10991824, 2010, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/we.380 by Universidad De Los Andes, Wiley Online Library on [26/11/2022].

See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
WIND ENERGY
Wind Energ. 2010; 13:559–572
Published online 15 December 2009 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/we.380

RESEARCH ARTICLE

Application of a LES technique to characterize the

wake deflection of a wind turbine in yaw
Ángel Jiménez, Antonio Crespo and Emilio Migoya
Laboratorio de Mecánica de Fluidos, Departamento de Ingeniería Energética y Fluidomecánica, ETSII, Universidad Politécnica de
Madrid, José Gutiérrez Abascal, 2 CP 28006, Madrid, Spain

ABSTRACT
When a wind turbine works in yaw, the wake intensity and the power production of the turbine become slightly smaller
and a deflection of the wake is induced. Therefore, a good understanding of this effect would allow an active control of
the yaw angle of upstream turbines to steer the wake away from downstream machines, reducing its effect on them. In
wind farms where interaction between turbines is significant, it is of interest to maximize the power output from the wind
farm as a whole and to reduce fatigue loads on downstream turbines due to the increase of turbulence intensity in wakes.
A large eddy simulation model with particular wind boundary conditions has been used recently to simulate and charac-
terize the turbulence generated by the presence of a wind turbine and its evolution downstream the machine. The simpli-
fied turbine is placed within an environment in which relevant flow properties like wind speed profile, turbulence
intensity and the anisotropy of turbulence are found to be similar to the ones of the neutral atmosphere. In this work, the
model is used to characterize the wake deflection for a range of yaw angles and thrust coefficients of the turbine. The
results are compared with experimental data obtained by other authors with a particle image velocimetry technique from
wind tunnel experiments. Also, a comparison with simple analytical correlations is carried out. Copyright © 2009 John
Wiley & Sons, Ltd.
KEYWORDS
wind turbine wake; wind turbine in yaw; large-eddy simulation; wake deflection; active control in wind farms

Correspondence
Á. Jiménez, Laboratorio de Mecánica de Fluidos, Departamento de Ingeniería Energética y Fluidomecánica, ETSII, Universidad
Politécnica de Madrid, José Gutiérrez Abascal, 2 CP 28006, Madrid, Spain
E-mail: a.jimenez@upm.es

Received 17 October 2008; Revised 6 October 2009; Accepted 20 October 2009

1. INTRODUCTION
The spectacular growth of wind energy production in
recent years has turned this renewable resource in a rec-
ognized alternative to more conventional sources of elec-
trical power generation. To reduce some of the costs of
wind power utilization, such as land and civil works, tur-
bines are usually grouped in wind farms, tending to be
located close to each other creating interference effects. A
momentum defect and an increased level of turbulence
intensity are induced in wakes formed by turbines, causing
a reduction in the power output of the wind farm and
unsteady loads on other machines. Specifically, turbulence
may increase in turbine arrays sufficiently to cause mea-
surable damage due to fatigue and dynamic loads that have
to be taken into account adequately; see Frandsen and
Thögersen.1 Suitable design criteria have to be developed
for the inter-spacing of wind turbines to minimize these
undesirable effects.
Different approaches, reviewed by Crespo et al.2 and
Vermeer et al.,3 have been adopted to study wind turbine
wakes. A large eddy simulation (LES) methodology for
studying wind turbine wakes is used in this work. It was
previously presented by Jiménez et al.4 as a continuation
of the Reynolds Average Navier Stokes (RANS) models
based on the k-e closure methods proposed by Crespo
et al.5 and Crespo and Hernández,6 and the explicit alge-
braic model for the components of the turbulent stress

Copyright © 2009 John Wiley & Sons, Ltd. 559

 10991824, 2010, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/we.380 by Universidad De Los Andes, Wiley Online Library on [26/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wake deflection of a wind turbine in yaw Á. Jiménez, A. Crespo and E. Migoya

tensor proposed by Gómez-Elvira et al.7 In all these previ-

ous methods, the Reynolds average over all turbulence
scales is imposed. Instead, LES will reproduce the unsteady
oscillations of the flow characteristics over all scales larger
than the grid size; consequently, a greater detail of the
turbulence characteristics is expected to be obtained. In
Jiménez et al.,4 a LES computation of the wake was per-
formed immersing an actuator disk-modelled turbine in an
environment with turbulence properties similar to the ones
of the atmosphere. A similar technique is used by other
authors like Masson,8 Kasmi and Mason,9 etc. Together
with the actuator disk approach, it must be highlighted the
actuator line representation given by Sørensen and Shen,10
recently used by Troldborg et al.11 to carry out a detailed
LES analysis of a wind turbine wake under uniform inflow
conditions.
Jiménez et al.4 gave a comparison of LES results with
experimental data obtained by Cleijne12 from the Sex-
bierum wind farm and with analytical correlations pre-
viously proposed by Crespo and Hernández13 and
Gomez-Elvira et al.7 In the present work, application of
the same technique to study the steady wake deflection of
a wind turbine in yaw is made.
Control of turbine parameters in wind farms has been
Figure 1. Recreation of the wake deflection due to a wind
suggested as a method to increase the power production of
turbine in yaw.
the whole wind farm and to reduce fatigue loads due to the
high level of turbulence in wakes. Corten and Shaak14
skew angle of the wake of a yawed turbine. Concerning
proposed a strategy based on decreasing the axial induc-
the interaction between machines, it is more relevant the
tion factor (through pitch angle control) at the upwind side
experimental study done by Parkin et al.18 They obtained
of the wind farm in order to get a higher wind speed in the
PIV images in the wake of a two bladed HAWT at dis-
wake and, consequently, a larger amount of available
tances from one to five diameters downstream, for differ-
kinetic energy for the turbines under the lee side. The
ent values of the yaw angle. Their PIV experiments were
research presented in this paper can give an illustration of
carried out with a two-bladed model wind turbine in a low
future possibilities to plan an active control to minimize
turbulence wind tunnel at Kungliga Tekniska Högskolan
the interference effects in wind farms, now based on the
(KTH), Sweden. Model turbines may not behave as full-
yaw of wind turbines.
sized ones and, consequently, it would be of interest to rely
When a wind turbine works in yaw, the wake intensity
also on experimental measurements obtained on field.
and the power production of the turbine become slightly
Unfortunately, data about full-sized turbines operating in
smaller and a deflection of the wake is induced. A suffi-
yaw conditions are scarce.
ciently good understanding of this effect would allow an
In this work, a preliminary analysis of wakes of wind
active control of the yaw angle of upstream turbines to
turbines in yaw is presented. The wake deflection and
steer the wake away from downstream machines, as illus-
trajectories are studied and compared to a simple analytical
trated in Figure 1, reducing its effect on them and giving
model and with experimental results.
as consequence an optimization of the power output from
the wind farm as a whole. Also a reduction of fatigue loads
on downstream turbines due to a lower increase of turbu-
lence intensity in wakes is achieved. However, possibly 2. LES MODEL WITH SIMPLIFIED
this also increases the fatigue in the first turbine, since BOUNDARY CONDITIONS
yawing itself may cause fatigue; accordingly, it should be
quantified if the net result is favourable. Only a direct simulation of turbulence would be able to
The gross wake deflection in yaw was shown by Clayton give us the full knowledge of the turbulence characteristics
and Filby15 who performed hot-wired measurements in the in the wake. In industrial or environmental applications,
wake of a wind turbine at a number of downstream posi- where Reynolds numbers are usually very high, direct-
tions. Particle image velocimetry (PIV) technique was numerical simulations (DNS) of turbulence are generally
used by Grant et al.16 and Grant and Parkin17 enabling a impossible because the very wide range that exists between
detailed understanding of vortex formation and expansion the largest and the smallest turbulent scales cannot be
phenomena in the near wake, both in yawed and non- explicitly simulated, even in the most powerful computers.
yawed conditions, and giving information about the initial Furthermore, DNS is not feasible near rough boundaries

