Characterisation of the Unsteady Flow in the

Nacelle Region of a Modern Wind Turbine

Frederik Zahle, Niels N. Sørensen

Wind Energy Division, Risø National Laboratory for Sustainable Energy

Technical University of Denmark, P.O. box 49, DK-4000 Roskilde, Denmark
frza@risoe.dtu.dk

Abstract
A 3D Navier-Stokes solver has been used to investigate the flow in the nacelle region of a
wind turbine where anemometers are typically placed to measure the flow speed and the turbine
yaw angle. A 500 kW turbine was modelled with rotor and nacelle geometry in order to capture
the complex separated flow in the blade root region of the rotor. A number of steady state and
unsteady simulations were carried out for wind speeds ranging from 6 m/s to 16 m/s as well as two
yaw and tilt angles. The flow in the nacelle region was found to be highly unsteady, dominated by
unsteady vortex shedding from the cylindrical part of the blades which interacted with the root
vortices from each blade, generating high velocity gradients. As a consequence, the nacelle wind
speed and the nacelle flow angle were found to vary significantly with height above the nacelle
surface. The nacelle anemometry showed significant dependence on both yaw and tilt angles with
yaw errors of up to 10◦ when operating in a tilted inflow.

1 Introduction
On modern wind turbines nacelle anemometry measurements are used primarily in the yaw control of
the turbine, for start-up and shut-down, but can also be used to establish the power curve of the turbine.
The industry standard for measuring the flow speed and yaw angle is to place a sonic anemometer
at the rear of the nacelle which measures the horizontal velocity components. It is, however, well-
known that there is considerable uncertainty associated with this measurement technique. One factor
of uncertainty is due to the highly unsteady flow in the near wake of the rotating blades. Another
factor is that a measurement in hub height might not be representative of the average flow speed
and yaw angle across the entire rotor disc. The variation in inflow could be caused by the upstream
terrain, wakes from neighbouring turbines, atmospheric turbulence and velocity shear. Another issue
is that standard practice is to establish the relationship between the freestream wind speed (FSWS)
and the nacelle wind speed (NWS) as well as the freestream flow angle (FSFA) and the nacelle flow
angle (NFA) based on one reference turbine, which is subsequently used on other turbines regardless
of local conditions such as terrain and park effects.
An analysis by Dahlberg et al. [2] based on experimental data from a number of wind farms identified
critical issues in relation to power curve measurements using the nacelle wind speed. The tilt angle of
the flow over the nacelle was identified as one issue that was critical to correct measurement. Masson
and Smaı̈li [7] used CFD to investigate the detailed flow over the nacelle establishing relationships
between the FSWS and the NWS. In this work an actuator disc was used in place of the actual blade
geometry.
The mechanical power of a wind turbine is approximately proportional to cos2 (ψ), where ψ is
the yaw angle [6]. This means that a yaw error of 10◦ results in a reduction of 3% in the power
production. Over the life span of a wind turbine this amounts to a significant loss in production,
as well as increases in fatigue loads on the turbine due to the cyclic variation in the loading. These
simple considerations suggest that correct measurement of the yaw angle should be of high priority
in the design of the turbine, and it could even be argued that with the continuous increase in size of
wind turbines, alternative methods for measuring the yaw angle should be investigated to overcome
the difficulties outlined above.
The aim of this work is to investigate the flow properties in the nacelle region of a turbine for a
number of wind speeds to assess whether characteristic flow patterns can be identified, which could
lead to improvements to the guidelines for placement of the nacelle anemometer or the corrections
applied to the measured flow quantities. The turbine that was used in this work is the 40 m diameter

1
Nordtank NKT 500/41 stall regulated turbine, which is equipped with LM 19.1 blades, and has a
nominal power output of 500 kW. Although this turbine is not pitch regulated, it is believed that it is
still representative of a modern turbine, and can provide useful insight into the basic flow mechanisms
that govern in the nacelle region of a turbine. The present turbine was also chosen because a recent
measurement campaign has been carried out on this turbine by Diznabi [3]. In the experiment, the
velocity was measured at four positions on the nacelle using sonic anemometers.

