Renewable Energy 39 (2012) 329e338

Contents lists available at SciVerse ScienceDirect

Renewable Energy
journal homepage: www.elsevier.com/locate/renene

Wake effect in wind farm performance: Steady-state and dynamic behavior

F. González-Longatt*, P. Wall, V. Terzija
School of Electrical and Electronic Engineering, The University of Manchester, Manchester M60 1QD, United Kingdom

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

Article history: The aim of this paper is to evaluate the impact of the wake effect on both the steady-state operation and
Received 14 May 2011 dynamic performance of a wind farm and provide conclusions that can be used as thumb rules in generic
Accepted 19 August 2011 assessments where the full details of the wind farms are unknown. A simplified explicit model of the
Available online 16 September 2011
wake effect is presented, which includes: the cumulative impact of multiple shadowing, the effects of
wind direction and the wind speed time delay. The model is implemented in MATLABÒ and then inte-
Keywords:
grated into a power system simulation package to describe the wake effect and its impact on a wind farm,
Wind farm
particularly in terms of the wake coefficient and overall active power losses. Results for two wind farm
Dynamic behavior
Steady-state behavior
layouts are presented to illustrate the importance of wind turbine spacing and the directionality of wind
Power system speeds when assessing the wake effect during steady-state operation and dynamic behavior.
Steady-state wake effect Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction models. They are based on approximations of either the Naviere-

Stokes or vorticity transport equations [8e10].
Wind turbines extract energy from the wind to produce elec- The exact modeling of the wind speed distribution within
tricity; therefore, the wind leaving the turbine must have a lower a wind park is a fairly complicated task and many of the necessary
energy content than the wind upstream of the turbine [1]. As parameters are not routinely available [1]. The choice of a suitable
a consequence, the wind downstream of a wind turbine has model depends on three factors: the desired computational time,
reduced speed and is turbulent; this downstream wind is the wake the necessary accuracy of prediction, and the available wind
of the turbine. As the wind flow proceeds further downstream this modeling parameters.
wake will begin to spread and gradually return to free stream In this paper, a simplified approach for the simulation of the
conditions. If a wake intersects with the swept area of a downwind wake effect is presented and used to investigate the impact of the
turbine the downwind turbine is said to be shadowed by the wake effect upon the steady-state and dynamic behavior of a wind
turbine producing the wake. farm.
The two main effects of a wake are: (i) a reduction in the wind The model presented here only uses those parameters and data
speed, which in turn reduces the energy production of the wind that are commonly available for a wind turbine and wind park.
farm; (ii) an increase in the turbulence of the wind, potentially Several aspects of the wake effect, such as: the cumulative impact
increasing the dynamic mechanical loading on downwind turbines. of multiple shadowing, the effect of wind direction, and the wind
It is important to consider these wake effects in the design of speed time delay are all taken into consideration. The model has
a wind farm in order to maximize the energy output and lifetime of been used to evaluate the effect of turbine wakes upon the energy
the machines [2]. production, calculated using several indexes, of two different
A large number of numerical models, of varying complexity, layouts of wind farm. This paper is novelty because it presents an
have been developed to describe a wake. In general, they can be exhaustive analysis of the effect of the wake effect on the steady-
classified as either explicit or implicit [2]. The explicit, or kinematic, state and dynamic behavior. The conclusions presented here can
wake models are the earliest and use self-similar velocity profiles be used as thumb rules in generic assessments where the full
determined semi-empirically [3e7]. The implicit wake models details of the wind farms are unknown.
were developed as elaborate alternatives to the explicit wake

2. Wake model
* Corresponding author. Tel.: þ44 (0)779 5634298.
E-mail addresses: [email protected], [email protected] (F. González- The development of models to describe wind turbine wakes
Longatt). began in the 1980s [11] for the purpose of estimating the fall in

0960-1481/$ e see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.renene.2011.08.053
330 F. González-Longatt et al. / Renewable Energy 39 (2012) 329e338

output power they cause [12]. Some of these models include: Now defining a thrust coefficient, CT ¼ 4a(1  a), the axial thrust
Ainslie’s model [11], Frandsen’s model [13], the Mosaic Tile model force can be expressed as:
[14] and Jensen’s model [15]. In some studies it is necessary that
any wake models used are: straightforward, dependent on rela- 1
T ¼ rA C v2 (6)
tively few wake measurements and economic in terms of the 2 0 T 0
necessary computing power [16]. However, despite their relative Finally, it is convenient to define the relationship between the
simplicity, these methods tend to give results that are in reasonable downstream wind velocity (v1) and the free wind speed (v0) in
agreement with the available data in the case of a single wake terms of the turbine thrust coefficient (CT):
within a small wind farm, and a simple meteorological environ-
v1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ment [12,16e18]. In addition, a comparison of different wake ¼ ð1  2aÞ ¼ 1  CT (7)
models, presented in Ref. [17], does not suggest any particular v0
difference, in terms of accuracy, between the sophisticated and
The above relationship is based on the assumption of idealized
simplified models.
downstream wake expansion; this is the most simple turbine wake
In this section the simple model for both a single wake and
model and is computationally very efficient.
multiple wakes is presented. The effects of the wind speed time
delay are included in these models.
2.2. N.O. Jensen wake model