Wind Energ. 2010; 13:559–572 © 2009 John Wiley & Sons, Ltd.
560
DOI: 10.1002/we

Á. Jiménez, A. Crespo and E. Migoya
when the exact geometry of roughness cannot be described
in detail. A LES approach is then conceptually more suit-
able for studying the turbulence evolution downstream of
a wind turbine, because the larger scales of turbulent
motion, those that control turbulent diffusion of momen-
tum and energy, are computed explicitly, whereas only the
effect of the small-scale motion that tends to be more
isotropic and dissipative has to be modelled.
A LES model with periodic boundary conditions has
been implemented in a CFD code to simulate and char-
acterize the turbulence generated in a wind turbine wake.
This model was previously employed by Jiménez et al.4
to obtain important turbulent properties of the flow in the
wake, mainly the velocity profiles and the components of
the Reynolds stress tensor, finding a good agreement with
experimental data. The same model was applied in
Jiménez et al.19 to the study of the characteristics of spec-
tral coherence of turbulence fluctuations in a wind turbine
wake. In this work, the same methodology is proposed to
characterize the wake deflection of a wind turbine in
yaw.
We will denote the filtered field variables without any
additional notation (usually a bar over variables is
employed for this purpose). Then, in this paper, ui repre-
sents the components of the instantaneous filtered veloc-
ity. Time and space (over homogeneous directions)
averaging is indicated with the symbol 〈·〉, and the dif-
ference between the instantaneous filtered value of any
variable and its average is denoted by an apostrophe,
‘: ui = 〈ui〉 + ui’.
The subgrid-scale (SGS) technique used is the
classical of Smagorinsky,20 with the dynamical calcula-
tion of the constant developed by Germano et al.21
The SGS stresses are found to be negligible compared
with the resolved ones: of the order of
⎛ ∂u ∂u j ⎞
τ ij(SGS ) = vT ⎜ i + ≈ 0.01 ρ u′iu′ j , where vT
⎝ ∂x j ∂xi ⎟⎠
represents the eddy viscosity given by the SGS model. This
is accomplish in the whole computational domain, except
for the component txy in a very small region close to the
walls, since at them, uy = 0 is imposed as boundary condi-
tion, and then u′y = 0 at the walls at any time. Thus, it is
justified to assume that most of the energy and momentum
transport rate are contained in the resolved fluctuations,
and only the filtered values are considered in the following
discussion.
The Reynolds number does not any significant influence
on the problem, as surface roughness is clearly dominant,
and molecular viscous effects are negligible compared to
those due to eddy viscosity, even in the proximity of the
ground (the bottom wall of the domain).
3. NUMERICAL SCHEME
The numerical method used to solve LES equations for
incompressible flow is a semi-implicit iterative solution
Wind Energ. 2010; 13:559–572 © 2009 John Wiley & Sons, Ltd.
DOI: 10.1002/we
10991824, 2010, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/we.380 by Universidad De Los Andes, Wiley Online Library on [26/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wake deflection of a wind turbine in yaw
procedure that is economical and simultaneously retains
most of the stability and accuracy properties of the fully
implicit scheme. It was developed by Charles D. Pierce at
Center for Turbulence Research, Stanford University, and
is conveniently described in Pierce and Moin22 and Wall
et al.23 The most important numerical properties of the
scheme are the following ones:
• Staggered Cartesian grid is used: Velocity compo-
nents are evaluated in the centre of the cell faces and
pressure is in the cell centres.
• Second order accuracy in space and time with a
central difference discretization scheme.
• A semi-implicit iterative solution procedure of the
discrete equations (implicit in y direction for the
advective and diffusive terms) is used.
• Simulations run at constant Courant-Friedrich-Lewy
number (CFL) of 0.75 in accordance with the stabil-
ity properties of the semi-implicit scheme, see Pierce
and Moin22 and Wall et al.23
Our calculations have been performed with a grid of
256 × 96 × 96 points in x, y and z directions, respectively,
which, considering the size of the computational domain
described below, ensures a good quality in the shape of the
cells. By ‘good quality’, we mean that the three sides of
the Cartesian cells are not highly unequal, that would be
undesirable for the filtering operation involved in LES
calculation of eddy viscosity. In the vertical direction, the
grid points are non-uniformly distributed; particularly,
they have been clustered around the turbine. The first grid
point is placed at a distance of 0.018 h over the ground,
where h is the wind turbine hub height. Grid convergence
has been verified by checking that time average results do
not change significantly with an increase of the number of
grid points. In addition, the resolution of the grid and its
ability to resolve the flow in the near and far wake regions
(not in the extended region right behind the wind turbine
rotor) was tested in previous works,4,19 where predictions
of LES calculations were compared with experimental
measurements, finding an acceptable agreement.
4. FLOW MODEL
The computational domain is a rectangular channel with
the shape of a parallelepiped whose sides have lengths
Lx = 34.9 D, Ly = 5.6 D and Lz = 10.7 D. Inside the box,
there is the wind turbine, which is simulated by several
momentum sinks. On the side walls of the channel, z =
±Lz/2, periodic boundary conditions are applied. The
mathematical implication of periodic boundary conditions
consist in the equality of the value of f, fz=−Lz/2 = fz=+Lz/2,
and also of its derivative, (∂f/∂z)z=−Lz/2 = (∂f/∂z)z=+Lz/2,
where function f represents the three velocity components
and the pressure. For the upper wall, a boundary condition
giving a fixed value of the shear stress is imposed.
561
 10991824, 2010, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/we.380 by Universidad De Los Andes, Wiley Online Library on [26/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wake deflection of a wind turbine in yaw Á. Jiménez, A. Crespo and E. Migoya