2 Computational Setup
2.1 Flow Solver
For all computations the EllipSys3D pressure based incompressible Reynolds averaged Navier-Stokes
flow solver written by Michelsen [9, 10] and Sørensen [15] is used. The code uses the finite volume
method, solving for the primitive variables u, v, w, and p, in general curvilinear coordinates. The
variables are stored in a collocated grid arrangement, and odd/even pressure decoupling is avoided using
the Rhie-Chow interpolation [13]. The iterative SIMPLE [11] or PISO [4] algorithm is used to advance
the solution in time using a second-order accurate scheme. The convective terms are discretised using
either the second order upwind difference scheme, SUDS [17], or the Quadratic Upstream Interpolation
for Convective Kinematics Scheme, QUICK [5], and the viscous terms are discretised using central
differencing. The momentum equations are solved decoupled from each other using a red/black Gauss-
Seidel point solver. To accelerate the convergence of the pressure-correction equation a multigrid
solution strategy is implemented and the code is fully parallelised using the MPI library with a non-
overlapping multiblock decomposition of the solution domain. The block-block communication is done
through one layer of ghost cells around each block. To further accelerate the convergence of the
solution, grid and time step sequencing is used. In each level in the grid sequence every second grid
point is removed, reducing the number of cells by a factor eight. For rotor simulations, three grid
levels are typically used. Steady state rotor simulations are carried out using the so-called steady state
moving mesh method implemented in EllipSys3D [14].
For computations of flow over aerofoils and wind turbine blades the EllipSys3D code uses the k − ω
SST model by Menter [8], because of its good performance in wall bounded adverse pressure gradient
flows.

2.2 Computational Mesh

The stall regulated Nordtank 500 turbine is equipped with LM 19.1 blades, and has an axi-symmetric
nacelle with a maximum diameter of 2.0 m and a total length of 8.9 m including the spinner. The
rotor and nacelle are tilted by approximately 2◦ . The computational model of the turbine has been
simplified compared to the actual geometry. The model contains the three blades, spinner and nacelle,
however, the tower has been omitted since this simplifies the meshing considerably. Furthermore, the
rotor was completely rigid, and the tilt angle was set to 0◦ in order to be able to carry out steady state
simulations. All simulations were carried out with uniform inflow, therefore no ground boundary was
included in the simulations.
The computational surface mesh was generated using Gridgen [1] and contained a total of 108
blocks of 32×32 cells. The blades were resolved with 256 cells in the chordwise direction and 96 cells
in the spanwise direction, with a 64×64 cells tip cap. The nacelle was resolved with 24 blocks of 32×32
cells. The volume mesh was generated using the hyperbolic mesh generator HypGrid [16] and grown
out to form a sphere with a radius of approximately 280 m corresponding to seven rotor diameters
with 128 cells in the normal direction. The first cell height in the boundary layer was set to 1 × 10−6
m to ensure y + values of less than 2. The mesh thus contained a total of 432 blocks of 32×32×32 cells,
totalling 14.2 million cells. Figures 1 and 2 show sideviews of the mesh as well as the detailed surface
mesh on the nacelle.
In Figure 2 the four positions of the nacelle anemometer used in [3] are indicated by red circles and
assigned the names ME2A, ME2B, ME2C, and ME1. The exact positions are listed in Table 1. The
orientation of the coordinate system is also indicated in Figure 2, which is a standard right-handed

2
 Figure 1: Sideview of the mesh showing only every second grid point.

system with x and U pointing into the page, y and V pointing vertically upwards, and z and W
pointing in the flow direction.

3 Y
ME1
ME2A ME2B ME2C
2

X Z

Figure 2: Detailed view of the surface mesh on the nacelle and root sections of the blade showing only
every second grid point also showing the positions of the four anemometers.

3
 Probe name x [m] y [m] z [m]
ME2A 0.00 1.60 1.7225
ME2B 0.00 1.60 3.4225
ME2C 0.00 1.60 4.4225
ME1 0.00 2.10 7.17

Table 1: Probe positions in the experiment of Diznabi [3].

Name Mech. Power [kW] Thrust [kN]

Grid level L3 L2 L1 L3 L2 L1
Steady state 94.3 183.6 182.3 39.8 37.7 37.8
Unsteady 88.3 186.0 182.7 39.6 38.5 37.8

Table 2: Mechanical power and thrust force for the steady state and unsteady computations at 8 m/s
and 0◦ yaw computed on grids L3, L2, and L1.