2.1. Wake behind an idealized wind turbine

The N.O. Jensen wake model is a simple single wake model. The
model is well documented in Ref. [6] and it is based on the
Assuming an idealized turbine, where the air flow around and
assumption of a wake with linearly expanding diameter [20].
behind the turbine is without rotation and friction, it is possible to
derive some important general equations describing the wake wind
2.2.1. Single wake model
speeds [15,19]. The simplified Bernoulli equation stating that the
If the near field behind a wind turbine is neglected the resulting
mechanical energy, per unit mass, along a streamline is conserved:
wake behind the wind generator can be treated as a turbulent
wake. The spread of a feature like this is such that the linear
rV 2
þp ¼ H (1) dimension (radius r) is proportional to the down-wind distance, x.
2 A balance of momentum gives (see Fig. 2):
where r is the air density, V is the wind speed, p is the pressure and

H is the total energy (constant along any streamline). Applying the pr02 u þ p r2  r02 v0 ¼ pr2 v1 (8)
Bernoulli equation for the wind just in front of and just behind the
rotor allows the drop in pressure over the rotor plane to be calcu- Assuming a linear expansion of the wake, the path taken by the
lated (see Fig. 1): wind that has passed through the turbine blades is represented by
a cone. The radius of this (shadow) cone r, can be calculated using
1

the following expression:
Dp ¼ r v20  v21 (2)
2
r ¼ r0 þ ax (9)
The axial thrust force (the force acting in the direction of the
wind), denoted as T, is calculated using the pressure difference: The dimensionless scalar a determines how quickly the wake
expands with distance and it is defined as:
T ¼ DpA0 (3)
1
where Dp is the pressure difference and A0 is the rotor area. Now a ¼   (10)
defining an ‘axial interference factor’, a, which is used to define the
z
2 ln
ratio between the reduction in speed at the plane of the blade disc z0
(u) and the undisturbed wind speed well upstream of the turbine
(v0):

u ¼ ð1  aÞv0 and thus v1 ¼ ð1  2aÞv0 (4)

A0 A( x)
v0 v0
The substitution of (4) for v1 in the thrust definition (3) yields:

T ¼ 2rað1  aÞv20 A0 (5)

α
u 1 v1
A1
A0
A

u v1 r0 rx = r0 + α x
v0
p0
p
p + Δp p1 = p0

Fig. 1. Air flow near an idealized turbine: velocity and pressure. This model allows Fig. 2. The N.O. Jensen wake model is a simple single wake model that assumes linear
some key wake equations to be derived. expansion of the wake cone.
 F. González-Longatt et al. / Renewable Energy 39 (2012) 329e338 331

where z is the hub height of the turbine generating the wake and z0 Ax ,i
is a constant called surface roughness, which depends on the char-
Ashadow,i
acteristics of the local terrain. The scalar a can assume many values Wake stream

depending on the local terrain and/or wind climate conditions. For

free wind, i.e. not affected by any upstream turbine a ¼ 0.04 is ri ( xij ) r0
usually suitable, otherwise a value of 0.08 could be used for a [6].
The velocity in the wake at a distance x from the wind turbine θ zij
can be obtained by solving (8) in terms of v1:
Rotor

pffiffiffiffiffiffiffiffiffiffiffiffiffiffi 
r 2 Swap
Area
0
v1 ¼ v0 þ v0 1  CT  1 (11)
r
A0
This equation provides the wake speed at the down-wind
location as a function of the incoming wind speed. Lij
dij
2.2.2. Multiple wake model
WTj
The effect discussed in the previous section is a single wake
vi ( xij )
Downwind turbine
model. In a wind farm every upstream wind turbine that shadows
a studied downstream turbine will have an effect on the studied
turbines performance [21]. This multiple wake effect is shown in v0 v0
Fig. 3, where the direction of free wind is shown for the undis- Wake α
Stream
xij
turbed wind speed v0. WTi vi 0 1
In this wind farm configuration, the wind speed at wind turbine Upwind turbine
j is affected not only by the upstream wind turbine that is directly in
front of it, wind turbine i, but also by other upstream wind turbines, v0
such as 1, 2 and 3.
In order to obtain a usable result for wind farms with many Fig. 4. A detailed example of partial shadowing that can be used to calculate the part
turbines, the effects of the multiple single wakes must be combined of a turbine’s swept area that is shadowed by another turbine’s wake. The effect a wake
into a single effect. A purely empirical mean is usually used to has on a turbine can be weighted using this value.