Periodic boundary conditions are also applied at the 5. BOUNDARY CONDITIONS

entrance plane, x = 0, and the exit, x = Lx. Periodic bound-
ary conditions guarantee that in the absence of wind tur- The boundary condition in the bottom side of the compu-
bines, the turbulent flow in which the wake diffuses is tational domain, y = 0, is expressed by the following equa-
independent on the x coordinate; this cannot be accom- tions, frequently used in meteorology:
plished in general by imposing Dirichlet inlet conditions.
Thus, the flow field that is represented corresponds to a uy = 0 (2)
whole cascade of infinite number of wind turbines, sepa-
rated the distance Lx and Lz. It could be argued that if the ∂ui ( x, y, z )
ρ ( v + vT ) = τ iy ( x, z ) (i = x, z ) (3)
interaction is, in fact, equivalent to that of an array of an ∂y
infinite number of turbines, the periodic boundary condi-
tions will work appropriately for the analysis of an iso- where
lated wind turbine wake only if the wake does not impinge
on the boundaries, in order to recover an unperturbed κ2
τ iy ( x, z ) = ρ 2
uh ( x, z ) ui ( x, d , z ) (i = x, z )
wind field at the domain limits. If the wake hit any bound- ⎡ ⎛ d ⎞⎤
ln ⎜ ⎟
⎣⎢ ⎝ y0 ⎠ ⎦⎥
ary, it would just reappear at the corresponding periodic
one, so then the condition of unperturbed incident flow
over the wind turbine will never be satisfied unless the (4)
wake effect disappears inside the domain without reach-
ing the boundaries. Calculations have been carried out and
both, with and without turbines, and it is found that the
lengths Lx and Lz are large enough so that the flow field uh ( x, z ) = ux ( x, d , z )2 + uz ( x, d , z )2 (5)
incident over the wind turbine is equivalent to the unper-
turbed flow without any turbine (with regard to velocity These equations are proposed by Senocak et al.24 In expres-
profile and turbulence intensity) and then the periodic sion (4) tiy are the instantaneous global shear stresses and
boundary conditions do not result in an alteration or skew d is the distance from the first node of the grid to the wall.
of the inflow conditions. Very similar approaches are used by Moeng25 and by Pio-
It is expected that this incident flow field will have the melli and Balaras.26 Note that relation (4) is consistent with
characteristics of a real atmospheric flow. In particular, we the law of the wall, equation (1), but it should be regarded
expect that the average flow velocity in the x direction will that, in the latter, the average values of velocity and shear
be given by the classical law of the wall, which in the case stress take part, whereas in condition (4) we are operating
of rough wall is expressed as: with instantaneous and local values.
At the top boundary of the domain, y = Ly, we adopt:
u* y
ux = ln (1) uy = 0 (6)
κ y0

where u* is the friction velocity, k is the von Karman ∂u x

ρ ( v + vT ) = ρu*2 (7)
constant of value 0.4 and y0 represents the terrain rough- ∂y
ness. We will check later whether this law of the wall is
really satisfied. ∂ uz
=0 (8)
The constant CFL condition allows a slight variation of ∂y
the time step, that leads to an average value of Δt ≈
0.035 D/U0, where U0 = 〈ux〉y=h is the average wind speed where u* is previously calculated from the results of equa-
at hub height. The averaging period of time to obtain a tion (4) according to its classical definition in the law of
sufficiently converged solution is around 5·104 time steps, the wall:
what, considering the longitudinal size of the domain, rep-
resents approximately 50 flow-through times at hub height ρu*2 = τ xy ( x, 0, z ) (9)
(50 Lx/U0). Before beginning to compute the statistical
data, simulation was maintained for 104 time steps, in order In other words, the value of the shear stress obtained in
to attain a turbulent-steady flow situation. the bottom boundary is imposed at the top boundary. In
The information about the dimension of the problem is both positions, the global stress is equal to the viscous plus
summarized in a dimensionless form in Table I. the subgrid stresses, since the resolved one is null there,

Table I. Summary of problem dimensions.

D/h y0/h Lx/h Ly/h Lz/h u*/U0 ΔtU0/h Av. time
−3
0.85714 1.4286·10 29.920 4.7619 9.1429 0.061956 ≈0.03 ≈5·104Δt