2.3 Computational Parameters

A series of simulations with the inflow velocity ranging from 6 m/s to 16 m/s in steps of 2 m/s were
run as both steady state and unsteady simulations. In these simulations the turbine had zero yaw and
tilt. Another two unsteady simulation were run at 8 m/s at 5◦ , and 10◦ yaw. Since the nacelle was
rotationally symmetric, these simulations could also be used to investigate the effect of tilt by also
extracting the probe velocities in the y=0 plane (i.e. interchanging the x and y coordinates of the
probe positions). As such, the effect of two tilt angles of 5◦ , and 10◦ are also presented in this paper.
The computations were all carried out using the QUICK scheme to discretise the convective terms
while the SIMPLE algorithms was used to solve the coupled velocity/pressure equations for the steady
state and unsteady simulations. All simulations were carried out assuming fully turbulent flow using the
k − ω SST model. The grid sequencing capability in EllipSys3D was used to check mesh independence
with a total of three meshes, level 1 (L1) being the original fine mesh and level 2 (L2) and level 3 (L3)
being the meshes successively coarsened by a factor of two in each coordinate direction. The Nordtank
turbine rotates at a speed of 27.1 RPM or 2.8379 rad/s. To obtain a reasonable temporal resolution,
the time step in the unsteady simulations was set to ∆t = 1.006376 × 10−3 s at the finest grid level
corresponding to 2200 time steps per revolution or 0.164◦ per time step. At the two coarser grid levels
the time steps were increased to ∆t = 4.025509 × 10−3 s at grid level 2 and ∆t = 8.051018 × 10−3 s
at grid level 3.
Since the nacelle rotated along with the spinner and rotor, the rotating nacelle wall boundary
had to be treated in a special manner to eliminate the effect of the rotation. A no-slip zero-rotation
boundary condition was thus enforced on the nacelle surface of the turbine for both the steady state
and the unsteady simulations.

3 Results
To validate the mesh and time step independence of the simulations the steady state and the unsteady
solutions were compared for the three grid levels L3, L2, and L1 for the 8 m/s 0◦ yaw case. Table 2
summarises the results. As is evident, the coarse grid level is not sufficiently resolved to predict the
power correctly, whereas the thrust is in better agreement with the solutions on grids L2 and L1. This
is because the drag component on the blade, that enters into the driving force, is more sensitive to
mesh resolution than the lift component, which is the main contributor to the thrust. The solutions on
grids L2 and L1, however, differ by less than 1%, which validates the mesh and time step independence
of the solutions.
Figure 3 shows the mechanical power as a function of wind speed for the steady state and unsteady
computations compared to experimental results by Paulsen [12]. As can be seen there is good agreement
at the lower wind speeds, whereas EllipSys3D overestimates the power at higher wind speeds, which

4
indicates that the simulations fail to predict the stall characteristics of the rotor correctly. Figure 4
shows the tangential and normal force distributions on the blade for the six wind speeds for the steady
state simulations. Indeed, the normal force continues to rise up to 14 m/s inflow velocity indicating
that stall does not occur on the outer part of the blades, which leads to the considerable over-estimation
of power compared to the experimental results. The present results also show that there is no gain
in accuracy when using unsteady simulations compared to steady state simulations for power curve
prediction. The poor prediction of stall has also been observed with simulations on 3D profile sections,
and can mainly be attributed to the fact that the complex three-dimensional flow patterns in stall are
supressed by the RANS methodology. Table 3 summarises the computed mechanical power and thrust
force of the turbine.

800

700
Mechanical Power [kW]

600

500

400

300

200
EllipSys3D ss
100 EllipSys3D uns
Experiment
0
4 6 8 10 12 14 16 18 20 22
Wind speed [m/s]

Figure 3: Power curve calculated using steady state (ss) and unsteady (uns) EllipSys3D simulations
compared to experimental results [12].

