You are on page 1of 15

Energy 166 (2019) 819e833

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Experimental investigation of the performance and wake effect of a


small-scale wind turbine in a wind tunnel
Bingzheng Dou a, b, Michele Guala b, Liping Lei a, *, Pan Zeng a
a
Department of Mechanical Engineering, Tsinghua University, Beijing 100084, China
b
St. Anthony Falls Laboratory, Department of Civil, Environmental, & Geo-Engineering, University of Minnesota, Minneapolis, MN 55414, USA

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

Article history: The wake of upstream wind turbines is known to affect the operation of downstream turbines and the
Received 21 May 2018 efficiency of the wind farm. In this study, a systematic experimentation on performance and wake spatial
Received in revised form evolution was carried out using a wind turbine model varying tip speed ratio, pitch and yaw angles. The
20 September 2018
change of pitch angle was observed to induce a greater effect on the wake velocity as compared to the tip
Accepted 16 October 2018
Available online 18 October 2018
speed ratio. This is interpreted in terms of “force viewpoint”, which describes more quantitatively the
relationship between the turbine performance and the wake, as compared to the “power viewpoint”,
based on the sole energy conversion. The turbine yaw angle is observed to cause not only a decrease in
Keywords:
Wind turbine
power and thrust, but also an offset and an asymmetry in the wake. The offset, quantified using the
Wind tunnel spatial distribution of the velocity minima, is modeled analytically. Comparisons of model estimations
S826 airfoil with the experimental measurements show that the proposed model can acceptably predict the wake
Wake effect offset of a yawed turbine. The observed dependencies of the mean velocity deficit and wake turbulence
Turbulence intensity on power, thrust, and yaw angle, may suggest new derating strategies for wind farm optimization.
© 2018 Elsevier Ltd. All rights reserved.

1. Introduction TI in the wake depends on the upstream turbine performance [4]


and thus on its tip speed ratio (which is defined as the ratio be-
In recent years, the global cumulative install wind capacity tween the speed of the tip blade and the velocity of the incoming
increased constantly [1]. Wind energy has constituted around 30% flow), pitch angle, and yaw angle [7].
of the renewable capacity [2]. Maximizing the efficiency of wind The wake of a wind turbine can be divided into two regions
power generation at the power plant scale has therefore gained which are the near wake and the far wake [8]. The near wake, still
more attention and resources. Therefore, for siting and controlling governed by the turbine geometry [9], the tip vortex, and the hub
wind turbines in the wind farm, understanding the spatial evolu- vortex structure, typically ends in the range between 2D (D is the
tion of the wake and the functional dependencies involved, are of diameter of the turbine) and 4D [10]. The large-scale turbulence is
great importance. Under most operating conditions, in the turbine generated in the annular shear layer of the wake and spreads into
wake, the mean downstream wind velocity reduces [3] and the the core of the wake, reaching the centerline of the wake between
Turbulence Intensity (TI) increases, affecting the unsteady loads on 3D to 5D [10]. In most modern wind farms, the turbines are sepa-
downwind turbines. The deficit in the power of the downstream rated by at least 5D to 7D [8] along the dominant wind direction.
turbines in the wake varies from about 20% to 46% compared with Therefore, a turbine in a wind farm is likely located in the far wake
the power output of an unobstructed single turbine operating [4,5]. of the upstream turbine, which makes far wake evolution critical
Thomsen et al. [6] proposed that an increase of fatigue loading of for wind farm optimization. Then, Jensen wake model [11], Larsen
wind turbine blade in a wind farm was found to be between 5% and wake model [12], Frandsen wake model [13,14], and Bastankhah
15% compared to an unobstructed turbine on flat terrain, and the wake model [13] which are mostly related to the thrust force of the
turbulence intensity was one of the most important load- turbine, are proposed as the semi-empirically wake models to
generating parameters. The variability of the velocity deficit and describe the flow in the wake. In addition, a lot of effort has been
put into studying the features of wake and the turbine performance
via the use of developed Computational Fluid Dynamics method
* Corresponding author. -Agel et al. [17,18], Stevens
(e.g. Sørensen et al. [15,16], Wu and Porte
E-mail address: leilp@tsinghua.edu.cn (L. Lei).

https://doi.org/10.1016/j.energy.2018.10.103
0360-5442/© 2018 Elsevier Ltd. All rights reserved.
820 B. Dou et al. / Energy 166 (2019) 819e833

Nomenclature Fdf force normal to the chord [N]


b pitch angle [ ]
g yaw angle [ ]
Variables drt distance between the rotor center and the tower
TI turbulence intensity center [m]
D turbine rotor diameter [m] s projection of drt on the y-coordinate [m]
s velocity standard deviation [m/s] Yoffset spanwise location of the minimum wake velocity [m]
U∞ free stream velocity [m/s] d parameter of the offset model
R rotor radius [m] z parameter of the offset model
c chord length [m]
r radius position [m] Repeated experiments
l tip speed ratio sCp(l) standard deviation of Cp at l
Cp power coefficient sCT(l) standard deviations of CT at l
P generated power [W] Cp ðlÞ overall average power coefficient at l
r air density [kg/m3]
CT ðlÞ average thrust coefficient at l
A rotor swept area [m2]
RSD relative standard deviation
CT thrust coefficient
m number of downstream spanwise profiles
T thrust force [N]
sU(x,y) wind velocity standard deviation at (x, y) [m/s]
a angle of attack [ ]
sTI(x,y) turbulence intensities standard deviation at (x, y)
Ft tangential force [N]
N number of independent time series
Fn normal force [N]
Ui(x,y) mean velocity of experiment i at (x, y) [m/s]
Ul local velocity on blade element [m/s]
TIi(x,y) turbulence intensity of experiment i at (x, y)
q twist angle [ ]
Uðx; yÞ averaged velocity of repeated experiments [m/s]
Fl lift force [N]
Fd drag force [N] TIðx; yÞ averaged TI of repeated experiments