Wind Energ. 2010; 13:559–572 © 2009 John Wiley & Sons, Ltd.
562
DOI: 10.1002/we

Á. Jiménez, A. Crespo and E. Migoya
as previously discussed. The resolved stress grows up very
quickly as one moves away from the upper and lower
boundaries, and for the major part of the domain, the
resolved stress is the only meaningful stress term. As a
consequence of the force balance over the flow, a steady
situation (in a turbulent sense) is achieved, and a com-
pletely developed turbulent flow with uniform shear, is
obtained. In the case without a wake, this would be equiv-
alent to a Couette stream.
Equation (7) represents a condition in the velocity
derivative. The SGS eddy viscosity varies with time and,
consequently, the velocity gradient at each point of the top
boundary also has to change, in order to achieve the correct
shear.
These boundary conditions guarantee, in the regions
that are not affected by the wake, a uniform value (no y-
variation) of the average global shear stress.
6. COMPARISON OF THE
CHARACTERISTICS OF THE
INCIDENT FLOW OVER THE
TURBINE WITH THOSE OF THE
ATMOSPHERIC FLOW
To check if the flow model described above reproduces in
absence of a wake the atmospheric turbulence parameters,
we performed a set of channel simulations without wind
turbines, getting a fully developed Couette stream flow.
Figure 2. Incident flow over the turbine. (a) Velocity profile and law of the wall. (b) Reynolds stress components profiles
compared with those measured at Sexbierum wind farm.
Wind Energ. 2010; 13:559–572 © 2009 John Wiley & Sons, Ltd.
DOI: 10.1002/we
10991824, 2010, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/we.380 by Universidad De Los Andes, Wiley Online Library on [26/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wake deflection of a wind turbine in yaw
Then, a comparison of the turbulence results with the
atmospheric values was made. In this case, mean flow
properties are homogeneous in the x and z directions, and
they are obtained averaging over all points in the xz-plane
and over a time period of 1750 D/U0 (approximately 50
flow-through times).
The calculated velocity profile obtained in the channel
simulation is given in Figure 2(a), and it is compared with
the logarithmic law expressed by equation (1). The agree-
ment is good in the zone where the wind turbine is placed.
Our wind turbine hub height has been set to 700 y0, and
the rotor diameter value is 600 y0 (e.g. y0 = 0.1 m, h = 70 m,
D = 60 m). Then, the rotor takes up a region going from
y = 400 y0 to y = 1000 y0. The largest deviation from the
theoretical logarithmic profile in this zone is achieved at
the bottom part of the rotor, position y = h − D/2, where
we have recorded a relative error of 4.39%. In the upper
part of the domain, the calculated value of the velocity is
larger than the one given by the law of the wall. This is
due to the boundary conditions (6) to (8) imposed in the
upper side of the computational domain. With other bound-
ary conditions, the velocity profile could be closer to the
logarithmic one, but the components of the atmospheric
turbulent stress tensor would be poorly reproduced. In
Figure 2(b), the distributions of all the calculated compo-
nents of the resolved stress tensor 〈tij〉 = −r〈ui′uj′〉 are
shown. The components 〈txz〉 and 〈tyz〉 are zero because of
symmetry, and as consequence of the boundary conditions
imposed, equations (4) to (9), the mean turbulent shear
563
 10991824, 2010, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/we.380 by Universidad De Los Andes, Wiley Online Library on [26/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wake deflection of a wind turbine in yaw Á. Jiménez, A. Crespo and E. Migoya

stress of the incident flow should be −r〈ui′uj′〉 = ru*2, • Turbulent kinetic energy
which is in excellent agreement with the calculated value. • Anisotropy (distribution of the kinetic energy
The three diagonal components are somewhat smaller than between the three diagonal components of Reynolds
the values estimated by Panofsky and Dutton,27 from a set stress tensor)
of experimental results, which are also indicated in inter-
national standards:28
7. WIND TURBINE MODEL
τ xx = −2.402 ρu*2 (10)
A simplified model of a wind turbine, which can be
τ yy = −1.252 ρu*2 (11) included in the sort of models known as actuator disk
models, is introduced inside the wind flow. The wind
τ zz = −1.922 ρu*2 (12) turbine rotor is approximated by a set of rectangular cells
in a Cartesian grid, which all together configure a disk
However, the numerically predicted values of the three
(Figure 3). We then introduce in those cells a momentum
diagonal components are quite close to those measured in
sink that guarantees that the extraction of momentum by
the positions away from wakes in the Sexbierum wind
the whole disk is equal to the one predicted by the actuator
farm, reported by Cleijne,12 being representative of the
disk theory. For a turbine in yaw, the rectangular cells will
unperturbed atmospheric turbulence in this site (see Figure
not be in a plane normal to x-axis, as will be discussed
2(b)). This wind farm placed in the north of the Nether-
afterwards. Such a model of a wind turbine is like a porous
lands has a total number of 18 turbines of 300 kW rated
disk, which has occasionally been used to simulate wind
power with a rotor diameter of D = 30 m and a hub height
turbines. On the actual resolution of the rotor disk, it is
of h = 35 m. Data shown in Figure 2 is referred only to the
composed by a total number of 167 cells in the correspond-
unperturbed flow, but Sexbierum measurements include
ing yz-plane. The whole force is distributed in a volume
both mean velocity profiles and Reynolds stresses distribu-
three cells thick in x-direction to avoid the introduction of
tions obtained al various positions in wakes. These exper-
a big instability on the flow.
imental data were chosen by Jiménez et al.4 to validate the
In the present study, the real blades of the turbine are
accuracy of the code and the grid resolution used. Figure
not taken into account because we are not really interested
2 allows us to conclude that the turbulent properties of the
in the local properties of the flow directly interacting with
simulated wind flow are satisfactorily close to those of the
the wind turbine blades, but in the downstream evolution
atmospheric boundary layer in the following aspects:
of the flow characteristics. Larsen et al.29 present the
• Velocity profile results from 3D CFD computations, indicating that the
• Uniform shear stress near wake field may be separated in two regions, according

Figure 3. Schematic view of the computational domain.