100 2000
6 m/s
1800 8 m/s
0 10 m/s
1600
Tangential Force [N/m]

12 m/s
Normal Force [N/m]

-100 1400 14 m/s

16 m/s
-200 1200
1000
-300 800
6 m/s
-400 8 m/s 600
10 m/s 400
-500 12 m/s
14 m/s 200
16 m/s
-600 0
0 2 4 6 8 10 12 14 16 18 20 22 0 2 4 6 8 10 12 14 16 18 20 22
Radius [m] Radius [m]

Figure 4: Tangential and normal force distributions along the blade for the six wind speeds for the
steady state simulations.

The additional two unsteady computations in yaw were carried out at a wind speed of 8 m/s at

5
 Steady state Unsteady
Wind speed [m/s] Mech. Power [kW] Thrust [kN] Mech. Power [kW] Thrust [kN]
6 65.2 24.80 67.20 25.30
8 182.3 37.80 182.7 37.80
10 330.4 49.20 337.7 50.15
12 480.0 57.50 477.9 57.66
14 610.5 63.30 611.4 63.95
16 720.0 67.80 719.4 68.66

Table 3: Mechanical power and thrust force for the steady state and unsteady computations.

Yaw angle [deg] Mech. Power [kW] Thrust [kN]

0 182.7 37.8
5 181.9 37.7
10 178.4 37.2

Table 4: Mechanical power and thrust force for the unsteady computations at 8 m/s at three yaw
angles.

5◦ and 10◦ yaw, respectively. Table 4 summarises the results of these simulations compared to the 0◦
yaw case. The results show that the yawed inflow conditions result in a reduction in power production
proportional to cos1.6 (ψ) and a reduction in thrust proportional to cos0.6 (ψ), which is somewhat lower
than the findings in Madsen et al. [6].
Turning to the analysis of the flow characteristics in the nacelle region, both the steady state
simulations and unsteady simulations will be analysed. These simulations were all carried out with
zero yaw and tilt angles. Since there is no true time variation in the steady state simulations azimuthally
averaged quantities were used. To obtain azimuthally varying flow quantities, the Cartesian velocity
components were transformed to a polar coordinate system, and based on the tangential velocity, Vφ ,
and the axial velocity, W , the nacelle flow angle (NFAφ ) could be computed as a function of the
azimuthal angle φ.

Vφ = −U sin(φ) + V cos(φ) (1)

N F Aφ = atan(Vφ /W ) (2)
The nacelle wind speed (NWSφ ) was defined as the velocity magnitude composed by Vφ and W at
the given azimuthal position φ: q
N W Sφ = Vφ2 + W 2 (3)

Following a right handed coordinate system, Vφ is, when viewed from downstream, defined positive
for an anti-clockwise rotation around the z-axis. NFAφ is defined positive for an anti-clockwise rotation
around the y-axis, when viewed from above. To obtain an azimuthal average of the NWS and NFA,
these quantities were integrated along a circular path with a radius corresponding to the vertical
position of the probe locations.
2π
1
Z
r=const.
N W Sss = N W Sφ dφ|z=const (4)
2π 0

2π
1
Z
r=const.
N F Ass = N F Aφ dφ|z=const (5)
2π 0
where r is the radial distance of the probe from the nacelle center. The average NWS and NFA
extracted from the unsteady simulations were evaluated as follows:
t=T
1
Z p
N W Suns = U 2 + W 2 dt (6)
T t=0

6
 t=T
1
Z
N F Auns = atan(U/W )dt (7)
T t=0
integrated over the time period t=0 to t=T. In the present work, the unsteady simulations were
averaged over 10 revolutions, corresponding to 22.2 s.
Figure 5 shows the azimuthally averaged NWS versus the FSWS for the steady state simulations as
well as the time averaged values extracted from the unsteady simulations at the four nacelle anemome-
ter positions compared to the experimental results of Diznabi [3]. The experimental results all show
an increasing speed-up of the NWS with increasing FSWS. This is to some extent also seen in the
computations, however, the computations under-predict the NWS for the three probe positions closest
to the rotor, whereas better agreement is found at the rearmost probe position ME1. The variation
of axial induction with wind speed has limited influence in the nacelle region where the blades are
cylindrical and non-lifting. It is also evident that the unsteady simulations are not in significantly
better agreement with the experiment compared to the steady state simulations.

