Energy Conversion and Management 123 (2016) 35–45

Contents lists available at ScienceDirect

Energy Conversion and Management

journal homepage: www.elsevier.com/locate/enconman

Aerodynamic optimization of the blades of diffuser-augmented wind

turbines
Jerson R.P. Vaz a,⇑, David H. Wood b
a
Faculty of Mechanical Engineering, Institute of Technology, Federal University of Pará, Av. Augusto Correa, N 1, Belém, PA 66075-900, Brazil
b
Department of Mechanical and Manufacturing Engineering, Schulich School of Engineering, University of Calgary, Calgary T2N 1N4, Alberta, Canada

a r t i c l e i n f o a b s t r a c t

Article history: Adding an exit diffuser is known to allow wind turbines to exceed the classical Betz–Joukowsky limit for a
Received 26 February 2016 bare turbine. It is not clear, however, if there is a limit for diffuser-augmented turbines or whether the
Received in revised form 13 May 2016 structural and other costs of the diffuser outweigh any gain in power. This work presents a new approach
Accepted 6 June 2016
to the aerodynamic optimization of a wind turbine with a diffuser. It is based on an extension of the
Available online 16 June 2016
well-known Blade Element Theory and a simple model for diffuser efficiency. It is assumed that the same
conditions for the axial velocity in the wake of an ordinary wind turbine can be applied on the flow far
Keywords:
downwind of the diffuser outlet. An algorithm to optimize the blade chord and twist angle distributions
Blade optimization
DAWT
in the presence of a diffuser was developed and implemented. As a result, an aerodynamic improvement
Diffuser of the turbine rotor geometry was achieved with the blade shape sensitive to the diffuser speed-up ratio.
BET In order to evaluate the proposed approach, a comparison with the classical Glauert optimization was
performed for a flanged diffuser, which increased the efficiency. In addition, a comparative assessment
was made with experimental results available in the literature, suggesting better performance for the
rotor designed with the proposed optimization procedure.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction
The addition of an exit diffuser to a horizontal-axis wind
turbine is one of the few ways in which power output may be
increased in cost-effective manner, as recently noted by
Al-Sulaiman and Yilbas [1]. There is an extensive literature on
diffuser-augmented wind turbine (DAWT) performance. Rio Vaz
et al. [2] developed an innovative approach for the performance
analysis of DAWTs based on blade element theory (BET), in which
a more general semi-empirical one-dimensional analysis was car-
ried out. Glauert’s correction for the induction at high thrust was
also employed. Their results yielded good agreement with experi-
mental data. Most recently, Kosasih and Hudin [3] investigated the
impact of turbulence intensity on micro wind turbine efficiency in
converting the wind energy into power. The performance of bare
micro wind turbine (MWT) and diffuser-augmented micro wind
turbine (DAMWT) models subject to different levels of turbulence
was reported. It was shown that shrouding the turbine with dif-
fuser increases the power coefficient C P by a factor of almost
two. Beyond a certain tip-speed ratio, the performance of both
MWT and DAMWT was shown to decrease with turbulence
⇑ Corresponding author.
E-mail address: jerson@ufpa.br (J.R.P. Vaz).
intensity, however the C P of the DAMWT was still greater than that
of the bare MWT wind indicating the diffuser augmentation was
still achievable even at high freestream turbulence.
Jafari and Kosasih [4] performed a Computational Fluid
Dynamics (CFD) study, where the augmentation is strongly
dependent on the geometry of the diffuser, such as length and
expansion angle. Also, they reported that a higher area ratio creates
greater pressure reduction at the diffuser exit, which increases the
mass flow rate, agreeing with Hansen et al. [5]. They employed a
one-dimensional CFD analysis to evaluate a DAWT composed of a
NACA 0015 airfoil on an ideal turbine, concluding that the power
coefficient for a shrouded turbine is proportional to the mass flow,
and the increasing flow through the rotor induced by the diffuser
increases the extracted power for the same thrust coefficient
compared to a bare wind turbine.
Wang et al. [6] measured the influence of a flanged diffuser on a
shrouded wind turbine. Their experimental results revealed that
the rotational speed and the dynamic strain of the blade are much
higher than those without a flanged diffuser. Abe and Ohya [7]
combined a numerical and experimental investigation of flanged
DAWTs. They suggested that the loading coefficient for the best
performance of a flanged diffuser is considerably smaller than for
a bare wind turbine. In addition, it was necessary to avoid
boundary-layer separation and maintain a high pressure-recovery

http://dx.doi.org/10.1016/j.enconman.2016.06.015
0196-8904/Ó 2016 Elsevier Ltd. All rights reserved.
36 J.R.P. Vaz, D.H. Wood / Energy Conversion and Management 123 (2016) 35–45

Nomenclature

Latin symbols Pc pitch of the vortex sheets (m)

a; a0 axial and tangential induction factors at the rotor r radial position at the rotor plane (m)
aopt optimum axial induction factor R radius of the rotor (m)
A; A3 area of the disc and the cross section at the diffuser s spacing between the lamina (m)
outlet u axial velocity at the rotor plane, u ¼ V 1 ¼ V 2 (m/s)
B number of blades u1 axial velocity in the wake, u1 ¼ V 4 , (m/s)
c chord (m) v0 velocity of the external stream, v 0 ¼ cV 0
CD drag coefficient V0 freestream wind velocity (m/s)
CL lift coefficient V3 axial velocity at the diffuser outlet (m/s)
CM torque coefficient w total induced velocity (m/s)
Cn normal force coefficient W relative velocity (m/s)
cp3 pressure coefficient at the diffuser outlet x local-speed ratio
CP power coefficient
C Popt optimum power coefficient Greek symbols
Ct tangential force coefficient a angle of attack (rad)
CT thrust coefficient b area ratio
dA elementary area (m2) c diffuser velocity ratio
dM elementary torque (N m) e velocity ratio
dP elementary power (W) gd diffuser efficiency
dT local thrust (N) h twist angle (rad)
F Prandtl tip-loss factor k tip-speed ratio
p0 pressure in the external flow (Pa) q density of the fluid (kg/m3)
p2 pressure at the turbine upstream (Pa) r solidity of the turbine
p3 pressure at the diffuser outlet (Pa) / flow angle (rad)
P output power (W) X angular speed of the turbine (rad/s)