Wind Energ. 2010; 13:559–572 © 2009 John Wiley & Sons, Ltd.
564
DOI: 10.1002/we
 10991824, 2010, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/we.380 by Universidad De Los Andes, Wiley Online Library on [26/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Á. Jiménez, A. Crespo and E. Migoya Wake deflection of a wind turbine in yaw

to whether or not the flow structures around the turbine

blades have developed to an azimuth uniform stage. The
extent of the inner region (in which blade flow structures
can be identified) is only of the order one rotor diameter.
Instead, since our purpose is to study the possibility of
interaction effect between different turbines, we are clearly
and specifically interested in the characterization of the
turbulence generated in the annular shear layer around the
wake at bigger distances downstream the turbine than one
diameter, as this is the main contribution to the whole
turbulence intensity that a wind turbine sited downstream
would see. Figure 4. Body forces introduced by the rotor in a yawed
In absence of yaw, the volume force introduced in the wind turbine according to the actuator disk theory.
disk cells is proportional to the area of each cell, giving a
global introduced force on the flow of:

1
f x = −CT ⋅ ρ AU 02 (13)
2 8. SIMPLE ANALYTICAL MODEL
TO EVALUATE THE WAKE
CT represents the thrust coefficient of the wind turbine, A
DEFLECTION FOR A TURBINE
is the frontal area of the rotor disk and U0 is the average
NON-ALIGNED WITH THE
velocity at the disk centre position of the incident unper-
INCIDENT WIND
turbed flow. It may look more appropriate to use the local
velocity immediately upstream of the turbine; however, The model is shown schematically in Figure 5. The turbine
classically, the thrust coefficient is referred to the unper- has a yaw angle q relative to the incident wind mean direc-
turbed mean wind velocity, not considering its turbulent tion. A simple top hat model is used, assuming that inside
variation. The minus sign means that the force has the a cross-section of the wake, the velocity is uniform and
opposite direction to the mean flow’s one and, therefore, inclined an angle a with respect to the ambient velocity.
decelerates the incident stream. That inclination is also what we call the skew angle. We
When the wind turbine axis is not aligned with the wind denote by d the wake width. Taking a control volume
direction, only the normal velocity component crosses the sufficiently far from the turbine so that the pressure
rotor. The force over the turbine is supposed to be due only field is in first approximation uniform and applying the
to the velocity component perpendicular to it and its direc- momentum conservation, the following relationships are
tion is supposed to be perpendicular to the turbine plane obtained for the force exerted by the turbine on the flow
as well. Then, according to the actuator disk theory and (see Appendix):
ignoring any possible redistribution of the load and induc-
tion over the disk due to the yaw situation, a new momen- πδ 2
tum deficit in that direction is induced as a result (Figure f x ≈ − ρU 0 ΔU (17)
4
4). Then, equation (13) must be replaced by:
πδ 2
1
f = CT ⋅ ρ A (U 0 cosθ )2 (14) fz ≈ − ρU 02 α (18)
2 4

The projection of this body force into the coordinate where the first equation corresponds to the projection in
system directions implies a diminution of the x-component the incident wind direction, and the second one is the
and the apparition of a z-component, as it is shown in projection in the other horizontal direction perpendicular
Figure 4: to it. ΔU is the velocity deficit in the wake. Additional
assumptions to derive (17) and (18) are that the error of
1 considering U0 − ΔU ≈ U0 is not important, and a should
f x = − f cosθ = −CT ⋅ ρ A (U 0 cosθ )2 cosθ (15)
2 be small enough to approximate cos a ≈ 1 and sin a ≈ a.
The value of a can be obtained as a function of d compar-
1 ing equations (18) and (16):
fz = − f sin θ = −CT ⋅ ρ A (U 0 cosθ )2 sin θ (16)
2
(δ )
2
D CT
α≈ cos2 θ sin θ (19)
The body force finally introduced in the cells configuring 2
the rotor disk is given by expressions (15) and (16). Of
course, the force exerted by the wind flow over the turbine Also ΔU could be estimated from (15) and (17). For the
is then the opposite of this one. deflection angle just behind the turbine, what we call the

Wind Energ. 2010; 13:559–572 © 2009 John Wiley & Sons, Ltd.
565
DOI: 10.1002/we
 10991824, 2010, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/we.380 by Universidad De Los Andes, Wiley Online Library on [26/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wake deflection of a wind turbine in yaw Á. Jiménez, A. Crespo and E. Migoya

Figure 5. Schematic showing of the model.

initial skew angle of the wake, it may be assumed that

d ≈ D, so that:

CT
α x =0 ≈ cos2 θ sin θ (20)
2

Otherwise, if the wake cross-section can be supposed to

increase linearly with downstream distance x, so that
d = D + bx, then expression (19) leads to:

CT
cos2 θ sin θ
α≈ 2 (21)
( )
2
x
1+ β
D

However, this is only correct for the far wake, where the
wake spread is dominated by the ambient turbulence.
Besides, it is not clear whether the usual values of b found
in the literature (see, for example, Lissaman,30 Vermeulen,31
Katic et al.32 and Voutsinas et al.33) will be valid in this case,
where there will be a strong entrainment in the wake due to
the ambient velocity component normal to the wake, U0 sin
a, as shown by Escudier34 and Servert et al.35 For this par-
ticular case, a reasonable agreement with the results of the
LES model has been found, taking for b a range of values
going from 0.09 to 0.125 as we will see later.
9. RESULTS OF CALCULATIONS
AND COMPARISON WITH THE
ANALYTICAL MODEL
We present a parametric study of the skew angle of the
wake a, analysing the effect of the two major governing
parameters: the thrust coefficient CT and the yaw angle of
the wind turbine q. We performed a set of LES calcula-
tions for a range of CT going from 0.7 to 0.9 and for a
range of q from 10° to 30°. Figure 6 gives a snapshot in
time showing the vortex structure of the wake generated
by the actuator disk model in a horizontal plane at hub
Figure 6. Instantaneous non-dimensional streamwise velocity
field, ux/U0 in the wake in a horizontal plane containing the
wind turbine hub for CT = 0.75 and yaw angle of 0°.
height for a yaw angle of 0° and CT = 0.75. In Figure 7,
we can see the general behaviour of the wake when
increasing the yaw angle progressively. Time-averaged
velocity contours (in the presence of wake, the x and z
axis are no longer homogeneous directions, so xz-planes
average is not possible here) are presented. As previously
indicated, the averaging time for these plots is about 50
flow-through times for the domain. It is clear that for
bigger yaw angles, a bigger skew angle is achieved, but
the difference is smaller as the yaw angle increases.
According to equation (20), the skew angle would reach
its maximum value for a yaw angle of qmax = 35.26°, which
means that it has not much sense to think of a wind turbine
working with yaw angles greater than 30°. This was also
noticed by Parkin et al.,18 whose experimental results
showed that the skew angle curve flattens off for large yaw
angles, giving a minimal gain in deflection by going to
higher yaw angles beyond 30°. In addition, as the yaw
angle increases, a lower quantity of momentum is extracted
from the airflow; as can be seen in Figure 7, the wake
width looks smaller and the velocity defect tends to vanish