Probe position ME2A Probe position ME2B

16 16
NWS=FSWS NWS=FSWS
EllipSys3D SS EllipSys3D SS
Nacelle Wind Speed [m/s]

Nacelle Wind Speed [m/s]

14 EllipSys3D uns 14 EllipSys3D uns
Experiment Experiment
12 12

10 10

8 8

6 6

4 4

4 6 8 10 12 14 16 4 6 8 10 12 14 16
Free Stream Wind Speed [m/s] Free Stream Wind Speed [m/s]

Probe position ME2C Probe position ME1

16 16
NWS=FSWS NWS=FSWS
EllipSys3D SS EllipSys3D SS
Nacelle Wind Speed [m/s]

Nacelle Wind Speed [m/s]

14 EllipSys3D uns 14 EllipSys3D uns

Experiment Experiment
12 12

10 10

8 8

6 6

4 4

4 6 8 10 12 14 16 4 6 8 10 12 14 16
Free Stream Wind Speed [m/s] Free Stream Wind Speed [m/s]

Figure 5: Nacelle wind speed (NWS) versus freestream wind speed (FSWS) extracted at the nacelle
probe positions computed using steady state (SS) and unsteady (uns) simulations compared to the
experimental results by Diznabi [3].

The turbine was in the experiment tilted by 2◦ , which due to the modelling approach in the steady
state simulations was not possible. This discrepancy might well influence the comparison between the
experiment and the simulations.
Figures 6 and 7 show a time series of the NWS and the NFA for 10 consecutive revolutions at the
ME1 probe position for the 8 m/s case, as well as the azimuthally binned curve. The NWS signal is
highly unsteady and the blade passages are smeared considerably but still detectable in the azimuthally
binned curve. There is also large variation in the flow angle, with variations of ±15◦ . In the flow angle
signal a 3P frequency is clearly visible. The steady state solution is also shown in the figures, which
captures the 3P frequency for both the velocity and flow angle and is less smeared.
Steady state simulations provide fast and reliable results for flows that are physically steady state.
However, for inherently unsteady flows, a steady state solution does not necessarily provide an average

7
 Average
10 Steady State

NWS [m/s]
8

0 60 120 180 240 300 360

Blade 1 azimuth [deg.]

Figure 6: Nacelle wind speed (NWS) as a function of blade 1 azimuthal angle extracted at the ME1
position.

Average
10 Steady State
NWS [m/s]

0 60 120 180 240 300 360

Blade 1 azimuth [deg.]

Figure 7: Nacelle flow angle (NFA) as a function of blade 1 azimuthal angle extracted at the ME1
position.

solution equivalent to the average of an unsteady solution. A steady state solution converges towards
one solution that is steady, which depends on the exact initial conditions and numerical settings. As
such it was interesting to compare the steady state and averaged unsteady flowfields to investigate to
what extent the steady state solutions were representative of the mean flow in the nacelle region of
the turbine.
To gain further insight into the flow characteristics, a number of flow quantities were extracted
from the simulations and analysed. Figures 8 to 10 show contour plots of the z-vorticity, nacelle
flow angle, and tangential velocity in an x-y plane at z=7.17 m, corresponding to the nacelle probe
position ME1 for both the 8 m/s steady state and unsteady simulations. As mentioned previously,
the unsteady flowfield was averaged over 10 rotor revolutions. Each blade appears to generate three
counter-rotating vortices that induce high tangential velocities, thus locally increasing the tangential
flow angles. The large negative flow angles are generated at the interfaces between the outermost
vortices of negative vorticity and the neighbouring vortices of positive vorticity. The strongest inner
vortex of positive vorticity generates the positive flow angles close to the nacelle surface. Notice also
that the flow angle as expected is positive in the region where the blade generates a driving force due
to the wake rotation which is opposite to the blade rotation. It is evident taht the vortical structures
are more distinct in the steady state flowfield compared to the averaged unsteady flowfield. This is
likely due to the fact that the wake vortices in the steady state simulations are stationary, whereas
the unsteady simulations contain higher frequency shedding of the vortices, which on average will
cause a smearing of gradients in the flow. However, this comparison indica tes that although there
are differences between the steady state and unsteady simulations, qualitatively, much insight can be

8
gained from steady state simulations.