model the interaction between these multiple wakes. A detailed

model of wake effects considers the shadowed areas of the
2.2.3. Wind speed delay between two successive wind turbines
upstream wind turbines. This shadowing is a measure of the degree
From the previous section, the relationship between the
of overlap between the area spanned by the wakes shadow cone
incoming wind speeds at the two units can be obtained based on
(Ashadow,i) and the area swept by the turbine experiencing shad-
the wake flow. Fig. 5 gives a diagram of the incoming wind speed
owing (A0); an example of which is depicted in Fig. 4.
for two successive units. The wake speed at the down-wind turbine
There are four distinct shadowing possibilities; namely:
location is a function of the incoming wind speed at the upwind
complete shadowing, quasi-complete shadowing, partial shadow-
location.
ing and no shadowing.
A relationship for the wind speed delay between two successive
If the wind turbines have the same diameter (2r0) then the area
wind turbines (sij) that are separated by a distance xij, can be
of the turbine being shadowed can be calculated using the
derived considering, vi, the incoming wind speed at turbine i and,
following basic trigonometric relationships:
vj, the incoming wind speed at wind turbine j.
! ! Assuming the air acceleration between the two successive wind
  2 L dij  Lij
Ashadow;i ¼ ri xij 1
cos  ij þ r02 cos 1   dij zij turbines, aij, is constant then:
ri xij ri xij
(12) d2 x d2 v
¼ ¼ aij (14)
dt dt
" !2 #
 n

X pffiffiffiffiffiffiffiffiffiffiffiffiffiffi The boundary conditions of the above equation (14) are:
r Ashadow;i
vj xij ¼ vi 1  1  1  CT 0 (13)
ri xij A0
i¼1

vi ( t ) ui ( t ) v j (t ) u j (t )

v0
2

1 i j
WTi WT j
3 xij

Fig. 5. The case of two successive turbines is used to calculate the wind speed delay for
Fig. 3. This example of multiple wakes shows why multiple wake models must the turbines; that is, the time delay between a wake being created at turbine i and the
combine the influence of every upstream wake, which shadows a turbine, into a single same wake then reaching the blades of turbine j. (a) 4  4 wind turbines; (b) 10  10
effect to obtain a usable result. wind turbines.
332 F. González-Longatt et al. / Renewable Energy 39 (2012) 329e338


x t ¼ sij ¼ xij a
vðt
 ¼ 0Þ ¼ vi (15)
v t ¼ sij ¼ vj 0.95

Wake Coefficient (p.u.)

0.9
The incoming wind’s flow time, or delay time, between the two 0.85
successive wind turbines can be calculated using the following
0.8
equation:
0.75 4D
6D
2xij 0.7
sij ¼ (16) 8D
vi þ vj 0.65 10D

Finally, the wind speed at the shadowed wind turbine j is 0 10 20 30 40 50 60 70 80 90

Angle (deg)
calculated in terms of the speed of the wind approaching the
shadowing turbine i. b
1
   
vj t þ sij ¼ v0 t þ sij þ ui ðtÞ  v0 t þ sij bij (17)
0.9

Wake Coefficient (p.u.)

The coefficient bij is the shadowing factor; defined as the swept 0.8
area of turbine j under the shadow of turbine i, normalized to the
swept area of turbine j. 0.7
4D
0.6
2.2.4. Wake combination 6D
0.5 8D
In order to obtain a usable result for wind farms with many
10D
turbines the effects of multiple single wakes must be combined into 0.4
a single effect. Wind turbines in a wind farm may experience 0 10 20 30 40 50 60 70 80 90
Angle (deg)
varying degrees of shadowing from upstream turbines. The result
of being exposed to many single wakes, with different degrees of Fig. 6. The effect of tower spacing upon the wake coefficient can be seen. Tower
shadowing and time delay, is a non-uniform distribution of velocity spacing values are defined in terms of the turbine rotor diameter, D. Two array layouts
that may place increased mechanical loading on the turbine blades. are considered; regular grids of 4  4 and 10  10 turbines. (a) Tower spacing: 4D. (b)
Tower spacing: 10D.
There are a variety of methods for combining these separate wakes
[20,22]: sum of squares of velocity deficits, energy balance,
geometric sum and linear superposition.
The following wind turbine details are also necessary: geometrical
In this paper the law of momentum conservation is used to
characteristics (radius and hub height), the power coefficient curve
determine the resultant wind speed at each single wind turbine
Cp(vw) and the thrust coefficient curve CT(vw).
location. This approach combines each single wake into a single
The calculation procedure is as follows:
equivalent wind speed, using the following relationship [1,6]:
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u n h
 i2
uX
vj ðtÞ ¼ vj0 ðtÞ þ u
u bij vj;k xkj ; t  vj0 ðtÞ (18) a 1
u
t k¼1
0.9
Wake Coefficient (p.u.)