Wind Energ. 2010; 13:559–572 © 2009 John Wiley & Sons, Ltd.
566
DOI: 10.1002/we

Á. Jiménez, A. Crespo and E. Migoya
Figure 7. Non-dimensional averaged-velocity contours, |u | /U0, in a horizontal plane containing the wind turbine hub for CT = 0.8
and yaw angles of 0, 10, 20 and 30°.
at a shorter distance. This effect is not so significative for
the smallest yaw angles, but quite important for a yaw
angle of 30°, as shown in Figure 8, where some time-aver-
aged velocity profiles have been plotted. This is related to
the cubic dependence of volume forces in x direction with
cos q, equation (15). This quicker decay could be also
of interest for the study on interaction effects between
turbines.
Figure 9 gives a comparison between the results obtained
with the LES calculation for the initial skew angle of the
wake and the analytical model described above equation
(20). In the simulations, the initial skew angle is deter-
mined as the angle formed by the mean flow direction and
Wind Energ. 2010; 13:559–572 © 2009 John Wiley & Sons, Ltd.
DOI: 10.1002/we
10991824, 2010, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/we.380 by Universidad De Los Andes, Wiley Online Library on [26/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wake deflection of a wind turbine in yaw
the segment joining the centre of the turbine rotor with the
centre of the wake at the very short downstream distance
of 0.5D. The method to find the centre of the wake at any
downstream position is explained in the next section. We
can see that the agreement is quite good for the ranges of
yaw angles and thrust coefficients studied. LES results
showed in Figure 9 are not easily comparable with the
measurements obtained by Parkin et al.18 They carried out
a PIV experiment with a 0.25 m diameter two-bladed
model wind turbine in a low turbulence wind tunnel. There
is no information about the value of the wind turbine thrust
coefficient. Otherwise, the wind turbine was run at is
optimum tip speed ratio for a free-stream wind velocity of
567
 10991824, 2010, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/we.380 by Universidad De Los Andes, Wiley Online Library on [26/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wake deflection of a wind turbine in yaw Á. Jiménez, A. Crespo and E. Migoya

Figure 8. Non-dimensional averaged streamwise velocity profiles, ux/U0, at hub height and several distances from turbine for
CT = 0.8. (a) x/D = 2.5; (b) x/D = 5.5; (c) x/D = 8.

Figure 9. Comparison of the initial skew angle of the wake between LES calculation and the analytical model given
by equation (20).

Table II. Initial skew angle of the wake, α⎪x = 0. 10. THE TRAJECTORY OF THE WAKE
q 10° 20° 30°

Parkin et al.18 2.4° 3.9° 4.5°

Equation (20) 1.59° 2.86°
6.3 m s−1, which leads the turbine to achieve a power coef-
ficient value of about 0.3. A quick calculation based on the
actuator disk theory allows us to obtain the induced veloc-
ity factor a ≈ 0.09, which would give a result of CT ≈ 0.33
for the thrust coefficient, which is far from the typical
values of a commercial wind turbine generating power in
a wind farm. However, this is only a rough estimation; it
is of interest to compare their experimental results with the
predictions of the previous analytical correlation. The
comparison of the initial skew angle is presented in Table
II. In all three cases, the initial skew angle of the wake
seems to be under-predicted in about one degree by the
model.
The agreement would be better if we adopt the value
CT ≈ 0.45. The uncertainty in the real value of CT makes it
difficult to know the real accuracy of the model in this
case.
We calculate the centre of the wake, at each downstream
3.55°
distance from the turbine, by taking the midpoint between
the two positions with wind speed equal to the 95% of the
free-stream velocity, following the same procedure as
Parkin et al.18 As such, we can follow the trajectory of the
wake axis. The trajectories of the wake calculated through
the LES procedure described above in the nine cases under
study are shown in Figure 10. With this information, we
would be able to check if an active control of the yaw angle
of turbines is of interest in each case or it is not. The
separation of the wake centre line from the non-yawed
(symmetric) case line is not so big even for the larger thrust
coefficient and yaw angles. For example, if the second
turbine would be positioned five diameters directly down-
stream from the first one (let’s suppose it with CT = 0.8) a
yaw angle of 30° would only steer the wake away about
one half of a diameter (see Figure 10), causing a reduction
in power in the first turbine, which in the worst case
(assuming no regulation in power, so it would be propor-
tional to the cube of the normal wind speed, and then to
cos330° ≈ 0.65) could reach the 35%. Then, active control

Wind Energ. 2010; 13:559–572 © 2009 John Wiley & Sons, Ltd.
568
DOI: 10.1002/we

Á. Jiménez, A. Crespo and E. Migoya
Figure 10. Trajectories of the wake centre obtained with LES for the nine cases under study.
based in the first turbine yawing may be probably undesir-
able. Now, we can imagine a completely different situa-
tion: if the downstream turbine would be sited at span
position of z/D = +1, taking into account the wake engross-
ment (the simple formula given previously would give
d/2 ≈ D/2 + 0.25 D at this position for the b values that we
will see below), a reduction in power will be expected in
the second turbine. In addition, the half wake situation that
may result on the downstream turbine may be very adverse
with respect to unsteady loads. Then, the upstream turbine
could only be yawed by 10°, pushing the wake completely
away from the second turbine, as an additional separation
of about −0.3 D is obtained (see Figure 10), compensating
the extra +0.25 D due to wake expansion. With a yaw
angle of 10°, at least 96% of the available energy would
be extracted, and active controlling may be an interesting
option.
Wind Energ. 2010; 13:559–572 © 2009 John Wiley & Sons, Ltd.
DOI: 10.1002/we
10991824, 2010, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/we.380 by Universidad De Los Andes, Wiley Online Library on [26/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wake deflection of a wind turbine in yaw
As clearly shown in Figures 8 and 10, the trajectory of
the wake centre experiences a deflection downstream the
wind turbine. The wake angle, determined in the LES
computation as the angle formed by the mean flow direc-
tion and the tangent to the trajectory of the wake centre at
any position, decays with the distance and the trajectory
tends to become parallel to the main flow direction. In
order to validate the model for the skew angle given by
equation (21) and to get the best value of the parameter b,
we make a comparison between the model predictions and
the results of our LES calculations. This is shown in Figure
11 for the three yaw angles considered with CT = 0.8. In
the three cases, a qualitative agreement between calcula-
tions exists when using the values 0.1, 0.1 and 0.125,
respectively (for yaw angles 10°, 20° and 30°), for b. The
agreement is slightly better for the lowest yaw angle. For
the other values of CT, a very similar situation is found,
569

Wake deflection of a wind turbine in yaw
Figure 11. Variation of the skew angle of the wake with the distance with the wind turbine.
and in all cases, the optimum b is the range 0.09–0.125.
The best region to perform the optimum fit for b ends at
a distance of about 8 D from the wind turbine. For higher
distances, the velocity defect in the wake has decayed from
its initial value and the wake width has increased. Then, it
is not possible to determine the wake centre position with
the same precision through the method described above,
and some uncertainty in the skew angle is unavoidable, as
it is revealed by the oscillations observed in curves pre-
sented in Figure 11 for high distances. In any case, the
wake skew angles for those far distances are very small.
11. CONCLUSIONS AND
FUTURE WORK
In this work, a preliminary analysis of wakes of wind
turbines in yaw is presented. The wake deflection and
trajectories are studied and compared to a simple analytical
model and with experimental results. The agreement with
both experiments and the analytical model is satisfactory
in order to predict the wake trajectory deviation. Although
no definitive conclusions can yet be drawn and more
experimental work is needed, a new possibility of active
wake controlling based in these simple analytical correla-
tions is proposed.
In future work, a more detailed analysis of the near
wake aerodynamics is needed to check whether the
assumptions made about the magnitude and direction of
the thrust force are correct.
The analysis will also be extended to include the study
of turbulence characteristics that are really important to
see how the downstream machines are affected if the
upstream one is in yaw. Specifically, the effect of yaw on
spectra, spectral coherence, meandering, turbulent lengths
and turbulence intensities will be analysed, as in Jiménez
et al.4,19
The approach presented here can also be of interest to
study wake meandering by using a quasi-steady model of
wakes considering an oscillating incident wind.
570
10991824, 2010, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/we.380 by Universidad De Los Andes, Wiley Online Library on [26/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Á. Jiménez, A. Crespo and E. Migoya
Also, the UPMWAKE6 and UPMPARK13 models
developed at our laboratory, which are less accurate than
the one presented here but less time-consuming, will be
adapted to include the yaw of the machine that originates
the wake. Specifically, UPMPARK can be used to study
the behaviour of a whole wind farm.
ACKNOWLEDGEMENTS
We would like to express our gratitude to Professor P.
Moin and to Professor J. Alonso (Center for Turbulence
Research, Stanford University) for their help and advice.
The authors greatly acknowledge the Spanish Ministry of
Science and Education for their financial support of the
research Project with reference number CGL2005-06966-
C07-03/CLI, ‘Estudio de microescala’ (INVENTO) and
the scholarship of Ángel Jiménez in the FPU research
support program. We are also grateful to the European
Project ‘Next Generation Design Tool for Optimisation of
Wind Farm Topology and Operation’ (TOPFARM) for
their collaboration.
REFERENCES
1. Frandsen S, Thögersen ML. Integrated fatigue loading
for wind turbines in wind farms by combining ambient
turbulence and wakes. Wind Energy 1999; 23: 327–
340.
2. Crespo A, Hernández J, Frandsen S. A survey of
modelling methods for wind turbine wakes and wind
farms. Wind Energy 1999; 2: 1–24.
3. Vermeer LJ, Sørensen JN, Crespo A. Wind turbine
wake aerodynamics. Progress in Aerospace Sciences.
2003; 6–7: 467–510.
4. Jiménez A, Crespo A, Migoya E, García J. Advances
in large-eddy simulation of a wind turbine wake.
Wind Energ. 2010; 13:559–572 © 2009 John Wiley & Sons, Ltd.
DOI: 10.1002/we