Figure 8: Contour plot of the z-vorticity at the ME1 position based on the steady state flowfield (left)
and averaged unsteady flowfield (right).

Figure 9: Contour plot of the NFAφ at the ME1 position based on the steady state flowfield (left) and
averaged unsteady flowfield (right).

In relation to positioning of nacelle anemometers, it is relevant to investigate the sensitivity of the

NWS and the NFA to the vertical position of the nacelle anemometer. Looking firstly at the steady
state results, Figure 11 shows the normalized NWS as a function of vertical distance at z=7.17 m,
corresponding to the plane where the ME1 nacelle probe is positioned. The NWS varies significantly
with vertical distance as much as 15% from y=1.5 m to y=2.5 m, where the velocity gradient is
strongest. Figure 5 suggested that the normalised NWS did not vary significantly with wind speed.
However, Figure 11 clearly shows that this is not the case. It so happens that at a height of y=2.1 m,
where the ME1 anemometer is located, the normalized NWS is quite close to 1 for all wind speeds.
However, as is also evident, the NWS varies significantly at other vertical positions, with as much as
8% at y=2.5 m. Figure 12 shows the NFA as a function of vertical distance. As for the NWS the flow
angle is also highly sensitive to the vertical position with a variation of as much as 12◦ from z=1.5 m
to z=2.5 m.
Figures 13 and 14 show equivalent profiles extracted from the unsteady simulations. Looking firstly

9
Figure 10: Contour plot of the NWSφ at the ME1 position based the steady state flowfield (left) and
averaged unsteady flowfield (right).

5 5
6 m/s 6 m/s
4.5 8 m/s 4.5 8 m/s
10 m/s 10 m/s
4 12 m/s 12 m/s
14 m/s 4
14 m/s
3.5 16 m/s 16 m/s
3.5
y [m]

y [m]

3 3
2.5 2.5
2 2
1.5 1.5
1 1
0.75 0.8 0.85 0.9 0.95 1 1.05 1.1 -10 -5 0 5 10 15 20
NWS/FSWS [-] NFA [deg.]

Figure 11: Azimuthally averaged nacelle wind Figure 12: Azimuthally averaged nacelle flow
speed (NWS/FSWS) as a function of vertical angle (NFA) as a function of vertical distance
distance extracted at the ME1 position from extracted at the ME1 position from the steady
the steady state simulations. state simulations.

at the NWS it can be seen that the unsteady solutions qualitatively agree quite well with the steady
state solutions shown in Figure 11, predicting the minimum and maximum positions of velocity at
approximately the same heights. The steady state solutions do, however, in general over-predict the
wind speed compared to the unsteady solutions. The same trend is visible for the time averaged NFA
in Figure 14, where the steady state flow angles plotted in Figure 12 are also lower than those extracted
in the unsteady simulations. However, the location of minimum flow angle is predicted approximately
in the same vertical position in both types of simulations.
The last part of the present investigation concerns the effect of tilt and yaw on the measured
flow quantities on the nacelle. Looking firstly at the case of tilt, Figure 15 shows the time averaged
normalized NWS for three tilt angles, 0◦ , 5◦ , and 10◦ . The tilt angle clearly has a strong effect on the
measured flow speed, with a 10% reduction in the NWS from 0◦ tilt to 10◦ tilt at the ME1 position.
The reason for the large shift in flow speed is that the root vortex which gives rise to a speed-up effect
is shifted upwards due to the tilt angle. Likewise, the NFA is strongly affected by tilt, giving rise to a
10◦ difference in the measured flow angle when the tilt is increased from 0◦ tilt to 10◦ .
When the turbine operates in yaw the NWS profile changes, however, not as drastically as for tilt.
As is evident the speedup region shifts downwards with increasing yaw angle, due to the movement

10
 5 5
W=6 m/s W=6 m/s
4.5 W=8 m/s 4.5 W=8 m/s
W=10 m/s W=10 m/s
4 W=12 m/s W=12 m/s
W=14 m/s 4
W=14 m/s
3.5 W=16 m/s W=16 m/s
3.5
y [m]

y [m]
3 3
2.5 2.5
2 2
1.5 1.5
1 1
0.75 0.8 0.85 0.9 0.95 1 1.05 1.1 -4 -2 0 2 4 6 8 10 12 14
NWS/FSWS [-] NFA [deg.]