coefficient, to give high performance. Ohya and Karasudani [8]
developed a turbine within a diffuser shroud with a broad-ring
brim at the exit. The shrouded wind turbine power was increased
by a factor between 2 and 5 over a bare wind turbine with the
same blade diameter and wind speed. This was because a low-
pressure region, due to a strong vortex formation behind the broad
brim, draws more mass flow through the blades. These aspects
highlight the importance of the development of models capable
of optimizing DAWTs [9]. Although there are several works avail-
able in the current literature on DAWTs, the authors are unaware
of any study of blade optimization with a diffuser.
Rio Vaz et al. [10] suggested that the optimum design of a
DAWT might be achieved through three different approaches.
The first one uses classical BET, which models the energy conver-
sion by the torque generated on the blade elements, as demon-
strated by Glauert [11]. Fletcher [12] applied this analysis to
DAWTs, including wake rotation and blade Reynolds number
effects. Igra [13] compared BET with data obtained from a
shrouded wind turbine with a rotor diameter of 3 m and exit-
area ratio of 1.6, producing 0.75 kW at 5 m/s with a power aug-
mentation ratio (over the bare turbine) of 2.4.
The second approach is based on vortex theory, where each of
the rotor blades is replaced by a lifting line and a vortex sheet is
continuously shed from the trailing edge, as further described by
Okulov and Sørensen [14]. The bound vorticity produces the local
lift on the blades while the trailing vortices induce the velocity
field in the rotor plane and in the wake. A recent example of this
approach, Wood [15], determined the maximum performance of
a bare turbine at low tip-speed ratio. Bontempo and Manna [16]
used a single ring vortex to analyze the flow around a ducted actu-
ator disk, in order to describe the flow around DAWTs.
The last methodology is CFD, e.g., Nobile et al. [17]. They
undertook a computationally-expensive simulation of an
augmented vertical axis wind turbine. CFD can be also coupled
with optimization methods using genetic algorithms and
evolutionary computation [18]. Global optimization using CFD
can be very time-consuming, e.g. [19], and suggests the use of
alternative optimization procedures, mainly those with cheaply
computed objective functions. The present paper describes such
a method, based on BET and the semi-empirical approach
described by Rio Vaz et al. [2], where the conditions to extend
the BET for the diffuser were detailed. In further support of this
approach, it is noted that BET generally agrees well with experi-
mental data [13]. In this paper the geometry of the rotor is deter-
mined using a one-dimensional analysis, assuming that, as for bare
turbines, the gauge pressure in the far-wake is zero. This hypothe-
sis aims at developing a theory having the closest equivalence to
the momentum relations for bare turbines, as described, for exam-
ple, by van Bussel [20]. In order to evaluate the proposed approach,
comparisons with the classical Glauert optimization [11] and
experimental results available in the literature were made.
The remainder of this paper is organized as follows. The next
section introduces the simple one-dimensional axial momentum
theory with a diffuser, showing the expressions for the maximiza-
tion of DAWT power coefficient. Section 3 provides a detailed
explanation of BET and its extension to DAWT. Section 4, shows
the proposed optimization, and depicts the relationship for the
optimum axial induction factor as a function of the diffuser
speed-up ratio. Results and discussion are stated in Section 5,
where the optimum conditions proposed for DAWTs are presented.
Section 6 shows the conclusions of this study.
2. Axial momentum theory with diffuser
Simple momentum theory considers the air frictionless and
ignores the rotational velocity component [2]. To model a diffuser
with losses, an approach similar to that used to determine duct
flow in the presence of losses [12,21] is developed. Fig. 1 shows
the control volume for an ideal DAWT.
 J.R.P. Vaz, D.H. Wood / Energy Conversion and Management 123 (2016) 35–45 37

Fig. 1. Simplified illustration of the velocities at the rotor plane and in the wake [2]. The rotor lies between stations 1 and 2 and is surrounded by the diffuser.

The power coefficient, C P , for the classical theory with diffuser is C P ¼ c4að1  aÞ2 ð8Þ
given by [21]:
" # Applying a momentum balance, as in classical theory, using the

2
V4   control volume shown in Fig. 1, the thrust coefficient is given by
CP ¼ e 1  ð1  gd Þ 1  b2 e2 ð1Þ C T ¼ 4að1  aÞ, which combined with Eq. (8) gives C P ¼ eC T . Note,
V0
if c ¼ 1, Eq. (8) describes a bare turbine. It is also noted that C P
where V 0 is the freestream velocity, V 4 is the velocity in the far- has a maximum value, where the optimum axial induction factor,
wake, e ¼ V 1 =V 0 is the speed-up ratio, V 1 ¼ V 2 is the velocity at aopt , becomes 1/3, giving the optimum thrust coefficient,
the rotor plane, b ¼ A=A3 , where A is the rotor area (assumed equal C Topt ¼ 8=9, and the optimum power coefficient, C Popt ¼ c16=27.
to the diffuser area at the rotor), A3 is the cross sectional area of the Therefore, the theoretical limit for a DAWT is a function of the dif-
diffuser outlet, and gd is the diffuser efficiency, defined as fuser geometry which controls c. Similar conclusions were also
p  p2 obtained by other authors, e.g., van Bussel [20] who used energy
gd ¼ 3  ð2Þ conservation to determine C P ¼ bw4að1  aÞ2 , where w represents
1
2
q V 22  V 23
the back pressure at the diffuser outlet. In this case, both parame-
where p2 and p3 are the static pressures at the rotor plane and at the ters b and w depend only the diffuser geometry. Similarly, Dick’s
diffuser outlet which are positions 2 and 3, respectively in Fig. 1, as [22] equation for the C P of a DAWT was the product of a mass con-
well as V 2 and V 3 , and q is the air density. centration coefficient, an energy augmentation coefficient and an
Applying the energy balance downstream to the diffuser exit, extraction coefficient, all strongly influenced by the diffuser geom-
gives the relationship for the velocity V 4 : etry. Although the thrust equation is mathematically similar to that
in the conventional actuator disc theory, the responses are different,

2
V4 because the axial induction factor at the rotor plane is influenced
¼ b2 e2 þ cp3 ð3Þ
V0 significantly by the diffuser. Maximum power can be obtained by
setting dC p =da ¼ 0 in Eq. (7). The result is an expression for the opti-
The pressure coefficient, cp3 , at the diffuser outlet is [2]
mum axial induction factor, aopt , which depends on c; gd , and b:
p3  p0  
cp3 ¼ ð4Þ 4 þ 3c2 ðgd  1Þ b2  1
1
qV 20 aopt ¼  2  ð9Þ
12 þ 3c2 ðgd  1Þ b  1
2

where p0 is the static pressure in the freestream flow. Substituting

Eq. (3) in Eq. (1), yields Substituting Eq. (9) in Eq. (7), the optimum power coefficient is
     given by
C P ¼ e 1  cp3 þ e3 gd 1  b2  1 ð5Þ
256c
Note that C P and cp3 depend on V 4 . It is assumed that C Popt ¼    2 ð10Þ
V 4 ¼ ð1  2aÞV 0 , where a is the conventional induction factor at 27 4 þ c gd  1Þ b2  1
2ð

the rotor, and the same condition for the axial velocity in the wake
of an ideal bare wind turbine is applied in the flow after the diffuser If c ¼ gd ¼ 1 Eqs. (9) and (10) reduce to the Betz–Joukowsky limit
outlet. The diffuser speed-up ratio, c, is defined as the ratio between (aopt ¼ 1=3 and C Popt ¼ 16=27). Thus, Eqs. (9) and (10) are general-
the maximum axial flow velocity in the diffuser and the freestream izations of the Betz–Joukowsky relation for a DAWT.
velocity. Thus, the velocity approaching the disc becomes Another important effect on DAWT design is the wake rotation.
V 1 ¼ V 2 ¼ cð1  aÞV 0 and e ¼ cð1  aÞ. Eqs. (4) and (5) give As is well-known from the literature, there is no rotation in the
   wake of a conventional actuator disk, but rotation is an essential
cp3 ¼ 1  b2 c2 1  a2 þ að3a  2Þ ð6Þ part of power extraction, in that the elemental torque, dM, is
and obtained directly from the angular momentum equation applied
   to an infinitesimal control volume of area, dA ¼ 2prdr [23]:
C P ¼ cð1  aÞ2 4a þ c2 ð1  aÞ 1  b2 ðgd  1Þ ð7Þ

If no losses occur in the diffuser (gd ¼ 1), Eq. (7) results in

dM ¼ qV 1 wr 2 dA ¼ 2qa0 cð1  aÞV 0 Xr 2 dA ð11Þ
38 J.R.P. Vaz, D.H. Wood / Energy Conversion and Management 123 (2016) 35–45

where r is the radius, w ¼ 2Xa0 is the angular velocity in the blade section. The flow angle, /, is obtained from the velocity dia-
near-wake, X is the rotor angular velocity, and a0 is the tangential gram shown in Fig. 2 as
induction factor. The torque coefficient is:
ð1  aÞV 0
tan / ¼ c ð18Þ
dM 4a0 cð1  aÞXr 2 ð1 þ a0 ÞXr
CM ¼ 1 ¼ ð12Þ
2
q V 2
0 dA V0 Consequently, it is possible to express the thrust and torque
coefficients for the B blades in the rotor as
The element power is obtained from
rC n
dP ¼ XdM ¼ 2qa0 cð1  aÞV 0 X2 r 2 dA ð13Þ C T ¼ c2 ð1  aÞ2 2
ð19Þ
sin /
By integrating this expression across the rotor, the power coeffi-
and
cient is given by
Z cð1  aÞð1 þ a0 ÞrX rC t
P 8 k CM ¼ ð20Þ
CP ¼ ¼ ca ð1  aÞx dx
0 3
ð14Þ V0 sin / cos /
1
qAV 30 2
k 0
2 where r ¼ Bc=2pr is the local solidity, and c is the local blade chord.
where x ¼ Xr=V 0 and k ¼ XR=V 0 are the local-speed ratio and the The axial induction factor can be obtained by combining
tip-speed ratio, respectively. C T ¼ 4að1  aÞ with Eq. (19),
a BcC n
¼ c2 ð21Þ
1a 8pr sin /
3. Blade element theory with a diffuser 2

In this section, the diffuser is incorporated into BET by using the while the tangential induction factor is obtained combining
control volume shown in Fig. 1. The induced velocities, V 1 and V 2 , Eqs. (12) and (20), yielding:
shown in Fig. 1, are given by cð1  aÞV 0 , while the flow velocity at a BcC t
¼ ð22Þ
the diffuser outlet, V 3 , using the continuity equation, can be writ- 1 þ a0 8pr sin / cos /
ten as
Clifton-Smith [24] showed that tip loss corrections are important in
V 3 ¼ bV 1 ¼ bcð1  aÞV 0 ð15Þ BET when determining optimum blade shape for maximum power
The elemental normal and tangential force coefficients at any production, besides being one of the most important corrections
location on a blade are defined as in BET for a bare turbine, see in the BET analysis. Prandtl’s tip loss factor, F, is defined as the ratio
Fig. 2, by of the total bound circulation of the blades and the circulation of a
rotor with an infinite number of blades. BET calculations made with
Fn the Prandtl factor show good agreement with free wake vortex the-
Cn ¼ 1 ¼ C l cos / þ C d sin / ð16Þ
2
qW 2 c ory and test data [25]. Prandtl’s factor arises from an approximate
solution for the potential flow a set of translating helicoidal surfaces
and representing the wake of each blade by the flow around the edges of
Ft a two-dimensional set of equally spaced semi-infinite laminae, as
Ct ¼ 1 ¼ C l sin /  C d cos / ð17Þ illustrated in Fig. 3. To include the diffuser influence, the external
qW 2 c
2 velocity v 0 ¼ cV 0 is assumed outside the blade tips within the
where C l and C d are, respectively, the lift and drag coefficients of the diffuser.
airfoil comprising the element. W is the relative velocity on each Wald [26] showed that, if all fluid between the laminae has the
same velocity, the circulation would be v 0 s, where v 0 is the velocity
of the external stream, and s is the spacing of the laminae, taken as
the normal distance between two helical sheets at the same radius.
Consequently, the circulation must be reduced by a factor equal to
2v 0 ðs=pÞ cos1 ½expðpDr=sÞ divided by v 0 s which leads to
2
F¼ cos1 ef ð23Þ
p

Fig. 2. Velocity diagram for the section of the rotor blade. Fig. 3. Illustration of two-dimensional velocity field around the vortex lines.
 J.R.P. Vaz, D.H. Wood / Energy Conversion and Management 123 (2016) 35–45
where f ¼ pDr=s, and Dr ¼ R  r is the distance from the edge of the
lamina. Furthermore, the distance s is a function of B and the pitch
of the vortex sheets, equal to 2pr tan /, as detailed by Branlard [27],
hence
Pc 2pr
s¼ cos / ¼ sin /
B B
Note that Eqs. (23) and (24) are the classical Prandtl’s factor. How-
ever, as sin / ¼ cð1  aÞV 0 =W from the velocity diagram (Fig. 2), the
parameter f becomes:
B WðR  rÞ
f ¼
2 rcð1  aÞV 0
Clearly the diffuser influences the flow angle /, which has a strong
dependence of the diffuser speed-up ratio, as described in Eq. (18).
To consider Prandtl’s tip-loss factor in the optimized approach,
Eq. (23) must be incorporated into Eqs. (21) and (22). If the tip loss
is regarded as the ratio of the induced velocity at the blade to the
streamtube average, e.g. Clifon-Smith [24], then there two impor-
tant observations to be made for DAWTs. First, the diffuser is
assumed to induce a circumferentially uniform axial flow, but no
azimuthal flow, through the rotor. This axial flow should reduce
the tip loss, as should the proximity of the blade tip to the diffuser
wall, in a way that is not possible to incorporate into F as defined
above. Second, there should be different tip loss factors for the axial
and circumferential motion. This paper regards the present formu-
lation as an expedient until a better model is developed.
4. Optimization model for diffuser-augmented wind turbines
The aerodynamic optimization is by maximizing the power
coefficient through maximizing the integrand ca0 ð1  aÞ in
Eq. (14). This requires
 Compute error ¼ j/iterþ1  /iter j.
d da0
½ca0 ð1  aÞ ¼ c ð1  aÞ  a0 ¼ 0
da da
Note that c – 0, and is independent of the induction factors. So,
Eq. (26) can be simplified to an equation that also applies to bare
turbines as used by Clifton-Smith [24]:
da0
ð1  aÞ ¼ a0
da
According to Hansen [23], if the local angles of attack are below
stall, a and a0 are not independent since the force according to
potential flow theory is perpendicular to the local velocity seen
by the blade. The total induced velocity, w, must be in the same
direction as the force, as illustrated in Fig. 2. On the other hand,
when the local angle of attack is above stall, Eq. (27) becomes inva-
lid since the drag, which is ignored in the potential theory, becomes
large. As recently noted by Wood [15], Eq. (27) is only strictly true if
the vortex pitch is independent of r. In particular, for k < 1, the
behavior of the induced velocity field seems to be heavily depen-
dent on the radius. Wood [15] showed that numerical optimization
of bare turbines gave constant pitch only when k was around one or
greater. Therefore, it is important to note that the present optimiza-
tion procedure is valid for k > 1 approximately, for which
x2 a0 ð1 þ a0 Þ ¼ c2 að1  aÞ
derived from the angle / in Fig. 2 in terms of
a0 Xr
tan / ¼
caV 0
and Eq. (18). Eq. (28) differentiated with respect to a yields

2
da0 x
ð1 þ 2a0 Þ ¼ 1  2a
da c
39
If Eqs. (27) and (30) are combined with Eq. (28), the optimum
relationship between a and a0 becomes:
a0 ¼ ð1  3aÞ=ð4a  1Þ ð31Þ
Eq. (31) was obtained by Glauert [11] for bare turbines. For DAWTs,
ð24Þ the optimum relationship between x and a can be obtained substi-
tuting Eq. (31) in Eq. (28) resulting in
"
2 #
2
16a3  24a2 þ 9  3
x
aþ
x
1¼0 ð32Þ
c c
In the high k limit, Eq. (32) requires a ! 1=3, as expected. As
ð25Þ
k # 0; a ! 1=4, whereas the correct limit is 1/2, see Wood [15].
Because Eq. (32), the blade optimization procedure can be
expressed as a function of the induction factors once the blade ele-
ment lift and drag are available. The iterative algorithm for the cal-
culation of optimum chord and twist angle, starting at its most
external section, is described below, using the following as input:
r; c; C L ðaopt Þ; C D ðaopt Þ and V 0 for a given k.
Algorithm 1. Blade optimization procedure.
Set initial values for a and a0 . In this work a ¼ 1=3 and a0 = 0;
for i ¼ 1 to Ns (Number of sections) do
while error > TOL do
iter ¼ iter þ 1
Compute /, using Eq. (18);
Compute C n and C t , using Eqs. (16) and (17);
Compute the optimum values for a and a0 , using
Eqs. (32) and (31);
Compute the optimum chord, using Eq. (21);
Compute the optimum twist angle h ¼ /  a;
Compute the new value for /, using Eq. (18);
ð26Þ end while
end for
Compute the chord and twist angle distributions.
ð27Þ
5. Results and discussion
5.1. Optimizing DAWT power output
The optimizations were done for area ratio b ¼ 0:278. This value
was chosen only for the purpose of evaluating the model behavior.
It is important to note that a large area ratio may cause flow sep-
aration within diffuser, altering its efficiency. The impact of area
ratio upon DAWT performance is still not completely clear, since
increasing area ratio typically results in an increasing recirculation
zone behind diffuser [7], leading in general to a decrease in pres-
sure, but this does not lead to more energy extraction by the tur-
bine. More detail on this regard can be found in Barbosa et al.
[28], who developed a mathematical model based on the Biot–
Savart law to describe the internal velocity for three conical dif-
fusers, taking into account the characteristics of flow around them.
ð28Þ Even for a higher area ratio, the internal velocity remained low.
In order to analyze the influence of the diffuser efficiency, gd , on
the power coefficient, C P , and the behavior of the optimum axial
ð29Þ induction factor, aopt , of a DAWT, Fig. 4 was generated using the
axial induction factor, a, and the diffuser speed-up ratio, c, as inde-
pendent variables through Eq. (7). Furthermore, Eqs. (9) and (10)
were used to calculate aopt , and the optimum power coefficient,
C Popt . Also, Fig. 4 is important in understanding how gd affects c,
ð30Þ
which is still a challenge due to the dependence on the
40 J.R.P. Vaz, D.H. Wood / Energy Conversion and Management 123 (2016) 35–45

1 1
vs. a vs. a
opt
vs. aopt vs. aopt
opt
0.8 0.14 0.8 0.15
0.14 0.15
Axial induction factor, a

Axial induction factor, a

0.24 0.25
0.24 0.25
0.35
0.34
0.35 0.45
0.6 0.34 0.6 0.55
0.44 0.45
0.55 0.65
0.44 0.44
0.4 0.34 0.4
0.24 0.6
0.34 5

0.
0.14

55
0.24 0.65
0.14 0.55 0.55
0.2 0.2 0.45 0.45
0.35 0.35
0.25 0.25
0.15
0.15

0 0
1 1.1 1.2 1.3 1.4 1.5 1.6 1 1.1 1.2 1.3 1.4 1.5 1.6
Diffuser speed-up ratio, Diffuser speed-up ratio,
(a) (b)
Fig. 4. Power coefficient (C P , solid line) as a function of a and c: (a) gd ¼ 0:2 and (b) gd ¼ 0:8.

recirculation in the wake and the diffuser geometry. However, it is gd ¼ 1, the flow rate is increased, contributing for an improved
noteworthy that the increasing diffuser efficiency results in an Cp. The optimum conditions computed using Eqs. (9) and (10) are
increasing diffuser speed-up ratio. Fig. 4a shows the results for ðC T2opt ; C Popt Þ ¼ fð0:78; 0:95Þ; ð1:06; 0:76Þ; ð1:33; 0:62Þ; ð1:61; 0:52Þ;
low diffuser efficiency (gd ¼ 0:2), yielding aopt ¼ 0:5; copt ¼ 1:34 ð1:9; 0:44Þg for gd ¼ f1:0; 0:8; 0:6; 0:4; 0:2g, respectively. The model
and C Popt ¼ 0:44. Note that aopt exceeds the conventional optimum searches the optimum point on each power coefficient curve, ensur-
value of 1/3 for a bare turbine. This occurs because the increasing ing a suitable convergence even if the design parameters are
energy loss along the diffuser by friction reduces the turbine power modified.
coefficient. Since e ¼ cð1  aÞ, the high value of aopt indicates that e To assess the diffuser effect on the optimization procedure, a
decreases for c constant, as observed by Phillips [21], who showed small DAWT was considered, whose blades use the NACA 0012
a significant decrease in e for diffuser efficiencies about 20%. For symmetrical airfoil, with the lift and drag obtained experimentally
gd ¼ 0:8; aopt ¼ 0:4; copt ¼ 1:6 and C Popt ¼ 0:76, as shown in by Sheldahl and Klimas [29] for a Reynolds number of 1:6  105 . In
Fig. 4b, indicating that at high gd ; aopt decreases towards the opti- terms of lift:drag ratio, this is a moderate airfoil chosen to focus on
mum value of a bare turbine. When gd ! 1, the power coefficient the optimization process, rather than airfoil selection. All simula-
becomes dependent only the diffuser speed-up ratio c, demon- tions were done for the optimum angle of attack 8 , with a maxi-
strating that the diffuser efficiency has a strong influence on C P mum C l =C d ¼ 44. The 4 m diameter rotor had 3 blades and hub
for DAWTs. of 0.2 m diameter.
Fig. 5 shows the power coefficient as a function of the thrust As the optimization procedure uses BET, c may vary with radius,
coefficient C T2 (C T2 ¼ C T =e2 , where C T2 is defined using the axial but this possibility is not considered. The average value of c was
velocity at the DAWT rotor, V 1 ). The maximum performance was used: c ¼ f1:0; 1:4; 1:8g for results shown in Fig. 6, and
obtained for gd ¼ 1 (no losses in the diffuser), leading to a c ¼ f1:0; 1:1; 1:2; 1:3g for the results depicted in Figs. 7 and 8. As
C Pmax ¼ 0:95, showing that theoretically the maximum energy previously described, the conditions of zero loss in the diffuser
extracted by a DAWT is C Pmax ¼ c16=27 according to Eq. (8). For and V 4 ¼ ð1  2aÞV 0 are imposed, so c ¼ 1:0 represent Glauert’s

1.2 1.4
ndiff = 0.2
Power coefficient with diffuser, Cp

Betz-Joukowsky Limit with Diffuser = 0.4

Velocity at the rotor plane, u(r)/V0

1 = 1.6, Cpmax = 95% = 0.6

1.3
= 0.8
= 1.0
1.2
0.8
Cp opt
1.1
0.6
1
= 1.0
0.4 0.9 = 1.4
= 1.8
0.8
0.2
0.7
0
0 0.5 1 1.5 2 0.6
2 0 0.2 0.4 0.6 0.8 1
Thurst coefficient, CT2 = CT / Radial position, r/R
Fig. 5. Power coefficient as a function of the thrust coefficient C T2 . Fig. 6. Axial wind velocity at the rotor plane.
 J.R.P. Vaz, D.H. Wood / Energy Conversion and Management 123 (2016) 35–45 41

0.3 0.8
γ = 1.0 γ = 1.0
γ = 1.1 γ = 1.1
0.25
γ = 1.2 0.6 γ = 1.2
γ = 1.3 γ = 1.3

Twist angle, θ − (rad)

0.2
Chord, c/R

0.4
0.15
0.2
0.1

0
0.05

0 −0.2
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Radial position, r/R Radial position, r/R
(a) (b)
Fig. 7. (a) Chord and (b) twist angle distributions.

0.8 3.5

0.7 3
0.6 Betz-Joukowsky limit Output power, P − (kW)
Power coefficient, C P

2.5
0.5
2
0.4
1.5
0.3
1 γ = 1.0
0.2 = 1.0
= 1.1
γ = 1.1
0.1 = 1.2 0.5 γ = 1.2
= 1.3 γ = 1.3
0 0
5 5.5 6 6.5 7 7.5 8 5 6 7 8 9 10
Tip-speed ratio, Wind velocity, V − (m/s)
0

(a) (b)
Fig. 8. (a) Power coefficient as a function of tip-speed ratio and (b) output power as a function of wind velocity.

optimization for a bare turbine as shown in Eq. (32). Note that
Fig. 6 shows the to the diffuser increases the axial wind velocity
at the rotor plane. The same observation was made by Hansen
et al. [5].
The chord, c, and twist angle, h, distributions as a function of
radius are shown in Fig. 7 for gd ¼ 1:0, and V 4 ¼ ð1  2aÞV 0 , for
the reasons given in Rio Vaz et al. [2]. Based on the work of van
Bussel [20], they incorporated diffuser effects in BET in order to
obtain a theory closest to the momentum relations for bare wind
turbines. The diffuser speed-up ratio causes a significant increase
in c and a smaller one in h along the blade. These increases occur
because the increased axial velocity at the rotor plane changes
the angle of attack, flow angle, and the induced velocities. Similar
behavior occurs when Glauert’s optimization is used, keeping the
rotational speed constant and increasing the wind speed. This
allows an improvement in the turbine efficiency, as shown later.
Fig. 8 shows a comparison between the performance character-
istics of the rotors described in Fig. 7. The results were generated
for average c ¼ f1:0; 1:1; 1:2; 1:3g, and c ¼ 1:0 indicates a rotor
designed by Glauert’s optimization. The rotational speed was held
constant (240 rpm); only the wind velocity was altered (5–10 m/s),
leading to a variable tip-speed ratio. Clearly, the present analysis
allows a better optimized blade for all c > 1:0. For c ¼ 1:3 the
power coefficient can exceed the Betz–Joukowsky limit by a small
amount. These results were generated for an uniform c, whereas, in
practice, c can vary with radius as shown in the next subsection for
a simple flanged diffuser.
5.2. Application using a flanged diffuser
To demonstrate a realistic application of the present optimiza-
tion methodology, a DAWT using a flanged diffuser with a rotor
diameter of 4.0 m, hub diameter of 0.4 m, and 3 blades with a
NACA 0012 airfoil was considered. The rated V 0 ¼ 7 m/s and
X ¼ 240 rpm, with k  7. The diffuser Reynolds number (based
on the diffuser inlet radius) is 2:6  104 . To apply the proposed
methodology, it was first necessary to establish c. For that, a CFD
computation carried out as described by Rio Vaz et al. [2] using
ANSYS-Fluent software with the k  x SST (Shear Stress Transport)
turbulence model, validated against experimental data of Abe and
Ohya [7]. The standard deviation of the differences between the
experimental and simulated data was 0.02. The relative error at
each axial position did not exceed 4%. Fig. 9a shows the computa-
tional conditions and the diffuser geometry (D ¼ 4:05 m,
L ¼ 1:5D; h ¼ 0:125D) utilized, and Fig. 9b presents the results
obtained for c in the present case. The good agreement between
numerical and experimental results justify the use of c computed
by CFD.
The radial profile of the diffuser speed-up ratio is presented in
Fig. 10a, and the optimized blade geometry in Fig. 10b. The present
optimization used the average value of cðrÞ which was  1:3. The
comparison between the optimized rotor using the proposed
methodology and the classical Glauert optimization is shown in
Fig. 11. The performance analysis was carried out using the model
42 J.R.P. Vaz, D.H. Wood / Energy Conversion and Management 123 (2016) 35–45

2
Experimental (Abe & Ohaya [7])

Diffuser speed-up ratio, (r = 0)

1.8 Present work

1.6

1.4

1.2

0.8

1 2 3 4 5 6
Axial position, x/D
(a) (b)
Fig. 9. (a) Diffuser geometry [7]. (b) Validation of the simulation.

0.8
0.1
Radial position, r/R

z−(m)
0
0.6
−0.1
2
0.4
1.5

1
0.2
0.5
y−(m) 0 0.2
0 −0.2
0 0.5 1 1.5 2 x−(m)
Diffuser speed-up ratio, *(r/R)
(a) (b)
Fig. 10. (a) Diffuser speed-up ratio c ðr=RÞ. (b) Blade geometry.

1 3.5
Glauert's optimization
Present work
3
0.8
Output power, P − (kW)
Power coefficient, C P

2.5
Betz-Joukowsky limit
0.6
2
0.4
1.5

0.2
1
Glauert’s optimization
Present work
0 0.5
5 6 7 8 9 10 5 6 7 8 9 10
Tip-speed ratio, Wind velocity, V − (m/s)
0

(a) (b)
Fig. 11. (a) Power coefficient as a function of tip-speed ratio and (b) output power as a function of wind speed.

of Rio Vaz et al. [2]. Note that the power coefficient for the turbine 5.3. Comparison with experimental data
designed using the proposed optimization is 1:2 times greater than
the Glauert optimal at the rated kð¼ 7Þ. For k ¼ 6 the difference is The present optimization is now compared with the experiment
far greater, about 1.9 times. For the rated wind speed (7 m/s), the by Hoopen [30] using a DAWT with a three-bladed rotor of diam-
output power for the present work is about 26% higher than the eter of 1.5 m. The diffuser was a circular airfoil designed by
Glauert’s optimization, increasing to 100% at 8.3 m/s. However, National Aerospace Laboratory (NLR) [31], as shown in Fig. 12a,
for high tip-speed ratio (k > 7:7 and V 0 < 6:5 m/s), both optimiza- where the suction side is pointed inwards and the area ratio,
tions give the same performance. A3 =A, is 1.7286. The blades and diffuser were designed using CFD,
 J.R.P. Vaz, D.H. Wood / Energy Conversion and Management 123 (2016) 35–45 43

0.8

Radial position, r/R

0.6

0.4
= 1.08
0.2

0
0.8 0.9 1 1.1 1.2 1.3 1.4
Diffuser speed-up ratio, *

(a) (b)
Fig. 12. (a) DAWT equipped with vortex generators [30] and (b) diffuser speed-up ratio as a function of the radial position.

Table 1 [31]. The exit plane of the diffuser had a diameter of 2 m and was
Design parameters used in the simulation of the DAWT. equipped with a Gurney flap of 0.04 m. Hoopen [30], provide data
Parameters Values for a thrust coefficient of 0.8, wind speed of 10 m/s and X ¼ 75 rad/
Turbine diameter 1:5 m s (716.2 rpm). The diffuser Reynolds number was 6  105 . Delta-
Hub diameter 0:3 m shaped vortex generators were installed on the diffuser trailing
Number of blades 3 edge in order to promote mixing. The diffuser speed-up ratio is
Diffuser airfoil NLR [31] shown in Fig. 12b. The average c was 1.08. The DAWT geometry
Blade airfoil NACA 2207
and the experimental parameters are summarized in Table 1.
Power output 531 W
Freestream velocity ðV 0 Þ 10 m/s Fig. 13 shows c and h developed by NLR [31] and the blade
Air density ðqÞ at 20 °C 1.2 kg/m3 geometry optimized in this work. Fig. 13a depicts the optimized
Rotational speed 716.2 rpm blade shape designed using Algorithm 1 and the original blade

0.7
0.6
0.02
z-(m)

0 0.5
-0.02
0.4
-0.05 0.3
0
0.05 0.2
y-(m)
0.1
x-(m)
(a)

0.3 0.5
NLR blade [31] NLR blade [31]
Optimized blade Optimized blade
0.25
0.4
Twist angle, - (rad)

0.2
Chord, c/R

0.3
0.15
0.2
0.1

0.1
0.05

0 0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Radial position, r/R Radial position, r/R
(b) (c)
Fig. 13. (a) Optimized blade geometry, (b) chord, and (c) twist angle distributions as a function of the radial position.
44 J.R.P. Vaz, D.H. Wood / Energy Conversion and Management 123 (2016) 35–45

Table 2 Table 2 presents the output power and torque for the original
Comparison between the proposed optimization and experimental data rotor and the present optimization for V 0 ¼ 10 m=s. For the origi-
(V 0 ¼ 10 m=s).
nal rotor, Table 1 shows good agreement in the torque and power
Angular speed Output power Torque with a 6% error at the same X of 75 rad/s. The CFD results of Hoo-
(rad/s) (W) (N m) pen [30] used X ¼ 137 rad/s, which is much larger than the exper-
Experimental [30] 75 531 7.1 imental one. Fig. 14 shows the diffuser geometry designed and
Present work (original 75 564 7.5 optimized with CFD software using the k   turbulence model.
rotor)
Present work 75 601 8.0
The diffuser airfoil and details upon the code can be find in
(optimized) Refs. [30,31].
CFD [30] 137 545 4.0 A significant improvement in C P for the optimized rotor is
CFD [30] 155 246 1.6 shown in Fig. 15a. The increase in C P is 35% compared to the orig-
inal design, demonstrating the usefulness of the present optimiza-
tion procedure. Furthermore, the Betz–Joukowsky limit is
exceeded, and consequently the output power improved, as
depicted in Fig. 15b for V 0 exceeding 11.5 m/s. For low wind veloc-
ities, the output power is close to that of the original rotor because
the power coefficients for both cases are very close, 57.5% for the
original rotor and 60.7% for the optimized one. Furthermore, as
the power output is directly proportional to V 30 , at low wind veloc-
ity both power outputs are very close. The optimization depends
strongly on the diffuser speed-up ratio, which also depends on
the diffuser efficiency. Thus, to a complete the DAWT optimization,
it is necessary to determine the relationship between diffuser
geometry and the flow characteristics through the rotor. Another
important issue is that increasing the diffuser speed-up ratio can
promote stall on the rotor, because the augmented wind velocity
increases the effective angle of attack.

6. Conclusions

This paper presents an optimization procedure for the design of

Fig. 14. Diffuser geometry [31].
the blades of a diffuser-augmented wind turbine. It is based on
classical blade element theory, with the main additions being the
designed by NLR [31]. Note that the chord is increased over most of diffuser efficiency and velocity ratio which determines the increase
the blade, Fig. 13b, due to the increasing effect of the velocity at the in wind speed at the rotor plane due to the diffuser. The present
rotor plane, which is a consequence of pressure drop in the diffuser methodology extends Glauert’s optimization for diffuser-
increasing the mass flow through the rotor, as well as to the augmented wind turbines, DAWTs.
increased wake area due to the disturbed flow behind the turbine. The model has low computational cost and easy numerical
As the chord at each blade section is directly influenced by the flow implementation. It was demonstrated that the prediction of output
angle, /, the increasing velocity results in an increasing chord. This power agrees well with experimental data. Another important fea-
effect could be also reproduced using Glauert’s optimization for ture of the proposed methodology is that the diffuser velocity ratio
bare turbines by increasing the wind speed to increase the flow can be calculated using computational fluid dynamics, as per-
rate at the rotor plane. In other words, the blade geometry is very formed by Abe and Ohya [7] or through experiments, as done by
sensitive to the variation in V 1 . The large difference in c is not mir- Hoopen [30].
rored in h, Fig. 13c, except close to the hub where little power is The performance of the proposed optimization was demon-
produced. strated through comparison with DAWT experimental data of

0.65 1800

0.6 Betz-Joukowsky limit 1600

Output power, P - (W)

1400
Power coefficient, Cp

0.55
1200
0.5
1000
0.45
800
0.4 600

0.35 400 NLR rotor [31]

NLR rotor [31]
Optimized rotor
Optimized rotor
0.3 200
3.5 4 4.5 5 5.5 6 6.5 7 7.5 7 8 9 10 11 12 13 14 15
Tip-speed ratio, Freestream velocity, V0 - (m/s)

(a) (b)
Fig. 15. (a) Power coefficient as a function of the tip-speed ratio and (b) output power as a function of the freestream wind velocity.
 J.R.P. Vaz, D.H. Wood / Energy Conversion and Management 123 (2016) 35–45
Hoopen [30]. The present scheme suggests that the power output
could be increased by 35%.
It is necessary to consider some limitations of the present
model, such as the diffuser velocity ratio, being an input rather
than an outcome of the model. Besides, c must depend on the dif-
fuser efficiency, gd , and this relationship requires further investiga-
tion. The assumption for the velocity in the far-wake,
V 4 ¼ ð1  2aÞV 0 , also needs to be confirmed through experimental
tests, since V 4 , directly affects the power coefficient of a DAWT
through Eq. (1). Finally, the tip loss formulation for DAWTs must
be improved. Despite such limitations, the results obtained in this
work present physically consistent behavior, showing that the dif-
fuser produces a significant increase on the power coefficient, jus-
tifying the use of diffusers as a technology to improve wind turbine
power output, especially small ones.
Acknowledgments
The authors would like to thank the NSERC Industrial Research
Chairs program, the ENMAX Corporation, CNPq, CAPES, Eletronorte,
and PROPESP/UFPA for financial support.
References
[1] Al-Sulaiman FA, Yilbas BS. Thermoeconomic analysis of shrouded wind
turbines. Energy Convers Manage 2015;96:599–604.
[2] Rio Vaz DATD, Mesquita ALA, Vaz JRP, Blanco CJC, Pinho JT. An extension of the
blade element momentum method applied to diffuser augmented wind
turbines. Energy Convers Manage 2014;87:1116–23.
[3] Kosasih B, Hudin HS. Influence of inflow turbulence intensity on the
performance of bare and diffuser-augmented micro wind turbine model.
Renew Energy 2016;87:154–67.
[4] Jafari SAH, Kosasih B. Flow analysis of shrouded small wind turbine with a
simple frustum diffuser with computational fluid dynamics simulations. J
Wind Eng Ind Aerodynam 2014;125:102–10.
[5] Hansen MOL, Sorensen NN, Flay RGJ. Effect of placing a diffuser around a wind
turbine. Wind Energy 2000;3:207–13.
[6] Wang WX, Matsubara T, Hu J, Odahara S, Nagai T, Karasutani T, et al.
Experimental investigation into the influence of the flanged diffuser on the
dynamic behavior of CFRP blade of a shrouded wind turbine. Renew Energy
2015;78:386–97.
[7] Abe K, Ohya Y. An investigation of flow fields around flanged diffusers using
CFD. J Wind Eng Ind Aerodynam 2004;92:315–30.
[8] Ohya Y, Karasudani T. A shrouded wind turbine generating high output power
with wind-lens technology. Energies 2010;3:634–49.
[9] Lubitz WD, Shomer A. Wind loads and efficiency of a diffuser augmented wind
turbine. In: Proc. CSME Intl Congress, Toronto.
[10] Rio Vaz DATD, Vaz JRP, Mesquita ALA, Pinho JT, Brasil ACP. Optimum
aerodynamic design for wind turbine blade with a rankine vortex wake.
Renew Energy 2013;55:296–304.
45
[11] Glauert H. Aerodynamic theory. In: Durand WF, editor. Division L. Airplanes
Propellers, vol. 4. p. 191–5 [reprinted, Dover, NewYork, 1963, Chapter XI].
[12] Fletcher CAJ. Computational analysis of diffuser-augmented wind turbines.
Energy Convers Manage 1981;21:175–83.
[13] Igra O. Research and development for shrouded wind turbines. Energy Convers
Manage 1981;21:13–48.
[14] Okulov V, Sørensen JN. Refined Betz limit for rotors with a finite number of
blades. Wind Energy 2008;11:415–26.
[15] Wood DH. Maximum wind turbine performance at low tip speed ratio.
Submitted J Renew Sustain Energy 2015;7:053126. http://dx.doi.org/10.1063/
1.4934805.
[16] Bontempo R, Manna M. Performance analysis of open and ducted wind
turbines. Appl Energy 2014;136:405–16.
[17] Nobile R, Vahdati M, Barlow JF, Mewburn-Crook A. Unsteady flow simulation
of a vertical axis augmented wind turbine: a two-dimensional study. J Wind
Eng Ind Aerodynam 2014;125:168–79.
[18] Sayed MA, Kandil HA, Shaltot A. Aerodynamic analysis of different wind-
turbine-blade profiles using finite-volume method. Energy Convers Manage
2012;64:541–50.
[19] Benini E, Toffolo A. Optimal design of horizontal-axis wind turbines using
blade-element theory and evolutionary computation. ASME J Solar Energy Eng
2002;124:357e63.
[20] van Bussel GJW. An assessment of the performance of diffuser augmented
wind turbines (DAWT’s). In: 3th ASME/JSME joint fluids engineering
conference, July 18–23, San Francisco, California, USA.
[21] Philipis DG. An investigation on diffuser augmented wind turbine design PhD
thesis. Department of Mechanical Engineering, School of Engineering, The
University of Auckland; 2003.
[22] Dick E. Momentum analysis of wind energy concentrator systems. Energy
Convers Manage 1984;24:19–25.
[23] Hansen M. Aerodynamics of wind turbines. 2nd ed. Earthscan; 2008.
[24] Clifton-Smith MJ. Wind turbine blade optimisation with tip loss corrections.
Wind Eng 2009;33(5):477–96.
[25] Wilson RE, Lissaman PBS. Applied aerodynamics of wind power
machines. Corvallis, Oregon: Oregon State University; 1974.
[26] Wald QR. The aerodynamics of propellers. Progress Aerospace Sci
2006;42:85–128.
[27] Branlard E. Wind turbine tip-loss corrections: review, implementation and
investigation of new models, Master’s thesis. Master of Science in Wind Energy
at the Technical University of Denmark; 2011 (Post-print 2013).
[28] Barbosa DLM, Vaz JRP, Figueiredo SWO, Silva MO, Lins EF, Mesquita ALA. An
investigation of a mathematical model for the internal velocity profile of
conical diffusers applied to DAWTs. Ann Braz Acad Sci 2015;87(2). Available
from: <http://dx.doi.org/10.1590/0001-3765201520140114>.
[29] Sheldahl R, Klimas P. Aerodynamic characteristics of seven symmetrical airfoil
sections through 180-degree angle of attack for use in aerodynamic analysis of
vertical axis wind turbines. Sandia National Laboratories, Report SAND80-
2114; 1981.
[30] Hoopen PDC M.Sc. thesis. An experimental and computational investigation of
a diffuser augmented wind turbine: with an application of vortex generators
on the diffuser trailing edge. Faculty of Aerospace Engineering, Delft University
of Technology; 2009.
[31] National Aerospace Laboratory (NLR). Evaluatie en verbetering van de
prestaties van eenkleinschalige diffuser augmented wind turbine (DAWT);
1e fase; 2008.