et al. [19], and Yang et al. [20,21]). configurations of the turbine to conduct a wake effect experiment
In order to explore wake evolution under monitored or for presenting the wind velocity profile in wake under controlled
controlled boundary conditions and provide a reference for simu- variation of inflow turbulence intensity [38]. Xie et al. [39]
lation, many researchers conducted wind tunnel or field turbine measured the power performance of an innovative pitched wind
experiments. Howard et al. conducted wake experiments by Par- turbine at various tip speed ratio in a wind tunnel. In addition, some
ticle Image Velocimetry (PIV) to compare the wakes produced by a previous studies are about the performance and the near wake in
single turbine operating at two different tip speed ratios [22], and various tip speed ratios and yaw angles [40,41], or reported the
to characterize the response of a wind turbine model under com- asymmetric distribution of the wake skew angle with respect to the
plex inflow conditions including roughness and thermal stability wake center and then proposed a simple analytical models to
effects [23]. Wang et al. [24] introduced a novel dual-rotor wind predict the wake deflection [42,43].
turbine and measured its performance and wake flow character- To provide a new design of turbine blades and airfoil profiles, it
istics using a PIV system. Cal et al. [25] investigated the wake is important to test a scaled rotor model under controlled wind
interaction of wind turbine models in turbine array by PIV. turbine operating conditions. The major goals are to quantify both
Different from the PIV in a wind tunnel, Hong et al. [26,27] used the the turbine performance and the wake spatial evolution, in
snow-PIV to quantify the evolution of blade generated coherent particular for varying tip speed ratio, pitch angle, and yaw angle. A
motions in near wake, such as the tip and trailing sheet vortices, few laboratory experiments investigated the relation between
identify their instability mechanisms and correlate them with turbine performance and the far wake evolution, at not only various
turbine operation, control and performance. Also measuring the tip speed ratios but also pitch angles and yaw angles. And little is
wake effect produced by a utility-scale wind turbine, Iungo [28] known on the effect of yaw angle on the distortion of the wake at
conducted some experiments with the Doppler Light Detection the same time. In this study, we provide high frequency measure-
And Ranging. Due to the PIV shooting range limitations and the ments of the streamwise velocity and turbulence intensity in the
uncontrollable natural conditions, Schümann et al. [29], Howard wake of a new turbine model, together with angular velocity and
et al. [30], Hamilton et al. [31], Iungo et al. [32], and Singh et al. [33] generated power measurements. We first vary the tip speed ratio
carried out wake experiments in an atmospheric boundary layer and pitch angle to determine the optimal performance. Then, under
wind tunnel using a hot-wire anemometer. such operating conditions, we explore the spatial evolution of the
For further studying the wake effect in different turbine oper- wake under prescribed yaw misalignment and we derive a simpler
ating configurations, researching the relation between the wake model able to predict the location of the maximum deficit (or
characteristics and the turbine performance, Howard et al. [22], minimum wake velocity) in the wake.
Whale et al. [34], and Hu et al. [35] compared the wake of a turbine In section 2, a description of the experimental system including
operating at different tip speed ratios. Medici et al. [36] reported wind tunnel, the experimental setup and wind turbine model is
the wake meandering and the turbine performance in different tip given. The results and discussion are detailed in section 3, including
speed ratios and pitch angles. Li et al. [37] analyzed the effect of the turbine power and thrust performance, wind velocity and tur-
turbulence intensity on the power and thrust characteristics of a bulence intensity in the wake. A simple offset model of the
0.5 m diameter wind turbine with different tip speed ratios, pitch maximum deficit in yawed condition is also presented in section 3.
and yaw angles in a wind tunnel. Then they selected an optimum The last section contains the discussion and conclusion.
B. Dou et al. / Energy 166 (2019) 819e833 821

2. Experiment setup pitch angle of the blades in Fig. 3 is 0 . To explore the wake evo-
lution and the turbine performance under pitched condition,
2.1. Wind tunnel and the measurement device various blades with different pitch angles were manufactured and
tested.
This experiment was conducted in a low-speed and negative- The turbine rotor consists of two blades, a nose cone, and a hub.
pressure wind tunnel (LSS wind tunnel, shown in Fig. 1, also can The hub was divided into two pieces in order to clamp and fix the
be seen in Ref. [44]) which was designed and constructed by the two blades. The assembly is shown in Fig. 4. The rotational inertia of
Laboratory for Super-lightweight Structure in Department of Me- the rotor is about 4.076  105 kg m2, and the flexural modulus of
chanical Engineering at Tsinghua University. The test section is the blade is about 2200 MPa.
2.2 m (length)  1.5 m (width)  1.5 m (height). The experimental The driving system includes a turbine rotor, DC motor, encoder,
wind flows across a honeycomb net and a rectifier net in order the and a tower. The hub was set at a height of 0.6 m from the bottom of
experimental condition is in a steady turbulence intensity. The free the wind tunnel. The encoder has a resolution of 500 steps per
stream wind velocity is 6 m/s in this experiment. In addition, the TI revolution, and its maximum radius is 16 mm. A six-component
I ¼ s=U∞ , here defined as the standard deviation of the local wind force sensor installed on the tower measures the thrust of the
velocity (s) divided by the free stream wind velocity (U∞) [45,46], is turbine.
about 1%.
A movable platform (including three sliding rails and a foun- 3. Results and discussion
dation support) was installed in the wind tunnel. The wind velocity
downstream in the turbine wake was measured by a Dantec hot- In this investigation, experimental data were obtained in
wire anemometer with a single wire probe. The probe positioning different operating conditions, such as different tip speed ratios,
system was mounted on the foundation support and computer- pitch angles, and yaw angles. We present the wake statistics in
controlled to move automatically on the horizontal measurement terms of the mean wind velocity and the TI. First, the characteristics
plane. The sampling frequency of this probe is 80000 Hz, and the of the wake under different tip speed ratios are investigated, with
samples number is set to 262,144. The environmental temperature that both the pitch angle and yaw angle are set to 0 .
is 22.5  C, which is controlled by two air conditioners.
The probe measures the wind velocities ranges from 3.5D to
3.1. Effect of tip speed ratio
8.5D downstream. Based on previous literature results [25,38,47],
the lateral extent of the measurement domain started at 3.6R
3.1.1. Turbine performance at various tip speed ratios
with an increment of 0.2Re2.6R. The full layout of the measure-
Fig. 5 shows the power coefficient of the turbine when the
ment points is shown in Fig. 2 (a) where R is the radius of the rotor.
turbine model is driven by the wind as a generator. Its tip speed
The foundation on the movable platform was able to sequentially
ratio l can ranges between 3.49 and 6.11. The maximum of power
move to the prescribe measuring points and collect high frequency
coefficient (Cp ¼ 0.342) is estimated as l ¼ 3.84.
velocity time series. In order to explore the wake evolution under
The power coefficient is defined as:
yawed condition, the turbine rotor orientation was changed and
measured systematically with 8 counterclockwise step in the P
horizontal plane (as shown in Fig. 2 (b)). Cp ¼ (1)
0:5rAU 3∞

2.2. Design of the blade and wind rotor where P is the generated power, r is air density, A is the swept area
of the rotor.
The diameter of wind turbine rotor is 0.2 m. The rotor was The thrust coefficient is defined as:
designed with two blades using S826 profile throughout the blade
span and was manufactured by a 3D printer using the DSM Somos T
CT ¼ (2)
lmagine 8000 resin. The experimental and numerical data of S826 0:5rAU 2∞
airfoil can be seen in Refs. [48e50]. In order to make the Reynolds
number and the attack angle of every blade elements to be equal, where T is the thrust force exerted on turbine model, directly
the Schmitz theory [51,52] was employed to design the blade chord measured by the load cell.
length and twist angle distributions which are shown in Fig. 3. The In this experiment, the rotor was controlled by a servo motor

Fig. 1. The position of wind turbine model and the measurement device in wind tunnel.
822 B. Dou et al. / Energy 166 (2019) 819e833

Fig. 2. Layout of the measurement points (a), image of the yawed turbine (b), force sensor (c) and rotatable platform used to hold the support tower(d).

Between l ¼ 3.5 and l ¼ 6, the difference of thrust coefficient is


small which means that the drag induced by the turbine used as a
generator (fixed torque) and a motor (fixed angular velocity) is
minimally different, and so the little difference must be in wake.
Therefore, in the following experiment, the power is measured in
the passive driving mode. And for controlling the rotational speed
of the rotor accurately, the thrust is measured in the active driving
mode. The relative standard deviation of the power and thrust
measurements were calculated to be below 6% (see Fig. A.1 in
Appendices).

3.1.2. Characteristics of the wake effect


Fig. 7 shows the contours of the mean wind velocity and TI in the
Fig. 3. Chord-radius ratio and twist-angle distribution of the wind turbine model
turbine wake for varying tip speed ratios. The relative standard
blades against the radius position r.
deviations of the mean wind velocity and turbulence intensity, in
the turbine wake, were estimated to be below 2% and 8%, respec-
(defined here as active driving mode) to measure the thrust under a tively (see Fig. A.2 in Appendices). The turbine extracts kinetic
given incoming wind velocity, as performed by Ref. [29]. Under the energy from the wind, through the lift and the torque generated
same flow conditions but in a slightly different configuration, the along the blades, while the rotor induces an additional drag force,
rotational speed was controlled by the manually adjusted electric which results in the mean flow decreasing in the wake, with a
load (defined as passive driving mode) to measure the power, as maximum deficit immediately downwind of the rotor plane.
described in Ref. [53]. Fig. 6 shows the difference in the thrust co- Fig. 8 represents the directions of the velocity and of the
efficient (CT) between the active and passive driving modes. resultant force on the airfoil of a generic blade element. With an
increase in tip speed ratio, both the angle of attack a and tangential

Fig. 4. Photograph of the blade rotor model (a) and the size of the wind turbine where the unit is mm (b).
B. Dou et al. / Energy 166 (2019) 819e833 823

value as reported by Refs. [4,55], consistent with a reduced deficit.


This law of determining wind velocity deficit can be called “power
viewpoint” which means that the velocity deficit mainly depends
on the turbine power extracted from the wind. Based on the above
analysis, Fdf mainly contributes to TI in the wake; the pressure force
is decomposed as Fdf ¼ Fn cos q þ Ft sin q, which is approximately
equal to the normal force Fn when the twist angle q is small.
Therefore, for increasing Fn on the airfoil section and resulting in a
larger thrust force on the entire rotor, Fdf will increase leading to
larger TI in the wake. This implies that the TI can be directly related
to the thrust coefficient.

3.2. Effect of pitch angle

The wake characteristics for different pitch angle b [ 4 , 2 ,


0 , 2 , 4 ] are explored now. In this experiment, l is set to 4, 5, and
Fig. 5. Performance curve: power coefficient as a function of the tip speed ratio. 6, while the yaw angle remain invariant 0 . The turbine rotor is
modified by imposing a different pitch angles to the two blades, as
shown in Fig. 9.

3.2.1. Turbine performance at various pitch angles


The power coefficient and thrust coefficients are reported in
Fig. 10. The maximum power coefficient is 0.335 at l ¼ 4 and b ¼ 0 ,
while the thrust coefficient CT increases with increasing tip speed
ratio, and decreases with increasing pitch angle.

3.2.2. Characteristics of the wake effect


The wake velocity field at different pitch angles are shown in
Fig. 11, Figs. A.3 and A.4 (in Appendices). The large pitch angle can
cause low wind velocity deficit and weak TI in the wake. This
feature is the result of the change in Fl and Fd of the airfoil. The
change of pitch angle has a greater effect on the wind velocity
(Fig. 12 and Fig. 13) as compared to the tip speed ratio, in both near
and far wake (consistent with Fig. 11(a)). For example, in Fig. 10(a),
when b ¼ 4 , Cp (l ¼ 6) is lower than Cp (l ¼ 4); Accordingly, in
Fig. 6. Thrust coefficient of the turbine model in passive drive and active drive method Figs. 12 and 13, the wake velocity at l ¼ 6 is larger than the one at
respectively. l ¼ 4. When changing the pitch angle, Cp (b ¼ 0 ) exhibits the
highest value, as compared to other pitch angles for the same tip
speed ratio (Fig. 10(a)). The corresponding wake velocity, based on
force (in y-coordinate) Ft are reduced, while the normal force (in x- the “power viewpoint” should be the smallest, however, in Fig. 12
coordinate) Fn grows. As a result, as the tip speed ratio exceeds the (x/D  5) and Fig. 13, we recognize this is not the case (the mini-
optimum value, the power coefficient decreases and the thrust mum velocity is observed for b ¼ 2 and 4 ). It means Cp is not
coefficient increases. the only factor determining the wind velocity deficit in the wake.
For operating conditions defined by l below the optimum tip The reason for the above mechanism is that, CT incorporates the
speed ratio, the angle of attack is comparatively reduced and both drag effect in x-coordinate, while Cp incorporates the drag effect in
Cp and CT are smaller than the peak along with weak wake effects. the tangential direction of the blade motion in the rotor plane,
Actually, the wake velocity deficit is controlled by both power resulting in the velocity deficit dependency on both Cp and CT. This
conversion and rotor drag, depending on the specific airfoil aero- interpretation can be called “force viewpoint” that is different from
dynamics and thus on both l and the pitch angle b (explored in the aforementioned “power viewpoint”.
section 3.2). It is should be noted that the pressure difference be- With increasing pitch angle, the dual turbulence intensity peaks
tween the two sides of the airfoil element of the blade results in the associated with the tip vortex are more evident, while the near-far-
force normal to the chord Fdf, (Fig. 8) contributing to aerodynamic wake merging region translates downstream (consistent with
lift and circulation. The circulation is (ideally) conserved from the derating operation past optimal conditions). It should be noted
blade to the tip vortices and thus results in the TI in the wake. The that, at x/D ¼ 3.5 (Figs. 12 and 13), there is a significant difference
top tip and bottom tip turbulence intensity peaks (Fig. 7(b)) gov- among the wind velocities under varying pitch angles, but this
erned by the strength of the tip vortexes gradually merged into a difference is not so evident at x/D ¼ 8.5 because of the spread of the
single peak along the downstream coordinate, consistent with turbulence peaks, which means the far wake will be slightly
Ref. [34], marking the transition between the near and the far wake. affected by the pitch angle. In the near wake the effect is obvious,
The position of the merging location, highlighting potential in- because the angle of attack a and the lift force Fl of the blade
teractions between the hub and the tip vortices, depends on the element become smaller but the drag force Fd changes little (the
magnitude of the TI in the wake: it is at x/D ¼ 6.5 for l ¼ 4, 5 experimental data of coefficient of lift force and drag force for S826
increasing to x/D ¼ 7.5 for l ¼ 6. Derated operating conditions airfoil can be found in Ref. [56]) with increasing pitch angle.
appear to delay the transition to wake meandering, as suggested by Then the Fl will play a major role in contributing to the pressure
Foti et al. [54]. force Fdf normal to the chord when the attack angle is small. As
As for the wake velocity, it is higher when l is not at its optimum mentioned in Section 3.1, Fdf can determine the magnitude of the TI.
824 B. Dou et al. / Energy 166 (2019) 819e833

Fig. 7. Contours of the wind velocity (a) and the TI (b) distribution against the different tip speed ratios. In order to place these contours better, the y-ordinate has been compressed.

3.3. Effect of yaw angle

In addition to the above investigation, the effects of different


yaw angles on the wake were studied. The tip speed ratio l is set at
values 4, 5, and 6. The yaw angle g is changed from 0 to 32 , for a
total of 15 operating conditions. The position of yawed turbine is
shown in Fig. 14 with g ¼ 16 , as an example.
In Fig. 14, the distance between the rotational center of the
tower and the hub center in the y-direction is s ¼ drt sin g where
drt ¼ 60 mm is the linear distance between the center of the rotor
and the center of the tower.

3.3.1. Turbine performance at various yaw angles


The power coefficient and thrust coefficients are reported in
Fig. 15. With increasing in yaw angle, both the Cp and CT experience
significant decreases.
Fig. 8. Force component acting on the airfoil element. Ul is the wind velocity in the
rotor plane, Ur ¼ lU∞ r=R, a is the attack angle, q is the twist angle of this airfoil
element. Fl is the lift force, Fd is the drag force (with respect to the airfoil geometry), Fn
3.3.2. Characteristics of the wake effect
is the normal force, Ft is the tangential force (with respect to the turbine rotor plane), The wind velocity and the TI distribution are shown in Fig. 16,
Fdf is the force normal to the chord. Figs. A.5, and A.6. With increasing the yaw angle, the wake region
becomes progressively distorted and asymmetric, while both the
wind velocity deficit and TI weaken. The wake asymmetry, with
respect to the hub vertical plane, characterize both the mean ve-
locity and the TI, resulting in asymmetrical effects on the down-
stream turbines in a wind farm.
In yawed operating conditions, there is not only a drag force on
the wind in the x-direction (contributing to CT), but also a thrust
Fig. 9. Diagram of the pitch angle.
component in the skew direction of the rotor plane, which acts on
the wind that pass through the rotor and distorts the wake. This
skew will make the downstream turbines to withstand a non-
The TI at b ¼ 4 is larger than at b ¼ 4 , and a similar trend is noted uniform wind velocity and turbulence levels (dynamic wake
for the CT, suggesting again that the magnitude of TI can be meandering effect was introduced by Larsen et al. [57]). Besides, the
approximately estimated as a function of the thrust (“force view- tip speed ratio has a significant effect on the mean velocity deficit
point”), even under different pitch configurations. (Fig. 17) in the near wake of the yawed turbine, with decreasing
influence in the far wake region.
B. Dou et al. / Energy 166 (2019) 819e833 825

Fig. 10. Power coefficient (a) and thrust coefficient (b) against pitch angle at different tip speed ratio.

Fig. 11. Contours of the wind velocity (a) and the TI (b) distribution with different pitch angles at l ¼ 4.

3.3.3. Wake structure with a yaw angle The scatter plot in Fig. 19 shows the positions of the maximum
The yaw variation results in the change of the wake direction, velocity deficit (or minimum velocity, the acquisition method is
the shape, and the offset of the minimum velocity cross stream described in Appendices, Fig. A.7 and eq. (A.3)) at different down-
location (see Fig. 18, where Yoffset is the distance between the po- stream distances and yaw angles. As the yaw angle increases, the
sition of the maximum velocity deficit and the y/R ¼ 0 line). position of maximum wind velocity deficit skews to the yaw side
826 B. Dou et al. / Energy 166 (2019) 819e833

Fig. 12. Wind velocity with different pitch angle at y/R ¼ 0 at l ¼ 4 (a), l ¼ 5 (b) and l ¼ 6 (c) respectively.

Fig. 13. Wind velocity with different pitch angle at y/R ¼ ±1 at l ¼ 4(a), l ¼ 5(b) and l ¼ 6(c) respectively.

Fig. 14. Diagram of wind turbine yaw and the wind velocity distribution downstream.

and expands along the downstream direction. We note that the between the wake and coflowing jets, the potential core region
trend of wake distortion is gradually slowing down, suggesting that (where the skew trend is linear) of this model is more than 8D for
the spatial evolution of the maximum velocity deficit locations can our case which is too long. Besides, the Bastankhah skew model
be approximately by a power law. may underestimate the offset for our case. The skew angle
Describing the change in the position of the maximum velocity magnitude predicted by Bastankhah skew model happened to be
deficit will be of great help to predict the downstream wake. The approximately half as large as that obtained in the simulation of
Jimenez skew model [43], considerably overestimates the wake Shapiro et al. [58]. In our case, by comparing the maximum wind
trajectory for all cases and specificity highlights the need for more velocity deficit at different tip speed ratios, the offset is observed to
realistic and robust modelling of turbine wakes in yawed conditions grow with increasing CT, even at the same yaw angle. This means
[42]. The Bastankhah skew model [42] is based on the analogy that Yoffset is not only related to the yaw angle but also to the thrust
B. Dou et al. / Energy 166 (2019) 819e833 827

Fig. 15. Power coefficient (a) and thrust coefficient (b) against yaw angle at different tip speed ratio.

Fig. 16. Contours of the wind velocity (a) and the TI (b) distribution with different yaw angles at l ¼ 4.

coefficient. As mentioned earlier, “force viewpoint”, both Cp and CT The general offset model is expressed as:
contribute to define the velocity deficit in the wake. In the yaw rffiffiffiffi
process, we argue that both Cp and CT determine the wake deficit, Yoffset x drt
¼ dðCT sin gÞz þ sin g (3)
but only CT contribute to the offset position. D D D
Based on the characteristics of Yoffset, eq. (3) is proposed, where:
CT sing was adopted to represent the effect of the wind thrust in y- where, d ¼ 0.54, z ¼ 0.75 and CT is estimated as a function of both g
coordinate; d and z are fitting parameter depending on the geo- and l.
metric characteristics of the wind turbine and the incoming wind Note that the values of the fitting parameters provided here
conditions. should be confirmed by further experiments conducted with tur-
bulent boundary layer inflow conditions. We acknowledge that in
828 B. Dou et al. / Energy 166 (2019) 819e833

Fig. 17. Maximum of wind velocity deficit at different yaw angles at l ¼ 4 (a), l ¼ 5 (b), and l ¼ 6 (c).

model. Otherwise, the simplified equation will slightly underesti-


mate the offset position.
rffiffiffiffi
Yoffset x drt
¼ d½CT ðg ¼ 0 ; lÞsin g cos22 g
z
þ sin g (4)
D D D
It is important to note that the two parameters remain invariant
in the two models.

4. Discussion and conclusion

In this experimental investigation, the tip speed ratio, pitch


angle, and yaw angle of a wind turbine model, exposed to a uniform
Fig. 18. Wake velocity distribution at x/D ¼ 5 and g ¼ 16 . flow, are varied in different combinations to assess the turbine
performance, and to characterize the wake field in terms of mean
velocity deficit and TI, in the range of x/D ¼ 3.5e8.5. The turbine
the near wake, there is a slight overestimation of the offset. performance is evaluated in terms of power and thrust coefficients,
In the absence of thrust coefficient estimated in yawed oper- both estimated through direct measurements of electrical power
ating conditions, a simplified offset model is proposed. Adaramola and drag force, respectively. The turbine reaches its maximum
and Krogstad [4] proposed that the power output closely follows power coefficient Cp ¼ 0.342 at l ¼ 3.84.
the theoretical curve Cp ðgÞ ¼ Cp ðg ¼ 0 Þcos3 ðgÞ. Consistently, we By varying tip speed ratio, the largest wind velocity deficit
assume the thrust force follows CT ðgÞ ¼ CT ðg ¼ 0 Þcos2 ðgÞ, which occurred at the optimum value (Fig. 7), corresponding to maximum
is found in agreement with the experimental data. The simplified power coefficient (Fig. 5). This finding is consistent with the studies
model can be thus expressed as eq. (4), and the fitting results are of Adaramola and Krogstad [4] and Steinbuch et al. [55], and in this
shown as the dotted lines in Fig. 20. When g < 30 , the fitting result work, has been referred to as “power viewpoint” (see section 3.1.2).
of the simplified model agrees well with the proposed general However, with increasing pitch angle, it should be noted that

Fig. 19. The general offset model and the position of the greatest wind velocity deficit at different yaw angle, (a) is at l ¼ 4, (b) is at l ¼ 5, (c) is at l ¼ 6. The solid line represents the
proposed general offset model eq. (3).
B. Dou et al. / Energy 166 (2019) 819e833 829

Fig. 20. The simplified offset model and the position of the greatest wind velocity deficit at different yaw angle, (a) is at l ¼ 4, (b) is at l ¼ 5, (c) is at l ¼ 6. The dotted line represents
the simplified offset model eq. (4).

although Cp ¼ 0.335 is maximum for l ¼ 4 and b ¼ 0 (Fig. 10(a)), variations can be implemented in a full control strategy devoted to
the wind velocity deficit is not the largest (Figs. 11(a)e13). In other obtain the maximum total power at the farm scale.
words, the largest wind velocity deficit did not occur at the
maximum power coefficient conditions (under varying both l and
b). This observation evidently does not agree with the “power
viewpoint”. In fact, CT, under maximal power coefficient operating Acknowledgements
condition, is lower as compared to e.g. b ¼ 4 or b ¼ 2 operating
condition (Fig. 10(b)), which means that the turbine exerts a weaker The work was supported by National Natural Science Founda-
drag force on the wind. Therefore, the wind velocity in the wake of tion of China (No. 51575296). Besides, the support provided by
a turbine cannot be predicted only as a function of the power co- China Scholarship Council (201706210200) during a visit of
efficient, e.g. simply according to the “power viewpoint”. The po- Bingzheng Dou to University of Minnesota is acknowledged. Zhe
wer coefficient and the thrust coefficient act together contributing Ma and Zhanpei Yang also gave some valuable suggestions.
to the wind velocity deficit. This approach, which we define as a
“force viewpoint” (see section 3.2.2), is also different from currently
existing wake models (such as Jensen wake model [11,59] and
Appendices
Larsen wake model [12]) that are only based on the thrust coeffi-
cient and wake growth rate to estimate the mean wind velocity
deficit. The importance of the thrust coefficient clearly emerges
when comparing TI in the turbine wake model under various
operating conditions. TI can be approximately expressed as a
function of only the thrust coefficient (see section 3.1.2), which is
interpreted here as another supporting factor to the “force view-
point”. Based on our observations, the increase of the thrust coef-
ficient results in the reinforcement of the TI in the wake.
Turbulence in the wake of the upstream turbine could be estimated
by the thrust coefficient, and could be taken into account in con-
trolling the fatigue load of downstream turbines. These findings
suggest that derating options to maximize energy conversion and
minimize unsteady loads within the whole wind farm could be also
driven by drag reduction, rather than by power reduction only.
In addition to tip speed ratio and blade pitch angle, the turbine
model performance and corresponding wake have been also
studied in yawed conditions: with increasing yaw angle, not only
the whole wake velocity deficit and the TI was reduced, but also a
marked asymmetry was observed to characterize the wake struc-
tures. The position of the maximum velocity deficit was observed to
skew to one side, varying non-linearly along the wind direction. In Fig. A.1. Relative Standard Deviation (RSD) of experimental measurements of the
order to describe such offset, a general model primarily depending power coefficient and thrust coefficient. sCp(l) are the standard deviations calculated
on the yaw angle and the thrust coefficient is proposed and verified. at a specific tip speed from 4 repeated experiments. sCT(l) are the standard deviations
calculated at a specific tip speed over N ¼ 12 independent datasets resulting from 4
A simpler model is also proposed for cases where CT cannot be
repeated experiments each divided into 3 temporal non overlapping windows. Cp ðlÞ
estimated under yawed operating conditions. and CT ðlÞ are the overall average power coefficient and thrust coefficient as a function
This study demonstrated that the change of the tip speed ratio, of the tip speed ratio.
the pitch angle or the yaw angle has a different effect on the turbine
performance and the downstream wake. In a wind farm, those
830 B. Dou et al. / Energy 166 (2019) 819e833

Fig. A.2. Relative Standard Deviation of experimental measurements in the wake: mean velocity (a) and the turbulence intensity (b) at l ¼ 5, b ¼ 0 , and g ¼ 0 . For the sake of
distinguishing the experimental error between the near turbine region and the far wake region, the domains 3.5D-5.5D and 6.5D-8.5D are selected, respectively. RSDU (3.5D-5.5D)
represents the standard deviation sU(x,y) and sTI(x,y) spatially averaged in a range from 3.5D to 5.5D at constant y (m ¼ 5 is the number of downstream experimental spans from
3.5D to 5.5D, or from 6.5D to 8.5D, with an increment of 0.5D). sU(x,y) and sTI(x,y) are standard deviations at location (x, y) of the wake velocity and the turbulence intensity,
calculated over N ¼ 12 independent time series, each providing Ui(x,y) and TIi(x,y) (see Eqs. A.1 and A.2).

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

XN h i.
sU ðx; yÞ ¼ i¼1
U i ðx; yÞ  Uðx; yÞ ðN  1Þ (A.1)

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

XN h i.
sTI ðx; yÞ ¼ i¼1
TIi ðx; yÞ  TIðx; yÞ ðN  1Þ (A.2)

P PN
where Uðx; yÞ ¼ N i¼1 Ui ðx; yÞ=N and TIðx; yÞ ¼ i¼1 TIi ðx; yÞ=N are
the average wake velocity and the average turbulence intensity of
repeated measurements, Ui(x,y) and TIi(x,y).

Fig. A.3. Contours of the wind velocity (a) and the TI (b) distribution with different pitch angles at l ¼ 5.
B. Dou et al. / Energy 166 (2019) 819e833 831

Fig. A.4. Contours of the wind velocity (a) and the TI (b) distribution with different pitch angles at l ¼ 6.

Fig. A.5. Contours of the wind velocity (a) and the TI (b) distribution with different yaw angles at l ¼ 5.
832 B. Dou et al. / Energy 166 (2019) 819e833

Fig. A.6. Contours of the wind velocity (a) and the TI (b) distribution with different yaw angles at l ¼ 6.

Fig. A.7. Diagram of the acquisition method for the maximum wind velocity deficit position. Point A, B, and C are three adjacent measurement points, and B is the minimum wind
velocity position. DUl is the velocity difference between A and B. DUr is the velocity difference between C and B. Point O is the intersection of the perpendicular bisectors of AB and
BC. j is the angle between AB and BO. c is the angle between CB and BO. x is the angle between BO and the vertical line. d is the offset correction. (a) and (b) are two cases due to the
difference in wind velocity between points A and C. The relationship among these parameters is as follows:
B. Dou et al. / Energy 166 (2019) 819e833 833

8 snowfall reveals large-scale flow structures in the wake of a 2.5-MW wind


>
> DUl DUr turbine. Nat Commun 2014;5:4216.
>
> j þ c þ arctan þ arctan ¼p [27] Hong J, Guala M, Chamorro LP, Sotiropoulos F. Probing wind-turbine/
>
> 0:2R 0:2R
>
> atmosphere interactions at utility scale: novel insights from the EOLOS
>
> 1
>
> OB cos j ¼ AB
wind energy research station. J Phys Conf 2014;524, 012001.
>
> [28] Iungo GV. Experimental characterization of wind turbine wakes: wind tunnel
>
> 2
>
> tests and wind LiDAR measurements. J Wind Eng Ind Aerod 2016;149:35e9.
>
< 1 [29] Schümann H, Pierella F, Sætran L. Experimental investigation of wind turbine
OB cos c ¼ BC (A.3) wakes in the wind tunnel. Energy Proced 2013;35:285e96.
> 2
>
>
[30] Howard KB, Chamorro LP, Guala M. A comparative analysis on the response of
>
> AB ¼ DU l þ ð0:2RÞ2
2 2 a wind-turbine model to atmospheric and terrain effects. Boundary-Layer
>
>
>
> Meteorol 2015;158(2):229e55.
> 2
> n Cal R. Statistical analysis of
>
> BC ¼ DU 2r þ ð0:2RÞ2 [31] Hamilton N, Suk Kang H, Meneveau C, Bayoa
>
> kinetic energy entrainment in a model wind turbine array boundary layer.
>
>
> J Renew Sustain Energy 2012;4(6), 063105.
: c þ arctan DUr  p ¼ arcsin d
> [32] Iungo GV, Viola F, Camarri S, Porte-Agel F, Gallaire F. Linear stability analysis
0:2R 2 OB of wind turbine wakes performed on wind tunnel measurements. J Fluid
Mech 2013;737:499e526.
Then, d and the modified offset Yoffset can be obtained through [33] Singh A, Howard KB, Guala M. On the homogenization of turbulent flow
the above equations: Yoffset ¼ YB þ d, where YB is the distance be- structures in the wake of a model wind turbine. Phys Fluids 2014;26(2),
tween point B and y/R ¼ 0. 025103.
[34] Whale J, Anderson CG, Bareiss R, Wagner S. An experimental and numerical
study of the vortex structure in the wake of a wind turbine. J Wind Eng Ind
References Aerod 2000;84(1):1e21.
[35] Hu H, Yang Z, Sarkar P. Dynamic wind loads and wake characteristics of a
[1] Update AM. Global wind report. Global Wind Energy Council; 2016. wind turbine model in an atmospheric boundary layer wind. Exp Fluid
[2] The J, Yu H. A critical review on the simulations of wind turbine aerodynamics 2011;52(5):1277e94.
focusing on hybrid RANS-LES methods. Energy 2017;138:257e89. [36] Medici D, Alfredsson PH. Measurements behind model wind turbines: further
[3] de Prada Gil M, Gomis-Bellmunt O, Sumper A, Bergas-Jane  J. Power generation evidence of wake meandering. Wind Energy 2008;11(2):211e7.
efficiency analysis of offshore wind farms connected to a SLPC (single large [37] Li Qa, Murata J, Endo M, Maeda T, Kamada Y. Experimental and numerical
power converter) operated with variable frequencies considering wake ef- investigation of the effect of turbulent inflow on a Horizontal Axis Wind
fects. Energy 2012;37(1):455e68. Turbine (Part I: power performance). Energy 2016;113:713e22.
[4] Adaramola MS, Krogstad PÅ. Experimental investigation of wake effects on [38] Li Qa, Murata J, Endo M, Maeda T, Kamada Y. Experimental and numerical
wind turbine performance. Renew Energy 2011;36(8):2078e86. investigation of the effect of turbulent inflow on a Horizontal Axis Wind
[5] Lee J, Son E, Hwang B, Lee S. Blade pitch angle control for aerodynamic per- Turbine (part II: wake characteristics). Energy 2016;113:1304e15.
formance optimization of a wind farm. Renew Energy 2013;54:124e30. [39] Xie W, Zeng P, Lei L. Wind tunnel experiments for innovative pitch regulated
[6] Thomsen K, Sorensen P. Fatigue loads for wind turbines operating in wakes. blade of horizontal axis wind turbine. Energy 2015;91:1070e80.
J Wind Eng Ind Aerod 1999;80(1e2):121e36. [40] Krogstad P-Å, Adaramola MS. Performance and near wake measurements of a
[7] Li Qa, Kamada Y, Maeda T, Murata J, Yusuke N. Effect of turbulence on power model horizontal axis wind turbine. Wind Energy 2012;15(5):743e56.
performance of a Horizontal Axis Wind Turbine in yawed and no-yawed flow [41] Haans W, Sant T, van Kuik G, van Bussel G. Measurement of tip vortex paths in
conditions. Energy 2016;109:703e11. the wake of a HAWT under yawed flow conditions. J Sol Energ-T Asme
[8] Mehta D, van Zuijlen AH, Koren B, Holierhoek JG, Bijl H. Large Eddy Simulation 2005;127(4):456e63.
of wind farm aerodynamics: a review. J Wind Eng Ind Aerod 2014;133:1e17. [42] Bastankhah M, Porte -Agel F. Experimental and theoretical study of wind
[9] Vermeer LJ, Sørensen JN, Crespo A. Wind turbine wake aerodynamics. Prog turbine wakes in yawed conditions. J Fluid Mech 2016;806:506e41.
Aero Sci 2003;39(6e7):467e510. [43] Jime  Crespo A, Migoya E. Application of a LES technique to characterize
nez A,
[10] Ainslie JF. Calculating the flowfield in the wake of wind turbines. J Wind Eng the wake deflection of a wind turbine in yaw. Wind Energy 2009;13(6):
Ind Aerod 1988;27(1e3):213e24. 559e72.
[11] Katic I, Højstrup J, Jensen NO. A simple model for cluster efficiency. Confer- [44] Xie W, Zeng P, Lei L. Wind tunnel testing and improved blade element mo-
ence A simple model for cluster efficiency. p. 407-410. mentum method for umbrella-type rotor of horizontal axis wind turbine.
[12] Larsen GC. A simple wake calculation procedure. 1988. Energy 2017;119:334e50.
[13] Bastankhah M, Porte -Agel F. A new analytical model for wind-turbine wakes. [45] Chamorro LP, Porte -Agel F. A wind-tunnel investigation of wind-turbine
Renew Energy 2014;70:116e23. wakes: boundary-layer turbulence effects. Boundary-Layer Meteorol
[14] Frandsen S. On the wind speed reduction in the center of large clusters of 2009;132(1):129e49.
wind turbines. J Wind Eng Ind Aerod 1992;39(1e3):251e65. [46] Hassan U, Taylor GJ, Garrad AD. The dynamic-response of wind turbines
[15] Sørensen JN, Myken A. Unsteady actuator disc model for horizontal axis wind operating in a wake flow. J Wind Eng Ind Aerod 1988;27(1e3):113e26.
turbines. J Wind Eng Ind Aerod 1992;39(1e3):139e49. [47] Ali N, Kadum HF, Cal RB. Focused-based multifractal analysis of the wake in a
[16] Sørensen JN, Shen WZ. Numerical modeling of wind turbine wakes. J Fluid Eng wind turbine array utilizing proper orthogonal decomposition. J Renew Sus-
2002;124(2):393. tain Energy 2016;8(6), 063306.
[17] Wu Y-T, Porte -Agel F. Large-Eddy simulation of wind-turbine wakes: evalu- [48] Sagmo KF, Bartl J, Sætran L. Numerical simulations of the NREL S826 airfoil.
ation of turbine parametrisations. Boundary-Layer Meteorol 2010;138(3): J Phys Conf 2016;753, 082036.
345e66. [49] Sarmast S, Chivaee HS, Ivanell S, Mikkelsen RF. Numerical investigation of the
[18] Wu Y-T, Porte -Agel F. Modeling turbine wakes and power losses within a wake interaction between two model wind turbines with span-wise offset.
wind farm using LES: an application to the Horns Rev offshore wind farm. J Phys Conf 2014;524, 012137.
Renew Energy 2015;75:945e55. [50] Krogstad PA, Lund JA. An experimental and numerical study of the perfor-
[19] Stevens RJAM, Martínez-Tossas LA, Meneveau C. Comparison of wind farm mance of a model turbine. Wind Energy 2012;15(3):443e57.
large eddy simulations using actuator disk and actuator line models with [51] Schmitz G. Theorie und entwurf von windra €dern optimaler leistung. Wis-
wind tunnel experiments. Renew Energy 2018;116:470e8. senschaftliche Zeitschrift der Universita€t Rostock 1955;5(5):379e91.
[20] Yang X, Kang S, Sotiropoulos F. Computational study and modeling of turbine [52] Wu B, Zhang X, Chen J, Xu M, Li S, Li G. Design of high-efficient and universally
spacing effects in infinite aligned wind farms. Phys Fluids 2012;24(11), applicable blades of tidal stream turbine. Energy 2013;60:187e94.
115107. [53] Wang Z, Tian W, Hu H. A Comparative study on the aeromechanic perfor-
[21] Yang X, Hong J, Barone M, Sotiropoulos F. Coherent dynamics in the rotor tip mances of upwind and downwind horizontal-axis wind turbines. Energy
shear layer of utility-scale wind turbines. J Fluid Mech 2016;804:90e115. Convers Manag 2018;163:100e10.
[22] Howard KB, Singh A, Sotiropoulos F, Guala M. On the statistics of wind turbine [54] Foti D, Yang X, Sotiropoulos F. Similarity of wake meandering for different
wake meandering: an experimental investigation. Phys Fluids 2015;27(7), wind turbine designs for different scales. J Fluid Mech 2018;842:5e25.
075103. [55] Steinbuch M, de Boer WW, Bosgra OH, Peters SAWM, Ploeg J. Optimal control
[23] Howard KB, Hu JS, Chamorro LP, Guala M. Characterizing the response of a of wind power plants. J Wind Eng Ind Aerod 1988;27(1e3):237e46.
wind turbine model under complex inflow conditions. Wind Energy [56] Sarmast S. Numerical study on instability and interaction of wind turbine
2015;18(4):729e43. wakes. KTH Royal Institute of Technology; 2014.
[24] Wang Z, Ozbay A, Tian W, Hu H. An experimental study on the aerodynamic [57] Larsen GC, Madsen HA, Thomsen K, Larsen TJ. Wake meandering: a pragmatic
performances and wake characteristics of an innovative dual-rotor wind approach. Wind Energy 2008;11(4):377e95.
turbine. Energy 2018;147:94e109. [58] Shapiro CR, Gayme DF, Meneveau C. Modelling yawed wind turbine wakes: a
[25] Cal RB, Lebro  n J, Castillo L, Kang HS, Meneveau C. Experimental study of the lifting line approach. J Fluid Mech 2018:841.
horizontally averaged flow structure in a model wind-turbine array boundary [59] Shakoor R, Hassan MY, Raheem A, Wu Y-K. Wake effect modeling: a review of
layer. J Renew Sustain Energy 2010;2(1), 013106. wind farm layout optimization using Jensen‫׳‬s model. Renew Sustain Energy
[26] Hong J, Toloui M, Chamorro LP, Guala M, Howard K, Riley S, et al. Natural Rev 2016;58:1048e59.

You might also like