Figure 13: Time averaged nacelle wind speed Figure 14: Time averaged nacelle flow angle
(NWS/FSWS) as a function of vertical dis- (NFA) as a function of vertical distance ex-
tance extracted at the ME1 position from the tracted at the ME1 position from the unsteady
unsteady simulations. simulations.

5 5
0 deg. 0 deg.
4.5 5 deg. tilt 4.5 5 deg. tilt
10 deg. tilt 10 deg. tilt
4 4

3.5 3.5
y [m]

y [m]
3 3

2.5 2.5

2 2

1.5 1.5

1 1
0.8 0.85 0.9 0.95 1 1.05 -6 -4 -2 0 2 4 6 8 10 12 14 16
NWS/FSWS [-] NFA [deg.]

Figure 15: Effect of tilt on the nacelle wind Figure 16: Effect of tilt on the nacelle flow
speed (NWS/FSWS) as a function of vertical angle (NFA) as a function of vertical distance
distance at the ME1 position. at the ME1 position.

of the root vortices. The measured NWS at the ME1 anemometer position shifts approximately 4%
when operating in yaw. The NFA is predicted quite accurately relative to the zero yaw computation
with only little error. As for the NWS, the minimum flow angle position is also shifted downwards for
increasing yaw angle.

4 Discussion
In the present work both steady state and unsteady simulations have been used to investigate the
flow characteristics in the nacelle region of a rotor. It was relevant to investigate the applicability of
the steady state simulations to this type of problem, since the majority of rotor simulations carried
out today are steady state. Although the assumption of steady state flow is not correct when the
flow is inherently unsteady, the present results, do however, indicate that steady state simulations are
to a large extent capable of capturing the most dominant flow characteristics in the nacelle region,
identifying the areas of high gradients in the NWS and the NFA. At the ME1 position, the unsteady
solutions predict a spread of 2.2◦ of the NFA for the computed range of wind speeds, which suggests
that a nacelle anemometer can be calibrated to account for the average turning of the flow at a given
anemometer position. These conclusions, however, do not hold if the turbine operates under non-
ideal conditions. A tilted inflow causes the flow pattern to change considerably, which for the present

11
 5 5
0 deg. 0 deg.
4.5 5 deg. yaw 4.5 5 deg. yaw
10 deg. yaw 10 deg. yaw
4 4
y [m] 3.5 3.5

y [m]
3 3

2.5 2.5

2 2

1.5 1.5

1 1
0.86 0.88 0.9 0.92 0.94 0.96 0.98 1 1.02 1.04 1.06 -5 0 5 10 15 20 25
NWS/FSWS [-] NFA [deg.]

Figure 17: Effect of yaw on the nacelle wind Figure 18: Effect of yaw on the nacelle flow
speed (NWS/FSWS) as a function of vertical angle (NFA) as a function of vertical distance
distance at the ME1 position. at the ME1 position.

turbine for the 10◦ tilt case results in a 10% reduction in the predicted NWS and an 8◦ increase in
the predicted NFA. Since the present paper only deals with one particular turbine, it still remains to
be shown that the present findings also apply to other turbines. However, the flow patterns that were
identified appear to be driven by the interaction between the blade root vortex and the horse shoe
vortex released at the intersection between the blade root and the spinner. As such it is likely that
similar mechanisms will dominate the flow on turbines with a similar blade root layout regardless of
whether it is stall or pitch regulated.