ksj

0.8
where vj(t) is the resultant wind speed for an arbitrary turbine j,
vj,k(xkj,t) is the wind speed approaching turbine j (with turbine k as 0.7 2x2
the shadowing turbine) and vj0 is the incoming wind speed at j 4x4
0.6 6x6
without shadowing.
8x8
0.5
10x10
3. Implementation of the wake model
0.4
0 10 20 30 40 50 60 70 80 90
The aim of this paper is to evaluate the effects which wakes have Angle (deg)
upon the steady-state and dynamic behavior of wind farms. For this
purpose the wake model briefly outlined above has been devel- b 1
oped, firstly, as a MatlabÒ [23] program. This program was used to
Wake Coefficient (p.u.)

directly evaluate the effects in steady-state and a later improve- 0.95

ment in the program allowed the inclusion of the time delay effects
of wakes. The features developed in MatlabÒ were also integrated
into the DIgSILENTÔ [24] simulation package in order to perform 0.9 2x2
analysis of the dynamic behavior of a wind farm. This section 4x4
outlines the calculation method and the quantitative indicator used 0.85 6x6
8x8
to evaluate the overall impact of the wake effect on the output
10x10
power of the wind farm.
0.8
0 10 20 30 40 50 60 70 80 90
3.1. Outline of the calculation method Angle (deg)

Fig. 7. The effect of the number of wind turbines upon the wake coefficient can be
The wake model used here requires a site plan of the wind farm seen. Regular arrays with between 2  2 and 10  10 wind turbines are used. The tower
with the coordinates of each wind turbine location within the farm. spacing values are defined in terms of the rotor diameter, D.
 F. González-Longatt et al. / Renewable Energy 39 (2012) 329e338 333

1 performed. The tests described in this section deal with the

effect the wind direction and wind farm layout, specifically the
0.9
Wake Coefficient (p.u.)

2x2 → number of turbines and their spacing, have upon the wake
0.8 coefficient.
The wind direction and wind farm layout have an important
0.7 4x4 →
effect on wind farm performance. This is because they modify the
0.6 6x6 → orientation and location of the wake cones; therefore, changes in
8x8 →
wind direction or wind farm layout produce changes in the wakes
0.5
10x10 → interaction (overlapping area) and consequently their effects on the
0.4 power output of each individual turbine.
3 4 5 6 7 8 9
In order to evaluate the effect of wind direction upon wind farm
Tower spacing (rotor diameters)
performance the wake coefficient has been plotted in terms of the
Fig. 8. The decrease in the wake coefficient as tower spacing falls, due to increased wind direction between 0 and 90 for a square array with uniform
wake interaction, can be seen. Regular arrays with 2  2, 4  4, 6  6, 8  8 and 10  10 spacing. Fig. 6 shows how changes in the wind direction modify the
turbines are used. wake coefficient for several different values of tower spacing (in
terms of the rotor diameter, D) and for two wind farm arrays; one
1. The wind speed calculation is initiated with the turbine posi-
array has sixteen turbines in a 4  4 grid whilst the other has one
tioned at the most upstream position.
hundred turbines in a 10  10 grid. In both arrays the turbines are
2. Calculate the wind speeds downstream of this turbine for all
regularly spaced so the spacing between each turbine in the same
downwind turbine positions.
row and each row of turbines is the same.
3. Calculate the wind speed for all downstream turbine positions
The lowest values of wake coefficient occur for the smallest
relative to the free wind speed v0(t).
angles (0 e9 ) and the largest angles (85 e90 ); under these
4. If the downstream turbine is in a partial wake reduce the
conditions the wakes of wind turbines in the same row, or column
velocity according to the relevant shadowing factor (bij).
respectively, will produce either complete or quasi-complete
5. Calculate the resultant wind speed in the downstream turbine,
shadowing. A similar condition is found again for wind directions
combining all single wakes into a uniform wind speed
between 42 and 48 and the wake coefficient values are the
according to the law of momentum conservation.
second lowest values. This is because whilst the downwind
6. Continue with the next turbine (using step 1), by summing the
turbines will still be under complete or quasi-complete shadowing,
law of momentum conservation.
as in the case of (0 e9 ; 85 e90 ), the apparent distances between
a turbine and those turbines shadowing it are larger. The variation
in the wind park performance for wind directions between these
3.2. Quantifying the wake effect zones of high shadowing is better. The wake coefficient has a ripple;
the amplitude of which is dependent on tower spacing whilst the
The wake coefficient (cwake) was used to evaluate the overall frequency is determined by the number of turbines.
impact of the wake effect on the output power of the wind farm. It An increase in the number of wind turbines in the wind park
has been defined as follows: decreases the wake coefficient when the spacing is maintained.
Given a fixed number of wind turbines inside the park smaller
total output with wake effect spacing between the turbines increases the shadow interference
cwake ¼ (19)
total output power neglecting wake effect compromising the performance of the wind farm.
Fig. 7 plots the variation of the wake coefficient with wind
This coefficient combines the local effect of the interactions
direction for wind farms with different numbers of turbines and
between every individual wake at each individual turbine into
two values of turbine spacing (4D and 10D). From this the depen-
a single measure of the wakes effect on the wind farm output.
dence of the wake coefficient upon wind direction is clear. In
addition, the changes in the wake coefficient tend to be larger for
4. Influence of wind direction and wind farm layout smaller values of tower spacing.
Changes in the wind direction have a more significant effect on
In order to evaluate the effect of the wakes upon the steady- the wake coefficient in tightly packed wind farms than in wind
state output power of a wind farm some tests have been farms with increased turbine spacing. If the number of turbines