Á. Jiménez, A. Crespo and E. Migoya
Journal of Physics: Conference Series 2007; 75:
012041. DOI: 10.1088/1742-6596/75/1/010241.
5. Crespo A, Manuel F, Moreno D, Fraga E, Hernández
J. Numerical analysis of wind turbine wakes.
Proceedings of the Delphi Workshop on Wind Energy
Applications (Bergeles G, Chadjivassiliadis J, eds),
1985: 15–25.
6. Crespo A, Hernández J. Numerical modelling of the
flow field in a wind turbine wake. Proceedings of
the 3rd Joint ASCE/ASME Mechanics Conference
Forum on Turbulent Flows ASME FED, 1989, 76:
121–127.
7. Gómez-Elvira R, Crespo A, Migoya E, Manuel F,
Hernández J. Anisotropy of turbulence in wind turbine
wakes. Journal of Wind Engineering and Industrial
Aerodynamics. 2005; 93: 797–814.
8. Masson C. Viscous differential/actuator disk method
and its applications. Proceedings of the 15th IEA
Symposium on the Aerodynamics of Wind Turbines.
Thor S-E (ed.). FOI, Swedish Defence Research
Agency, Stockholm, 2001; 65–80.
9. Kasmi AE, Masson C. An extended k-e model for
turbulent flow through horizontal-axis wind turbines.
Journal of Wind Engineering and Industrial Aero-
dynamics. 2008; 96: 103–122.
10. Sørensen JN, Shen WZ. Numerical modeling of wind
turbine wakes. Journal of Fluids Engineering. 2002;
124: 393–399.
11. Troldborg N, Sørensen JN, Mikkelsen R. Numerical
Simulations of wake characteristics of a wind turbine
in uniform inflow. Wind Energy 2009; DOI: 10.1002/
we.345 (online).
12. Cleijne JW. Results of Sexbierum Wind Farm Report.
MT–TNO, Apeldoorn 1992; 92–388.
13. Crespo A, Hernández J. Turbulence characteristics in
wind turbine wakes. Journal of Wind Engineering and
Industrial Aerodynamics. 1996; 61: 71–85.
14. Corten GP, Schaak P. Heat and Flux, increase of wind
farm production by reduction of the axial induction.
Proc. EWEC’2003, Madrid 2003.
15. Clayton BR, Filby P. Measurements effects of oblique
flows and change in blade pitch angle on performance
and wake development of model wind turbines. Proc.
4th BWEA Wind energy Conference, Cranfield, BHRA
Fluid Engineering, 469–484.
16. Grant, I., Parkin, P. and Wang, X., 1997. Optical
vortex tracking studies of a horizontal axis wind
turbine in yaw using laser-sheet flow visualization.
Experiments in Fluids 1982; 23: 513–119.
17. Grant I, Parkin P. A DPIV study of the trailing vortex
elements from the blades of a horizontal axis wind
turbine in yaw. Experiments in Fluids 2000; 28: 368–
376.
Wind Energ. 2010; 13:559–572 © 2009 John Wiley & Sons, Ltd.
DOI: 10.1002/we
10991824, 2010, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/we.380 by Universidad De Los Andes, Wiley Online Library on [26/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wake deflection of a wind turbine in yaw
18. Parkin P, Holm R, Medici D. The application of
PIV to the wake of a wind turbine in yaw. Proc.
4th International Symposium on particle image
Velocimetry, Göttingen 2001.
19. Jiménez A, Crespo A, Migoya E, García J. Large-eddy
simulation of spectral coherence in a wind turbine
wake. Environmental Research Letters. 2008; 3:
015004. DOI: 10.1088/1748-9326/3/1/015004.
20. Smagorinsky J. General circulation experiments with
the primitive equations. part i: the basic experiment.
Monthly Weather Review 1963; 91: 99–164.
21. Germano M, Piomelli U, Moin P, Cabot WH. A Dynamic
Subgrid-Scale Eddy Viscosity Model. Proceedings of
Summer Workshop, Center of Turbulence Research,
Stanford University, CA, 1990.
22. Pierce CD, Moin P. Progress-variable approach for
Large-Eddy simulation of turbulent combustion.
Report No. TF–80, Center for Turbulence Research,
Stanford University, CA, 2001.
23. Wall D, Pierce CD, Moin P. A semi-implicit method
for resolution of acoustic waves in low Mach number
flows. Journal of Computational Physics 2002; 181,
545–563
24. Senocak I, Ackerman AS, Kirkpatrik MP, Stevens
DE, Mansour NN. Study of near-surface models
in large-eddy simulations of neutrally stratified
atmospheric boundary layer. Annual Research Briefs
2004, Center for Turbulence Research, Stanford
University, CA, 2004.
25. Moeng CH. A large-eddy simulation model for the
study of planetary boundary-layer turbulence. Journal
of Atmospheric Sciences 1984; 41: 2052–2062.
26. Piomelli U, Balaras E. Wall-Layer Models for Large-
Eddy Simulations. Annual Review of Fluid Mechanics
2002; 34: 349–374.
27. Panofsky HA, Dutton JA. Atmospheric Turbulence.
Wiley, New York, 1984.
28. International Standard IEC 61400–1 (Ed.3.0), 2005.
29. Larsen GC, Madsen HA, Sørensen NN. Mean wake
deficit in the near field. Proceedings of European
Wind Energy Conference EWEC’2003, Madrid,
2003.
30. Lissaman PBS. Energy effectiveness of arbitrary
arrays of wind turbines. AIAA Paper 79–0114, 1979.
31. Vermeulen PEJ. An experimental analysis of wind
turbine wakes. Proc. 3rd Int. Symp. on Wind Energy
Systems, Lyngby 1980; 431–450.
32. Katic I, Højstrup J, Jensen NO. A simple model for
cluster efficiency. Proc. EWEC’86, Rome, 1986; 1:
407–410.
33. Voutsinas SG, Rados KG, Zervos A. On the analysis
of wake effects in wind parks. Wind Engineering,
1990; 14; 204–219.
571
 10991824, 2010, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/we.380 by Universidad De Los Andes, Wiley Online Library on [26/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wake deflection of a wind turbine in yaw Á. Jiménez, A. Crespo and E. Migoya

34. Escudier MP. Aerodynamics of a burning turbulent
gas jet in a crossflow. Combustion Science and
Technology, 1972; 4: 293–301.
35. Servert J, Crespo A, Hernández J. A One-Dimensional
Model of a Turbulent Jet Diffusion Flame in an
Ambient Atmospheric Flow, derived from a Three-
Dimensional Model. Combustion, Science and Tech-
nology 1997; 124: 83–114.

APPENDIX: DERIVATION OF
MOMENTUM CONSERVATION
EQUATIONS FOR A SKEWED WAKE
Let’s establish the control volume shown in Figure 12. m1
is defined as the mass flow crossing the wind turbine, m2
is the entrainment mass flow introduced into the wake
through its lateral contour and m3 is the global mass flow
crossing the outlet of the control volume considered. In
order to achieve mass conservation, we must admit that
m3 = m1 + m2. In addition, according to the model proposed
in the figure, we have m3 = r (U0 − ΔU)pd 2/4.
Considering that except in a much-reduced region very
close to the turbine rotor, the pressure has recovered its
original atmospheric value in the whole contour, its con-
tribution to the momentum equation will be null. Then, the
total force exerted by the wind turbine on the flow must
exactly compensate the momentum flow rate across the
boundaries of our control volume:
   
f = m3u3 − m1u1 − m2u2 (22)
At inlet and lateral boundaries, we assume an unperturbed
field of wind velocity, of magnitude U0 and parallel to x-
axis direction, as shown in the figure. Then, projecting
equation (22) on the two horizontal axes:

f x = m3 (U 0 − ΔU ) cos α − m1U 0 − m2U 0 − = m3 [(U0 − ΔU ) cos α − U 0 ]

fz = m3 (U0 − ΔU ) ( − sin α ) (23)

The skew angle a is expected to be small, and assuming with m3 ≈ rU0pd 2/4, leading finally to:
ΔU << U0, we can approximate this relations by:
πδ 2
f x ≈ − ρU 0 ΔU
f x ≈ − m3ΔU 4 (25)
(24)
fz ≈ − m3U 0α πδ 2
fz ≈ − ρU 02 α
4

Figure 12. Simplified representation of the momentum conservation-based model for the wake angle.

Wind Energ. 2010; 13:559–572 © 2009 John Wiley & Sons, Ltd.
572
DOI: 10.1002/we