5 Conclusions
In this work the characteristics of the flow in the nacelle region of a stall regulated turbine has been
investigated. It has been shown that a distinct flow pattern exists with a complex set of vortical
structures, which induce high tangential velocities in the region where nacelle anemometers are typically
placed. It was found that the flow pattern persists for a number of different wind speeds, yaw and tilt
angles. From the simulations it was found that the nacelle wind speed was predicted quite accurately
compared to the free stream wind speed, whereas the nacelle flow angle was more sensitive. This
sensitivity was shown to be related to the very high gradients in the tangential wind speed resulting in
a significant variation in nacelle flow angle with height above the nacelle. Comparing the steady state
and unsteady simulations, it was found that despite the fact that the flow is highly unsteady in the
nacelle region, the steady state solutions were in quite good qualitative agreement with the unsteady
time averaged solutions. When operating in tilt the flow around the nacelle influenced the nacelle
wind speed and flow angle considerably. Yaw misalignment did not have as strong an influence on the
measurement of wind speed and flow angle. In general these findings suggest that a detailed numerical
analysis of the flow in the nacelle region of a turbine could be very useful for positioning the nacelle
anemometer. Furthermore, the results indicate that it is very difficult to make simple corrections to
the measured quantities that will be valid for all flow conditions, particularly off-design conditions.

6 Acknowledgements
The work was funded by the Danish Energy Agency under contract ENS-33033-0055. Computations
were made possible by the use of the Thyra PC-cluster at Risø National Laboratory DTU.

References
[1] URL http://www.pointwise.com.

12
 [2] J.-A. Dahlberg, S. Frandsen, H. A. Madsen, I. Antoniou, T. Friis Pedersen, R. Hunter, H. Klug,
and A. Albers. Is the nacelle mounted anemometer an acceptable option in performance testing.
In Proceedings of the European Wind Energy Conference, pages 624–637, Nice, France, 1-5 March
1999.
[3] B. Diznabi. Investigation of the flow relation to nacelle anemometry. Master’s thesis, Danish
Technical University, Lyngby, Denmark, 2009.
[4] R. I. Issa. Solution of the implicitly discretised fluid flow equations by operator-splitting. journal
of Computational Physics, 62:40–65, 1985.
[5] B. Leonard. A stable and accurate convective modelling procedure based on quadratic upsteam
interpolation. Computer Methods in Applied Mechanics and Engineering, 19:59–98, 1979.
[6] H. A. Madsen, N. N. Sørensen, and S. Schreck. Yaw aerodynaics analyzed with three codes in
comparison with experiment. In 41st Aerospace Sciences Meeting and Exhibit, number AIAA
2003-519, Reno, Nevada, USA, 6-9 January 2003.
[7] C. Masson and A. Smaı̈li. Numerical study of turbulent flow around a wind turbine nacelle. Wind
Energy, 9:281–298, 2006.
[8] F. R. Menter. Zonal two-equation k − ω models for aerodynamic flows. AIAA paper 93-2906,
1993.
[9] J. A. Michelsen. Basis3D—a platform for development of multiblock PDE solvers. Technical
Report AFM 92-05, Technical University of Denmark, 1992.
[10] J. A. Michelsen. Block structured multigrid solution of 2D and 3D elliptic PDEs. Technical Report
AFM 94-06, Technical University of Denmark, 1994.
[11] S. V. Patankar and D. B. Spalding. A calculation procedure for heat, mass and momentum
transfer in three-dimensional parabolic flows. International Journal of Heat and Mass Transfer,
15:1787–1806, 1972.
[12] U. S. Paulsen. Konceptundersøgelse Nordtank NTK 500/41 måling af effektkurve. Technical
Report Risø-I-889(DA), Risø National Laboratory, Roskilde, Denmark, November 1995.
[13] C. M. Rhie and W. L. Chow. Numerical study of the turbulent flow past an aerofoil with trailing
edge separation. AIAA journal, 21:1525–1532, 1983.
[14] N. N. Sørensen. Rotor computations using a ’steady state’ moving mesh. IEA Joint Action
Committee on aerodynamics. Annex XI and 20. Aero experts meeting, Pamplona (ES), 25-26
May 2005.
[15] N. N. Sørensen. General purpose flow solver applied to flow over hills. Technical Report Risø-R-
827(EN), Risoe National Laboratory, 1995.
[16] N. N. Sørensen. HypGrid2D—a 2-D mesh generator. Technical report, Risø-R-1035(EN), Risoe
National Laboratory, 1998.
[17] R. W. Yeo, P. Wood, and A. N. Hrymak. A numerical study of laminar 90-degree bend duct flow
with different discretization schemes. Journal of Fluids Engineering, 112:563–568, 1991.

13