0° 0% calm
20 337.5° 22.5°
max
315° 45°

292.5° 67.5°
15 mean
Value (m/s)

270° 90°
min
28%
10 112.5°
247.5°
56%
225° 135°
84%

J F M A M J J A S O N D A 202.5° 157.5°
180°

Fig. 9. The monthly statistics for the wind speed and wind speed frequency rose of the year long wind regime used in the long term simulations.
334 F. González-Longatt et al. / Renewable Energy 39 (2012) 329e338

1
a b
16 16
13 14 15 16
0.8
Output Power(p.u)

13 15

10 12
v0 11 12
9 11
0.6 φ
6 8
0° 7 8 Δy
0.4 Δy 5 7

2 4
1 2 3 4
1 3
0.2 Δx Δx
50 100 150 200 250 300 350
Time (days) Array I: Regular Array II: Irregular
Fig. 10. The output power for an ideal wind farm layout, which experiences no wake Fig. 12. The two non-ideal wind farm layouts compared in these simulations. (a) Fixed
effects, exposed to the wind regime described in Appendix. direction at 0 . (b) Considering changes in the direction.

increases then the mean value of the wake coefficient falls; as does more than 75% of rated power during a year and generation above
the magnitude of the changes in the wake coefficient as the angle this level occurs for an average of 16.74 h/day.
varies. In order to evaluate the effect of the wind direction and layout of
The effect of the number of wind turbines, and the spacing of a wind farm upon overall performance two different configurations
those turbines, upon the wake coefficient is relevant when were used (Fig. 12): (a) Array I: regular distribution with equally-
attempting to select an appropriate layout for a proposed wind spaced wind turbines every 4-rotor-diameters (4D  4D), and (b)
farm. Fig. 8 illustrates the degradation of performance when wind Array II: the wind farm has tower spacing of 4-rotor-diameters
turbines are too close together. This degradation is more significant along its rows and 7-diameter spacing between rows (4D  7D) in
for an array with more turbines because their will be more wake a configuration of staggered towers as shown in Fig 12b.
cones shadowing each turbine. The wake coefficient and probability distribution functions for
Array I are shown in Figs. 13 and 14 respectively for two wind
5. Steady-state (long term effects) regimes; the first regime has a fixed direction of 0 whilst the
second has variable direction. The wind turbine layout of Array I
The long term effects of wakes in a wind farm are evaluated for provides the maximum shadowing possible when the wind direc-
one wind farm with 16 wind turbines (2 MW turbines, see details in tion is fixed equal to 0 . As a consequence of this the wind farm
Appendix) using a 365-day time series of wind speed and wind performance is lower in this scenario (Fig. 13a).
direction. The mean wind speed is 13.9407  2.032 m/s and the The average output power is 10.825% less than the no-wake
prevailing wind direction is from 0 , some statistical details about wind farm case; with a minimum wake coefficient of 0.6972 p.u.
these time series are shown in Fig. 9 (more details in Appendix). The rated power is only expected during 80 days/year with a daily
Initially, all wind turbines are placed in a straight line that is average of 8.72 h/day. When the effect of wind direction is
perpendicular to the prevailing wind direction. This wind farm considered in the wake model the average power output of Array I
layout effectively eliminates the wake effect for the wind regime increases by 8.25% and the time at rated output increases to 168
considered. Fig. 10 shows the output power in per unit based on the days/year; the minimum wake coefficient is 0.7361 p.u. This
nominal power of the wind farm; during the time period of the improvement is caused by the reduction of the shadowing factor as
evaluation the wind farm would have produced rated power for
approximately 224 days/year.
The wake coefficient for this wind farm geometry is the
maximum possible (Cwake ¼ 1.00 p.u.) because no wake interactions
a
Wake Coefficient (p.u)

occurred. The probability distribution function (PDF) is plotted in 1

Fig. 11; the best-fit Weibull distribution shows that the output
power falls within the range of 0.975e1.00 p.u. for 12.97% of the
0.8
time. This design provides a high probability (93.40%) of generating

0.6
14 50 100 150 200 250 300 350
Actual Data Time (day)
12
b
Probability Density (%)

Best-fit Weibull Distribution

10
Wake Coefficient (p.u)

8 1

6
0.8
4
k = 0.9661 p.u
2 c = 10.6032
0.6
0 50 100 150 200 250 300 350
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Time (day)
Output Power (p.u)
Fig. 13. The wake coefficient for Array I when exposed to the wind speed of the year
Fig. 11. The probability density function of the output power considering an ideal wind long wind regime with both fixed, 0 , (a) and actual (b) wind direction. (a) Fixed wind
farm layout that experiences no wake effects. direction at 0 . (b) Considering wind direction.
 F. González-Longatt et al. / Renewable Energy 39 (2012) 329e338 335

a 14 b 14 a 14 b 14
Actual Data Actual Data
Actual Data Actual Data 12 12
12

Probability Density (%)

Probability Density (%)
12 Best-fit Weibull Best-fit Weibull
Probability Density (%)

Best-fit Weibull

Probability Density (%)

Best-fit Weibull
10 k = 0.9411 p.u 10 k = 0.9642 p.u
10 k = 0.8844 p.u 10 k = 0.9443 p.u
c = 5.8609 c = 8.0382 c = 10.1052
c = 8.8879 8 8
8 8
6 6 6
6

4 4 4
4

2 2 2
2

0 0 0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Output Power (p.u) Output Power (p.u) Output Power (p.u) Output Power (p.u)

Fig. 14. The probability density function of the output power of Array I when exposed Fig. 16. The probability density function of the output power of Array II when exposed
to the wind speed of the year long wind regime with both fixed, 0 , (a) and actual (b) to the wind speed of the year long wind regime with both fixed, 0 , (a) and actual (b)
wind direction. (a) Fixed direction at 0 . (b) Considering changes in the direction. wind direction.

a consequence of better interaction between the wakes for different of a wind farm based on Array II is the highest; in terms of both
wind directions. In this particular case the average wind direction mean and minimum power output, and annual energy output.
is 7.3461 with a rather large standard deviation of 58.2169 , A better view of the rated power and energy delivered for the
this is more favorable than the scenario with a fixed direction of different cases is evaluated considering the capacity factor (CF); and
0 for the geometric configuration of Array I; as is shown in Figs. 13 from the results presented, the superior performance of Array II
and 14. (0.1454) is evident. The offsetting, or staggering, of one row of
The wake coefficient for Array II is shown in Fig. 15. It is evident towers behind another, provides a better wake coefficient because
that the best wake coefficient behavior occurs when the wind it reduces the amount of wake interaction within the array. This
direction is considered and, despite the minimum wake coefficient means that a significant increase in the annual energy output of the
(0.6982 p.u) being obtained in this case, the average value is same set of turbines, exposed to the same wind regime, has been
0.9948 p.u. If wind direction is not included the average output is achieved simply by adjusting the array layout in an attempt to
3.77% less than when direction is considered. The mean wake reduce the wake effects occurring within the array.
coefficient is only 0.9572 p.u. and the minimum value is 0.8646 p.u..
The probability density function of the output power is shown in 6. Dynamic (short-term effect)
Fig. 16. Array II exhibits superior behavior when the changes in
wind direction are included with rated power output for 222 days/ The wake effect is a factor during every time-scale of power
year at an average of 22.2 h/day. If fixed direction wind is consid- system behavior. An assessment of the impact of wakes upon the
ered the number of days/year at rated power is only 172. dynamic behavior of a single wind farm is performed through time
Finally, an overall picture of the power and energy production domain simulations using the wake model presented in this paper;
for each case is shown in Table 1. An array without wake effects is with the inclusion of the wind time delay between wind turbines
a favorable condition in terms of the total power production, and inside the wind farm. The variability of wind speed has been
mechanical stress on individual wind turbines. In this case the considered by using a 60 s time series of wind data, with an average
average output power obtained was highest when wake effects did value of 14.46078 m/s and standard deviation of 0.62808 (more
not occur; however, this design is neither technically nor details in Appendix), for the simulations. Wind direction is
economically feasible. When accounting for wake effects the output considered fixed during the 60 s simulation but multiple simula-
tions are performed to allow several different wind directions to be
considered.
a The wind farm consists of 16 constant-speed wind turbines
(2 MW each, details in Appendix) connected through a medium-
Wake Coefficient (p.u)

1
voltage underground distribution system with four feeders. Two
step up transformers in parallel feed the output power of the wind
0.8 farm into an external grid. The network shown in Fig. 17 has been
used as the test network.
The wind farm layouts previously defined as Array I and II, in
0.6 Section 5, have been used for a comprehensive evaluation of the
50 100 150 200 250 300 350

b Table 1
Power and energy output for the arrays considered.
Wake Coefficient (p.u)

1
Variable No wake Array I Array II

0.8 Wind direction Wind direction

Fixed Variable Fixed Variable

0.6 Mean output power (MW) 29.273 26.086 28.424 26.086 29.163
50 100 150 200 250 300 350 Minimum power (MW) 9.170 6.393 8.970 6.393 9.170
Time (day) Energy output 241.48 228.02 248.43 228.02 254.83
(GW h/year)
Fig. 15. The wake coefficient for Array II when exposed to the wind speed of the year CF 0.1348 0.1301 0.1417 0.1314 0.1454
long wind. (a) Fixed wind direction at 0 . (b) Considering wind direction.
336 F. González-Longatt et al. / Renewable Energy 39 (2012) 329e338

10 kV 0.69 kV a
1
13 14 15 16 0
External 66 kV 10
0.8

Wake Coefficient (p.u)

grid 20
9 10 11 12 30
0.6
40

v0 5 6 7 8
0.4

θ 0.2
1 2 3 4
0° 0
0 10 20 30 40 50 60
Fig. 17. Schematic layout of the distribution system supporting the wind farm; the Time (s)
collecting point is a substation. The specific network for each array layout is designed
b
-3
using the actual distances between the turbines in that layout. x 10
6
0
5 10
impact of the wake effect upon the wake coefficient and active 20
power losses in the wind farm. The internal network has been 4 30

Losses (p.u)
designed considering the actual distances between the wind 40
3
turbines in the each of the array layouts (Fig. 12).
Figs. 18 and 19 depict the variation of the wake coefficient over 2
the 60 s simulation period for five separate wind directions for both
Array I and Array II. 1
These figures show that the wind direction has a significant
0
influence upon the impact of the wake effect on the output power 0 10 20 30 40 50 60
and active power losses of the wind farm. Array I exhibits worse Time (s)
behavior, in terms of wake coefficient, at 0 than Array II because
Fig. 19. The influence of wind direction on the variation in the wake coefficient and
the layout of Array II maximizes the number of wind turbines power losses of Array II is clear. The wind direction is fixed for each simulation and
with undisturbed wind for this wind direction. Whilst the same a range of values from 0 to 40 are used.
design feature ensures that the opposite is true for a wind
direction of 40 .

a a 1
1 Array I
0 Array II
0.8
Wake Coefficient (p.u)

10
0.8
Wake Coefficient (p.u)

20
30 0.6
0.6 40
0.4
0.4
0.2
0.2
0
0 10 20 30 40 50 60
0
0 10 20 30 40 50 60 Time (s)
Time (s) -3

b x 10
b 6
x 10
-3 6
No wake
0 5 Array I
5 10 Array II
20 4
Losses (p.u)

4 30
Losses (p.u)

3
40
3
2
2
1
1
0
0 0 10 20 30 40 50 60
0 10 20 30 40 50 60 Time (s)
Time (s)
Fig. 20. A comparison of the wake coefficient and power losses for a no wake array;
Fig. 18. The influence of wind direction on the variation in the wake coefficient and Array I and Array II show the improved performance as the wake effects are reduced. A
power losses of Array I is clear. The wind direction is fixed for each simulation and fixed wind direction of 0 is used for the entire simulation and therefore the wake
a range of values from 0 to 40 are used. coefficient of the ideal array is always 1.
 F. González-Longatt et al. / Renewable Energy 39 (2012) 329e338 337

Delays in Array I increase as the wind direction changes 2000 1

between 0 and 40 and the maximum delays are reached at 45 . P
Ct
This is because the increased apparent distance between the wind Cp
turbines increases the time delays between each turbine; which in

Power (kW)

Cp and Ct
turn reduces the interactions between the wakes. The layout of 1000 0.5
Array II provides the same tower spacing along its rows as Array I,
however the spacing between each row is larger and therefore for
small changes in the wind direction the behavior of each array is
similar.
0 0
The comparison of the short-term behavior of the wake 0 5 10 15 20 25
Wind speed (m/s)
coefficient for Array I and Array II, depicted in Fig. 20 for a fixed
wind direction of 0 , shows that Array II offers superior Fig. A.1. Power curve (P), power coefficient (Cp) and thrust coefficient (CT) of 2 MW
performance. Active power losses inside Array II are higher than turbine.
Array I, due to the increased spacing between rows. However,
the maximum difference is less than 0.1% of the rated power
output.

7. Conclusion

The primary impact of the wake effect is to disturb the flow of

the wind inside a wind farm. As a consequence of a single wake or
interaction between multiple separate wakes the wind down-
stream from a wind turbine has reduced speed and becomes highly
turbulent. This reduces the energy production of the wind farm
when compared to that expected if free wind existed throughout
the farm. In this paper a simple explicit model for wakes was
presented, and used to evaluate the impact of the wake effect on
both the steady-state operation and dynamic performance of
a wind farm.
Simulations of the effect of wakes upon wind farm operation
indicate that the array efficiency depends upon the spacing
between turbines and the nature of the wind regime they are
exposed to, in terms of the forcing conditions and distribution of
wind directionality and speeds. Wind conditions depend on nature Fig. A.2. Wind speed and direction time series for long term evaluation (365 days).
and can therefore not be modified; however, the wind farm layout
can be modified to reduce the influence of the wake effect and 16

thereby optimize wind farm performance. 15.5

When attempting this modification the geometric distribu-
Wind Speed (m/s)

