Informes Tcnicos Ciemat

910
diciembre, 1999

CFD Calculations ofthe Flow Around

aWindTurbineNacelle

J. Vrela
D. Bercebal

Departamento de Energas Renovables

Toda correspondenica en relacin con este trabajo debe dirigirse al Servicio de

Informacin y Documentacin, Centro de Investigaciones Energticas, Medioambientales y
Tecnolgicas, Ciudad Universitaria, 28040-MADRID, ESPAA.

Las solicitudes de ejemplares deben dirigirse a este mismo Servicio.

Los descriptores se han seleccionado del Thesauro del DOE para describir las materias
que contiene este informe con vistas a su recuperacin. La catalogacin se ha hecho
utilizando el documento DOE/TIC-4602 (Rev. 1) Descriptive Cataloguing On-Line, y la
clasificacin de acuerdo con ei documento DOE/TIC.4584-R7 Subject Categories and Scope
publicados por el Office of Scientific and Technical Information del Departamento de Energia
de los Estdos Unidos.

Se autoriza la reproduccin de los resmenes analticos que aparecen en esta

publicacin.

Depsito Legal: M -14226-1995

ISSN: 1135-9420
IPO: 238-99-003-5

Editorial CIEMAT

CLASIFICACIN DOE Y DESCRIPTORES

170600

CALCULARON METHODS; ANEMOMETERS; FLOWMETERS; WIND TURBINES;

AERODYNAMICS; FLUID MECHANICS

"CFD Calculations of the Flow Around a Wind Trubine Nacelle"

Vrela, J.; Bercebal, D.
41 pp. 38 fig. 5 refs.
Abstract:
The purpose of this work is to identify the influence of a MADE AE30 wind turbine nacelle on the site calibration
aiiemometer placed on the upper back of the nacelle by means of flow simulations around the nacelle using
FLUENT, a Commercial Computational Fluid Dynamics code (CFD), which provides modeling capabilities for
the simulation of wide range laminar and turbulent fluid flow problems. Different 2D and 3D simulations were
accomplished in order to estimate the effects of the complex geometry on the flow behavior. The speed up and
braking vales of the air flow at the anemometer position are presented for different flow conditions. