industry. The simple and robust concept includes a three bladed wind turbine rotor, a
gearbox, an induction machine directly connected to the grid and a control system.
It was cheap and very robust but the power quality was poor and, in some cases, it
influenced the voltage level on the grid. During the 80s, most of wind power installations
were limited to few hundreds kilowatts to the existing distribution grids. The size of those
installations did not threaten the overall power system stability and the voltage quality
assessment was simple (when connected to the conventional power system). In this thesis,
the wind power installations to small isolated networks are not included. During the period,
the analysis concentrated on development of the wind turbine technology and investigation
of the dynamic behaviour of the wind turbines.
The 90’s represented an important break through; new concepts emerged because of a
demand for more efficient power production and to comply with power quality
requirements. During the 90s, the wind turbines (and farms) grew in size and ratio from the
few hundreds kilowatts to the megawatt size. The increased rated power of the wind farms
to areas with good wind resources leads to concerns on: “to which extent the wind power
interferes to the power system?” Most of the decade has been dedicated to voltage quality
analysis of wind power and to economics of the power systems including wind power.
In the late 90s, with the wind farms rating hundreds megawatts, the concerns start to
focus on the transient voltage stability of the power system. The studies focus on the
dynamic behaviour of the induction machines during disturbances, where the dynamic
effects of the turbulence were neglected. During the same period an International
Electrotechnical Commission (IEC) task force issued a standard procedure: the IEC 61400
21[1] to fillin the lack of technical standards on assessment of power quality from wind
turbines.
Nowadays, some power systems start to face problems of integrating thousands
megawatts of wind power, which are decentralised spread over large extensions. At this
17/152
moment, the problems of planning, operation and control of the power systems with large
wind power become very important [2]. On these problems, the main challenges are
classified in longterm planning, operation and energy management systems, and power
system performance.
The long term planning focuses on several topics, such as the adjustment of
agreements between transmission system operators to cope with the stochastic nature of the
wind power, this also includes economical and financial issues. Another important aspect
included in this topic relates the distribution system reliability with wind power.
The operation and energy management systems focus on the forecast of the wind
power and its relation to trade agreements. It also includes the security analysis and the
reserve of power to ensure reliable operation of the entire power system.
The last topic, the power system performance, focuses on the control of voltage and
frequency, power quality issues, and on the dynamic behaviour of the wind power sources.
This report focuses on the dynamic behaviour of the wind power resources to an extensive
area and its influences on the power quality and voltage stability issues.
1.2 Literature Review
Integration of wind power into the power system has been studied by many authors
before but most of them focused on different characteristics and issues of the power system.
Wind power influences several power system characteristics from economic dispatch to
stability and quality issues.
Wan and Brian in [3] pointed out main factors to utility integration of solar and wind
power where studies from late 1970s until 1980s provided a starting point and general
classification of the most relevant power system aspects. One of the main aims was to
define technical limits to intermittent power integration. One of the conclusions was that
there were no clear limits on wind power integrations and that penetration limits from many
studies were economic rather than technical limits. It also clearly concluded that the spatial
distribution of the wind must be taken into account and wind speed data are needed.
A project report (ALTENER) in [2] aimed to establish insight and to orient works in
the integration of renewable power in the European Network. It characterized the impacts
of large amount of renewable energy on European conventional utility practice (not only
operation but the institutional aspects also). This latter report related potential problems on
the power exchange agreements, stability of the network and power quality. Some of the
conclusions are that dynamic studies of the power system characteristics are required in
largescale renewable energy penetration and the stochastic nature of the wind speed is
relevant to the power system control and quality.
A more recent report by Nielsen et all in [4] reviews the technical options and
constrains of integration of distributed power generation. One of the focuses was on a new
power system structure to deal with the imbalance on consumption and production from the
stochastic nature of the wind power. Similar work was presented by Jørn et all in [5] where
the focus were on the trading/economic aspects from operation of large–scale wind power
in the Danish power system.
On the dynamic subject, several works have been done on the analysis of power
quality and on the transient stability from wind turbines to the power systems.
18/152
Tande et all in [6] developed an extensive analysis of the potential impacts of wind
turbines on the power quality, where the work focused on the small scale integration.
Similar, Sørensen in [7] and Larsson in [8] characterized the moderns wind turbines and
classified the most relevant characteristics that supported the IEC 6140021[1]. Tande in [9]
explained how to assess the voltage quality from wind turbines using wind turbine
characteristics. All works have focused on the power quality of wind turbines.
Akhmatov et all in [10], Brunt et all in [11] and Wiik et all in [12] presented transient
stability analysis of the power system with large amount of wind power. On those works,
the transient voltage collapse was investigated as well as the wind turbines behaviour
during short circuits.
On modelling wind turbines/farms, some textbooks present the wind turbine
characteristics. Freris in [13] and Hau in [14] both detailed the wind energy conversion
systems presenting a general overview of the wind turbine and Heier in [15] addressed
specifically the integration of wind power where much effort was given on the
characterization of the wind turbine/farms components.
Wilkie et all in [16] presented simple models to wind turbines and argued that the
contributions of wind turbines components must be modelled into the overall simulation but
the total accuracy was not essential to obtain an adequate representation. Estanqueiro in
[17] and Petru in [18] presented wind turbine models that could be used to power quality
studies.
Sørensen et all in [19] and Estanqueiro et all in [20] showed that the assessment of
power quality from wind farms depends on appropriate representation of the wind speed. In
wind farms, the turbulence spatial correlation must be included.
Giebel in [21] showed that the spatial distribution of the wind turbines gives benefits
in long term because it reduces the power variations from the wind energy. Beyer et all in
[22] showed that the spatial correlation of the wind speed influences the power production
of wind farms. From both works, the power variations from wind farms can be reduced
significantly particularly in large scale applications. However, none of them analysed the
results to the dynamic power systems.
The dynamic power system analyses have been extensively investigated. Kundur et
all in [23] and Anderson and Fouad in [24] present dynamic models to the power system
components and means to analyse the stability of it. Some of the analyses include the
simulations of the entire power system operation under normal operation and faults. The
Power System Engineering Committee of the IEEE in [25] resume several works on power
system analysis specifically on modal analyses for system dynamic performance that can
identify the possible problems on normal operation of the power system, however the
modal analysis must be complemented with dynamic simulations.
Dynamic simulations of very large power systems are very expensive. Lei et all in
[26] and Eliasson in [42] present dynamic reduction of large power systems for stability
studies aggregating machines with similar (coherent) dynamic characteristics. With the
reduced model, the overall dynamic analyses are time feasible. Aggregate dynamic models
are used to simulate integration of wind power to the dynamic operation of the power
system in this thesis.
Voltage stability has been pointed out as another problem to large integration of wind
power because wind farms demand reactive power. Taylor in [27] and Custem in [28]
19/152
presented an extensive explanation of the voltage stability problem and means of copying
with it. In addition, the Power System Stability Subcommittee of the IEEE in [29] and
Cañizares in [30] suggest several tools to analyse the voltage stability, one of them is the
loadability curves to characterize the maximum load that can be installed before the voltage
collapses. The loadability curves are used in this thesis in connection with the wind power.
1.3 Wind Power Basics
The wind turbines are composed of an aerodynamic rotor, a mechanical transmission
system, an electrical generator, a control system, limited reactive power compensation and
a stepup transformer. The conventional wind turbine is even at the present time, the most
common type of wind turbine installed. Figure 1.1 presents the basic components of a
conventional wind turbine.
Rotor
Blades
Generator
Wind Sensors
Gear Box
Nacelle
Brake
Tower
Yaw
Control
Tip Break
Rotor
Blades
Generator
Wind Sensors
Gear Box
Nacelle
Brake
Tower
Yaw
Control
Tip Break
Network
Stepup
transformer
Figure 1.1 Basic components of a wind turbine unity.
The conventional wind turbine is connected directly to the grid and the generator is
“synchronized” to the network. This technology has been named “fixed” rotational speed
wind turbine because the induction generator allows small mechanical speed variations.
The main power system problems from this wind turbine technology come from the
lack of control on the active and reactive powers. The active and reactive power control is
very important to keep the frequency and voltage stable within limits. Lack of reactive
power can lead to voltage problems and no control in the active power can cause frequency
deviations.
This report focus on this wind turbine technology influence on the power system
voltage stability and on the power system quality. Because of lacking controls on active and
reactive power, this wind turbine technology is considered the poorest power quality when
harmonics problems are not concerned.
In addition, this report focus on the largescale integration hence a substantial part of
this report is dedicated to model the power fluctuations from large groups of wind turbines.
The power produced from a large number of wind turbines will vary relatively less than the
power produced from a single wind turbine due to the cancellation effect from the poor
20/152
spatial correlation of the wind acting on each wind turbine. Figure 1.2 illustrates the power
“smoothing” effect when increasing the number of wind turbines.
150 Wind Turbines
30 Wind Turbines 300 Wind Turbines
0 100 200 300 400 500 600
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
Time [s ]
P
o
w
e
r [p
.u
.]
0 100 200 300 400 500 600
0. 3
0. 4
0. 5
0. 6
0. 7
0. 8
0. 9
1
1. 1
Time [s ]
P
ow
er [p.u.]
0 100 200 300 400 500 600
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
Time [s ]
P
o
w
e
r [p
.u
.]
0 100 200 300 400 500 600
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
Time [s ]
P
o
w
e
r [p
.u
.]
1 Wind
Turbine
Power(p.u.) Power(p.u.) Power(p.u.) Power(p.u.)
150 Wind Turbines
30 Wind Turbines 300 Wind Turbines
0 100 200 300 400 500 600
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
Time [s ]
P
o
w
e
r [p
.u
.]
0 100 200 300 400 500 600
0. 3
0. 4
0. 5
0. 6
0. 7
0. 8
0. 9
1
1. 1
Time [s ]
P
ow
er [p.u.]
0 100 200 300 400 500 600
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
Time [s ]
P
o
w
e
r [p
.u
.]
0 100 200 300 400 500 600
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
Time [s ]
P
o
w
e
r [p
.u
.]
1 Wind
Turbine
Power(p.u.) Power(p.u.) Power(p.u.) Power(p.u.)
Figure 1.2 Power smoothing effect from wind farms.
The power variation from wind turbines is very complex and demand special
techniques to cope with the spatial distribution of the wind turbines than a simple scale up
from a single wind turbine.
In addition to the problems of dynamic power fluctuations, another important issue
investigated in this report is the voltage stability from connecting large amount of wind
power.
The voltage stability in the power system can be classified in slow dynamic and
transient. The slow dynamic is related to slow increase in load in the power system and
deals with the reactive and active power supply. Once the wind turbines have limited
reactive power compensation and usually demands reactive power from the power system,
here its influences of the reactive power demand and the active power injection are
investigated.
Transient voltage stability problems have also been related to large integration of
wind turbines. The transient voltage stability deals with the voltage stability after the power
system being subjected to large disturbances, normally short circuits. Here transient
stability is not studied.
1.4 Thesis Outline
Chapter 2 addresses the main problems of the power system that are related to the
wind turbines. It starts introducing the main power quality characteristics of the wind
turbines and proceeds to present the main problems from integrating single wind turbines in
the power system are presented. After having presented the power system interactions with
21/152
wind turbines, the possible problems to large integration of wind turbines in to the power
system, based on a scale up from a single wind turbine, are discussed.
Chapter 3 presents the main characteristics of the wind acting on the rotor of wind
turbines. The wind is classified in two main components, and the most relevant
characteristics of each part are described. In the final section, a suitable wind speed model
to assess the power quality from wind turbines is presented.
Chapter 4 presents the main components from wind turbines. It starts dividing the
wind turbine in aeroelastic and electrical components. The aeroelastic components present
the relevant characteristics and models to the wind turbine aerodynamic rotor and drive
train. In this thesis, the dynamic wind turbine model is implemented in SIMPOW/ABB
1
,
where available dynamic models to the electrical components are used. The dynamic wind
turbine model however can be directly applied to other simulation tools. In the last part of
the chapter, the wind turbine model to power quality assessment is presented and compared
against measurements.
Chapter 5 presents the turbulence coherence effects on the wind farm production. The
main characteristics of coherence and its influences on the dynamic wind farm power
production are presented. Chapter 5 also presents an Aggregate Wind Farm (AWF) model
that can be used to power quality assessment of wind farms and to dynamic stability
analysis of power systems with large number of wind turbines. The AWF model is a single
equivalent wind turbine that replaces several wind turbines in the wind farm. The wind
speed to the AWF takes into account the wind farm layout and the spatial coherence.
Chapter 6 presents some illustration cases of the voltage stability analysis and power
quality from large integration of wind turbines. First, the large wind farms impacts on the
voltage stability to a part of a power system are illustrated. In the second part, the
connection of large amount of wind power to a Brazilian network is presented in terms of
voltage stability and quality. Finally, the integration of more than 4GW of wind power to
the Nordic Power System illustrates application of the aggregate wind farm.
1
SIMPOW
®
is a dedicated digital simulation tool to power system dynamic analysis developed by ABB
©
,
www.abb.se
22/152
Chapter 2
2 Wind Power Integration
Large integration of wind power can lead to problems on the voltage control or on the
stability of the power systems as mentioned in chapter 1. Chapter 2 presents the main
impacts from wind power on the power system with emphasis on the voltage stability and
power quality.
The power quality and stability problems, and means of coping with them, are not
new to the power system engineers. However, those problems related to wind power are not
well described when it comes to largescale integration.
Here the main problems to the voltage stability and to the power quality related to
largescale integration of wind power are presented. The problem is introduced by pointing
out the relevant power quality characteristics of a small wind farm (or a single wind
turbine) and after it is scaled up to represent the largescale case.
First, the scales of integration are defined as follow:
• Smallscale wind power integration – the wind power installed is relatively
small compared to the conventional power system. The wind energy counts
for small part of the total energy production in the power system (i.e. up to
few percents e.g. 2%). In smallscale integration, the power system is assumed
to have enough spinning reserve of active power and the frequency is kept
constant therefore only voltage problems are concerned.
• Largescale wind power integration – the wind power installed sizes the
conventional power stations. The wind power counts for large part of the total
energy production in the power system (e.g. above 10%). The largescale
integration can cause power quality or stability problems and, in some
particular cases, the frequency ca be affected by the wind turbines. Hence, the
voltage and frequency problems are concerned.
2.1 SmallScale Integration of Wind Power
In this case, the power system is considered strong and the main problems from
connecting wind farms come from the voltage control. The wind farm is composed by
several wind turbines. Each wind turbine has as basic electrical components: an induction
generator, local reactive power compensation and a stepup transformer. The wind farm
limit is defined by the Point of Common Coupling (PCC), additionally the wind farm may
use an integration transformer to connect to a higher voltage level e.g. transmission
systems.
Figure 2.1 presents the relevant electrical components of a conventional wind farm. In
this thesis, the focus is on the direct connected wind turbines type (so called “fixed” rotor
speed). On those types of wind turbines, the active power that comes from the wind is
transferred to the power system without storage devices.
23/152
WTG
WTG
To the next WT
Network
PCC
StepUp
transformer
StepUp
transformer
Integration
transformer
Capacitor
Capacitor
To the next WT
Figure 2.1 Basic components of a wind farm.
Induction (or asynchronous) machines applied as generators demand reactive power
from the network (chapter 4), which is partially compensated with shunt capacitor banks. In
special configurations, special reactive power compensation is demanded and installed at
the PCC (e.g. variable reactive power compensation). In special installations, the voltage
level/variations at the PCC can demand a variable tap change transformer.
Wind farms have very little control of the active power due to the stochastic
behaviour of the wind, in addition, the voltage control on this type of wind turbine can be
done only by changing the amount of reactive power compensation (shunt capacitors
installed). The lack of control on the active and reactive powers can disturb the voltage on
the PCC.
The disturbances on the voltage from wind turbines are classified in different time
scales. The classification presented as follow is in agreement with IEC 6140021 [1]:
• Steadystate – does not include dynamics (very slow dynamics representing
periods above 10 minutes to hours);
• Dynamic – include the dynamics in the time frame from milliseconds to 10
minutes;
• Harmonics – includes voltage variations in high frequency (e.g. above 50Hz
to Europe and 60Hz to US and Brazil, periods less than 20 milliseconds) due
to the electronic equipments installed in the wind turbines (this last part is not
related to the wind speed).
Here, the harmonics are not an important issue because this report focus on the direct
connected wind turbine type that does not emit harmonics components on current, therefore
it is not included in the following subsections.
2.1.1 Steady – state operation
The steadystate operational analysis assures that the:
• The currents will not exceed thermal limits nor will the protections act during
extreme powers;
• The voltage levels will not exceed limits.
24/152
In near future, wind turbines can be certified to power quality [1]. From these
certifications, a set of data will help to verify the steadystate operation of the wind turbines
and wind farms. Table 2.1 presents the main steadystate parameters to wind turbines
certified to power quality [1].
Table 2.1 Main steady state parameters defined in IEC 6140021 [1].
Parameters
Rated active power
Rated reactive power
Rated apparent power
Rated current of the wind turbine at rated Voltage
Rated voltage of the wind turbine
Maximum permitted power setup in the controller
Maximum measured power in 60 seconds average period
Maximum recorded power in 0,2 seconds average period
Reactive power demand/supply as function of the active power
Reactive power measured or estimated to the P
mc
Reactive power measured or estimated to the P
60
Reactive power measured or estimated to the P
0,2
Based on the parameters specified on Table 2.1 and with the electrical characteristics
of the network it is possible to determine the impacts on the voltage quality as well as the
maximum currents on the cables and transformers.
The impacts on the voltage quality to the different conditions as expressed in Table
2.1 can be computed with help of a load flow program or by simple equations. Following,
simple equations to determine the voltage levels are introduced. Figure 2.2 presents the
electrical representation of the wind turbine and the power system, where the reference
node and the equivalent impedance represent the entire power system at the wind turbine
terminals.
Reference
Node
S=P + j Q
Z = R + j X
Wind turbine terminals
U∠δ
~
U
0
∠0
Figure 2.2 Single line equivalent for a wind turbine connection.
P and Q are the active and reactive powers respectively from the wind turbine, there
are no load or shunt elements installed at the wind turbine terminals. The voltage at the
wind turbine terminals (U) can be determined as follow:
U U U ∆ + =
0
(2.1)
where, U
0
is the voltage at the reference node, U is the voltage at the wind turbine
terminals and ∆U can be computed as:
25/152
( ) ( )
0 0
U
j QR PX
U
QX PR
U
⋅ −
+
+
= ∆
(2.2)
where, R and X are the resistance and reactance inductive characteristics of the electrical
network respectively. The voltage can increase or decrease depending on the amount of the
reactive and active power flux and on the network characteristics. Using Equation (2.2) it is
possible to compute the voltage levels and compare to preset limits imposed by the local
network operator.
However, in most cases, it is important to detail the network and include the loads
installed and to use a load flow program to compute the voltage and currents on the relevant
nodes and lines respectively.
2.1.2 Dynamic Operation
The wind turbines dynamically produce power that varies in a broad range of
frequencies and amplitudes. These continuous variations of active and reactive powers from
the wind farm cause dynamic voltage variations. The dynamic voltage variations from the
wind turbines during operation are quantified by flicker and step change [1].
The flicker emissions during continuous and switching operations and the voltage
step change are the voltage quality indicators influenced from small number of wind
turbines connected to the grid. The flicker emission is computed from flicker coefficients
measured from wind turbines during the power quality data sheet.
The flicker emission is a measure of the human perception of the bulb light variation
consequent of the voltage low frequency variation. The value is computed to short term (10
minutes) and long term (120 minutes). The flicker emission includes voltage variations in
frequencies up to 25 Hertz that are weighted with an eye perception function according to
[32] and its posterior amendments [33] and [34].
Figure 2.3 presents the voltage fluctuations as a function of frequency that will
represent a unity of shortterm flicker perceptivity (P
st
) to two different conditions:
sinusoidal voltage fluctuations and rectangular voltage fluctuations based on [34].
26/152
Figure 2.3 Voltage fluctuations corresponding to flicker emission unity [34].
From Figure 2.3, the maximum flicker perception comes from around 8Hz where the
voltage fluctuations must be reduced in order to respect the flicker limits. The limits on
Figure 2.3 considers that voltage variations leads to light intensity variations in light bulbs.
In order evaluate the flicker contribution from wind turbines, Figure 2.4 presents the
Power Spectral Distribution (PSD) of the power produced from a three bladed wind turbine
(225kW).
0.00E+00
5.00E+01
1.00E+02
1.50E+02
2.00E+02
2.50E+02
3.00E+02
3.50E+02
4.00E+02
4.50E+02
5.00E+02
0. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.
Frequency (Hz)
Srot*f (kW²)
Figure 2.4 Measured power spectra of the electrical power from a 225kW pitch regulated wind turbine.
The PSD in Figure 2.4 includes contributions from deterministic and stochastic parts.
The fundamental frequency of rotation (1p – one time the rotational speed of the rotor) is
approximately 0.7Hz, at this frequency there is a small contribution related to some
asymmetry in the rotor. At the frequency of 2.1Hz (3p – three times the rotational speed of
the rotor) there is a large contribution to the power variation. The 3p effect is related to
27/152
rotational turbulence and the blades passing the tower in a threebladed rotor type of wind
turbines. In the frequency of 8.4Hz, corresponding to 12p, a small amount of energy is also
presented that has been related to the flexible aeroelastic part of the wind turbine in addition
to the induction generator [66]. PSD of power measurements from different threebladed
wind turbines show similar pattern, i.e. the power variations reduce significantly above the
frequency of 3p. Although one could expect high power variations in a broad frequency
range, a three bladed rotor cancels the multiples harmonics different from the 3np and in
addition, the dynamic components of the wind turbines damp the high frequency power
oscillations.
The power variations are consequence of the wind field on the rotor area and the wind
turbine dynamics. The turbulence and tower shadow influence the wind field on the rotor
area and the three blades crossing the wind field transfer the power variations to the main
shaft. The power on the main shaft will dynamically interact with the wind turbine
components, e.g. drive train torsional moments, and finally the generator will convert the
power to the network.
Figure 2.3 and Figure 2.4 indicate that the main flicker contributions from wind
turbines comes from the 3p power variations. The power variation in very low frequency
below 0.7 Hz is caused by the simple turbulence acting on the rotor area but has small
influence on the flicker. The highenergy content on 3p frequency comes from the effect of
the blades rotating on the turbulent field added together to the tower shadow. These effects
will be more detailed in chapters to come where each component of the power fluctuation
from a wind turbine is explained and models presented to simulate them.
The flicker defined in the previous paragraphs is related to the continuous operation
of the wind turbines. In addition, wind turbines also generate flicker due to switching
operation and startup. The flicker during continuous operation is caused by the power
fluctuation from the turbulence added to the wind turbine dynamics. The flicker due to
switching operations is caused by startup or switching of generators of wind turbines
because the high inrush currents cause voltage dips. Associated to the flicker emissions
during switching operations, the voltage dip is relevant because the voltage will drop
instantaneously due to the inrush current.
This report focus on the analysis of the continuous operation of wind turbines,
therefore it is restricted to the flicker emission during continuous operation. The flicker and
voltage dips from switching operations are not treated in this report.
The flicker emission in short term and long term can be estimated from the power
quality tests [1]. The power quality tests of wind turbines express a flicker coefficient for
each wind turbine for different network phase angle condition and different annual mean
wind speeds for a wind farm. The shortterm flicker emission (P
st
) and the longterm flicker
emission (P
lt
) to a wind farm can be estimated according to [1]:
( ) ( )
∑
=
⋅ = =
wt
N
i
i n a k i
k
lt st
S v c
S
P P
1
2
,
,
1
ψ
(2.3)
where S
k
is the short circuit capacity, c
i
is the flicker coefficient of wind turbine i to specific
network impedance phase angle (ψ
k
) and annual average wind speed v
a
from the site, S
n,i
is
the rated power of wind turbine i and N
wt
is the number of wind turbines in the wind farm.
28/152
The P
st
and P
lt
are assumed the same because it is assumed that the mean wind speed and
turbulence will be maintained in 10 minutes average as well as in 120 minutes.
Equation (2.3) takes into account the “cancellation” effects, which comes from the
wind dynamics in the wind farm that is not correlated, so the flicker is not a linear sum of
all flicker produced from each wind turbine. Because the 3p is the main flicker contribution
and these relatively high frequencies are approximately uncorrelated this is a reasonable
assumption.
2.2 LargeScale Integration of Wind Energy
As introduced before, here the largescale integration problems are based on the
smallscale ones. The largescale integration means a relatively high wind power compared
to the local power system. The large integration can occur in two main conditions:
• Large wind farms connected to the transmission system or;
• Several small wind farms connected to the distribution systems in one area of
the power system.
In either condition, the power quality and system stability assessment become more
complex and depending on the sizes, they demand special investigations of voltage and
frequency variations.
In the smallscale integration, the frequency was assumed constant. With high wind
power capacity installed, the large active power variations can interact with the frequency
controllers in the conventional power stations, so frequency variations can happens. In
addition, large reactive power demanded by the wind farms can reduce the reactive power
supply, hence the voltage stability limits can be reduced and must be analysed too.
There are several issues arising from largescale wind power integration, but here the
focus is on the voltage stability and the dynamic power oscillations during normal operation
of the power system. In order to introduce the maim issues of the power system and wind
turbines, Figure 2.5 presents a simple single line with the basic structure of the power
system, where it is possible to distinguish the main system components.
Generation System Transmission System Distribution System
Loads
Loads
Loads
Loads
Figure 2.5 Basic Power System Structure.
The generation system is mainly composed by synchronous machines that are usually
large. The transmission system is composed by transmission lines that extend for large
29/152
distances and interconnect different generation units. The transmission lines demand special
consideration in controlling the voltage at the terminals due to reactive power flow (in AC
type lines). Distribution systems delivery power to the loads where the voltage level is
lower. The distribution lines require special attention to control the voltage at the loads.
The power system must supply a reliable and quality electrical power to the loads. In
order to achieve reliability, the power system must have reserves and controllers that can
deliver the power when it is demanded, task mainly supplied by conventional generators
and controllers installed throughout the power system. On the other hand, active controllers
compensate the voltage and frequency variations keeping the power quality within limits.
The power system quality and stability depends mainly on the power system
controllability [23], assuming that the power is available. Figure 2.6 presents a simple
equivalent of the entire power system including the main controllers where only
conventional equipments are included.
GS
P +j Q
R
tran
X
tran
Frequency
control Voltage control
Voltage control
V
term
U
field
f
term
q
term
V
term
X
control
Figure 2.6 Single line equivalent of a Power System.
In Figure 2.6, the apparent power supplied to the load (P+jQ) flows through the
transmission and distribution system from the generators stations – GS. The power flux
results in voltage variations compensated near to the loads with decentralized voltage
controllers – by adding or reducing reactive power – and in the generation stations with
voltage controllers that change the excitation level of the synchronous machines. Variations
on the active power result in frequency deviation that speed governors act to keep the
balance on consumption and production increasing or reducing the primemover power.
The voltage controllers are mainly related to the reactive power while the frequency
controllers to the active power ([23] and [27]).
Wind turbines are a special kind of generators, which has none or little voltage and
frequency control capabilities and they supply an intermittent power. In addition, wind
turbines in general use asynchronous generators that demand reactive power from the
network to its excitation. The reactive power demanded to the wind farms is partially
compensated by capacitor banks and the network supplies the rest of the reactive power.
The active power produced from wind farms varies all the time and leads to
continuous power flux variations. On the generation stations, it leads to continuous action
of the frequency controllers to keep the balance on production and consumption (and the
frequency constant).
30/152
2.2.1 Voltage Stability Problem
A definition to the voltage stability phenomenon has not been widely accepted yet.
Nevertheless, several task forces have worked on basic definitions of voltage stability. For
instance, the IEEE Task Force Report in [29] defines the voltage stability in terms of the
ability of maintain voltage so that when load is increased, load power will increase hence
voltage and power are controlled.
Although voltage stability definition is not widely accepted, the voltage collapse is
well recognized. Here, the voltage collapse occurs if after an increase in load or power
injection, the voltages are below acceptable levels followed by a progressive and
uncontrollable decline in voltage [29].
The voltage stable operation means that the voltages near to loads are identical or
close to the predisturbance values [27], where disturbances may be a simple load increase
or a variation in power from a wind farm.
The voltage collapse in general results from an incident of voltage instability. The
voltage instability phenomenon is defined here as having crossed the maximum deliverable
power limit, the mechanism of load power restoration becomes unstable, reducing instead
of increasing the power consumed [29]. The voltage instability event can grow to voltage
collapse leading to entire or a large part of the power system with very low voltage profile.
Here, in order to illustrate the voltage collapse, a simple formula based on the load
flow calculations is introduced. Figure 2.7 presents a single line diagram used to define
simple analytical equations to voltage stability (please note that the shunt elements are not
included).
U∠ δ °
Z = R + j X
S=P + j Q
Reference Node
U 0
∠ 0 °
Figure 2.7 Simplified transmission line equivalent diagram.
above, U
0
is the infinite node voltage, Z is the impedance characteristic to the specific node
(also called short circuit impedance). The voltage difference between the two nodes can be
defined as:
∗
∗
⋅ = −
U
S
Z U U
0
(2.4)
Assuming U
0
real and rewriting Equation (2.4) as:
( ) ( ) ( jQ P jX R jU U U U U
S Z U U U U
I R I R
− ⋅ + + − ⋅ = +
⋅ + ⋅ = ⋅
∗ ∗ ∗
0
2 2
0
)
(2.5)
31/152
and remembering that U* = U
R
 jU
I
. In addition, a function H is defined as H =ZS*= H
R
+
jH
I
. Isolating the imaginary part of the voltage (U
I
) as:
( )
0 0
U
H
U
QR PX
U
I
I
=
−
=
(2.6)
and inserting in Equation (2.5):
( )
( )
0
0
0
2
0
2
0
2
0
2
= − ⋅ −


.

\

+
= + − ⋅ −


.

\
 −
+
R R
I
R
R R
H U U
U
H
U
XQ RP U U
U
QR PX
U
(2.7)
The real part of the voltage in the node can be defined as:
(
(
(
¸
(
¸


.

\

−


.

\

− ± =
R
I
R
H
U
H
U U U
2
0
2
0 0
4
2
1
(2.8)
For the sake of simplicity, it is assumed that the infinite node voltage U
0
=1 p.u., then
Equation (2.8) becomes very simple as:
( )
(
¸
(
¸
− − ± =
R I R
H H U
2
4
1
2
1
(2.9)
Adding the real part of the voltage in Equation (2.9) and the imaginary in Equation
(2.6) results in the voltage at the node as follow:
( )
I R I
jH H H U +
(
¸
(
¸
− − ± =
2
4
1
2
1
(2.10)
Now it is possible to state that:
• If H
R
– H
I
2
< –1/4 – there is no physical solution.
• If H
R
– H
I
2
= –1/4 – both solutions coalesce and the point of voltage collapse
is reached.
• If H
R
– H
I
2
> –1/4 – the solution is double: one physical and one spurious
(unstable).
The voltage stability is complex and, even to a very simple case, includes the load,
network characteristics, and the voltage at the sending node.
Using the diagram in Figure 2.7, it is possible to illustrate the maximum power
transferred to a specific node as a relation of the voltage at that node (Figure 2.8) to
different load factor conditions.
32/152

U

P/Pmax
Figure 2.8 Power transfer to a node as function of the voltage (“nose curve”).
Figure 2.8 is a simple illustration of the relation between the power transmitted to a
node and its voltage. This curve is the Voltage vs. Power characteristic of the node also
called “nosecurve”. Using Figure 2.8, the voltage stability limit is characterised by the
vertical tangent at the nose point that is in fact the maximum power transmitted to the node
in agreement with definitions above.
When transferring the problem to the complex power system, the relevant factors that
lead to voltage instability are: the transmission lines and power transfer strength of the
power system; generator reactive power/voltage control limits [35]; load characteristics
[36]; characteristic of reactive power compensation; and the action of voltage control
devices such as under load tap transformers [37].
The dispersed voltage controllers acting on the distribution grid also influence the
voltage stability. Distribution grids uses under load tap changer transformer, which under
voltage instability events, tends to increase the problem by increasing the current flow in
order to reestablish the set voltage level.
In addition, the characteristics of the reactive power compensation devices with the
voltage contribute to the voltage instability. Usually, shunt capacitor banks compensate the
reactive power in the power system. The reactive power supplied from shunt capacitors is
related to the squared of the voltage. Hence, when started a voltage decline it will reduce
the local reactive power production stressing even further the transmission lines and
reducing further the voltage level. Similarly, the loads response to voltage changes
influences the voltage stability.
Finally, large number of wind farms on power systems (high penetration) demands
reactive power and in addition some synchronous generators (generation stations) are shut
down in order to achieve cheaper energy production. Under this condition, the excessive
demand for reactive power can be a problem.
2.2.1.1 Analysis of Voltage Stability
The analysis of voltage stability for a given power system involves examination of
several aspects, e.g. distance to voltage collapse, mechanisms that lead to voltage collapse
33/152
among others [27]. The voltage stability problem has been discussed in a large number of
papers and always analysed by means of expensive and complicated models. A
comprehensive reference list can be found in [28].
The voltage stability here includes periods from 15 minutes to hours, being a dynamic
problem rather than static [29]. However, as the dynamics involved in the voltage stability
problem are very slow, the voltage instability is analysed by static models.
The load flow problem is very closely associated with voltage stability analysis [27].
The load flow programs determine the operational characteristics of the power system
based on the load schedule and voltage reference in the generation units.
In load flow problems, the Jacobian (NewtonRaphson algorithm) represents a linear
relation between the power and voltage at a specific operational point. When the voltage
collapses, the maximum power transferred was reached and the Jacobian becomes singular.
At that point, there is no solution to the load flow problem and that is the maximum
transmissible load. The use of the Jacobian properties has been pointed by several authors
where modal analysis and voltage collapse proximity indicators have been proposed [29].
Here, the voltage stability is defined in terms loadability curves to a specific node
before voltage collapses [28]. The loadability curves are similar to the “nosecurve” in
Figure 2.8, however, the lower part (unstable part) of the curve is not simulated here. The
load to a specific node of the power system is stressed until the Jacobian matrix becomes
singular and, at that point, the maximum load is defined.
The loadability curves are computed using a loadability computation tool based on a
load flow program using NewtonRaphson algorithm, which was implemented in Matlab as
part of this project.
The loadability curve indicates the maximum load increase in the power system under
specific conditions. Here, the loadability curve is also used to define the maximum wind
power to the power system. The wind power to a node is increased until the maximum
power transfer is reached. The loadability curves are similar to the injection of wind power
but the power to the node is injected instead of a drained.
2.2.2 Frequency Control Problem
The power system has a nominal frequency for which all generators are synchronized.
In a synchronized system, the power is naturally shared between different generators based
on the rate of the rating of the generators or as defined by the system operators.
The frequency control in 10 minutes can be classified in primary and secondary [38].
The primary control is fast control actions to keep the instantaneous balance between
production and consumption. Secondary control is slow control actions to reestablish
nominal system frequency and scheduled power interchanges. This thesis focuses on the
primary control of frequency because it is the controller acting instantaneously to avoid
frequency deviations related to stochastic variations from wind turbines.
The primary frequency control is done by speed governors, which automatically
adjust the prime movers driving the generator to keep the balance on consumption and
generation. The speed governor acts based on speed deviation where, “(…) most automatic
controls use high gain negative feedback, which, by its active nature, can cause oscillations
to grow in amplitude with time.”[39].
34/152
Originally, the interconnected generators were fairly close to one another, and
oscillations were at frequencies of the order of 1 to 2Hz ([23] and [24]). Dampers windings
on the generators were used to prevent the oscillations to grow. Increasing the demands for
reliability makes the rapid automatic voltage controllers that are used to prevent the
generator loosing synchronism following a system fault. This fast action tends to reduce the
damping of the system oscillations hence special Power System Stabilizers – PSS – were
designed to damp those oscillations [39]. The power system is very complex and specific
components can interact with other causing oscillations.
From an operating point of view, oscillations are acceptable as long as they decay.
Oscillations are a characteristic of the power system, which are initiated by the normal
small changes in the system load and, in this report, from wind farms power variations. The
wind farm power produced can be viewed as a continuous negative load variation that
demands the speed governors to act all the time to keep the balance on the system.
The power oscillations can be sustained due to the controller natural characteristic.
Although the oscillations are not expected to increase in time, the sustained variations can
become a problem.
2.2.2.1 Analysis of Power System Oscillations
The power system oscillations analysis may be done by modal analysis
complemented by dynamic simulations ([23], [24], [25] [39] [40] and [41]). The modal
analysis uses a linear representation to the entire power system being suitable to analyse
small disturbances in the power system (it is also called small signal stability analysis of the
power system). The modal analysis however must be used with caution because it does not
include the nonlinear behaviour of equipments on the electrical power system. Hence, it is
also important to include dynamic simulations to analyse the power system oscillations
during normal operation.
The modal analysis has been recognized as one of the most reliable tools to analyse
power systems ([23] and [25]). With the modal analysis it is possible to define the
eigenvalues of the power system hence it is possible to identify the electromechanical
oscillation modes.
The power system can be described by equations that include: all electromechanical
characteristics, the network equations and the controllers. These equations can be linearised
on an operational point. Where, the linear model put on the form of a set of first order
differential equations with constant coefficients (state equations) have the form:
         
          u d x c y
u B x A x
⋅ + ⋅ =
⋅ + ⋅ = &
(2.11)
where, x is a vector of the dynamic state variables, A is the state matrix, B is the input
matrix, y is the output (or measured) variables, c is the output coefficient matrix and d is a
matrix of coefficients describing the direct connection between the input and output
variables: the vector u.
The eigenvalues of A represents the roots of the state equations defined as the values
of s that satisfies:
35/152
  0 det = − I s A
i
(2.12)
where each solution s
i
defines a time function e
(si t)
that satisfies the state equations [25].
The function is called a mode of oscillation of the system.
If all eigenvalues have a negative real part all modes decay with time and the system
is said to be stable. If one eigenvalue has a positive real part, the corresponding mode will
grow exponentially in time being unstable.
The modal analysis characterizes the modes of oscillation of the power system and
the less damped oscillation modes, i.e. the modes that the respective eigenvalues are close
to the positive plane. Hence, the possible problems in the frequency control in the power
system are characterized.
However, the modal analysis must be accompanied with dynamic simulations in
order to include the nonlinear characteristic of the power system [39]. Hence, dynamic
models to largescale wind power must be applied to the power system dynamic
simulations to analyse the interactions between the wind power and frequency controllers
during normal operation of the power system, in addition to the modal analysis.
Dynamic simulations including the entire power systems are very expensive. In order
to analyse the main characteristics of the power system, dynamic reduced models are used
as pointed in [26] and [42]. The dynamic reduction is an aggregation of machines and loads
with similar dynamic performance that reduces the number of equations to describe the
power system and does not lose important information [26]. In this thesis, dynamic reduced
models are used to analyse the power system performance with wind power.
2.3 Remarks on Wind Energy Integration
The integration problems caused by wind power have been discussed in this chapter.
The problems were classified in smallscale and largescale integration. Table 2.2 presents
the main integrations problems from wind power [20].
Table 2.2 Main power system influences from the wind energy integration
Integration
scale
Problems Causes
Steady state voltage rise Wind speed variation
Overcurrent Peaks of wind speed
Protection error action Peaks of wind speed
Flicker emission during
continuous operation
Dynamic operation of wind turbines
Flicker emission during
switching operations
Switching/start up operation of generators
Voltage drop In rush current due to switching operations of generators
S
m
a
l
l
S
c
a
l
e
Harmonics Power electronic converters
Power system Oscillations Inability of the power system controllers to cope with the
power variations from the wind farm and loads
L
a
r
g
e
S
c
a
l
e
Voltage stability Reactive power limitations and excessive reactive power
demand from the power system
The smallscale wind power integration into power system is well investigated and
there are plenty of tools available to analyse its interaction. The problems from connecting
36/152
small numbers of wind turbines to the power system are related to the voltage quality and
thermal capacity of the lines and cables.
The largescale integration of wind farms, however, it is not properly investigated yet
and it is very complex. The largescale problems include all problems from the smallscale
and in addition problems of voltage stability and power system quality. The Voltage
stability problem is related to the limitation of power transfer. The extensive use of
capacitors banks to compensate reactive power can result in larger problems in event of
islanding e.g. selfexcitation of the induction generators.
Large disturbances such as short circuits in sites with large amount of wind turbines
can result in voltage transient instability due to the reduced capacity in transferring reactive
power and the huge demand from the asynchronous generators as well as transformers after
the fault is removed.
The power system quality can deviate from limits. The voltage quality deviates
because of the power flux in the network, and in largescale integration, the frequency can
deviate from its limits as well.
The main conclusions are:
• Smallscale integration of wind energy deals mainly with voltage quality and
reliability;
• Large integration of wind turbines includes the smallscale problems and
overall stability problems;
• The analysis of largescale integration demands suitable models to represent
wind power;
• Network characteristics play an important role on the power system stability
and quality;
• Wind turbines power characteristics play also another important role in the
overall power system stability and quality
In order to investigate the wind power influences on the power system quality and
stability it is imperative the use of appropriate models to represent the wind farm and the
power system. The wind turbine models must have the following characteristics:
• Dynamics models for the wind speed acting on each wind turbine;
• The spatial wind coherence in largescale;
• Detailed model to the wind turbine dynamics;
• Proper representation of the power flux of the wind turbine generator;
• Time feasible in order to analyse large power system.
Moreover, the power system models must have the following characteristics:
• Include proper models to the electrical components;
• Include models to the loads;
• Be flexible to allow the implementation of wind turbine/farm models and
interact with other simulation tools;
37/152
• Input/output capabilities in order to analyse the data;
• Allow the implementation of large power systems;
• In large integration, represent properly the electromechanical modes in the
power system.
38/152
Chapter 3
3 Wind Speed Model
Wind turbines produce a complex and continuously fluctuating power. A large part of
the complexity resides on the input: the wind. The main source of power variation on
conventional wind turbines is the wind speed variation.
The wind is complex and the blades crossing the wind field modify the power
fluctuations. The main objective of this chapter is to present a dynamic wind model for
power quality assessment of a three bladed upwind horizontal axis wind turbine type. The
wind speed model includes the turbulence and tower shadow in the rotor area.
Figure 3.1 illustrates an example of the wind field acting on the rotor area of a wind
turbine (borrowed from [43]).
Wind speed
Spatially correlated
turbulence
Mean wind speed
profile
Figure 3.1 Illustration of the wind on the rotor area of a wind turbine [43].
The wind is classified in two main parts: the first part represents a mean wind speed
profile over the rotor area, here defined as deterministic; and the second part is turbulence
on top of the deterministic (see Figure 3.1), here assumed as stochastic.
The deterministic part of the wind over the rotor area is assumed “constant” in 10
minutes period and the wind variations are the time variant part that has a stochastic
behaviour. This assumption is valid only in periods up to few minutes because above this
there is a slow wind variation in the mean wind speed due to continuous change in the
atmosphere.
Returning to Figure 3.1, the wind acting on each rotating blade section is different
from the wind in a stationary reference frame. The blades pass through different wind
39/152
speeds in each revolution. The process of passing several times through this wind speed
field results in power variations at n times the rotational speed revolution of the rotor called
np’s.
Figure 3.2 presents a measured time series of active power produced from a
500kWstall regulated, threebladed wind turbine type (direct connected to the grid). Figure
3.2 also shows the power variations classified in fast and slow where the mean value is
removed.
0 100 200 300 400 500 600
150
100
50
0
50
100
150
200
250
300
350
Time (s)
A
c
t
i
v
e
p
o
w
e
r
(
k
W
)
Measurement
Slow variations
Fast variations
Measurements
Low frequency
power variations
High frequency
power variations
Figure 3.2 Power produced by a 500kW stall regulated wind turbine.
In Figure 3.2 the active power measured from a three bladed wind turbine vary in a
broad frequency range. It is possible to decompose the measurements in slow and fast
power variations where the slow power variation includes frequencies up to 0.5Hz and the
fast power variation includes all frequencies above 0.5Hz.
The low frequency power variations are the largest variations and they are related to
turbulence on the rotor area as it has been recorded in anemometers. In this particular case,
the slow power variations are within +80kW and–110kW, which is the main responsible for
high standard deviation of power produced from wind turbines.
The high frequency power variations are much smaller, in this case within ±10kW.
The high frequency power variations cannot be directly related to the turbulence but to the
dynamics of the wind turbine including the blade rotation in the wind field. The relatively
high frequency power variation has been related as one of the main responsible for the
flicker problems [60] as introduced in chapter 2.
This chapter presents the general main windrelated power output variations from
wind turbines. The aim is to reproduce the mechanical torque using an equivalent wind
speed model to the entire rotor. The Equivalent Wind Speed (EWS) model simulates the
wind by an equivalent time series that when applied to a specific aerodynamic model
reproduces the aerodynamic torque from a wind turbine, which applied to a drive
train/generator model simulates the electrical power from a real wind turbine (Figure 3.3).
The dynamics of the wind turbines are explained in chapter 4 where the wind turbine
components are introduced.
40/152
W
Aerodynamic model
T(t)
W(t)
Aerodynamic torque
Drive train/
generator P
el
(t)
Wind turbine Network
Electric Power
Wind speed
Figure 3.3 General overview of wind turbine models.
In Figure 3.3, W is the wind speed over the rotor area, T is the aerodynamic torque
and P
el
is the electrical power transferred to the utility network.
3.1 Model Description
Measurements of turbulence in a stationary reference frame show that the amplitudes
of the wind speed variation reduce with the frequency. Above a few Hertz, the wind speed
variations are insignificant [44]. However, the rotation of the blades causes variations in the
effective wind acting on each blade section on n times the rotor speed frequencies.
Figure 3.4 presents the Power Spectral Density (PSD) of the wind measured on a
rotating blade section [45]. The wind speed variation on multiples frequencies of the rotor
(that in this case is 0.5 Hz) is clear.
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
10
−3
10
−2
10
−1
10
0
10
1
10
2
Frequency (Hz)
(
(
m
/
s
)
2
/
H
z
)
PSD of wind measured on rotating blade section
Figure 3.4 Wind speed measured on a section of a rotating blade [45].
The PSD (in Figure 3.4) shows that each blade section of the wind turbine
experiences high wind speed variations on the rotational speed of the rotor. In the low
frequency range from 0 to 1.5Hz are the main wind variations. Above that frequency, the
wind speed variations are very reduced mostly because of the dynamics of turbulence.
41/152
The measurements were done on a single section of a rotating blade, hence when
adding up the wind effects on the remaining others two blades in a symmetrical three
bladed rotor of a wind turbine, all the frequencies different from 3 times the rotor speed
(3np) are cancelled. The cancellation effect is expected due to the symmetric position of the
threebladed rotor and has been supported by power measured from wind turbines. Figure
3.5 presents the PSD of the measured electrical power from a three bladed wind turbine,
500kW, stall regulated and operating with in constant speed.
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
10
−2
10
−1
10
0
10
1
10
2
10
3
10
4
Frequency (Hz)
P
S
D
o
f
a
c
t
i
v
e
p
o
w
e
r
(
k
W
2
/
h
z
)
3p
1p
Figure 3.5 PSD of measured electrical power output of a 500kW stall regulated wind turbine.
The PSD in Figure 3.5 correspond to the measured time series of power presented in
Figure 3.2. The fundamental frequency (1p) is approximately 0.4Hz (indicated in the
figure). At this frequency, a small contribution is present that is related to some asymmetry
in the rotor. At the frequency of 0.8Hz (2p), a small power variation is presented that is also
related to some asymmetry on the rotor. The torsional drivetrain mode (dynamic from the
wind turbine) is around 1.08Hz that gives a small power variation. The most relevant power
variation is at 3p frequency (1.2Hz), which is relevant to the power quality analysis hence it
is retained in the dynamic wind model. At the frequency range from 6 to 11Hz, a small
amount of energy is presented, which is related to a complex mode of the wind turbine. The
complex mode is related to a dynamic interaction between the electrical generator and the
flexible wind turbine, it is not related to the wind turbulence.
In this chapter, only the windinduced power variations that are relevant to power
quality are treated hence the dynamics of the wind turbine is not included in this chapter but
explained in chapter 4. Considering the measured PSD of power produced from wind
turbines and the measured wind speed on a rotating blade, the main wind related power
variations are from the 0 and 3 times the revolution of the rotor.
The wind speed model represents the power variations from the wind in the frequency
range of interest as a time domain model. It generates 10 minutes period time series. The 10
minutes period was selected for several reasons, first of all the 10 minutes period avoid
most of the cyclical variations related to meteorological phenomena (e.g. diurnal variations
related to periods of hours). Second, the period matches the time frame of the primary
42/152
frequency controllers from power systems, which its performance with large amount of
wind power is one of the aims of this report. Moreover, the voltage quality assessment (e.g.
statistics and flicker) in [46] as well as wind turbines characteristics, e.g. power curve, are
based on 10 minutes measurements
In 10 minutes period, the stochastic and deterministic parts can be combined to
express the total wind at one position as:
) , , , ( ) , , ( ) , , , ( t z y x g z y x U t z y x W + = (3.1)
where W(x,y,z,t) is the total wind speed in position xyz at time t, U(x,y,z) is the deterministic
wind speed that is assumed time independent and g(x,y,z,t) is the stochastic part that has
mean value zero. Figure 3.6 presents the references axis and notation of the wind turbines.
The origin of the coordinate system is the centre of the nacelle precisely at the height of the
centre of the rotor, so the zaxis is negative in the direction to ground.
x
y
z
Top view Front view
x
z
⊕y
z
x
y
Figure 3.6 Reference axis used in the wind turbine.
3.1.1 Equivalent Wind Speed Model
The Equivalent Wind Speed (EWS) model is an equivalent for the entire rotor that
takes into account the stochastic and deterministic wind speed actions on the rotor area. The
EWS applied to an aerodynamic function simulates the real torque on the main shaft of
wind turbines. Figure 3.7 illustrates the principle of the EWS.
43/152
Wind Speed
W
Aerodynamic forces
T(t)
W(t)
Aerodynamic torque
Drive train
and generator P
el
(t)
Equivalent
wind speed Aerodynamic model
W
eq
Ψ
eq
T(t)
Weq(t)
Drive train
and generator
P
el
(t)
Electric Power
Wind Wind turbine Network
Electric Power
Aerodynamic torque
Equivalent
Aerodynamic
torque
Figure 3.7 Equivalent Wind Speed Model principle.
The upper part of Figure 3.7 illustrates how the power is produced in a real wind
turbine. The wind speed (W), which is different at each rotor position for each time, acts on
the blades of the wind turbine. The aerodynamic blades convert the wind by complex
aerodynamic effects on mechanical power on the mechanical shaft of the wind turbine
(aerodynamic torque (T) times the rotor speed). The mechanical power is then converted
into electrical power in the electrical generator. The lower part of Figure 3.7 illustrates the
main idea of the EWS, which is to reproduce the same mechanical torque using a single
equivalent wind speed time series applied to an aerodynamic model. Because the rotational
speed depends on the electrical generator, it is assumed that the electrical power produced
by the EWS is the same as the one produced in a real wind turbine.
3.1.2 Description of the Equivalent Wind Model
The wind acting on the rotor area of a wind turbine can be modelled in several ways.
The objective of the model defines the type and details to be retained of the wind model.
The EWS aims to simulate the electrical power output variations from a threebladed wind
turbine.
Usually the wind models assumes that the continuous wind field on the rotor area can
be replaced by discrete time series of wind speeds in a grid of points on the rotor area. This
process is very common and reliable, however, it generates many time series that are only
used partly and generally, it consumes much simulation time [47].
Yet, in the grid wind model, an aerodynamic routine interpolates between the
appropriate points of the wind grid to generate a continuous wind speed at a given section
(radius) of the rotor blade. Based on the continuous wind speeds in a number of blade
sections, an aerodynamic module calculates the aerodynamic torque on the main shaft.
Finally, the aerodynamic torque enters a dynamic model of the drive train and electrical
generator, which will output the electrical power.
The EWS replaces all wind time series on the rotor area and a relatively simple
aerodynamic function supplies the aerodynamic torque. The main advantages of the
44/152
equivalent wind speed model are: fast computation and reduced memory requirements. The
single equivalent wind speed model is also very suitable for simultaneous simulation of a
large number of wind turbines making it possible to efficiently estimate the impact of a
large wind farm on the power quality.
The Equations (3.2) to (3.4) below provide a more formal description of the
aerodynamic model for a 3bladed wind turbine shown in Figure 3.7. Applying blade
element theory, an aerodynamic coefficient (Ψ(r)) on a given blade section r is calculated.
Ψ(r) and W (t, r, φ) are used to calculate the contribution from a single blade to the
aerodynamic torque M(t,φ) according to Equation (3.2). It must be noticed that in the
following equations polar coordinates is used: r,φ are the rotor position in polar coordinates
where r is the radial position and φ is the angular position of the blade section.
∫
⋅ Ψ =
R
r
dr r t W r t M
0
) , , ( ) ( ) , ( φ φ
(3.2)
where R is the length of the blade, and r
0
is the inner radius, where the aerodynamic forces
start to develop. The effect of all three blades are then added to give the total aerodynamic
torque T(t,φ) using:
∑
=
=
3
1
) , ( ) , (
b
b
t M t T φ φ
(3.3)
For the summation in Equation (3.3), it can be assumed that the azimuth position φ
b
of blade number b is given by:
3 , 2 , 1
), 1 (
3
2
=
− + =
b
b
b
π
φ φ
(3.4)
The Equations (3.2) to (3.4) describe how the aerodynamic torque is generated
physically. The equivalent wind speed is based on an expansion of the wind speed field in
the azimuth angle (φ). A similar expansion was done by Madsen and Rasmussen [48] using
real expansion coefficients. Using complex expansion coefficients, the expansion of the
wind speed field formally becomes much simpler as given in Equation (3.5).
∑
∞
−∞ =
=
n
jn
n
e r t W r t W
φ
φ ) , (
~
) , , (
} {
(3.5)
) , (
~
} {
r t W
n
is the complex azimuth expansion coefficient. This azimuth expansion coefficient
indicates the amplitude of the n
th
harmonic in the rotational speed [49]. Inserting (3.5) into
(3.2) provides the expanded expression for M(t,φ) in:
∑
∞
−∞ =
⋅ =
n
jn
n
e t M t M
φ
φ ) (
~
) , (
} {
(3.6)
45/152
the azimuth expansion coefficients ) (
~
} {
t M
n
are given in:
∫
⋅ Ψ =
R
r
n n
dr r t W r t M
0
) , (
~
) ( ) (
~
} { } {
(3.7)
Inserting Equation (3.7) into Equation (3.3), and further using Equation (3.4) yields
the azimuth expansion of the torque T(t,φ) according to:
∑
∞
−∞ =
⋅ =
n
n j
n
e t T t T
φ
φ
3
} 3 {
) (
~
) , (
(3.8)
where the azimuth expansion coefficients of the torque ) (
~
} 3 {
t
n
T are given as:
) (
~
3 ) (
~
} 3 { } 3 {
t M t T
n n
⋅ = (3.9)
From Equation (3.8), the summation of the contributions from the three blades
remove harmonics which are not multiple of three. This effect will be complete if the rotor
is symmetrical, which e.g. requires that the blades have the same tip pitch angle. Substantial
first harmonics have been measured on wind turbines with unsymmetrical rotors typically
due to pitch misalignment.
Now defining the equivalent wind speed W
eq
(t,φ) according to:
∑
∞
−∞ =
=
n
n j
n
eq eq
e t W t W
φ
φ
3
} 3 {
) (
~
) , (
(3.10)
with the weighted azimuth expansion coefficients ) (
~
} 3 .{
t
n eq
W given in:
∫
∫
Ψ
⋅ Ψ
=
R
r
R
r
n
n eq
dr r
dr r t W r
t W
0
0
) (
) , (
~
) (
) (
~
} 3 {
} 3 .{
(3.11)
Finally, the equivalent torque T
eq
(t,φ) is defined as:
) , ( ) , ( φ φ t W t T
eq eq eq
⋅ Ψ =
(3.12)
with the equivalent aerodynamic influence coefficient Ψ
eq
defined according to:
∫
Ψ = Ψ
R
r
eq
dr r
0
) ( 3
(3.13)
Using the definitions in the Equations (3.10) to (3.13), it can be shown that the
equivalent wind speed W
eq
(t,φ) generates the same torque T
eq
(t,φ) as the wind speed field
v(t,φ), i.e. T
eq
(t,φ) = T(t,φ).
46/152
In principle, the aerodynamic load influence coefficient Ψ(r) has to be determined
individually based on the geometry and hence the distribution forces along the blade radius
r. In the present model, it has been assumed that the contribution to the torque is
proportional to the radius, i.e. Ψ(r)=k⋅r and r
0
is 10 % of the radius R. These assumptions
are based on the physical aerodynamic performance of the blades that excluding the region
in the beginning of the blade (root) and the region near to the tip presents the torque
proportional to the radius, fact that has been observed in different blades and have been
justified in [50].
The equivalent wind speed is the sum of the deterministic and stochastic components,
i.e.
) , ( ) ( ) , (
, det ,
φ φ φ t g U t W
sto eq eq eq
+ =
(3.14)
The deterministic part, as explained before, is independent of the time. It will only
influence the dynamics of the wind turbine, because the blades rotate.
In the following sections, the implementation of EWS model is presented after
having introduced the main deterministic and stochastic components. The EWS
implementation is based on Equation (3.10), however, the EWS only includes the 0p and
the 3p harmonics components because they are the most relevant to power quality
assessment, an assumption that is supported by measurements.
3.1.3 Deterministic Part of the Wind
3.1.3.1 Tower Shadow
Tower shadow along with mean wind speed are the most relevant deterministic effect
on the electrical power output from threebladed wind turbines. Horizontal axis wind
turbines always have some form of tower support structure. The tower can be lattice or
cylindrical. Usual, small wind turbines use lattice towers and large new wind turbines use
cylindrical towers.
Towers are obstacles to the free wind that modifies the wind flow. The upstream flow
is reduced in front the tower and increased laterally. Downstream the tower makes a wake
effect that reduces the horizontal wind (Figure 3.8).
Here the model is limited to the effects on horizontal upwind wind turbines type,
hence the relevant effect is the upstream flow. Mathematically the wind field can be
calculated by approximating the tower to a cylinder and assuming the existence of a two
dimensional potential flow ([13], [61] and [51]). The flow is then decomposed in the
longitudinal and lateral wind components, i.e. the wind in the y direction and x direction
respectively.
The horizontal longitudinal (yaxis) is the most relevant to power variations because
it turns to be the wind converted into active power. The wind far from the wind turbine is
called the ambient wind speed (U
h
) and it is not interfered by the tower. When the wind
comes close to the tower, the tower deviates the wind as shown in Figure 3.8.
The potential flow for the wind is constant all places around the tower, and can be
expressed as [51]:
47/152
( )
ts
d
a
ts h wind
d U
2
) sin( − ⋅ = β ψ
(3.15)
ψ
wind
is the potential flow of the wind in polar coordinates where d
ts
is the distance of the
blade section to the tower centre and β is the angle between the blade section and the
horizontal flux. Decomposing the wind in the rotor area in radial – U
hr
– and tangential –
U
ht
– components as illustrated in Figure 3.8.The radial and tangential wind speed
components can be derived from the potential flow as Equations (3.16) and (3.17) (from
[51]):
( ) ) cos( 1
1
2
2
β
β
ψ
ts
d
a
h
wind
ts
hr
U
d
U − − =
∂
∂
− =
(3.16)
( ) ) sin( 1
2
2
β
ψ
ts
d
a
h
wind
ht
U
d
U + =
∂
∂
=
(3.17)
a
Upstream
Tower
U
hr
U
ht
U
h
y
x
d
ts
β
Downstream
Figure 3.8 The tower shadow effects on the horizontal wind (top view).
Finally, the wind speed components presented in Equations (3.16) and (3.17) are
converted to the xy axis (longitudinal and lateral components) of the wind turbine.
( )
¦
¹
¦
´
¦
≤ ≤
≤ ≤ ⋅
=
π φ π
π φ β
2 for 0
0 for ) 2 sin(
2
2
ts
d
a
h
hx
U
U
(3.18)
( )
¦
¹
¦
´
¦
≤ ≤
≤ ≤ ⋅ −
=
π φ π
π φ β
2 for
0 for ) 2 cos( 1
2
2
h
d
a
h
hy
U
U
U
ts
(3.19)
where φ is the rotor azimuth position angle, β is the angle between the blade section and the
rotor axis and d
ts
is the distance from the tower centre to the blade section expressed as
(3.20).
2 2 2
y x d
ts
+ = (3.20)
The tower is finite, thus the tower shadow will only influences the semi plane up to
the tower height in terms of azimuth angles (φ) the tower influences the range from [0 π].
48/152
Figure 3.9 illustrates the tower shadow effect on the wind speed field in front of the rotor
on an upwind type rotor and a cylindrical tower.
Figure 3.9 Wind speed field interference by the tower shadow.
In this particular case, the tower reduces 20% the wind speed in front of the tower and
the tower influences only the lower semi plane. Above the tower, it has been assumed that
the wind is not modified by the tower presence because only the horizontal component of
the wind is modelled. Only the horizontal part of the wind was modelled in the first place
because it is the most relevant to the aerodynamic process of generating lift hence
converting the wind power into mechanical power.
Figure 3.10 illustrates the effects of tower shadow on the main shaft torque of three
bladed wind turbines rotors.
Harmonic components (np s)
F
F
T
o
f
t
h
e
n
o
r
m
a
l
i
z
e
d
t
o
r
q
u
e
0 pi/2 pi 3pi/2 2pi
0.94
0.95
0.96
0.97
0.98
0.99
1
1.01
M
o
m
e
n
t
(
p
.
u
.
)
Equivalent Moment
Single blade moment
0 5 10 15 20 25 30 35 40
0
0.5
1
1.5
2
2.5
3
x 10
−3
Figure 3.10 Normalised torque influenced by the tower shadow.
49/152
The upper part of Figure 3.10 presents the time series of normalised moment in a
three bladed wind turbine rotor. The tower influences each one of the blades generating
several harmonic components on the torque. The lower part of Figure 3.10 presents a
Fourier transformation of the normalised moment influenced by tower shadow, which
reveals that practically only the 3np components are presented in it. The wind turbine
dynamics filter out the higher 3np frequency elements in Figure 3.10 in accordance to the
PSD of the measured power output from wind turbines that shows small fluctuations in
frequencies different from 3p (Figure 3.5)
3.1.3.2 Wind Shear
The friction between the ground and the moving air generates a vertical wind profile
where the mean wind speed increases with the altitude. This process is called wind shear.
The ground/wind friction influences the wind speed up to hundreds meters, acting on the
rotor area of all commercial wind turbines [13].
The wind shear has a strong relation with the site. A smooth site, with small
obstacles, e.g. sea with small waves, has a small wind speed variation with the height. On
the other hand, the urban areas present a high wind speed variation with the height. That
phenomenon is expected because in smooth areas the friction is much lower.
In general, the electrical power outputted from a three bladed wind turbine has very
little influences from the wind shear because it is quite linear. However, the wind shear is
very important to blade loads analyses but it is not transmitted to the electrical power.
The wind speed relation between two different heights can be expressed according to
Equation (3.21) [13], derived from the Prandtl logarithm law model [52].
( )
( )
0
2
0
1
ln
ln
) (
) (
2
1
z
z h
z
z h
z h U
z h U
+
+
=
+
+
(3.21)
where z
1
and z
2
are the heights from the rotor centre (see Figure 3.6 for reference axis), h is
the height of the centre of the rotor, U is the wind speed and z
0
is the roughness length that
characterizes the terrain. The wind shear modifies only the vertical profile i.e. there is no
influences on lateral space xaxis or to the wind speed related to the position of the wind
turbine –yaxis.
The z
0
of a terrain is a measurement of the roughness, which can be determined in
various ways, one very common is to make comparisons with sites for which z
0
has been
determined from measurements and decide the most suitable one [13]. Typical values for
various types of surfaces are given in Table 3.1 [53].
Table 3.1 Typical values of surface roughness length z
0
for various types of terrain [53].
Type of terrain Z
0
(m)
Smooth sea 2.0 – 3.0 x 10
4
Sand 0.2 – 1.0 x 10
3
Low grass 1.0 – 4.0 x 10
2
High grass 0.4 – 1.0 x 10
1
Forest 0.1 – 1.0
City 1.0 – 4.0
50/152
Using Equation (3.21), Figure 3.11 illustrates wind shears for different sites with an
average wind speed of 10 m/s at 10 m height.
0 2 4 6 8 10 12 14 16 18 20
0
10
20
30
40
50
60
70
80
Wind speed (m/s)
H
e
i
g
t
h
(
m
)
Smooth sea
High grass site
Urban site
Figure 3.11 Wind shear for different sites (ground is reference for height).
The wind speed difference with the height depends mainly on the site. The wind
turbine blades rotating on the wind sheared leads to cyclical loads on the mechanical
torque. In smooth sites the effects are reduced leading to reduced power variations, on the
other hand in sites with tall obstacles, the wind shear leads to higher wind speed variation
hence higher cyclic torque variations. A wind turbine applied to wind field sheared (e.g.
Figure 3.11) will suffer cyclic loads because when the blades rotate on the field, the height
a blade section (point p in Figure 3.12) changes according to:
) cos(φ ⋅ − = r h z
p
(3.22)
Hence, each blade section in position p is subjected to the wind speed U
p
:
( ) ( ) ( )
( )


.

\



.

\
 −
⋅ = −
0
0
ln
cos
ln
cos
z
h
z
r h
h U r h U
p
φ
φ
(3.23)
where the variables are presented in Figure 3.12 as well as the reference axis used here.
51/152
Blade
rotation
h
π/2
X p
r
z
φ
X p
y
Blade
representation
Swept area
Front view
x
z
Side view
Figure 3.12 Reference axis and angles used in the wind turbine.
Where, φ is the angular position of a blade, p is a point in the rotating blade section
distant r from the root and h is the height of the wind turbine.
Using Equation (3.23), Figure 3.13 illustrates the normalised mechanical torque of a
three bladed wind turbine from the wind shear in a site with z
0
=0.05 (the wind turbine is the
same used to compute the tower shadow effects in Figure 3.10).
Harmonic components (np s)
F
F
T
o
f
t
h
e
t
o
r
q
u
e
0 pi/2 pi 3pi/2 2pi
0.9
0.95
1
1.05
1.1
Azimuth angle φ (Rad)
M
o
m
e
n
t
/
E
q
u
i
v
a
l
e
n
t
w
i
n
d
s
p
e
e
d
(
p
.
u
.
)
Blade 1
Blade 2
Blade 3
Torque at the main shaft
0 5 10 15 20 25 30 35
0
2
4
6
8
x 10
−4
Figure 3.13. Normalised torque influenced by wind shear (site with a medium z
0
).
The upper part in Figure 3.13 presents the normalised torque variations due to wind
shear. The 120° displacements between the three blades cancel the symmetrical torque
resulting in cancellation of torque contributions in frequencies that are different from 3np.
The FFT of the resulting torque presented in the lower part of Figure 3.13 shows that in
52/152
symmetrical rotors the total torque on the main shaft presents mostly 3p frequency
variations. The variation in the 3p frequency is very small.
The effects from the tower shadow were much more relevant (the FFT in 3p was
around 2.5x10
3
) compared to the effects from wind shear (6x10
4
). The wind shear effect is
very important in mechanical and fatigue analysis of blade loads but to the power quality, it
is neglected. Currently, the deterministic part of the EWS model only includes the tower
shadow and the mean wind speed as it is presented in the following section.
3.1.4 Implementation of the Deterministic Component
The deterministic part represents only the tower shadow effect, in particular the
interference from tubular towers in up wind turbines type. The implemented deterministic
model is illustrated as block diagrams in Figure 3.14, where “detr3rd” is the amplitude
from the contribution of the 3
rd
harmonic calculated based in the geometric parameters of
the wind turbine, “um” is the mean wind speed specified and “Wsdet” is the output
deterministic wind time series [54].
1
WSdet
Product1
Product
um
Mean
Wi nd
s
1
Integrator
detr3rd
Effect due tower shadow
cos
Cos(3u)
1
w rotor
Figure 3.14 Implementation of the deterministic model in Simulink/MATLAB.
The implementation in Figure 3.14 first generates the rotor position φ as the integral
of the rotor speed, which is a feedback from the wind turbine model. The remaining
implementation in principle corresponds to Equation (3.10). However, Equation (3.10) is
based on complex numbers and for the implementation only the mean value and the third
harmonics are included. In addition, only the cos part of the third harmonic is included
because the tower shadow is symmetric, so the imaginary part of the third harmonic
expansion coefficient is zero.
3.1.5 Stochastic Part – Turbulence
The stochastic part of the wind, in this thesis called turbulence, is the time variant part
of the wind acting on the rotor area. Turbulence is the wind speed variations in a broad
range from seconds to minutes. The variations have naturally a random behaviour but the
air dynamics creates a main pattern on the wind speed variations. Low frequency variations
have large amplitudes and higher frequency variations have lower amplitudes. Recording
the definition in Equation (3.1), the turbulence can be measured in terms of its variance:
  ( )  
2 2
W W Ε − Ε = σ (3.24)
where E[ ] is the mean expectation operator, and W is the wind speed. The turbulence is
also expressed in terms of standard deviation divided by the average wind speed (U
0
=
E[W]) – so named turbulence intensity – I
u
, defined as:
53/152
0
U
I
u
σ
=
(3.25)
3.1.5.1 Power Spectral Density of Turbulence
The turbulence varies in a broad frequency range. Spectral density function of the
turbulence has been investigated for several years and several models have been proposed.
In this report, a socalled Kaimal turbulence model is used to represent the power spectral
of turbulence [44], defined as:
3
5
0
0
2
5 . 1 1
) (


.

\



.

\
 ⋅
⋅ +
⋅
=
⋅
U
x f
U
x f
f S f
L
L
σ
(3.26)
where f is the frequency of the turbulence, σ
2
is the variance, x
L
is the turbulence length
scale (proposed maximum 600 m) and U
0
is the average wind speed. Figure 3.15 presents
the logarithm power spectra (S(f)*f) of the turbulence simulated with Kaimal model to an
average wind speed of 10m/s, turbulence intensity of 10% and turbulence length scale
600m.
10
−3
10
−2
10
−1
10
0
10
1
10
−11
10
−10
10
−9
10
−8
10
−7
10
−6
10
−5
10
−4
10
−3
10
−2
10
−1
Frequency (Hz)
P
S
D
*
f
(
(
m
/
s
)
2
)
Figure 3.15. Kaimal PSD of Turbulence.
The Kaimal power spectra pattern was fitted based on several experimental data
collected with neutral atmosphere over flat homogeneous terrain in Kansas as explained in
[44]. The Kaimal power spectra has a peak when x
L
⋅f/U
0
=1 and decreases with the –5/3 of
the frequency, hence, the turbulence variations tend to zero in high frequency.
Several studies (e.g. the review presented in [55]) try to reveal the more accurate
turbulence spectral model. Here, the Kaimal spectrum was selected to model turbulence.
However, the application of other model is straightforward. The choice for Kaimal is done
because it is widely accepted and because the Kaimal is suggested according to the Danish
standard for loads and safety of wind turbines construction [56].
54/152
3.1.5.2 Coherence of the Wind
The coherence of the wind is relevant because it shows the correlation between wind
speeds in different positions. The correlation of the wind in two different locations depends
on the average wind speed, the distance between the positions and the frequency.
Near positions are well correlated but it reduces as the distance increases. In similar
way, turbulence is highly correlated in lower frequency and decreases as the frequency
increases, these characteristics have been measured as presented in [57].
The EWS uses the Davenport coherence type to simulate the coherence in the rotor
area. The Davenport type coherence has the following form [57]:
( )
0
,
U
f d
k
e d f
⋅
−
= γ
(3.27)
where γ is the coherence, U
0
is the mean wind speed, f the frequency, d is the distance
between two positions and k is a decay factor. The decay factor has originally been
expressed as a constant equal to 7.7 [57]. Measurements in other sites however reveal a
different constant value that is in the range from 2 to 27 that is mostly related to the ratio
d/h (distance divided between the two points and the height of the measurements (h)) as
presented in [55].
Considering the application to wind turbines and the Danish standard for loads and
safety of wind turbines construction [56], the EWS uses 12 for the decay factor constant.
3.1.5.3 Power Spectral Density of a Rotating Blade – Rotational Turbulence
The blades rotation generates special power variations. As for the deterministic part
of the wind, the blades rotation will move energy contents of the turbulence from low
frequency to multiple frequencies of the rotor speed, a process usually called “rotational
sampling of turbulence”[59].
Wind turbine blades pass through different turbulences several times. It results that
the power spectra of the turbulence on a blade section of a wind turbine has high energy
content in the multiples of the rotational speed of the wind turbine. The rotational sampling
of turbulence has been reported as the main cause of flicker from wind turbines, which
makes its inclusion on wind speed model relevant to power quality assessments [60].
The rotational sampling of turbulence has been studied for several years and means to
simulate it has been proposed by different ways. One of the methods includes the
generation of several correlated time series of wind speed on the rotor area of the wind
turbine, so called wind grid model. Then interpolation of appropriate points generates the
wind on a rotating blade section, where the appropriate points comes from the expected
position of the rotating wind turbine blade as presented in [47]. This method demands
generation of several correlated time series of wind speeds, which result in large memory
usage.
The method used here is based on equivalent wind speeds to each rotational sampling
frequency, which is explained in the following subsection.
55/152
3.1.6 Implementation of the Stochastic Component
The basic component of the stochastic component is the turbulence wind speed. The
turbulence wind speed is simulated by applying a “Kaimal” filter to a random number
generator. The Kaimal filter is fitted to output the Kaimal power spectral. The filter used in
the EWS was fitted in [50].
The stochastic component includes the rotational sampling turbulence that is
simulated the same way as the deterministic component, i.e. first, the azimuth expansion
coefficients are simulated, and then they are used as amplitudes to the harmonics according
to Equation (3.10) (only 0 and 3p). However, the azimuth expansion coefficients are time
dependent in this case.
The azimuth expansion coefficients are random signals [49], and the PSD of the n
th
azimuth expansion coefficient
{ }
) (
3
ω ∆
n w
S is equal to the PSD of the wind speed in a fixed
point ) ( ω ∆
w
S (here Kaimal), multiplied by an admittance function ) (
} 3 {
~ ω ∆
n w
F , i.e.:
{ }
) ( ) ( ) (
} 3 {
~
3
ω ω ω ∆ ⋅ ∆ = ∆
w n w n w
S F S
(3.28)
where S
w
is the Kaimal distribution and the admittance function can be determined by
a triple integral, which is [49]:
} 3 {
~
n
w
F
( ) ( ) ( )
( )
2
0
2
0
1
0
2 1 2 1
~
, ,
) (
} 3 {


.

\

= ∆
∫
∫ ∫
R
R R
k
w
dr r
dr dr r r f F r r
F
n
ψ
ψ ψ
ω
(3.29)
where ψ is the aerodynamic function, f is the frequency, r is the radial position along the
blade on the rotor area and F
k
defined as:
( ) ( ) φ φ γ
π
π
d d f r r f F
k
cos ,
2
1
) , , (
2
0
2 1
∫
=
(3.30)
where γ(f,d) is the square root coherence according to Equation (3.27) where the distance d
is the is ( ) φ cos 2
2 1
2
2
2
1
r r r r − + , r
1
and r
2
are two radial positions and φ is the angular
difference between them.
In the EWS, the admittance function is represented by a second order filter with a
transfer function H
adm{0p}
fitted in [50]. The EWS only accounts to the 0p and 3p as they are
the most relevant to three bladed wind turbine rotor types.
The filter applied to the 0p harmonic component (from [50]) is:
1 3518 . 7 6823 . 7
9904 . 0 7869 . 4
) (
2 2
2 2
} 0 {
+ ⋅ ⋅ + ⋅ ⋅
+ ⋅ ⋅
=
s d s d
s d
s H
TF TF
TF
p adm
(3.31)
56/152
where d
TF
= R/U
0
is the normalization parameter, R is the total radius of the wind turbine
rotor and U
0
is the average wind speed. Figure 3.16 presents the normalised numerical and
the fitted admittance function to the 0p (source: [50]).
10
−3
10
−2
10
−1
10
0
10
1
10
2
10
3
10
−5
10
−4
10
−3
10
−2
10
−1
10
0
Frequency normalised w (1)
A
d
m
it
t
a
n
c
e
F
u
n
c
t
io
n
F
(
w
)
(
1
)
Numerical
Fitted
10
−3
10
−2
10
−1
10
0
10
1
10
2
10
3
10
−5
10
−4
10
−3
10
−2
10
−1
Frequency normalised w (1)
A
d
m
it
t
a
n
c
e
F
u
n
c
t
io
n
F
(
w
)
(
1
)
Numerical
Fitted
Figure 3.16 Normalised admittance function to 0p.
To find the real rotational frequency (in rad/s) in Figure 3.16 is necessary to scale the
normalised frequency (w) with the constant of normalisation (d
T.F.
), i.e. w⋅d
T.F.
. The
admittance function to the 0p harmonic component can be understood as a smoothing
function to the variations of the turbulence on a stationary reference frame that are
transferred to the mechanical power. In low frequency, the turbulence variations are totally
transferred to the aerodynamic power while on high frequency, the variations are filtered
out due to the coherence of the turbulence.
To the 3p harmonic component, the transfer function H
adm{3p}
is fitted to [50]:
1 7722 . 1 3691 . 0
0307 . 0 2766 . 0
) (
2 2 } 3 {
+ ⋅ ⋅ + ⋅ ⋅
+ ⋅ ⋅
=
s d s d
s d
s H
TF TF
TF
p adm
(3.32)
Figure 3.17 presents the normalised numerical and fitted admittance function to the
0p (source: [50]).
It is important to remind that the normalised frequency to the 3p harmonic component
is shifted by a frequency of 3p from Equation (3.10). Hence, to find the real frequency it is
necessary to scale the normalised frequency (w) with the normalisation factor (d
T.F.
) and
with the 3p rotational speed (in rad/s – w
3p
), i.e. w
3p
±w⋅d
T.F
⋅
Figure 3.17 Normalised admittance function to 3p.
57/152
The admittance function to the 3p harmonic component can be understood as a
smoothing procedure to the variations of the turbulence that are transferred to the
mechanical power similar to the 0p. However, the reduction effect is much higher
compared to the 0p admittance function because the coherence decays much more due to
the higher frequency (3 times the rotational speed). At lower and above than 3p
frequencies, the variations from turbulence are filtered out due to the coherence of the
turbulence.
From Figure 3.17 it is possible to verify a poor agreement between the numerical
result and the fitted admittance function at very low frequencies. The difference at very low
frequency happens because the numerical results were fitted to a second order filter.
However, at these lower frequencies, the 3p admittance function is quite small and,
therefore not so important. In addition, a fourth order filter does an excellent fit but the
improvements in terms of power variations on the results did not justify the use of higher
order filter [50].
Finally, Figure 3.18 illustrates the total implementation of the stochastic part of the
EWS implemented in Simulink/MATLAB
®
.
cos(u[1])
cos3azm
Sum
Product
Kaimal_Filter_2 Admit3_Filter_1 3 Harmonic
sin(u[1])
sin3azm
s
1
integ
3
Gain
1
Rotor
Velocity
Kaimal_Filter_1 Admit0_Filter
Random
Random
Product1
Kaimal_Filter_3 Admit3_Filter_2
Random
Stochastic
Wind
Figure 3.18 Implementation of the stochastic model in Simulink/MATLAB.
In Figure 3.18, the rotor position is first computed by integrating the rotor speed of
the wind turbine. Kaimal filters applied to the random numbers simulate the turbulence
wind speed, which is inputted to the admittance functions representing the smoothing effect
on the rotor area to 0p and 3p. Finally, the three components are added resulting in the
stochastic part of the EWS. The complete EWS is the sum of the stochastic and
deterministic parts as presented in Equation (3.14).
3.2 Validation of the Equivalent Wind Speed Model
This section presents tests of the EWS, which is done in two parts: in the first part the
deterministic and stochastic models are verified separately and in the second part the EWS
is compared with measurements.
58/152
In the first part, the output of the EWS is compared to the ones from the Design
Bases Program 2 – DBP2. The DBP2 is an integrated model of a horizontal axis wind
turbine that the main aim is to analyse structural performance of wind turbines [61]. In this
chapter, the DBP2 generates time series of the wind speed on blade sections of the wind
turbine that are compared to the EWS.
The deterministic part of EWS is compared in terms of the normalised deterministic
wind influenced by the tower shadow. Two winds are simulated using the EWS, the first
one with the 0p and 3p harmonic components only and the second with all 3np harmonic
components. The wind speed simulated in DBP2 is seen from a single blade section located
at 2/3 of the total blade length. Therefore, the simulated wind speed in DBP2 was processed
to reflect the average wind speed seen from threebladed rotor. This was done by averaging
the measured wind speed at times t∆t, t and t+∆t, where ∆t is the time corresponding to
120° in the rotational speed of the rotor of the wind turbine (40 rpm). Figure 3.19 presents
the comparison and Table 3.2 presents the tower parameters from the wind turbine tested.
0.94
0.95
0.96
0.97
0.98
0.99
1.00
1.01
1.02
180.00 150.00 120.00 90.00 60.00 30.00 0.00 30.00 60.00 90.00 120.00 150.00 180.00
Azimuth Angle (deg.)
N
o
r
m
a
l
i
s
e
d
W
i
n
d
S
p
e
e
d
0p+3p Model All 3p multiples components DPB2
Figure 3.19 Normalised simulated deterministic wind component compared to a DBP2 model.
Figure 3.19 shows a good agreement between the “all 3np” simulation and the DBP2
calculation, whereas the agreement of the “0p+3p” simulation model is poor. This
illustrates that the equivalent deterministic wind speed can replace the deterministic wind
speed simulated with DBP2 but it requires that orders higher than 3p are included. Still,
only 0p and 3p are included in the model because the mechanical structure reduces the
corresponding simulated mechanical torque as discussed previously in this chapter.
Table 3.2 Parameters used in the simulation for tower shadow.
Physical geometric parameters Size
Rotor diameter 29 m
Distance from tower to rotor 1.15 m
Tower diameter 1.70 m
The stochastic part of the model was simulated for an average wind speed of 8m/s,
turbulence intensity of 0.10. In the EWS, the dynamics of the wind turbine was not
modelled, so the rotational speed of the rotor was assumed constant 40 rpm while in the
DBP2 the stochastic wind speed was simulated to a single blade section. The comparison of
the stochastic behaviour is shown in Figure 3.20 in terms of PSD’s.
59/152
10
−1
10
0
10
1
10
−6
10
−5
10
−4
10
−3
10
−2
10
−1
10
0
Frequency (Hz)
(
(
m
/
s
)
2
/
H
z
)
EWS
DBP2
Figure 3.20 PSD comparisons of the stochastic model.
As previously mentioned, only the contributions from zero and 3
rd
harmonics are
taken into account in the EWS and the frequency range of interest is from 0p to 4p. In this
frequency range, the EWS model is close to the DBP2 model with only minor deviations. In
the frequency range above 3Hz, the EWS differs from the DBP2 because the DBP2 shows
the turbulence acting on a single blade section of a wind turbine including all physical
aspects that are not presented in the electrical power measurements because they are filtered
out by the wind turbines dynamics.
In the second part of the validation, the EWS output was compared with measured
data. The wind speed was measured in a rotating blade with a five holes Pitot tube. The
Pitot tube was mounted at 15 m of the rotor centre at an angle of 14° to the local chord and
in a distance of about one chord length in front of the blade leading edge [48].
The wind measured at the hub height on a meteorological mast in front of the wind
turbine was used to compute the mean wind speed and turbulence intensity (input
parameters of the model). The turbulence intensity and the mean wind speed measured in a
fixed point on a meteorological mast were 0.16 and 10.51m/s respectively and the rotor
speed was 30 rpm. The main parameters of the wind turbine are shown in Table 3.3.
Table 3.3. Parameters of the wind turbine used in the measurement comparisons.
Physical geometric parameters Size
Rotor diameter 41 m
Distance from tower to rotor disc 2.9 m
Tower diameter 1.70 m
The measured wind speed is seen from a single point on a rotating blade, whereas the
reduced model describes the summed effect of the wind speed on all three blades.
Therefore, the measured wind speed was processed to reflect the average wind speed from
three blades before it was compared to the simulation. This was done by averaging the
measured wind speed at times t–∆t, t and t+∆t, where ∆t is the time corresponding to 120°
60/152
in 30rpm. The comparison of the measurement and simulation are shown in Figure 3.21
where the 1p frequency is approximately 0.5Hz.
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00
Frequency (Hz)
P
S
D
*
f
(
(
m
/
s
)
²
)
Srot*fmeasured Srot*f EWS
Figure 3.21 Measured and the simulated equivalent wind speeds on rotating blade section.
The comparison presents a reasonable agreement in the range from 0 to 3p. However,
two aspects are different for the measurement and simulation, which could explain most of
the difference.
First, the measurements are done in a single section of the blade, whereas the model
intends to include the wind averaged along the blades. Hence, the standard deviation of the
measured rotating wind speed is 1,7m/s, which is approximately equal to the standard
deviation measured in the fixed point in the mast. This complies with the theory for
rotational sampling described by Kristensen and Frandsen [59], where it is described how
the rotation moves standard deviation from lower frequencies to higher frequencies. The
standard deviation of the simulated equivalent wind speed is 1.08m/s, which is 40% lower
than the fixedpoint standard deviation as expected from the averaging process along the
[50].
The second difference between the measurement and simulation is, that the wind
speed is only measured on a single blade, so it was necessary to assume that the wind speed
seen by blade 2 is the same at time t+∆t as the wind speed seen by blade 1 at time t as
described above. In addition, the same assumption is made for the blade 3. it is important to
quote that the dynamic wind turbine was not modelled, so to the EWS (in this chapter only)
the rotor speed f the wind turbine was assumed constant what is also pointed as one more
reason to the difference between the model and measurements.
3.3 Equivalent Wind Speed Model Remarks
In this section, the EWS model was introduced. It has been reduced in complexity
although is based on complex theory.
The EWS consists of a deterministic part and a stochastic part. Both of them have
been compared with a verified model – DBP2. The EWS has also been compared to
measurements.
61/152
The deterministic part does not present a good approximation using only the zero and
the third harmonics. The higher components can be added to the model but measurements
of the power indicate that they are filtered by the dynamics of the wind turbine. The
deterministic part of the model was developed for tubular towers. In the case of a lattice
tower, an approximate tower effect can be obtained by specifying an equivalent fictitious
tubular tower.
The stochastic part was compared with a spectrum derived from the DBP2 model. It
showed good agreement in the frequency range from zero to 4p. Finally, there is no
significant difference considering the energy content of the DBP2 and the proposed model
for the stochastic part.
The comparisons with measurements presented a good agreement in the range from
0P to 4P. The small difference of amount of energy in the range from 1P to 4P is associated
to the single point measured compared with an equivalent wind simulated. The difference
also presented in the range above 4p was expected as a result from the initial assumptions
that the wind turbine acts as a low order pass filter.
The implementation of the model as a function of the rotor position enables the
program to be applied for variable or constant rotor speed wind turbine models without
additional modifications.
The fast computation and reduced memory usage are very suitable for simultaneous
simulation of a large number of wind turbines making it possible to efficiently estimate the
impact of a large wind farm on the power quality.
The increased speed and reduced memory usage could be compared in terms of wind
time series generated. As the conventional models generate several wind time series, the
presented model only generates one resulting in a small memory usage and a decrease in
the time computation.
62/152
Chapter 4
4 Wind Turbine Model
The produced electrical power from wind turbines does not have the same behaviour
in terms of variation as the wind. Wind turbines are dynamic generators with several
components that influence the power conversion from the wind. As mentioned before, the
dynamics of the wind turbine filter out the high frequency power variations but it also
includes new components due to its dynamics itself.
There are several types of wind turbines commercially available. Each one of them
has a special configuration and specific power dynamics. Here a dynamic wind turbine
model to a conventional wind turbine is presented. The conventional wind turbine has a
stallregulated rotor connected to an induction generator through a gearbox. The electrical
generator is directly connected to the network. A softstart is used to startup the wind
turbine. The reactive power demanded from the generator is partial compensated using a
capacitor bank. Finally, the integration of the electrical power into the grid is done at
medium voltage (e.g. 10kV), so a stepup transformer is used.
From the basic components of the conventional wind turbine, the relevant ones are
classified as:
• Aerodynamic rotor – converts the wind power into mechanical power;
• Transmission system – connects the aerodynamic rotor to the electrical
generator using shafts and gearbox;
• Generator – converts the mechanical to electrical power using an
asynchronous generator synchronized with the network;
• Controller – controls and optimise the wind turbine operation;
• Reactive power compensation – partial or total compensation of the reactive
power demanded by the generator;
• Stepup transformer – integrates the power produced in the medium voltage
electrical network;
• Network – medium voltage grid that transmit the power.
Each one of the components influences the dynamic operation of the wind turbines. In
addition, the elements of the wind turbine interact with each other, e.g. the aerodynamic
rotor output depends on the wind speed and on the rotor speed.
The Equivalent Wind Speed model (presented is chapter 03) is the input to the
dynamic wind turbine model. The dynamic wind turbine model is intended for power
quality assessment of wind turbines. This chapter presents a dynamic wind turbine model
for a threebladed stall regulated wind turbine with induction generator connected directly
to the network.
4.1 Simulation Tool
The dynamic wind turbine model can be implemented in any simulation program.
Here, the wind turbine model is implemented in the Power System Simulation software
63/152
SIMPOW from ABB because it has built in an extensive library of electrical components
and it can solve the network equations outputting the relevant characteristics of the
electrical power system. Therefore, the focus is on developing appropriate models of the
aeroelastic parts of the wind turbine. The dynamic aeroelastic model must simulate the
power conversion from the aerodynamic rotor and the transmission to the electrical
generator.
Once an appropriate induction machine model is chosen, the mechanical torque from
the wind turbine model is the input and the SIMPOW simulates all the electrical and
mechanical characteristics of the power system equipments modelled. SIMPOW has some
limitations in handling stochastic processes. Random numbers are not available in the
library. Hence, the wind speeds are time series, which are input to SIMPOW.
The total dynamic wind turbine model is composed of several modules: the first
module is the wind speed generation (described in chapter 3) that is implemented in
Matlab
®
, the Matlab module outputs time series of wind speed that is read by SIMPOW.
The second module is the aeroelastic model of the wind turbine that reads the wind speed
from a file and outputs the mechanical torque on the electrical machine shaft. The third
module comprises the load flow and dynamic simulation of all electrical equipments. The
second and third modules are implemented in SIMPOW. Finally, the last module is the data
processing that is implemented in Matlab, so SIMPOW exports time series of selected
variables
4.2 Wind Turbine Model
Each component of the wind turbine has some relations with other components.
Figure 4.1 presents the main relations among different components. The aerodynamic rotor
depends on the transmission system; the transmission system depends on the electrical
generator and aerodynamic rotor; and so on.
Aerodynamic
Rotor
Mechanical
Transmission
Electrical
Generator
Stepup
Transformer
Network
Reactive Power
Compensation
Controller
Wind Speed
T
aero ω
aero
ω
gen
T
gen
U
net.
U
term.
f
net.
f
net.
I
gen
I
comp
I
wtu
Wind
ω
aero
Figure 4.1 Interaction between each components of a wind turbine unity.
In Figure 4.1, ω is the rotational speed, T is the torque, U is the voltage, I is the
current and f is the frequency where the subscripts aero means to the aerodynamic rotor and
gen means electrical generator, comp means compensation unity, net means network and
wtu means wind turbine unity.
The wind turbine components are classified in electrical and aeroelastic parts. The
aeroelastic part comprises the aerodynamic rotor and the mechanical transmission system.
64/152
The electrical part comprises all the electrical components, i.e. the generator, reactive
power compensation, stepup transformer, and network.
4.2.1 Aeroelastic Components
4.2.1.1 Aerodynamic Rotor
The aerodynamic rotor converts the wind into mechanical power. Aerodynamic
effects throughout the blades convert the wind flow in aerodynamic torque. Each blade
section contributes to the total aerodynamic torque in a single blade. Integrating the torque
along the blade sections and adding up the effects of all 3 blades results in the aerodynamic
torque on the main shaft.
The aerodynamic model uses the Equivalent Wind Speed (EWS) as input to compute
the available power and it uses the speed of the rotor to compute the torque on the main
shaft.
The aerodynamic power on the main shaft can be expressed in per unit as:
( )
baserotor
pitch p
aerdynamic
P
C W R
P
θ λ π ρ ,
2
1
3 2
⋅ ⋅ ⋅ ⋅
=
(4.1)
where ρ is the air density; R is the rotor radius, W is the wind speed, C
p
is aerodynamic
power coefficient and P
baserotor
is the rated power of the rotor in order to give the torque in
per unit. The power coefficient is a normalized function of the pitch angle θ
pitch
and the tip
speed ratio λ, which is defined as the ratio between blade tip speed and the wind speed as
follows:
W
R
rotor
⋅
=
ω
λ (4.2)
where ω
rotor
is the rotational speed of the rotor. Using Equation (4.1), and the power
coefficient of a stall regulated wind turbine, the aerodynamic torque in per unit can be
expressed as:
( )
baserotor rotor
p
baserotor rotor
c aerodynami
c aerodynami
P
C W R
P
P
T
⋅ ⋅
⋅ ⋅ ⋅ ⋅
=
⋅
=
ω
λ π ρ
ω 2
3 2
(4.3)
where in this case the C
p
is a function only of the λ because stall regulated wind turbines
have a fixed pitch angle. This model fairly represents the power conversion in wind speeds
below the stall region [62]. However, during stall conditions the aerodynamics has been
reported to be quite different from static conditions because Equation (4.3) corresponds to
steady state aero loads. It underestimates the power fluctuations in the stall condition
because the stall effect does not happen instantaneously as the wind changes but after a
time lag [62].
A dynamic stall model is implemented in the aerodynamic module. The model is
based on time lags of separation [62]. The power conversion in an aerodynamic rotor can
be decomposed in three conditions:
65/152
• Static power conversion condition – is the steadstate power conversion curve
of an aerodynamic;
• Totally attached condition – is an extreme dynamic condition that considers
that the flow does not stall in any wind speed, hence the power converted is
proportional to the cubic of the wind speed;
• Totally separated condition – another extreme dynamic condition where in all
wind speeds the flow is totally separated of the wind turbine blades.
So, during stall condition, when the wind speed changes instantaneously, the power
conversion follow instantaneously the present degree of separation and after the time delay
it come back to the steady state curve. The power coefficient computation is done by
dedicated aerodynamic program. Figure 4.2 illustrates the three power coefficients to a
660kW stall regulated machine.
0 5 10 15 20 25
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Wind speed (m/s)
Static
Attached
Totally separated
Power Coeficient C
P
Figure 4.2 Power coefficients to compute the dynamic power coefficient.
The static (or steady state) power coefficient tends in low wind speed to follow the
totally attached curve and in high wind speed to follow the totally separated curve. In low
wind speeds, the blade profiles are designed to extract the power based on aerodynamic
coefficients that depend on the attached flow to generate lift. On high wind speeds, the
blade profiles are designed to start to stall (or separate) hence reducing the lift generated
and the extracted power from the wind, which is important to avoid overload of the wind
turbine.
The first step to compute the dynamic power coefficient, C
p
(λ) is to find the static
interpolation factor (f
static
) that represents the static curve as a combination of the others two
as follow
( ) ( ) ( ) ( ) λ λ λ
separeted attached static
p static p static p
C f C f C ⋅ − + ⋅ = 1
(4.4)
Then the dynamic interpolation factor (f) is computed as a time lag according to:
66/152
( )
ds
static
f f
t
f
τ
−
=
∂
∂
(4.5)
where the time constant τ
ds
=4/U
0
is computed from experimental tests as defined in [62].
U
0
is the average wind speed. Finally, the dynamic power coefficient is defined as:
( ) ( ) ( ) ( ) λ λ λ
separeted attached
p p p
C f C f C ⋅ − + ⋅ = 1
(4.6)
The dynamic power coefficient applied to Equation (4.3) gives the torque.
4.2.1.2 Drive Train
The aerodynamic torque, in a wind turbine, varies all the time due to wind turbulence
and tower shadow. The variations however are not directly transferred to the electrical
power output because of the dynamic mechanical transmission system.
The electrical generators run in a relatively high speed compared to the aerodynamic
rotor. The drive train connects high speed in the electrical generator side and slow speed in
the aerodynamic rotor using a gearbox. The drive train includes a gearbox; shafts and disc
brakes that can be positioned in the low or in the highspeed shaft depending on the size of
the wind turbine. On small wind turbines the brakes are positioned in the low speed shaft
that reduces the stresses on the gear box during shut down of wind turbines, on large wind
turbines, however, the torques in low speed shaft are very high hence the disc brakes are
located in the high speed shaft. Figure 4.3 illustrates the drive train of a small wind turbine
size.
Generator
Disc Brake
Gearbox
Lowspeed shaft Highspeed shaft
Figure 4.3 Example of drivetrain components.
In the left side of Figure 4.3, the aerodynamic rotor is presented that is connected to
the electrical generator in the right side through a gearbox. The gearbox is analogue to an
electrical transformer where the torques are like the currents and the rotational speed are
like the voltages. An ideal gearbox has no losses and does not modify the dynamics of the
system. The drive train modifies the dynamics of the system because they include torsional
modes. The shaft torsional modes are related to the aerodynamic rotor mass swinging with
the induction generator mass through the flexible transmission shaft.
Figure 4.4 presents a simplified dynamic drive train model where the gearbox is
considered ideal hence omitted here and the speeds and torques are in p.u.
67/152
ω
rotor
T
aerodyn
.
∆δ
T
electromechanical
ω
generator
d
K
H
rotor
H
generator
Stifness
Figure 4.4 Dynamic representation of the drive train model.
In Figure 4.4, T
aerodyn
is the aerodynamic torque computed from the aerodynamic
module, ω
rotor
is the rotational speed of the aerodynamic rotor, ω
generator
is the rotational
speed of the induction generator, T
electromechanical
is the electromechanical torque in the
induction generator, H
rotor
is the inertia of the aerodynamic rotor, H
generator
is the inertia of
the induction generator, K
stifness
is the equivalent stiffness of the shaft connecting the two
masses, d is the equivalent damping coefficient that includes the losses in the system and an
aerodynamic damping and Damping losses are the torque losses related to the damping
coefficient.
The equations of motion in to the dynamic drive train in Figure 4.4 can be expressed
according to:
( )
( ) ( )
rotor generator generator rotor aerodyn
rotor
rotor
d K T
dt
d
H ω ω δ δ
ω
− − − − =
.
2 (4.7)
( )
0
ω ω
δ
⋅ =
rotor
rotor
dt
d
(4.8)
( )
( ) ) ( 2
rotor generator hanical electromec generator rotor
generator
generator
d T K
dt
d
H ω ω δ δ
ω
− + − − = (4.9)
( )
0
ω ω
δ
⋅ =
generator
generator
dt
d
(4.10)
where the constants are explained as follow:
H
rotor
– is the total inertia of the aerodynamic rotor expressed in per unit. The rotor
inertia includes the blades and the hub in the low speed shaft. The inertia is then converted
to the high speed side of the shaft and finally converted to the per unit system using the
following equation (4.11):
( )
base gear
rotor
rotor
S q
J
H
⋅
⋅ =
2
2
0
2
1 ω
(4.11)
where J
rotor
is the rotor inertia in physical units on the low speed side of the gearbox, q
gear
is
the ratio between low and high speeds in the gearbox, ω
0
= 2πf
0
/poles is the mechanical
68/152
synchronous speed on the generator side and S
base
is the base power of the electrical
generator.
H
generator
– is the inertia of the electrical generator in per unit defined as:
( )
base
generator
generator
S
J
H
2
0
2
1
ω
⋅ = (4.12)
where J
generator
is the generator inertia in physical units.
K is the equivalent stiffness of the shaft connecting the two masses and can be
expressed as:
base
stiffness
T
k
K =
(4.13)
where T
base
is the mechanical base torque and k
stiffness
is the stiffness constant in physical
units, which can be computed using the first torsional eigenfrequency (ω
transmission
) as
follow:
( )
rotor on transmissi stiffness
J k ⋅ =
2
ω (4.14)
where ω
transmission
is the first torsional eigenfrequency. The transmission system eigen
frequency is measured in tests of the wind turbine. In the tests, the generator side has the
brakes on (i.e. fixed end) and the torque is measured.
d is the damping coefficient expressed as:
base
damping
T
c
d =
(4.15)
where c
damping
is damping coefficient computed in physical units, which is computed as
follow:
( ) ( )
2
log
2
log
2
2
δ π
δ
+
⋅
=
rotor stiffness
damping
J k
c
(4.16)
where δ
log
is the logarithmic decrement in terms of few percents, which is very difficult to
determine and in general it is measured from the tests of the wind turbines.
The main intention is to model the first torsional mode of the mechanical
transmission system but actually, the model includes the first and second modes due to the
use of two masses system (second order model). There are several other models to simulate
the drive train module (e.g. [10], [63] and [64]). The complexity of the model depends on
the aim of the model. Considering the power quality assessment of wind turbines, a two
masses model and a flexible shaft simulate the most relevant torsional moment from wind
turbines [63].
69/152
The first torsional mode of a medium sized wind turbine is around 1Hz and it is
considered the most relevant for power quality from wind turbines. The second mode is
related to the interaction between the rotor and generator masses oscillating against each
other due to the stiffness of the shaft, which is around 2 Hz.
4.2.2 Electrical Components
The electrical components comprise the electrical generator, the reactive power
compensation, control system, stepup transformer. In this thesis, they are modelled using
the standard library of SIMPOW/ABB, which has verified models of most of the
conventional electrical components in power systems including electrical machines,
transmission lines, transformers and power electronics and controllers.
Details of the electrical power system including the wind farm depend on the purpose
of the simulation. Here only some characteristics of the electrical generator are presented
and Annex 9.1 presents the some characteristics of the models used here from the
SIMPOW.
4.2.2.1 Electrical generator
The asynchronous generator is the most common type of electrical machine. It is
mostly applied as motor. As generator, however, it has been used for many years in wind
turbines. The reliability, low price, and low maintenance made this type of machine the
most suitable for wind turbines. Allied to those characteristics, the speed flexibility (slip),
when compared to the synchronous machines, reduces the current spikes due to wind gusts.
In the asynchronous generators, the rotational speed of the rotor is not fixed as it is in
the synchronous machines, but it is related to the torque from the prime mover and to the
network frequency.
The rotational speed of the asynchronous generator must be above the synchronous
speed in order to force the power flux to the network. Figure 4.5 presents an example of
power relation to the speed and voltage terminals in an asynchronous generator based on
static models [65].
Figure 4.5 Active power as function of speed and voltage terminals for an asynchronous generator.
In Figure 4.5, positive signal means power produced from the machine. The active
power is mainly related to the prime mover torque on the mechanical shaft and the voltage
70/152
influences the electrical characteristic of the induction machine, i.e. the excitation (reactive
power) modifies with the voltage. The reactive power of induction machines as generator
varies with the voltage. Figure 4.6 illustrates the reactive power from an induction
generator.
Figure 4.6 Reactive power as a function of the speed and voltage for an asynchronous generator (1p.u. = rated
reactive power at 1pu volts).
In Figure 4.6, negative power means the power system supplies power and 1p.u. of
reactive power means the rated reactive power. The asynchronous machine does not have
an independent excitation control system, hence the excitation is supplied from the power
system, which means reactive power consumption. The reactive power demanded of an
induction machine depends mainly on the voltage at its terminals and on the active power
supplied to the power system.
Keeping the active power constant and increasing the voltage lead to an increase in
the reactive power because the excitation is a shunt element proportional to the voltage.
Still keeping the active power constant and at this time reducing the voltage lead to a small
reduction on the reactive power consumption, however decreasing further the voltage leads
to an increase in the reactive power because of the losses that are increased due to the
higher currents flowing in the machine.
4.3 Verification of the Complete Wind Turbine Model
In order to verify the Dynamic Wind Turbine Model, a 500kW stall regulated wind
turbine directly connected to the grid is modelled and compared to measurements. The
wind turbine is installed at the test station facility of the RISØ laboratory. Table 4.1
presents a resume of the most relevant information to the wind turbine.
71/152
Table 4.1 Basic characteristics of the Nortank 500kW wind turbine modelled.
Nominal Power 500kW
Rotor diameter 35 m
Generator speed 1500 rpm
Rotor speed 27 rpm
Capacitor Bank 300kVAr
Generator terminal voltage 400V
Electrical frequency 50Hz
Tower height 36 m
Gearbox ratio 55.55
First torsional frequency 1.08Hz
Estimated logarithmic damping to drive train 5%
Number of blades 3
Constant of inertia of the aerodynamic rotor 1.15 s
Constant of inertia to the generator 0.15 s
Tower shadow 3p effect 1.5%
Figure 4.7 presents the static power coefficient to this wind turbine.
0
0,1
0,2
0,3
0,4
0,5
0,6
0 2 4 6 8 10
Tip Speed Ratio
C
p
12
Figure 4.7 C
p
(λ) static characteristic of the wind turbine Nortank 500kW.
To this verification, the wind speed measured in the hub high is used as input to the
dynamic wind turbine model and admittance filters (as introduced in chapter 2)
representing the coherence smoothing effect on the rotor area of the wind turbine are
applied. In addition, time series of wind speeds to the 3p effect are simulated based on the
average wind speed and turbulence intensity measured.
The average wind speed measured is approximately 8m/s and turbulence intensity of
13.3%. The low wind speed is far from the stall condition and the dynamic stall module is
not implemented in this case. Figure 4.8 presents the measured wind speed applied to the
dynamic wind turbine model.
72/152
0 100 200 300 400 500 600
4
5
6
7
8
9
10
11
Time (s)
W
i
n
d
s
p
e
e
d
(
m
/
s
)
Figure 4.8 Measured wind speed.
Figure 4.9 presents the time series of powers simulated and measured in the wind
turbines terminals during 600s. A general good agreement is seen between the simulated
and measured power from the wind turbine.
0 100 200 300 400 500 600
0
100
200
300
400
time (s)
A
c
t
i
v
e
p
o
w
e
r
(
k
W
)
Measured Active Power
0 100 200 300 400 500 600
0
100
200
300
400
time (s)
A
c
t
i
v
e
p
o
w
e
r
(
k
W
)
Simulated Active Power
Figure 4.9 Time series of simulated and measured power to the Nortank 500kW.
The mean wind speed was adjusted to simulate the same mean power in the period of
10 minutes. The measured wind speed directly applied to the model resulted in a mean
power difference of 75kW (–15% related to the rated power) when compared to the
measured power. One possible reason for the difference can be that the ambient wind speed
is measured 100 meters from the wind turbine because it cannot be measured on the rotor
73/152
plane. Hence, the wind speed acting on the rotor of the wind turbine is not necessarily the
same as the measured one. The adjustments in the wind speed are composed of modifying
the mean wind speed and to scale the turbulence to modify the standard deviation. Another
possible reason for the difference may be that the actual aerodynamic characteristics are
different from the modelled ideal characteristics.
With the mean wind speed adjusted (Figure 4.9), the mean and maximum powers and
the standard deviation are overestimated while the minimum power is underestimated. The
simulated mean power is 1.45% higher than the measured power the normalisation used in
this section is related to the rated power. The simulated instantaneous maximum power is
0.92% higher than the measured and the simulated standard deviation is 1.07% higher than
the measured. The simulated instantaneous minimum power is 3.19% lower than the
measurements. Those results indicate that the mean wind speed and the turbulence could
still require some adjustments, however, the differences are assumed reasonable here.
The power spectral distribution (PSD) of the active power is one of the relevant
characteristics from the wind turbines because of the stochastic nature of the process, hence
Figure 4.10 presents PSD of the measured and simulated power that corresponds to the time
series presented in Figure 4.9. In the upper part of Figure 4.10 the spectral power, i.e. S(f)⋅f,
has been chosen because it shows the power variance from each frequency that helps to
illustrate the significance of each frequency component. The middle part shows the
cumulative standard deviation of the measured and simulated active power computed from
the power spectra in the upper part. Finally, the lower part of Figure 4.10 shows the
normalised difference between the standard deviation simulated and measured, where the
normalisation refers to the rated power of the wind turbine.
(
k
W
2
)
Power spectra S(f)*f of active power
Simulation
Measurement
0 1 2 3 4 5 6 7 8 9 10
0
10
20
30
40
50
(
k
W
)
Simulation
Measurement
0 1 2 3 4 5 6 7 8 9 10
0.6
0.8
1
Frequency (Hz)
(
%
)
Normalized cumulative difference of standard deviation
0 1 2 3 4 5 6 7 8 9 10
0
100
200
300
Figure 4.10 Verification of the standard deviation of the dynamic wind turbine model.
From 0 to 6Hz there is a general good agreement between the simulated and
measured power spectra that is verified in terms of the cumulate standard deviation (middle
part of the figure). In the power spectra, at the frequency of approximately 0.8Hz, the
measurements have a power variation related to the first tower bending moment. This mode
74/152
was not included in the dynamic wind turbine model because it is small the power
contribution. The power on the frequency around 1.08 is related to the first torsional mode
from the drive train. The simulated drive train has more variations than the measured one.
In the frequency range below 3p (1.5Hz), the measured and the simulated active powers
agree.
In the frequency range from 6.7Hz up 9.5Hz, the power spectral distribution of the
measured power shows a power variation related to complex modes of oscillation that have
been related to the flexible complex aerodynamic rotor and the induction generator [66].
These effects have not been implemented in the dynamic wind turbine model because they
are small compared to the relevant power variations for power quality assessments as the
difference between the simulated and measured standard deviations accumulated is less
than 0.8% (normalisation to the rated power).
The last relevant parameter compared in this section is the flicker emission. The
shortterm flicker emission can be computed based on the time series of active and reactive
power [34]. The flicker shortterm emission (P
st
) from the measurements is 0.1834 and
from the simulation, P
st
is 0.1681, the total difference was small and because it is not linear,
it cannot be directly compared.
In order to detail the flicker estimation using the dynamic wind turbine model, the P
st
flicker is computed again where different power contributions in frequency domain are
included as presented in Figure 4.11. This is done by applying a sharp cutoff filter to the
measured and simulated active and reactive powers at different frequencies and then those
time series are applied to the flicker program that computes the P
st
.
0.05
0
0.05
0.1
0.15
0.2
0 1 2 3 4 5 6 7 8 9
Frequency (Hz)
P
s
t
10
Pst_measured Pst_simulated Difference
Figure 4.11. Verification of the flicker P
st
to different frequencies.
The simulated flicker is very similar to the measured one. The flicker as a function of
the power contributions bellow different frequencies shows that the 3p is the most relevant
effect on the flicker while in the frequency range from 5Hz to 10Hz the contribution to the
total flicker is reduced (less than 0.02 compared to 0.18). The dynamic wind turbine model
only includes the effects up to 4Hz that is the most relevant to power quality assessment.
75/152
4.4 Dynamic Wind Turbine Model Remarks
A dynamic wind turbine model has been presented, which is intended to power
quality assessment of wind turbines. This chapter focus on the aeroelastic model because
the electrical components are modelled by a dedicated conventional power system
simulation program.
The model was applied to a 500kW machine and the simulated electrical power was
compared to the power measured. The results showed a general good agreement. The power
spectral density, however, present some differences.
The main differences are related to basic assumptions done in the model. The first
difference is related to the tower bending moment that is not included in the aeroelastic
model therefore the differences were expected.
Finally, the model is suitable to power quality assessment. The model can be applied
to different simulation tools with small modifications. The model also is fast and permits
the simultaneous simulation of several wind turbines representing wind farms.
76/152
Chapter 5
5 Aggregate Wind Farm Model
Proper and successful integration of wind farms into the power systems rely on good
understanding of the power variations from wind turbines. Wind turbines deliver a
fluctuating power caused by turbulence and machine dynamics. Several wind turbines
compose a wind farm where each wind turbine contributes to the net power output
variations.
Power utilities usually are concerned about power quality deviations and stability
issues from wind farms operation. The power produced by wind turbines varies in a broad
range of frequencies and can cause undesirable voltage variations as well as frequency
variations in largescale integration.
Recently, the power quality assessment from wind turbines has been addressed in
IEC6140021 [1]. In IEC 6140021 the most relevant power quality characteristics from
wind turbines has been indicated. It gives also tools to estimate power quality deviations
due to the wind turbines connection as pointed in chapter 2.
The tools presented in [1] can estimate the power quality deviations, however,
simulations of the wind farms are still recommended to some specific conditions, e.g. when
large amount of wind energy is connected to the local capacity or when the networks are
weak. In those cases, the wind farms power production must be simulated and power
quality, power stability, and proper operation of the protections must be assured.
A wind farm model to support power quality assessment of wind farms based on
aggregate representation of the wind farm is presented. The Aggregate Wind Farm (AWF)
is a scale up of a wind turbine where the wind speed is an average of the wind acting on
each wind turbine in the wind farm. The aggregate wind farm represents the entire wind
farm.
The AWF reduces the simulation time and it has been implemented in a commercial
power system simulation tool SIMPOW/ABB
®
using available library models of the
electrical components as presented in chapter 4. The time inexpensive implementation of
the AWF makes the analysis of integration of largescale wind farms in the power system
more effective and feasible.
5.1 Aggregate Wind Speed Model
An appropriate wind speed model is essential to obtain realistic simulations of the
power fluctuations from wind farms [67]. The turbulence model was defined in chapter 3,
and here the main aspects of the wind in the park scale are presented.
The Aggregate Farm Wind Speed (AWFWS) model is an average of the wind speed
of the wind acting on each wind turbine in the wind farm. The AWFWS is based on a linear
system assumption. Considering the power from the wind farm is the sum of the power
produced from each wind turbine in the wind farm, the AWFWS is defined as the average
of the wind speeds in each wind turbine that applied to the scaled wind turbine simulates
the wind farm power produced.
77/152
Currently, the AWFWS represents the stochastic part of the longitudinal wind. The
longitudinal wind is the most relevant for power production in wind turbines, which is a
reasonable assumption. Here, it is also assumed that all wind turbines have the same
average wind speed and there is neither deterministic influences nor wake effect in the wind
farm.
5.2 Coherence of Turbulence in a Park Scale
The turbulence between two positions is correlated. The degree of correlation
depends on the distance between the two positions, the average wind speed and the
turbulence intensity very similar to the case on the rotor area of a wind turbine as presented
in chapter 3.
Similar to the coherence on the rotor plane in Equation (3.27), the Davenport type
coherence [57] also represents the coherence between two wind turbines however the decay
factors are modified to account larger distances. The Davenport type coherence between
two positions x and y is rewritten here as:
( )


.

\
 ⋅
−
=
0
,
U
f d
a
xy
xy
e
xy
d f γ
(5.1)
where a
xy
is the decay factor; d
xy
is the distance between the wind turbine in position x and
wind turbine in position y as presented in Figure 5.1, f is the frequency of the turbulence
and U
0
is the mean wind speed. In this case, the prevailing wind direction influences the
coherence factor, where the inflow wind speed angle for each wind turbine (α
xy
) is defined
as in Figure 5.1.
y
x
α
xy d
xy
U
0
Figure 5.1 Spatial disposition of the two wind turbines (α
xy
= 90° means lateral disposition).
where d
xy
is the distance between the wind turbines in positions x and y, α
xy
is the inflow
wind speed angle to wind turbine in x that it is the relative angle of the wind speed to the
wind turbine in the position y and U
0
is the mean wind speed.
The decays factors a
xy
for the park scale includes the distances up to kilometres and
have been experimentally studied in [58], which suggests the following expression:
( ) ( ) ( ) ( )
2 2
sin cos
xy lat xy long xy
a a a α α + = (5.2)
where α
xy
is the wind inflow angle (see Figure 5.1) and a
lat
is the lateral component:
( )
s m
a
lat
/ 1
5 5 . 17
σ
⋅ ± = (5.3)
78/152
where σ is the standard deviation of the wind speed in m/s. Moreover, a
long
is the
longitudinal component:
( )
0
5 15
U
a
long
σ
⋅ ± =
(5.4)
Figure 5.2 presents an example of coherence simulated with different separations
from 100m up to 1300m all with average wind speed of 15m/s and turbulence intensity of
15%, inflow angle = 0°, a
lat
= 17.5 and a
long
= 15.
10
−4
10
−3
10
−2
10
−1
10
−20
10
−18
10
−16
10
−14
10
−12
10
−10
10
−8
10
−6
10
−4
10
−2
10
0
C
o
h
e
r
e
n
c
e
γ
Frequency (Hz)
100 m
500 m
900
1300 m
Figure 5.2 Coherence factor for different distances between two points.
The separation and the frequency have strong influences on the coherence. The
distances in a wind farm are in the range from hundred meters up to kilometres, in which
the turbulence will be highly correlated at very low frequencies up to 0.001Hz. At the 3p
frequency, which is around 1Hz in medium sized wind turbines, the turbulence can be
considered uncorrelated because of the strong reduction from the coherence. These facts are
important to the implemented AWFWS as presented in the following section.
5.3 Aggregate Turbulence
The wind field acting on the rotor of a wind turbine can be expressed as an equivalent
time series of wind speed as explained in chapter 3.
The equivalent wind speed (w
eq,i
(t)) for the wind turbine i can be expressed as sum of
harmonics on the rotor azimuth angle as follow:
∑
∞
−∞ =
=
k
k j
i eq k i eq
i rotor
e t w t w
,
3
, , 3 ,
) (
~
) (
φ
(5.5)
79/152
where the
i eq k
w
, , 3
~
is the 3k
th
term of the expansion of the wind, k is the multiple for the third
harmonic component in terms of azimuth angle φ
rotor
for wind turbine i.
Power spectral analysis of power measurements from threebladed wind turbines
revealed that the harmonic components at three times the azimuth speed are the most
significant that are converted to the electrical power (chapter 3). Equation (5.5) is then
rewritten as:
{ } { } ) 3 sin( ) ( Im 2 ) 3 cos( ) ( Re 2 ) ( ) (
3 3 0 rotor p rotor p p eq
t w t w t w t w φ φ + + ≈
(5.6)
where, the PSD of the w
np
’s are computed as admittance functions applied to the PSD of
turbulence (e.g. Kaimal [44]) as pointed in chapter 3.
Using the Aggregate Wind Farm Wind Speed (AWFWS), its harmonic components can
be rewritten as:


.

\

=
∑
=
. .
1
,
. .
) (
1
) (
T W
N
i
i eq
T W
AWFWS
t w
N
t w
(5.7)
Inserting Equation (5.6) in Equation (5.7):
( ) { } { }
∑ ∑ ∑
= = =
⋅ + ⋅ +
=
. . . . . .
1
, , 3
. . 1
, , 3
. . 1
, 0
. .
) 3 sin( ) ( Im
2
) 3 cos( ) ( Re
2
) (
1
) (
T W T W T W
N
i
i rotor i p
T W
N
i
i rotor i p
T W
N
i
i p
T W
AWFWS
t w
N
t w
N
t w
N
t w
φ φ
(5.8)
where N
W.T.
is the number of wind turbines in the wind farm. Rewriting (5.8):
{ } { } ) 3 sin( ) ( Im 2 ) 3 cos( ) ( Re 2 ) (
) (
, , 3 , , 3 , 0 AWF rotor AWFWS p AWF rotor AWFWS p AWFWS p
AWFWS
t w t w t w
t w
φ φ ⋅ + ⋅ +
≈
(5.9)
Here, it is assumed that the azimuth position of the aggregate wind farm φ
rotor,AWF
can
replace the azimuth position of each wind turbine. Since the winds at 3p frequency are
totally uncorrelated because of the higher frequency and distances between the wind
turbines, the assumption is valid. w
0k,AWFWS
and the w
3k,AWFWS
are the harmonic amplitudes
of the AWFWS, defined as:
( )


.

\

=
∑
=
. .
1
, 0
. .
, 0
) (
1
) (
T W
N
i
i p
T W
AWFWS p
t w
N
t w (5.10)
( )


.

\

=
∑
=
. .
1
, 3
. .
, 3
) (
1
) (
T W
N
i
i p
T W
AWFWS p
t w
N
t w
(5.11)
5.4 Simulation of Wind Speeds
The wind speed simulator has six main inputs parameters:
80/152
• Wind speed parameters – e.g. average wind speed, wind direction turbulence
length scale, turbulence intensity, coherence factors, etc.
• Wind turbine diameter.
• Wind farm layout.
• Turbulence Power Spectral Density model, e.g. Kaimal [44].
• Coherence model – e.g. Davenport [57].
• Admittance functions – to 0p and 3p harmonic components.
The model outputs three time series that compose the Aggregate Wind Farm Wind
Speed – AWFWS – as defined in Equation(5.9). Figure 5.3 presents the main structure of
the AWFWS.
Aggregate Wind
Speed Time series
generator
3p time series 0p time series 3p time series
Wind Speed
Parameters
Wind Turbine
Rotor Diameter
Wind Farm
Layout
PSD turbulence
model
Coherence
model
Admittance
Functions
Figure 5.3 Structure of the AWFWS generator.
The aggregate wind speed times series generator in Figure 5.3 is based on the
Shinozuka method using a crossspectral matrix similar to the method used in [68]. The
Shinozuka method purpose is to compute a realization of a stochastic process given the
spectral density function of the process.
Here a crossspectral matrix represents the correlated turbulence processes. The
crossspectral matrix contains all energy from the turbulence on each wind turbine position
in the wind farm and the contributions due to the correlation between two positions. The
crossspectral matrix S(f) is NxN, which corresponds to a wind farm with N wind turbines.
The S(f) has the following structure
( ) ( )






.

\

=
−
− − − −
−
) ( ) ( ) ( ) (
) ( ) ( ) (
) ( ) ( ) (
) ( ) ( ) ( ) (
, 1 , 2 , 1 ,
, 1 1 , 1 1 , 1
. 2 2 , 2 1 , 2
, 1 1 , 1 2 , 1 1 , 1
f S f S f S f S
f S f S f S
f S f S f S
f S f S f S f S
f S
n n n n n n
n n n n n
n
n n
L
M
M O M M
L K
K
(5.12)
81/152
where the diagonal elements are the spectral density of turbulence on each wind turbine.
The nondiagonal elements are the crossspectral densities of the turbulence for the pairs
x,y. The cross power spectral density includes a time delay and it is written as [19]:
( ) ( )
xy
f
xy xy
e f S f S
τ π 2 −
=
(5.13)
where , S
xy
(f) is the cross spectrum between the turbulences in x and y, f is the frequency, γ
xy
is the coherence function between x and y defined in Equation (5.1) and τ
xy
is the travelling
time of the wind travel from wind turbine x to y, which can be determined according to
Figure 5.1 as
0
) cos(
U
d
xy xy
xy
α
τ =
(5.14)
where U
0
is the mean wind speed. Thus, the crossspectral density can be expressed as
( )
xy
f j
yy xx xy xy
e f S f S d f f S
τ π
γ
2
) ( ) ( , ) (
−
⋅ =
(5.15)
Hence, using the spectral matrix and taking the definition of the AWSWF from
Equation (5.10) and Equation (5.11) the spectral density terms to the Aggregate Wind Farm
Wind Speed (S
AWFWS,k
(f)) for each harmonic component is written as:
∑ ∑
= =
=
. . . .
1 1
0 , 2
. .
0 ,
) (
1
) (
T W T W
N
x
N
y
p xy
T W
p AWFWF
f S
N
f S
(5.16)
∑
=
=
. .
1
3 ,
2
. .
3 ,
) (
1
) (
T W
N
x
p xx
T W
p AWFWF
f S
N
f S
(5.17)
Where S
xy,k
are the cross spectral elements of the k
th
harmonic component and S
xx,k
is
the spectral density of each k
th
harmonic component for each wind turbine. The main
difference between the terms to the 0p and 3p is that for the 3
rd
harmonic the cross spectral
elements S
xy,k
are assumed to be zero due to the high frequency and relative large distance
between wind turbines x and y (coherence is zero).
After having defined the spectral terms of the aggregate wind speed, it is necessary to
discretise the frequency in order to use a computer code to generate the wind speeds. The
discrete step of the frequency to simulate a time series with length T
p
is ∆f=1/T
p
, so the i
th
frequency is f[i]=i∆f. The corresponding discrete value of S
xy
(f) is S
xy
[i]=S
xy
(∆f i)⋅∆f.
The corresponding time series is discrete by the sampled representation of the wind
speed with time steps of ∆t=1/f
s
where f
s
is the sampling frequency in Hz. The sampling
frequency limits the frequency to ±f
s
/2 and consequently the frequency index i to ±Ns/2,
where Ns=T
p
×f
s
is the number of samples in the simulated time series. Ns must be integer
and in order to use FFT it should be an exponent of 2.
The next step is for each frequency index k to generate random complex numbers
c
k
=A
k
+jB
k
to finally use the inverse Fourier Transformation with Fourier coefficients c
k
, i.e.
82/152
∑ ∑
= =
∆ ⋅ ⋅ − ∆ ⋅ ⋅ + =
N
k
k
N
k
k
f k B f k A A t w
1 1
0
) sin( 2 ) cos( 2 ) (
(5.18)
Then the energy content to the k
th
component from Equation (5.18) can be rewritten
as:
( )
2 2
2 ) (
k k k
B A w E + = (5.19)
It is imperative that this method preserves the variance σ² of the process. Thus, using
a single sided spectrum, the variance can be defined as:
∑
∫ ∫
=
∞
∆ ⋅ ∆ ⋅ ≈ ≈ =
2 /
0
0
2 /
0
2
) ( ) ( ) (
s
s N
K
xx
f
xx xx
f f k S df f S df f S σ
(5.20)
The discrete part of the energy content of the frequency k⋅∆f must then have the same
mean value of the energy content E
k
,
{ } f f k S E E
xx k
∆ ⋅ ∆ ⋅ = ) ( (5.21)
Assuming E{A
k
2
}= E{B
k
2
} then,
{ } { }
( )
4
2 2
f f k S
B E A E
xx
k k
∆ ∆ ⋅
= =
(5.22)
Assuming that the mean value of A
k
and B
k
are zero, i.e. E{A
k
}= E{B
k
}=0, the
standard deviation σ{A
k
} and σ{B
k
} can be determined as
{ } { } { } { } ( )
( )
2
2 2
f f k S
A E A E B A
xx
k k k k
∆ ∆ ⋅
= − = = σ σ (5.23)
Thus, A
k
and B
k
can be determined from a normally distributed stochastic process
with mean value zero and standard deviation:
( )
2
f f k S
xx
∆ ∆ ⋅
. Figure 5.4 presents a
comparison between the formal Shinozuka random phase angle method and the use of the
two random numbers
83/152
−3 −2 −1 0 1 2 3
−3
−2
−1
0
1
2
3
Distribution of A
k
D
is
t
r
ib
u
t
io
n
o
f
B
k
AWFWS method
Shinozuka´s method
Figure 5.4 Distribution of random constants.
From Figure 5.4 it is possible to see that the terms to the Inverse Fourier
Transformation are more random than the random phase angle method.
5.5 Aggregate Wind Turbine Machine
The aggregate wind turbine replaces the entire wind farm with a single wind turbine.
The single wind turbine is modelled as presented in chapter 4.
The aggregation procedures have been extensively applied to power system analysis
particularly to stability analysis of large power systems. The method has been applied to
represent groups of similar machines in areas of the power system.
The aggregation procedures of induction motors applied to power quality analysis
have been successfully reported, e.g. [69], [70] and [71].
Here, the aggregation procedure is applied to similar wind turbines, hence the
equivalent wind turbine to the entire wind farm is a scale up of a single wind turbine, i.e.
the base power becomes N
wt
times the base power of a single wind turbine in the farm,
where N
wr
is the number of turbines in the wind farm.
5.6 Results and Discussions
Here, the simulated power using the AWF is compared with a simulation that models
individually each wind turbine in a wind farm, hereafter called IWF. The IWF simulates all
wind turbines in the farm applying to each of them a correlated time series of wind.
The IWT models each wind turbine in the wind farm with accurate models to the
wind turbine dynamics. The Parksimu model [19] generates correlated time series of wind
speed to each wind turbine in the wind farm.
The AWF model uses a single aggregate wind turbine that replaces all wind turbines
in the wind farm.
5.6.1 Case Description
The wind farm is composed of six 660kW stall regulated directly connected wind
turbines type. The wind turbine is composed of an aerodynamic module, a mechanical
84/152
transmission system, an asynchronous generator, a capacitor bank, and a stepup
transformer.
The aerodynamic module uses power coefficients as introduced in chapter 4 and
dynamic stall effects are included. The drive train of the wind turbine has the first torsional
moment in 1Hz and inertias of 3.05 and 0.2 per units (seconds) of the aerodynamic rotor
and induction generator respectively. The stiffness and damping constants are computed
using the definitions from chapter 4. The logarithmic decrement damping coefficient is
assumed to be 5%. Table 5.1 presents a resume of the most relevant information to the
single wind turbine.
Table 5.1 Basic characteristics of the 660kW wind turbine modelled.
Nominal Power 660kW
Rotor diameter 48 m
Generator speed 1500 rpm
Rotor speed 21 rpm
Capacitor Bank 200kVAr
Generator terminal voltage 400V
Electrical frequency 50Hz
Tower height 36 m
Gearbox ratio 71.43
First torsional frequency 0.8Hz
Estimated logarithmic damping to drive train 5%
Number of blades 3
Constant of inertia of the aerodynamic rotor 3.05 s
Constant of inertia to the generator 0.2 s
The electrical generator is a squirrel cage induction machine, modelled with a 3
rd
order model where current displacement module is not active (i.e. the rotor resistance is
constant).
0.00
0.20
0.40
0.60
0.80
1.00
1.20
0.00 5.00 10.00 15.00 20.00 25.00 30.00
Wind speed (m/s)
P
o
w
e
r
c
u
r
v
e
(
p
.
u
.
)
Figure 5.5. Static power curve of the wind turbine.
Figure 5.6 presents the wind farm layout. Where the lateral space between wind
turbines is 4 times the rotor diameter and the longitudinal space is 10 times. The wind
direction is measured relatively to the North as indicated.
85/152
4 x D
D
4 x D
10 x D
4 x D
4 x D
10 x D
N
θ
ws
WS
Figure 5.6 Wind farm layout.
The power system is very strong at the Point of Common Coupling (PCC), 400 times
the rated power of the wind farm. The Thevenin equivalent at the PCC has an angle of
approximate 50°. The PCC is the same to both simulations: AWF and IWF.
5.6.2 Results
5.6.2.1 Wind Speed Simulator
The wind speed simulator is verified here compared to Parksimu [19]. The Parksimu
is a program to simulate correlated time series of wind speeds of each wind turbine in a
given wind farm based on the layout of the wind farm, the mean wind speed and turbulence
intensity as well as on the spatial coherence. The admittance filters are disabled and only
one wind speed is simulated in order to compare to the wind speed for a single point as in
the Parksimu.
Figure 5.7 presents a general comparison between the wind speed simulated with the
Aggregate Wind Farm Wind Speed (AWFWS) and the Parksimu. In the upper part, time
series of wind speed simulated with both models are presented. The middle part presents
the power spectral of the simulated wind speed using the AWFWS simulator and the one
produced from Parksimu described in [19]. In addition, the Kaimal turbulence model is
plotted. Kaimal is the target power spectral of turbulence used in both simulation programs.
In the lower part of the figure, the cumulative normalised differences of standard deviation
of the AWFWS and Parksimu to the Kaimal target spectrum are presented. The same
characteristics are applied for both simulations, i.e. average wind speed of 10 m/s at 30
meters height, turbulence intensity of 10%, and turbulence length scale of 600m.
86/152
0 100 200 300 400 500 600
5
10
15
W
i
n
d
s
p
e
e
d
(
m
/
s
)
Time(s)
Time series of wind speed
AWFWS
Parsimu
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
10
−3
10
−2
10
−1
10
0
P
o
w
e
r
s
p
e
c
r
t
a
l
(
m
/
s
)
2
AWFWS
Parksimu
Kaimal
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
−0.4
−0.2
0
Frequency (Hz)
(
%
)
Cumulative difference of stanrdard deviation to Kaimal target
AWFWS
Parksimu
Figure 5.7 Comparisons of the wind speed simulator.
The upper part of Figure 5.7 shows two wind speeds uncorrelated with similar range
of wind speed variations. The power spectra reveal general agreement of both wind speed
simulations with the target power spectra (Kaimal), in addition, the AWFWS presents more
variations than the Parksimu around the target power spectrum. The high variation around
the target spectral is a consequence of using two random numbers instead of a random
phase angle as used on the Parksimu.
In the lower part of Figure 5.7, the normalised cumulative differences of standard
deviation show small difference with a total difference less than 0.2%. In this case, the
normalisation factor is the mean wind speed (10m/s). Despite the high variation of the
power spectrum of the AWSWF, both simulated wind speeds have similar energy content
and slightly difference to the target standard deviation. From Figure 5.7, the AWFWS
appears to represent better the turbulence, the difference relative to the target power
spectral distribution (Kaimal) is mostly related to the random seeds, and in this particular
case the random seed chose to the AWSWF represented better the turbulence than the
Parksimu, anyway the total accumulated difference in terms of standard deviation is small.
5.6.2.2 Wind Farm Power Production
This section presents comparisons of the wind farm power simulated using the AWF
to the IWT. Six cases are discussed: three different average wind speeds (10,13 and 16m/s)
and two different turbulence intensities (10 and 20%). Those relative high wind speeds
have been selected because the power curve at those wind speeds is nonlinear.
The average power, maximum power in 0.2 seconds and flicker emission in short
term were the parameters selected to compare the different models, in accordance to the
IEC 6140021 [1], in addition the standard deviation and minimum power are presented in
order to give more information on the AWF usage. Table 5.2 presents the main power
characteristics and flicker emission in short term (P
st
) simulated using the AWF and the
IWF.
87/152
Table 5.2 Representative power characteristic values of all simulations
Mean Wind speed 10m/s 13m/s 16m/s
Turbulence intensity
(%)
10 20 10 20 10 20
Mean Value 2419.73 2403.85 3501.37 3411.19 3870.91 3769.75
Max Value 2804.95 3158.64 3791.55 3970.79 4014.85 4107.99
Min Value 1921.04 1602.02 2992.91 2395.15 3589.11 2786.66
I
W
F
(
k
W
)
Standard
Deviation
131.36 248.62 114.66 244.47 54.65 164.07
Mean Value 2430.11 2454.74 3533.61 3519.13 3903.02 3868.42
Max Value 2892.29 3477.01 3914.87 4152.74 4061.64 4256.20
Min Value 1830.25 1492.88 3153.48 2615.11 3713.48 3358.08
A
W
F
(
k
W
)
Standard
Deviation
162.93 370.92 132.61 255.30 41.251 96.01
IWT 0.094 0.185 0.071 0.164 0.0474 0.120
F
l
i
c
k
e
r
P
s
t
AWF 0.086 0.181 0.064 0.140 0.0468 0.100
In low turbulence intensity, the power characteristics simulated with the AWF agree
well with the ones simulated with the IWF. Figure 5.8 illustrates the power characteristics
simulated on low turbulence intensity.
Turbulence intensity 10%
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
8 9 10 11 12 13 14 15 16 17 18
Wind speed (m/s)
P
o
w
e
r
(
M
W
)
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
S
t
a
n
d
a
r
d
D
e
v
i
a
t
i
o
n
(
M
W
)
.
Individual mean Aggregation mean
Individual max Aggregation max
Individual min Aggregation min
StD Individual StD Aggregate
Figure 5.8 Power characteristics evolution with the wind speed (10% turbulence intensity).
In general, the power simulated with the AWF is slightly higher than the IWF. A
possible reason for the power differences may be that the wind speed simulated with the
AWF has slightly high standard deviation then the ones simulated with the Parksimu as
introduced in Figure 5.7. Both simulations show that the power variations reduce with the
increased wind speed because of the nonlinearity of the aerodynamic power conversion
(Figure 5.5).
In high turbulence intensity, the differences are large than the ones in the low
turbulence but it is still presenting good agreement except to minimal power. Figure 5.9
illustrates the power characteristics from both simulations on high turbulence intensity.
88/152
Turbulence intensity 20%
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
8 9 10 11 12 13 14 15 16 17 18
Wind speed (m/s)
P
o
w
e
r
(
M
W
)
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90
S
t
a
n
d
a
r
d
D
e
v
i
a
t
i
o
n
(
M
W
)
.
Individual mean Aggregation mean
Individual max Aggregation max
Individual min Aggregation min
StD Individual StD Aggregate
Figure 5.9 Power characteristics evolution with the wind speed (20% turbulence intensity).
Similar to the characteristics presented in low turbulence intensity, the AWF in
general simulates higher power than the IWF and the power variations reduces with high
wind speed. The differences are related to the same reasons as in the low turbulence
intensity case. However, in this case, the high turbulence intensity in high wind speed
strongly influences the simulated minimum power and the standard deviation in the AWF.
The high turbulence intensity means that the wind speed will vary in a broad wind
speed range on top of the mean wind speed. In the high mean wind speed (16m/s), the
power conversion on the aerodynamic rotor is extremely nonlinear (see Figure 5.10).
Hence, the power variation in a single wind turbine in the IWF can be very different from
the one simulated with the AWF because the AWF is an average wind speed that smoothes
the wind variations leading to less power variation and a concentration around the high
mean value.
0.00
0.20
0.40
0.60
0.80
1.00
1.20
0.00 5.00 10.00 15.00 20.00 25.00 30.00
Wind speed (m/s)
A
g
g
r
e
g
a
t
e
P
o
w
e
r
(
p
.
u
.
)
Power curve
10m/s 10% ti
16m/ 20%ti
Figure 5.10. Non linear effects on the power variations.
89/152
Following, samples of time series of active power simulated are presented that
illustrates the processes described above. Figure 5.11 presents the power produced by IWF
and AWF with 13 m/s average wind speed and 10% turbulence intensity.
0 100 200 300 400 500 600
2800
3000
3200
3400
3600
3800
4000
P
(
k
W
)
Time (s)
Ws= 13m/s and Turbulence = 10%
Individual Wind Farm
0 100 200 300 400 500 600
2800
3000
3200
3400
3600
3800
4000
P
(
k
W
)
Time (s)
Aggregate Wind Farm
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
0
1000
2000
3000
4000
Figure 5.11 Power simulated 13 m/s and 10% turbulence intensity.
The simulated active powers presented in Figure 5.11 are uncorrelated and have
similar ranges of power variations. At 13m/s and 10% turbulence intensity, the
aerodynamic power conversion is reasonable linear. Power spectral analyses reveal small
differences in frequency that are presented in Figure 5.12.
P
o
w
e
r
s
p
e
c
t
r
a
(
k
W
2
)
Power spectra S(f)*f of active power
Individual Wind Farm
Aggregate Wind Farm
0 1 2 3 4 5
0
50
100
150
(
k
W
)
Cumulative standard deviation
Individual Wind Farm
Aggregate Wind Farm
0 1 2 3 4 5
−0.8
−0.6
−0.4
−0.2
0
Frequency (Hz)
(
%
)
Normalized cumulative difference of standard deviation
Figure 5.12 Power spectral comparisons at 13 m/s and turbulence intensity 10%.
90/152
At 13m/s, the power spectrum of the AWF presents a good agreement with the IWF
as showed in the upper part of Figure 5.12. The cumulative difference of the standard
deviation from both simulations agree well as presented in the middle part of Figure 5.12
with a steady difference that does not significantly changes with the frequency, that means
that the main difference of the standard deviation (power variations) are concentrated on the
low frequency range. The lower part in Figure 5.12 presents the normalised difference of
the standard deviation between the powers simulated with the AWF and with the IWF,
where the negative signal means that the AWF overestimates the standard deviation and the
normalisation factor is the rated power of the wind farm.
As one can notice, the overall difference is small (less than 0.2%) and the main
difference is on the lower frequency range (less than 1Hz). At frequencies above 1Hz, the
power variations from the IWF are slightly higher than the power variations from the AWF
that reduces the cumulative standard deviation difference. The reduction on the standard
deviation difference is very small.
When the wind speed and the turbulence intensity are increased, the differences
become more noticeable. Figure 5.13 presents the active power simulated at 16m/s and
turbulence intensity 20%.
0 100 200 300 400 500 600
2500
3000
3500
4000
4500
P
(
k
W
)
Time (s)
Individual Wind Farm
0 100 200 300 400 500 600
2500
3000
3500
4000
4500
P
(
k
W
)
Time (s)
Aggregate Wind Farm
Figure 5.13 Power simulated at 16 m/s and turbulence intensity 20%.
In Figure 5.13, the IWF simulates a steady power reduction in 120 seconds while the
AWF does not. This is related to the averaging wind speed process of the AWF added to
the nonlinear rotor as explained before. The power spectra of the powers reveal more
information as presented in Figure 5.14.
91/152
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
0
2
4
6
x 10
4
P
o
w
e
r
s
p
e
c
t
r
a
(
k
W
2
)
Power spectra S(f)*f of active power
Individual Wind Farm
Aggregate Wind Farm
0 1 2 3 4 5
0
50
100
150
200
(
k
W
)
Cumulative standard deviation
Individual Wind Farm
Aggregate Wind Farm
0 1 2 3 4 5
0
1
2
3
Frequency (Hz)
(
%
)
Normalized cumulative difference of standard deviation
Figure 5.14 Power spectral comparisons at 16 m/s and turbulence intensity 20%.
The power produced from the AWF is smother than the power produced by the IWF
that is related to the high nonlinear power curve of the wind turbine. The standard
deviation of the AWF is very different of the one in the IWF. The differences are very
noticeable in the lower frequency range, i.e. less than 1Hz. In the lower frequency range,
the wind speed variations are significant, which in this case acts on the high nonlinear area
of the power curve leading to greater differences on the standard deviation and minimum
power.
The difference of standard deviation actually decreases with the frequency. At the
frequency range above 3p, the differences on the standard deviation change very little
because at this frequency range the wind speed variations are relatively small hence the
dynamic stall provides a better power relation between the active power variation and the
wind speed variation.
The power variations on low frequency range presented the most significant
differences between the AWF and IWF. In the higher frequency rage, the power differences
were relatively small that support the use of the aggregate model to flicker estimation from
large wind farms. Following, the flicker estimated in the AWF is compared to the one in the
IWF.
Using the shortterm flicker presented in Table 5.2, the rated power of the wind farm
and the number of wind turbines in the wind farm, the flicker coefficient (c) to the wind
turbine in the wind farm is computed. The flicker coefficient from a wind turbine is a
normalised measure of the flicker from a wind turbine as introduced in chapter 2 that can be
computed using Equation (2.3) [1]. Following, Figure 5.15 presents the flicker coefficients
for the wind turbine using the P
st
simulated with the AWF and IWF.
92/152
0.00
0.50
1.00
1.50
2.00
2.50
10 13 16
Wind speed
(m/s)
F
l
i
c
k
e
r
c
o
e
f
f
i
c
i
e
n
t
(
c
)
Individual 10% TI Aggregate 10% TI
Individual 20% TI Aggregate 20% TI
Turbulence
Intensity(%)
10% 20% 10% 20% 10% 20%
10 13 16
Figure 5.15 Flicker coefficients comparisons (computed according to [34]).
The flicker coefficients computed using the AWF are quite similar to those simulated
with the IWF, particularly on the low turbulence intensity. In high turbulence intensity, the
nonlinear behaviour of the power conversion leads to greater differences between the
AWF and the IWF but still god agreement.
In order to verify the contributions from the power variations in frequency domain to
the flicker coefficient, following, the flicker estimation is detailed to two different wind
speed conditions (13m/s, 20% turbulence intensity and 16m/s 10% turbulence intensity).
The flicker estimated at 13m/s and 20% turbulence intensity with the AWF presented
the higher difference to the one simulated with the IWF. Figure 5.16 presents the evolution
of the flicker coefficient computed with the P
st
that included the contributions of the power
variations with different frequencies. The active and reactive powers simulated in the AWF
and IWF were filtered out with different lowpass cutoff frequencies and then applied to
the flicker code that simulates the shortterm flicker emission.
Flicker coefficient Wind speed 13 m/s
Turbulence intensity 20%
0
0.5
1
1.5
2
2.5
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Frequency (Hz)
F
l
i
c
k
e
r
c
o
e
f
f
i
c
i
e
n
t
c
Ck_IWF Ck_AWF Difference
Figure 5.16 Evolution of flicker coefficients at 13m/s 20% turbulence intensity.
The flicker estimated in the AWF poorly agrees with the flicker coefficient estimated
in the IWF at 13m/s and 20% turbulence intensity. The main difference appears to be
93/152
related to the 3p area, but the flicker procedure is not linear and the curve in Figure 5.16
must be used with caution.
The flicker estimated at 16m/s and 10% turbulence intensity presented the best
agreement. The evolution of the flicker coefficient presents very small differences as
showed in Figure 5.17.
Flicker coefficient Wind speed 16 m/s
Turbulence intensity 10%
0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Frequency (Hz)
F
l
i
c
k
e
r
c
o
e
f
f
i
c
i
e
n
t
c
Ck_IWF Ck_AWF Difference
Figure 5.17 Evolution of flicker coefficients at 16m/s 10% turbulence intensity.
The relevant differences encountered here are related to the nonlinear rotor under
stall conditions and high turbulence intensity. The high turbulence intensity leads to large
wind variations that when added to high wind speeds lead to large incursions on the non
linear region of the power conversion. The dynamic stall improved the aggregation model
because it generates a relation between wind speed variation and power variation on the
stall condition. However, it is not enough to compensate to the 6 wind turbines in high
turbulence intensity.
5.6.2.3 Extension to Large Wind Farms and Different Random Seeds
In order to verify the usage of the AWF to large number of wind turbines, wind farms
with sizes of 6, 12, 15, 21 and 30 wind turbines have been implemented in
Matlab/Simulink
®
. Moreover, it had also been indicated that the random seeds can modify
the results, and until now, the AWF has been applied to a small wind farm keeping the
same random seeds in the simulations in order to trace the results. In this subsection, the
model is extended to a large number of wind turbines and different random seeds are
applied to two different wind speed conditions (13 and 16m/s 20% turbulence intensity).
In this case, the wind farm model does not have the electrical dynamic components
and only the aerodynamic power is computed using the dynamic stall module. This is done
because on the aerodynamic rotor resides larger part of the nolinear conversion and the
simulation of larger wind farms including the electrical components would demand
extensive simulations not giving much information because the main differences reside on
the wind speed conversion. In addition, the cases presented in the previous section
supported that the differences found were related to the wind speed variations at low
frequency range (lower than 1Hz).
In this section, The wind farm layouts are similar to Figure 5.6 where the new wind
turbines are included in groups of three in the end of the rows keeping the same distance
94/152
between them and the same wind speed inflow angle, which makes the power extremely
correlated.
The mean power, standard deviation, maximum and minimum powers are the power
characteristics investigated in the following section, where they are normalised with the
rated power of the wind farm. Positive values stand for the AWF lower than the IWF.
Starting with the mean wind speed at 13m/s, Figure 5.18 presents the normalised
power characteristics as a function of different number of wind turbines.
Wind speed 13 m/s and Turbulence intensity 20%
2
1
0
1
2
3
4
0 5 10 15 20 25 30 35
Number of Wind Turbines
N
o
r
m
a
l
i
z
e
d
d
i
f
f
e
r
e
n
c
e
(
%
)
Average Power Standard deviation Poly. (Standard deviation) Poly. (Average Power)
Wind speed 13 m/s and Turbulence intensity 20%
4
2
0
2
4
6
8
0 5 10 15 20 25 30 35
Number of wind turbines
N
o
r
m
a
l
i
s
e
d
d
i
f
f
e
r
e
n
c
e
(
%
)
Maximum Power Minimum Power Poly. (Maximum Power) Poly. (Minimum Power)
Figure 5.18 Influences of different sizes of aggregate wind farms to the power characteristics at 13m/s, 20 %
turbulence intensity
The upper part of Figure 5.18 presents the normalised differences of the standard
deviation and mean powers. Increasing the number of wind turbines in the aggregation
procedure improves the performance of the AWF. In addition, the random seeds have
strong influences on the results that reduce with the number of wind turbines in the
aggregation procedure. These two main factors are related to the fact that increasing the
number of wind turbines in this particular nonlinear part of the power curve decreases the
influence of a single wind turbine on the total results.
One can notice in this subsection, that the standard deviation in the AWF is higher
than the IWF in agreement with the complete model presented in the previous subsection,
95/152
however, the mean power in the AWF is lower than the IWF, which is different from the
complete model in SIMPOW (Figure 5.9). This difference is related to the strong influence
of the random seed in the mean value. Probably in the complete model in SIMPOW,
random seeds used to generate the wind speeds in both Parksimu and AWFWS were two
extreme cases.
The lower part of Figure 5.18 presents the normalised differences of the maximum
and minimum power. In general, the differences between the IWF and AWF reduce when
increasing the number of wind turbines in the aggregation procedure. These conclusions are
also related to the reduced significance of a single wind turbine when increasing the high
number of wind turbines.
Following the analyses with the mean wind speed at 16m/s are presented.
Wind speed 16 m/s and Turbulence intensity 20%
1.5
1
0.5
0
0.5
1
1.5
2
0 5 10 15 20 25 30 35
Number of Wind Turbines
N
o
r
m
a
l
i
s
e
d
d
i
f
f
e
r
e
n
c
e
(
%
)
Average Power Standard deviation Poly. (Standard deviation) Poly. (Average Power)
Wind speed 16 m/s and Turbulence intensity 20%
8
6
4
2
0
2
4
6
8
0 5 10 15 20 25 30 35
Number of Wind Turbines
N
o
r
m
a
l
i
s
e
d
d
i
f
f
e
r
e
n
c
e
(
%
)
Maximum Power Minimum Power Poly. (Maximum Power) Poly. (Minimum Power)
Figure 5.19 Influences of different sizes of aggregate wind farms to power characteristics at 16m/s, 20 %
turbulence intensity
Similar to Figure 5.18, the upper part of Figure 5.19 presents the normalised
differences of the standard deviation and the mean power. Increasing the number of wind
turbines in the aggregation procedure improves the performance of the AWF. The random
seeds have strong influences on the results that in general reduce with the number of wind
turbines in the aggregation procedure. These two main factors are also related to the fact
96/152
that increasing the number of the wind turbines decreases the influence of a single wind
turbine on the total results.
The lower part of Figure 5.19 presents the normalised differences maximum and
minimum power differences between the AWF and the IWF. In general, the differences
between the IWF and AWF as well as the random seeds influences on the results tend to
reduce when increasing the number of wind turbines in the aggregation procedure. These
conclusions are also related to the reduced significance of a single wind turbine when
increasing the high number of wind turbines.
5.7 Aggregate Wind Farm Model Remarks
Simulation of large wind farms can be extremely expensive, depending on the size of
the wind farm and on the extension of the power system network. Simulation of all wind
turbines in a large extension of the power system is practically impossible. Therefore, there
is a demand for equivalent models to analyse the power system interaction with the wind
power.
The Aggregate Wind Farm model can replace the entire wind farm in power system
simulation programs. It can be applied to replace large wind farms in order to simulate the
impact on the power system characteristics, i.e. the power oscillations, power quality
deviations and to analyse the control or to optimise the power system operation.
The AWF is time inexpensive and easy to implement in a conventional power system
simulation tool used by the utilities. Using SIMPOW
®
installed in a Pentium III
®
 750MHz
(with 128Mb of ram memory), the simulation of the power produced from 6 wind turbines
with the IWF took around 5 hours while using the AWF, in the same program installed in
the same computer, the simulation time was reduced to 20 minutes.
The aggregate procedure focuses on the wind speed simulation. The wind speed
model accounts to spatial coherence of turbulence on the entire wind farm representing the
smooth effects on the power production in the wind farm.
The AWF has some limitations. The minimum power and standard deviation of the
power in high wind speed and high turbulence intensity do not agree with the simulation of
individual wind turbines.
However, the flicker agrees well. The flicker is mostly caused by the 3p effect from
wind turbines, which region the AWF seems to reproduce well the IWF.
Finally, The AWF simulates reasonably the power produced from wind farms. The
AWF makes possible power quality assessment of large number of wind turbines.
Respecting the AWF limitations, the model supports the design and operation of the power
system where the wind energy contributes to large part of the power production.
97/152
98/152
Chapter 6
6 Large Scale Integration – Case Analysis
The wind power impact on the power system quality and stability has been introduced
and specific models have been presented. Here, three cases are investigated to illustrate
interactions between wind power and power system in a largescale integration. Figure 6.1
shows the cases studied here and the tools used to perform the analyses.
Dynamic Analysis
Static Analysis
Case 3
Power Quality:
Nordic Power
System
Case 2
Voltage Stability &
Quality:
Brazilian Grid
Case 1
Voltage Stability:
Fictitious Grid
Figure 6.1 Case Studies.
6.1 Case Study 1: Voltage Stability in a Modern Power System
Large amount of wind energy can modify the voltage stability of a power system as
introduced in the chapter 2. Here, the impacts of wind power to the voltage stability
characteristics are illustrated.
A Loadflow program computes the loadability curves to the case analyse based on
static models. The loadability curves are computed by increasing the load in specific buses
of the power system (chapter 2). This way the power transfer capabilities is stressed until
the voltage collapses.
6.1.1 Power System Characteristics
The power system studied is a small part of an existing power system. Figure 6.2
shows a single line diagram of the modelled power system where the rated voltage is
220kV. Figure 6.2 includes the total impedance of the lines and the loads installed on the
network.
In Figure 6.2, the generators are installed on the buses 6 and 7. The bus 7 is the slack
bus and bus 6 is a voltagecontrolled bus (synchronous generator). Here, they have no
restriction on reactive or active powers. The balance on bus 6 is an active power production
of 110MW that explains why the loads are not expressed in the diagram. Bus 2 is supplying
20MVAr of reactive power to the network as indicated in the diagram and in addition, a
capacitor bank with rated power of 50MVAr is installed on bus 3 to reduce the reactive
power flux on the network.
99/152
GS
GS
40+j*30
P + j*Q
110  j*20
P + j*Q
100 + j*100
100 + j*48.83
Wind Power
Bus 3
Bus 7
Bus 6
Bus 2
Bus 1
Bus 4
Bus 5
Swing Bus
Voltage Controlled
PV  Bus
(2+j*65) Ohms
(22.8 + j*62.6) Ohms
(6.7 + j*15) Ohms
(6.75 + j*25)
Ohms
(6.7 + j*70)
Ohms
(27 + j*35)
Ohms
(6.7 + j*65) Ohms
Figure 6.2 Diagram of the Power System used in analysis (loads in MW and MVAr).
Simulations show that bus 3 is the weakest node in the system due to long lines
connecting to buses 7 and 4, in addition, bus 3 has a large reactive load installed that leads
to a power factor of 0.7 largely compensated with shunt capacitor banks. Figure 6.3 shows
the loadability curve to bus 3.
50 100 150 200 250 300 350 400 450 500 550
130
140
150
160
170
180
190
200
210
220
230
Loadability with cos(φ)=0.5
Load in the bus 3 (MVA)
V
o
l
t
a
g
e
l
e
v
e
l
(
k
V
)
bus 1
bus 2
bus 3
bus 4
Figure 6.3 Loadability curve to bus 3 without wind turbines.
As expected, the lowest voltage is achieved on bus 3 because it has the highest load
installed. The voltage on bus 2 is reduced due to high power flowing in the power system
although it is located between two voltagecontrolled buses. The voltage on bus 4 follows
the behaviour on bus 3 because the power flow to bus 3 causes voltage drop in the network.
Finally, the voltage on bus 1 remains constant because it is only connected to the bus 7
hence there is no change in the power flux in its transmission lines.
100/152
The voltage curves to bus 6 and 7 are not plotted because they are voltage fixed
amplitude (synchronous generators without restrictions). The synchronous generator keeps
the voltage constant on bus 6 by inserting or drawing reactive power. In this particular case,
there is no reactive power limitation on the synchronous machine on bus 6 hence the
voltage will always be constant. Bus 7 is the reference bus to the load flow calculations,
which has constant voltage amplitude and the angle is the reference for the calculations on
all other buses voltages. In addition, all active and reactive power balances are done on the
reference bus, which has no restrictions. In reality, bus 1 would also suffer voltage
variations once the voltage on bus (#7) starts to reduce because the power limits were
reached on the synchronous generator on bus 7, similar behaviour will happen to bus 6.
In Figure 6.3, the active and reactive load to bus 3 were increased linearly, i.e. the
active and reactive powers to bus 3 had fixed load factor, in this condition, the maximum
load that can be installed is 424MVA, i.e. 300MW + j300MVAr. The load flow problem is
complex and the maximum active and reactive powers that can be delivered through the
lines are different. Hence, the loadability procedure (load increase until the voltage
collapses) is modified to two: in the first, the reactive power is kept constant then the active
power increased linearly until the voltage collapses; in the second, the active power is kept
constant then the reactive power increased linearly until the voltage collapses.
Using the new loadability procedure, the independent maximum loads to active and
reactive powers are modified. In the first case, the maximum active power increased from
300MW to 540MW (the reactive power was kept constant 100MVAr). In the second case,
the maximum reactive power increased from 300MVAr to 380MVAr (the active load was
kept constant 100MW). The power system is weaker to transfer reactive power than active
power, which is expected and related to the electrical characteristics of the transmission
lines.
The maximum load that can be installed on bus 3 is basically limited to the network
strength because there are no limitations on the synchronous generators and the loads are
constant independent of the voltage. Inclusion of the reactive power limits on the
synchronous generators is likely to reduce the maximum load supplied to bus 3 because
when the synchronous machine reaches the maximum reactive power limit it loses the
voltage control capability.
The loads dependency on the voltage is complex and constant power related to the
voltage is a neutral assumption. An inverse dependency on the reactive, i.e. increase of
power demanded as the voltage decrease, reduces the limit of maximum load that can be
installed to bus 3, on the other hand, a direct power relation can increase the voltage
stability limits because the loads (on the power system) reduce with the voltage reduction.
In addition, the reactive power compensation with shunt capacitor banks is relevant to the
voltage stability because of the high sensitivity to the voltage (i.e. reactive power
proportional to V²). In this power system, the loadability to bus 3 is influenced by shunt
capacitors banks, which heavily compensate the reactive power (50MVAr). The voltage
reduction leads to an increase in the flow of reactive power trough the lines that reduce
further the voltage as introduced in chapter 2.
101/152
6.1.2 Wind Power Representation
In this section, a static model simulates the wind farm power production because the
voltage stability problem is slow. The wind farm static model must simulate the reactive
power demanded based on the active power and the voltage at the wind turbine bus.
In this thesis, the active (P) and the reactive (Q) powers are specified to the load flow
program (that computes the loadability curves). The P component is the rated power of the
wind farm to be installed on a specific site and Q is computed based on the voltage
terminals with a polynomial function. Because the wind turbines have no active control on
the voltage, this is a reasonable assumption that is very similar to the procedure suggested
in [28].
The polynomial function used here was fitted based on three different sizes of
induction machines as presented in Figure 6.4. Assuming the rated active power, the
reactive power consumption depends on the voltage (as introduced in chapter 4). An
increase in the voltage leads to an increase in the reactive power because the induction
machine excitation (i.e. the reactive power consumption) is related to V² on the other hand
the voltage decrease leads to an increase in the reactive losses that is related to V
2
increasing the total reactive power consumption.
0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2
0.95
1
1.05
1.1
1.15
1.2
1.25
Voltage terminals (p.u.)
N
o
r
m
a
l
i
s
e
d
d
e
m
a
n
d
o
f
r
e
a
c
t
i
v
e
p
o
w
e
r
(
p
.
u
.
)
Reactive power demand to rated active power
0.7841*X
−2
+ 0.7178*X
2
− 0.5036
500kW
Ploynomial fit
2.4MW
400kW
Figure 6.4 Loadability curve to bus 3 without wind turbines.
In Figure 6.4, the polynomial function to the reactive power for induction machines
does not include the noload capacitor bank installed at the wind turbines and the
normalisation factor is the nominal reactive power at rated voltage. The reactive power
compensation of the entire wind farm is modelled as shunt capacitor banks in the load flow
program.
6.1.3 Wind Power Impacts on Voltage Stability
In order to investigate the wind power impacts on the voltage stability, bus 5 is
connected to the power system. All the wind power in the power system is concentrated on
102/152
it. Bus 5 is connected to the weakest part of the network (bus 3) through a short radial line
(Figure 6.2). Connecting large wind power to the weakest bus is expected to be the worst
case scenario to the voltage stability because this bus already faces limited power transfer
capabilities. Figure 6.5 shows the behaviour of the voltage of the entire power system as a
function of the wind power injected in bus 5.
50 100 150 200 250 300 350 400 450 500 550
130
140
150
160
170
180
190
200
210
220
230
Wind power to bus 5 (MW)
V
o
l
t
a
g
e
l
e
v
e
l
(
k
V
)
bus 1
bus 2
bus 3
bus 4
bus 5
Figure 6.5 Maximum wind power integration concerning voltage stability.
Figure 6.5 is computed in a similar way to the loadability curve. The installed wind
power in bus 5 is increased until the voltage collapses. The maximum wind power that can
be installed before voltage collapses is 429MW being smaller than the maximum load that
can be installed to bus 3 (537MW). The reactive power demanded by the wind power
explains the small limit to wind power integration.
In order to investigate the wind power impacts on the voltage stability, a 100 MW
wind farm is installed to bus 5 and new loadability curves to bus 3 are computed. This size
of wind power represents 28% of power penetration in the power system.
Table 6.1 and Table 6.2 present the wind power impacts on the loadability curves
concerning voltage stability. Table 6.1 presents the loadability limits increasing only the
active load in bus 3. Two conditions are presented: in the first, the wind turbine reactive
power is constant (i.e. the reactive power polynomial function is not used); in the second,
the reactive power of the wind farm is represented with the polynomial function of the
voltage.
Table 6.1 Bus 3 loadability limits keeping the reactive power constant.
Condition P (MW) P (MW) Q (MVAr)
Without wind power 537 537 100
With wind power constant 580  100
With wind power reactive power polynomial
representation
 534 100
Difference +43 –3 0
103/152
The row presenting the difference in Table 6.1 shows the impact of the wind power
on the voltage stability. The wind power modifies the limits of voltage stability. At first
sight, the wind power local production shall improve the loadability on bus 3 as the fist
column presents an increase of 43MW. However, when the voltage dependency of reactive
power demanded by the wind power is included, the loadability to bus 3 is reduced. Those
results show that the reactive power from wind farms play an important role in the voltage
stability.
Table 6.2 shows the loadability limits with the active load in bus 3 kept constant.
Table 6.2 Bus 3 loadability limits keeping the active power constant.
Condition Q (MVAr) Q (MVAr) P (MW)
Without wind power 385 385 100
With wind power constant 350  100
With wind power reactive power polynomial
representation
 335 100
Difference –35 –50 0
100 200 300 400 500 600 700 800 900
160
170
180
190
200
210
220
230
Max wind power in the bus 5 (MW)
V
o
l
t
a
g
e
l
e
v
e
l
(
k
V
)
Wind turbine cos(φ) =1 with power electronics
bus 1
bus 2
bus 3
bus 4
bus 5
In this condition, the voltage stability limit is reduced because the reactive power to
the wind farms flows through the same transmission lines that supply bus 3. Similar to the
first case, the dependency on voltage of the reactive power from the wind farm strongly
influences the voltage stability. Considering constant power the wind power reduces in 9%
the loadability to bus 3 and including the voltage dependency, the reduction is 13%.
The demand of reactive power is the key factor on voltage stability from wind farms,
hence wind turbines with power electronics, which can actively control the reactive power,
can be used to regulate the voltage and improve the voltage stability. Following, Figure 6.8
illustrates the integration limits of wind power to bus 3 (concerning voltage stability) of
wind turbines with power converters that can keep power factor unity independent of the
voltage.
Figure 6.6 Maximum wind power integration concerning voltage stability (load factor unity).
In this case, the absence of reactive power flux to the wind farm improves the voltage
stability. The voltage on bus 5 is higher than on bus 3 because it is injecting power, in this
case the voltage tends to increase at first moment but as the power flux increases, the
104/152
associate losses increases leading to voltage reduction and finally to the voltage collapse. In
this new condition, the maximum wind power to the power system is 811MW, which is
much higher than the original case 429MW.
Another important feature using power electronics, as cited before, is that they can be
used to regulate the voltage by injecting or draining reactive power. However, the
maximum apparent power of the power converter must be respected. Figure 6.6 illustrates
the loadability curve to the power system with 100MW wind farm installed in bus 5 using
power electronics that can control the voltage (in this illustration, the voltage was set to 2%
above the nominal).
0 100 200 300 400 500 600 700
130
140
150
160
170
180
190
200
210
220
230
Installed power in the bus 3 (MW)
V
o
l
t
a
g
e
l
e
v
e
l
(
k
V
)
bus 1
bus 2
bus 3
bus 4
bus 5
Figure 6.7. Loadability curve to bus 3 with wind turbines using power electronics.
As the maximum power from the power electronics must be respected, the wind farm
is limited to 100MVA. The active power of the wind farm is 100MW that is reduced as the
reactive power is injected in the power system. The power converter is limited to operate at
minimum capacitive power factor of 0.8. When the wind farm reaches the maximum
reactive power production, the wind turbines lose the voltage control capability hence they
behave as fixed power production. Figure 6.8 illustrates the power production of the wind
farm with the loadability of bus 3 as well as the power factor of the wind farm.
105/152
100 200 300 400 500 600
0
10
20
30
40
50
60
70
80
90
100
Active power installed in Bus 3
A
c
t
i
v
e
a
n
d
r
e
a
c
t
i
v
e
p
o
w
e
r
f
r
o
m
w
i
n
d
t
u
r
b
i
n
e
s
100 200 300 400 500 600
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
c
o
s
(
φ
)
c
a
p
a
c
i
t
i
v
e
Active Power
Reactive Power
Power factor
Figure 6.8 Evolution of the power production from the wind turbines (with electronic power converter).
The active power produced from the wind farm is reduced to respect the thermal
limits of the power electronics. In particular example, the controllers were not implemented
and the reactive power from the wind farm was increased linearly until it reaches the
maximum of 60MVAr (corresponding to a power factor of 0.8 capacitive) and the active
power is reduced to 80MW in order to respect the 100MVA limit.
6.2 Case Study 2: Voltage Stability and Quality in a Brazilian
Power System
Large amount of wind power has been proposed to Brazil. The Brazilian power
system is mainly composed by hydropower stations. However, in some regions, it is not
possible to install more hydropower stations. In addition, during the last few years, a long
dry season leaded the entire power system to a critical minimum water level on the
reservoirs resulting in an national energy crisis [72].
The Brazilian government has introduced new regulations to diversify the power
matrix. The wind power is one of them, which is expected to contribute to the energy in a
large extend in years to come. The Northeast coast of Brazil presents good wind conditions
and complementary wind characteristics to the main river flow along the year [73].
Therefore, to this region alone, more than 2000MW of wind power have been proposed.
This section presents some investigations of the wind power integration to a part of the
Northeast region.
6.2.1 Power System Characteristics
The interconnected Brazilian electrical power system is divided in four regional
systems: North (N); Northeast (NE); Southeast (SE) and South (S). There are connections
between all four regional systems. The connections between the NE system and the N are
strong with a transfer capacity of few thousand Megawatts, however to the other regions
there is a single link in 500kV with limited power transfer. Figure 6.9 illustrates the
Brazilian interconnected system.
106/152
Transmission Lines
Legend
Wind power plans
Notheast System
Figure 6.9 Brazilian interconnected system power system [72]
In Figure 6.9, the NE system is circled. The regional NE power system has
approximately 10270MW of hydropower stations installed and 435MW of diesel power
stations. The hydropower stations are concentrated in the middle of the region (marked as
number 5 in Figure 6.9) far from the load consumption centres. All transmission lines in the
region are AC lines with voltage levels from 500kV to long distances down to 138kV to
short distances.
More than 2000 MW of wind power has been planned to the region [73]. The plans
account to install large amount of wind turbines in the coast area from Natal to Fortaleza
(NorthNortheast coast marked in Figure 6.9).
The installation of 100MW to a part of the system is investigated with respect to the
voltage stability and voltage quality during continuous operation. The 100MW is divided in
5 wind farms where all of them are installed to MOSSORO bus. Figure 6.10 presents a
schematic representation of the network to which the wind farms are installed.
107/152
100MW of
conventional wind
turbines
BANABUIU
S.CRUZ
Figure 6.10 Brazilian network studied [72].
The case studied focus on the region between BANABUIU and S.CRUZ (Figure
6.10). In Figure 6.10, the rated voltage between BANABUIU and ACUII is 230kV and
between ACUII and S.CRUZ is 138kV. The installation of the wind farm is planned to near
future when reinforcement of the power system is planned. The simulations take into
account the future power system condition [74]. In the future condition, between
BANABUIU and MOSSORO will have a 230kV line and from ACUII to S.CRUZ will
have an additional line in 230kV, where the link in 138kV will be open.
The load characteristics of the local network are presented in Table 6.3, which
presents the load to relevant nodes in the study case to high and light conditions.
BANABUIU and S.CRUZ are modelled as infinite nodes and all other buses are modelled
as constant load, i.e. voltage independent.
Table 6.3 Relevant loads in the Brazilian power system studied.
Node Heavy Load Light Load
MOSSORO 93.66MW+j⋅20.72MVAr 58.73MW+j⋅12.99MVAr
RUSSAS 52.80MW+j⋅13.70MVAr 42.40MW+j⋅3.30MVAr
ACU 52.88MW+j⋅3.81MVAr 45.82MW+j⋅3.31MVAr
6.2.2 Wind Power Representation
The total wind power has 100MW of conventional wind turbines, i.e. directly
connected to the network, that uses induction machines and partial reactive power
compensation with shunt capacitor banks. The total wind power is divided in 5 wind farms
with 30 identical wind turbines, all of them installed to MOSSORO bus on high voltage
level (230kV).
The voltage stability analyses are similar to section 6.1 Case Study 1: Voltage
Stability in a Modern Power System, where fixed active power and voltage dependent
reactive power model the wind power. The reactive power dependency on the voltage is
modelled with a polynomial function of voltage.
Here, the dynamic voltage quality is also assessed through the Aggregate Wind Farm
(AWF) model (presented in chapter 5), which simulates the power produced from the entire
wind farm. The aggregate wind farm represents identical 150 wind turbines (in five wind
farms with 30 wind turbines each) of 680kW. Actually, the 680 kW wind turbine does not
108/152
exist, hence the wind turbine used in this section has the same normalised data from a
similar wind turbine presented in chapter 5 (660kW).
The 100MW wind power is divided in 5 wind farms with identical layout and
identical 30 wind turbines each. There is no wind correlation between the five wind farms
and the aggregate wind speed model generates one time series to each wind farm.
Figure 6.11 presents the layout of each wind farm. The distance between wind
turbines is 4 times the rotor diameter (i.e. 4x48m = 192m). With this wind speed direction,
this is assumed the worstcase scenario because a large number of wind turbines are aligned
in a row with the wind speed. The distance between the wind turbines is small that is
acceptable because the wake effect is not simulated in the AWF.
0 200 400 600 800 1000 1200 1400 1600 1800 2000
150
200
250
300
350
400
450
500
550
600
L
a
t
e
r
a
l
d
i
s
t
a
n
c
e
(
m
)
Longitudinal distance (m)
Wind farm layout
N
Wind speed
Figure 6.11 Layout of a single wind farm applied to the Brazilian power system studied.
6.2.3 Wind Power Impacts on the Voltage Stability
The loadability curves characterize the voltage stability in the study cases similar to
the previous section. The power system in Figure 6.10 is implemented in the loadability
computation tool (chapter 2) as presented in Figure 6.12.
In Figure 6.12, the arrows represent loads installed to the buses. The line between
ACUII and S.CRUZ is modelled as 230KV according to the reinforcement plans. Similarly,
the ring connection between BANABUIU and MOSSORO that is not present in Figure
6.10, here is included according to the reinforcement plans.
109/152
BANABUIU
MOSSORO
RUSSAS
ACUII
MOSSORO
 LOAD
S.CRUZ
230KV
230KV
230KV
230KV
230KV
WIND FARMS
69KV
Reference
PV node
Figure 6.12 Network topology of Brazilian power system studied.
In order to characterize the original installation, loadability limits are computed to the
MOSSORO bus. The loadability limits uses the same procedure presented in the previous
section, which is to increase the active and reactive load individually until the voltage
collapses. Table 6.4 presents the loadability limits to MOSSORO to different load
conditions.
Table 6.4 Loadability limits to MOSSORO.
Load Conditions
Maximum
Active load (MW)
Maximum
Reactive load (MVAr)
Heavy Load 888 534
Light Load 733 460
The classification of the loadability limits to MOSSORO presented in Table 6.4
represents the maximum active and reactive loads that can be installed to MOSSORO
before the voltage collapses independent of each other, as explained in the previous section.
The active load that can be installed to MOSSORO is higher than the reactive because
of the electrical characteristics of the transmission lines in the local power system. In
addition, the differences between heavy load and light load conditions happen because
under heavy load, several shunt capacitors were switched on to compensate the reactive
power, while under light load condition some shunt reactors were switched on to limit the
voltage increase. However, the differences in terms of reactive power limits are very small
because the reactive power produced by the shunt capacitors is very sensitive to voltage
variation.
The highest value (888MW, 364MVAr) characterizes the loadability limit to
MOSSORO because with the load evolution all capacitors available in the power systems
must be switched on to prevent high reactive power flow on the network.
Before analysing the impacts of wind power on the voltage stability, the limits to
integration of wind power are drawn concerning the voltage stability. To this power system,
the maximum wind power that can be installed to MOSSORO (respecting the voltage
stability) are 780MW in heavy load and 727MW in light load. In both cases, the maximum
wind power to the bus is small than the maximum load that can be installed because the
wind turbines demand reactive power as explained in the previous section.
The difference between the heavy and light load conditions are related to the shunt
capacitors installed in the system under heavy load condition similar to the loadability
limits presented in Table 6.4.
110/152
Figure 6.13 presents the evolution of the voltage in the relevant buses along the wind
power installed to MOSSORO (230kV) in light load condition.
Wind Power Installed in MOSSORO 230kV(MW)
V
o
l
t
a
g
e
(
p
.
u
.
)
MOSSORO 230kV
Mossoro 69kV
RUSSAS
ACUII
0 100 200 300 400 500 600 700 800
0.7
0.75
0.8
0.85
0.9
0.95
1
1.05
Figure 6.13 Maximum wind power to MOSSORO bus (light load condition).
The voltage on the power system reduces with the wind power inserted in
MOSSORO. The closest buses present the high voltage variation while ACUII follow the
general behaviour but with less severity. The MOSSORRO 69kV is included in Figure 6.13
just to present the voltage on the main load connected to the MOSSORO bus.
In order to assess the impacts of 100MW on the voltage stability, the wind power is
installed to MOSSORO and new loadability curves are computed. Table 6.5 presents the
loadability limits computed to different load conditions.
Table 6.5 Loadability limits to MOSSORO with wind power.
Conditions Load condition
Maximum
Active load (MW)
Maximum
Reactive load (MVAr)
No wind power Heavy Load 888 534
With wind power Heavy Load 905
432
(+110MVAr to the wind farm)
No wind power Light Load 733 460
With wind power Light Load 758
380
(+114MVAr to the wind farm)
In Table 6.5, the loadability limits to MOSSORO in high and low load conditions are
compared to cases with and without wind power installed (at rated power). In this power
system, the wind power improves the voltage stability concerning the active. However, as
expected, the net loadability limit concerning reactive power is reduced as the wind farms
demand reactive power, but the total reactive power flowing to MOSSORO is slightly
increased (7.4% in the light load and 1.5% in heavy load) because the wind farms have
shunt capacitors that help to improve the voltage stability.
6.2.4 Wind Power Impact on the Voltage Quality
The impacts of 100MW wind farm on the voltage quality are analysed by means of
dynamic simulations using the AWF presented in chapter 5. The AWF is applied to a
111/152
conventional fixed speed wind turbine connected to the transmission level at 230kV with
dedicated lines and transformers.
The voltage is analysed under different wind speeds with two load conditions: heavy
load and light load. The wind speeds are simulated (using the aggregate wind farm wind
speed simulator from chapter 5) to 8, 10, 12 and 14 m/s all of them with 20% turbulence
intensity. Those wind speeds represent the wind speed range with the most relevant power
variations from wind turbines. Above 14 m/s the power conversion of the wind turbine
limits the active power and the wind speed variations lead to relative small power variations
and below 8 m/s the level of power is small and the power variations are also reduced.
Figure 6.14 shows the 10 minutes average values of the active power produced by the
wind farms and its influences on the average voltages on MOSSORO under heavy and light
load conditions. Figure 6.14 also includes the maximum and minimum values to each load
condition as envelop curves.
8 9 10 11 12 13 14
0
50
100
150
W
i
n
d
p
o
w
e
r
(
M
W
)
Mean power
Max power
Min power
8 9 10 11 12 13 14
234
236
238
240
242
244
V
o
l
t
a
g
e
a
t
M
o
s
s
o
r
o
(
k
V
)
Mean voltage − Heavy load
Max voltage − Heavy load
Min voltage − Heavy load
8 9 10 11 12 13 14
234
236
238
240
242
244
V
o
l
t
a
g
e
a
t
M
o
s
s
o
r
o
(
k
V
)
Wind speed (m/s)
Mean voltage − Light load
Max voltage − Light load
Min voltage − Light load
Figure 6.14 Wind power influences on the voltage to different wind speeds.
The power production by the wind power increases with the wind speed and the
voltages on both load conditions reduce with the increased wind speed. As the mean wind
speed increases, the wind power produced and the reactive power demands increase that
causes voltage drops due to losses in the network. The main voltage differences between
the heavy and light load are caused because of the reactive power compensation schemes to
each load condition (i.e. large shunt capacitors installed under heavy load condition).
In Figure 6.14, the aggregation of 150 wind turbines leaded to a very smooth wind
power production where the maximum and minimum powers are very close to the average.
As an illustration, in the average wind speed of 10m/s, the difference between the
maximum and minimum active powers is approximately 10MW while the mean power
produced is approximately 70MW (next section presents the time series).
Another relevant characteristic, presented as follow, is the standard deviation of the
power and voltage to different wind speeds as presented in Figure 6.15, where the reactive
power is also plotted.
112/152
8 9 10 11 12 13 14
0.8
1
1.2
1.4
1.6
A
c
t
i
v
e
P
.
(
M
W
)
Std dev. Active P. Heavy load
Std dev. Active P. Light load
8 9 10 11 12 13 14
0
0.05
0.1
0.15
0.2
V
o
l
t
a
g
e
(
k
V
)
Std dev. Voltage Heavy load
Std dev. Voltage Light load
8 9 10 11 12 13 14
0.2
0.4
0.6
0.8
1
R
e
a
c
t
i
v
e
P
.
(
M
V
A
R
)
Wind speed (m/s)
Std dev. Reactive Heavy load
Std dev. Reactive Light load
Figure 6.15 General wind power influences on MOSSORO bus.
The standard deviation (STD) gives an idea of the processes’ fluctuations, from
Figure 6.15, the SDT of the active power produced from the wind farm has a maximum on
the mean wind speed of 10 m/s then reduces because of the nonlinear characteristic of the
power curve. The voltage and the reactive power are strongly related. Both STD of voltage
and reactive power increases with the wind speed because on high wind speeds the high
active power reduces the voltage that consequently reduces the reactive power produced by
shunt capacitors leading to higher reactive power fluxes.
The main differences between the heavy and light load conditions are on high wind
speeds where the voltage reduction influences the reactive power produced by the
capacitors banks installed in the power system. Hence, the heavy load condition, which has
more capacitors installed, presents the higher standard deviation of voltage and reactive
power.
The curves above support that probabilistic load flow studies should be used for
analysis in the future in order to assess the impact of the wind turbines in the whole
operating range.
Following, the wind farm influences on the voltage quality to the wind speed of
10m/s are detailed to each load condition.
6.2.4.1 Light Load Condition
Figure 6.16 presents the time series of the wind power produced and the voltage on
bus MOSSORO 230kV at mean wind speed 10 m/s and turbulence intensity of 20% in light
load condition.
113/152
0 100 200 300 400 500 600
242.7
242.8
242.9
243
243.1
243.2
Time (s)
V
o
l
t
a
g
e
(
k
V
)
Voltage at PCC
0 100 200 300 400 500 600
65
70
75
80
Wind Power Production in Mossoro
Time (s)
A
c
t
i
v
e
P
o
w
e
r
(
M
W
)
Figure 6.16. Voltage at MOSSORO and power flux from the wind farm in light load condition (mean wind
speed 10m/s).
Most of the time, the voltage variations are within 0.3kV (0.125% of the mean
voltage 240kV). The deepest voltage levels occur on the highest active power production
because the total power flow in the network cause voltage drops as introduced in chapter 2.
The voltage is 5.6% above the nominal value (230kV) because the system is designed to
operate a little above the nominal in order to compensate the voltage drops in lower voltage
levels. Following, Figure 6.17 presents the wind speed, active and reactive powers and
voltage statistics in terms of histograms in light load condition.
O
c
u
r
r
e
n
c
e
s
(
%
)
Wind speed (m/s)
33.5 34 34.5 35 35.5 36 36.5 37 37.5 38 38.5
0
2
4
6
O
c
u
r
r
e
n
c
e
s
(
%
)
Reactive Power (MVAr)
64 66 68 70 72 74 76
0
2
4
6
O
c
u
r
r
e
n
c
e
s
(
%
)
Active Power (MW)
242.75 242.8 242.85 242.9 242.95 243 243.05 243.1 243.15 243.2
0
2
4
6
O
c
u
r
r
e
n
c
e
s
(
%
)
Voltage (kV)
9.7 9.8 9.9 10 10.1 10.2 10.3 10.4
0
2
4
6
Figure 6.17 Statistics wind speed, active power, reactive power and voltage variations at MOSSORO (light
load).
114/152
The voltage and reactive power variations are related to the active power production
from the wind farm that depends on the aggregate wind speed. The reactive power
demanded from the wind farm depends on the active power and on the voltage on
MOSSORO. Higher wind speed leads to higher active power production that leads to
higher reactive power demand that leads to lower voltages. The relation between the
different parameters is not linear and it becomes more complex from the upper to the lower
part of Figure 6.17, which explain the different statistical distribution of the voltage.
6.2.4.2 Heavy Load Condition
Following, Figure 6.18 presents the time series of the wind power produced and the
voltage on bus MOSSORO 230kV at mean wind speed 10 m/s and turbulence intensity of
20% in heavy load condition.
0 100 200 300 400 500 600
235.9
236
236.1
236.2
236.3
236.4
Time (s)
V
o
l
t
a
g
e
(
k
V
)
Voltage at PCC
0 100 200 300 400 500 600
65
70
75
80
Wind Power Production in Mossoro
Time (s)
A
c
t
i
v
e
P
o
w
e
r
(
M
W
)
Figure 6.18 Voltage at MOSSORO and power flux from the wind farm in heavy load condition (mean wind
speed 10m/s).
The time series to heavy load condition are very similar to the ones with light load
condition. However, the mean value of voltage is different and the nonlinear relation of the
voltage and reactive power leads to a different statistical distribution of the voltage and
reactive power as presented in Figure 6.17.
115/152
9.7 9.8 9.9 10 10.1 10.2 10.3 10.4
0
2
4
6
O
c
u
r
r
e
n
c
e
s
(
%
)
Wind speed (m/s)
33.5 34 34.5 35 35.5 36 36.5 37 37.5 38
0
2
4
6
O
c
u
r
r
e
n
c
e
s
(
%
)
Reactive Power (MVAr)
65 66 67 68 69 70 71 72 73 74 75
0
2
4
6
O
c
u
r
r
e
n
c
e
s
(
%
)
Active Power (MW)
235.9 235.95 236 236.05 236.1 236.15 236.2 236.25 236.3 236.35
0
2
4
6
O
c
u
r
r
e
n
c
e
s
(
%
)
Voltage (kV)
Figure 6.19 Statistics wind speed, active power, reactive power and voltage variations on MOSSORO (Heavy
load).
The wind speed is exactly the same as the one applied to the light load condition,
similarly the active power produced in heavy load condition is very similar to the light load
condition. However due to the lower mean voltage, the distributions of the reactive power
and voltage at this time are different. The voltage is more concentrated to the mean value
and the reactive power more spread. Those differences can be related to the nonlinear
relation between the active power, reactive power and the voltage that in this case, in
addition, the lower voltage level increases the losses.
The dynamic voltage variations in this power system are very small. This is related to
two main conditions, first, the wind power is connected to transmission lines (high voltage),
which are relatively strong and have lower losses that reduce the voltage drop. A second
explanation resides on the power smoothing effects from the spatial distribution of the wind
turbines. The low correlation between different wind speeds (acting on each wind turbine)
smoothes the power produced by the wind farm that reduces the power flow variation hence
lower voltage variation.
6.3 Case Study 3: Power System Interactions – NORDEL
The Nordic countries are expected to face large wind energy integration in the years
to come. This section illustrates the impacts of largescale integration of wind power to the
frequency and voltage regulation of large power systems.
The Nordic power system has generators spread over large distances including
Denmark, Finland, Iceland, Norway, and Sweden. This power system includes hydro,
thermo (oil, coal fired and even geothermal) and nuclear power stations supplying in
parallel electricity to more than 20 million inhabitants.
116/152
6.3.1 Power System Description
Finland, Sweden, Norway, and a part of the Denmark are synchronized in the Nordel
system. The Nordel, hereby defined, permits power exchange between the Nordic countries
(Denmark, Finland, Iceland, Norway and Sweden), whose primary aims are optimise and
efficient electricity market.
Here, the objectives are the dynamic power quality and voltage assessment of the
synchronized interconnected Nordel power system. Figure 6.20 illustrates the Nordic Power
System [75].
Figure 6.20 High voltage Nordic power network [75].
The synchronous Nordel power system includes the eastern part of Denmark
(Zealand), Finland, Norway, and Sweden. The western part of Denmark is not synchronized
to the Nordel but to the European System. There are no interconnections between Iceland
and the rest of the Nordel power system. There are HVDC links between the synchronized
Nordel and Germany, Poland and the western part of Denmark.
In the synchronized Nordel, the frequency and voltage are regulated in order to
remain within certain limits. The frequency regulation includes stability requirements and
power exchange agreements. The frequency in normal conditions on the 400kV network
must be within 49.9 and 50.1Hz [76]. Under exceptional conditions, the frequency can vary
within 47.5 and 52Hz, however those conditions are related to transient and island
problems. The voltage regulation in the Nordel includes optimised reactive power flux as
well as stability requirements. In addition, the voltages in the Nordel during continuous
operations can be within 90 and 105% of the nominal values [77].
The harmonious operation of the Nordic power system means that all machines
contribute to the voltage and frequency control avoiding power oscillations and voltage
deviations. The control characteristics are mainly based on agreements in power exchange
and stability requirements. The size and characteristics of the Nordic power system makes
it an interesting study case. In the Nordel, the primary frequency control automatically
117/152
keeps the balance of consumption and production and the secondary frequency control
manually regulates the power production agreements in certain areas of the power system.
The secondary frequency control is not studied here.
The combined primary frequency control, in order to keep the system frequency
constant, shall react to the frequency variations with 6000MW/Hz. Table 6.6 presents the
requirements of frequency regulation to Nordel.
Table 6.6. Requirements on frequency response in the Nordel power system.
Country Frequency Response (MW/Hz)
Denmark 270
Finland 1050
Norway 2220
Sweden 2460
Nordel 6000
To maintain contractual voltage quality and avoid voltage collapse, the Nordel power
system must be capable of keeping the voltage within permissible range. The voltage
controllers are specified to keep the voltage within permissible level and to provide reactive
power reserve to ensure reliable operation.
6.3.1.1 Reduced Nordel Model
In the following analysis, a reduced network simulates the dynamic electrical
characteristics of the synchronized Nordel power system. The reduced Nordel power
system model is explained in [38]. The reduced network represents the synchronized
Nordel with 34 aggregate nodes. The aggregate nodes represent the loads and the
generation units with the frequency and voltage controllers included. The power system
representation is done as follow:
• Finnish system: 2 nodes in 400kV;
• Swedish system: 08 nodes in 400kV and 2 nodes in 300kV;
• Norwegian system: 11 nodes in 400kV and 11 nodes in 300kV;
• Danish system is included in the model as constant loads installed to the
neighbours systems (these loads can be negative or positive).
The reduced model takes into account the electromechanical dynamic representation
using aggregate models to the synchronous generators. The reduced model has 21
equivalent synchronous generators and includes equivalents controllers to the frequency
and voltage. The HVDC links are modelled as constant PQ loads to the proper buses.
Figure 6.21 presents the network topology of the reduced Nordel system.
118/152
F
i
n
l
a
n
d
S
w
e
d
e
n
N
o
r
w
a
y
Figure 6.21 Reduced model to the Nordic power system [38].
In Figure 6.21, the network in red represents the 400kV grid and the network in green
represents the 300kV grid. The aggregate synchronous generators and loads are also plotted
in Figure 6.21. Table 6.7 presents the synchronous machines and the aggregate wind farms
(AWF) connecting nodes.
119/152
Table 6.7 Nordel reduced machines connection nodes.
Machine # Generators Name Bus Name AWF
1 Finland SOFGG1_ SOFIN Yes
2 Finland NOFGG1_ NOFIN Yes
3 Sweden: STCKGG1_ STCKH Yes
4 Sweden: SVSGG1_ SVSW Yes
5 Sweden: SVWGG12 SVW
2
Yes
2
6 Sweden: STRFGG1_ STRFN No
7 Sweden: SVNGG11 SVN1 No
8 Sweden: SVMGG1_ SVM1 No
9 Sweden: SVNGG12 SVN2 No
10 Sweden: SVWGG11 SVW
2
Yes
2
 No sync. Machine installed KRLSK Yes
11 Norway: EASTGG1_ EAST3 No
12 Norway: HDALGG1_ HDAL3 No
13 Norway: CNTRGG11 CNTR3A No
14 Norway: CNTRGG12 CNTR3B No
15 Norway: SOUTGG11 SOUTGG11 No
16 Norway: WESTGG1_ WEST300 No
17 Norway: NWESGG1_ NWEST3 No
18 Norway: MID3GG1_ MID300 Yes
19 Norway: NOR3GG1_ NOR300 Yes
6.3.1.2 Nordel Characteristics
Before connecting the large wind farms, the Nordel power system is characterized
with modal analysis to indicate the main oscillation modes, the phase angle of the modes
and the machines participating on those modes.
Using the dynamic simulation tool (SIMPOW), the relevant eigenfrenquencies to the
Nordel are expressed in Figure 6.22 (without wind power installed).
In Figure 6.22, the xaxis presents the real part of the eigenvalue: smaller absolute
values in the negative plane means close to instability. The yaxis presents the damped
frequency (in Hz) of the mode (imaginary part of the eigenvalue). All eigenvalues are in the
stable region of the plane.
2
There is only one AWF installed to SVW.
120/152
0.317934
1.296631
1.112578
0.5362
0.551629
1.5
1
0.5
0
0.5
1
1.5
1.2 1 0.8 0.6 0.4 0.2 0
Real part of the Eigenvalue
I
m
a
g
p
a
r
t
o
f
t
h
e
E
i
g
e
n
v
a
l
u
e
(
H
z
)
Figure 6.22 Relevant eigenvalues of the reduced model to the Nordel.
Considering that wind turbines produce relevant power variation up to few hertz, five
modes (see marked modes in Figure 6.22) are characterized using modal analysis. The
relevant characteristics of the selected modes are:
1. 0.318Hz – oscillation of the South Finnish system against to the entire power system;
2. 0.552Hz – general oscillation in the entire power system, see Figure 6.23.
3. 0.536Hz – oscillation of Stockholm node against the entire power system;
4. 1.113Hz – oscillation of the North Sweden against the power system;
5. 1.297Hz – oscillation of the East part of the system (Norway) system against West
part of Sweden
Figure 6.23 presents the interarea phase and amplitudes of participation from the
synchronous machines in the second electromechanical mode selected (0.552Hz). In Figure
6.23, the vectors near to the generators represent the participation of the machines in the
electromechanical mode. The modulus of the vector indicates the magnitude of the
participation, i.e. to how extent this machine is involved in the electromechanical mode.
The phase angle of the vectors indicates the tendency of rotor angle deviation of each
machine, i.e. the power exchange between different machines (power system oscillation).
121/152
Figure 6.23 Modal analysis of the eigenvalue 0.34962 + 0.55164 Hz –Nordic Power System.
The entire power system can be excited in the frequency of 0.55Hz. At this particular
electromechanical mode, the south Finnish system (machine installed to SOFIN__), the
Swedish system (machines installed to SVN1__, SVN2__ and STCKH__) and the
Norwegian system (particularly the machines installed to NOR300__, SOUTH3B_,
HDAL3_, NWEST3__) all have strong participation in this electromechanical mode. The
phase angles from this graphical presentation shows that the South and Southwest of
122/152
Norway oscillates in phase with the SOFIN__ and both of them oscillate against (i.e.
almost opposite phase directions) the north of Norway and to some extent to the north of
Sweden. Moreover, the Swedish machine installed in node STCKH oscillates in a different
direction because of the complex power system operation.
All electromechanical modes are stable. Therefore, all the amplitudes of the power
oscillations shall decay in time. The rotor angle stability in this power system
representation from modal analysis does not present any unstable operation problem. The
Finnish system is in most of the electromechanical modes, in addition, it has the less
damped electromechanical mode. Hence, the Finnish system is expected to suffer severe
oscillations compared to the other machines although the oscillations shall not increase in
time.
6.3.2 Wind Power Projects and Representation
The wind power plants were chosen based on national wind power prognosis. A total
of 4400MW of wind turbines are simulated to Finland, Norway, and Sweden. Eight sets of
550MW wind farms simulate the wind power integration to the Nordel. Table 6.8 presents
the wind power plans simulated in this section.
Table 6.8 Wind power plans simulated (Figure 6.21 identifies the buses’ name).
Country Wind Power Installed
Finland 1100MW – installed 50% to NOFIN and 50% to SOFIN
Norway 1100MW – installed 50% to MID300 and 50% to NOR300
Sweden 2200MW – installed as four sets of 550MW to: SVW, SVSW, KRLSK, and STCK.
The dynamic AWF presented in chapter 5 models the wind power to the Nordel
where sets of 550MW are installed to specific nodes on the Nordel (Table 6.8). The
550MW AWF represents 300 wind turbines of approximately 1.83MW. The wind turbines
are a conventional type, i.e. directly connected to the network with induction generator and
without power electronics.
The 550MW aggregate wind farm is composed of 10 wind farms with identical layout
and identical 30 wind turbines each. There is no correlation between the 10 wind farms.
The aggregate wind speed model generates time series to each wind farm and the 10 time
series are averaged to produce a single aggregate wind speed.
The layout of each small wind farm is similar to Figure 6.11, where to this wind
turbine, the distance between each wind turbine is approximately four times the rotor
diameter (256m) where the rotor diameter is 64m. The wind speed angle is 90°, which
means that 10 wind turbines will face strong correlated wind speed.
6.3.2.1 Aggregate Wind Farm Model
The aggregate wind farm is scale up of a 1.83MW wind turbine that formally does not
exists, however the data is fitted based on similar size of wind turbines. Table 6.9 presents
the main characteristics of the 1.83MW wind turbine used in the simulations.
123/152
Table 6.9 Basic characteristics of the 1.83MW wind turbine.
Nominal Power 1.83MW
Rotor diameter 64 m
Generator speed 1500 rpm
Rotor speed 21 rpm
Capacitor Bank 660kVAr
Electrical frequency 50Hz
Tower height 60 m
Gearbox ratio 71.43
First torsional frequency 1.0Hz
Estimated logarithmic damping to drive train 5%
Number of blades 3
3p frequency 1.05Hz
Eight different aggregate wind speeds are generated: two of 8m/s; two of 10m/s; two
of 12m/s; and two of 16m/s all of them with turbulence intensity of 20% that is considered
a strong turbulence intensity. The wind power produced from each aggregate wind farm is
presented in Figure 6.24.
0 100 200 300 400 500 600
320
340
360
380
P
o
w
e
r
(
M
W
)
10 m/s Average wind speed
0 100 200 300 400 500 600
180
200
220
240
P
o
w
e
r
(
M
W
)
08 m/s Average wind speed
0 100 200 300 400 500 600
420
440
460
480
P
o
w
e
r
(
M
W
)
12 m/s Average wind speed
0 100 200 300 400 500 600
545
550
555
Time (s)
P
o
w
e
r
(
M
W
)
16 m/s Average wind speed
Figure 6.24 Aggregate wind farm power simulation.
In Figure 6.24, two time series of power produced from the AWF with mean wind
speed at 8, 10 12 and 16m/s are presented. The larger standard deviation of the power
occurs in the average wind speed of 10 m/s followed by the powers at 12m/s and 8m/s. At
16 m/s is the lowest standard deviation of power produced. Figure 6.25 presents the time
series of the active power variations from different wind speeds.
124/152
0 100 200 300 400 500 600
−25
−20
−15
−10
−5
0
5
10
15
20
25
Time (s)
P
o
w
e
r
v
a
r
i
a
t
i
o
n
(
M
W
)
Aggregate wind speed 8 m/s
Aggregate wind speed 10 m/s
Aggregate wind speed 12 m/s
Aggregate wind speed 16 m/s
Figure 6.25 Aggregate wind farm power variation.
The power variation from the aggregate wind farm is very smooth compared to the
power produced from a single wind turbine mainly because of the low wind correlation on
large distances. The largest difference between maximum and minimum powers is at 10m/s
(7% of the wind farm rated power) followed by the ones at 8m/s (6% of the rated power), at
12m/s (5.2%) and at 16m/s, which the difference is reduced to 1.6%.
The differences between the maximum and minimum powers (in less than 10
minutes) are relevant to the frequency controllers of the power system because they specify
the maximum power range. However, the standard deviation can be even more relevant to
express the power variation of the entire process than the range of power variation because
the wind power does not vary instantaneously between the maximum and minimum. Figure
6.26 presents the power characteristics: mean, maximum, minimum and standard deviation
produced by the AWF with the 8 time series of wind speed.
0
100
200
300
400
500
600
6 8 10 12 14 16 18
Wind speed (m/s)
A
g
g
r
e
g
a
t
e
W
i
n
d
P
o
w
e
r
(
M
W
)
0.00
2.00
4.00
6.00
8.00
10.00
12.00
S
t
a
n
d
a
r
d
D
e
v
i
a
t
i
o
n
(
M
W
)
max min Meanvalue Standard Deviation
Figure 6.26 Aggregate wind farm power characteristics (at 20% turbulence intensity).
125/152
The power characteristics of the aggregate wind farms in different wind speeds are as
expected. The standard deviation presented two quite different values because to each
aggregate wind farm, new random seeds are used in order to avoid correlated time series of
wind speeds and, as presented in chapter 5, the random seeds have influences on the
standard deviation.
Each one of the 8 cases presented above, are installed to the Nordel where they
simulate the largescale integration of wind power. Different wind speeds are applied
because it is not expected that the entire power system would have the same average wind
speed.
6.3.3 Wind Power Impacts on the Power System Voltage and Frequency
Regulation
The mean wind speed to each AWF has been assigned, to some extend, arbitrarily.
However, because Finland has presented the less damped electromechanical modes of
oscillations, to the two nodes, two cases are simulated: Case I which simulates large power
variation from wind power (12m/s) and Case II which simulates small wind power variation
(16m/s).
The average wind speed simulated to Norway is 10m/s, which have the largest power
variation, was selected because from modal analysis, these nodes have a considerable
participation on power oscillations. Table 6.10 presents the wind applied to the NORDEL
where it is simulated large wind power variation to FINLAND.
Table 6.10 Aggregate wind farms average wind speeds.
Wind speed (m/s)
AWF #
Country/
Near Generator
Bus Name
Case I Case II
1 Finland SOFGG1_ SOFIN 16 12
2 Finland NOFGG1_ NOFIN 16 12
3 Sweden: STCKGG1_ STCKH 8 8
4 Sweden: SVSGG1_ SVSW 8 8
5 Sweden: SVWGG12 SVW
3
12 16
5 Sweden: SVWGG11 SVW
3
12 16
6 Sweden: No sync. machine KRLSK 12 16
7 Norway: MID3GG1_ MID300 10 10
8 Norway: NOR3GG1_ NOR300 10 10
6.3.3.1 Frequency controllers
The performance of the frequency controllers and of the power systems depends on
several aspects. One of the relevant aspects here is the characteristic of the speed governors
from all synchronous generators. The power oscillations are generated by the power
unbalance and the properties of the speed regulators among other thinks. Figure 6.27
presents the simulated power balance of the entire Nordel as well as the total power
variation from the wind turbines and from the synchronous generators.
3
There is only one AWF installed to SVW.
126/152
0 100 200 300 400 500 600
−60
−40
−20
0
20
40
60
P
o
w
e
r
(
M
W
)
Time (s)
total synchronous machines power variation
total wind power variation
power balance
Figure 6.27 Power variations and power balance in the Nordel case studied.
In Figure 6.27, the total instantaneous variations of active power produced from all
synchronous machines, from all AWF and the total instantaneous power balance are
plotted. The total instantaneous variations are computed by adding the power from all
machines at each time. The power balance is the difference between the power variation
from the synchronous machines and the wind farms. The secondary frequency control is not
modelled, so the set points of the synchronous generators were adjusted aiming to keep the
regional power balances. In principle, the dynamic power balance presented in Figure 6.27
depends on the set points and characteristics of all generators and changing the average
wind speed from different places did not significantly modify the total results presented in
Figure 6.27.
In general, the power balance varies in a narrow range (within ±10MW) meaning a
good combined response of the speed governors. However, the speed controllers are acting
individually and independently in each machine, therefore some power oscillations between
different machines cannot be presented in the total instantaneous power balance because
they can be cancelled from those machines that the power oscillations are out of phase.
In order to verify the dynamic operation of the power system, Figure 6.28 presents the
time series of the speed and the total standard deviation of the power from each machine in
the Nordel, in addition, the total wind power production is included.
127/152
0 100 200 300 400 500 600
3050
3100
3150
3200
time (s)
W
i
n
d
P
o
w
e
r
(
M
W
)
Total wind power to NORDEL
0 100 200 300 400 500 600
49.98
49.99
50
50.01
50.02
time (s)
F
r
e
q
u
e
n
c
y
(
H
z
)
Speed of each machine
0 2 4 6 8 10 12 14 16 18 20
0
100
200
300
400
Machine #
S
t
d
o
f
P
o
w
e
r
(
M
W
)
STD of the power of each machine
Figure 6.28 Wind power, frequency and standard deviation of power (mean wind speed at 12m/s to Finland).
The frequency in Figure 6.28 represents the speed in each machine that is well
regulated and the variations are within 49.99 and 50.01Hz. An increase in the wind power
leads to an increase on the frequency because the load on synchronous machines is reduced
and the speed drop characteristics of the speed governors lead to an operational frequency
slightly above the rated. Similarly, a reduction in the wind power leads to a decrease in the
frequency.
The slow frequency variations are related to the total slow wind speed variation and
the fast speed variations are related to the dynamic power variations from the wind turbines
(e.g. 3p and torsional moments).
In the lower part of Figure 6.28, the standard deviation of the power from each
synchronous machine is presented, where Table 6.7 presents the machine names. The
higher power variation was presented by machine #9 that is connected to bus SVN2__.
This bus is connected to the Finnish system, to the Norwegian system and to the STCKH_
both of them with the higher wind power variations that explains the higher power
variation.
In order to analyse in detail the dynamic behaviour of the power system with large
wind power integration, the power spectral distribution of speed and power to the machines
with high standard deviation (#9 and #7 from Figure 6.31) and the machine #1 (Finland that
showed a strong electromechanical participation in the oscillation modes) are presented in
Figure 6.29.
128/152
0 0.5 1 1.5 2 2.5 3
10
−2
10
0
10
2
Frequency (Hz)
P
o
w
e
r
s
p
e
c
t
r
a
(
M
W
2
)
Active power @ machine # 9
Active power @ machine # 7
Active power @ machine # 1
0 0.5 1 1.5 2 2.5 3
10
−9
10
−8
10
−7
10
−6
10
−5
Frequency (Hz)
P
o
w
e
r
s
p
e
c
t
r
a
(
H
z
2
)
Speed @ machine # 9
Speed @ machine # 7
Speed @ machine # 1
Figure 6.29 Power spectral distribution of the power and speed of selected machines (mean wind speed at
12m/s to Finland).
The wind power variations excite the power system in several modes. The
electromechanical mode at 0.318Hz is excited in the Finish system (machine #1). The
electromechanical mode around 0.55Hz is also excited. In this case, the wind power excites
the power system with a frequency around 1Hz (3p) that influences several machines
because at this frequency there are some electromechanical modes related to several
machines as presented in the modal analysis.
The power spectral distributions of the speed variations in the machines are very
small. The machine #1 presents a strong variation at 0.318Hz as expected because of its
electromechanical mode of oscillation. All the other machines present higher power
spectral speed variations at the 3p frequency because that is the frequency that the wind
turbine excites the power system. The machine #7 presents a higher speed variation at
1.1Hz because that is the electromechanical mode associated to this machine.
As some important electromechanical modes are near to 0.5 Hz and taking into
account that as the wind turbines grow in size there is a tend to reduce the speed of rotation,
the rotational speed of the AWF is modified to 10 rpm. This slow speed of rotation leads to
a 3p effect around 0.5Hz that will excite less damped modes of the entire power system.
Figure 6.30 presents the power spectral distributions for the same machines similar to
Figure 6.29 at this new condition, i.e. 3p from the wind turbines at low frequency 0.5Hz.
129/152
0 0.5 1 1.5 2 2.5 3
10
−2
10
0
10
2
Frequency (Hz)
P
o
w
e
r
s
p
e
c
t
r
a
(
M
W
2
)
Active power @ machine # 9
Active power @ machine # 7
Active power @ machine # 1
0 0.5 1 1.5 2 2.5 3
10
−9
10
−8
10
−7
10
−6
10
−5
Frequency (Hz)
P
o
w
e
r
s
p
e
c
t
r
a
(
H
z
2
)
Speed @ machine # 9
Speed @ machine # 7
Speed @ machine # 1
Figure 6.30 Power spectral distribution of power and speed to selected machines (AWF modified to lower
rotational speed (3p~0.5Hz)).
In Figure 6.30, the wind power strongly excites the power system controllers. The
power variations on the 3p effect are very similar to the ones presented in Figure 6.29,
however at the frequency of 0.318Hz, the power variations by machine #1 are much higher
than the previous case. In this case, it is possible to verify the isolated excitation of the
machine #7 at 1.1Hz that is related to its electromechanical mode. The power spectral
distributions of the speed variations on the selected machines are much higher particularly
to machine #1 mostly because the proximity to the less damped electromechanical modes.
After having analysed the simulations to high wind power variation by the AWF in
the Finish power system, here the mean wind speed applied to Finland is modified from 12
m/s to 16m/s in order to identify larger modifications on the results. Figure 6.31 presents
the results of the dynamic simulation for case II (i.e. high wind speed to Finland).
130/152
0 100 200 300 400 500 600
3050
3100
3150
3200
time (s)
W
i
n
d
P
o
w
e
r
(
M
W
)
Total wind power to NORDEL
0 100 200 300 400 500 600
49.98
49.99
50
50.01
50.02
time (s)
F
r
e
q
u
e
n
c
y
(
H
z
) Speed of each machine
0 2 4 6 8 10 12 14 16 18 20
0
100
200
300
400
Machine #
S
t
d
o
f
P
o
w
e
r
(
M
W
)
STD of the power of each machine
Figure 6.31 Wind power, frequency and standard deviation of power (average wind speed 16m/s to Finland).
There are no significant differences between Case I and II. The average frequency of
the power system is slightly higher than the Case I meaning that it could be necessary an
adjustment of the set points of the frequency controllers in order to keep the power
production within the country areas. In the case II, the standard deviations of power to each
synchronous machine have slightly changed. This is related to the fact that the speed
governors of the synchronous machines shall react to the power variations from the nearest
wind power avoiding the power flow between different areas of the power system.
6.3.3.2 Voltage Quality
Using the case I wind speed distribution and the original AWF (i.e. 3p frequency
around 1Hz), the voltage on the Nordel network is analysed. Figure 6.32 presents the
voltage simulated on all machines in the Nordel network. The voltages are in p.u. The
differences in mean values come from the voltage set points specified to the machines in
the power system.
131/152
0 100 200 300 400 500 600
3050
3100
3150
3200
time (s)
W
i
n
d
P
o
w
e
r
(
M
W
)
Total wind power to NORDEL
0 100 200 300 400 500 600
0.95
1
1.05
time (s)
V
o
l
t
a
g
e
(
p
.
u
.
)
Voltage for all machines
0 2 4 6 8 10 12 14 16 18 20
0
0.005
0.01
0.015
Machine #
S
t
d
o
f
V
o
l
t
a
g
e
(
%
)
STD of the voltage of each machine
Figure 6.32 Wind power and voltage deviations simulated in Nordel system (mean wind speed 12m/s to
Finland).
The total wind power installed to the power system is the same as presented in the
previous cases. The voltage levels are not easy to identify in the graph, and the middle part
of Figure 6.32 is only presented to show that the variations are very small meaning that the
voltage regulators operate properly. The voltage quality does not seem to be an important
issue here, anyway, similar to the frequency control analysis, the voltage variation to
selected machines with high standard deviation of voltage are detailed as follow (Figure
6.33).
Frequency (Hz)
P
o
w
e
r
s
p
e
c
t
r
a
(
V
2
)
Voltage deviation @ machine #4
Voltage deviation @ machine #7
Voltage deviation @ machine #19
0 100 200 300 400 500 600
−0.04
−0.02
0
0.02
0.04
Time (s)
V
o
l
t
a
g
e
d
e
v
i
a
t
i
o
n
(
%
)
Voltage deviation @ machine #4
Voltage deviation @ machine #7
Voltage deviation @ machine #19
0 0.5 1 1.5 2 2.5 3
10
0
10
2
10
4
Figure 6.33 Power spectral distribution of voltage and voltage deviation to selected machines in the Nordel
(Finland AWF at 12m/s).
132/152
In Figure 6.33, the power spectral distribution and time series of the voltage
deviations to machines #4, 7 and 19 are presented. Those machines were selected because
they presented higher standard deviation of voltage. The power spectral distributions of
voltage show that the high variations occur at the 3p frequency from the wind turbine. It
also shows that the electromechanical modes also influence the voltage controllers. The
excitations of the synchronous machines are intrinsically related to the speed of the
machine hence the presences of the modes were expected. The voltage deviations in time
domain are within ±0.04% of the rated voltage that is very small.
Similarly to the frequency control analysis, the AWF was modified to rotate at 10rpm
(3p=0.5Hz) in order to analyse the influences of the wind power on exciting the modes of
the power system. Figure 6.34 presents the power spectral distribution of the voltage
deviation as well as the time series of the voltage deviations on machines # 4, 7 and 19.
Frequency (Hz)
P
o
w
e
r
s
p
e
c
t
r
a
(
V
2
)
Voltage deviation @ machine #4
Voltage deviation @ machine #7
Voltage deviation @ machine #19
0 100 200 300 400 500 600
−0.05
0
0.05
Time (s)
V
o
l
t
a
g
e
d
e
v
i
a
t
i
o
n
(
%
)
Voltage deviation @ machine #4
Voltage deviation @ machine #7
Voltage deviation @ machine #19
0 0.5 1 1.5 2 2.5 3
10
0
10
2
10
4
Figure 6.34 Power spectra distribution of voltage and voltage deviation simulated in Nordel System (mean
wind speed 12m/s to Finland AWF low frequency (3p=0.5Hz)).
In this case, the power spectral distributions of the voltage are much higher at the 3p
frequency. In the original AWF, the voltage power spectral at 1Hz were around 1000V²,
now that the 3p is around 0.5Hz, the corresponding power spectral peak is near to 10000V².
This is explained because at the 0.5Hz an eigenfrequency of the power system is excited.
The time series of the voltage deviations in this new condition agrees with the power
spectral distribution, where the amplitudes now are limited to ±0.05%. However, it is
important to note that the voltage variations simulated are very small (160V at 400kV) and
it can be influenced by the precision of the simulations. Hence, the last three figures must
be analysed with caution.
6.4 Case Analysis Remarks
Power quality and stability problems from large wind farms to the network have been
investigated here. Three cases were analysed to illustrate the problems on the power quality
and stability when integrating large amount of wind power, namely the voltage stability,
voltage quality and the frequency control were investigated.
133/152
The voltage stability was illustrated by means of loadability curves drawn to the
power system. The studied cases showed that the wind power can contribute to improve the
voltage stability and that the reactive power is an important issue from large wind power
concerning the voltage stability.
The voltage quality has been investigated by means of dynamic simulations using the
Aggregate Wind Farm. The frequency control has been investigated by modal analysis and
dynamic simulation using the AWF model. From the studied cases, the dispersion of the
wind power over extensive areas smoothed the power variations from the wind turbines.
The smoothed power variations caused small than expected voltage and frequency
deviations in the high voltage network.
Following detailed remarks to each case are presented.
6.4.1 Remarks on Case 1
The first case presented only voltage stability analysis. The voltage stability was
analysed with loadability curves to the power system with and without large amount of
wind power.
The conventional wind turbines reduce the reactive power availability in the power
system, fact that is deeply related to the voltage stability. In this power system studied, the
voltage stability was in general reduced with wind power because of the reactive power
demanded by the wind turbine generators. However, new wind turbine technologies using
electronic power converters, which can regulate the reactive power from the wind turbine,
improved the voltage stability and it can actively contribute to regulate the voltage.
The voltage stability depends also on the network characteristics. The case represents
the integration to the transmission level (220kV) where the reactances inductive of the lines
and the reactive power compensation schemes played an important role.
6.4.2 Remarks on Case 2
In the second case, the analysis of voltage stability and dynamic voltage quality of a
Brazilian power system was presented. The voltage stability was studied with similar
methodology as the one used in the case 1, however different load conditions were supplied
improving the analysis.
The voltage stability was addressed in heavy and light load conditions, where the
voltages on the network were modified in each condition due to the reactive power
compensation schemes. The wind power had small impact on the voltage stability, where it
increased the loadability limits of the local power system, fact that is different from the
previous case. In this case, a relative low reactive power is flowing in the network when
compared to case 1 that explains the improvements on the voltage stability.
The dynamic voltage variation has been studied using dynamic simulations with
aggregate wind farm. The smoothing effects on the power fluctuation from the wind farm
cased small voltage fluctuations on the power system. The worstcase scenario showed
voltage variations within 0.3 kV at 230kV nominal voltage. The wind farms will be
installed to the transmission level hence the voltage variations on the low voltage levels
area can be influenced. When distribution grid data is available, the model must be
extended to include the distribution grids and the voltage quality on the consumers must be
assessed.
134/152
6.4.3 Remarks on Case 3
Case 3 analysed largescale integration of wind power influence on the voltage and
frequency controllers. The voltage control was related to the voltage quality and in this
case, only the dynamic voltage variations were presented. The frequency control was
related to dynamic stability and power quality.
The Nordic power system (here called Nordel) is very strong. Aggregate wind farms
(AWF) represented the largescale wind power in the Nordel based on the national plans of
wind power.
Here some assumptions were taken:
• Two cases for each average wind speeds of 8, 10, 12 and 16 m/s were
simulated with turbulence intensity of 20%. The cases covers a large extend of
the power curve from wind turbines in addition, the 20% turbulence intensity
is expected to be the worst scenario to normal operation of wind turbines.
• The wind farms were aggregate in similar blocks of 550MW. This assumption
made ease to change the assigned average wind speeds to different AWFs, i.e.
the time series of the aggregated wind speeds can be shifted to different AWF
s without generating new time series. The application of the AWF to the real
wind farms layout is straightforward when it is available.
• The layout of the wind farms, similar to case 2, assumed 90° wind direction
leading to rows of 10 wind turbines with high correlated turbulence that might
be considered one worstcase scenario.
• The connection points of the AWFs were assumed to buses in Sweden and
Finland, where in Norway, EFI Sintef, from Norway, proposed specific buses
to which the large scale wind power should be simulated.
• Except for the AWFs installed in Finland and Norway, the average wind
speeds were assigned arbitrarily. The wind speeds in Finland and Norway
were chose with high wind power variations because in some specific
machines installed in these countries, which the wind power is planned,
presented high participation on electromechanical modes at low frequency.
The Finnish system presented the less damped mode so the high wind power
variations were assigned to both buses in Finland (12m/s) and because of the
same reason 10m/s was assigned to Norway. However, a sensitive analyse
was carried, where the high wind power variation (12m/s) in Finland was
replaced with the smallest at mean wind speed 16 m/s.
• The automatic secondary generation control unit was not implemented.
Hence, the set points of the nearest synchronous machines were adjusted to
maintain the areas (countries) balances.
The main conclusions from the simulations are classified on voltage and frequency
control as following.
6.4.3.1 Frequency control
The frequency control was primarily analysed with modal techniques. Using modal
techniques, some of the most significant electromechanical modes were identified and
135/152
characterized. The less damped modes were below 0.6Hz and all electromechanical modes
were stable. With modal analysis was possible to characterize also the most important
machines participating on each electromechanical mode and how was the direction of the
relative rotor angle speed variation of the machines.
The AWF completes the analysis because it provided means of dynamic simulation to
verify the behaviour of the frequency controllers (speed governors) under normal operation.
From the dynamic simulations, the electromechanical modes were also identified. They are
well damped (frequency controllers well designed) so the power oscillations caused from
the wind power in the Nordel did not increase in time.
One of the machines that presented high participation on the electromechanical
modes from the modal analysis was the machine #9 connected to bus SVN2__ that was
also identified on the dynamic simulation with higher speed and power variations during
normal operation with wind power. This bus did not have wind turbines installed, but its
specific location in addition to the electromechanical characteristics of the power system
made this machine the most sensible in the simulations. This observation might indicate
that the power variations from the wind power do not necessarily influence only locally but
it can be transferred to remote parts of the power system.
In order to verify the extreme impacts of wind power in the less damped
electromechanical modes of the power system, the aggregate wind turbine was modified to
have the 3p effect at 0.5Hz. Under this new condition the power system still stable,
however the wind turbines directly excite the less damped electromechanical modes hence
higher power and speed variations at low frequencies were encountered, particularly on the
machines with high participation on the electromechanical modes that was expected.
6.4.3.2 Voltage controllers
The voltage controller’s performances were analysed by dynamic simulation using
the AWF. The voltage controllers worked properly compensating the reactive power
demanded from the AWF and keeping the voltage within 0.10% variations (±0.05%).
The wind power simulated did not represent a problem to the voltage quality.
However, the dynamic simulation revealed that some modes that were excited from the
wind power on the frequency analyses were also present on the voltages. Anyway, the
voltage variations were small.
The dynamic analyses of largescale wind power indicated that the voltage quality
was not a concern on high voltage networks and it might be more relevant in lower voltage
levels including the distribution networks. At those grids, the limited voltage regulation and
the losses might lead to large voltage variations from wind power and become a problem
so, when necessary, the grid must be detailed and the voltage analysed properly.
136/152
Chapter 7
7 Conclusions
A methodology to investigate the wind power influences on the power system was
presented in this thesis. It includes analysis of the wind power influences on the voltage
stability, power system stability and power quality characteristics.
The voltage stability was analysed with loadability curves to the power system. The
voltage stability was influenced by the wind power integration, where the reactive power
was the main factor. A loadability computation tool was developed in this thesis and a static
model to the wind power on the voltage stability was presented. Modifications of the wind
turbine characteristics, i.e. application of power electronics, were simulated improving the
voltage stability.
The main conclusions for the voltage stability are that although the wind power
alleviates the active power fluxes in the network, the reactive power flux to the wind farms
will reduce the voltage stability limits. Wind turbine technologies with power converters
that can actively control the reactive power consumption increased the voltage stability (i.e.
extended the power limit of the voltage collapse) of the power system.
The power system stability and power quality were investigated with dynamic
simulations. Dynamic models to largescale wind power that can assess the power system
stability and quality were presented. The dynamic models are based on aggregation
procedures of wind turbines.
The aggregate wind farm model permits an easy assessment of the power system
dynamic operation from largescale integration of wind power. The assessment of power
quality parameters such as voltage and frequency deviations due to the inclusion of large
amount of wind power demands simulations of the wind farms power production and its
interaction with the power systems. Dynamic simulations of a high number of wind
turbines are unpractical for very large power systems. Therefore, the aggregate model is
very suitable to identify the impact of largescale wind power in the dynamic operation of
the power system because it is time inexpensive and it includes the most relevant
characteristics of the wind turbines power dynamics, particularly the turbulence and its
spatial correlation. Using the aggregate models, the identification of voltage deviations and
inadequate frequency regulation can be done in few minutes even for large power systems.
The aggregate wind speed that is applied to the aggregate wind turbine model was
also developed and presented. It includes relevant characteristics of the turbulence and its
spatial correlation effects on the power produced from wind farms. The power smoothing
effect from the wind farms power production have been verified in the aggregate wind
speed model to correctly represent the large wind farms power production.
In addition, the problems for the power systems stability that are mainly related to
power oscillations were studied with modal analysis to identify the most relevant
characteristics. The applications of the modal analysis with the dynamic simulations are
recommended in order to assess possible power system stability problems.
Related to the power quality problems, the studied cases show that higher number of
wind turbines smoothes the power variations consequently the flicker phenomena can be
137/152
considerably reduced. The simulated dynamic voltages were within limits and the flicker
was not a concern however in the cases studied here, the network was strong and the wind
power was connected to the high voltage network (transmission lines).
The frequency control was also investigated using a reduced model of the Nordel
power system. The largescale wind power integration interacts with the power system
controllers. The power variation in broad frequency range excite some electromechanical
modes of the power system however the frequencies variations were so low that it does not
endanger the dynamic stability or cause serious power oscillations in the power system.
Nevertheless, the dynamic simulations showed that the wind power do not impact only
local machines but it can influence remote parts because of the power system
electromechanical characteristics hence it is necessary to investigate the entire power
system operation with largescale integration of wind power.
Before introducing the aggregate wind farm, dynamic wind turbine models have been
presented. The dynamic wind turbine model represents the most relevant wind speed and
wind turbine characteristics to power quality characteristics. The frozen turbulence,
rotational sampling turbulence and tower shadow effects are the most relevant to power
quality hence included in the wind speed model. The wind turbine model described the
most relevant aeroelastic components of wind turbines for simulation of power fluctuations.
The aerodynamic power conversion including the dynamic stall is included in the
aeroelastic model. In addition, the drive train dynamic operation is simplified to include the
most relevant modes of oscillation of the power variations.
The aggregate wind farm presented some limitations that can be improved in future
works. The inclusion of the wake effect phenomenon in the wind farm and the deterministic
parts from the wind speed can be new tasks. Moreover, the extension of the model to
variable speed wind turbines is also envisaged. In the near future, the variable wind turbines
will account to large power production.
In addition, as future work, here the integration of wind speed forecasting in
extensive areas with the aggregate wind speed model is proposed. The inclusion of this
wind speed forecasting and geographical disposition of wind farms improves analyses of
the dynamic integration of the largescale wind power in the power systems. The forecast
model can generate the mean wind speed (10 minutes) and turbulence intensity to specific
locations that can be used as input to the aggregate wind speed model. The integration of
different tools will improve the analysis of the power system characteristics with benefits to
the analysis of the frequency secondary control in the power system and analysis of real
conditions as well as storm fronts.
The inclusion of the loads characteristics and its time variation to the analysis of the
dynamic power system operation it also proposed here as future work. The interaction of
the loads in addition to the wind power variation on the power system dynamic operation
can be relevant to characterize the power oscillations.
Finally, it is important to validate and improve the aggregate wind farm model by
obtaining experimental measurements from large wind farms.
138/152
8 Reference List
[1] IEC 6140021 Measurement and assessment of power quality characteristics of grid
connected wind turbines. International Electrotechnical Commission, FDIS 2001.
[2] “Integration of Renewable Energy in the Electrical Network in 2010”, draft final
report, EU ALTENER PROJECT XVII/4.1030/Z/99115, BrusselsBelgium, 2001.
[3] YihHuei Wan and Brian K. Parsons, “Factors Relevant to Utility Integration of
Intermittent Renewable Technologies”, NREL Report NREL/TP4634953,
Colorado, USA, August, 1993.
[4] John Eli Nielsen, Søren Varming and Christian Gaardesatrup, “Review of Technical
Options and Constraints for Integration of Distributed Generation in Electricity
Networks”, Report SUSTELNET, (online: www.sustelnet.net/documents.html)
Netherlands, December, 2002.
[5] Jørn Saxow, Svend Skovgaard, Peggy Friis, “Power Plant Operation Correlated
with wind energy production within the Elsam Utility Area – status 1995”,
Elsamprojekt A/S Power Station Engineering Report EP96/404a, Denmark, June,
1996.
[6] John Olave Tande, Per Nørgaard, Poul Sørensen, Lars Søndergaard, Preben
Jørgensen, Anders Vikkelsø, Jacob Dyring Kledal, Jørgen S. Christensen, “Power
Quality and grid Connection of Wind Turbines”, Summary report, Vol. 1, 2 and 3,
Risø report RisøR853, Roskilde, Denmark, October, 1996.
[7] Poul Sørensen, Troels Friis Pedersen, Gehart Gerdes, Rainer Klosse, Niel
Robertson, Wily Davy, Maria Koulouvari, Evangelis Morfiadakis and Åke Larsson,
“European Wind Turbine Testing Procedure Developments, Task 2: Power
Quality”, Risø report Risø R1093(EN) (online: www.risoe.dk/rispubl/VEA/
veapdf/risr1093.pdf), Roskilde, Denmark, May 2001.
[8] Åke Larsson, “The Power Quality of Wind Turbines”, PhD dissertation (online:
www.elkraft.chalmers.se/Publikationer/EMKE.publ/Abstracts/LarssonD.abstr.html),
Department of Electrical Power Engineering, Chalmers University of Technology,
Göteborg, Sweden, 2000.
[9] John Olave Giæver Tande, “Applying Power Quality Characteristics of Wind
Turbines for Assessing Impact on Voltage Quality”, In Wind energy, 5, 2002.
[10] Vladislav Akhmatov, Arne Hejde Nielsen, Hans Knudsen, “Electromechanical
interaction and stability of power grids with windmills”, in proceedings of the
IASTED International Conference Power and Energy Systems, Marbela , Spain,
September, 2000.
139/152
[11] Marianne Bruntt, Jan Havsager, Hans Knudsen, “Incorporation of Wind Power in
the East Danish Power System”, in proceedings of IEEE Power Tech’99 conference,
Budapest, Hungary, September, 1999.
[12] Jan Wiik, Jan Ove Gjerde and Terje Gjengedal, “Impacts from Large Scale
Integration of Wind Farms into Weak Power Systems”, in Proceedings of
PowerCon 2000, Power System Technology, Australia, 2000.
[13] L.L. Freris, “Wind Energy Conversion Systems”. Prentice Hall, London, UK, 1990.
[14] Erich Hau, WindTurbines Fundamentals, Techonlogies, Application, Economics,
Spring Verlag, Germany, 2000.
[15] Siegfried Heier, Grid Integration of Wind Energy Conversion Systems, John Willey
and Sons, England, 1998.
[16] J. Wilkie, W.E. Leithead and C. Anderson, “Modelling of Wind Turbines by Simple
Models”, Wind engineering vol. 14, No. 4, 1990.
[17] Estanqueiro, A. I. “Dynamic Models of Wind Parks” (in Portuguese: Modelação
Dinamica de Parques Eólicos), PhD dissertation, Instituto Superior Técnico,
Universidade Técnica de Lisboa, Portugal, 1997.
[18] Thomas Petru, “Modelling of Wind Turbines for Power System Studies”, Licentiate
technical report Nr. 391L (online: www.elkraft.chalmers.se/Publikationer/
EMKE.publ/Abstracts/Petru2001.abstr.html), Department of Electric Power
Engineering, Chalmers University of Technology, Sweden, 2001.
[19] Poul Sørensen, Anca Hansen, Lorand Janosi, Jonh Bech and Birgite BakJensen,
“Simulation of interaction between wind farm and power system”. RISØ report R
1281(EN) (online: www.risoe.dk/rispubl/VEA/veapdf/risr1281.pdf), Roskilde,
Denmark, December 2001.
[20] Ana I. Estanqueiro, R. Aguiar, J. A. Gil Saraiva, Rui M G. Castro and J. M Ferreira
de Jesus, “Development and Application of a model for power output fluctuations in
a wind farm”, in proceedings of the European Wind Energy Conference, Lübeck
Travemund, Germany, March, 1993.
[21] Gregor Giebel, “On the Benefits of Distributed Generation of Wind Energy in
Europe”, PhD dissertation (online: gregie.hjem.wanadoo.dk/Ggiebel_Distributed
WindEnergyInEurope.pdf), Reihe 6, report Nr. 444,VDI Verlag, Düsseldorf,
Germany, 2001.
[22] H.G. Beyer, J. Luther, R. SteinbergerWillms, “Fluctuations in the Combined Power
Output from Geografically Distributed Grid Coupled Wind Energy Conversion
Systems – An Analysis in the Frequency Domain”, Wind engineering Vol 14, No. 3,
1990.
140/152
[23] Prabha Kundur Ed, “Power System Stability and Control”, EPRI Power
Engineering Series, McGraw Hill, 1993.
[24] Paul M. Anderson, A. A. Fouad, “Power System Control and Stability”, IEEE
Power System Engineering Series, New York, 1993.
[25] System dynamic performance subcommittee of the Power system engineering
committee of the Power Engineering Society “Eigenanalysis and Frequency Domain
Methods for Power Systems: Dynamic Performance”, IEEE 90TH02923PWR.
IEEE, Piscataway, NJ, USA, 1989.
[26] Lei Wang, Meir Klein, Solomon Yirga and Prabha Kundur, “Dynamic Reduction of
Large Power System for Stability Studies”, IEEE transactions on Power System,
vol. 12 No. 2, May, 1997.
[27] Carson W. Taylor, “Power System Voltage Stability”, EPRI Power Engineering
Series, McGraw Hill, 1994.
[28] Thierry Van Custem, Costas Vournas, “Voltage Stability of Electric Power System”,
Kluvwer’s Power Electronic and Power Systems Series. Series Editor: Thomas A
Lipo and M.A. Pai, Brussels 1998.
[29] System dynamic performance subcommittee of the Power system engineering
committee of the Power Engineering Society, “Voltage Stability of Power Systems:
Concepts, Analytical tools, and industry experience”, IEEE 90TH03582PWR,
Piscataway, NJ, USA, 1990.
[30] Claudio Cañizares, “Voltage Stability Assessment: Concepts, Practices and Tools”
Power System Stability Subcommittee Special Publication IEEE/PES, Final
Document (thunderbox.uwaterloo.ca/~claudio/claudio.html#VSWG), August 2002.
[31] S. Persaud, B. Fox and D. Flynn, Impact of remotely connected wind turbines on
steady state operation of radial distribution networks, in Proceedings of Generation
Transmission and Distribution, Vol. 147, N3, May, 2000.
[32] IEC 8680 Flickermeter Functional and design specifications CEI/IEC 8680, 1986;
[33] IEC 8680 Amendment 1 to Flickermeter Evaluation of flicker severity. CEI/IEC
8680, 1990;
[34] IEC 8680 Flickermeter Evaluation of flicker severity. CEI/IEC 8680, 1991;
[35] Johansson, S., "Longterm Voltage Stability in Power Systems  Alleviating the
Impact of Generator Current Limiters". Technical Report No. 335, Chalmers
University of Technology, Göteborg, Sweden. 1998.
141/152
[36] Karlsson, D., "Voltage stability simulations using detailed models based on field
measurements". Technical Report No. 230, Chalmers University of Technology,
Göteborg, Sweden, 1992.
[37] Mats Larsson, “Coordinated Voltage Control in Electric Power Systems”, Ph.D.
thesis (www.iea.lth.se/publications/point/Theses/pdf_files/LTHIEA1025.pdf),
Department of Industrial Electrical Engineering and Automation, Lund Institute of
Technology, Lund, Sweden, 2001.
[38] Bjørn Harald Bakken, “Technical and Economical Aspects of operation of thermal
and hydro power systems”, Ph.D. Dissertation, NTNU, report Nr. 16, Trondheim,
Norway, 1997.
[39] Graham Rogers, “Power System Oscillations”, Kluvwer’s Power Electronic and
Power Systems Series. Series Editor: M.A. Pai, Canada, 2000.
[40] K. R. Padiyar, “Analysis of Subsynchronous Resonance in Power System”,
Kluvwer’s Power Electronic and Power Systems Series. Series Editor: M.A. Pai,
India, 1999.
[41] Olof Samuelsson. “Power System Damping  Structural Aspects of Controlling
Active Power”, Ph.D. dissertation (www.iea.lth.se/~ielolof/olof/PhDthesis.pdf),
Department of Industrial Electrical Engineering and Automation, Lund Institute of
Technology, Lund, Sweden, 1997.
[42] Bo Eliasson. “Damping of Power Oscillations in Large Power Systems”, PhD
dissertation, Department of Automatic Control, Lund Institute of Technology, Lund,
Sweden, 1990.
[43] Sundar, R. M. and Sulivan, J. P., “Performance of Wind Turbines in a Turbulent
Atmosphere”. Journal of Solar Energy, Vol. 31, No. 6 pp. 567575, 1983.
[44] Kaimal, J. C., J. C. Wyngaard, Y. Izumi, and O. R. Cote, “Spectral Characteristics
of Surface Layer Turbulence”. Quarterly Journal of the Royal Meteorological
Society, vol. 98, 1972.
[45] Jorgen Thirstrup Petersen and Helge Aagaard Madsen, “Local Inflow and Dynamics
– measured and simulated on a rotating Wind Turbine Blade”, Risø report, DRAFT,
RisøR993 (EN), Roskilde, Denmark September 1997.
[46] “Voltage characteristics of electricity supplied by public systems”, European
standard, CENELEC EN 50160, November, 1994.
[47] Winkelaar, D., “Fast threedimensional wind simulation of stochastic blade loads”.
Paper presented at the 10
th
ASME Wind Energy Symposium, Houston, 2024
January 1991.
142/152
[48] Madsen and Rasmussen, “Rotor loading on a threebladed wind turbine”, in
proceedings of the European Wind Energy Conference EWEC’89, Glasgow 1989.
[49] Poul Sørensen, “Frequency Domain Modelling of Wind Turbine Structures”. Risø
publication RisøR749(EN), Roskilde, Denmark, April 1994.
[50] Wiebke Lagrander, “Models for Variable Speed Wind Turbines”, M.Sc.
dissertation, CREST Loughborough University, UK, 1996.
[51] Madsen, Peter Hauge, Sten Fradsen, William E. Holley and Jens Carsten Hansen,
“Dynamics and Fatigue Damage of Wind Turbine Rotors during Steady Operation”.
Risø report R512, Risø National Laboratory, Roskilde, Denmark 1984.
[52] Panofsky, H. A., “The atmospheric boundary layer below 150 metres”. Annual
Review of Fluids Mechanics, vol 6 pp.147177, 1974.
[53] Engineering Science Data Unit, Characteristics of wind speed in the lower layers of
the atmosphere near the ground: Strong wind (neutral atmosphere). Data Item
72026, ESDU London, 1972.
[54] Pedro Rosas, Poul Sørensen, Henrik Bindner, “Fast Wind Modelling”, in
Proceedings of European Wind Energy Conference, Kassel, Germany, 2000;
[55] Solari, G., “Turbulence Modelling for Gust Loading”. ASCE Journal of Structural
Engineering, Vol. 113, No. 7, July 1987.
[56] “last og sikkerhed for vindmøllekonstruktioner”) (in English: Danish standard for
loads and safety of wind turbines construction, 1
st
edition may 1992, Danish
Standard DS 472.
[57] Davenport, A. G., “The spectrum of horizontal gustiness near the ground in high
winds”. Quarterly Journal of Meteorology Society 87, pgs 194 – 211.
[58] W. Schelz and D. Infield, “Horizontal two points coherence for separations greater
than the measurement height”. BoundaryLayer Meteorology 87,Kluwer Academic
Publishers Netherlands, 1998.
[59] Leif Kristensen and Sten Frandsen, “Model for power spectra of the blade of a wind
turbine measured from the moving frame of reference”, Journal of Wind
Engineering and Industrial Aerodynamics, 10 (1982) 249262;
[60] Poul Sørensen, Methods for Calculation of the Flicker Contributions from Wind
Turbines, Risø report Risø – I – 939 (EN), Roskilde, Denmark, December 1995.
[61] Gunner C. Larsen, Sten Frandsen, Poul Sørensen and Michael S. Courtney, “Design
Basis for Horizontal Axis Turbines – Theoretical Background”. Risø publication
RisøM2836, Roskilde, Denmark, December 1989.
143/152
[62] Stig Øye, “Dynamic Stall – simulated as time lag of separation” in Proceedings of
the 4
th
IEA Symposium on the aerodynamics of wind turbines. Mc Anulty (Ed.),
Rome Italy 1991.
[63] Henrik Bindner, “Active Control: Wind Turbine Model”. Report number: R
920(EN) Risø National Laboratory, Denmark, 1999;
[64] Hans Ganader, Torbjörb Thiringer, “An extended model of drive train dynamics”.
Report number 6R, Department of Electric Power Engineering, Chalmers University
of Technology, Sweden, 1997.
[65] Paul C. Krause, Oleg Wasynczuk and Scott D. Sundhoff. “Analysis of Electric
Machinery and Drive Systems”, IEEE Press, WilleyInterscience, second edition,
2002;
[66] Torben J. Larsen, Morten Hartvig Hansen, “Generator dynamics – analysis and
simulations”, Risø report, RisøR1395(EN), Roskilde, Denmark, February, 2003.
[67] Poul Sørensen, Anca Hansen, Pedro Rosas, Wind models for prediction of power
fluctuations from wind farms. Journal of Wind Engineering, n. 89 October 2001.
[68] A.I. Estanqueiro, R. F. Aguiar, J. A. G. Saraiva, R. M. G. Castro and J. M. F. D.
Jesus. “On the effect of utility grid characteristics on wind park power output
fluctuations”, in Proceedings of British Wind Energy Conference BWEA’15, York
1993;
[69] T. Y. J lem, R.T.H Alden, “Comparison of experimental and aggregate induction
motor responses”. IEEE Transactions on Power Systems Vol. 9, n. 4, November
1994, pgs:18951900;
[70] M. Akbaba, S. Q. Fakhro, “New model for single unit representation of induction
motor loads, including skin effect, for power system transient stability studies”. IEE
Proceedings – B Vol. 139, n. 6, November 1992;
[71] D.C. Franklin, A. Morelato, “Improving dynamic aggregation of induction motor
models”. IEEE Transactions on Power Systems Vol. 9, n. 4, November 1994, pg:
19341941;
[72] DU_Rede Oper.NNE.R20_261202, North – Northeast Power System Diagram –
from December 2002 (in Portuguese). Power System National Operator (Operador
Nacional do Sistema) – online: www.ons.com.br/ons/sin/index.htm, Brasilia,
Brazil, 2000.
[73] Approved wind projects to Brazil on line www.eolica.com.br/index_por.html,
Recife, Brazil, 2002.
[74] Data base – Power system analysis in Portuguese (Base de Dados  Estudos
Elétricos), reference case to the year 2004. National Operator of the Power System
144/152
ONS: 06/09/01 CICLO 0204, online: www.ons.com.br/ons/sin/index.htm,
Brasilia, Brazil, 2002.
[75] “Nordic High Voltage Network” (“Det Nordiske Høyspentnette”) on line diagram
at www.nordel.org/Content/Default.asp?PageID=125, 2002.
[76] “Recommendations for frequency, time deviation, regulation and reserve”
(Rekommandasjon for frekvens, tidsavvik, reguérstyrke og reserve) NORDEL
available online at www.nordel.org/Content/Default.asp?PageID=130, 1996.
[77] “Operational Performance Specifications for Thermal Power Units larger than
100MW” NORDEL available online: www.nordel.org, 1995.
[78] Torbjörn Thiringer and Jorma Luomi, Comparison of reduced order dynamic
models of induction machines, IEEE transactions on Power Systems, vol 16, No 1,
February, 2001.
[79] Torbjörb Thiringer, “Measurements and Modelling of LowFrequency Disturbances
in Induction Machines”. Technical Report no.: 293, PhD dissertation, Department of
Electric Power Engineering, Chalmers University of Technology, Sweden, 1996.
[80] Communication from ABB/SIMPOW Induction Machine Model – September 2002;
[81] Rosas, P. A. C., “Wind Power Influences on the Voltage Stability”, in proceedings
2001 European Wind Energy Conference, Copenhagen, 2001.
145/152
146/152
9 Annexes
9.1 Electrical Components Model in SIMPOW
The SIMPOW has been developed since 1977 in ABB. It is a dynamic simulation
program dedicated to the power system problem. It is divided into 5 programs:
• OPTPOW – solves the network equations (static solution) using iterative
method;
• STATPOW – solves the short circuit equations;
• DYNPOW – solves the differential equations to the power system in time;
• DSL – Dynamic Simulation Language;
• DYNPOST – plots the outputs;
To power quality assessment from wind turbines the OPTPOW, DYNPOW, DSL and
DYNPOST are used. Any dynamic simulation demands at least the use of the upper four
modules.
The OPTPOW computes the load flow in the power system. The wind turbine model
in this program is an asynchronous machine, which the electrical characteristics of the
machine and the active power produced are specified. The program automatically computes
the voltage in all nodes, the power fluxes, and the mechanical speed and torque in the
electrical machine.
The DYNPOW solves the differential equations to the power system in the dynamic
simulation. In the DYNPOW module, the dynamic models of the electrical machinery and
controllers of the machines are specified.
The DSL module is the dynamic simulation language. In the DSL, we implement the
mechanical model of the wind turbine. The interface of the DSL and the DYNPOW
modules is the mechanical torque. The wind turbine model is mostly done in the DSL
module where the DYNPOW send the speed of the machine and the DSL return the torque
on the main shaft of the electrical generator.
The DYNPOST module provides the graphical output of the simulation. Using
DYPOST, it is possible to either analyse the electrical characteristics or export to an ASCII
file that can be analyse in any other program.
The electrical components comprise the electrical generator, the reactive power
compensation, control system, stepup transformer, transmission lines, and the entire power
system. The transmission lines formally does not belong to the wind turbine unity but it is
an important component in wind farm simulations therefore it is included here.
The electrical components are modelled using the standard library of SIMPOW/ABB.
The SIMPOW library has verified models to most of the conventional electrical
components in the power system including electrical machines, transmission lines,
transformers and power electronics and controllers.
The electrical components in a conventional wind turbine are:
147/152
• Asynchronous generators,
• Stepup transformers,
• Lines (in a wind farm) and,
• Reactive power compensation
• Power system representation
Details of the electrical power system including the wind farm depend on the purpose
of the simulation.
9.1.1 Electrical generator
The asynchronous generator is the most common type of electrical machine. It is
mostly applied as motor. As generator, however, it has been used for many years in wind
turbines.
The dynamic asynchronous generator model was implemented using the SIMPOW
library. Figure 9.1 presents the main model parameters to the generator.
Electrical parameters: R, X
Current (p.u.)
Electrical Generator
Model
Voltage and frequency
(p.u.)
Air gap Torque
(p.u.)
Speed (p.u.)
Figure 9.1. Electrical generator model parameters.
The inputs to the model are the torque that is computed from the aeroelastic module
(in DSL), the voltage and the frequency that comes from the next electrical component. The
outputs are the mechanical speed (feed in the aeroelastic and wind model) and the current
(feed in the next electrical component). The electrical parameters derive from the
equivalent circuit presented in Figure 9.2.
R
1
R'
2
/s
X
1
X'
2
X
m
R
m
V
1
I
1
I
2
Figure 9.2. Electrical generator equivalent.
Where R
1
is the stator resistance the X
1
is the stator reactance inductive; X’
2
is the
rotor equivalent reactance inductive referred to the stator, R’
2
is the rotor equivalent
resistance referred to the stator; R
m
is the resistance representing losses and X
m
is the
magnetization reactance. The rotor resistance is modelled as a function of the rotor speed in
148/152
order to model the skin effect in the rotor. The saturation can be also stated. All units are in
per unit and the voltages and currents are vectors with an angle and amplitude.
The model itself is a socalled park model in d and q rotating reference frames where
the transients of the stator are neglected. This model is also called as a third order model
reduced from a fifth order. The third order model to induction machines represents
reasonable the power quality characteristics [78] and [79].
The SIMPOW converts all parameters to a direct – d – and a quadrature – q – rotating
axis so called park transformation. The dq reference frame is rotating at synchronous speed
as determined by the electrical angular frequency of the impressed stator voltage, i.e.
synchronously rotating axis [65] and [80].
9.1.2 Stepup Transformer
In these studies, a simple a symmetrical 3phase without saturation transformer model
is used. The configuration of the transformer depends on the installation. The
characteristics and parameters of the model used are presented in Figure 9.3.
High Voltage side
Figure 9.3 Structure of the transformer model.
The inputs to the model are the current from one side, the voltage from the other side
and the reference frequency to the rotating dqframe. Here, the wind turbine is connected to
the low voltage side (stepup transformer), the current inputted comes from the wind
turbine generator and from the reactive power compensation unity on the wind turbine
terminals.
The model outputs the current in the high voltage side and the voltage on the low
voltage side.
The transformer model uses the electrical parameters from a “T” electrical equivalent
transformer model (see Figure 9.4).
V
2
V
1
R'
2 X'
2
I
2
I
1
R
2
X
2
R
m
X
m
Ideal Transformer
Figure 9.4 Electrical transformer model.
Where R
1
is the primary resistance the X
1
is the primary reactance inductive; X’
2
is the
secondary equivalent reactance inductive referred to the primary side, R
2
is the secondary
Low Voltage side
Electrical parameters: R,
X
Stepup Transformer
Model
Current (p.u.)
Voltage and frequency (p.u.)
Voltage (p.u.) Current (p.u.)
149/152
side equivalent resistance referred to the primary side; R
m
is the resistance representing
losses and X
m
is the magnetization reactance.
9.1.3 Reactive Power Compensation.
A switched capacitor bank usually supplies the reactive power in the wind turbines. A
small capacitor bank supplies the noload reactive power to the asynchronous generator
and, during normal operation, small steps of capacitors can be connected to follow the
reactive power demanded from the generator. Other technologies uses power electronics to
compensate the reactive power, but here only the noload capacitor bank supplies the
reactive power during continuous operation.
9.1.4 Lines and Cables
Lines and cables can be modelled in different ways. Here, a concentrated π models
the electrical behaviour of the lines and cables where depending on the distances and type
of cables/lines it can be modified. Figure 9.5 presents the main structure of the basic
cable/lines models.
Electrical parameters: R, X, B, d
Receiving node
Sending node
Current (p.u.)
Line Model
Voltage (p.u.)
Current (p.u.)
Voltage (p.u.)
Figure 9.5 Electrical Transmission structure model.
The model inputs are the one voltage and one current and it outputs the other voltage
and current. The frequency reference is used to transform all variable to the dqrotating
frame and to compute the reactance in the lines. The electrical parameters come from the π
electrical equivalent for the lines (see Figure 9.6)
R X
V
2
I
2
I
1
V
1
B
c
/2 B
c
/2
Figure 9.6 Electrical transmission model.
Where R
is the equivalent resistance and the X is the equivalent reactance inductive of
the line; B
C
is the equivalent susceptance capacitive.
9.1.5 Slack bus
The slack bus is also reefed as reference node. It is modelled as a synchronous
machine or by an infinite node connected through the short circuit impedance. Figure 9.8
150/152
presents the use of an infinite node to replace the power system with the short circuit
impedance at the PCC.
Electrical short circuit parameters: R, X
Reference
angle (Deg.)
Voltage (p.u.)
Reference node Model
Current (p.u.)
Voltage (p.u.)
Figure 9.7 Slack bus structure.
The inputs to the model are: the voltage, the reference angle, and the currents
computed from the other models. The electrical characteristics are the equivalent
impedance at the PCC also called short circuit impedance. The model outputs the voltage at
the PCC. A representation of the positive sequence of the slack bus is presented in Figure
9.8.
R
sc
X
sc
V
1
I
1
V
0
 Constant
Infinite Node
Figure 9.8 Slack bus model (positive sequence).
where R
sc
is the short circuit resistance, the X
sc
is the short circuit reactance inductive;
V
0
is the voltage reference in the infinite node that is fixed.
151/152
152/152
ISBN : 8791184169
ISBN : 8791184169
DYNAMIC INFLUENCES OF WIND POWER ON THE POWER SYSTEM
By Pedro Rosas
Thesis submitted to Ørsted Institute, Section of Electric Power Engineering Technical University of Denmark In partial fulfilment of the requirements for the degree of Doctor of Philosophy Technical Report RISØ R1408 ISBN: 8791184169
Ørsted Institute, Risø National Laboratory & Brazilian Wind Energy Centre Denmark, March 2003
iii
iv
DYNAMIC INFLUENCES OF WIND POWER ON THE POWER SYSTEM Pedro André Carvalho Rosas Risø National Laboratory, Wind Energy Department & Ørsted Institute – Section of Electric Power Engineering, Technical University of Denmark
Abstract
The thesis first presents the basics influences of wind power on the power system stability and quality by pointing out the main power quality issues of wind power in a smallscale case and following, the expected largescale problems are introduced. Secondly, a dynamic wind turbine model that supports power quality assessment of wind turbines is presented. Thirdly, an aggregate wind farm model that support power quality and stability analysis from large wind farms is presented. The aggregate wind farm model includes the smoothing of the relative power fluctuation from a wind farm compared to a single wind turbine. Finally, applications of the aggregate wind farm model to the power systems are presented. The power quality and stability characteristics influenced by largescale wind power are illustrated with three cases. In this thesis, special emphasis has been given to appropriate models to represent the wind acting on wind farms. The wind speed model to a single wind turbine includes turbulence and tower shadow effects from the wind and the rotational sampling turbulence due to the rotation of the blades. In a park scale, the wind speed model to the wind farm includes the spatial coherence between different wind turbines. Here the wind speed model is applied to a constant rotational speed wind turbine/farm, but the model is suitable to variable speed wind turbine/farm as well. The cases presented here illustrate the influences of the wind power on the power system quality and stability. The flicker and frequency deviations are the main power quality parameters presented. The power system stability concentrates on the voltage stability and on the power system oscillations. From the cases studied, voltage and the frequency variations were smaller than expected from the largescale wind power integration due to the low spatial correlation of the wind speed. The voltage quality analysed in a Brazilian power system and in the Nordel power system from connecting large amount of wind power showed very small voltage variations. The frequency variations analysed from the Nordel showed also small variations in the frequency but it also showed that the wind turbines excites the power system in the electromechanical modes. Concerning the stability analysis, the study cases showed that largescale wind power modifies the voltage stability of the power system and can cause power oscillations. It is showed here that the reactive power from the wind farms is the key factor on the voltage stability problem. During continuous operation, the distributed wind power variations did not give any problems to the power system stability concerning the power oscillations.
v
DYNAMIC INFLUENCES OF WIND POWER ON THE POWER SYSTEM Pedro André Carvalho Rosas Risø National Laboratory, Wind Energy Department & Ørsted Institute – Section of Electric Power Engineering, Technical University of Denmark
Resumé
Først i afhandlingen introduceres de vigtigste elkvalitetsproblemstillinger, elsystemets stabilitet og elkvalitet, i relation til vindkraft, når vindmøller tilsluttes i mindre og stor skala. Efterfølgende præsenteres en dynamisk vindmøllemodel, der er specielt egnet til analyse af elkvalitet fra en vindmølle. Herefter introduceres en aggregeret vindmølleparkmodel til brug ved analyse af elkvalitet og netstabilitet fra store vindmølleparker. Den aggregerede vindmølleparkmodel indeholder den udglatning af de relative effekfluktuationer fra en vindmøllepark sammenlignet med en enkelt vindmølle. Endelig præsenteres nogle anvendelser af den aggregerede vindmølleparkmodel i forskellige cases. I afhandlingen er der lagt specielt vægt på at opstille relevante og anvendelige modeller af vinden i en vindmøllepark. Modellen for vindhastigheden for enkelt vindmølle inkluderer turbulens og tårnskygge herunder især roterende sampling af turbulensen som fremkommer pga. vingernes rotation. I parkskala modellen er der taget hensyn til den rummelige koherens mellem de forskellige vindmøller. Vindhastighedsmodellen er i afhandlingen anvendt på vindmøller med konstant omløbstal, men den kan også umiddelbart anvendes på møller med variabelt omløbstal. De cases, der præsenteres, illustrerer indflydelsen af vindmølleparker på elsystemets stabilitet og på elkvaliteten. Flicker og frekvensafvigelser er de vigtigste elkvalitetsparameter, der benyttes til vurderingen af indflydelsen. Stabiliteten af elsystemet vurderes vha. spændingsstabilitet og effekt og frekvenssvingninger. Resultaterne fra de forskellige cases viser at spændings og frekvensvariationerne er mindre end man kunne forvente på grund af den lille rumlige korrelation af vindhastighederne ved de forskellige vindmøller. Undersøgelsen af spændingskvaliteten i den brasilianske case og og i Nordelcasen viser at der kan forventes små spændingsvariationer selv når der tilsluttes meget vindkraft. Undersøgelsen af frekvensvariationerne i Nordelcasen viser også kun små variationer i frekvensen, men den viser også at der kan være tilfælde, hvor vindmøllerne anslår elektromekaniske egensvingninger i elsystemet. Med hensyn til stabilitets analyserne viser de forskellige cases at storskala integration af vindkraft kan ændre grænsen for spændingstabilitet og kan forårsage effektfluktuationer. Det fremgår endvidere af analyserne at den reaktive effekt der forbruges i parkerne spiller en nøglerolle med hensyn til spændingstabiliteten. Effektfluktuationerne under normal drift af vindmølleparkerne i Nordelsystemet i de undersøgte tilfælde viser at vindfluktuationerne ikke giver anledning til noget problem med stabiliteten af elsystemet.
vi
vii . I am also grateful to EFISintef where I did my external research in special to John Olav Tande. I would like also to tank my friends and supervisors Poul Sørensen and Henrik Bindner. which I am thankful for. Maria de Lourdes. Claudino Araújo and Alexandre Rosas. I would like to tank my wife Alexsandra Rosas for the long support in my dreams and for understanding and handling long lone periods while I was working to finish this phase of our life. Ørsted Institute at DTU for the general assistance in different ways. Technical University of Denmark. for the supervision. and Dr. family. You have contributed to this thesis more than you can imagine. Risø. guidance. Arne Hejde. Risø National Laboratory and Technical University of Denmark. This work has been also economically supported by Risø National Laboratory. Everaldo Feitosa. Last. in special to my mother (Elaine Carvalho). 30 March 2003 Pedro Rosas. I would like to thank all my friends. Kjetil Uhlen and Trond Toftevaag for the great opportunity. technical discussions and support in Trondheim. who also strongly supported and initiated the project. Mr. In addition. my brother (Gustavo Rosas) and relatives in special to Ms. Brazilian Wind Energy Centre. Jan RønneHansen (In memoriam).. I would like to thank the staff and management at the Wind Energy and Atmospheric Department from Risø National Laboratory and at the Section of Electric Power Engineering. who strongly supported and helped me and to Prof. Brazil. First. through a doctoral scholarship.. The work is a fruit of an international cooperation between the Brazilian Wind Energy Centre. I would like also to express my gratitude to Prof. but not least. Technical University of Denmark. helpful discussions and valuable time.Acknowledgment This work has been carried out at the Risø National Laboratory and Technical University of Denmark and supported financially by the CAPES (Coordenação de Aperfeiçoamento de Pessoal de Nível Superior). Norway.
8/152 .
5.3 Remarks on Wind Energy Integration __________________________ 36 3 Wind Speed Model _______________________________________________ 39 3.1 1.3 Validation of the Equivalent Wind Speed Model __________________ 58 Equivalent Wind Speed Model Remarks ________________________ 61 4 Wind Turbine Model______________________________________________ 63 9/152 .1.3.1.1 Power Spectral Density of Turbulence ______________________ 54 3.3 Deterministic Part of the Wind ______________________________ 47 3.1 Model Description ___________________________________________ 41 3.3.2 3.2.1.2.1 Tower Shadow ________________________________________ 47 3.5.4 Implementation of the Deterministic Component ________________ 53 3.1.2.1 Analysis of Voltage Stability _____________________________ 33 2.2 Wind Shear ___________________________________________ 50 3.1.1 SmallScale Integration of Wind Power _________________________ 23 2.1.1.Table of Contents Abstract ____________________________________________________________ v Resumé____________________________________________________________ vi Acknowledgment ____________________________________________________ vii 1 Introduction ____________________________________________________ 17 1.1.2.4 2 Motivation _________________________________________________ 17 Literature Review ___________________________________________ 18 Wind Power Basics __________________________________________ 20 Thesis Outline ______________________________________________ 21 Wind Power Integration ___________________________________________ 23 2.1.2 Coherence of the Wind __________________________________ 55 3.2 LargeScale Integration of Wind Energy ________________________ 29 2.2 Dynamic Operation _______________________________________ 26 2.1 Analysis of Power System Oscillations _____________________ 35 2.5.1 Voltage Stability Problem __________________________________ 31 2.1.5 Stochastic Part – Turbulence ________________________________ 53 3.1 Equivalent Wind Speed Model ______________________________ 43 3.2 Frequency Control Problem_________________________________ 34 2.2 Description of the Equivalent Wind Model_____________________ 44 3.3 Power Spectral Density of a Rotating Blade – Rotational Turbulence _____________________________________________________ 55 3.6 Implementation of the Stochastic Component___________________ 56 3.3 1.2.1.2 1.1 Steady – state operation____________________________________ 24 2.1.1.1.
2.2.3.4 Wind Power Impact on the Voltage Quality ___________________ 111 6.3.1 Wind Speed Simulator __________________________________ 86 5.2 5.2 Drive Train ___________________________________________ 67 4.2 Wind Turbine Model ________________________________________ 64 4.3.2.3 4.3 Wind Power Impacts on the Voltage Stability _________________ 109 6.2.1 Reduced Nordel Model_________________________________ 118 6.2 Wind Power Representation _______________________________ 102 6.1.2.1 Case Study 1: Voltage Stability in a Modern Power System ________ 99 6.2.1 5.2.1 Simulation Tool _____________________________________________ 63 4.2 Wind Power Representation _______________________________ 108 6.2.2 Electrical Components ____________________________________ 70 4.1 Electrical generator_____________________________________ 70 4.2.4.2.2 Nordel Characteristics _________________________________ 120 6.5 Verification of the Complete Wind Turbine Model________________ 71 Dynamic Wind Turbine Model Remarks ________________________ 76 Aggregate Wind Speed Model _________________________________ 77 Coherence of Turbulence in a Park Scale________________________ 78 Aggregate Turbulence _______________________________________ 79 Simulation of Wind Speeds ___________________________________ 80 Aggregate Wind Turbine Machine _____________________________ 84 Aggregate Wind Farm Model ______________________________________ 77 5.7 6 Aggregate Wind Farm Model Remarks _________________________ 97 Large Scale Integration – Case Analysis _____________________________ 99 6.2 Case Study 2: Voltage Stability and Quality in a Brazilian Power System __________________________________________________________ 106 6.1 Aerodynamic Rotor ____________________________________ 65 4.2.2.1 Light Load Condition __________________________________ 113 6.6.3.1.6 Results and Discussions ______________________________________ 84 5.1.6.1 Aeroelastic Components ___________________________________ 65 4.1.4.1.6.1.3.2.3 Wind Power Impacts on Voltage Stability ____________________ 102 6.1 Power System Description ________________________________ 117 6.2 Wind Power Projects and Representation _____________________ 123 6.1.2 Heavy Load Condition _________________________________ 115 6.3 Extension to Large Wind Farms and Different Random Seeds ___ 94 5.2.3 5.1 Power System Characteristics ______________________________ 106 6.1 Aggregate Wind Farm Model____________________________ 123 10/152 .2.3 Case Study 3: Power System Interactions – NORDEL ____________ 116 6.4.2 Wind Farm Power Production ____________________________ 87 5.4 5 5.4 5.6.2 Results _________________________________________________ 86 5.1 Power System Characteristics _______________________________ 99 6.2.6.1 Case Description _________________________________________ 84 5.
1.3 Reactive Power Compensation.2 Voltage controllers ____________________________________ 136 7 8 9 Conclusions____________________________________________________ 137 Reference List __________________________________________________ 139 Annexes _______________________________________________________ 147 9._____________________________ 150 9.6.3.1.1 Remarks on Case 1 ______________________________________ 134 6.2 Remarks on Case 2 ______________________________________ 134 6.2 Stepup Transformer _____________________________________ 149 9.4.1 Electrical Components Model in SIMPOW _____________________ 147 9.1.3.3 Wind Power Impacts on the Power System Voltage and Frequency Regulation ______________________________________________________ 126 6.1 Frequency controllers __________________________________ 126 6.5 Slack bus ______________________________________________ 150 11/152 .4.3.3.4.2 Voltage Quality _______________________________________ 131 6.1.3 Remarks on Case 3 ______________________________________ 135 6.4 Lines and Cables ________________________________________ 150 9.3.3.4 Case Analysis Remarks ______________________________________ 133 6.3.1 Electrical generator ______________________________________ 148 9.4.1.4.1 Frequency control _____________________________________ 135 6.
.......1 Basic components of a wind farm........................49 Figure 3...................................................................52 Figure 3.............................8 The tower shadow effects on the horizontal wind (top view).....48 Figure 3.. .3 Voltage fluctuations corresponding to flicker emission unity [34]...41 Figure 3..................5 PSD of measured electrical power output of a 500kW stall regulated wind turbine....................7 Equivalent Wind Speed Model principle...............................40 Figure 3.....................................................6 Single line equivalent of a Power System......5 Basic Power System Structure......... ....................................................................39 Figure 3................... .... ................52 Figure 3.....41 Figure 3..................... Normalised torque influenced by wind shear (site with a medium z0)........ ...............49 Figure 3................................................33 Figure 3..............................1 Basic components of a wind turbine unity.......................................57 Figure 3............................. ........ ....................18 Implementation of the stochastic model in Simulink/MATLAB.......4 Wind speed measured on a section of a rotating blade [45].20 Figure 1....... ....................... .....17 Normalised admittance function to 3p..............51 Figure 3..... ..........54 Figure 3...............3 General overview of wind turbine models.....................31 Figure 2...............27 Figure 2.......57 Figure 3....... Kaimal PSD of Turbulence.42 Figure 3..................... ..................................... . ...................................... ..44 Figure 3...................................................30 Figure 2.......4 Measured power spectra of the electrical power from a 225kW pitch regulated wind turbine..............25 Figure 2............ .....................................13...........2 Power smoothing effect from wind farms............ ......53 Figure 3.......8 Power transfer to a node as function of the voltage (“nose curve”).........................58 12/152 .....................................11 Wind shear for different sites (ground is reference for height).......10 Normalised torque influenced by the tower shadow......................43 Figure 3.......List of Figures Figure 1.............24 Figure 2....... ...................29 Figure 2...............12 Reference axis and angles used in the wind turbine......15.9 Wind speed field interference by the tower shadow.............................6 Reference axis used in the wind turbine...................21 Figure 2.........................2 Power produced by a 500kW stall regulated wind turbine...............27 Figure 2................................ ...........................................................14 Implementation of the deterministic model in Simulink/MATLAB........................ .....................................2 Single line equivalent for a wind turbine connection........16 Normalised admittance function to 0p... .....7 Simplified transmission line equivalent diagram.1 Illustration of the wind on the rotor area of a wind turbine [43]..................
..........6 Reactive power as a function of the speed and voltage for an asynchronous generator (1p....... .................................... = rated reactive power at 1pu volts)..3 Example of drivetrain components............................................... ............13 Power simulated at 16 m/s and turbulence intensity 20%...................................11.......... 86 Figure 5...............8 Measured wind speed.............. 61 Figure 4. .................................9 Time series of simulated and measured power to the Nortank 500kW......................... 93 13/152 .................. .10 Verification of the standard deviation of the dynamic wind turbine model.......................... ...................... 73 Figure 4.................................................................15 Flicker coefficients comparisons (computed according to [34]). ........................ .............20 PSD comparisons of the stochastic model........ ....6 Wind farm layout......... 74 Figure 4.............................9 Power characteristics evolution with the wind speed (20% turbulence intensity)............................................................................... Verification of the flicker Pst to different frequencies............ 92 Figure 5......... 88 Figure 5. ................................................................... 67 Figure 4...................................14 Power spectral comparisons at 16 m/s and turbulence intensity 20%....1 Spatial disposition of the two wind turbines (αxy = 90° means lateral disposition). 64 Figure 4...... Non linear effects on the power variations....u..................................4 Dynamic representation of the drive train model............ ..... ..................... 90 Figure 5.....................19 Normalised simulated deterministic wind component compared to a DBP2 model..........8 Power characteristics evolution with the wind speed (10% turbulence intensity)... 68 Figure 4................................................ 85 Figure 5...................................... 71 Figure 4..................................11 Power simulated 13 m/s and 10% turbulence intensity........... 66 Figure 4........................................ 75 Figure 5......................................................................... Static power curve of the wind turbine...................................................... .................. ......................... ................. 60 Figure 3....................1 Interaction between each components of a wind turbine unity............. ................................................... .. 91 Figure 5................................................... .Figure 3...............10........................................5 Active power as function of speed and voltage terminals for an asynchronous generator.......................................................7 Comparisons of the wind speed simulator.5.4 Distribution of random constants. 70 Figure 4....................2 Power coefficients to compute the dynamic power coefficient.... ... ......................2 Coherence factor for different distances between two points... 73 Figure 4.........................................21 Measured and the simulated equivalent wind speeds on rotating blade section............... 89 Figure 5.....................7 Cp(λ) static characteristic of the wind turbine Nortank 500kW.. ............................... 81 Figure 5............. ......................................................... 89 Figure 5... 87 Figure 5..... 59 Figure 3....... 78 Figure 5.... 72 Figure 4.................... 79 Figure 5............................................. 90 Figure 5................................................3 Structure of the AWFWS generator............ 84 Figure 5.......12 Power spectral comparisons at 13 m/s and turbulence intensity 10%..........................................
........ Voltage at MOSSORO and power flux from the wind farm in light load condition (mean wind speed 10m/s).....16 Evolution of flicker coefficients at 13m/s 20% turbulence intensity... ............ .............................. reactive power and voltage variations at MOSSORO (light load)................................... active power.................................. ............6 Maximum wind power integration concerning voltage stability (load factor unity)...................................... ......... Loadability curve to bus 3 with wind turbines using power electronics.... .111 Figure 6........................... ..................105 Figure 6...............93 Figure 5..........13 Maximum wind power to MOSSORO bus (light load condition)...........18 Voltage at MOSSORO and power flux from the wind farm in heavy load condition (mean wind speed 10m/s).11 Layout of a single wind farm applied to the Brazilian power system studied.........106 Figure 6.......................................96 Figure 6....2 Diagram of the Power System used in analysis (loads in MW and MVAr).......................... .................107 Figure 6.......................................................................94 Figure 5..................20 High voltage Nordic power network [75]..................117 Figure 6................ ..............................119 Figure 6......19 Statistics wind speed............... ............................12 Network topology of Brazilian power system studied............................................103 Figure 6.....110 Figure 6....21 Reduced model to the Nordic power system [38]....... .......15 General wind power influences on MOSSORO bus... reactive power and voltage variations on MOSSORO (Heavy load)....23 Modal analysis of the eigenvalue 0..............................10 Brazilian network studied [72].....Figure 5... .........1 Case Studies...............................................116 Figure 6...................................3 Loadability curve to bus 3 without wind turbines.......................................34962 + 0............................114 Figure 6.....................7.............................. ............113 Figure 6......14 Wind power influences on the voltage to different wind speeds............18 Influences of different sizes of aggregate wind farms to the power characteristics at 13m/s.....95 Figure 5..............................4 Loadability curve to bus 3 without wind turbines....108 Figure 6............................112 Figure 6....................99 Figure 6.............................114 Figure 6.104 Figure 6...................... 20 % turbulence intensity .......................16.... ............5 Maximum wind power integration concerning voltage stability..........................................8 Evolution of the power production from the wind turbines (with electronic power converter)...100 Figure 6................17 Statistics wind speed...............................................................17 Evolution of flicker coefficients at 16m/s 10% turbulence intensity...................................... ..115 Figure 6............22 Relevant eigenvalues of the reduced model to the Nordel..................... ........55164 Hz –Nordic Power System..............121 Figure 6.....................109 Figure 6.................102 Figure 6......... ..........................9 Brazilian interconnected system power system [72] ...100 Figure 6................. 20 % turbulence intensity ..................19 Influences of different sizes of aggregate wind farms to power characteristics at 16m/s.............. active power.122 14/153 .......................................
.........................5Hz)).............. .......................5Hz))............................6 Electrical transmission model......................... 133 Figure 9................................................................................. 125 Figure 6........... ............... ..................30 Power spectral distribution of power and speed to selected machines (AWF modified to lower rotational speed (3p~0......................... Electrical generator model parameters......................................... 124 Figure 6...................... frequency and standard deviation of power (mean wind speed at 12m/s to Finland).27 Power variations and power balance in the Nordel case studied......32 Wind power and voltage deviations simulated in Nordel system (mean wind speed 12m/s to Finland)........................................... 148 Figure 9.................. 132 Figure 6............................4 Electrical transformer model................................29 Power spectral distribution of the power and speed of selected machines (mean wind speed at 12m/s to Finland).... 149 Figure 9. ........ 150 Figure 9.............. 130 Figure 6.........33 Power spectral distribution of voltage and voltage deviation to selected machines in the Nordel (Finland AWF at 12m/s).. 132 Figure 6...................Figure 6.2...28 Wind power....................................... 151 15/153 .......................................................................... 127 Figure 6..........................24 Aggregate wind farm power simulation...31 Wind power....................................................8 Slack bus model (positive sequence)........................................................................... 151 Figure 9.................................34 Power spectra distribution of voltage and voltage deviation simulated in Nordel System (mean wind speed 12m/s to Finland AWF low frequency (3p=0...... ......................... ........................ 148 Figure 9...... 150 Figure 9...7 Slack bus structure....................................................................................................3 Structure of the transformer model.........25 Aggregate wind farm power variation.....................1...... ...................... frequency and standard deviation of power (average wind speed 16m/s to Finland).....5 Electrical Transmission structure model...... ................... 125 Figure 6..... 129 Figure 6... 149 Figure 9.................................. ............. Electrical generator equivalent...... 128 Figure 6....................... 131 Figure 6.....26 Aggregate wind farm power characteristics (at 20% turbulence intensity).............................................................................
.............1 Bus 3 loadability limits keeping the reactive power constant.... Requirements on frequency response in the Nordel power system.1 Basic characteristics of the Nortank 500kW wind turbine modelled....................110 Table 6.....103 Table 6............50 Table 3........59 Table 3.88 Table 6.......1 Basic characteristics of the 660kW wind turbine modelled...........................83MW wind turbine.60 Table 4.............118 Table 6..........123 Table 6.......2 Representative power characteristic values of all simulations..................................................................25 Table 2... ........2 Bus 3 loadability limits keeping the active power constant......... ...........4 Loadability limits to MOSSORO.....72 Table 5...111 Table 6........104 Table 6........................10 Aggregate wind farms average wind speeds...............................6.....85 Table 5.124 Table 6....List of Tables Table 2.....3 Relevant loads in the Brazilian power system studied............................7 Nordel reduced machines connection nodes............................................................21 identifies the buses’ name)...... .......120 Table 6.........2 Main power system influences from the wind energy integration...108 Table 6........................1 Typical values of surface roughness length z0 for various types of terrain [53]....1 Main steady state parameters defined in IEC 6140021 [1]..............................2 Parameters used in the simulation for tower shadow.............36 Table 3................................................8 Wind power plans simulated (Figure 6..3........5 Loadability limits to MOSSORO with wind power.................... .....9 Basic characteristics of the 1................................. Parameters of the wind turbine used in the measurement comparisons.........126 16/152 ........
During the 90s. the wind turbines (and farms) grew in size and ratio from the few hundreds kilowatts to the megawatt size. In the late 90s. The studies focus on the dynamic behaviour of the induction machines during disturbances. Wind power has to overcome some technical as well as economical barriers if it should produce a substantial part of the electricity. some of the technical aspects are treated. In this report. the focus is on the wind power as it is said to hit large integration in the near future. The size of those installations did not threaten the overall power system stability and the voltage quality assessment was simple (when connected to the conventional power system). During the period. where the dynamic effects of the turbulence were neglected. particularly those regarding the power system quality and stability. The 90’s represented an important break through. the analysis concentrated on development of the wind turbine technology and investigation of the dynamic behaviour of the wind turbines. At this 17/152 . a simple and robust wind turbine concept emerged and became very popular pulling the wind power industry. Wind turbines along with solar energy and fuel cells are possible solutions for the environmentalfriendly energy production. In this report. 1. which raises some technical problems concerning grid integration. Nowadays. it influenced the voltage level on the grid. During the 80s. a gearbox. During the period. some power systems start to face problems of integrating thousands megawatts of wind power. the wind power installations to small isolated networks are not included. This technology has already reached a penetration level in some areas. most of wind power installations were limited to few hundreds kilowatts to the existing distribution grids. In this thesis. During the same period an International Electrotechnical Commission (IEC) task force issued a standard procedure: the IEC 6140021[1] to fillin the lack of technical standards on assessment of power quality from wind turbines. The simple and robust concept includes a three bladed wind turbine rotor. in some cases. with the wind farms rating hundreds megawatts. The increasing concerns to environmental issues demand the search for more sustainable electrical sources.1 Motivation It is possible to state that the significant impact of wind power started in the beginning of 80s very much related to the mid 70s oil crises. It was cheap and very robust but the power quality was poor and. an induction machine directly connected to the grid and a control system. new concepts emerged because of a demand for more efficient power production and to comply with power quality requirements. The increased rated power of the wind farms to areas with good wind resources leads to concerns on: “to which extent the wind power interferes to the power system?” Most of the decade has been dedicated to voltage quality analysis of wind power and to economics of the power systems including wind power. the concerns start to focus on the transient voltage stability of the power system.Chapter 1 1 Introduction Wind energy is said to be one of the most prominent sources of electrical energy in years to come. which are decentralised spread over large extensions.
The operation and energy management systems focus on the forecast of the wind power and its relation to trade agreements.moment. several works have been done on the analysis of power quality and on the transient stability from wind turbines to the power systems. On the dynamic subject. the main challenges are classified in longterm planning. It characterized the impacts of large amount of renewable energy on European conventional utility practice (not only operation but the institutional aspects also). One of the conclusions was that there were no clear limits on wind power integrations and that penetration limits from many studies were economic rather than technical limits. The last topic. This report focuses on the dynamic behaviour of the wind power resources to an extensive area and its influences on the power quality and voltage stability issues. One of the main aims was to define technical limits to intermittent power integration. such as the adjustment of agreements between transmission system operators to cope with the stochastic nature of the wind power. Similar work was presented by Jørn et all in [5] where the focus were on the trading/economic aspects from operation of large–scale wind power in the Danish power system. Wind power influences several power system characteristics from economic dispatch to stability and quality issues. A project report (ALTENER) in [2] aimed to establish insight and to orient works in the integration of renewable power in the European Network. Another important aspect included in this topic relates the distribution system reliability with wind power. focuses on the control of voltage and frequency. the power system performance.2 Literature Review Integration of wind power into the power system has been studied by many authors before but most of them focused on different characteristics and issues of the power system. The long term planning focuses on several topics. operation and control of the power systems with large wind power become very important [2]. On these problems. operation and energy management systems. the problems of planning. It also clearly concluded that the spatial distribution of the wind must be taken into account and wind speed data are needed. power quality issues. this also includes economical and financial issues. This latter report related potential problems on the power exchange agreements. and on the dynamic behaviour of the wind power sources. 18/152 . Some of the conclusions are that dynamic studies of the power system characteristics are required in largescale renewable energy penetration and the stochastic nature of the wind speed is relevant to the power system control and quality. A more recent report by Nielsen et all in [4] reviews the technical options and constrains of integration of distributed power generation. stability of the network and power quality. and power system performance. Wan and Brian in [3] pointed out main factors to utility integration of solar and wind power where studies from late 1970s until 1980s provided a starting point and general classification of the most relevant power system aspects. One of the focuses was on a new power system structure to deal with the imbalance on consumption and production from the stochastic nature of the wind power. It also includes the security analysis and the reserve of power to ensure reliable operation of the entire power system. 1.
On those works. the power variations from wind farms can be reduced significantly particularly in large scale applications. Estanqueiro in [17] and Petru in [18] presented wind turbine models that could be used to power quality studies. Aggregate dynamic models are used to simulate integration of wind power to the dynamic operation of the power system in this thesis. Giebel in [21] showed that the spatial distribution of the wind turbines gives benefits in long term because it reduces the power variations from the wind energy. Beyer et all in [22] showed that the spatial correlation of the wind speed influences the power production of wind farms. With the reduced model. Similar. The Power System Engineering Committee of the IEEE in [25] resume several works on power system analysis specifically on modal analyses for system dynamic performance that can identify the possible problems on normal operation of the power system. Tande in [9] explained how to assess the voltage quality from wind turbines using wind turbine characteristics. the transient voltage collapse was investigated as well as the wind turbines behaviour during short circuits. However. Wilkie et all in [16] presented simple models to wind turbines and argued that the contributions of wind turbines components must be modelled into the overall simulation but the total accuracy was not essential to obtain an adequate representation. where the work focused on the small scale integration. Akhmatov et all in [10]. however the modal analysis must be complemented with dynamic simulations. Some of the analyses include the simulations of the entire power system operation under normal operation and faults. Brunt et all in [11] and Wiik et all in [12] presented transient stability analysis of the power system with large amount of wind power. From both works. Lei et all in [26] and Eliasson in [42] present dynamic reduction of large power systems for stability studies aggregating machines with similar (coherent) dynamic characteristics. none of them analysed the results to the dynamic power systems. some textbooks present the wind turbine characteristics. All works have focused on the power quality of wind turbines. Taylor in [27] and Custem in [28] 19/152 . Freris in [13] and Hau in [14] both detailed the wind energy conversion systems presenting a general overview of the wind turbine and Heier in [15] addressed specifically the integration of wind power where much effort was given on the characterization of the wind turbine/farms components. On modelling wind turbines/farms. Voltage stability has been pointed out as another problem to large integration of wind power because wind farms demand reactive power. the turbulence spatial correlation must be included. In wind farms. Sørensen in [7] and Larsson in [8] characterized the moderns wind turbines and classified the most relevant characteristics that supported the IEC 6140021[1]. Kundur et all in [23] and Anderson and Fouad in [24] present dynamic models to the power system components and means to analyse the stability of it. Sørensen et all in [19] and Estanqueiro et all in [20] showed that the assessment of power quality from wind farms depends on appropriate representation of the wind speed. The dynamic power system analyses have been extensively investigated. the overall dynamic analyses are time feasible. Dynamic simulations of very large power systems are very expensive.Tande et all in [6] developed an extensive analysis of the potential impacts of wind turbines on the power quality.
the most common type of wind turbine installed. Blades Nacelle Gear Box Wind Sensors Network Rotor Generator Brake Tower Stepup transformer Yaw Control Tip Break Figure 1. In addition. a control system. The active and reactive power control is very important to keep the frequency and voltage stable within limits. this wind turbine technology is considered the poorest power quality when harmonics problems are not concerned. This report focus on this wind turbine technology influence on the power system voltage stability and on the power system quality. limited reactive power compensation and a stepup transformer. Lack of reactive power can lead to voltage problems and no control in the active power can cause frequency deviations. the Power System Stability Subcommittee of the IEEE in [29] and Cañizares in [30] suggest several tools to analyse the voltage stability. The conventional wind turbine is connected directly to the grid and the generator is “synchronized” to the network. an electrical generator. The main power system problems from this wind turbine technology come from the lack of control on the active and reactive powers. one of them is the loadability curves to characterize the maximum load that can be installed before the voltage collapses.1 presents the basic components of a conventional wind turbine. The conventional wind turbine is even at the present time. a mechanical transmission system. Because of lacking controls on active and reactive power. this report focus on the largescale integration hence a substantial part of this report is dedicated to model the power fluctuations from large groups of wind turbines. This technology has been named “fixed” rotational speed wind turbine because the induction generator allows small mechanical speed variations. The loadability curves are used in this thesis in connection with the wind power.3 Wind Power Basics The wind turbines are composed of an aerodynamic rotor.1 Basic components of a wind turbine unity. The power produced from a large number of wind turbines will vary relatively less than the power produced from a single wind turbine due to the cancellation effect from the poor 20/152 .presented an extensive explanation of the voltage stability problem and means of copying with it. Figure 1. In addition. 1.
4 0.3 0 100 200 300 Time [s ] 400 500 600 1 Wind Turbine 30 Wind Turbines 150 Wind Turbines 300 W ind Turbines Figure 1.] P owe r [p.3 0 100 200 300 Time [s ] 400 500 600 0. In addition to the problems of dynamic power fluctuations.8 0.3 0.) 1.2 Power smoothing effect from wind farms.5 0.u.5 0. After having presented the power system interactions with 21/152 .1 1.spatial correlation of the wind acting on each wind turbine. Power(p.4 0.6 0.4 0. Here transient stability is not studied.) 1 Power(p.4 0. 1.2 illustrates the power “smoothing” effect when increasing the number of wind turbines.1 1 Power(p.u.8 P owe r [p. The voltage stability in the power system can be classified in slow dynamic and transient.7 0. Figure 1. The power variation from wind turbines is very complex and demand special techniques to cope with the spatial distribution of the wind turbines than a simple scale up from a single wind turbine.7 0.) 1.7 0.6 0.u.u.9 0.8 P owe r [p. The transient voltage stability deals with the voltage stability after the power system being subjected to large disturbances.4 Thesis Outline Chapter 2 addresses the main problems of the power system that are related to the wind turbines. Once the wind turbines have limited reactive power compensation and usually demands reactive power from the power system.) 1.u.u.8 0.1 Power(p. normally short circuits.7 0.u.5 0.5 0.1 1 1 0. Transient voltage stability problems have also been related to large integration of wind turbines. The slow dynamic is related to slow increase in load in the power system and deals with the reactive and active power supply.9 0.] 0 100 200 300 Time [s ] 400 500 600 0. It starts introducing the main power quality characteristics of the wind turbines and proceeds to present the main problems from integrating single wind turbines in the power system are presented.6 0.6 0. here its influences of the reactive power demand and the active power injection are investigated.] 0.3 0 100 200 300 Time [s ] 400 500 600 0.9 0.] P ower [p.9 0. another important issue investigated in this report is the voltage stability from connecting large amount of wind power.u.
Finally.se 22/152 . In this thesis.wind turbines. The main characteristics of coherence and its influences on the dynamic wind farm power production are presented. First. and the most relevant characteristics of each part are described. The dynamic wind turbine model however can be directly applied to other simulation tools. It starts dividing the wind turbine in aeroelastic and electrical components. The AWF model is a single equivalent wind turbine that replaces several wind turbines in the wind farm. Chapter 5 presents the turbulence coherence effects on the wind farm production. The wind is classified in two main components. the integration of more than 4GW of wind power to the Nordic Power System illustrates application of the aggregate wind farm. are discussed. a suitable wind speed model to assess the power quality from wind turbines is presented. www. the dynamic wind turbine model is implemented in SIMPOW/ABB1. Chapter 6 presents some illustration cases of the voltage stability analysis and power quality from large integration of wind turbines. Chapter 5 also presents an Aggregate Wind Farm (AWF) model that can be used to power quality assessment of wind farms and to dynamic stability analysis of power systems with large number of wind turbines. Chapter 3 presents the main characteristics of the wind acting on the rotor of wind turbines. where available dynamic models to the electrical components are used. based on a scale up from a single wind turbine.abb. The aeroelastic components present the relevant characteristics and models to the wind turbine aerodynamic rotor and drive train. In the last part of the chapter. the wind turbine model to power quality assessment is presented and compared against measurements. the large wind farms impacts on the voltage stability to a part of a power system are illustrated. In the final section. In the second part. Chapter 4 presents the main components from wind turbines. 1 SIMPOW® is a dedicated digital simulation tool to power system dynamic analysis developed by ABB©. the connection of large amount of wind power to a Brazilian network is presented in terms of voltage stability and quality. the possible problems to large integration of wind turbines in to the power system. The wind speed to the AWF takes into account the wind farm layout and the spatial coherence.
in some particular cases. The problem is introduced by pointing out the relevant power quality characteristics of a small wind farm (or a single wind turbine) and after it is scaled up to represent the largescale case. Chapter 2 presents the main impacts from wind power on the power system with emphasis on the voltage stability and power quality. 2%). The wind energy counts for small part of the total energy production in the power system (i. above 10%). The largescale integration can cause power quality or stability problems and. are not new to the power system engineers. Figure 2. and means of coping with them. 23/152 . transmission systems.1 SmallScale Integration of Wind Power In this case. the scales of integration are defined as follow: • Smallscale wind power integration – the wind power installed is relatively small compared to the conventional power system. In this thesis. Each wind turbine has as basic electrical components: an induction generator. up to few percents e. additionally the wind farm may use an integration transformer to connect to a higher voltage level e. The wind power counts for large part of the total energy production in the power system (e. Hence.g.g. the focus is on the direct connected wind turbines type (so called “fixed” rotor speed).e. The power quality and stability problems. the voltage and frequency problems are concerned. the power system is assumed to have enough spinning reserve of active power and the frequency is kept constant therefore only voltage problems are concerned. Largescale wind power integration – the wind power installed sizes the conventional power stations.Chapter 2 2 Wind Power Integration Large integration of wind power can lead to problems on the voltage control or on the stability of the power systems as mentioned in chapter 1. the active power that comes from the wind is transferred to the power system without storage devices. On those types of wind turbines. The wind farm limit is defined by the Point of Common Coupling (PCC). • 2.1 presents the relevant electrical components of a conventional wind farm. the frequency ca be affected by the wind turbines. those problems related to wind power are not well described when it comes to largescale integration.g. In smallscale integration. the power system is considered strong and the main problems from connecting wind farms come from the voltage control. The wind farm is composed by several wind turbines. Here the main problems to the voltage stability and to the power quality related to largescale integration of wind power are presented. First. However. local reactive power compensation and a stepup transformer.
The classification presented as follow is in agreement with IEC 6140021 [1]: • • • Steadystate – does not include dynamics (very slow dynamics representing periods above 10 minutes to hours). Here. periods less than 20 milliseconds) due to the electronic equipments installed in the wind turbines (this last part is not related to the wind speed).1. the voltage control on this type of wind turbine can be done only by changing the amount of reactive power compensation (shunt capacitors installed).g. the voltage level/variations at the PCC can demand a variable tap change transformer. 24/152 . Induction (or asynchronous) machines applied as generators demand reactive power from the network (chapter 4). The voltage levels will not exceed limits. variable reactive power compensation). in addition.1 Basic components of a wind farm. which is partially compensated with shunt capacitor banks. The lack of control on the active and reactive powers can disturb the voltage on the PCC. special reactive power compensation is demanded and installed at the PCC (e. Harmonics – includes voltage variations in high frequency (e.g. Wind farms have very little control of the active power due to the stochastic behaviour of the wind. Dynamic – include the dynamics in the time frame from milliseconds to 10 minutes. 2. above 50Hz to Europe and 60Hz to US and Brazil.1 Steady – state operation The steadystate operational analysis assures that the: • • The currents will not exceed thermal limits nor will the protections act during extreme powers. In special configurations. In special installations. the harmonics are not an important issue because this report focus on the direct connected wind turbine type that does not emit harmonics components on current. The disturbances on the voltage from wind turbines are classified in different time scales. therefore it is not included in the following subsections.To the next WT StepUp transformer WTG Capacitor StepUp transformer WTG Integration transformer Network PCC Capacitor To the next WT Figure 2.
In near future. From these certifications. Parameters Rated active power Rated reactive power Rated apparent power Rated current of the wind turbine at rated Voltage Rated voltage of the wind turbine Maximum permitted power setup in the controller Maximum measured power in 60 seconds average period Maximum recorded power in 0. U is the voltage at the wind turbine terminals and ∆U can be computed as: 25/152 .1 presents the main steadystate parameters to wind turbines certified to power quality [1]. The impacts on the voltage quality to the different conditions as expressed in Table 2. wind turbines can be certified to power quality [1]. Figure 2.1 can be computed with help of a load flow program or by simple equations.2 Single line equivalent for a wind turbine connection.2 presents the electrical representation of the wind turbine and the power system. Table 2. there are no load or shunt elements installed at the wind turbine terminals. S=P + j Q U0∠0 Reference Node Z=R+jX ~ U∠δ Wind turbine terminals Figure 2. Table 2. where the reference node and the equivalent impedance represent the entire power system at the wind turbine terminals. a set of data will help to verify the steadystate operation of the wind turbines and wind farms.1) where.2 Based on the parameters specified on Table 2.2 seconds average period Reactive power demand/supply as function of the active power Reactive power measured or estimated to the Pmc Reactive power measured or estimated to the P60 Reactive power measured or estimated to the P0. Following. P and Q are the active and reactive powers respectively from the wind turbine.1 and with the electrical characteristics of the network it is possible to determine the impacts on the voltage quality as well as the maximum currents on the cables and transformers.1 Main steady state parameters defined in IEC 6140021 [1]. U0 is the voltage at the reference node. simple equations to determine the voltage levels are introduced. The voltage at the wind turbine terminals (U) can be determined as follow: U = U 0 + ∆U (2.
∆U = (PR + QX ) + (PX − QR ) ⋅ j U0 U0 (2. Using Equation (2. 26/152 . The voltage can increase or decrease depending on the amount of the reactive and active power flux and on the network characteristics. The dynamic voltage variations from the wind turbines during operation are quantified by flicker and step change [1]. Figure 2. it is important to detail the network and include the loads installed and to use a load flow program to compute the voltage and currents on the relevant nodes and lines respectively. The value is computed to short term (10 minutes) and long term (120 minutes).2) where.2) it is possible to compute the voltage levels and compare to preset limits imposed by the local network operator. in most cases.3 presents the voltage fluctuations as a function of frequency that will represent a unity of shortterm flicker perceptivity (Pst) to two different conditions: sinusoidal voltage fluctuations and rectangular voltage fluctuations based on [34]. However. R and X are the resistance and reactance inductive characteristics of the electrical network respectively. These continuous variations of active and reactive powers from the wind farm cause dynamic voltage variations. The flicker emissions during continuous and switching operations and the voltage step change are the voltage quality indicators influenced from small number of wind turbines connected to the grid.2 Dynamic Operation The wind turbines dynamically produce power that varies in a broad range of frequencies and amplitudes. The flicker emission is computed from flicker coefficients measured from wind turbines during the power quality data sheet. The flicker emission is a measure of the human perception of the bulb light variation consequent of the voltage low frequency variation. The flicker emission includes voltage variations in frequencies up to 25 Hertz that are weighted with an eye perception function according to [32] and its posterior amendments [33] and [34]. 2.1.
7Hz.50E+02 3.00E+00 0. the maximum flicker perception comes from around 8Hz where the voltage fluctuations must be reduced in order to respect the flicker limits. At the frequency of 2. 5. 1. The PSD in Figure 2.3 considers that voltage variations leads to light intensity variations in light bulbs. 8.00E+01 0.4 Measured power spectra of the electrical power from a 225kW pitch regulated wind turbine. The limits on Figure 2.00E+02 2. 10.50E+02 4.1Hz (3p – three times the rotational speed of the rotor) there is a large contribution to the power variation. at this frequency there is a small contribution related to some asymmetry in the rotor.00E+02 5. 4.Figure 2. 7. From Figure 2. Figure 2.4 includes contributions from deterministic and stochastic parts.3.50E+02 1. 5.50E+02 2. Frequency (Hz) 6.00E+02 Srot*f (kW²) 4.3 Voltage fluctuations corresponding to flicker emission unity [34].00E+02 3.4 presents the Power Spectral Distribution (PSD) of the power produced from a three bladed wind turbine (225kW). The 3p effect is related to 27/152 . 3. 2. 9. Figure 2.00E+02 1. In order evaluate the flicker contribution from wind turbines. The fundamental frequency of rotation (1p – one time the rotational speed of the rotor) is approximately 0.
The flicker and voltage dips from switching operations are not treated in this report. corresponding to 12p. 28/152 . the dynamic components of the wind turbines damp the high frequency power oscillations. drive train torsional moments. The turbulence and tower shadow influence the wind field on the rotor area and the three blades crossing the wind field transfer the power variations to the main shaft. The power variations are consequence of the wind field on the rotor area and the wind turbine dynamics.e. The shortterm flicker emission (Pst) and the longterm flicker emission (Plt) to a wind farm can be estimated according to [1]: 1 Sk Pst = Plt = ∑ (c (ψ i =1 i N wt k . The power on the main shaft will dynamically interact with the wind turbine components. These effects will be more detailed in chapters to come where each component of the power fluctuation from a wind turbine is explained and models presented to simulate them. therefore it is restricted to the flicker emission during continuous operation. Sn. the power variations reduce significantly above the frequency of 3p.3 and Figure 2. e.4Hz. PSD of power measurements from different threebladed wind turbines show similar pattern. The highenergy content on 3p frequency comes from the effect of the blades rotating on the turbulent field added together to the tower shadow. ci is the flicker coefficient of wind turbine i to specific network impedance phase angle (ψk) and annual average wind speed va from the site. The flicker due to switching operations is caused by startup or switching of generators of wind turbines because the high inrush currents cause voltage dips. The power variation in very low frequency below 0.g.i is the rated power of wind turbine i and Nwt is the number of wind turbines in the wind farm. and finally the generator will convert the power to the network. Associated to the flicker emissions during switching operations.7 Hz is caused by the simple turbulence acting on the rotor area but has small influence on the flicker.4 indicate that the main flicker contributions from wind turbines comes from the 3p power variations. In the frequency of 8. Figure 2. i. va ) ⋅ S n .3) where Sk is the short circuit capacity. In addition. wind turbines also generate flicker due to switching operation and startup. This report focus on the analysis of the continuous operation of wind turbines. a three bladed rotor cancels the multiples harmonics different from the 3np and in addition. Although one could expect high power variations in a broad frequency range. a small amount of energy is also presented that has been related to the flexible aeroelastic part of the wind turbine in addition to the induction generator [66]. the voltage dip is relevant because the voltage will drop instantaneously due to the inrush current. The flicker during continuous operation is caused by the power fluctuation from the turbulence added to the wind turbine dynamics. The power quality tests of wind turbines express a flicker coefficient for each wind turbine for different network phase angle condition and different annual mean wind speeds for a wind farm. The flicker defined in the previous paragraphs is related to the continuous operation of the wind turbines.i ) 2 (2. The flicker emission in short term and long term can be estimated from the power quality tests [1].rotational turbulence and the blades passing the tower in a threebladed rotor type of wind turbines.
The large integration can occur in two main conditions: • • Large wind farms connected to the transmission system or. Several small wind farms connected to the distribution systems in one area of the power system. the frequency was assumed constant. hence the voltage stability limits can be reduced and must be analysed too. The generation system is mainly composed by synchronous machines that are usually large.2 LargeScale Integration of Wind Energy As introduced before. In addition. In the smallscale integration. large reactive power demanded by the wind farms can reduce the reactive power supply. where it is possible to distinguish the main system components.3) takes into account the “cancellation” effects. In order to introduce the maim issues of the power system and wind turbines. here the largescale integration problems are based on the smallscale ones.5 Basic Power System Structure. 2.5 presents a simple single line with the basic structure of the power system. The transmission system is composed by transmission lines that extend for large 29/152 . Equation (2. Because the 3p is the main flicker contribution and these relatively high frequencies are approximately uncorrelated this is a reasonable assumption. the large active power variations can interact with the frequency controllers in the conventional power stations. There are several issues arising from largescale wind power integration.The Pst and Plt are assumed the same because it is assumed that the mean wind speed and turbulence will be maintained in 10 minutes average as well as in 120 minutes. so frequency variations can happens. so the flicker is not a linear sum of all flicker produced from each wind turbine. they demand special investigations of voltage and frequency variations. In either condition. the power quality and system stability assessment become more complex and depending on the sizes. The largescale integration means a relatively high wind power compared to the local power system. but here the focus is on the voltage stability and the dynamic power oscillations during normal operation of the power system. Loads Loads Loads Loads Generation System Transmission System Distribution System Figure 2. Figure 2. With high wind power capacity installed. which comes from the wind dynamics in the wind farm that is not correlated.
task mainly supplied by conventional generators and controllers installed throughout the power system. The voltage controllers are mainly related to the reactive power while the frequency controllers to the active power ([23] and [27]). which has none or little voltage and frequency control capabilities and they supply an intermittent power. Voltage control Vterm Ufield GS Rtran Xtran P +j Q Vterm q term Frequency control fterm Xcontrol Voltage control Figure 2. The distribution lines require special attention to control the voltage at the loads. On the other hand. the apparent power supplied to the load (P+jQ) flows through the transmission and distribution system from the generators stations – GS. The power system must supply a reliable and quality electrical power to the loads. On the generation stations. wind turbines in general use asynchronous generators that demand reactive power from the network to its excitation. Wind turbines are a special kind of generators. In addition. active controllers compensate the voltage and frequency variations keeping the power quality within limits. the power system must have reserves and controllers that can deliver the power when it is demanded.6 Single line equivalent of a Power System. In order to achieve reliability.6 presents a simple equivalent of the entire power system including the main controllers where only conventional equipments are included. Figure 2. The transmission lines demand special consideration in controlling the voltage at the terminals due to reactive power flow (in AC type lines). assuming that the power is available. In Figure 2. The reactive power demanded to the wind farms is partially compensated by capacitor banks and the network supplies the rest of the reactive power. it leads to continuous action of the frequency controllers to keep the balance on production and consumption (and the frequency constant). 30/152 . Distribution systems delivery power to the loads where the voltage level is lower. The active power produced from wind farms varies all the time and leads to continuous power flux variations. The power flux results in voltage variations compensated near to the loads with decentralized voltage controllers – by adding or reducing reactive power – and in the generation stations with voltage controllers that change the excitation level of the synchronous machines.distances and interconnect different generation units.6. The power system quality and stability depends mainly on the power system controllability [23]. Variations on the active power result in frequency deviation that speed governors act to keep the balance on consumption and production increasing or reducing the primemover power.
Z is the impedance characteristic to the specific node (also called short circuit impedance). U0 is the infinite node voltage. the voltage collapse occurs if after an increase in load or power injection. The voltage instability event can grow to voltage collapse leading to entire or a large part of the power system with very low voltage profile. above. the IEEE Task Force Report in [29] defines the voltage stability in terms of the ability of maintain voltage so that when load is increased.5) 31/152 .2. reducing instead of increasing the power consumed [29]. The voltage difference between the two nodes can be defined as: U −U0 = Z ⋅ S∗ U∗ (2. The voltage collapse in general results from an incident of voltage instability. Although voltage stability definition is not widely accepted.7 Simplified transmission line equivalent diagram. several task forces have worked on basic definitions of voltage stability.1 Voltage Stability Problem A definition to the voltage stability phenomenon has not been widely accepted yet. where disturbances may be a simple load increase or a variation in power from a wind farm. Here. a simple formula based on the load flow calculations is introduced. the voltages are below acceptable levels followed by a progressive and uncontrollable decline in voltage [29]. Nevertheless. For instance. the voltage collapse is well recognized.2. load power will increase hence voltage and power are controlled. The voltage stable operation means that the voltages near to loads are identical or close to the predisturbance values [27].4) Assuming U0 real and rewriting Equation (2.7 presents a single line diagram used to define simple analytical equations to voltage stability (please note that the shunt elements are not included).4) as: U ⋅U ∗ = U 0 ⋅U ∗ + Z ⋅ S ∗ 2 U R + U I2 = U 0 ⋅ (U R − jU I ) + (R + jX ) ⋅ (P − jQ ) (2. The voltage instability phenomenon is defined here as having crossed the maximum deliverable power limit. the mechanism of load power restoration becomes unstable. in order to illustrate the voltage collapse. Here. S=P + j Q Reference Node U0 ∠0° Z=R+jX U∠δ° Figure 2. Figure 2.
In addition.6) results in the voltage at the node as follow: 1 1 2 U = ± − H I − H R + jH I 4 2 ( ) (2. • If HR – HI2 = –1/4 – both solutions coalesce and the point of voltage collapse is reached. even to a very simple case. and the voltage at the sending node.6) and inserting in Equation (2.8) For the sake of simplicity.. a function H is defined as H =ZS*= HR + jHI. The voltage stability is complex and.7. Isolating the imaginary part of the voltage (UI) as: UI = (PX − QR ) = H I U0 U0 (2. • If HR – HI2 > –1/4 – the solution is double: one physical and one spurious (unstable). it is assumed that the infinite node voltage U0 =1 p.jUI. then Equation (2.7) The real part of the voltage in the node can be defined as: H 2 1 2 I − H U R = U0 ± U0 − 4 R U0 2 (2.u.8) to different load factor conditions.9) Adding the real part of the voltage in Equation (2.10) Now it is possible to state that: • If HR – HI2 < –1/4 – there is no physical solution. 32/152 .8) becomes very simple as: 1 1 2 UR = ± − HI − HR 4 2 ( ) (2.9) and the imaginary in Equation (2. includes the load.and remembering that U* = UR . it is possible to illustrate the maximum power transferred to a specific node as a relation of the voltage at that node (Figure 2. network characteristics. Using the diagram in Figure 2.5): (PX − QR) − U 0 ⋅ U R − (RP + XQ) = 0 U + U0 2 2 R H U + I − U0 ⋅U R − H R = 0 U 0 2 R 2 (2.
Hence.U P/Pmax Figure 2. generator reactive power/voltage control limits [35]. load characteristics [36]. The reactive power supplied from shunt capacitors is related to the squared of the voltage. distance to voltage collapse. e.8 Power transfer to a node as function of the voltage (“nose curve”). This curve is the Voltage vs. 2. when started a voltage decline it will reduce the local reactive power production stressing even further the transmission lines and reducing further the voltage level. When transferring the problem to the complex power system. shunt capacitor banks compensate the reactive power in the power system.1 Analysis of Voltage Stability The analysis of voltage stability for a given power system involves examination of several aspects. Usually.8 is a simple illustration of the relation between the power transmitted to a node and its voltage. Finally.g. Power characteristic of the node also called “nosecurve”. Similarly. Under this condition.8. the relevant factors that lead to voltage instability are: the transmission lines and power transfer strength of the power system. the loads response to voltage changes influences the voltage stability. large number of wind farms on power systems (high penetration) demands reactive power and in addition some synchronous generators (generation stations) are shut down in order to achieve cheaper energy production. the voltage stability limit is characterised by the vertical tangent at the nose point that is in fact the maximum power transmitted to the node in agreement with definitions above. Figure 2. and the action of voltage control devices such as under load tap transformers [37]. Distribution grids uses under load tap changer transformer. the characteristics of the reactive power compensation devices with the voltage contribute to the voltage instability. The dispersed voltage controllers acting on the distribution grid also influence the voltage stability. the excessive demand for reactive power can be a problem.1.2. tends to increase the problem by increasing the current flow in order to reestablish the set voltage level. characteristic of reactive power compensation. which under voltage instability events. mechanisms that lead to voltage collapse 33/152 . Using Figure 2. In addition.
Secondary control is slow control actions to reestablish nominal system frequency and scheduled power interchanges. the loadability curve is also used to define the maximum wind power to the power system. The loadability curve indicates the maximum load increase in the power system under specific conditions. however. being a dynamic problem rather than static [29]. A comprehensive reference list can be found in [28]. the Jacobian (NewtonRaphson algorithm) represents a linear relation between the power and voltage at a specific operational point. However. the voltage stability is defined in terms loadability curves to a specific node before voltage collapses [28]. the voltage instability is analysed by static models. 34/152 . which. The voltage stability here includes periods from 15 minutes to hours.2. The load flow programs determine the operational characteristics of the power system based on the load schedule and voltage reference in the generation units. Here. The speed governor acts based on speed deviation where. This thesis focuses on the primary control of frequency because it is the controller acting instantaneously to avoid frequency deviations related to stochastic variations from wind turbines. the maximum load is defined. The voltage stability problem has been discussed in a large number of papers and always analysed by means of expensive and complicated models. The loadability curves are similar to the injection of wind power but the power to the node is injected instead of a drained. The loadability curves are similar to the “nosecurve” in Figure 2. The loadability curves are computed using a loadability computation tool based on a load flow program using NewtonRaphson algorithm. the power is naturally shared between different generators based on the rate of the rating of the generators or as defined by the system operators. Here.among others [27]. at that point. In a synchronized system. The load flow problem is very closely associated with voltage stability analysis [27]. the maximum power transferred was reached and the Jacobian becomes singular. The use of the Jacobian properties has been pointed by several authors where modal analysis and voltage collapse proximity indicators have been proposed [29]. At that point. The primary control is fast control actions to keep the instantaneous balance between production and consumption. as the dynamics involved in the voltage stability problem are very slow.8. The frequency control in 10 minutes can be classified in primary and secondary [38]. The wind power to a node is increased until the maximum power transfer is reached. When the voltage collapses.”[39]. The load to a specific node of the power system is stressed until the Jacobian matrix becomes singular and. 2. “(…) most automatic controls use high gain negative feedback. which automatically adjust the prime movers driving the generator to keep the balance on consumption and generation. which was implemented in Matlab as part of this project.2 Frequency Control Problem The power system has a nominal frequency for which all generators are synchronized. In load flow problems. the lower part (unstable part) of the curve is not simulated here. by its active nature. can cause oscillations to grow in amplitude with time. The primary frequency control is done by speed governors. there is no solution to the load flow problem and that is the maximum transmissible load.
y is the output (or measured) variables. These equations can be linearised on an operational point. [24]. The power system can be described by equations that include: all electromechanical characteristics. The power oscillations can be sustained due to the controller natural characteristic. This fast action tends to reduce the damping of the system oscillations hence special Power System Stabilizers – PSS – were designed to damp those oscillations [39].11) where. Although the oscillations are not expected to increase in time. The modal analysis has been recognized as one of the most reliable tools to analyse power systems ([23] and [25]). Where. Hence. [25] [39] [40] and [41]).Originally. 2. the sustained variations can become a problem. B is the input matrix. the interconnected generators were fairly close to one another. The eigenvalues of A represents the roots of the state equations defined as the values of s that satisfies: 35/152 . A is the state matrix. Oscillations are a characteristic of the power system. which are initiated by the normal small changes in the system load and. c is the output coefficient matrix and d is a matrix of coefficients describing the direct connection between the input and output variables: the vector u. the network equations and the controllers. Dampers windings on the generators were used to prevent the oscillations to grow. from wind farms power variations. x is a vector of the dynamic state variables. The power system is very complex and specific components can interact with other causing oscillations. and oscillations were at frequencies of the order of 1 to 2Hz ([23] and [24]). Increasing the demands for reliability makes the rapid automatic voltage controllers that are used to prevent the generator loosing synchronism following a system fault. The modal analysis however must be used with caution because it does not include the nonlinear behaviour of equipments on the electrical power system. The wind farm power produced can be viewed as a continuous negative load variation that demands the speed governors to act all the time to keep the balance on the system. From an operating point of view.2. it is also important to include dynamic simulations to analyse the power system oscillations during normal operation. in this report. oscillations are acceptable as long as they decay. With the modal analysis it is possible to define the eigenvalues of the power system hence it is possible to identify the electromechanical oscillation modes. The modal analysis uses a linear representation to the entire power system being suitable to analyse small disturbances in the power system (it is also called small signal stability analysis of the power system).2.1 Analysis of Power System Oscillations The power system oscillations analysis may be done by modal analysis complemented by dynamic simulations ([23]. the linear model put on the form of a set of first order differential equations with constant coefficients (state equations) have the form: & [x] = [A]⋅ [x] + [B]⋅ [u ] [ y] = [c]⋅ [x] + [d ]⋅ [u ] (2.
Hence. The modal analysis characterizes the modes of oscillation of the power system and the less damped oscillation modes. the corresponding mode will grow exponentially in time being unstable. The problems were classified in smallscale and largescale integration. Dynamic simulations including the entire power systems are very expensive.2 presents the main integrations problems from wind power [20]. the modes that the respective eigenvalues are close to the positive plane. If all eigenvalues have a negative real part all modes decay with time and the system is said to be stable. the modal analysis must be accompanied with dynamic simulations in order to include the nonlinear characteristic of the power system [39]. In order to analyse the main characteristics of the power system. The dynamic reduction is an aggregation of machines and loads with similar dynamic performance that reduces the number of equations to describe the power system and does not lose important information [26]. i.12) where each solution si defines a time function e(si t) that satisfies the state equations [25].3 Remarks on Wind Energy Integration The integration problems caused by wind power have been discussed in this chapter.e. the possible problems in the frequency control in the power system are characterized.2 Main power system influences from the wind energy integration Integration scale Problems Causes Wind speed variation Peaks of wind speed Peaks of wind speed Dynamic operation of wind turbines Switching/start up operation of generators In rush current due to switching operations of generators Power electronic converters Inability of the power system controllers to cope with the power variations from the wind farm and loads Reactive power limitations and excessive reactive power demand from the power system Steady state voltage rise Overcurrent Protection error action Flicker emission during continuous operation Flicker emission during switching operations Voltage drop Harmonics Power system Oscillations Voltage stability Small Scale The smallscale wind power integration into power system is well investigated and there are plenty of tools available to analyse its interaction. The function is called a mode of oscillation of the system. In this thesis. The problems from connecting 36/152 Large Scale . However. dynamic reduced models are used as pointed in [26] and [42]. 2. in addition to the modal analysis.det[A − si I ] = 0 (2. dynamic models to largescale wind power must be applied to the power system dynamic simulations to analyse the interactions between the wind power and frequency controllers during normal operation of the power system. Table 2. Hence. If one eigenvalue has a positive real part. dynamic reduced models are used to analyse the power system performance with wind power. Table 2.
The largescale problems include all problems from the smallscale and in addition problems of voltage stability and power system quality. Time feasible in order to analyse large power system. the frequency can deviate from its limits as well. Large disturbances such as short circuits in sites with large amount of wind turbines can result in voltage transient instability due to the reduced capacity in transferring reactive power and the huge demand from the asynchronous generators as well as transformers after the fault is removed. Network characteristics play an important role on the power system stability and quality. Proper representation of the power flux of the wind turbine generator.small numbers of wind turbines to the power system are related to the voltage quality and thermal capacity of the lines and cables.g. The voltage quality deviates because of the power flux in the network. the power system models must have the following characteristics: 37/152 . The main conclusions are: • • • • • Smallscale integration of wind energy deals mainly with voltage quality and reliability. Moreover. selfexcitation of the induction generators. The wind turbine models must have the following characteristics: • • • • • • • • Dynamics models for the wind speed acting on each wind turbine. The Voltage stability problem is related to the limitation of power transfer. Include proper models to the electrical components. Detailed model to the wind turbine dynamics. however. Large integration of wind turbines includes the smallscale problems and overall stability problems. Be flexible to allow the implementation of wind turbine/farm models and interact with other simulation tools. The power system quality can deviate from limits. Include models to the loads. Wind turbines power characteristics play also another important role in the overall power system stability and quality In order to investigate the wind power influences on the power system quality and stability it is imperative the use of appropriate models to represent the wind farm and the power system. The largescale integration of wind farms. and in largescale integration. The spatial wind coherence in largescale. The analysis of largescale integration demands suitable models to represent wind power. it is not properly investigated yet and it is very complex. The extensive use of capacitors banks to compensate reactive power can result in larger problems in event of islanding e.
Allow the implementation of large power systems.• • • Input/output capabilities in order to analyse the data. In large integration. represent properly the electromechanical modes in the power system. 38/152 .
A large part of the complexity resides on the input: the wind. Returning to Figure 3. The wind is classified in two main parts: the first part represents a mean wind speed profile over the rotor area. here assumed as stochastic. The main objective of this chapter is to present a dynamic wind model for power quality assessment of a three bladed upwind horizontal axis wind turbine type. The blades pass through different wind 39/152 . Wind speed Spatially correlated turbulence Mean wind speed profile Figure 3.1. and the second part is turbulence on top of the deterministic (see Figure 3. here defined as deterministic.1). Figure 3. The wind is complex and the blades crossing the wind field modify the power fluctuations. The main source of power variation on conventional wind turbines is the wind speed variation. The deterministic part of the wind over the rotor area is assumed “constant” in 10 minutes period and the wind variations are the time variant part that has a stochastic behaviour. the wind acting on each rotating blade section is different from the wind in a stationary reference frame.Chapter 3 3 Wind Speed Model Wind turbines produce a complex and continuously fluctuating power. This assumption is valid only in periods up to few minutes because above this there is a slow wind variation in the mean wind speed due to continuous change in the atmosphere. The wind speed model includes the turbulence and tower shadow in the rotor area.1 Illustration of the wind on the rotor area of a wind turbine [43].1 illustrates an example of the wind field acting on the rotor area of a wind turbine (borrowed from [43]).
in this case within ±10kW. It is possible to decompose the measurements in slow and fast power variations where the slow power variation includes frequencies up to 0. Figure 3.2 presents a measured time series of active power produced from a 500kWstall regulated. which is the main responsible for high standard deviation of power produced from wind turbines.5Hz and the fast power variation includes all frequencies above 0. The high frequency power variations cannot be directly related to the turbulence but to the dynamics of the wind turbine including the blade rotation in the wind field. In this particular case.3). The process of passing several times through this wind speed field results in power variations at n times the rotational speed revolution of the rotor called np’s. 40/152 . threebladed wind turbine type (direct connected to the grid). The Equivalent Wind Speed (EWS) model simulates the wind by an equivalent time series that when applied to a specific aerodynamic model reproduces the aerodynamic torque from a wind turbine.2 the active power measured from a three bladed wind turbine vary in a broad frequency range. Figure 3.speeds in each revolution. The high frequency power variations are much smaller. The aim is to reproduce the mechanical torque using an equivalent wind speed model to the entire rotor. This chapter presents the general main windrelated power output variations from wind turbines.2 Power produced by a 500kW stall regulated wind turbine. In Figure 3. which applied to a drive train/generator model simulates the electrical power from a real wind turbine (Figure 3.2 also shows the power variations classified in fast and slow where the mean value is removed. 350 300 250 200 Active power (kW) 150 100 50 0 50 100 150 0 Measurement Slow variations Fast variations 100 200 300 Time (s) 400 500 600 Measurements Low frequency power variations High frequency power variations Figure 3. The low frequency power variations are the largest variations and they are related to turbulence on the rotor area as it has been recorded in anemometers. The relatively high frequency power variation has been related as one of the main responsible for the flicker problems [60] as introduced in chapter 2. the slow power variations are within +80kW and–110kW. The dynamics of the wind turbines are explained in chapter 4 where the wind turbine components are introduced.5Hz.
In Figure 3. W is the wind speed over the rotor area.W T(t) W(t) Drive train/ generator Pel(t) Aerodynamic model Wind speed Aerodynamic torque Electric Power Wind turbine Network Figure 3.5 2 2. Above that frequency.1 Model Description Measurements of turbulence in a stationary reference frame show that the amplitudes of the wind speed variation reduce with the frequency. the wind speed variations are insignificant [44].3. In the low frequency range from 0 to 1. The wind speed variation on multiples frequencies of the rotor (that in this case is 0.4) shows that each blade section of the wind turbine experiences high wind speed variations on the rotational speed of the rotor.5 5 Figure 3.4 presents the Power Spectral Density (PSD) of the wind measured on a rotating blade section [45].5Hz are the main wind variations. the rotation of the blades causes variations in the effective wind acting on each blade section on n times the rotor speed frequencies. Above a few Hertz. The PSD (in Figure 3. T is the aerodynamic torque and Pel is the electrical power transferred to the utility network.3 General overview of wind turbine models. However.5 3 Frequency (Hz) 3.5 Hz) is clear. 3.5 1 1. Figure 3.5 4 4. 10 2 PSD of wind measured on rotating blade section 10 1 10 0 ((m/s)2/Hz) 10 −1 10 −2 10 −3 0 0.4 Wind speed measured on a section of a rotating blade [45]. 41/152 . the wind speed variations are very reduced mostly because of the dynamics of turbulence.
The torsional drivetrain mode (dynamic from the wind turbine) is around 1.5 PSD of measured electrical power output of a 500kW stall regulated wind turbine. At the frequency of 0. hence when adding up the wind effects on the remaining others two blades in a symmetrical threebladed rotor of a wind turbine. first of all the 10 minutes period avoid most of the cyclical variations related to meteorological phenomena (e. which is relevant to the power quality analysis hence it is retained in the dynamic wind model. At this frequency. The 10 minutes period was selected for several reasons.5 correspond to the measured time series of power presented in Figure 3. The complex mode is related to a dynamic interaction between the electrical generator and the flexible wind turbine. the period matches the time frame of the primary 42/152 . Second. stall regulated and operating with in constant speed. diurnal variations related to periods of hours).2. it is not related to the wind turbulence. the main wind related power variations are from the 0 and 3 times the revolution of the rotor.2Hz). a small contribution is present that is related to some asymmetry in the rotor. a small power variation is presented that is also related to some asymmetry on the rotor. 10 4 1p 3p 10 3 PSD of active power (kW2/hz) 10 2 10 1 10 0 10 −1 10 −2 0 1 2 3 4 5 6 7 8 9 Frequency (Hz) 10 11 12 13 14 15 Figure 3. all the frequencies different from 3 times the rotor speed (3np) are cancelled. The PSD in Figure 3. At the frequency range from 6 to 11Hz.5 presents the PSD of the measured electrical power from a three bladed wind turbine. Considering the measured PSD of power produced from wind turbines and the measured wind speed on a rotating blade. only the windinduced power variations that are relevant to power quality are treated hence the dynamics of the wind turbine is not included in this chapter but explained in chapter 4. Figure 3.8Hz (2p). It generates 10 minutes period time series. The cancellation effect is expected due to the symmetric position of the threebladed rotor and has been supported by power measured from wind turbines. The fundamental frequency (1p) is approximately 0.4Hz (indicated in the figure). 500kW.g. In this chapter. a small amount of energy is presented. which is related to a complex mode of the wind turbine. The wind speed model represents the power variations from the wind in the frequency range of interest as a time domain model. The most relevant power variation is at 3p frequency (1.The measurements were done on a single section of a rotating blade.08Hz that gives a small power variation.
z) is the deterministic wind speed that is assumed time independent and g(x. e.t) is the stochastic part that has mean value zero. 43/152 . z . Moreover.7 illustrates the principle of the EWS.y. y . Figure 3. are based on 10 minutes measurements In 10 minutes period. which its performance with large amount of wind power is one of the aims of this report.y.frequency controllers from power systems. so the zaxis is negative in the direction to ground.y.z. t ) = U ( x. y . z ⊕y z x y x y z x Front view Figure 3. Top view 3.z.1) where W(x. The EWS applied to an aerodynamic function simulates the real torque on the main shaft of wind turbines. statistics and flicker) in [46] as well as wind turbines characteristics. power curve.g. z ) + g ( x .6 Reference axis used in the wind turbine. U(x. t ) (3. y . the stochastic and deterministic parts can be combined to express the total wind at one position as: W ( x .1 Equivalent Wind Speed Model The Equivalent Wind Speed (EWS) model is an equivalent for the entire rotor that takes into account the stochastic and deterministic wind speed actions on the rotor area. The origin of the coordinate system is the centre of the nacelle precisely at the height of the centre of the rotor. Figure 3. the voltage quality assessment (e.6 presents the references axis and notation of the wind turbines.1. z .g.t) is the total wind speed in position xyz at time t.
which is to reproduce the same mechanical torque using a single equivalent wind speed time series applied to an aerodynamic model.7 illustrates how the power is produced in a real wind turbine. Because the rotational speed depends on the electrical generator. the aerodynamic torque enters a dynamic model of the drive train and electrical generator. acts on the blades of the wind turbine. Finally.2 Description of the Equivalent Wind Model The wind acting on the rotor area of a wind turbine can be modelled in several ways. an aerodynamic module calculates the aerodynamic torque on the main shaft. it is assumed that the electrical power produced by the EWS is the same as the one produced in a real wind turbine. The aerodynamic blades convert the wind by complex aerodynamic effects on mechanical power on the mechanical shaft of the wind turbine (aerodynamic torque (T) times the rotor speed). The lower part of Figure 3. The upper part of Figure 3. in the grid wind model. which is different at each rotor position for each time. This process is very common and reliable. The EWS replaces all wind time series on the rotor area and a relatively simple aerodynamic function supplies the aerodynamic torque. The mechanical power is then converted into electrical power in the electrical generator. Based on the continuous wind speeds in a number of blade sections. however. The EWS aims to simulate the electrical power output variations from a threebladed wind turbine.7 illustrates the main idea of the EWS. 3. The objective of the model defines the type and details to be retained of the wind model.1. The main advantages of the 44/152 . which will output the electrical power. Yet. it consumes much simulation time [47].7 Equivalent Wind Speed Model principle. an aerodynamic routine interpolates between the appropriate points of the wind grid to generate a continuous wind speed at a given section (radius) of the rotor blade. The wind speed (W). Usually the wind models assumes that the continuous wind field on the rotor area can be replaced by discrete time series of wind speeds in a grid of points on the rotor area.Aerodynamic torque W T(t) W(t) Drive train and generator Pel(t) Wind Speed Aerodynamic forces Equivalent Aerodynamic torque Electric Power W eq Equivalent wind speed Ψeq Aerodynamic model T(t) Weq(t) Aerodynamic torque Drive train and generator Pel(t) Electric Power Wind Wind turbine Network Figure 3. it generates many time series that are only used partly and generally.
It must be noticed that in the following equations polar coordinates is used: r.φ) using: T (t .7.equivalent wind speed model are: fast computation and reduced memory requirements.2) provides the expanded expression for M(t. the expansion of the wind speed field formally becomes much simpler as given in Equation (3.φ) in: M (t . r )e jnφ (3. The single equivalent wind speed model is also very suitable for simultaneous simulation of a large number of wind turbines making it possible to efficiently estimate the impact of a large wind farm on the power quality. The effect of all three blades are then added to give the total aerodynamic torque T(t.3) For the summation in Equation (3.φ are the rotor position in polar coordinates where r is the radial position and φ is the angular position of the blade section. 3 (3.3). r ) is the complex azimuth expansion coefficient. r.2). and r0 is the inner radius.φ ) = ∑ M (t . it can be assumed that the azimuth position φb of blade number b is given by: φb = φ + b = 1.4) describe how the aerodynamic torque is generated physically. r . an aerodynamic coefficient (Ψ(r)) on a given blade section r is calculated.5) into (3.4) The Equations (3.3 2π (b − 1).2. The equivalent wind speed is based on an expansion of the wind speed field in the azimuth angle (φ). φ) are used to calculate the contribution from a single blade to the aerodynamic torque M(t. W (t .φ ) = n = −∞ ∑W ∞ ~ { n} (t . Applying blade element theory.5). The Equations (3. where the aerodynamic forces start to develop. r . M (t .φ) according to Equation (3. A similar expansion was done by Madsen and Rasmussen [48] using real expansion coefficients. Ψ(r) and W (t. Using complex expansion coefficients. This azimuth expansion coefficient indicates the amplitude of the nth harmonic in the rotational speed [49].6) 45/152 .2) where R is the length of the blade. Inserting (3.φ ) = n = −∞ ∑M ∞ ~ { n} (t ) ⋅ e jnφ (3.2) to (3.4) below provide a more formal description of the aerodynamic model for a 3bladed wind turbine shown in Figure 3.φ ) = ∫ Ψ (r ) ⋅ W (t .5) ~ W{n} (t .φ )dr r0 R (3.2) to (3.φb ) b =1 3 (3.
φ ) = n = −∞ ∑T ∞ ~ {3 n} (t ) ⋅ e j 3nφ (3.9) From Equation (3.φ) according to: T (t .φ ) with the equivalent aerodynamic influence coefficient Ψeq defined according to: Ψeq = 3∫ Ψ (r )dr r0 R (3.~ the azimuth expansion coefficients M {n} (t ) are given in: ~ ~ M {n} (t ) = ∫ Ψ (r ) ⋅ W{n} (t .10) ~ with the weighted azimuth expansion coefficients Weq.8). φ ) = n = −∞ ∑W ∞ ~ eq{3 n} (t )e j 3nφ (3. and further using Equation (3. which e.7) into Equation (3.φ) generates the same torque Teq(t. This effect will be complete if the rotor is symmetrical.φ) = T(t. r )dr r0 R (3.φ) as the wind speed field v(t. Teq(t.13) Using the definitions in the Equations (3.7) Inserting Equation (3. the equivalent torque Teq(t.φ ) = Ψeq ⋅ Weq (t .e.3). Substantial first harmonics have been measured on wind turbines with unsymmetrical rotors typically due to pitch misalignment.{3n} (t ) given in: ~ Weq.8) ~ where the azimuth expansion coefficients of the torque T{3n} (t ) are given as: ~ ~ T{3n} (t ) = 3 ⋅ M {3n} (t ) (3.{3n} (t ) = r0 ∫ Ψ (r ) ⋅W R r0 R ~ {3 n } (t . r ) dr ∫ Ψ (r )dr (3.φ).φ). requires that the blades have the same tip pitch angle. the summation of the contributions from the three blades remove harmonics which are not multiple of three. it can be shown that the equivalent wind speed Weq(t. Now defining the equivalent wind speed Weq(t.g. i.φ) according to: Weq (t .11) Finally.12) (3.4) yields the azimuth expansion of the torque T(t. 46/152 .10) to (3.φ) is defined as: Teq (t .13).
the implementation of EWS model is presented after having introduced the main deterministic and stochastic components. i.1. i. When the wind comes close to the tower.φ ) (3. Horizontal axis wind turbines always have some form of tower support structure. it has been assumed that the contribution to the torque is proportional to the radius. Towers are obstacles to the free wind that modifies the wind flow. however. The potential flow for the wind is constant all places around the tower. fact that has been observed in different blades and have been justified in [50]. 3. [61] and [51]). hence the relevant effect is the upstream flow.e. the tower deviates the wind as shown in Figure 3. The horizontal longitudinal (yaxis) is the most relevant to power variations because it turns to be the wind converted into active power. sto (t . small wind turbines use lattice towers and large new wind turbines use cylindrical towers. The EWS implementation is based on Equation (3. It will only influence the dynamics of the wind turbine.In principle.φ ) = U eq .10). The tower can be lattice or cylindrical. The equivalent wind speed is the sum of the deterministic and stochastic components.8.1.8). Usual.14) The deterministic part. Mathematically the wind field can be calculated by approximating the tower to a cylinder and assuming the existence of a two dimensional potential flow ([13]. The wind far from the wind turbine is called the ambient wind speed (Uh) and it is not interfered by the tower. The upstream flow is reduced in front the tower and increased laterally. det (φ ) + g eq . the aerodynamic load influence coefficient Ψ(r) has to be determined individually based on the geometry and hence the distribution forces along the blade radius r. because the blades rotate. is independent of the time.3 Deterministic Part of the Wind 3. In the present model. an assumption that is supported by measurements. i. Weq (t .e. Here the model is limited to the effects on horizontal upwind wind turbines type. Ψ(r)=k⋅r and r0 is 10 % of the radius R. Downstream the tower makes a wake effect that reduces the horizontal wind (Figure 3. and can be expressed as [51]: 47/152 . the EWS only includes the 0p and the 3p harmonics components because they are the most relevant to power quality assessment.e.1 Tower Shadow Tower shadow along with mean wind speed are the most relevant deterministic effect on the electrical power output from threebladed wind turbines. as explained before. These assumptions are based on the physical aerodynamic performance of the blades that excluding the region in the beginning of the blade (root) and the region near to the tip presents the torque proportional to the radius.3. The flow is then decomposed in the longitudinal and lateral wind components. the wind in the y direction and x direction respectively. In the following sections.
Finally.16) and (3.19) where φ is the rotor azimuth position angle.16) ∂ψ wind = Uh 1 + ∂d ( a2 d ts 2 )sin(β ) Uht Uhr dts x a y (3.16) and (3.20) The tower is finite.8.8 The tower shadow effects on the horizontal wind (top view). U hx U h = 0 ( )sin(2 ⋅ β ) for 0 ≤ φ ≤ π a2 d ts 2 for π ≤ φ ≤ 2π (3. β is the angle between the blade section and the rotor axis and dts is the distance from the tower centre to the blade section expressed as (3. thus the tower shadow will only influences the semi plane up to the tower height in terms of azimuth angles (φ) the tower influences the range from [0 π].17) (from [51]): U hr = − U ht = 2 1 ∂ψ wind = −U h 1 − da 2 cos( β ) ts dts ∂β ( ) (3.17) are converted to the xy axis (longitudinal and lateral components) of the wind turbine.18) U hy U h 1 − a 2 cos(2 ⋅ β ) for 0 ≤ φ ≤ π d ts = U h for π ≤ φ ≤ 2π 2 ( ) (3.15) ψwind is the potential flow of the wind in polar coordinates where dts is the distance of the blade section to the tower centre and β is the angle between the blade section and the horizontal flux.17) Uh Upstream β Tower Downstream Figure 3.20). d ts = x 2 + y 2 2 (3.a ψ wind = U h sin( β ) ⋅ dts − d ( 2 ts ) (3. 48/152 .The radial and tangential wind speed components can be derived from the potential flow as Equations (3. Decomposing the wind in the rotor area in radial – Uhr – and tangential – Uht – components as illustrated in Figure 3. the wind speed components presented in Equations (3.
94 0 pi/2 pi Equivalent Moment Single blade moment 3pi/2 2pi 3 x 10 −3 FFT of the normalized torque 2. it has been assumed that the wind is not modified by the tower presence because only the horizontal component of the wind is modelled.96 0. Only the horizontal part of the wind was modelled in the first place because it is the most relevant to the aerodynamic process of generating lift hence converting the wind power into mechanical power.5 1 0.) 0.9 illustrates the tower shadow effect on the wind speed field in front of the rotor on an upwind type rotor and a cylindrical tower. 1.99 0. 49/152 .5 2 1.01 1 Moment (p.97 0. Figure 3. Figure 3.Figure 3.10 illustrates the effects of tower shadow on the main shaft torque of three bladed wind turbines rotors.10 Normalised torque influenced by the tower shadow.98 0.u. Above the tower.95 0. In this particular case.9 Wind speed field interference by the tower shadow.5 0 0 5 10 15 20 25 Harmonic components (np s) 30 35 40 Figure 3. the tower reduces 20% the wind speed in front of the tower and the tower influences only the lower semi plane.
has a small wind speed variation with the height.1 – 1.0 x 101 0. A smooth site. h is the height of the centre of the rotor. However. the electrical power outputted from a three bladed wind turbine has very little influences from the wind shear because it is quite linear. which can be determined in various ways. with small obstacles.21) [13].10 in accordance to the PSD of the measured power output from wind turbines that shows small fluctuations in frequencies different from 3p (Figure 3. The wind turbine dynamics filter out the higher 3np frequency elements in Figure 3. On the other hand.1. The lower part of Figure 3.5) 3. The wind shear modifies only the vertical profile i. The tower influences each one of the blades generating several harmonic components on the torque. there is no influences on lateral space xaxis or to the wind speed related to the position of the wind turbine –yaxis. U is the wind speed and z0 is the roughness length that characterizes the terrain.21) where z1 and z2 are the heights from the rotor centre (see Figure 3. The wind shear has a strong relation with the site.e.0 x 103 1.0 – 4.2 – 1.0 – 3.0 x 104 0.6 for reference axis). e.1 Typical values of surface roughness length z0 for various types of terrain [53]. acting on the rotor area of all commercial wind turbines [13].0 1. Type of terrain Smooth sea Sand Low grass High grass Forest City Z0 (m) 2. the urban areas present a high wind speed variation with the height. That phenomenon is expected because in smooth areas the friction is much lower. sea with small waves. In general. the wind shear is very important to blade loads analyses but it is not transmitted to the electrical power. U (h + z1 ) ln = U (h + z2 ) ln ( ) ( ) h + z1 z0 h+ z2 z0 (3.4 – 1. The ground/wind friction influences the wind speed up to hundreds meters.1 [53].3.10 presents the time series of normalised moment in a three bladed wind turbine rotor. Table 3.g.0 x 102 0. This process is called wind shear.0 50/152 .0 – 4. which reveals that practically only the 3np components are presented in it.2 Wind Shear The friction between the ground and the moving air generates a vertical wind profile where the mean wind speed increases with the altitude. Typical values for various types of surfaces are given in Table 3. The wind speed relation between two different heights can be expressed according to Equation (3. derived from the Prandtl logarithm law model [52]. one very common is to make comparisons with sites for which z0 has been determined from measurements and decide the most suitable one [13]. The z0 of a terrain is a measurement of the roughness.The upper part of Figure 3.10 presents a Fourier transformation of the normalised moment influenced by tower shadow.
In smooth sites the effects are reduced leading to reduced power variations. 51/152 .12 as well as the reference axis used here. the wind shear leads to higher wind speed variation hence higher cyclic torque variations.11 Wind shear for different sites (ground is reference for height).22) (3. Figure 3. The wind speed difference with the height depends mainly on the site.23) where the variables are presented in Figure 3. on the other hand in sites with tall obstacles.12) changes according to: z p = h − r ⋅ cos(φ ) Hence. The wind turbine blades rotating on the wind sheared leads to cyclical loads on the mechanical torque. Figure 3.Using Equation (3.11 illustrates wind shears for different sites with an average wind speed of 10 m/s at 10 m height.21). 80 Smooth sea High grass site Urban site 70 60 50 Heigth (m) 40 30 20 10 0 0 2 4 6 8 10 12 Wind speed (m/s) 14 16 18 20 Figure 3.11) will suffer cyclic loads because when the blades rotate on the field. the height a blade section (point p in Figure 3.g. each blade section in position p is subjected to the wind speed Up: h − r cos(φ ) ln z0 U p (h − r cos(φ )) = U (h ) ⋅ h ln z 0 (3. A wind turbine applied to wind field sheared (e.
05 (the wind turbine is the same used to compute the tower shadow effects in Figure 3.10). Figure 3.) 1.13 presents the normalised torque variations due to wind shear. Where. The FFT of the resulting torque presented in the lower part of Figure 3.13.12 Reference axis and angles used in the wind turbine. φ is the angular position of a blade. Using Equation (3.1 1.23).13 shows that in 52/152 .05 1 Blade 1 Blade 2 Blade 3 Torque at the main shaft 0 pi/2 pi Azimuth angle φ (Rad) 3pi/2 2pi 0. Moment/Equivalent wind speed (p. p is a point in the rotating blade section distant r from the root and h is the height of the wind turbine.9 8 x 10 −4 FFT of the torque 6 4 2 0 0 5 10 15 20 Harmonic components (np s) 25 30 35 Figure 3.95 0. Normalised torque influenced by wind shear (site with a medium z0). The upper part in Figure 3.13 illustrates the normalised mechanical torque of a three bladed wind turbine from the wind shear in a site with z0=0. The 120° displacements between the three blades cancel the symmetrical torque resulting in cancellation of torque contributions in frequencies that are different from 3np.u.z Xp Swept area Blade rotation x Xp r z y φ Blade representation h π/2 Front view Side view Figure 3.
is the time variant part of the wind acting on the rotor area. The effects from the tower shadow were much more relevant (the FFT in 3p was around 2.4 Implementation of the Deterministic Component The deterministic part represents only the tower shadow effect.14 Implementation of the deterministic model in Simulink/MATLAB. Low frequency variations have large amplitudes and higher frequency variations have lower amplitudes.14. only the cos part of the third harmonic is included because the tower shadow is symmetric. The variation in the 3p frequency is very small. in particular the interference from tubular towers in up wind turbines type. it is neglected. so the imaginary part of the third harmonic expansion coefficient is zero. the turbulence can be measured in terms of its variance: σ 2 = Ε (W − Ε[W ])2 [ ] (3. 3.5 Stochastic Part – Turbulence The stochastic part of the wind. The implementation in Figure 3. The turbulence is also expressed in terms of standard deviation divided by the average wind speed (U0 = E[W]) – so named turbulence intensity – Iu. defined as: 53/152 . 3. Currently.symmetrical rotors the total torque on the main shaft presents mostly 3p frequency variations. Equation (3. the deterministic part of the EWS model only includes the tower shadow and the mean wind speed as it is presented in the following section. which is a feedback from the wind turbine model. The variations have naturally a random behaviour but the air dynamics creates a main pattern on the wind speed variations.10). in this thesis called turbulence. The remaining implementation in principle corresponds to Equation (3.14 first generates the rotor position φ as the integral of the rotor speed. where “detr3rd” is the amplitude from the contribution of the 3rd harmonic calculated based in the geometric parameters of the wind turbine. Recording the definition in Equation (3. In addition. However. Turbulence is the wind speed variations in a broad range from seconds to minutes.1). 1 w rotor 1 s Integrator cos Cos(3u) 1 detr3rd Effect due tower shadow Product Product1 WSdet um Mean Wind Figure 3.1.1.5x103) compared to the effects from wind shear (6x104). “um” is the mean wind speed specified and “Wsdet” is the output deterministic wind time series [54].24) where E[ ] is the mean expectation operator. The implemented deterministic model is illustrated as block diagrams in Figure 3. and W is the wind speed. The wind shear effect is very important in mechanical and fatigue analysis of blade loads but to the power quality.10) is based on complex numbers and for the implementation only the mean value and the third harmonics are included.
the review presented in [55]) try to reveal the more accurate turbulence spectral model.Iu = σ U0 3. Several studies (e. 54/152 .1. The Kaimal power spectra pattern was fitted based on several experimental data collected with neutral atmosphere over flat homogeneous terrain in Kansas as explained in [44]. The choice for Kaimal is done because it is widely accepted and because the Kaimal is suggested according to the Danish standard for loads and safety of wind turbines construction [56].15. defined as: f ⋅ S( f ) f ⋅ xL U0 3 1 + 1. Kaimal PSD of Turbulence. a socalled Kaimal turbulence model is used to represent the power spectral of turbulence [44].g.26) where f is the frequency of the turbulence. Spectral density function of the turbulence has been investigated for several years and several models have been proposed.1 Power Spectral Density of Turbulence (3.25) The turbulence varies in a broad frequency range. the Kaimal spectrum was selected to model turbulence.15 presents the logarithm power spectra (S(f)*f) of the turbulence simulated with Kaimal model to an average wind speed of 10m/s. Figure 3. Here. hence. the turbulence variations tend to zero in high frequency.5.5 ⋅ f ⋅ x L U 0 5 σ2 = (3. In this report. the application of other model is straightforward. σ2 is the variance. turbulence intensity of 10% and turbulence length scale 600m. However. The Kaimal power spectra has a peak when xL⋅f/U0=1 and decreases with the –5/3 of the frequency. 10 −1 10 −2 10 −3 10 −4 ((m/s)2) PSD*f 10 −5 10 −6 10 −7 10 −8 10 −9 10 −10 10 −11 10 −3 10 −2 10 Frequency (Hz) −1 10 0 10 1 Figure 3. xL is the turbulence length scale (proposed maximum 600 m) and U0 is the average wind speed.
The rotational sampling of turbulence has been studied for several years and means to simulate it has been proposed by different ways.3 Power Spectral Density of a Rotating Blade – Rotational Turbulence The blades rotation generates special power variations. which is explained in the following subsection. the blades rotation will move energy contents of the turbulence from low frequency to multiple frequencies of the rotor speed. This method demands generation of several correlated time series of wind speeds. a process usually called “rotational sampling of turbulence”[59].27) where γ is the coherence.5.5. 55/152 . One of the methods includes the generation of several correlated time series of wind speed on the rotor area of the wind turbine.3. Wind turbine blades pass through different turbulences several times. The method used here is based on equivalent wind speeds to each rotational sampling frequency. The Davenport type coherence has the following form [57]: γ ( f . Near positions are well correlated but it reduces as the distance increases.d)= e −k d⋅ f U0 (3. so called wind grid model. Then interpolation of appropriate points generates the wind on a rotating blade section. Considering the application to wind turbines and the Danish standard for loads and safety of wind turbines construction [56]. which result in large memory usage. Measurements in other sites however reveal a different constant value that is in the range from 2 to 27 that is mostly related to the ratio d/h (distance divided between the two points and the height of the measurements (h)) as presented in [55]. U0 is the mean wind speed. d is the distance between two positions and k is a decay factor.2 Coherence of the Wind The coherence of the wind is relevant because it shows the correlation between wind speeds in different positions. these characteristics have been measured as presented in [57]. The rotational sampling of turbulence has been reported as the main cause of flicker from wind turbines.7 [57]. f the frequency. It results that the power spectra of the turbulence on a blade section of a wind turbine has high energy content in the multiples of the rotational speed of the wind turbine. the EWS uses 12 for the decay factor constant. turbulence is highly correlated in lower frequency and decreases as the frequency increases. The EWS uses the Davenport coherence type to simulate the coherence in the rotor area.1. the distance between the positions and the frequency.1. which makes its inclusion on wind speed model relevant to power quality assessments [60]. The correlation of the wind in two different locations depends on the average wind speed. where the appropriate points comes from the expected position of the rotating wind turbine blade as presented in [47]. As for the deterministic part of the wind. 3. In similar way. The decay factor has originally been expressed as a constant equal to 7.
multiplied by an admittance function Fw{3n} (∆ω ) .e.6823 ⋅ dTF ⋅ s 2 + 7.7869 ⋅ dTF ⋅ s 2 + 0. d ) cos(φ )dφ 0 (3. the admittance function is represented by a second order filter with a transfer function Hadm{0p} fitted in [50].9904 H adm{0 p} ( s ) = 2 7. The turbulence wind speed is simulated by applying a “Kaimal” filter to a random number generator.10) (only 0 and 3p). and the PSD of the nth azimuth expansion coefficient S w{3n} (∆ω ) is equal to the PSD of the wind speed in a fixed point S w (∆ω ) (here Kaimal). The azimuth expansion coefficients are random signals [49]. first. the azimuth expansion coefficients are simulated. The filter used in the EWS was fitted in [50].: ~ S w{3n} (∆ω ) = Fw{3n} (∆ω ) ⋅ S w (∆ω ) ~ (3.3518 ⋅ dTF ⋅ s + 1 2 (3. which is [49]: R R Fw{3 n} (∆ω ) = ~ ∫ ∫ψ (r )ψ (r )F ( f . f is the frequency.1.6 Implementation of the Stochastic Component The basic component of the stochastic component is the turbulence wind speed.29) where ψ is the aerodynamic function. r1 . the azimuth expansion coefficients are time dependent in this case. However.e. r1 and r2 are two radial positions and φ is the angular difference between them. i. The stochastic component includes the rotational sampling turbulence that is simulated the same way as the deterministic component. The Kaimal filter is fitted to output the Kaimal power spectral.3. i. and then they are used as amplitudes to the harmonics according to Equation (3. The EWS only accounts to the 0p and 3p as they are the most relevant to three bladed wind turbine rotor types. r )dr dr 1 2 k 1 2 1 0 0 2 R ∫ ψ (r )dr 0 2 (3.31) 56/152 .30) is the is r12 + r22 − 2r1r2 cos(φ ) .27) where the distance d In the EWS. The filter applied to the 0p harmonic component (from [50]) is: 4. r2 ) = 1 2π 2π ∫ γ ( f . where γ(f.d) is the square root coherence according to Equation (3. r is the radial position along the blade on the rotor area and Fk defined as: Fk ( f . r .28) where Sw is the Kaimal distribution and the admittance function Fw{3 n} can be determined by ~ a triple integral.
where dTF = R/U0 is the normalization parameter. To the 3p harmonic component. w⋅dT.F.7722 ⋅ dTF ⋅ s + 1 (3. The admittance function to the 0p harmonic component can be understood as a smoothing function to the variations of the turbulence on a stationary reference frame that are transferred to the mechanical power.) and with the 3p rotational speed (in rad/s – w3p).F⋅ 10 −1 Numerical Fitted 10 −2 Admittance Function F(w) (1) 10 −3 10 −4 10 −5 10 −3 10 −2 10 −1 10 10 Frequency normalised w (1) 0 1 10 2 10 3 Figure 3. to find the real frequency it is necessary to scale the normalised frequency (w) with the normalisation factor (dT. the variations are filtered out due to the coherence of the turbulence. 57/152 . In low frequency.17 Normalised admittance function to 3p. To find the real rotational frequency (in rad/s) in Figure 3.16 is necessary to scale the normalised frequency (w) with the constant of normalisation (dT.3691 ⋅ dTF ⋅ s 2 + 1. i.0307 2 0. Hence.10). the transfer function Hadm{3p} is fitted to [50]: H adm{3 p} ( s ) = 0. 10 0 Numerical Fitted 10 −1 Admittance Function F(w) (1) 10 −2 10 −3 10 −4 10 −5 10 −3 10 −2 10 −1 10 10 Frequency normalised w (1) 0 1 10 2 10 3 Figure 3.e.17 presents the normalised numerical and fitted admittance function to the 0p (source: [50]).2766 ⋅ dTF ⋅ s + 0. the turbulence variations are totally transferred to the aerodynamic power while on high frequency..32) Figure 3. Figure 3.16 presents the normalised numerical and the fitted admittance function to the 0p (source: [50]).F.).e.16 Normalised admittance function to 0p.F. w3p ±w⋅dT. i. R is the total radius of the wind turbine rotor and U0 is the average wind speed. It is important to remind that the normalised frequency to the 3p harmonic component is shifted by a frequency of 3p from Equation (3.
In Figure 3. the three components are added resulting in the stochastic part of the EWS.18 illustrates the total implementation of the stochastic part of the EWS implemented in Simulink/MATLAB®. Random Kaimal_Filter_1 Admit0_Filter Sum Stochastic Wind 1 Rotor Velocity 1 s integ 3 Gain cos(u[1]) cos3azm Random Kaimal_Filter_2 Admit3_Filter_1 sin(u[1]) sin3azm Product 3 Harmonic Random Kaimal_Filter_3 Admit3_Filter_2 Product1 Figure 3. 3. the 3p admittance function is quite small and. From Figure 3. which is inputted to the admittance functions representing the smoothing effect on the rotor area to 0p and 3p. the variations from turbulence are filtered out due to the coherence of the turbulence. the reduction effect is much higher compared to the 0p admittance function because the coherence decays much more due to the higher frequency (3 times the rotational speed). However. 58/152 .14). therefore not so important. The complete EWS is the sum of the stochastic and deterministic parts as presented in Equation (3. Figure 3. Finally. at these lower frequencies. At lower and above than 3p frequencies. The difference at very low frequency happens because the numerical results were fitted to a second order filter.18. In addition.The admittance function to the 3p harmonic component can be understood as a smoothing procedure to the variations of the turbulence that are transferred to the mechanical power similar to the 0p. a fourth order filter does an excellent fit but the improvements in terms of power variations on the results did not justify the use of higher order filter [50].18 Implementation of the stochastic model in Simulink/MATLAB. However.2 Validation of the Equivalent Wind Speed Model This section presents tests of the EWS. Finally. which is done in two parts: in the first part the deterministic and stochastic models are verified separately and in the second part the EWS is compared with measurements. Kaimal filters applied to the random numbers simulate the turbulence wind speed.17 it is possible to verify a poor agreement between the numerical result and the fitted admittance function at very low frequencies. the rotor position is first computed by integrating the rotor speed of the wind turbine.
00 90. Therefore. so the rotational speed of the rotor was assumed constant 40 rpm while in the DBP2 the stochastic wind speed was simulated to a single blade section.00 150. the dynamics of the wind turbine was not modelled.02 1.00 0.00 30. where ∆t is the time corresponding to 120° in the rotational speed of the rotor of the wind turbine (40 rpm). 1. Two winds are simulated using the EWS.99 0.2 presents the tower parameters from the wind turbine tested.00 60. the simulated wind speed in DBP2 was processed to reflect the average wind speed seen from threebladed rotor.19 shows a good agreement between the “all 3np” simulation and the DBP2 calculation.00 Normalised Wind Speed 0. The wind speed simulated in DBP2 is seen from a single blade section located at 2/3 of the total blade length.19 presents the comparison and Table 3. Physical geometric parameters Rotor diameter Distance from tower to rotor Tower diameter Size 29 m 1. only 0p and 3p are included in the model because the mechanical structure reduces the corresponding simulated mechanical torque as discussed previously in this chapter.00 90.20 in terms of PSD’s.00 30. the first one with the 0p and 3p harmonic components only and the second with all 3np harmonic components. the output of the EWS is compared to the ones from the Design Bases Program 2 – DBP2. the DBP2 generates time series of the wind speed on blade sections of the wind turbine that are compared to the EWS.) Figure 3. The deterministic part of EWS is compared in terms of the normalised deterministic wind influenced by the tower shadow. The comparison of the stochastic behaviour is shown in Figure 3. turbulence intensity of 0. Figure 3.97 0. 59/152 .00 120.00 60.00 150. Figure 3.2 Parameters used in the simulation for tower shadow. whereas the agreement of the “0p+3p” simulation model is poor.00 180. This illustrates that the equivalent deterministic wind speed can replace the deterministic wind speed simulated with DBP2 but it requires that orders higher than 3p are included.98 0.96 0. Still.95 0p+3p Model All 3p multiples components DPB2 0.In the first part.01 1.10.00 Azimuth Angle (deg. The DBP2 is an integrated model of a horizontal axis wind turbine that the main aim is to analyse structural performance of wind turbines [61].19 Normalised simulated deterministic wind component compared to a DBP2 model.70 m The stochastic part of the model was simulated for an average wind speed of 8m/s. t and t+∆t.00 120. Table 3.15 m 1. In this chapter. In the EWS.94 180. This was done by averaging the measured wind speed at times t∆t.
Table 3. the EWS model is close to the DBP2 model with only minor deviations.3. Therefore.3. whereas the reduced model describes the summed effect of the wind speed on all three blades. the EWS output was compared with measured data.20 PSD comparisons of the stochastic model.70 m The measured wind speed is seen from a single point on a rotating blade. The turbulence intensity and the mean wind speed measured in a fixed point on a meteorological mast were 0.51m/s respectively and the rotor speed was 30 rpm.10 0 EWS DBP2 −1 10 10 −2 ((m/s)2/Hz) 10 −3 10 −4 10 −5 10 −6 10 −1 10 Frequency (Hz) 0 10 1 Figure 3. the measured wind speed was processed to reflect the average wind speed from three blades before it was compared to the simulation.9 m 1. t and t+∆t. The main parameters of the wind turbine are shown in Table 3. In this frequency range. In the frequency range above 3Hz. where ∆t is the time corresponding to 120° 60/152 . Parameters of the wind turbine used in the measurement comparisons.16 and 10. The Pitot tube was mounted at 15 m of the rotor centre at an angle of 14° to the local chord and in a distance of about one chord length in front of the blade leading edge [48]. The wind speed was measured in a rotating blade with a five holes Pitot tube. This was done by averaging the measured wind speed at times t–∆t. only the contributions from zero and 3rd harmonics are taken into account in the EWS and the frequency range of interest is from 0p to 4p. The wind measured at the hub height on a meteorological mast in front of the wind turbine was used to compute the mean wind speed and turbulence intensity (input parameters of the model). the EWS differs from the DBP2 because the DBP2 shows the turbulence acting on a single blade section of a wind turbine including all physical aspects that are not presented in the electrical power measurements because they are filtered out by the wind turbines dynamics. Physical geometric parameters Rotor diameter Distance from tower to rotor disc Tower diameter Size 41 m 2. As previously mentioned. In the second part of the validation.
The EWS has also been compared to measurements.50 4.08m/s.00 1.05 0.5Hz.10 0.00 0.21 Measured and the simulated equivalent wind speeds on rotating blade section. Both of them have been compared with a verified model – DBP2. In addition.50 3. whereas the model intends to include the wind averaged along the blades.30 PSD*f((m/s)²) 0.25 0.50 2.00 0.00 3. The comparison of the measurement and simulation are shown in Figure 3. that the wind speed is only measured on a single blade.35 0. the same assumption is made for the blade 3.in 30rpm. which could explain most of the difference. 0.20 0. However.00 Frequency (Hz) 2. First. Hence.50 1. the measurements are done in a single section of the blade.7m/s.00 Figure 3. the EWS model was introduced. The standard deviation of the simulated equivalent wind speed is 1. It has been reduced in complexity although is based on complex theory. the standard deviation of the measured rotating wind speed is 1. 61/152 . The EWS consists of a deterministic part and a stochastic part. so it was necessary to assume that the wind speed seen by blade 2 is the same at time t+∆t as the wind speed seen by blade 1 at time t as described above. where it is described how the rotation moves standard deviation from lower frequencies to higher frequencies.15 0. it is important to quote that the dynamic wind turbine was not modelled. This complies with the theory for rotational sampling described by Kristensen and Frandsen [59].3 Equivalent Wind Speed Model Remarks In this section. The second difference between the measurement and simulation is.40 0.45 0. two aspects are different for the measurement and simulation. 3. which is 40% lower than the fixedpoint standard deviation as expected from the averaging process along the [50].21 where the 1p frequency is approximately 0. The comparison presents a reasonable agreement in the range from 0 to 3p. which is approximately equal to the standard deviation measured in the fixed point in the mast.50 Srot*fmeasured Srot*f EWS 0. so to the EWS (in this chapter only) the rotor speed f the wind turbine was assumed constant what is also pointed as one more reason to the difference between the model and measurements.
The fast computation and reduced memory usage are very suitable for simultaneous simulation of a large number of wind turbines making it possible to efficiently estimate the impact of a large wind farm on the power quality. In the case of a lattice tower. an approximate tower effect can be obtained by specifying an equivalent fictitious tubular tower. It showed good agreement in the frequency range from zero to 4p. The difference also presented in the range above 4p was expected as a result from the initial assumptions that the wind turbine acts as a low order pass filter. As the conventional models generate several wind time series. 62/152 . The higher components can be added to the model but measurements of the power indicate that they are filtered by the dynamics of the wind turbine. there is no significant difference considering the energy content of the DBP2 and the proposed model for the stochastic part. The stochastic part was compared with a spectrum derived from the DBP2 model. the presented model only generates one resulting in a small memory usage and a decrease in the time computation. The increased speed and reduced memory usage could be compared in terms of wind time series generated. The small difference of amount of energy in the range from 1P to 4P is associated to the single point measured compared with an equivalent wind simulated. Finally.The deterministic part does not present a good approximation using only the zero and the third harmonics. The deterministic part of the model was developed for tubular towers. The comparisons with measurements presented a good agreement in the range from 0P to 4P. The implementation of the model as a function of the rotor position enables the program to be applied for variable or constant rotor speed wind turbine models without additional modifications.
the aerodynamic rotor output depends on the wind speed and on the rotor speed. the dynamics of the wind turbine filter out the high frequency power variations but it also includes new components due to its dynamics itself. the integration of the electrical power into the grid is done at medium voltage (e. Controller – controls and optimise the wind turbine operation. so a stepup transformer is used. the elements of the wind turbine interact with each other.1 Simulation Tool The dynamic wind turbine model can be implemented in any simulation program. Each one of the components influences the dynamic operation of the wind turbines. Each one of them has a special configuration and specific power dynamics. Here. Stepup transformer – integrates the power produced in the medium voltage electrical network. The Equivalent Wind Speed model (presented is chapter 03) is the input to the dynamic wind turbine model. the wind turbine model is implemented in the Power System Simulation software 63/152 . The reactive power demanded from the generator is partial compensated using a capacitor bank. The electrical generator is directly connected to the network. The conventional wind turbine has a stallregulated rotor connected to an induction generator through a gearbox. Finally. Reactive power compensation – partial or total compensation of the reactive power demanded by the generator. A softstart is used to startup the wind turbine. the relevant ones are classified as: • • • • • • • Aerodynamic rotor – converts the wind power into mechanical power. From the basic components of the conventional wind turbine. e. Network – medium voltage grid that transmit the power. As mentioned before. This chapter presents a dynamic wind turbine model for a threebladed stall regulated wind turbine with induction generator connected directly to the network. The dynamic wind turbine model is intended for power quality assessment of wind turbines. Wind turbines are dynamic generators with several components that influence the power conversion from the wind. Here a dynamic wind turbine model to a conventional wind turbine is presented. 4. In addition.g.g. Generator – converts the mechanical to electrical power using an asynchronous generator synchronized with the network. 10kV). Transmission system – connects the aerodynamic rotor to the electrical generator using shafts and gearbox. There are several types of wind turbines commercially available.Chapter 4 4 Wind Turbine Model The produced electrical power from wind turbines does not have the same behaviour in terms of variation as the wind.
the last module is the data processing that is implemented in Matlab. The second module is the aeroelastic model of the wind turbine that reads the wind speed from a file and outputs the mechanical torque on the electrical machine shaft. In Figure 4.SIMPOW from ABB because it has built in an extensive library of electrical components and it can solve the network equations outputting the relevant characteristics of the electrical power system. so SIMPOW exports time series of selected variables 4. Unet. U is the voltage. Hence. Figure 4. The wind turbine components are classified in electrical and aeroelastic parts. ω is the rotational speed. the focus is on developing appropriate models of the aeroelastic parts of the wind turbine. comp means compensation unity.1 presents the main relations among different components. I is the current and f is the frequency where the subscripts aero means to the aerodynamic rotor and gen means electrical generator. The third module comprises the load flow and dynamic simulation of all electrical equipments. The second and third modules are implemented in SIMPOW. the Matlab module outputs time series of wind speed that is read by SIMPOW. The dynamic aeroelastic model must simulate the power conversion from the aerodynamic rotor and the transmission to the electrical generator. fnet. net means network and wtu means wind turbine unity. The aeroelastic part comprises the aerodynamic rotor and the mechanical transmission system. Mechanical Transmission ωgen Figure 4. Wind Speed Wind Aerodynamic Rotor Controller Reactive Power Compensation Taero Network ωaero ωaero Tgen Electrical Generator Igen fnet. which are input to SIMPOW. T is the torque. 64/152 .2 Wind Turbine Model Each component of the wind turbine has some relations with other components. Once an appropriate induction machine model is chosen. Stepup Transformer Icomp Iwtu Uterm. Therefore. The aerodynamic rotor depends on the transmission system. Finally. Random numbers are not available in the library. the wind speeds are time series. The total dynamic wind turbine model is composed of several modules: the first module is the wind speed generation (described in chapter 3) that is implemented in Matlab®.1 Interaction between each components of a wind turbine unity. and so on.1. the mechanical torque from the wind turbine model is the input and the SIMPOW simulates all the electrical and mechanical characteristics of the power system equipments modelled. SIMPOW has some limitations in handling stochastic processes. the transmission system depends on the electrical generator and aerodynamic rotor.
This model fairly represents the power conversion in wind speeds below the stall region [62].2.3) corresponds to steady state aero loads. R is the rotor radius.e. the generator.1) where ρ is the air density. Using Equation (4. However. Integrating the torque along the blade sections and adding up the effects of all 3 blades results in the aerodynamic torque on the main shaft.1 Aeroelastic Components 4. 4.The electrical part comprises all the electrical components. It underestimates the power fluctuations in the stall condition because the stall effect does not happen instantaneously as the wind changes but after a time lag [62].3) where in this case the Cp is a function only of the λ because stall regulated wind turbines have a fixed pitch angle. reactive power compensation.2) where ωrotor is the rotational speed of the rotor.1 Aerodynamic Rotor The aerodynamic rotor converts the wind into mechanical power. and network. Each blade section contributes to the total aerodynamic torque in a single blade. during stall conditions the aerodynamics has been reported to be quite different from static conditions because Equation (4. The model is based on time lags of separation [62].1. W is the wind speed. Cp is aerodynamic power coefficient and Pbaserotor is the rated power of the rotor in order to give the torque in per unit. θ pitch ) 2 = Pbaserotor Paerdynamic (4. and the power coefficient of a stall regulated wind turbine. stepup transformer. the aerodynamic torque in per unit can be expressed as: Taerodynamic = Paerodynamic ωrotor ⋅ Pbaserotor ρ ⋅ π ⋅ R 2 ⋅ W 3 ⋅ C p (λ ) = 2 ⋅ ωrotor ⋅ Pbaserotor (4. Aerodynamic effects throughout the blades convert the wind flow in aerodynamic torque. A dynamic stall model is implemented in the aerodynamic module. The aerodynamic model uses the Equivalent Wind Speed (EWS) as input to compute the available power and it uses the speed of the rotor to compute the torque on the main shaft. The power coefficient is a normalized function of the pitch angle θpitch and the tip speed ratio λ.1). i.2. which is defined as the ratio between blade tip speed and the wind speed as follows: λ= ωrotor ⋅ R W (4. The aerodynamic power on the main shaft can be expressed in per unit as: 1 ρ ⋅ π ⋅ R 2 ⋅ W 3 ⋅ C p (λ . The power conversion in an aerodynamic rotor can be decomposed in three conditions: 65/152 .
25 0. hence the power converted is proportional to the cubic of the wind speed. Figure 4. The first step to compute the dynamic power coefficient. which is important to avoid overload of the wind turbine.1 0. Cp(λ) is to find the static interpolation factor (fstatic) that represents the static curve as a combination of the others two as follow C pstatic (λ ) = f static ⋅ C pattached (λ ) + (1 − f static ) ⋅ C psepareted (λ ) (4. the power conversion follow instantaneously the present degree of separation and after the time delay it come back to the steady state curve.4) Then the dynamic interpolation factor (f) is computed as a time lag according to: 66/152 .15 0.3 0.2 Power coefficients to compute the dynamic power coefficient. On high wind speeds. 0.05 0 0 Power Coeficient CP Static Attached Totally separated 5 Wind speed (m/s) 10 15 20 25 Figure 4.45 0. Totally attached condition – is an extreme dynamic condition that considers that the flow does not stall in any wind speed.2 illustrates the three power coefficients to a 660kW stall regulated machine. the blade profiles are designed to start to stall (or separate) hence reducing the lift generated and the extracted power from the wind. during stall condition. • So. The power coefficient computation is done by dedicated aerodynamic program. The static (or steady state) power coefficient tends in low wind speed to follow the totally attached curve and in high wind speed to follow the totally separated curve. when the wind speed changes instantaneously.• • Static power conversion condition – is the steadstate power conversion curve of an aerodynamic.4 0. In low wind speeds. Totally separated condition – another extreme dynamic condition where in all wind speeds the flow is totally separated of the wind turbine blades. the blade profiles are designed to extract the power based on aerodynamic coefficients that depend on the attached flow to generate lift.2 0.35 0.
4 presents a simplified dynamic drive train model where the gearbox is considered ideal hence omitted here and the speeds and torques are in p.3. the aerodynamic rotor is presented that is connected to the electrical generator in the right side through a gearbox. On small wind turbines the brakes are positioned in the low speed shaft that reduces the stresses on the gear box during shut down of wind turbines. The drive train includes a gearbox.3 Example of drivetrain components. shafts and disc brakes that can be positioned in the low or in the highspeed shaft depending on the size of the wind turbine.3) gives the torque. 4. The drive train connects high speed in the electrical generator side and slow speed in the aerodynamic rotor using a gearbox.5) where the time constant τds =4/U0 is computed from experimental tests as defined in [62]. The electrical generators run in a relatively high speed compared to the aerodynamic rotor.∂f ( f static − f ) = ∂t τ ds (4. An ideal gearbox has no losses and does not modify the dynamics of the system. 67/152 .u. however. The shaft torsional modes are related to the aerodynamic rotor mass swinging with the induction generator mass through the flexible transmission shaft. Figure 4. on large wind turbines.3 illustrates the drive train of a small wind turbine size.1. The variations however are not directly transferred to the electrical power output because of the dynamic mechanical transmission system. The drive train modifies the dynamics of the system because they include torsional modes.2 Drive Train The aerodynamic torque.2.6) The dynamic power coefficient applied to Equation (4. in a wind turbine. The gearbox is analogue to an electrical transformer where the torques are like the currents and the rotational speed are like the voltages. U0 is the average wind speed. the dynamic power coefficient is defined as: C p (λ ) = f ⋅ C p attached (λ ) + (1 − f ) ⋅ C p separeted (λ ) (4. Disc Brake Gearbox Generator Lowspeed shaft Highspeed shaft Figure 4. Figure 4. Finally. the torques in low speed shaft are very high hence the disc brakes are located in the high speed shaft. In the left side of Figure 4. varies all the time due to wind turbulence and tower shadow.
In Figure 4. Hrotor is the inertia of the aerodynamic rotor.11) where Jrotor is the rotor inertia in physical units on the low speed side of the gearbox. The rotor inertia includes the blades and the hub in the low speed shaft.4. qgear is the ratio between low and high speeds in the gearbox. d is the equivalent damping coefficient that includes the losses in the system and an aerodynamic damping and Damping losses are the torque losses related to the damping coefficient.8) = K (δ rotor − δ generator ) − Telectromechanical + d (ω generator − ω rotor ) d (δ rotor ) = ω rotor ⋅ ω 0 dt 2 H generator d (ω generator ) dt (4. ωgenerator is the rotational speed of the induction generator.7) (4. ω0 = 2πf0/poles is the mechanical 68/152 . Hgenerator is the inertia of the induction generator. Kstifness is the equivalent stiffness of the shaft connecting the two masses.9) d (δ generator ) dt = ω generator ⋅ ω 0 (4. The inertia is then converted to the high speed side of the shaft and finally converted to the per unit system using the following equation (4. ωrotor is the rotational speed of the aerodynamic rotor. Telectromechanical is the electromechanical torque in the induction generator.4 can be expressed according to: 2 H rotor d (ω rotor ) = Taerodyn.Taerodyn. ∆δ d Telectromechanical K ωrotor Hrotor Stifness ωgenerator Hgenerator Figure 4. − K (δ rotor − δ generator ) − d (ω generator − ω rotor ) dt (4.10) where the constants are explained as follow: Hrotor – is the total inertia of the aerodynamic rotor expressed in per unit.11): H rotor = 1 J rotor (ω 0 ) ⋅ 2 2 qgear ⋅ Sbase 2 (4. The equations of motion in to the dynamic drive train in Figure 4. Taerodyn is the aerodynamic torque computed from the aerodynamic module.4 Dynamic representation of the drive train model.
a two masses model and a flexible shaft simulate the most relevant torsional moment from wind turbines [63].14) where ωtransmission is the first torsional eigenfrequency. The main intention is to model the first torsional mode of the mechanical transmission system but actually. fixed end) and the torque is measured. which is computed as follow: cdamping = 2δ log (2π )2 + (δ log )2 k stiffness ⋅ J rotor (4.13) where Tbase is the mechanical base torque and kstiffness is the stiffness constant in physical units. In the tests. 69/152 .12) where Jgenerator is the generator inertia in physical units.g. [10]. which can be computed using the first torsional eigenfrequency (ωtransmission) as follow: k stiffness = (ω transmission ) ⋅ J rotor 2 (4. The transmission system eigenfrequency is measured in tests of the wind turbine. Hgenerator – is the inertia of the electrical generator in per unit defined as: H generator = 1 J generator (ω 0 ) ⋅ 2 Sbase 2 (4. K is the equivalent stiffness of the shaft connecting the two masses and can be expressed as: K= k stiffness Tbase (4.15) where cdamping is damping coefficient computed in physical units. There are several other models to simulate the drive train module (e.16) where δlog is the logarithmic decrement in terms of few percents.e. which is very difficult to determine and in general it is measured from the tests of the wind turbines. [63] and [64]). Considering the power quality assessment of wind turbines. The complexity of the model depends on the aim of the model. the generator side has the brakes on (i. d is the damping coefficient expressed as: d= cdamping Tbase (4. the model includes the first and second modes due to the use of two masses system (second order model).synchronous speed on the generator side and Sbase is the base power of the electrical generator.
which is around 2 Hz. however. reduces the current spikes due to wind gusts.1 Electrical generator The asynchronous generator is the most common type of electrical machine.2. The active power is mainly related to the prime mover torque on the mechanical shaft and the voltage 70/152 . the rotational speed of the rotor is not fixed as it is in the synchronous machines.5 presents an example of power relation to the speed and voltage terminals in an asynchronous generator based on static models [65]. the reactive power compensation. it has been used for many years in wind turbines. control system. As generator. positive signal means power produced from the machine. The second mode is related to the interaction between the rotor and generator masses oscillating against each other due to the stiffness of the shaft. 4.2 Electrical Components The electrical components comprise the electrical generator. and low maintenance made this type of machine the most suitable for wind turbines. but it is related to the torque from the prime mover and to the network frequency. transmission lines.5 Active power as function of speed and voltage terminals for an asynchronous generator.1 presents the some characteristics of the models used here from the SIMPOW. Allied to those characteristics. when compared to the synchronous machines.5. In the asynchronous generators. the speed flexibility (slip). Figure 4. stepup transformer. Details of the electrical power system including the wind farm depend on the purpose of the simulation. In this thesis. Figure 4.The first torsional mode of a medium sized wind turbine is around 1Hz and it is considered the most relevant for power quality from wind turbines. The reliability.2. 4. transformers and power electronics and controllers. which has verified models of most of the conventional electrical components in power systems including electrical machines. The rotational speed of the asynchronous generator must be above the synchronous speed in order to force the power flux to the network.2. they are modelled using the standard library of SIMPOW/ABB. It is mostly applied as motor. In Figure 4. low price. Here only some characteristics of the electrical generator are presented and Annex 9.
Table 4. The wind turbine is installed at the test station facility of the RISØ laboratory.6 Reactive power as a function of the speed and voltage for an asynchronous generator (1p. hence the excitation is supplied from the power system. In Figure 4. i. The asynchronous machine does not have an independent excitation control system.e. The reactive power demanded of an induction machine depends mainly on the voltage at its terminals and on the active power supplied to the power system. Figure 4.u.1 presents a resume of the most relevant information to the wind turbine. of reactive power means the rated reactive power.6.u. 71/152 . Still keeping the active power constant and at this time reducing the voltage lead to a small reduction on the reactive power consumption. Keeping the active power constant and increasing the voltage lead to an increase in the reactive power because the excitation is a shunt element proportional to the voltage.influences the electrical characteristic of the induction machine.3 Verification of the Complete Wind Turbine Model In order to verify the Dynamic Wind Turbine Model. however decreasing further the voltage leads to an increase in the reactive power because of the losses that are increased due to the higher currents flowing in the machine. which means reactive power consumption.6 illustrates the reactive power from an induction generator. = rated reactive power at 1pu volts). Figure 4. The reactive power of induction machines as generator varies with the voltage. a 500kW stall regulated wind turbine directly connected to the grid is modelled and compared to measurements. 4. the excitation (reactive power) modifies with the voltage. negative power means the power system supplies power and 1p.
8 presents the measured wind speed applied to the dynamic wind turbine model.55 1. To this verification. time series of wind speeds to the 3p effect are simulated based on average wind speed and turbulence intensity measured. the 2) are the The average wind speed measured is approximately 8m/s and turbulence intensity of 13.Table 4.08Hz 5% 3 1. Nominal Power Rotor diameter Generator speed Rotor speed Capacitor Bank Generator terminal voltage Electrical frequency Tower height Gearbox ratio First torsional frequency Estimated logarithmic damping to drive train Number of blades Constant of inertia of the aerodynamic rotor Constant of inertia to the generator Tower shadow 3p effect 500kW 35 m 1500 rpm 27 rpm 300kVAr 400V 50Hz 36 m 55.1 Basic characteristics of the Nortank 500kW wind turbine modelled.4 Cp 0.5 0.15 s 0.3%.15 s 1.1 0 0 2 4 6 Tip Speed Ratio 8 10 12 Figure 4. Figure 4.5% Figure 4.7 Cp(λ) static characteristic of the wind turbine Nortank 500kW. 72/152 . In addition.3 0.2 0. the wind speed measured in the hub high is used as input to dynamic wind turbine model and admittance filters (as introduced in chapter representing the coherence smoothing effect on the rotor area of the wind turbine applied. 0.6 0.7 presents the static power coefficient to this wind turbine. The low wind speed is far from the stall condition and the dynamic stall module is not implemented in this case.
8 Measured wind speed.9 presents the time series of powers simulated and measured in the wind turbines terminals during 600s. Measured Active Power 400 Active power (kW) 300 200 100 0 0 100 200 300 time (s) Simulated Active Power 400 500 600 400 Active power (kW) 300 200 100 0 0 100 200 300 time (s) 400 500 600 Figure 4. One possible reason for the difference can be that the ambient wind speed is measured 100 meters from the wind turbine because it cannot be measured on the rotor 73/152 . Figure 4. A general good agreement is seen between the simulated and measured power from the wind turbine. The mean wind speed was adjusted to simulate the same mean power in the period of 10 minutes.9 Time series of simulated and measured power to the Nortank 500kW.11 10 9 Wind speed (m/s) 8 7 6 5 4 0 100 200 300 Time (s) 400 500 600 Figure 4. The measured wind speed directly applied to the model resulted in a mean power difference of 75kW (–15% related to the rated power) when compared to the measured power.
10 Verification of the standard deviation of the dynamic wind turbine model. The power spectral distribution (PSD) of the active power is one of the relevant characteristics from the wind turbines because of the stochastic nature of the process. at the frequency of approximately 0. Another possible reason for the difference may be that the actual aerodynamic characteristics are different from the modelled ideal characteristics. Those results indicate that the mean wind speed and the turbulence could still require some adjustments. The simulated instantaneous maximum power is 0. the mean and maximum powers and the standard deviation are overestimated while the minimum power is underestimated. With the mean wind speed adjusted (Figure 4.9).6 0 1 2 3 4 5 6 Frequency (Hz) 7 8 9 10 Figure 4. where the normalisation refers to the rated power of the wind turbine. From 0 to 6Hz there is a general good agreement between the simulated and measured power spectra that is verified in terms of the cumulate standard deviation (middle part of the figure). The adjustments in the wind speed are composed of modifying the mean wind speed and to scale the turbulence to modify the standard deviation. This mode 74/152 .07% higher than the measured.8 0.plane. hence Figure 4. The simulated mean power is 1. the measurements have a power variation related to the first tower bending moment. however. Hence.45% higher than the measured power the normalisation used in this section is related to the rated power. Power spectra S(f)*f of active power 300 Simulation Measurement (kW2) 200 100 0 50 40 0 1 2 3 4 5 6 7 8 9 10 (kW) 30 20 10 0 0 1 2 3 4 5 6 7 8 Simulation Measurement 9 10 1 Normalized cumulative difference of standard deviation (%) 0.e. the differences are assumed reasonable here.10 shows the normalised difference between the standard deviation simulated and measured. S(f)⋅f. The simulated instantaneous minimum power is 3. The middle part shows the cumulative standard deviation of the measured and simulated active power computed from the power spectra in the upper part. In the upper part of Figure 4.92% higher than the measured and the simulated standard deviation is 1. Finally.9. i. the lower part of Figure 4.19% lower than the measurements.10 the spectral power. has been chosen because it shows the power variance from each frequency that helps to illustrate the significance of each frequency component. In the power spectra. the wind speed acting on the rotor of the wind turbine is not necessarily the same as the measured one.8Hz.10 presents PSD of the measured and simulated power that corresponds to the time series presented in Figure 4.
it cannot be directly compared. The power on the frequency around 1.05 0 1 2 3 4 5 Frequency (Hz) 6 7 8 9 10 Pst_simulated Difference Figure 4. In the frequency range from 6. the measured and the simulated active powers agree. The simulated drive train has more variations than the measured one.was not included in the dynamic wind turbine model because it is small the power contribution.11. Verification of the flicker Pst to different frequencies. the Pst flicker is computed again where different power contributions in frequency domain are included as presented in Figure 4.1681.7Hz up 9. The shortterm flicker emission can be computed based on the time series of active and reactive power [34].5Hz.8% (normalisation to the rated power). 75/152 . These effects have not been implemented in the dynamic wind turbine model because they are small compared to the relevant power variations for power quality assessments as the difference between the simulated and measured standard deviations accumulated is less than 0. This is done by applying a sharp cutoff filter to the measured and simulated active and reactive powers at different frequencies and then those time series are applied to the flicker program that computes the Pst.11. The flicker as a function of the power contributions bellow different frequencies shows that the 3p is the most relevant effect on the flicker while in the frequency range from 5Hz to 10Hz the contribution to the total flicker is reduced (less than 0.08 is related to the first torsional mode from the drive train. the power spectral distribution of the measured power shows a power variation related to complex modes of oscillation that have been related to the flexible complex aerodynamic rotor and the induction generator [66]. Pst is 0. The simulated flicker is very similar to the measured one. The dynamic wind turbine model only includes the effects up to 4Hz that is the most relevant to power quality assessment. In order to detail the flicker estimation using the dynamic wind turbine model.15 0.1 Pst 0. The flicker shortterm emission (Pst) from the measurements is 0. The last relevant parameter compared in this section is the flicker emission. the total difference was small and because it is not linear. In the frequency range below 3p (1.1834 and from the simulation.05 0 Pst_measured 0. 0.2 0.18).5Hz).02 compared to 0.
4. which is intended to power quality assessment of wind turbines.4 Dynamic Wind Turbine Model Remarks A dynamic wind turbine model has been presented. The first difference is related to the tower bending moment that is not included in the aeroelastic model therefore the differences were expected. The model was applied to a 500kW machine and the simulated electrical power was compared to the power measured. Finally. The model also is fast and permits the simultaneous simulation of several wind turbines representing wind farms. however. present some differences. 76/152 . This chapter focus on the aeroelastic model because the electrical components are modelled by a dedicated conventional power system simulation program. The power spectral density. The main differences are related to basic assumptions done in the model. The model can be applied to different simulation tools with small modifications. the model is suitable to power quality assessment. The results showed a general good agreement.
and proper operation of the protections must be assured. when large amount of wind energy is connected to the local capacity or when the networks are weak. In IEC 6140021 the most relevant power quality characteristics from wind turbines has been indicated. Power utilities usually are concerned about power quality deviations and stability issues from wind farms operation. e. Wind turbines deliver a fluctuating power caused by turbulence and machine dynamics. Several wind turbines compose a wind farm where each wind turbine contributes to the net power output variations. the AWFWS is defined as the average of the wind speeds in each wind turbine that applied to the scaled wind turbine simulates the wind farm power produced. and here the main aspects of the wind in the park scale are presented. The Aggregate Farm Wind Speed (AWFWS) model is an average of the wind speed of the wind acting on each wind turbine in the wind farm. The tools presented in [1] can estimate the power quality deviations. A wind farm model to support power quality assessment of wind farms based on aggregate representation of the wind farm is presented. power stability.g. 5. The time inexpensive implementation of the AWF makes the analysis of integration of largescale wind farms in the power system more effective and feasible.Chapter 5 5 Aggregate Wind Farm Model Proper and successful integration of wind farms into the power systems rely on good understanding of the power variations from wind turbines. simulations of the wind farms are still recommended to some specific conditions. however. It gives also tools to estimate power quality deviations due to the wind turbines connection as pointed in chapter 2. The Aggregate Wind Farm (AWF) is a scale up of a wind turbine where the wind speed is an average of the wind acting on each wind turbine in the wind farm.1 Aggregate Wind Speed Model An appropriate wind speed model is essential to obtain realistic simulations of the power fluctuations from wind farms [67]. In those cases. the wind farms power production must be simulated and power quality. The turbulence model was defined in chapter 3. The AWFWS is based on a linear system assumption. The aggregate wind farm represents the entire wind farm. Recently. Considering the power from the wind farm is the sum of the power produced from each wind turbine in the wind farm. The power produced by wind turbines varies in a broad range of frequencies and can cause undesirable voltage variations as well as frequency variations in largescale integration. the power quality assessment from wind turbines has been addressed in IEC6140021 [1]. The AWF reduces the simulation time and it has been implemented in a commercial power system simulation tool SIMPOW/ABB® using available library models of the electrical components as presented in chapter 4. 77/152 .
3) 78/152 . the average wind speed and the turbulence intensity very similar to the case on the rotor area of a wind turbine as presented in chapter 3.1) and alat is the lateral component: alat = (17. d xy ) = e d ⋅f − a xy xy U0 (5. Here. the AWFWS represents the stochastic part of the longitudinal wind.1 Spatial disposition of the two wind turbines (αxy = 90° means lateral disposition).1. the Davenport type coherence [57] also represents the coherence between two wind turbines however the decay factors are modified to account larger distances. f is the frequency of the turbulence and U0 is the mean wind speed. the prevailing wind direction influences the coherence factor. 5.5 ± 5) ⋅ σ 1m / s (5. The longitudinal wind is the most relevant for power production in wind turbines.2) where αxy is the wind inflow angle (see Figure 5.1) where axy is the decay factor. The degree of correlation depends on the distance between the two positions.1. where the inflow wind speed angle for each wind turbine (αxy) is defined as in Figure 5. it is also assumed that all wind turbines have the same average wind speed and there is neither deterministic influences nor wake effect in the wind farm. y dxy x Figure 5. Similar to the coherence on the rotor plane in Equation (3.2 Coherence of Turbulence in a Park Scale The turbulence between two positions is correlated.27). which is a reasonable assumption. The decays factors axy for the park scale includes the distances up to kilometres and have been experimentally studied in [58]. dxy is the distance between the wind turbine in position x and wind turbine in position y as presented in Figure 5. αxy is the inflow wind speed angle to wind turbine in x that it is the relative angle of the wind speed to the wind turbine in the position y and U0 is the mean wind speed. which suggests the following expression: a xy = (a long cos(α xy )) + (alat sin (α xy )) 2 2 (5.Currently. In this case. αxy U0 where dxy is the distance between the wind turbines in positions x and y. The Davenport type coherence between two positions x and y is rewritten here as: γ ( f .
which is around 1Hz in medium sized wind turbines.2 presents an example of coherence simulated with different separations from 100m up to 1300m all with average wind speed of 15m/s and turbulence intensity of 15%. The separation and the frequency have strong influences on the coherence.4) Figure 5. 5. alat = 17. The distances in a wind farm are in the range from hundred meters up to kilometres. the turbulence can be considered uncorrelated because of the strong reduction from the coherence. These facts are important to the implemented AWFWS as presented in the following section. inflow angle = 0°.001Hz.i (5. in which the turbulence will be highly correlated at very low frequencies up to 0.where σ is the standard deviation of the wind speed in m/s.eq . 10 0 10 −2 100 m 500 m 900 1300 m 10 −4 10 −6 10 −8 Coherence γ 10 −10 10 −12 10 −14 10 −16 10 −18 10 −20 10 −4 10 Frequency (Hz) −3 10 −2 10 −1 Figure 5.i(t)) for the wind turbine i can be expressed as sum of harmonics on the rotor azimuth angle as follow: weq . The equivalent wind speed (weq. along is the longitudinal component: along = (15 ± 5) ⋅ σ U0 (5.2 Coherence factor for different distances between two points.i (t )e j 3 kφrotor .5 and along = 15. At the 3p frequency.3 Aggregate Turbulence The wind field acting on the rotor of a wind turbine can be expressed as an equivalent time series of wind speed as explained in chapter 3. Moreover.5) 79/152 .i (t ) = k = −∞ ~ ∑w ∞ 3 k .
8) where NW.T .8): w AWFWS (t ) ≈ w 0 p .T .i is the 3kth term of the expansion of the wind.g.T .6) where. AWF ) (5. i =1 ∑ Re{w3 p. it is assumed that the azimuth position of the aggregate wind farm φrotor. ∑ (w3 p .T . NW . Since the winds at 3p frequency are totally uncorrelated because of the higher frequency and distances between the wind turbines.9) Here.T . AWF ) + 2 Im {w3 p .7): wAWFWS (t ) = 1 NW . w0k. the assumption is valid. ∑ (w0 p .AWFWS are the harmonic amplitudes of the AWFWS. Kaimal [44]) as pointed in chapter 3. AWFWS (t ) = 1 NW .i ) + 2 NW . AWFWS (t ) + 2 Re {w3 p . i =1 (5. is the number of wind turbines in the wind farm. Power spectral analysis of power measurements from threebladed wind turbines revealed that the harmonic components at three times the azimuth speed are the most significant that are converted to the electrical power (chapter 3). Rewriting (5. ∑ weq .5) is then rewritten as: weq (t ) ≈ w0 p (t ) + 2 Re{w3 p (t )}cos(3φ rotor ) + 2 Im{w3 p (t )}sin(3φ rotor ) (5.T.T . AWFWS (t )}sin( 3 ⋅ φ rotor .i (t )}sin(3 ⋅ φ rotor .i (t ) ) NW . Equation (5.4 Simulation of Wind Speeds The wind speed simulator has six main inputs parameters: 80/152 .~ where the w3k . defined as: w0 p . i=1 (5.T .AWFWS and the w3k.T .AWF can replace the azimuth position of each wind turbine. AWFWS (t )}cos( 3 ⋅ φ rotor . k is the multiple for the third harmonic component in terms of azimuth angle φrotor for wind turbine i.T .T . NW .i (t ) ) + 2 NW . its harmonic components can be rewritten as: w AWFWS (t ) = 1 NW .11) 5.6) in Equation (5. Using the Aggregate Wind Farm Wind Speed (AWFWS).7) Inserting Equation (5. i =1 ∑ (w 0 p .10) 1 NW . i =1 ∑ Im{w 3 p .i ) (5. AWFWS (t ) = NW . NW .i (t ) N W .T .eq .T .i (t ) ) w3 p . the PSD of the wnp’s are computed as admittance functions applied to the PSD of turbulence (e.i (t )}cos(3 ⋅ φrotor . i=1 (5.
which corresponds to a wind farm with N wind turbines. The Shinozuka method purpose is to compute a realization of a stochastic process given the spectral density function of the process.g.1 ( f ) S 2. wind direction turbulence length scale. coherence factors. etc. The aggregate wind speed times series generator in Figure 5.9).n−1 ( f ) L S n−1.12) 81/152 .1 ( f ) M S n−1.1 ( f ) S1. e.3 presents the main structure of the AWFWS. Wind Speed Parameters Wind Turbine Rotor Diameter Wind Farm Layout Aggregate Wind Speed Time series generator PSD turbulence model Coherence model Admittance Functions 0p time series 3p time series 3p time series Figure 5. 2 ( f ) S 2. turbulence intensity.3 is based on the Shinozuka method using a crossspectral matrix similar to the method used in [68]. Coherence model – e.g. The S(f) has the following structure (S ( f )) = S1. The model outputs three time series that compose the Aggregate Wind Farm Wind Speed – AWFWS – as defined in Equation(5. The crossspectral matrix S(f) is NxN. Here a crossspectral matrix represents the correlated turbulence processes. Wind farm layout.3 Structure of the AWFWS generator. 2 ( f ) M M Sn. Figure 5.n ( f ) S n .1 ( f ) S n . average wind speed. Wind turbine diameter.n ( f ) S 2. Kaimal [44].n−1 ( f ) S n . n ( f ) M S n−1.• • • • • • Wind speed parameters – e. The crossspectral matrix contains all energy from the turbulence on each wind turbine position in the wind farm and the contributions due to the correlation between two positions.n ( f ) (5.n−1 ( f ) S1. Turbulence Power Spectral Density model.2 ( f ) K K O L S1. Admittance functions – to 0p and 3p harmonic components.g. Davenport [57].
y. The corresponding time series is discrete by the sampled representation of the wind speed with time steps of ∆t=1/fs where fs is the sampling frequency in Hz. The main difference between the terms to the 0p and 3p is that for the 3rd harmonic the cross spectral elements Sxy.k(f)) for each harmonic component is written as: S AWFWF .11) the spectral density terms to the Aggregate Wind Farm Wind Speed (SAWFWS.1 as τ xy = cos(α xy )d xy U0 (5. The corresponding discrete value of Sxy(f) is Sxy[i]=Sxy(∆f i)⋅∆f. The cross power spectral density includes a time delay and it is written as [19]: S xy ( f ) = S xy ( f ) e −2πfτ xy (5.10) and Equation (5.T . i. the crossspectral density can be expressed as S xy ( f ) = γ ( f . γxy is the coherence function between x and y defined in Equation (5. 3 p xy . x =1 ∑S (f) (5.T .3 p ( f ) = 1 NW .T . 2 NW .k are assumed to be zero due to the high frequency and relative large distance between wind turbines x and y (coherence is zero).where the diagonal elements are the spectral density of turbulence on each wind turbine.17) Where Sxy.T . where Ns=Tp×fs is the number of samples in the simulated time series. d xy ) S xx ( f ) S yy ( f ) ⋅ e − j 2πfτ xy (5.1) and τxy is the travelling time of the wind travel from wind turbine x to y.k are the cross spectral elements of the kth harmonic component and Sxx. Ns must be integer and in order to use FFT it should be an exponent of 2.k is the spectral density of each kth harmonic component for each wind turbine.e. Sxy(f) is the cross spectrum between the turbulences in x and y.15) Hence. The discrete step of the frequency to simulate a time series with length Tp is ∆f=1/Tp. Thus. 0 p (f) (5. it is necessary to discretise the frequency in order to use a computer code to generate the wind speeds.13) where . 2 NW . f is the frequency. The nondiagonal elements are the crossspectral densities of the turbulence for the pairs x. which can be determined according to Figure 5. NW . using the spectral matrix and taking the definition of the AWSWF from Equation (5. 82/152 . x =1 y =1 ∑ ∑S xx .16) S AWFWF .14) where U0 is the mean wind speed. The next step is for each frequency index k to generate random complex numbers ck=Ak+jBk to finally use the inverse Fourier Transformation with Fourier coefficients ck.T . After having defined the spectral terms of the aggregate wind speed. 0 p ( f ) = 1 NW . The sampling frequency limits the frequency to ±fs/2 and consequently the frequency index i to ±Ns/2. so the ith frequency is f[i]=i∆f.
Figure 5.e.20) The discrete part of the energy content of the frequency k⋅∆f must then have the same mean value of the energy content Ek.22) Assuming that the mean value of Ak and Bk are zero.19) It is imperative that this method preserves the variance σ² of the process.21) { } { } S xx (k ⋅ ∆f )∆f 4 (5. E{Ek } = S xx (k ⋅ ∆f ) ⋅ ∆f Assuming E{Ak2}= E{Bk2} then.w(t ) = A0 + ∑ 2 ⋅ Ak cos( k ⋅ ∆f ) − ∑ 2 ⋅ Bk sin( k ⋅ ∆f ) k =1 k =1 N N (5.18) can be rewritten as: E k ( w ) = 2 A k2 + B k2 ( ) (5. the variance can be defined as: σ 2 = ∫ S xx ( f )df ≈ 0 ∞ fs / 2 ∫ S xx ( f )df ≈ ∑ S xx (k ⋅ ∆f ) ⋅ ∆f 0 K =0 Ns / 2 (5. using a single sided spectrum.4 presents a 2 Thus. E Ak2 = E B k2 = (5.18) Then the energy content to the kth component from Equation (5. Thus. i. Ak and Bk can be determined from a normally distributed stochastic process comparison between the formal Shinozuka random phase angle method and the use of the two random numbers 83/152 .23) with mean value zero and standard deviation: S xx (k ⋅ ∆f )∆f . E{Ak}= E{Bk}=0. the standard deviation σ{Ak} and σ{Bk} can be determined as σ {Ak } = σ {Bk } = E {Ak2 } − (E{Ak })2 = S xx (k ⋅ ∆f )∆f 2 (5.
Here.1 Case Description The wind farm is composed of six 660kW stall regulated directly connected wind turbines type. the aggregation procedure is applied to similar wind turbines. The aggregation procedures have been extensively applied to power system analysis particularly to stability analysis of large power systems. The Parksimu model [19] generates correlated time series of wind speed to each wind turbine in the wind farm. The single wind turbine is modelled as presented in chapter 4. 5. The IWT models each wind turbine in the wind farm with accurate models to the wind turbine dynamics. [70] and [71]. [69].6. The method has been applied to represent groups of similar machines in areas of the power system. a mechanical 84/152 . e.e. The wind turbine is composed of an aerodynamic module. The IWF simulates all wind turbines in the farm applying to each of them a correlated time series of wind.3 AWFWS method Shinozuka´s method 2 1 Distribution of Bk 0 −1 −2 −3 −3 −2 −1 0 Distribution of Ak 1 2 3 Figure 5.4 Distribution of random constants. 5. hereafter called IWF. i. The AWF model uses a single aggregate wind turbine that replaces all wind turbines in the wind farm. From Figure 5.4 it is possible to see that the terms to the Inverse Fourier Transformation are more random than the random phase angle method. hence the equivalent wind turbine to the entire wind farm is a scale up of a single wind turbine.6 Results and Discussions Here. where Nwr is the number of turbines in the wind farm.g. 5. the simulated power using the AWF is compared with a simulation that models individually each wind turbine in a wind farm.5 Aggregate Wind Turbine Machine The aggregate wind turbine replaces the entire wind farm with a single wind turbine. The aggregation procedures of induction motors applied to power quality analysis have been successfully reported. the base power becomes Nwt times the base power of a single wind turbine in the farm.
1.2 per units (seconds) of the aerodynamic rotor and induction generator respectively. Figure 5.6 presents the wind farm layout.00 Power curve (p.00 10.u.00 5.8Hz 5% 3 3.transmission system. Where the lateral space between wind turbines is 4 times the rotor diameter and the longitudinal space is 10 times.43 0. a capacitor bank.00 15. The drive train of the wind turbine has the first torsional moment in 1Hz and inertias of 3.00 30.1 Basic characteristics of the 660kW wind turbine modelled.2 s The electrical generator is a squirrel cage induction machine.05 and 0. 85/152 .60 0.e. Nominal Power Rotor diameter Generator speed Rotor speed Capacitor Bank Generator terminal voltage Electrical frequency Tower height Gearbox ratio First torsional frequency Estimated logarithmic damping to drive train Number of blades Constant of inertia of the aerodynamic rotor Constant of inertia to the generator 660kW 48 m 1500 rpm 21 rpm 200kVAr 400V 50Hz 36 m 71. and a stepup transformer.1 presents a resume of the most relevant information to the single wind turbine. an asynchronous generator.00 Wind speed (m/s) 20.00 0. Table 5.80 0. The stiffness and damping constants are computed using the definitions from chapter 4.20 0.05 s 0. The wind direction is measured relatively to the North as indicated. The logarithmic decrement damping coefficient is assumed to be 5%.40 0.00 25. modelled with a 3rd order model where current displacement module is not active (i. The aerodynamic module uses power coefficients as introduced in chapter 4 and dynamic stall effects are included.20 1.00 Figure 5. Table 5. the rotor resistance is constant). Static power curve of the wind turbine.) 0.5.
In the upper part. the cumulative normalised differences of standard deviation of the AWFWS and Parksimu to the Kaimal target spectrum are presented. The admittance filters are disabled and only one wind speed is simulated in order to compare to the wind speed for a single point as in the Parksimu. average wind speed of 10 m/s at 30 meters height. i. The middle part presents the power spectral of the simulated wind speed using the AWFWS simulator and the one produced from Parksimu described in [19].2 Results 5. The same characteristics are applied for both simulations. Kaimal is the target power spectral of turbulence used in both simulation programs. 86/152 . the Kaimal turbulence model is plotted.1 Wind Speed Simulator The wind speed simulator is verified here compared to Parksimu [19]. turbulence intensity of 10%.2. The power system is very strong at the Point of Common Coupling (PCC). The Parksimu is a program to simulate correlated time series of wind speeds of each wind turbine in a given wind farm based on the layout of the wind farm. In addition.6. 400 times the rated power of the wind farm. the mean wind speed and turbulence intensity as well as on the spatial coherence. and turbulence length scale of 600m.7 presents a general comparison between the wind speed simulated with the Aggregate Wind Farm Wind Speed (AWFWS) and the Parksimu.e. 5.6.6 Wind farm layout. Figure 5. The Thevenin equivalent at the PCC has an angle of approximate 50°. time series of wind speed simulated with both models are presented. In the lower part of the figure. The PCC is the same to both simulations: AWF and IWF.10 x D D N θws WS 4xD 4xD 4xD 4xD 10 x D Figure 5.
Time series of wind speed 15 Wind speed (m/s) 10 AWFWS Parsimu 5 0 100 200 300 Time(s) 400 500 600 Power specrtal (m/s)2 10 10 10 10 0 −1 −2 −3 AWFWS Parksimu Kaimal 0 0. the AWFWS appears to represent better the turbulence.5 4 AWFWS Parksimu 4.5 3 Frequency (Hz) 3.2 Wind Farm Power Production This section presents comparisons of the wind farm power simulated using the AWF to the IWT. 87/152 .6. maximum power in 0.5 5 0 (%) −0. Table 5.2 −0. In this case. in accordance to the IEC 6140021 [1]. The average power. The power spectra reveal general agreement of both wind speed simulations with the target power spectra (Kaimal). the normalised cumulative differences of standard deviation show small difference with a total difference less than 0.13 and 16m/s) and two different turbulence intensities (10 and 20%). Those relative high wind speeds have been selected because the power curve at those wind speeds is nonlinear.2 seconds and flicker emission in short term were the parameters selected to compare the different models.7 Comparisons of the wind speed simulator. 5.5 2 2.4 0 0.2%. both simulated wind speeds have similar energy content and slightly difference to the target standard deviation.7. in addition. the normalisation factor is the mean wind speed (10m/s). the AWFWS presents more variations than the Parksimu around the target power spectrum.5 3 3.7. Despite the high variation of the power spectrum of the AWSWF. In the lower part of Figure 5.5 2 2. From Figure 5.5 1 1. Six cases are discussed: three different average wind speeds (10. in addition the standard deviation and minimum power are presented in order to give more information on the AWF usage.5 5 Figure 5.7 shows two wind speeds uncorrelated with similar range of wind speed variations. and in this particular case the random seed chose to the AWSWF represented better the turbulence than the Parksimu. anyway the total accumulated difference in terms of standard deviation is small. The high variation around the target spectral is a consequence of using two random numbers instead of a random phase angle as used on the Parksimu. the difference relative to the target power spectral distribution (Kaimal) is mostly related to the random seeds.5 4 Cumulative difference of stanrdard deviation to Kaimal target 4. The upper part of Figure 5.2.5 1 1.2 presents the main power characteristics and flicker emission in short term (Pst) simulated using the AWF and the IWF.
8 Power characteristics evolution with the wind speed (10% turbulence intensity).5 Individual min StD Individual 0.62 2454.9 illustrates the power characteristics from both simulations on high turbulence intensity.181 10 3501.48 132.8 illustrates the power characteristics simulated on low turbulence intensity.5 Turbulence intensity 10% 0.64 1602. the power characteristics simulated with the AWF agree well with the ones simulated with the IWF.094 0.140 10 3870.91 114.85 3158.05 0.47 3519.74 2615.251 0.064 13m/s 20 3411. the power simulated with the AWF is slightly higher than the IWF.25 2.164 0. 4.0 0.74 3477.37 3791.5 0.11 54.0 0.15 244.19 3970.66 164.61 3914.15 1.04 131.88 370.61 0.185 0.25 162.100 In low turbulence intensity. the differences are large than the ones in the low turbulence but it is still presenting good agreement except to minimal power.10 0.02 248. In high turbulence intensity.0 0.45 Flicker Pst AWF (kW) IWF (kW) 4.95 1921.55 2992.5 0.73 2804.91 4014.01 0.07 3868.29 1830.0468 16m/s 20 3769.02 4061.93 0.99 2786. In general.20 3358.65 3903.0474 0.66 3533.35 2.79 2395.Table 5.08 96.64 3713.40 3.0 8 9 10 11 12 13 Wind speed (m/s) 14 15 16 17 18 0.0 Individual mean Individual max Aggregation mean Aggregation max Aggregation min StD Aggregate 0.42 4256. 88/152 Standard Deviation (MW) .00 Figure 5.48 41.92 0.5).2 Representative power characteristic values of all simulations Mean Wind speed Turbulence intensity (%) Mean Value Max Value Min Value Standard Deviation Mean Value Max Value Min Value Standard Deviation IWT AWF 10m/s 10 2419. Figure 5.36 2430. Both simulations show that the power variations reduce with the increased wind speed because of the nonlinearity of the aerodynamic power conversion (Figure 5.87 3153.5 0.30 0.20 1.01 1492.11 255.75 4107. 3.85 3589.120 0.071 0.11 2892. Figure 5.086 20 2403. A possible reason for the power differences may be that the wind speed simulated with the AWF has slightly high standard deviation then the ones simulated with the Parksimu as introduced in Figure 5.7.30 Power (MW) .13 4152.
Non linear effects on the power variations.40 1.10 0.5 Individual min StD Individual 0. The high turbulence intensity means that the wind speed will vary in a broad wind speed range on top of the mean wind speed. the high turbulence intensity in high wind speed strongly influences the simulated minimum power and the standard deviation in the AWF. In the high mean wind speed (16m/s).00 0.00 Figure 5. the power conversion on the aerodynamic rotor is extremely nonlinear (see Figure 5.5 0.60 Standard Deviation (MW) 2. Power (MW) 3. 89/152 .10).40 0.00 5.9 Power characteristics evolution with the wind speed (20% turbulence intensity).60 0.80 0.00 15.20 0.0 0.80 3.90 4.20 Power curve 10m/s 10% ti 16m/ 20%ti 0.00 Figure 5.) 0.00 10.50 2. Similar to the characteristics presented in low turbulence intensity. the power variation in a single wind turbine in the IWF can be very different from the one simulated with the AWF because the AWF is an average wind speed that smoothes the wind variations leading to less power variation and a concentration around the high mean value.5 0. in this case.00 Aggregate Power (p.00 30.0 8 9 10 11 12 13 Wind speed (m/s) 14 15 16 17 18 0.u.0 0.00 25.Turbulence intensity 20% 4.0 Individual mean Individual max Aggregation mean Aggregation max Aggregation min StD Aggregate 0.0 0.10.20 1. 1. the AWF in general simulates higher power than the IWF and the power variations reduces with high wind speed.5 0.30 1. The differences are related to the same reasons as in the low turbulence intensity case. However. Hence.00 Wind speed (m/s) 20.5 0.70 .
5 2 2. 90/152 .11 presents the power produced by IWF and AWF with 13 m/s average wind speed and 10% turbulence intensity.5 4 4. samples of time series of active power simulated are presented that illustrates the processes described above. The simulated active powers presented in Figure 5.5 3 Cumulative standard deviation 3.11 Power simulated 13 m/s and 10% turbulence intensity.11 are uncorrelated and have similar ranges of power variations.12 Power spectral comparisons at 13 m/s and turbulence intensity 10%.4 −0. Ws= 13m/s and Turbulence = 10% 4000 3800 3600 P (kW) 3400 3200 3000 Individual Wind Farm 2800 0 100 200 300 Time (s) 400 500 600 4000 3800 3600 P (kW) 3400 3200 3000 Aggregate Wind Farm 2800 0 100 200 300 Time (s) 400 500 600 Figure 5.5 1 1. Figure 5. Power spectra S(f)*f of active power Power spectra (kW 2) 4000 3000 2000 1000 0 150 100 50 Individual Wind Farm Aggregate Wind Farm 0 0 −0. At 13m/s and 10% turbulence intensity.5 5 Individual Wind Farm Aggregate Wind Farm (%) (kW) −0. the aerodynamic power conversion is reasonable linear. Power spectral analyses reveal small differences in frequency that are presented in Figure 5.Following.12.6 −0.8 0 1 2 Frequency (Hz) 3 4 5 Figure 5.2 0 1 2 3 4 Normalized cumulative difference of standard deviation 5 0 0.
4500 4000 P (kW) 3500 3000 Individual Wind Farm 2500 0 100 200 300 Time (s) 400 500 600 4500 4000 P (kW) 3500 3000 Aggregate Wind Farm 2500 0 100 200 300 Time (s) 400 500 600 Figure 5.12 presents the normalised difference of the standard deviation between the powers simulated with the AWF and with the IWF. the power spectrum of the AWF presents a good agreement with the IWF as showed in the upper part of Figure 5. the differences become more noticeable. As one can notice. the IWF simulates a steady power reduction in 120 seconds while the AWF does not.At 13m/s.14.13 presents the active power simulated at 16m/s and turbulence intensity 20%. The reduction on the standard deviation difference is very small. This is related to the averaging wind speed process of the AWF added to the nonlinear rotor as explained before. where the negative signal means that the AWF overestimates the standard deviation and the normalisation factor is the rated power of the wind farm. 91/152 . the overall difference is small (less than 0.13. Figure 5. the power variations from the IWF are slightly higher than the power variations from the AWF that reduces the cumulative standard deviation difference. At frequencies above 1Hz. The cumulative difference of the standard deviation from both simulations agree well as presented in the middle part of Figure 5. that means that the main difference of the standard deviation (power variations) are concentrated on the low frequency range. In Figure 5.12. The lower part in Figure 5.13 Power simulated at 16 m/s and turbulence intensity 20%. When the wind speed and the turbulence intensity are increased.12 with a steady difference that does not significantly changes with the frequency.2%) and the main difference is on the lower frequency range (less than 1Hz). The power spectra of the powers reveal more information as presented in Figure 5.
In the higher frequency rage. The power variations on low frequency range presented the most significant differences between the AWF and IWF.5 3 Cumulative standard deviation 3. Figure 5.15 presents the flicker coefficients for the wind turbine using the Pst simulated with the AWF and IWF. 92/152 .3) [1]. the flicker estimated in the AWF is compared to the one in the IWF.2.Power spectra (kW 2) 6 4 2 0 x 10 4 Power spectra S(f)*f of active power Individual Wind Farm Aggregate Wind Farm 0 0. In the lower frequency range. The power produced from the AWF is smother than the power produced by the IWF that is related to the high nonlinear power curve of the wind turbine. the differences on the standard deviation change very little because at this frequency range the wind speed variations are relatively small hence the dynamic stall provides a better power relation between the active power variation and the wind speed variation. Following. the rated power of the wind farm and the number of wind turbines in the wind farm. The flicker coefficient from a wind turbine is a normalised measure of the flicker from a wind turbine as introduced in chapter 2 that can be computed using Equation (2. the power differences were relatively small that support the use of the aggregate model to flicker estimation from large wind farms.5 5 200 150 (kW) 100 50 0 3 Normalized cumulative difference of standard deviation 2 (%) Individual Wind Farm Aggregate Wind Farm 0 1 2 3 4 5 1 0 0 1 2 Frequency (Hz) 3 4 5 Figure 5. The standard deviation of the AWF is very different of the one in the IWF. Following.e. less than 1Hz. The difference of standard deviation actually decreases with the frequency.5 2 2. Using the shortterm flicker presented in Table 5.5 1 1. At the frequency range above 3p. The differences are very noticeable in the lower frequency range.5 4 4. the wind speed variations are significant. which in this case acts on the high nonlinear area of the power curve leading to greater differences on the standard deviation and minimum power. i. the flicker coefficient (c) to the wind turbine in the wind farm is computed.14 Power spectral comparisons at 16 m/s and turbulence intensity 20%.
15 Flicker coefficients comparisons (computed according to [34]).5 4 4.5 2 2.2.5 2 Flicker coefficient c 1. The flicker coefficients computed using the AWF are quite similar to those simulated with the IWF.00 Aggregate 10% TI Aggregate 20% TI Flicker coefficient (c) 1.50 Individual 10% TI Individual 20% TI 2.5 1 1.5 0 0 0. The main difference appears to be 93/152 . The active and reactive powers simulated in the AWF and IWF were filtered out with different lowpass cutoff frequencies and then applied to the flicker code that simulates the shortterm flicker emission. In high turbulence intensity. The flicker estimated at 13m/s and 20% turbulence intensity with the AWF presented the higher difference to the one simulated with the IWF.5 Frequency (Hz) 3 3.16 presents the evolution of the flicker coefficient computed with the Pst that included the contributions of the power variations with different frequencies. the nonlinear behaviour of the power conversion leads to greater differences between the AWF and the IWF but still god agreement.00 0. particularly on the low turbulence intensity. Flicker coefficient Wind speed 13 m/s Turbulence intensity 20% 2. following. In order to verify the contributions from the power variations in frequency domain to the flicker coefficient.5 Ck_IWF 1 Ck_AWF Difference 0. the flicker estimation is detailed to two different wind speed conditions (13m/s.00 10% 10 10 20% 10% 13 13 20% 10% 16 16 20% Wind speed (m/s) Figure 5.50 Turbulence Intensity(%) 0. The flicker estimated in the AWF poorly agrees with the flicker coefficient estimated in the IWF at 13m/s and 20% turbulence intensity.50 1. 20% turbulence intensity and 16m/s 10% turbulence intensity). Figure 5.5 5 Figure 5.16 Evolution of flicker coefficients at 13m/s 20% turbulence intensity.
0.5 1 1.3 Extension to Large Wind Farms and Different Random Seeds In order to verify the usage of the AWF to large number of wind turbines. The high turbulence intensity leads to large wind variations that when added to high wind speeds lead to large incursions on the nonlinear region of the power conversion. The flicker estimated at 16m/s and 10% turbulence intensity presented the best agreement. The evolution of the flicker coefficient presents very small differences as showed in Figure 5. but the flicker procedure is not linear and the curve in Figure 5. 5.6 where the new wind turbines are included in groups of three in the end of the rows keeping the same distance 94/152 .7 0. The wind farm layouts are similar to Figure 5. However.2 0. The dynamic stall improved the aggregation model because it generates a relation between wind speed variation and power variation on the stall condition.17.5 Flicker coefficient c Flicker coefficient Wind speed 16 m/s Turbulence intensity 10% 0. In this subsection.16 must be used with caution.1 0 0 0.4 0.6. In this section. The relevant differences encountered here are related to the nonlinear rotor under stall conditions and high turbulence intensity.5 5 Figure 5. This is done because on the aerodynamic rotor resides larger part of the nolinear conversion and the simulation of larger wind farms including the electrical components would demand extensive simulations not giving much information because the main differences reside on the wind speed conversion.5 4 4. it is not enough to compensate to the 6 wind turbines in high turbulence intensity. In this case. 15.3 Ck_IWF Ck_AWF Difference 0.17 Evolution of flicker coefficients at 16m/s 10% turbulence intensity. the AWF has been applied to a small wind farm keeping the same random seeds in the simulations in order to trace the results. the cases presented in the previous section supported that the differences found were related to the wind speed variations at low frequency range (lower than 1Hz). the model is extended to a large number of wind turbines and different random seeds are applied to two different wind speed conditions (13 and 16m/s 20% turbulence intensity).5 3 3.1 Frequency (Hz) 0. the wind farm model does not have the electrical dynamic components and only the aerodynamic power is computed using the dynamic stall module.6 0. In addition. 12. Moreover.related to the 3p area.5 2 2.2. 21 and 30 wind turbines have been implemented in Matlab/Simulink®. and until now. it had also been indicated that the random seeds can modify the results. wind farms with sizes of 6.
that the standard deviation in the AWF is higher than the IWF in agreement with the complete model presented in the previous subsection. 4 Wind speed 13 m/s and Turbulence intensity 20% 3 Normalized difference (%) 2 1 0 1 2 0 5 10 Standard deviation 15 20 25 30 35 Number of Wind Turbines Average Power Poly. One can notice in this subsection. (Standard deviation) Poly. Starting with the mean wind speed at 13m/s. (Maximum Power) Poly. Positive values stand for the AWF lower than the IWF.18 presents the normalised power characteristics as a function of different number of wind turbines. In addition. 95/152 .between them and the same wind speed inflow angle. where they are normalised with the rated power of the wind farm. These two main factors are related to the fact that increasing the number of wind turbines in this particular nonlinear part of the power curve decreases the influence of a single wind turbine on the total results. which makes the power extremely correlated. (Minimum Power) Figure 5.18 presents the normalised differences of the standard deviation and mean powers. standard deviation. Increasing the number of wind turbines in the aggregation procedure improves the performance of the AWF.18 Influences of different sizes of aggregate wind farms to the power characteristics at 13m/s. the random seeds have strong influences on the results that reduce with the number of wind turbines in the aggregation procedure. Figure 5. maximum and minimum powers are the power characteristics investigated in the following section. 20 % turbulence intensity The upper part of Figure 5. The mean power. (Average Power) Wind speed 13 m/s and Turbulence intensity 20% 8 6 Normalised difference (%) 4 2 0 2 4 0 5 Maximum Power 10 15 20 25 30 35 Number of wind turbines Minimum Power Poly.
Probably in the complete model in SIMPOW.19 presents the normalised differences of the standard deviation and the mean power.5 1 1. (Maximum Power) Poly. Following the analyses with the mean wind speed at 16m/s are presented.5 0 5 Average Power 10 Standard deviation 15 20 25 30 Poly. the mean power in the AWF is lower than the IWF. This difference is related to the strong influence of the random seed in the mean value.18. (Standard deviation) 8 6 Normalised difference (%) Wind speed 16 m/s and Turbulence intensity 20% 4 2 0 2 4 6 8 0 5 Maximum Power 10 Minimum Power 15 20 25 30 35 Number of Wind Turbines Poly. the upper part of Figure 5. The random seeds have strong influences on the results that in general reduce with the number of wind turbines in the aggregation procedure.9). 2 1.5 0 0. the differences between the IWF and AWF reduce when increasing the number of wind turbines in the aggregation procedure. 20 % turbulence intensity Similar to Figure 5. These two main factors are also related to the fact 96/152 . random seeds used to generate the wind speeds in both Parksimu and AWFWS were two extreme cases.however. (Average Power) 35 Number of Wind Turbines Poly.19 Influences of different sizes of aggregate wind farms to power characteristics at 16m/s. The lower part of Figure 5. which is different from the complete model in SIMPOW (Figure 5. These conclusions are also related to the reduced significance of a single wind turbine when increasing the high number of wind turbines. (Minimum Power) Figure 5. In general.5 Normalised difference (%) Wind speed 16 m/s and Turbulence intensity 20% 1 0. Increasing the number of wind turbines in the aggregation procedure improves the performance of the AWF.18 presents the normalised differences of the maximum and minimum power.
Simulation of all wind turbines in a large extension of the power system is practically impossible. The Aggregate Wind Farm model can replace the entire wind farm in power system simulation programs.7 Aggregate Wind Farm Model Remarks Simulation of large wind farms can be extremely expensive. The lower part of Figure 5. The wind speed model accounts to spatial coherence of turbulence on the entire wind farm representing the smooth effects on the power production in the wind farm. the simulation of the power produced from 6 wind turbines with the IWF took around 5 hours while using the AWF. 97/152 . Finally. which region the AWF seems to reproduce well the IWF. the differences between the IWF and AWF as well as the random seeds influences on the results tend to reduce when increasing the number of wind turbines in the aggregation procedure. in the same program installed in the same computer. The AWF makes possible power quality assessment of large number of wind turbines. The AWF has some limitations.e. there is a demand for equivalent models to analyse the power system interaction with the wind power. The minimum power and standard deviation of the power in high wind speed and high turbulence intensity do not agree with the simulation of individual wind turbines. The flicker is mostly caused by the 3p effect from wind turbines. The aggregate procedure focuses on the wind speed simulation. the flicker agrees well. power quality deviations and to analyse the control or to optimise the power system operation.19 presents the normalised differences maximum and minimum power differences between the AWF and the IWF. Respecting the AWF limitations. the simulation time was reduced to 20 minutes.750MHz (with 128Mb of ram memory). Using SIMPOW® installed in a Pentium III® . the power oscillations. the model supports the design and operation of the power system where the wind energy contributes to large part of the power production. It can be applied to replace large wind farms in order to simulate the impact on the power system characteristics. Therefore. i. In general. 5. The AWF simulates reasonably the power produced from wind farms. depending on the size of the wind farm and on the extension of the power system network.that increasing the number of the wind turbines decreases the influence of a single wind turbine on the total results. The AWF is time inexpensive and easy to implement in a conventional power system simulation tool used by the utilities. However. These conclusions are also related to the reduced significance of a single wind turbine when increasing the high number of wind turbines.
98/152 .
Chapter 6
6 Large Scale Integration – Case Analysis
The wind power impact on the power system quality and stability has been introduced and specific models have been presented. Here, three cases are investigated to illustrate interactions between wind power and power system in a largescale integration. Figure 6.1 shows the cases studied here and the tools used to perform the analyses. Case 1 Voltage Stability: Fictitious Grid Case 2 Voltage Stability & Quality: Brazilian Grid Case 3 Power Quality: Nordic Power System
Static Analysis Dynamic Analysis
Figure 6.1 Case Studies.
6.1 Case Study 1: Voltage Stability in a Modern Power System
Large amount of wind energy can modify the voltage stability of a power system as introduced in the chapter 2. Here, the impacts of wind power to the voltage stability characteristics are illustrated. A Loadflow program computes the loadability curves to the case analyse based on static models. The loadability curves are computed by increasing the load in specific buses of the power system (chapter 2). This way the power transfer capabilities is stressed until the voltage collapses.
6.1.1 Power System Characteristics
The power system studied is a small part of an existing power system. Figure 6.2 shows a single line diagram of the modelled power system where the rated voltage is 220kV. Figure 6.2 includes the total impedance of the lines and the loads installed on the network. In Figure 6.2, the generators are installed on the buses 6 and 7. The bus 7 is the slack bus and bus 6 is a voltagecontrolled bus (synchronous generator). Here, they have no restriction on reactive or active powers. The balance on bus 6 is an active power production of 110MW that explains why the loads are not expressed in the diagram. Bus 2 is supplying 20MVAr of reactive power to the network as indicated in the diagram and in addition, a capacitor bank with rated power of 50MVAr is installed on bus 3 to reduce the reactive power flux on the network.
99/152
40+j*30 Bus 1 (2+j*65) Ohms Bus 2 110  j*20
P + j*Q Swing Bus
GS
(22.8 + j*62.6) Ohms (6.7 + j*70)
Bus 7 (6.7 + j*65) Ohms
Ohms
Bus 4 (27 + j*35)
(6.7 + j*15) Ohms (6.75 + j*25)
Ohms
Bus 6
GS
P + j*Q Voltage Controlled
Bus 3 Wind Power Bus 5
Ohms
100 + j*48.83 100 + j*100
PV  Bus
Figure 6.2 Diagram of the Power System used in analysis (loads in MW and MVAr).
Simulations show that bus 3 is the weakest node in the system due to long lines connecting to buses 7 and 4, in addition, bus 3 has a large reactive load installed that leads to a power factor of 0.7 largely compensated with shunt capacitor banks. Figure 6.3 shows the loadability curve to bus 3.
Loadability with cos(φ)=0.5 230 220
210
200
Voltage level (kV)
190
180
170
160
150 bus 1 bus 2 bus 3 bus 4 50 100 150 200 250 300 350 Load in the bus 3 (MVA) 400 450 500 550
140
130
Figure 6.3 Loadability curve to bus 3 without wind turbines.
As expected, the lowest voltage is achieved on bus 3 because it has the highest load installed. The voltage on bus 2 is reduced due to high power flowing in the power system although it is located between two voltagecontrolled buses. The voltage on bus 4 follows the behaviour on bus 3 because the power flow to bus 3 causes voltage drop in the network. Finally, the voltage on bus 1 remains constant because it is only connected to the bus 7 hence there is no change in the power flux in its transmission lines.
100/152
The voltage curves to bus 6 and 7 are not plotted because they are voltage fixed amplitude (synchronous generators without restrictions). The synchronous generator keeps the voltage constant on bus 6 by inserting or drawing reactive power. In this particular case, there is no reactive power limitation on the synchronous machine on bus 6 hence the voltage will always be constant. Bus 7 is the reference bus to the load flow calculations, which has constant voltage amplitude and the angle is the reference for the calculations on all other buses voltages. In addition, all active and reactive power balances are done on the reference bus, which has no restrictions. In reality, bus 1 would also suffer voltage variations once the voltage on bus (#7) starts to reduce because the power limits were reached on the synchronous generator on bus 7, similar behaviour will happen to bus 6. In Figure 6.3, the active and reactive load to bus 3 were increased linearly, i.e. the active and reactive powers to bus 3 had fixed load factor, in this condition, the maximum load that can be installed is 424MVA, i.e. 300MW + j300MVAr. The load flow problem is complex and the maximum active and reactive powers that can be delivered through the lines are different. Hence, the loadability procedure (load increase until the voltage collapses) is modified to two: in the first, the reactive power is kept constant then the active power increased linearly until the voltage collapses; in the second, the active power is kept constant then the reactive power increased linearly until the voltage collapses. Using the new loadability procedure, the independent maximum loads to active and reactive powers are modified. In the first case, the maximum active power increased from 300MW to 540MW (the reactive power was kept constant 100MVAr). In the second case, the maximum reactive power increased from 300MVAr to 380MVAr (the active load was kept constant 100MW). The power system is weaker to transfer reactive power than active power, which is expected and related to the electrical characteristics of the transmission lines. The maximum load that can be installed on bus 3 is basically limited to the network strength because there are no limitations on the synchronous generators and the loads are constant independent of the voltage. Inclusion of the reactive power limits on the synchronous generators is likely to reduce the maximum load supplied to bus 3 because when the synchronous machine reaches the maximum reactive power limit it loses the voltage control capability. The loads dependency on the voltage is complex and constant power related to the voltage is a neutral assumption. An inverse dependency on the reactive, i.e. increase of power demanded as the voltage decrease, reduces the limit of maximum load that can be installed to bus 3, on the other hand, a direct power relation can increase the voltage stability limits because the loads (on the power system) reduce with the voltage reduction. In addition, the reactive power compensation with shunt capacitor banks is relevant to the voltage stability because of the high sensitivity to the voltage (i.e. reactive power proportional to V²). In this power system, the loadability to bus 3 is influenced by shunt capacitors banks, which heavily compensate the reactive power (50MVAr). The voltage reduction leads to an increase in the flow of reactive power trough the lines that reduce further the voltage as introduced in chapter 2.
101/152
4. In this thesis.e.2 Normalised demand of reactive power (p.15 0. the active (P) and the reactive (Q) powers are specified to the load flow program (that computes the loadability curves). the polynomial function to the reactive power for induction machines does not include the noload capacitor bank installed at the wind turbines and the normalisation factor is the nominal reactive power at rated voltage.25 500kW Ploynomial fit 2. Reactive power demand to rated active power 1.2 Wind Power Representation In this section.15 1.2 Figure 6. An increase in the voltage leads to an increase in the reactive power because the induction machine excitation (i. The P component is the rated power of the wind farm to be installed on a specific site and Q is computed based on the voltage terminals with a polynomial function. the reactive power consumption) is related to V² on the other hand the voltage decrease leads to an increase in the reactive losses that is related to V2 increasing the total reactive power consumption.6. Because the wind turbines have no active control on the voltage. All the wind power in the power system is concentrated on 102/152 . Assuming the rated active power. a static model simulates the wind farm power production because the voltage stability problem is slow.95 1 1.) 1.05 1 0. The wind farm static model must simulate the reactive power demanded based on the active power and the voltage at the wind turbine bus. 6. the reactive power consumption depends on the voltage (as introduced in chapter 4).95 0. The reactive power compensation of the entire wind farm is modelled as shunt capacitor banks in the load flow program.7178*X2 − 0.1.) 1.4MW 400kW 1. bus 5 is connected to the power system.1 1.u.4 Loadability curve to bus 3 without wind turbines.3 Wind Power Impacts on Voltage Stability In order to investigate the wind power impacts on the voltage stability.7841*X−2 + 0.8 0.4.05 Voltage terminals (p. In Figure 6.9 0. The polynomial function used here was fitted based on three different sizes of induction machines as presented in Figure 6.1.u.1 1.85 0.5036 1. this is a reasonable assumption that is very similar to the procedure suggested in [28].
The installed wind power in bus 5 is increased until the voltage collapses. 230 220 210 200 Voltage level (kV) 190 180 170 160 150 bus 1 bus 2 bus 3 bus 4 bus 5 50 100 150 200 250 300 350 Wind power to bus 5 (MW) 400 450 500 550 140 130 Figure 6. In order to investigate the wind power impacts on the voltage stability. The reactive power demanded by the wind power explains the small limit to wind power integration. Table 6. The maximum wind power that can be installed before voltage collapses is 429MW being smaller than the maximum load that can be installed to bus 3 (537MW). Condition Without wind power With wind power constant With wind power reactive power polynomial representation Difference P (MW) 537 580 +43 P (MW) 537 534 –3 Q (MVAr) 100 100 100 0 103/152 . Connecting large wind power to the weakest bus is expected to be the worstcase scenario to the voltage stability because this bus already faces limited power transfer capabilities.1 and Table 6. Bus 5 is connected to the weakest part of the network (bus 3) through a short radial line (Figure 6. Two conditions are presented: in the first. Table 6.5 shows the behaviour of the voltage of the entire power system as a function of the wind power injected in bus 5. the wind turbine reactive power is constant (i.5 Maximum wind power integration concerning voltage stability. Figure 6. the reactive power of the wind farm is represented with the polynomial function of the voltage.e. Table 6.2). a 100 MW wind farm is installed to bus 5 and new loadability curves to bus 3 are computed.1 presents the loadability limits increasing only the active load in bus 3.5 is computed in a similar way to the loadability curve. the reactive power polynomial function is not used).it. Figure 6.1 Bus 3 loadability limits keeping the reactive power constant.2 present the wind power impacts on the loadability curves concerning voltage stability. in the second. This size of wind power represents 28% of power penetration in the power system.
which can actively control the reactive power. Similar to the first case.6 Maximum wind power integration concerning voltage stability (load factor unity).The row presenting the difference in Table 6. In this case. can be used to regulate the voltage and improve the voltage stability. The demand of reactive power is the key factor on voltage stability from wind farms. the loadability to bus 3 is reduced. the absence of reactive power flux to the wind farm improves the voltage stability. hence wind turbines with power electronics. the wind power local production shall improve the loadability on bus 3 as the fist column presents an increase of 43MW. Wind turbine cos(φ) =1 with power electronics 230 220 210 Voltage level (kV) 200 190 180 170 bus 1 bus 2 bus 3 bus 4 bus 5 100 200 300 400 500 600 Max wind power in the bus 5 (MW) 700 800 900 160 Figure 6. Following. Those results show that the reactive power from wind farms play an important role in the voltage stability. when the voltage dependency of reactive power demanded by the wind power is included. the voltage stability limit is reduced because the reactive power to the wind farms flows through the same transmission lines that supply bus 3.1 shows the impact of the wind power on the voltage stability. in this case the voltage tends to increase at first moment but as the power flux increases. The wind power modifies the limits of voltage stability.2 shows the loadability limits with the active load in bus 3 kept constant. the reduction is 13%. However. Table 6. Condition Without wind power With wind power constant With wind power reactive power polynomial representation Difference Q (MVAr) 385 350 –35 Q (MVAr) 385 335 –50 P (MW) 100 100 100 0 In this condition. the 104/152 .8 illustrates the integration limits of wind power to bus 3 (concerning voltage stability) of wind turbines with power converters that can keep power factor unity independent of the voltage. At first sight.2 Bus 3 loadability limits keeping the active power constant. The voltage on bus 5 is higher than on bus 3 because it is injecting power. Considering constant power the wind power reduces in 9% the loadability to bus 3 and including the voltage dependency. the dependency on voltage of the reactive power from the wind farm strongly influences the voltage stability. Table 6. Figure 6.
The power converter is limited to operate at minimum capacitive power factor of 0. Figure 6. However. which is much higher than the original case 429MW. the voltage was set to 2% above the nominal). In this new condition. the maximum apparent power of the power converter must be respected. 105/152 . When the wind farm reaches the maximum reactive power production.8 illustrates the power production of the wind farm with the loadability of bus 3 as well as the power factor of the wind farm. Loadability curve to bus 3 with wind turbines using power electronics. Figure 6.8. As the maximum power from the power electronics must be respected. is that they can be used to regulate the voltage by injecting or draining reactive power. the maximum wind power to the power system is 811MW. the wind turbines lose the voltage control capability hence they behave as fixed power production.6 illustrates the loadability curve to the power system with 100MW wind farm installed in bus 5 using power electronics that can control the voltage (in this illustration. as cited before.associate losses increases leading to voltage reduction and finally to the voltage collapse.7. The active power of the wind farm is 100MW that is reduced as the reactive power is injected in the power system. 230 220 210 200 Voltage level (kV) 190 180 170 160 150 bus 1 bus 2 bus 3 bus 4 bus 5 0 100 200 300 400 Installed power in the bus 3 (MW) 500 600 700 140 130 Figure 6. Another important feature using power electronics. the wind farm is limited to 100MVA.
The wind power is one of them. during the last few years. This section presents some investigations of the wind power integration to a part of the Northeast region.3 10 0 100 200 300 400 Active power installed in Bus 3 500 600 0. the controllers were not implemented and the reactive power from the wind farm was increased linearly until it reaches the maximum of 60MVAr (corresponding to a power factor of 0.7 .2 Case Study 2: Voltage Stability and Quality in a Brazilian Power System Large amount of wind power has been proposed to Brazil.4 20 0. The Brazilian power system is mainly composed by hydropower stations.1 Power System Characteristics The interconnected Brazilian electrical power system is divided in four regional systems: North (N). more than 2000MW of wind power have been proposed.8 70 Power factor 60 Reactive Power 50 0. to this region alone. Northeast (NE). 106/152 cos( φ) capacitive 0.100 1 90 Active Power 0.2 Figure 6. However. The Northeast coast of Brazil presents good wind conditions and complementary wind characteristics to the main river flow along the year [73].9 illustrates the Brazilian interconnected system. in some regions. it is not possible to install more hydropower stations.2.8 capacitive) and the active power is reduced to 80MW in order to respect the 100MVA limit. 6. The connections between the NE system and the N are strong with a transfer capacity of few thousand Megawatts. a long dry season leaded the entire power system to a critical minimum water level on the reservoirs resulting in an national energy crisis [72]. In addition. which is expected to contribute to the energy in a large extend in years to come. however to the other regions there is a single link in 500kV with limited power transfer. 6.6 40 0. There are connections between all four regional systems. Therefore. Southeast (SE) and South (S). The active power produced from the wind farm is reduced to respect the thermal limits of the power electronics. In particular example.8 Evolution of the power production from the wind turbines (with electronic power converter). Figure 6.9 Active and reactive power from wind turbines 80 0.5 30 0. The Brazilian government has introduced new regulations to diversify the power matrix.
the NE system is circled. The regional NE power system has approximately 10270MW of hydropower stations installed and 435MW of diesel power stations. 107/152 . The installation of 100MW to a part of the system is investigated with respect to the voltage stability and voltage quality during continuous operation. The plans account to install large amount of wind turbines in the coast area from Natal to Fortaleza (NorthNortheast coast marked in Figure 6. Figure 6.10 presents a schematic representation of the network to which the wind farms are installed.9) far from the load consumption centres. The 100MW is divided in 5 wind farms where all of them are installed to MOSSORO bus. All transmission lines in the region are AC lines with voltage levels from 500kV to long distances down to 138kV to short distances.9. The hydropower stations are concentrated in the middle of the region (marked as number 5 in Figure 6.9 Brazilian interconnected system power system [72] In Figure 6.Notheast System Wind power plans Legend Transmission Lines Figure 6. More than 2000 MW of wind power has been planned to the region [73].9).
Here.2.10. The load characteristics of the local network are presented in Table 6. The simulations take into account the future power system condition [74]. all of them installed to MOSSORO bus on high voltage level (230kV).80MW+j⋅13. The aggregate wind farm represents identical 150 wind turbines (in five wind farms with 30 wind turbines each) of 680kW.10 Brazilian network studied [72].88MW+j⋅3. i.82MW+j⋅3.30MVAr 45. In Figure 6.10). the 680 kW wind turbine does not 108/152 .40MW+j⋅3. Table 6.99MVAr 42. the dynamic voltage quality is also assessed through the Aggregate Wind Farm (AWF) model (presented in chapter 5).e. BANABUIU and S.3 Relevant loads in the Brazilian power system studied. Node MOSSORO RUSSAS ACU Heavy Load 93. The installation of the wind farm is planned to near future when reinforcement of the power system is planned. between BANABUIU and MOSSORO will have a 230kV line and from ACUII to S.81MVAr Light Load 58.CRUZ is 138kV.e.31MVAr 6. which presents the load to relevant nodes in the study case to high and light conditions. that uses induction machines and partial reactive power compensation with shunt capacitor banks.CRUZ (Figure 6. i.CRUZ Figure 6.2 Wind Power Representation The total wind power has 100MW of conventional wind turbines.CRUZ will have an additional line in 230kV. In the future condition. which simulates the power produced from the entire wind farm. Actually. voltage independent. where the link in 138kV will be open. directly connected to the network. The case studied focus on the region between BANABUIU and S. the rated voltage between BANABUIU and ACUII is 230kV and between ACUII and S.3.70MVAr 52.73MW+j⋅12. The total wind power is divided in 5 wind farms with 30 identical wind turbines.100MW of conventional wind turbines BANABUIU S. The reactive power dependency on the voltage is modelled with a polynomial function of voltage.CRUZ are modelled as infinite nodes and all other buses are modelled as constant load. The voltage stability analyses are similar to section 6. where fixed active power and voltage dependent reactive power model the wind power.72MVAr 52.1 Case Study 1: Voltage Stability in a Modern Power System.66MW+j⋅20.
6. hence the wind turbine used in this section has the same normalised data from a similar wind turbine presented in chapter 5 (660kW).2.10.e. The 100MW wind power is divided in 5 wind farms with identical layout and identical 30 wind turbines each. The power system in Figure 6. here is included according to the reinforcement plans. 4x48m = 192m).11 presents the layout of each wind farm.12. The distance between wind turbines is 4 times the rotor diameter (i. the arrows represent loads installed to the buses. With this wind speed direction. 109/152 .11 Layout of a single wind farm applied to the Brazilian power system studied. There is no wind correlation between the five wind farms and the aggregate wind speed model generates one time series to each wind farm. this is assumed the worstcase scenario because a large number of wind turbines are aligned in a row with the wind speed. The line between ACUII and S.10 is implemented in the loadability computation tool (chapter 2) as presented in Figure 6.exist. In Figure 6.CRUZ is modelled as 230KV according to the reinforcement plans. The distance between the wind turbines is small that is acceptable because the wake effect is not simulated in the AWF.12. Wind farm layout 600 550 500 N Wind speed 450 Lateral distance (m) 400 350 300 250 200 150 0 200 400 600 800 1000 1200 Longitudinal distance (m) 1400 1600 1800 2000 Figure 6. Similarly. the ring connection between BANABUIU and MOSSORO that is not present in Figure 6.3 Wind Power Impacts on the Voltage Stability The loadability curves characterize the voltage stability in the study cases similar to the previous section. Figure 6.
loadability limits are computed to the MOSSORO bus. Load Conditions Heavy Load Light Load Maximum Active load (MW) 888 733 Maximum Reactive load (MVAr) 534 460 The classification of the loadability limits to MOSSORO presented in Table 6. 110/152 . However. the differences between heavy load and light load conditions happen because under heavy load. the maximum wind power that can be installed to MOSSORO (respecting the voltage stability) are 780MW in heavy load and 727MW in light load. the maximum wind power to the bus is small than the maximum load that can be installed because the wind turbines demand reactive power as explained in the previous section.CRUZ PV node 230KV 230KV 230KV Reference BANABUIU RUSSAS 230KV MOSSORO 230KV Figure 6.LOAD ACUII 69KV S. while under light load condition some shunt reactors were switched on to limit the voltage increase. The highest value (888MW. the differences in terms of reactive power limits are very small because the reactive power produced by the shunt capacitors is very sensitive to voltage variation. In both cases. as explained in the previous section.4 presents the loadability limits to MOSSORO to different load conditions. The active load that can be installed to MOSSORO is higher than the reactive because of the electrical characteristics of the transmission lines in the local power system. In order to characterize the original installation. which is to increase the active and reactive load individually until the voltage collapses.4 Loadability limits to MOSSORO. several shunt capacitors were switched on to compensate the reactive power. The difference between the heavy and light load conditions are related to the shunt capacitors installed in the system under heavy load condition similar to the loadability limits presented in Table 6.4. Before analysing the impacts of wind power on the voltage stability. The loadability limits uses the same procedure presented in the previous section.12 Network topology of Brazilian power system studied. To this power system. 364MVAr) characterizes the loadability limit to MOSSORO because with the load evolution all capacitors available in the power systems must be switched on to prevent high reactive power flow on the network. Table 6. Table 6. the limits to integration of wind power are drawn concerning the voltage stability.WIND FARMS MOSSORO . In addition.4 represents the maximum active and reactive loads that can be installed to MOSSORO before the voltage collapses independent of each other.
13 presents the evolution of the voltage in the relevant buses along the wind power installed to MOSSORO (230kV) in light load condition. Table 6.05 1 0. 6. the net loadability limit concerning reactive power is reduced as the wind farms demand reactive power.13 Maximum wind power to MOSSORO bus (light load condition).4% in the light load and 1.5 presents the loadability limits computed to different load conditions.Figure 6.75 MOSSORO 230kV Mossoro 69kV RUSSAS ACUII 0 100 200 300 400 500 600 Wind Power Installed in MOSSORO 230kV(MW) 700 800 0.5. In this power system.85 0. The voltage on the power system reduces with the wind power inserted in MOSSORO. The closest buses present the high voltage variation while ACUII follow the general behaviour but with less severity. In order to assess the impacts of 100MW on the voltage stability. 1.5% in heavy load) because the wind farms have shunt capacitors that help to improve the voltage stability.u. Conditions No wind power With wind power No wind power With wind power Load condition Heavy Load Heavy Load Light Load Light Load Maximum Active load (MW) 888 905 733 758 Maximum Reactive load (MVAr) 534 432 (+110MVAr to the wind farm) 460 380 (+114MVAr to the wind farm) In Table 6. Table 6. the loadability limits to MOSSORO in high and low load conditions are compared to cases with and without wind power installed (at rated power). the wind power improves the voltage stability concerning the active.9 0.4 Wind Power Impact on the Voltage Quality The impacts of 100MW wind farm on the voltage quality are analysed by means of dynamic simulations using the AWF presented in chapter 5. However.) 0. The AWF is applied to a 111/152 . the wind power is installed to MOSSORO and new loadability curves are computed. as expected.13 just to present the voltage on the main load connected to the MOSSORO bus. but the total reactive power flowing to MOSSORO is slightly increased (7.8 0.95 Voltage (p.2.7 Figure 6.5 Loadability limits to MOSSORO with wind power. The MOSSORRO 69kV is included in Figure 6.
Another relevant characteristic. As the mean wind speed increases. The main voltage differences between the heavy and light load are caused because of the reactive power compensation schemes to each load condition (i.14 also includes the maximum and minimum values to each load condition as envelop curves. Figure 6.e. the difference between the maximum and minimum active powers is approximately 10MW while the mean power produced is approximately 70MW (next section presents the time series). the wind power produced and the reactive power demands increase that causes voltage drops due to losses in the network. large shunt capacitors installed under heavy load condition). The power production by the wind power increases with the wind speed and the voltages on both load conditions reduce with the increased wind speed. presented as follow. The voltage is analysed under different wind speeds with two load conditions: heavy load and light load.15. is the standard deviation of the power and voltage to different wind speeds as presented in Figure 6. Figure 6.conventional fixed speed wind turbine connected to the transmission level at 230kV with dedicated lines and transformers. In Figure 6.14 Wind power influences on the voltage to different wind speeds. As an illustration. 150 Wind power (MW) 100 50 0 Mean power Max power Min power 8 9 10 11 12 13 14 Voltage at Mossoro (kV) 244 242 240 238 236 234 8 9 10 11 12 13 14 Mean voltage − Heavy load Max voltage − Heavy load Min voltage − Heavy load Voltage at Mossoro (kV) 244 242 240 238 236 234 8 9 10 11 Wind speed (m/s) 12 Mean voltage − Light load Max voltage − Light load Min voltage − Light load 13 14 Figure 6. Those wind speeds represent the wind speed range with the most relevant power variations from wind turbines. the aggregation of 150 wind turbines leaded to a very smooth wind power production where the maximum and minimum powers are very close to the average. The wind speeds are simulated (using the aggregate wind farm wind speed simulator from chapter 5) to 8.14 shows the 10 minutes average values of the active power produced by the wind farms and its influences on the average voltages on MOSSORO under heavy and light load conditions. Above 14 m/s the power conversion of the wind turbine limits the active power and the wind speed variations lead to relative small power variations and below 8 m/s the level of power is small and the power variations are also reduced. 12 and 14 m/s all of them with 20% turbulence intensity. in the average wind speed of 10m/s. 10. 112/152 . where the reactive power is also plotted.14.
2 8 9 10 11 Wind speed (m/s) 12 8 9 10 11 12 Std dev. Voltage Heavy load Std dev. Hence.6 0.6 Active P. 6. Active P. Reactive Heavy load Std dev.15. presents the higher standard deviation of voltage and reactive power.4 1. The voltage and the reactive power are strongly related. (MVAR) Std dev. which has more capacitors installed. 113/152 .4 0. the wind farm influences on the voltage quality to the wind speed of 10m/s are detailed to each load condition. (MW) 1. Heavy load Std dev.2 1 0.1. Light load 13 14 Voltage (kV) 0. the heavy load condition. The standard deviation (STD) gives an idea of the processes’ fluctuations. Active P. The main differences between the heavy and light load conditions are on high wind speeds where the voltage reduction influences the reactive power produced by the capacitors banks installed in the power system. The curves above support that probabilistic load flow studies should be used for analysis in the future in order to assess the impact of the wind turbines in the whole operating range.16 presents the time series of the wind power produced and the voltage on bus MOSSORO 230kV at mean wind speed 10 m/s and turbulence intensity of 20% in light load condition.8 0. Reactive Light load 13 14 Figure 6. Following. Both STD of voltage and reactive power increases with the wind speed because on high wind speeds the high active power reduces the voltage that consequently reduces the reactive power produced by shunt capacitors leading to higher reactive power fluxes.4.15 General wind power influences on MOSSORO bus. Voltage Light load 13 14 Reactive P.8 0.2 8 9 10 11 12 Std dev. the SDT of the active power produced from the wind farm has a maximum on the mean wind speed of 10 m/s then reduces because of the nonlinear characteristic of the power curve.1 Light Load Condition Figure 6.1 0. from Figure 6.2.15 0.05 0 1 0.
7 0 100 200 300 Time (s) 400 500 600 Wind Power Production in Mossoro 80 Active Power (MW) 75 70 65 0 100 200 300 Time (s) 400 500 600 Figure 6.75 .5 6 4 2 242.8 242.1 243.5 66 68 70 72 74 Reactive Power (MVAr) 76 9.1 10.5 35 35. Most of the time. active and reactive powers and voltage statistics in terms of histograms in light load condition.3 10.3kV (0. The voltage is 5.15 243.125% of the mean voltage 240kV).05 243. Ocurrences (%) 6 Wind speed (m/s) 4 2 0 9.95 243 243.17 presents the wind speed.8 9.9 242.16.6% above the nominal value (230kV) because the system is designed to operate a little above the nominal in order to compensate the voltage drops in lower voltage levels.243.2 Voltage at PCC 243.8 242. active power. reactive power and voltage variations at MOSSORO (light load).5 38 Voltage (kV) 38.2 10.4 Ocurrences (%) Active Power (MW) Figure 6.2 34 34. Following.9 10 10.7 6 4 2 0 64 6 4 2 0 33. Voltage at MOSSORO and power flux from the wind farm in light load condition (mean wind speed 10m/s).5 36 36. Figure 6.85 242.17 Statistics wind speed. 114/152 Ocurrences (%) Ocurrences (%) 0 242.9 242.1 Voltage (kV) 243 242. the voltage variations are within 0. The deepest voltage levels occur on the highest active power production because the total power flow in the network cause voltage drops as introduced in chapter 2.5 37 37.
The reactive power demanded from the wind farm depends on the active power and on the voltage on MOSSORO.2 236.17. However.9 0 100 200 300 Time (s) 400 500 600 Wind Power Production in Mossoro 80 Active Power (MW) 75 70 65 0 100 200 300 Time (s) 400 500 600 Figure 6. The relation between the different parameters is not linear and it becomes more complex from the upper to the lower part of Figure 6.The voltage and reactive power variations are related to the active power production from the wind farm that depends on the aggregate wind speed.18 presents the time series of the wind power produced and the voltage on bus MOSSORO 230kV at mean wind speed 10 m/s and turbulence intensity of 20% in heavy load condition.2 Heavy Load Condition Following.4 Voltage at PCC 236. Higher wind speed leads to higher active power production that leads to higher reactive power demand that leads to lower voltages. which explain the different statistical distribution of the voltage. 115/152 . 236.3 Voltage (kV) 236.17. Figure 6. 6. The time series to heavy load condition are very similar to the ones with light load condition. the mean value of voltage is different and the nonlinear relation of the voltage and reactive power leads to a different statistical distribution of the voltage and reactive power as presented in Figure 6.4.18 Voltage at MOSSORO and power flux from the wind farm in heavy load condition (mean wind speed 10m/s).1 236 235.2.
This section illustrates the impacts of largescale integration of wind power to the frequency and voltage regulation of large power systems. Norway.95 236 236. the wind power is connected to transmission lines (high voltage). in addition. The dynamic voltage variations in this power system are very small. reactive power and the voltage that in this case.5 36 36. active power. However due to the lower mean voltage. reactive power and voltage variations on MOSSORO (Heavy load).25 236.35 34 34. Those differences can be related to the nonlinear relation between the active power.9 235.2 236.1 236.3 Case Study 3: Power System Interactions – NORDEL The Nordic countries are expected to face large wind energy integration in the years to come. This power system includes hydro.3 236. The low correlation between different wind speeds (acting on each wind turbine) smoothes the power produced by the wind farm that reduces the power flow variation hence lower voltage variation. the distributions of the reactive power and voltage at this time are different. the lower voltage level increases the losses.5 6 4 2 0 235.Ocurrences (%) 6 Wind speed (m/s) 4 2 0 9. 6. 116/152 Ocurrences (%) .4 Ocurrences (%) Active Power (MW) Ocurrences (%) Reactive Power (MVAr) Figure 6.2 10.8 9. A second explanation resides on the power smoothing effects from the spatial distribution of the wind turbines. coal fired and even geothermal) and nuclear power stations supplying in parallel electricity to more than 20 million inhabitants.05 236. similarly the active power produced in heavy load condition is very similar to the light load condition.5 35 35. thermo (oil. and Sweden. which are relatively strong and have lower losses that reduce the voltage drop.3 10. The voltage is more concentrated to the mean value and the reactive power more spread.5 37 37. The Nordic power system has generators spread over large distances including Denmark. This is related to two main conditions.7 6 4 2 0 65 6 4 2 0 33. Iceland.9 10 10. Finland.1 10.5 Voltage (kV) 38 66 67 68 69 70 71 72 73 74 75 9. first. The wind speed is exactly the same as the one applied to the light load condition.19 Statistics wind speed.15 236.
6.3.1
Power System Description
Finland, Sweden, Norway, and a part of the Denmark are synchronized in the Nordel system. The Nordel, hereby defined, permits power exchange between the Nordic countries (Denmark, Finland, Iceland, Norway and Sweden), whose primary aims are optimise and efficient electricity market. Here, the objectives are the dynamic power quality and voltage assessment of the synchronized interconnected Nordel power system. Figure 6.20 illustrates the Nordic Power System [75].
Figure 6.20 High voltage Nordic power network [75].
The synchronous Nordel power system includes the eastern part of Denmark (Zealand), Finland, Norway, and Sweden. The western part of Denmark is not synchronized to the Nordel but to the European System. There are no interconnections between Iceland and the rest of the Nordel power system. There are HVDC links between the synchronized Nordel and Germany, Poland and the western part of Denmark. In the synchronized Nordel, the frequency and voltage are regulated in order to remain within certain limits. The frequency regulation includes stability requirements and power exchange agreements. The frequency in normal conditions on the 400kV network must be within 49.9 and 50.1Hz [76]. Under exceptional conditions, the frequency can vary within 47.5 and 52Hz, however those conditions are related to transient and island problems. The voltage regulation in the Nordel includes optimised reactive power flux as well as stability requirements. In addition, the voltages in the Nordel during continuous operations can be within 90 and 105% of the nominal values [77]. The harmonious operation of the Nordic power system means that all machines contribute to the voltage and frequency control avoiding power oscillations and voltage deviations. The control characteristics are mainly based on agreements in power exchange and stability requirements. The size and characteristics of the Nordic power system makes it an interesting study case. In the Nordel, the primary frequency control automatically
117/152
keeps the balance of consumption and production and the secondary frequency control manually regulates the power production agreements in certain areas of the power system. The secondary frequency control is not studied here. The combined primary frequency control, in order to keep the system frequency constant, shall react to the frequency variations with 6000MW/Hz. Table 6.6 presents the requirements of frequency regulation to Nordel.
Table 6.6. Requirements on frequency response in the Nordel power system. Country Denmark Finland Norway Sweden Nordel Frequency Response (MW/Hz) 270 1050 2220 2460 6000
To maintain contractual voltage quality and avoid voltage collapse, the Nordel power system must be capable of keeping the voltage within permissible range. The voltage controllers are specified to keep the voltage within permissible level and to provide reactive power reserve to ensure reliable operation. 6.3.1.1 Reduced Nordel Model In the following analysis, a reduced network simulates the dynamic electrical characteristics of the synchronized Nordel power system. The reduced Nordel power system model is explained in [38]. The reduced network represents the synchronized Nordel with 34 aggregate nodes. The aggregate nodes represent the loads and the generation units with the frequency and voltage controllers included. The power system representation is done as follow: • • • • Finnish system: 2 nodes in 400kV; Swedish system: 08 nodes in 400kV and 2 nodes in 300kV; Norwegian system: 11 nodes in 400kV and 11 nodes in 300kV; Danish system is included in the model as constant loads installed to the neighbours systems (these loads can be negative or positive).
The reduced model takes into account the electromechanical dynamic representation using aggregate models to the synchronous generators. The reduced model has 21 equivalent synchronous generators and includes equivalents controllers to the frequency and voltage. The HVDC links are modelled as constant PQ loads to the proper buses. Figure 6.21 presents the network topology of the reduced Nordel system.
118/152
Finland
Norway
Figure 6.21 Reduced model to the Nordic power system [38].
In Figure 6.21, the network in red represents the 400kV grid and the network in green represents the 300kV grid. The aggregate synchronous generators and loads are also plotted in Figure 6.21. Table 6.7 presents the synchronous machines and the aggregate wind farms (AWF) connecting nodes.
Sweden
119/152
2 There is only one AWF installed to SVW.1.Table 6.2 Nordel Characteristics Before connecting the large wind farms. the xaxis presents the real part of the eigenvalue: smaller absolute values in the negative plane means close to instability.7 Nordel reduced machines connection nodes. the Nordel power system is characterized with modal analysis to indicate the main oscillation modes. Using the dynamic simulation tool (SIMPOW). The yaxis presents the damped frequency (in Hz) of the mode (imaginary part of the eigenvalue). the relevant eigenfrenquencies to the Nordel are expressed in Figure 6.22 (without wind power installed).3. Machine # 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 Generators Name Finland SOFGG1_ Finland NOFGG1_ Sweden: STCKGG1_ Sweden: SVSGG1_ Sweden: SVWGG12 Sweden: STRFGG1_ Sweden: SVNGG11 Sweden: SVMGG1_ Sweden: SVNGG12 Sweden: SVWGG11 No sync. All eigenvalues are in the stable region of the plane. the phase angle of the modes and the machines participating on those modes. 120/152 .22. In Figure 6. Machine installed Norway: EASTGG1_ Norway: HDALGG1_ Norway: CNTRGG11 Norway: CNTRGG12 Norway: SOUTGG11 Norway: WESTGG1_ Norway: NWESGG1_ Norway: MID3GG1_ Norway: NOR3GG1_ Bus Name SOFIN NOFIN STCKH SVSW SVW2 STRFN SVN1 SVM1 SVN2 SVW2 KRLSK EAST3 HDAL3 CNTR3A CNTR3B SOUTGG11 WEST300 NWEST3 MID300 NOR300 AWF Yes Yes Yes Yes Yes2 No No No No Yes2 Yes No No No No No No No Yes Yes 6.
five modes (see marked modes in Figure 6.5 0.22 Relevant eigenvalues of the reduced model to the Nordel. 0.5 1.23 presents the interarea phase and amplitudes of participation from the synchronous machines in the second electromechanical mode selected (0.e. In Figure 6.23. 2.5362 0.112578 1 Imag part of the Eigenvalue (Hz) 0. Considering that wind turbines produce relevant power variation up to few hertz. The phase angle of the vectors indicates the tendency of rotor angle deviation of each machine.8 0. The relevant characteristics of the selected modes are: 1. 4.552Hz).297Hz – oscillation of the East part of the system (Norway) system against West part of Sweden Figure 6. 121/152 .318Hz – oscillation of the South Finnish system against to the entire power system. 0.551629 0.317934 0 1. 5. 0. to how extent this machine is involved in the electromechanical mode. 1. 3.6 0.1.536Hz – oscillation of Stockholm node against the entire power system.296631 1. i.5 1 1.552Hz – general oscillation in the entire power system. the power exchange between different machines (power system oscillation). i.113Hz – oscillation of the North Sweden against the power system.2 1 0.22) are characterized using modal analysis. the vectors near to the generators represent the participation of the machines in the electromechanical mode. see Figure 6.5 Real part of the Eigenvalue Figure 6.4 0.e.2 0 0. The modulus of the vector indicates the magnitude of the participation. 1.23.
SVN2__ and STCKH__) and the Norwegian system (particularly the machines installed to NOR300__. HDAL3_.Figure 6. At this particular electromechanical mode.55164 Hz –Nordic Power System.23 Modal analysis of the eigenvalue 0.55Hz.34962 + 0. SOUTH3B_. the south Finnish system (machine installed to SOFIN__). the Swedish system (machines installed to SVN1__. The phase angles from this graphical presentation shows that the South and Southwest of 122/152 . The entire power system can be excited in the frequency of 0. NWEST3__) all have strong participation in this electromechanical mode.
Eight sets of 550MW wind farms simulate the wind power integration to the Nordel.21 identifies the buses’ name).3. The Finnish system is in most of the electromechanical modes. the Finnish system is expected to suffer severe oscillations compared to the other machines although the oscillations shall not increase in time. it has the less damped electromechanical mode. Country Finland Norway Sweden Wind Power Installed 1100MW – installed 50% to NOFIN and 50% to SOFIN 1100MW – installed 50% to MID300 and 50% to NOR300 2200MW – installed as four sets of 550MW to: SVW.2. in addition. Therefore. Table 6. The aggregate wind speed model generates time series to each wind farm and the 10 time series are averaged to produce a single aggregate wind speed. i. The wind speed angle is 90°. Table 6.8 presents the wind power plans simulated in this section. however the data is fitted based on similar size of wind turbines. 6. which means that 10 wind turbines will face strong correlated wind speed. and Sweden.1 Aggregate Wind Farm Model The aggregate wind farm is scale up of a 1. the distance between each wind turbine is approximately four times the rotor diameter (256m) where the rotor diameter is 64m. The 550MW aggregate wind farm is composed of 10 wind farms with identical layout and identical 30 wind turbines each.Norway oscillates in phase with the SOFIN__ and both of them oscillate against (i. The dynamic AWF presented in chapter 5 models the wind power to the Nordel where sets of 550MW are installed to specific nodes on the Nordel (Table 6.3. where to this wind turbine. directly connected to the network with induction generator and without power electronics. A total of 4400MW of wind turbines are simulated to Finland. The 550MW AWF represents 300 wind turbines of approximately 1. The wind turbines are a conventional type.83MW. There is no correlation between the 10 wind farms.9 presents the main characteristics of the 1. almost opposite phase directions) the north of Norway and to some extent to the north of Sweden. 123/152 .8 Wind power plans simulated (Figure 6.e. The rotor angle stability in this power system representation from modal analysis does not present any unstable operation problem. Table 6.8). Moreover. 6. Hence. the Swedish machine installed in node STCKH oscillates in a different direction because of the complex power system operation.e. all the amplitudes of the power oscillations shall decay in time. KRLSK. and STCK.83MW wind turbine that formally does not exists.2 Wind Power Projects and Representation The wind power plants were chosen based on national wind power prognosis.83MW wind turbine used in the simulations.11. All electromechanical modes are stable. Norway. SVSW. The layout of each small wind farm is similar to Figure 6.
In Figure 6.83MW wind turbine. 124/152 .24 Aggregate wind farm power simulation.Table 6. two of 12m/s. and two of 16m/s all of them with turbulence intensity of 20% that is considered a strong turbulence intensity.83MW 64 m 1500 rpm 21 rpm 660kVAr 50Hz 60 m 71.24. two of 10m/s.05Hz Eight different aggregate wind speeds are generated: two of 8m/s.43 1. The wind power produced from each aggregate wind farm is presented in Figure 6. Figure 6. two time series of power produced from the AWF with mean wind speed at 8. Nominal Power Rotor diameter Generator speed Rotor speed Capacitor Bank Electrical frequency Tower height Gearbox ratio First torsional frequency Estimated logarithmic damping to drive train Number of blades 3p frequency 1.0Hz 5% 3 1. The larger standard deviation of the power occurs in the average wind speed of 10 m/s followed by the powers at 12m/s and 8m/s.25 presents the time series of the active power variations from different wind speeds. At 16 m/s is the lowest standard deviation of power produced. 240 Power (MW) 08 m/s Average wind speed 220 200 180 380 0 100 200 300 400 500 10 m/s Average wind speed 360 340 320 480 0 100 200 300 400 500 600 600 Power (MW) Power (MW) 460 440 12 m/s Average wind speed 420 555 0 100 200 300 400 500 16 m/s Average wind speed 550 600 Power (MW) 545 0 100 200 300 Time (s) 400 500 600 Figure 6.9 Basic characteristics of the 1. 10 12 and 16m/s are presented.24.
The power variation from the aggregate wind farm is very smooth compared to the power produced from a single wind turbine mainly because of the low wind correlation on large distances.00 300 6.00 500 Aggregate Wind Power (MW) 10.00 0.00 400 8.6%.00 Figure 6. However.26 presents the power characteristics: mean. maximum. the standard deviation can be even more relevant to express the power variation of the entire process than the range of power variation because the wind power does not vary instantaneously between the maximum and minimum.26 Aggregate wind farm power characteristics (at 20% turbulence intensity). Standard Deviation (MW) 125/152 . at 12m/s (5. 600 12.2%) and at 16m/s. Figure 6. The largest difference between maximum and minimum powers is at 10m/s (7% of the wind farm rated power) followed by the ones at 8m/s (6% of the rated power).00 200 4.00 100 max 0 6 8 10 12 Wind speed (m/s) 14 16 18 min Meanvalue Standard Deviation 2.25 Aggregate wind speed 8 m/s Aggregate wind speed 10 m/s Aggregate wind speed 12 m/s Aggregate wind speed 16 m/s 20 15 10 Power variation (MW) 5 0 −5 −10 −15 −20 −25 0 100 200 300 Time (s) 400 500 600 Figure 6. which the difference is reduced to 1. minimum and standard deviation produced by the AWF with the 8 time series of wind speed.25 Aggregate wind farm power variation. The differences between the maximum and minimum powers (in less than 10 minutes) are relevant to the frequency controllers of the power system because they specify the maximum power range.
The average wind speed simulated to Norway is 10m/s.3. as presented in chapter 5. arbitrarily. are installed to the Nordel where they simulate the largescale integration of wind power. these nodes have a considerable participation on power oscillations. Figure 6.1 Frequency controllers The performance of the frequency controllers and of the power systems depends on several aspects. the random seeds have influences on the standard deviation. Each one of the 8 cases presented above.10 Aggregate wind farms average wind speeds. However.3.3. Table 6. The power oscillations are generated by the power unbalance and the properties of the speed regulators among other thinks. 6. Table 6. was selected because from modal analysis. new random seeds are used in order to avoid correlated time series of wind speeds and. to some extend. which have the largest power variation. One of the relevant aspects here is the characteristic of the speed governors from all synchronous generators.10 presents the wind applied to the NORDEL where it is simulated large wind power variation to FINLAND.3 Wind Power Impacts on the Power System Voltage and Frequency Regulation The mean wind speed to each AWF has been assigned.The power characteristics of the aggregate wind farms in different wind speeds are as expected. because Finland has presented the less damped electromechanical modes of oscillations. AWF # 1 2 3 4 5 5 6 7 8 Country/ Near Generator Finland SOFGG1_ Finland NOFGG1_ Sweden: STCKGG1_ Sweden: SVSGG1_ Sweden: SVWGG12 Sweden: SVWGG11 Sweden: No sync. 126/152 . machine Norway: MID3GG1_ Norway: NOR3GG1_ Bus Name SOFIN NOFIN STCKH SVSW SVW3 SVW3 KRLSK MID300 NOR300 Wind speed (m/s) Case I Case II 16 12 16 12 8 8 8 8 12 16 12 16 12 16 10 10 10 10 6.27 presents the simulated power balance of the entire Nordel as well as the total power variation from the wind turbines and from the synchronous generators. two cases are simulated: Case I which simulates large power variation from wind power (12m/s) and Case II which simulates small wind power variation (16m/s). Different wind speeds are applied because it is not expected that the entire power system would have the same average wind speed. to the two nodes. 3 There is only one AWF installed to SVW. The standard deviation presented two quite different values because to each aggregate wind farm.
28 presents the time series of the speed and the total standard deviation of the power from each machine in the Nordel.27.27 depends on the set points and characteristics of all generators and changing the average wind speed from different places did not significantly modify the total results presented in Figure 6. 127/152 .27 Power variations and power balance in the Nordel case studied. from all AWF and the total instantaneous power balance are plotted. In principle. In general. However. the speed controllers are acting individually and independently in each machine. therefore some power oscillations between different machines cannot be presented in the total instantaneous power balance because they can be cancelled from those machines that the power oscillations are out of phase. In order to verify the dynamic operation of the power system. The power balance is the difference between the power variation from the synchronous machines and the wind farms. The secondary frequency control is not modelled. the dynamic power balance presented in Figure 6. the power balance varies in a narrow range (within ±10MW) meaning a good combined response of the speed governors. the total wind power production is included. the total instantaneous variations of active power produced from all synchronous machines. The total instantaneous variations are computed by adding the power from all machines at each time. in addition. In Figure 6.27.60 total synchronous machines power variation total wind power variation power balance 40 20 Power (MW) 0 −20 −40 −60 0 100 200 300 Time (s) 400 500 600 Figure 6. Figure 6. so the set points of the synchronous generators were adjusted aiming to keep the regional power balances.
to the Norwegian system and to the STCKH_ both of them with the higher wind power variations that explains the higher power variation.31) and the machine #1 (Finland that showed a strong electromechanical participation in the oscillation modes) are presented in Figure 6. frequency and standard deviation of power (mean wind speed at 12m/s to Finland). In the lower part of Figure 6. the standard deviation of the power from each synchronous machine is presented. 3p and torsional moments). where Table 6.28.29. the power spectral distribution of speed and power to the machines with high standard deviation (#9 and #7 from Figure 6. In order to analyse in detail the dynamic behaviour of the power system with large wind power integration.01Hz.99 and 50. Similarly. The slow frequency variations are related to the total slow wind speed variation and the fast speed variations are related to the dynamic power variations from the wind turbines (e.28 Wind power. This bus is connected to the Finnish system.3200 Wind Power (MW) 3150 3100 Total wind power to NORDEL 3050 50.g. a reduction in the wind power leads to a decrease in the frequency.98 400 STD of the power of each machine 300 200 100 0 0 2 4 6 8 10 Machine # 12 14 16 18 20 0 100 200 300 time (s) 400 500 600 Figure 6. The higher power variation was presented by machine #9 that is connected to bus SVN2__. 128/152 Std of Power (MW) . The frequency in Figure 6.01 50 49.28 represents the speed in each machine that is well regulated and the variations are within 49.99 Speed of each machine 49.02 0 100 200 300 time (s) 400 500 600 Frequency (Hz) 50. An increase in the wind power leads to an increase on the frequency because the load on synchronous machines is reduced and the speed drop characteristics of the speed governors lead to an operational frequency slightly above the rated.7 presents the machine names.
129/152 .29 Power spectral distribution of the power and speed of selected machines (mean wind speed at 12m/s to Finland).30 presents the power spectral distributions for the same machines similar to Figure 6. the rotational speed of the AWF is modified to 10 rpm.5 3 10 −5 Power spectra (Hz2) 10 −6 Speed @ machine # 9 Speed @ machine # 7 Speed @ machine # 1 10 −7 10 −8 10 −9 0 0. The wind power variations excite the power system in several modes.318Hz as expected because of its electromechanical mode of oscillation. The machine #1 presents a strong variation at 0.5 1 1. the wind power excites the power system with a frequency around 1Hz (3p) that influences several machines because at this frequency there are some electromechanical modes related to several machines as presented in the modal analysis. The power spectral distributions of the speed variations in the machines are very small.5 1 1. i. The electromechanical mode around 0. Figure 6. The machine #7 presents a higher speed variation at 1. All the other machines present higher power spectral speed variations at the 3p frequency because that is the frequency that the wind turbine excites the power system. As some important electromechanical modes are near to 0.5Hz.10 2 Power spectra (MW2) Active power @ machine # 9 Active power @ machine # 7 Active power @ machine # 1 10 0 10 −2 0 0. This slow speed of rotation leads to a 3p effect around 0.5 Frequency (Hz) 2 2.1Hz because that is the electromechanical mode associated to this machine. The electromechanical mode at 0.5 Hz and taking into account that as the wind turbines grow in size there is a tend to reduce the speed of rotation.55Hz is also excited.318Hz is excited in the Finish system (machine #1). 3p from the wind turbines at low frequency 0.5 3 Figure 6.5 Frequency (Hz) 2 2.5Hz that will excite less damped modes of the entire power system.29 at this new condition. In this case.e.
5 3 10 −5 Power spectra (Hz2) 10 −6 Speed @ machine # 9 Speed @ machine # 7 Speed @ machine # 1 10 −7 10 −8 10 −9 0 0. 130/152 . here the mean wind speed applied to Finland is modified from 12 m/s to 16m/s in order to identify larger modifications on the results. however at the frequency of 0.5 Frequency (Hz) 2 2.30.5Hz)). Figure 6. the power variations by machine #1 are much higher than the previous case.5 Frequency (Hz) 2 2.1Hz that is related to its electromechanical mode. the wind power strongly excites the power system controllers.5 1 1.5 1 1. After having analysed the simulations to high wind power variation by the AWF in the Finish power system. high wind speed to Finland).10 2 Power spectra (MW2) Active power @ machine # 9 Active power @ machine # 7 Active power @ machine # 1 10 0 10 −2 0 0.31 presents the results of the dynamic simulation for case II (i.30 Power spectral distribution of power and speed to selected machines (AWF modified to lower rotational speed (3p~0. The power spectral distributions of the speed variations on the selected machines are much higher particularly to machine #1 mostly because the proximity to the less damped electromechanical modes. In Figure 6.5 3 Figure 6. it is possible to verify the isolated excitation of the machine #7 at 1.29. In this case.318Hz. The power variations on the 3p effect are very similar to the ones presented in Figure 6.e.
the standard deviations of power to each synchronous machine have slightly changed.02 0 100 200 300 time (s) 400 500 600 Frequency (Hz) Speed of each machine 50.99 49.3. There are no significant differences between Case I and II.31 Wind power. The average frequency of the power system is slightly higher than the Case I meaning that it could be necessary an adjustment of the set points of the frequency controllers in order to keep the power production within the country areas. Std of Power (MW) 131/152 . frequency and standard deviation of power (average wind speed 16m/s to Finland).98 400 STD of the power of each machine 300 200 100 0 0 2 4 6 8 10 Machine # 12 14 16 18 20 0 100 200 300 time (s) 400 500 600 Figure 6.2 Voltage Quality Using the case I wind speed distribution and the original AWF (i.e. The differences in mean values come from the voltage set points specified to the machines in the power system. 3p frequency around 1Hz).01 50 49.3200 Wind Power (MW) Total wind power to NORDEL 3150 3100 3050 50. 6.32 presents the voltage simulated on all machines in the Nordel network. The voltages are in p. This is related to the fact that the speed governors of the synchronous machines shall react to the power variations from the nearest wind power avoiding the power flow between different areas of the power system. Figure 6.3.u. In the case II. the voltage on the Nordel network is analysed.
The total wind power installed to the power system is the same as presented in the previous cases.005 0 0 2 4 6 8 10 Machine # 12 14 16 18 20 Figure 6.015 0 100 200 300 time (s) 400 500 600 Std of Voltage (%) STD of the voltage of each machine 0. and the middle part of Figure 6.02 Voltage deviation @ machine #4 Voltage deviation @ machine #7 Voltage deviation @ machine #19 0 −0. anyway.3200 Wind Power (MW) Total wind power to NORDEL 3150 3100 3050 1.95 0.u.) 1 0.32 is only presented to show that the variations are very small meaning that the voltage regulators operate properly. The voltage levels are not easy to identify in the graph.33). similar to the frequency control analysis.02 −0.33 Power spectral distribution of voltage and voltage deviation to selected machines in the Nordel (Finland AWF at 12m/s).32 Wind power and voltage deviations simulated in Nordel system (mean wind speed 12m/s to Finland).04 Voltage deviation (%) 0. 132/152 . The voltage quality does not seem to be an important issue here. 10 4 Power spectra (V2) Voltage deviation @ machine #4 Voltage deviation @ machine #7 Voltage deviation @ machine #19 10 2 10 0 0 0.05 Voltage for all machines 0 100 200 300 time (s) 400 500 600 Voltage (p.04 0 100 200 300 Time (s) 400 500 600 Figure 6.5 Frequency (Hz) 2 2.5 3 0.01 0.5 1 1. the voltage variation to selected machines with high standard deviation of voltage are detailed as follow (Figure 6.
it is important to note that the voltage variations simulated are very small (160V at 400kV) and it can be influenced by the precision of the simulations. In the original AWF. The time series of the voltage deviations in this new condition agrees with the power spectral distribution. The excitations of the synchronous machines are intrinsically related to the speed of the machine hence the presences of the modes were expected. Similarly to the frequency control analysis. The voltage deviations in time domain are within ±0.34 Power spectra distribution of voltage and voltage deviation simulated in Nordel System (mean wind speed 12m/s to Finland AWF low frequency (3p=0. This is explained because at the 0. Three cases were analysed to illustrate the problems on the power quality and stability when integrating large amount of wind power. 133/152 . 7 and 19 are presented.5 3 0. voltage quality and the frequency control were investigated. However.05 Voltage deviation (%) Voltage deviation @ machine #4 Voltage deviation @ machine #7 Voltage deviation @ machine #19 0 −0. the power spectral distributions of the voltage are much higher at the 3p frequency.05 0 100 200 300 Time (s) 400 500 600 Figure 6.34 presents the power spectral distribution of the voltage deviation as well as the time series of the voltage deviations on machines # 4. the AWF was modified to rotate at 10rpm (3p=0. the power spectral distribution and time series of the voltage deviations to machines #4. 7 and 19.33.5Hz an eigenfrequency of the power system is excited. The power spectral distributions of voltage show that the high variations occur at the 3p frequency from the wind turbine. Figure 6. 6.04% of the rated voltage that is very small. the last three figures must be analysed with caution. It also shows that the electromechanical modes also influence the voltage controllers. In this case.5 1 1.05%.5Hz)).4 Case Analysis Remarks Power quality and stability problems from large wind farms to the network have been investigated here. 10 4 Power spectra (V2) Voltage deviation @ machine #4 Voltage deviation @ machine #7 Voltage deviation @ machine #19 10 2 10 0 0 0.5 Frequency (Hz) 2 2.In Figure 6. namely the voltage stability. where the amplitudes now are limited to ±0. Those machines were selected because they presented higher standard deviation of voltage.5Hz. the corresponding power spectral peak is near to 10000V². Hence. the voltage power spectral at 1Hz were around 1000V².5Hz) in order to analyse the influences of the wind power on exciting the modes of the power system. now that the 3p is around 0.
However.4. In this case. The frequency control has been investigated by modal analysis and dynamic simulation using the AWF model. which can regulate the reactive power from the wind turbine. Following detailed remarks to each case are presented. The voltage stability depends also on the network characteristics. The voltage quality has been investigated by means of dynamic simulations using the Aggregate Wind Farm. 6. The wind farms will be installed to the transmission level hence the voltage variations on the low voltage levels area can be influenced. the dispersion of the wind power over extensive areas smoothed the power variations from the wind turbines. The worstcase scenario showed voltage variations within 0. the voltage stability was in general reduced with wind power because of the reactive power demanded by the wind turbine generators. however different load conditions were supplied improving the analysis. The voltage stability was analysed with loadability curves to the power system with and without large amount of wind power. the model must be extended to include the distribution grids and the voltage quality on the consumers must be assessed. In this power system studied. fact that is deeply related to the voltage stability. fact that is different from the previous case. the analysis of voltage stability and dynamic voltage quality of a Brazilian power system was presented. where it increased the loadability limits of the local power system.1 Remarks on Case 1 The first case presented only voltage stability analysis. new wind turbine technologies using electronic power converters. 6. The smoothing effects on the power fluctuation from the wind farm cased small voltage fluctuations on the power system. a relative low reactive power is flowing in the network when compared to case 1 that explains the improvements on the voltage stability. When distribution grid data is available. The voltage stability was studied with similar methodology as the one used in the case 1. The wind power had small impact on the voltage stability.3 kV at 230kV nominal voltage.4.2 Remarks on Case 2 In the second case. where the voltages on the network were modified in each condition due to the reactive power compensation schemes. The smoothed power variations caused small than expected voltage and frequency deviations in the high voltage network.The voltage stability was illustrated by means of loadability curves drawn to the power system. improved the voltage stability and it can actively contribute to regulate the voltage. 134/152 . The voltage stability was addressed in heavy and light load conditions. From the studied cases. The conventional wind turbines reduce the reactive power availability in the power system. The dynamic voltage variation has been studied using dynamic simulations with aggregate wind farm. The studied cases showed that the wind power can contribute to improve the voltage stability and that the reactive power is an important issue from large wind power concerning the voltage stability. The case represents the integration to the transmission level (220kV) where the reactances inductive of the lines and the reactive power compensation schemes played an important role.
6. the set points of the nearest synchronous machines were adjusted to maintain the areas (countries) balances. proposed specific buses to which the large scale wind power should be simulated. Using modal techniques. a sensitive analyse was carried. only the dynamic voltage variations were presented.4. This assumption made ease to change the assigned average wind speeds to different AWFs. The layout of the wind farms. The application of the AWF to the real wind farms layout is straightforward when it is available. 6.e. The Nordic power system (here called Nordel) is very strong. where in Norway. i. similar to case 2. However. the time series of the aggregated wind speeds can be shifted to different AWF s without generating new time series. which the wind power is planned. Except for the AWFs installed in Finland and Norway. The wind speeds in Finland and Norway were chose with high wind power variations because in some specific machines installed in these countries. The frequency control was related to dynamic stability and power quality. Hence.4. where the high wind power variation (12m/s) in Finland was replaced with the smallest at mean wind speed 16 m/s. the 20% turbulence intensity is expected to be the worst scenario to normal operation of wind turbines.1 Frequency control The frequency control was primarily analysed with modal techniques. from Norway. The automatic secondary generation control unit was not implemented. The connection points of the AWFs were assumed to buses in Sweden and Finland. 10. The wind farms were aggregate in similar blocks of 550MW. Here some assumptions were taken: • Two cases for each average wind speeds of 8. presented high participation on electromechanical modes at low frequency.3. the average wind speeds were assigned arbitrarily. some of the most significant electromechanical modes were identified and 135/152 . The voltage control was related to the voltage quality and in this case. assumed 90° wind direction leading to rows of 10 wind turbines with high correlated turbulence that might be considered one worstcase scenario.3 Remarks on Case 3 Case 3 analysed largescale integration of wind power influence on the voltage and frequency controllers. The cases covers a large extend of the power curve from wind turbines in addition. • • • • • The main conclusions from the simulations are classified on voltage and frequency control as following. EFI Sintef. 12 and 16 m/s were simulated with turbulence intensity of 20%. The Finnish system presented the less damped mode so the high wind power variations were assigned to both buses in Finland (12m/s) and because of the same reason 10m/s was assigned to Norway. Aggregate wind farms (AWF) represented the largescale wind power in the Nordel based on the national plans of wind power.
6. The less damped modes were below 0. With modal analysis was possible to characterize also the most important machines participating on each electromechanical mode and how was the direction of the relative rotor angle speed variation of the machines. the grid must be detailed and the voltage analysed properly. particularly on the machines with high participation on the electromechanical modes that was expected.6Hz and all electromechanical modes were stable.3. They are well damped (frequency controllers well designed) so the power oscillations caused from the wind power in the Nordel did not increase in time. but its specific location in addition to the electromechanical characteristics of the power system made this machine the most sensible in the simulations. 136/152 .5Hz. In order to verify the extreme impacts of wind power in the less damped electromechanical modes of the power system. This observation might indicate that the power variations from the wind power do not necessarily influence only locally but it can be transferred to remote parts of the power system. the dynamic simulation revealed that some modes that were excited from the wind power on the frequency analyses were also present on the voltages. Anyway. the aggregate wind turbine was modified to have the 3p effect at 0. This bus did not have wind turbines installed. the voltage variations were small. however the wind turbines directly excite the less damped electromechanical modes hence higher power and speed variations at low frequencies were encountered. The dynamic analyses of largescale wind power indicated that the voltage quality was not a concern on high voltage networks and it might be more relevant in lower voltage levels including the distribution networks. The AWF completes the analysis because it provided means of dynamic simulation to verify the behaviour of the frequency controllers (speed governors) under normal operation. when necessary. the electromechanical modes were also identified.05%).10% variations (±0. However. The wind power simulated did not represent a problem to the voltage quality. One of the machines that presented high participation on the electromechanical modes from the modal analysis was the machine #9 connected to bus SVN2__ that was also identified on the dynamic simulation with higher speed and power variations during normal operation with wind power. At those grids. The voltage controllers worked properly compensating the reactive power demanded from the AWF and keeping the voltage within 0.4. the limited voltage regulation and the losses might lead to large voltage variations from wind power and become a problem so.2 Voltage controllers The voltage controller’s performances were analysed by dynamic simulation using the AWF.characterized. From the dynamic simulations. Under this new condition the power system still stable.
The aggregate wind farm model permits an easy assessment of the power system dynamic operation from largescale integration of wind power. the identification of voltage deviations and inadequate frequency regulation can be done in few minutes even for large power systems.e. extended the power limit of the voltage collapse) of the power system. Dynamic models to largescale wind power that can assess the power system stability and quality were presented. The assessment of power quality parameters such as voltage and frequency deviations due to the inclusion of large amount of wind power demands simulations of the wind farms power production and its interaction with the power systems. the problems for the power systems stability that are mainly related to power oscillations were studied with modal analysis to identify the most relevant characteristics. Modifications of the wind turbine characteristics. The voltage stability was influenced by the wind power integration. Wind turbine technologies with power converters that can actively control the reactive power consumption increased the voltage stability (i. the aggregate model is very suitable to identify the impact of largescale wind power in the dynamic operation of the power system because it is time inexpensive and it includes the most relevant characteristics of the wind turbines power dynamics. The voltage stability was analysed with loadability curves to the power system. It includes analysis of the wind power influences on the voltage stability. The aggregate wind speed that is applied to the aggregate wind turbine model was also developed and presented. The power smoothing effect from the wind farms power production have been verified in the aggregate wind speed model to correctly represent the large wind farms power production. The dynamic models are based on aggregation procedures of wind turbines. Dynamic simulations of a high number of wind turbines are unpractical for very large power systems. the reactive power flux to the wind farms will reduce the voltage stability limits. Using the aggregate models.Chapter 7 7 Conclusions A methodology to investigate the wind power influences on the power system was presented in this thesis. A loadability computation tool was developed in this thesis and a static model to the wind power on the voltage stability was presented.e. Related to the power quality problems. It includes relevant characteristics of the turbulence and its spatial correlation effects on the power produced from wind farms. power system stability and power quality characteristics. The applications of the modal analysis with the dynamic simulations are recommended in order to assess possible power system stability problems. The main conclusions for the voltage stability are that although the wind power alleviates the active power fluxes in the network. particularly the turbulence and its spatial correlation. In addition. were simulated improving the voltage stability. where the reactive power was the main factor. the studied cases show that higher number of wind turbines smoothes the power variations consequently the flicker phenomena can be 137/152 . application of power electronics. The power system stability and power quality were investigated with dynamic simulations. Therefore. i.
The inclusion of this wind speed forecasting and geographical disposition of wind farms improves analyses of the dynamic integration of the largescale wind power in the power systems. The frozen turbulence. it is important to validate and improve the aggregate wind farm model by obtaining experimental measurements from large wind farms. the drive train dynamic operation is simplified to include the most relevant modes of oscillation of the power variations. Finally. Before introducing the aggregate wind farm. The aerodynamic power conversion including the dynamic stall is included in the aeroelastic model. The simulated dynamic voltages were within limits and the flicker was not a concern however in the cases studied here. the dynamic simulations showed that the wind power do not impact only local machines but it can influence remote parts because of the power system electromechanical characteristics hence it is necessary to investigate the entire power system operation with largescale integration of wind power. The wind turbine model described the most relevant aeroelastic components of wind turbines for simulation of power fluctuations. In addition. the network was strong and the wind power was connected to the high voltage network (transmission lines). the extension of the model to variable speed wind turbines is also envisaged. rotational sampling turbulence and tower shadow effects are the most relevant to power quality hence included in the wind speed model. here the integration of wind speed forecasting in extensive areas with the aggregate wind speed model is proposed. In the near future. The inclusion of the loads characteristics and its time variation to the analysis of the dynamic power system operation it also proposed here as future work. as future work. the variable wind turbines will account to large power production. The largescale wind power integration interacts with the power system controllers. The aggregate wind farm presented some limitations that can be improved in future works. The forecast model can generate the mean wind speed (10 minutes) and turbulence intensity to specific locations that can be used as input to the aggregate wind speed model. Moreover.considerably reduced. The power variation in broad frequency range excite some electromechanical modes of the power system however the frequencies variations were so low that it does not endanger the dynamic stability or cause serious power oscillations in the power system. Nevertheless. The interaction of the loads in addition to the wind power variation on the power system dynamic operation can be relevant to characterize the power oscillations. The frequency control was also investigated using a reduced model of the Nordel power system. The dynamic wind turbine model represents the most relevant wind speed and wind turbine characteristics to power quality characteristics. The integration of different tools will improve the analysis of the power system characteristics with benefits to the analysis of the frequency secondary control in the power system and analysis of real conditions as well as storm fronts. The inclusion of the wake effect phenomenon in the wind farm and the deterministic parts from the wind speed can be new tasks. dynamic wind turbine models have been presented. 138/152 . In addition.
8
[1] [2] [3]
Reference List
IEC 6140021 Measurement and assessment of power quality characteristics of grid connected wind turbines. International Electrotechnical Commission, FDIS 2001. “Integration of Renewable Energy in the Electrical Network in 2010”, draft final report, EU ALTENER PROJECT XVII/4.1030/Z/99115, BrusselsBelgium, 2001. YihHuei Wan and Brian K. Parsons, “Factors Relevant to Utility Integration of Intermittent Renewable Technologies”, NREL Report NREL/TP4634953, Colorado, USA, August, 1993. John Eli Nielsen, Søren Varming and Christian Gaardesatrup, “Review of Technical Options and Constraints for Integration of Distributed Generation in Electricity Networks”, Report SUSTELNET, (online: www.sustelnet.net/documents.html) Netherlands, December, 2002. Jørn Saxow, Svend Skovgaard, Peggy Friis, “Power Plant Operation Correlated with wind energy production within the Elsam Utility Area – status 1995”, Elsamprojekt A/S Power Station Engineering Report EP96/404a, Denmark, June, 1996. John Olave Tande, Per Nørgaard, Poul Sørensen, Lars Søndergaard, Preben Jørgensen, Anders Vikkelsø, Jacob Dyring Kledal, Jørgen S. Christensen, “Power Quality and grid Connection of Wind Turbines”, Summary report, Vol. 1, 2 and 3, Risø report RisøR853, Roskilde, Denmark, October, 1996. Poul Sørensen, Troels Friis Pedersen, Gehart Gerdes, Rainer Klosse, Niel Robertson, Wily Davy, Maria Koulouvari, Evangelis Morfiadakis and Åke Larsson, “European Wind Turbine Testing Procedure Developments, Task 2: Power Quality”, Risø report Risø R1093(EN) (online: www.risoe.dk/rispubl/VEA/ veapdf/risr1093.pdf), Roskilde, Denmark, May 2001. Åke Larsson, “The Power Quality of Wind Turbines”, PhD dissertation (online: www.elkraft.chalmers.se/Publikationer/EMKE.publ/Abstracts/LarssonD.abstr.html), Department of Electrical Power Engineering, Chalmers University of Technology, Göteborg, Sweden, 2000. John Olave Giæver Tande, “Applying Power Quality Characteristics of Wind Turbines for Assessing Impact on Voltage Quality”, In Wind energy, 5, 2002. Vladislav Akhmatov, Arne Hejde Nielsen, Hans Knudsen, “Electromechanical interaction and stability of power grids with windmills”, in proceedings of the IASTED International Conference Power and Energy Systems, Marbela , Spain, September, 2000.
[4]
[5]
[6]
[7]
[8]
[9] [10]
139/152
[11]
Marianne Bruntt, Jan Havsager, Hans Knudsen, “Incorporation of Wind Power in the East Danish Power System”, in proceedings of IEEE Power Tech’99 conference, Budapest, Hungary, September, 1999. Jan Wiik, Jan Ove Gjerde and Terje Gjengedal, “Impacts from Large Scale Integration of Wind Farms into Weak Power Systems”, in Proceedings of PowerCon 2000, Power System Technology, Australia, 2000. L.L. Freris, “Wind Energy Conversion Systems”. Prentice Hall, London, UK, 1990. Erich Hau, WindTurbines Fundamentals, Techonlogies, Application, Economics, Spring Verlag, Germany, 2000. Siegfried Heier, Grid Integration of Wind Energy Conversion Systems, John Willey and Sons, England, 1998. J. Wilkie, W.E. Leithead and C. Anderson, “Modelling of Wind Turbines by Simple Models”, Wind engineering vol. 14, No. 4, 1990. Estanqueiro, A. I. “Dynamic Models of Wind Parks” (in Portuguese: Modelação Dinamica de Parques Eólicos), PhD dissertation, Instituto Superior Técnico, Universidade Técnica de Lisboa, Portugal, 1997. Thomas Petru, “Modelling of Wind Turbines for Power System Studies”, Licentiate technical report Nr. 391L (online: www.elkraft.chalmers.se/Publikationer/ EMKE.publ/Abstracts/Petru2001.abstr.html), Department of Electric Power Engineering, Chalmers University of Technology, Sweden, 2001. Poul Sørensen, Anca Hansen, Lorand Janosi, Jonh Bech and Birgite BakJensen, “Simulation of interaction between wind farm and power system”. RISØ report R1281(EN) (online: www.risoe.dk/rispubl/VEA/veapdf/risr1281.pdf), Roskilde, Denmark, December 2001. Ana I. Estanqueiro, R. Aguiar, J. A. Gil Saraiva, Rui M G. Castro and J. M Ferreira de Jesus, “Development and Application of a model for power output fluctuations in a wind farm”, in proceedings of the European Wind Energy Conference, LübeckTravemund, Germany, March, 1993. Gregor Giebel, “On the Benefits of Distributed Generation of Wind Energy in Europe”, PhD dissertation (online: gregie.hjem.wanadoo.dk/Ggiebel_Distributed WindEnergyInEurope.pdf), Reihe 6, report Nr. 444,VDI Verlag, Düsseldorf, Germany, 2001. H.G. Beyer, J. Luther, R. SteinbergerWillms, “Fluctuations in the Combined Power Output from Geografically Distributed Grid Coupled Wind Energy Conversion Systems – An Analysis in the Frequency Domain”, Wind engineering Vol 14, No. 3, 1990.
[12]
[13] [14] [15] [16] [17]
[18]
[19]
[20]
[21]
[22]
140/152
[23] [24] [25]
Prabha Kundur Ed, “Power System Stability and Control”, EPRI Power Engineering Series, McGraw Hill, 1993. Paul M. Anderson, A. A. Fouad, “Power System Control and Stability”, IEEE Power System Engineering Series, New York, 1993. System dynamic performance subcommittee of the Power system engineering committee of the Power Engineering Society “Eigenanalysis and Frequency Domain Methods for Power Systems: Dynamic Performance”, IEEE 90TH02923PWR. IEEE, Piscataway, NJ, USA, 1989. Lei Wang, Meir Klein, Solomon Yirga and Prabha Kundur, “Dynamic Reduction of Large Power System for Stability Studies”, IEEE transactions on Power System, vol. 12 No. 2, May, 1997. Carson W. Taylor, “Power System Voltage Stability”, EPRI Power Engineering Series, McGraw Hill, 1994. Thierry Van Custem, Costas Vournas, “Voltage Stability of Electric Power System”, Kluvwer’s Power Electronic and Power Systems Series. Series Editor: Thomas A Lipo and M.A. Pai, Brussels 1998. System dynamic performance subcommittee of the Power system engineering committee of the Power Engineering Society, “Voltage Stability of Power Systems: Concepts, Analytical tools, and industry experience”, IEEE 90TH03582PWR, Piscataway, NJ, USA, 1990. Claudio Cañizares, “Voltage Stability Assessment: Concepts, Practices and Tools” Power System Stability Subcommittee Special Publication IEEE/PES, Final Document (thunderbox.uwaterloo.ca/~claudio/claudio.html#VSWG), August 2002. S. Persaud, B. Fox and D. Flynn, Impact of remotely connected wind turbines on steady state operation of radial distribution networks, in Proceedings of Generation Transmission and Distribution, Vol. 147, N3, May, 2000. IEC 8680 Flickermeter Functional and design specifications CEI/IEC 8680, 1986; IEC 8680 Amendment 1 to Flickermeter Evaluation of flicker severity. CEI/IEC 8680, 1990; IEC 8680 Flickermeter Evaluation of flicker severity. CEI/IEC 8680, 1991; Johansson, S., "Longterm Voltage Stability in Power Systems  Alleviating the Impact of Generator Current Limiters". Technical Report No. 335, Chalmers University of Technology, Göteborg, Sweden. 1998.
[26]
[27] [28]
[29]
[30]
[31]
[32] [33] [34] [35]
141/152
Lund Institute of Technology. Chalmers University of Technology. 1997. Canada.lth. and O. 1990. Sundar. India. CENELEC EN 50160. “Power System Oscillations”. M.iea.Structural Aspects of Controlling Active Power”. D. Olof Samuelsson. “Performance of Wind Turbines in a Turbulent Atmosphere”. Dissertation. Sweden.. Technical Report No. Pai. November. Padiyar. “Analysis of Subsynchronous Resonance in Power System”. Department of Industrial Electrical Engineering and Automation. Journal of Solar Energy. 2001. Wyngaard. J.. Jorgen Thirstrup Petersen and Helge Aagaard Madsen. J. Izumi. Y. R. Roskilde. [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] 142/152 . Winkelaar.D. thesis (www. 1992.iea.A. dissertation (www. 2000. RisøR993 (EN). Norway. C. 1999. vol. Department of Automatic Control.pdf). Bjørn Harald Bakken. R. Lund. Paper presented at the 10th ASME Wind Energy Symposium. “Power System Damping . Lund. Series Editor: M. 1972.A. P.lth. “Fast threedimensional wind simulation of stochastic blade loads”. “Technical and Economical Aspects of operation of thermal and hydro power systems”. Trondheim. Houston. Quarterly Journal of the Royal Meteorological Society. Ph. Sweden. D. Kluvwer’s Power Electronic and Power Systems Series. C. PhD dissertation. 98. “Local Inflow and Dynamics – measured and simulated on a rotating Wind Turbine Blade”. “Voltage characteristics of electricity supplied by public systems”. Series Editor: M. Kluvwer’s Power Electronic and Power Systems Series. 16. Sweden. Lund Institute of Technology. Department of Industrial Electrical Engineering and Automation. R. NTNU.se/~ielolof/olof/PhDthesis. “Coordinated Voltage Control in Electric Power Systems”.D.. Cote. "Voltage stability simulations using detailed models based on field measurements". Ph. 2024 January 1991. Graham Rogers. 31. Kaimal. 1994. “Damping of Power Oscillations in Large Power Systems”. Ph. K. No. Göteborg. Sweden. J. Lund. report Nr. Pai. 6 pp. 567575. 1997. and Sulivan. Mats Larsson. “Spectral Characteristics of Surface Layer Turbulence”.D.. Vol. Risø report.se/publications/point/Theses/pdf_files/LTHIEA1025. 1983. Lund Institute of Technology. European standard.[36] Karlsson. Denmark September 1997. DRAFT. 230. Bo Eliasson.pdf).
December 1995. “Design Basis for Horizontal Axis Turbines – Theoretical Background”. Quarterly Journal of Meteorology Society 87. “Model for power spectra of the blade of a wind turbine measured from the moving frame of reference”. “The atmospheric boundary layer below 150 metres”. Methods for Calculation of the Flicker Contributions from Wind Turbines. Risø publication RisøR749(EN). pgs 194 – 211. 1974. Roskilde.Sc. Roskilde. Gunner C. 1st edition may 1992. Denmark 1984. “Turbulence Modelling for Gust Loading”. Henrik Bindner. Infield. Data Item 72026. H. Panofsky. “Dynamics and Fatigue Damage of Wind Turbine Rotors during Steady Operation”. Schelz and D.. Vol. Risø report R512. Sten Frandsen. 2000. UK. “Rotor loading on a threebladed wind turbine”. No. BoundaryLayer Meteorology 87. Courtney. 1998. Larsen. dissertation. [52] [53] [54] [55] [56] [57] [58] [59] [60] [61] 143/152 . Risø publication RisøM2836. M. Poul Sørensen. Pedro Rosas. Madsen. Kassel. 1972. Davenport. April 1994. Annual Review of Fluids Mechanics. CREST Loughborough University. 10 (1982) 249262. vol 6 pp.147177.. Poul Sørensen and Michael S. Wiebke Lagrander. 1996. Denmark. Leif Kristensen and Sten Frandsen. Peter Hauge. “Models for Variable Speed Wind Turbines”. Journal of Wind Engineering and Industrial Aerodynamics. W. G. July 1987.[48] [49] [50] [51] Madsen and Rasmussen. Germany. Roskilde. A. Denmark. “The spectrum of horizontal gustiness near the ground in high winds”. Characteristics of wind speed in the lower layers of the atmosphere near the ground: Strong wind (neutral atmosphere). Danish Standard DS 472. “last og sikkerhed for vindmøllekonstruktioner”) (in English: Danish standard for loads and safety of wind turbines construction. Risø National Laboratory. ASCE Journal of Structural Engineering. December 1989. in Proceedings of European Wind Energy Conference. G. 113. Denmark. Roskilde. in proceedings of the European Wind Energy Conference EWEC’89. 7. Risø report Risø – I – 939 (EN). Glasgow 1989. A. “Fast Wind Modelling”.Kluwer Academic Publishers Netherlands. “Frequency Domain Modelling of Wind Turbine Structures”.. Sten Fradsen. Engineering Science Data Unit. Poul Sørensen. ESDU London. Poul Sørensen. William E. Solari. “Horizontal two points coherence for separations greater than the measurement height”. Holley and Jens Carsten Hansen.
Denmark. February. Mc Anulty (Ed. Castro and J. 9. “New model for single unit representation of induction motor loads. A. Power System National Operator (Operador Nacional do Sistema) – online: www. Akbaba. “Improving dynamic aggregation of induction motor models”.NNE.com. R. M.[62] Stig Øye. Pedro Rosas.br/ons/sin/index.eolica. Journal of Wind Engineering. including skin effect. J. Report number: R920(EN) Risø National Laboratory. A.htm. Morten Hartvig Hansen. Morelato. “Active Control: Wind Turbine Model”. November 1994. n. M. Report number 6R. “Analysis of Electric Machinery and Drive Systems”. “Dynamic Stall – simulated as time lag of separation” in Proceedings of the 4th IEA Symposium on the aerodynamics of wind turbines. IEE Proceedings – B Vol. Henrik Bindner. 2002. Franklin. n. 4. M. Brasilia.com.ons. Wind models for prediction of power fluctuations from wind farms. 89 October 2001. RisøR1395(EN). Saraiva. for power system transient stability studies”. Data base – Power system analysis in Portuguese (Base de Dados . F. Sweden. Brazil. “Comparison of experimental and aggregate induction motor responses”. in Proceedings of British Wind Energy Conference BWEA’15. North – Northeast Power System Diagram – from December 2002 (in Portuguese). Torbjörb Thiringer.H Alden. n. 1997. Q. Aguiar. 6.Estudos Elétricos). 1999. Sundhoff. Denmark. R. 2000. 2002. pgs:18951900. G. Krause. Torben J. 2003. pg: 19341941. n.I. Risø report. Jesus. R. Estanqueiro.C.). WilleyInterscience. 4. Roskilde. Oleg Wasynczuk and Scott D. Anca Hansen. Paul C. November 1994. November 1992. reference case to the year 2004. Fakhro. “An extended model of drive train dynamics”. 139. Hans Ganader. IEEE Transactions on Power Systems Vol. A. DU_Rede Oper. Chalmers University of Technology. Poul Sørensen. Larsen. York 1993. “Generator dynamics – analysis and simulations”. Y.br/index_por. second edition. D. National Operator of the Power System [63] [64] [65] [66] [67] [68] [69] [70] [71] [72] [73] [74] 144/152 . D. 9. F. Rome Italy 1991. “On the effect of utility grid characteristics on wind park power output fluctuations”. IEEE Press.R20_261202. S. Recife.html. Approved wind projects to Brazil on line www. IEEE Transactions on Power Systems Vol. G.T. Department of Electric Power Engineering. Brazil. T. J lem.
tidsavvik. 2002. Department of Electric Power Engineering. online: www. 1995. Torbjörb Thiringer. reguérstyrke og reserve) NORDEL available online at www.nordel.ONS: 06/09/01 CICLO 0204. “Recommendations for frequency. February. A. Chalmers University of Technology. Communication from ABB/SIMPOW Induction Machine Model – September 2002. Technical Report no.nordel. P. 1996. 2001. regulation and reserve” (Rekommandasjon for frekvens.. PhD dissertation. IEEE transactions on Power Systems. Sweden.br/ons/sin/index. “Operational Performance Specifications for Thermal Power Units larger than 100MW” NORDEL available online: www. 2002. Rosas. C.ons. [75] [76] “Nordic High Voltage Network” (“Det Nordiske Høyspentnette”) on line diagram at www.org. time deviation. [77] [78] [79] [80] [81] 145/152 . Brasilia. vol 16. No 1.org/Content/Default. Copenhagen.: 293. Brazil.asp?PageID=130.nordel. 2001.asp?PageID=125. 1996. “Wind Power Influences on the Voltage Stability”. in proceedings 2001 European Wind Energy Conference. Torbjörn Thiringer and Jorma Luomi. “Measurements and Modelling of LowFrequency Disturbances in Induction Machines”. Comparison of reduced order dynamic models of induction machines.com.htm.org/Content/Default.
146/152 .
DYNPOW – solves the differential equations to the power system in time. DSL – Dynamic Simulation Language. DYNPOW. it is possible to either analyse the electrical characteristics or export to an ASCII file that can be analyse in any other program. The electrical components comprise the electrical generator. DYNPOST – plots the outputs. stepup transformer. The electrical components are modelled using the standard library of SIMPOW/ABB. It is a dynamic simulation program dedicated to the power system problem. Any dynamic simulation demands at least the use of the upper four modules. To power quality assessment from wind turbines the OPTPOW. transmission lines. The transmission lines formally does not belong to the wind turbine unity but it is an important component in wind farm simulations therefore it is included here. The interface of the DSL and the DYNPOW modules is the mechanical torque. The DYNPOST module provides the graphical output of the simulation. The DSL module is the dynamic simulation language. control system. The DYNPOW solves the differential equations to the power system in the dynamic simulation. STATPOW – solves the short circuit equations. transformers and power electronics and controllers. the dynamic models of the electrical machinery and controllers of the machines are specified. which the electrical characteristics of the machine and the active power produced are specified. we implement the mechanical model of the wind turbine. transmission lines. The wind turbine model in this program is an asynchronous machine.1 Electrical Components Model in SIMPOW The SIMPOW has been developed since 1977 in ABB. In the DSL. and the mechanical speed and torque in the electrical machine. It is divided into 5 programs: • • • • • OPTPOW – solves the network equations (static solution) using iterative method. the reactive power compensation.9 Annexes 9. the power fluxes. Using DYPOST. The electrical components in a conventional wind turbine are: 147/152 . The wind turbine model is mostly done in the DSL module where the DYNPOW send the speed of the machine and the DSL return the torque on the main shaft of the electrical generator. The program automatically computes the voltage in all nodes. In the DYNPOW module. The SIMPOW library has verified models to most of the conventional electrical components in the power system including electrical machines. DSL and DYNPOST are used. The OPTPOW computes the load flow in the power system. and the entire power system.
) Electrical Generator Model Current (p. Rm is the resistance representing losses and Xm is the magnetization reactance.u.1. It is mostly applied as motor. 9. R’2 is the rotor equivalent resistance referred to the stator.u. Where R1 is the stator resistance the X1 is the stator reactance inductive.) Figure 9.1. Electrical generator model parameters. X’2 is the rotor equivalent reactance inductive referred to the stator.1 presents the main model parameters to the generator. Electrical generator equivalent.u.u. however. The electrical parameters derive from the equivalent circuit presented in Figure 9.• • • • • Asynchronous generators. it has been used for many years in wind turbines. The inputs to the model are the torque that is computed from the aeroelastic module (in DSL). X Air gap Torque (p.2. The rotor resistance is modelled as a function of the rotor speed in 148/152 . Lines (in a wind farm) and. Stepup transformers. Electrical parameters: R.) Speed (p.1 Electrical generator The asynchronous generator is the most common type of electrical machine. the voltage and the frequency that comes from the next electrical component. Reactive power compensation Power system representation Details of the electrical power system including the wind farm depend on the purpose of the simulation. Figure 9. As generator. The dynamic asynchronous generator model was implemented using the SIMPOW library. R1 X1 X'2 I2 Rm Xm R'2/s V1 I1 Figure 9.2.) Voltage and frequency (p. The outputs are the mechanical speed (feed in the aeroelastic and wind model) and the current (feed in the next electrical component).
synchronously rotating axis [65] and [80].1.) Electrical parameters: R. The transformer model uses the electrical parameters from a “T” electrical equivalent transformer model (see Figure 9. R2 is the secondary 149/152 .) Current (p. All units are in per unit and the voltages and currents are vectors with an angle and amplitude. High Voltage side X Stepup Transformer Model Voltage and frequency (p.2 Stepup Transformer In these studies. The model itself is a socalled park model in d and q rotating reference frames where the transients of the stator are neglected.u. The SIMPOW converts all parameters to a direct – d – and a quadrature – q – rotating axis so called park transformation. The dq reference frame is rotating at synchronous speed as determined by the electrical angular frequency of the impressed stator voltage.4). the voltage from the other side and the reference frequency to the rotating dqframe. X’2 is the secondary equivalent reactance inductive referred to the primary side. i. Here.e. the current inputted comes from the wind turbine generator and from the reactive power compensation unity on the wind turbine terminals. 9.3 Structure of the transformer model. The characteristics and parameters of the model used are presented in Figure 9.u.3. The configuration of the transformer depends on the installation. the wind turbine is connected to the low voltage side (stepup transformer). a simple a symmetrical 3phase without saturation transformer model is used.order to model the skin effect in the rotor.) Voltage (p.u. Where R1 is the primary resistance the X1 is the primary reactance inductive. The saturation can be also stated. Low Voltage side Current (p. The inputs to the model are the current from one side. The third order model to induction machines represents reasonable the power quality characteristics [78] and [79]. The model outputs the current in the high voltage side and the voltage on the low voltage side.u. This model is also called as a third order model reduced from a fifth order.) Figure 9. I2 V2 R'2 X'2 Rm X2 Xm R2 I1 V1 Ideal Transformer Figure 9.4 Electrical transformer model.
during normal operation. d Sending node Current (p.side equivalent resistance referred to the primary side.u.) Figure 9.) Line Model Receiving node Voltage (p. X. Other technologies uses power electronics to compensate the reactive power. Rm is the resistance representing losses and Xm is the magnetization reactance. but here only the noload capacitor bank supplies the reactive power during continuous operation. small steps of capacitors can be connected to follow the reactive power demanded from the generator.6 Electrical transmission model.u. Electrical parameters: R. 9. BC is the equivalent susceptance capacitive.5 Electrical Transmission structure model. The model inputs are the one voltage and one current and it outputs the other voltage and current. Figure 9.1.1.) Voltage (p. It is modelled as a synchronous machine or by an infinite node connected through the short circuit impedance. 9. Where R is the equivalent resistance and the X is the equivalent reactance inductive of the line.1. The frequency reference is used to transform all variable to the dqrotating frame and to compute the reactance in the lines. A small capacitor bank supplies the noload reactive power to the asynchronous generator and.u.3 Reactive Power Compensation.5 Slack bus The slack bus is also reefed as reference node.8 150/152 .u. Figure 9. Here.6) I2 R X I1 V2 Bc/2 Bc/2 V1 Figure 9. 9. The electrical parameters come from the π electrical equivalent for the lines (see Figure 9.) Current (p. B. A switched capacitor bank usually supplies the reactive power in the wind turbines.4 Lines and Cables Lines and cables can be modelled in different ways.5 presents the main structure of the basic cable/lines models. a concentrated π models the electrical behaviour of the lines and cables where depending on the distances and type of cables/lines it can be modified.
the reference angle.u. the Xsc is the short circuit reactance inductive.8. A representation of the positive sequence of the slack bus is presented in Figure 9.) Figure 9.) Voltage (p. X Current (p.u.) Reference angle (Deg. The electrical characteristics are the equivalent impedance at the PCC also called short circuit impedance.7 Slack bus structure. I1 Rsc Xsc Infinite Node V1 V0 . The model outputs the voltage at the PCC.8 Slack bus model (positive sequence). The inputs to the model are: the voltage.) Reference node Model Voltage (p. where Rsc is the short circuit resistance. V0 is the voltage reference in the infinite node that is fixed.u. and the currents computed from the other models. 151/152 . Electrical short circuit parameters: R.presents the use of an infinite node to replace the power system with the short circuit impedance at the PCC.Constant Figure 9.
152/152 .
ISBN : 8791184169 .
This action might not be possible to undo. Are you sure you want to continue?