15
tion of wind turbines inside the wind farm is the most sensible
design parameter to adjust. Both the steady-state and dynamic 14.5

simulations performed here have demonstrated that consid- 14

ering the wake effect when developing a turbine layout can
13.5
offer considerable improvements in the power output of a wind
farm. 13
5 10 15 20 25 30 35 40 45 50 55 60
Furthermore the dynamic simulations performed have Time (s)
demonstrated that even if the wake influenced wind farm layout
Fig. A.3. Wind speed time series for short-term evaluation (60.00 s).
increases the distance between the turbines the resulting increase
in electrical losses will be minimal compared to the increase in the
power drawn from the wind. References

[1] Koch F, Gresch M, Shewarega F, Erlich I, Bachmann U. Consideration of wind

farm wake effect in power system dynamic simulation. In: Power tech, 2005
Appendix IEEE Russia; 2005. p. 1e7.
[2] Kiranoudis CT, Maroulis ZB. Effective short-cut modelling of wind park effi-
ciency. Renewable Energy 1997;11:439e57.
[3] Lissaman PBS. Energy efficiencies of arbitrary of wind turbines. Journal of
Energy 1979;3.
Table A.1 [4] Vermenulen PEJ. An experimental investigation of wind-turbines wakes. In:
Characteristics of 2 MW wind turbine. 3rd International symposium on wind energy systems, Copenhagen, Den-
mark; 1980. p. 214e23.
Description Parameter Value
[5] Milborrow DJ. The performance of arrays of wind turbines. Journal of Wind
Hub height h 78 Engineering and Industrial Aerodynamics 1980;5.
Rotor diameter r0 80 [6] Katic I, Højstrup J, Jensen NO. A simple model for cluster efficiency. In:
Area swept Awt 4536 m2 European wind energy conference and exhibition, Rome; 1986. p. 407e10.
Cut-in speed Vin 4 m/s [7] Voutsinas S, Rados K, Zervos A. On the analysis of wake effects in wind parks.
Cut-out speed Vout 25 m/s Wind Engineering 1990;14.
[8] Smith D, Taylor GJ. Further analysis of turbine wake development and inter-
Standing trust coefficient Ct,s 0.158
action data. In: 13th BWEA conference, Swansea, Wales; 1991. p. 280e92.
338 F. González-Longatt et al. / Renewable Energy 39 (2012) 329e338

[9] Zervos A, Huberson S, Hemon A. Three-dimensional free-wake calculation of [17] Barthelmie RJ, Folkerts L, Larsen GC, Rados K, Pryor SC, Frandsen ST, et al.
wind turbine wakes. In: International conference on wind farms, Leeuwarden, Comparison of wake model simulations with offshore wind turbine wake
Netherlands; 1987. p. 52e9. profiles measured by sodar. Journal of Atmospheric and Oceanic Technology
[10] Zervos A, Huberson S, Hermom A. Three-dimensional free-wake calculation of 2005;23.
wind turbine wake. Journal of Wind Engineering and Industrial Aerodynamics [18] Magnussin M, Smedsam AS. Air flow behind wind turbines. Journal of Wind
1988;27:65. Engineering and Industrial Aerodynamics 1999;80:169e89.
[11] Ainslie JF. Calculating the flow field in the wake of wind turbines. Journal of [19] Andersen PS, Krabbe U, Lundsager P, Petersen H. Basis material for wind
Wind Engineering and Industrial Aerodynamics 1988;27:213e24. turbine design. Risø National Laboratory; 1980.
[12] Mortensen NG, Hearthfield H, Landbergg H. Getting started with WASP 7.0. In: [20] Sørensen T, Thøgersen ML, Nielsen P. Adapting and calibration of existing
Risø-I-1532(EN); 2000. wake models to meet the conditions inside offshore wind farms. Aalborg:
[13] Frandsen S, Barthelmie R, Pryor S, Rathmann O, Larsen S, Højstrup J, et al. EMD International A/S; 2008. p. 53.
Analytical modeling of wind speed deficit in large offshore wind farms. Wind [21] Zhang XY, Wang WQ. Wind farm and wake effect modeling for simulation of
Energy 2006;9:39e53. a studied power system. In: 2009 IEEE/PES power systems conference and
[14] Rathmann O., Frandsen S.T., Barthelmie R.J. Wake modeling for intermediate exposition, vols. 1e3; 2009. p. 403e8.
and large wind farms. European Wind Energy Conference and Exhibition, [22] Lange B, Waldl HP, Barthelmie R, Gil A, Heinemann D. Modelling of offshore
Milan, May 2007, p 8. wind turbine wakes with the wind farm program FLaP. Wind Energy 2003;6:
[15] Jensen NO. A note on wind generator interaction. Risø National Laboratory; 87e104.
1983. [23] MATLAB, 7.8.0.347 (R2009a) ed. Natick, Massachusetts: The MathWork, Inc.;
[16] Barthelmie RJ, Hansen K, Frandsen ST, Rathmann O, Schepers JG, Schlez W, 2009.
et al. Modeling and measuring flow and wind turbine wakes in large wind [24] DigSILENT PowerFactory, 14.0.519.1 ed. Gomaringen, Germany: DIgSILENT
farms offshore. Wind Energy 2009;12:431e44. GmbH; 2010.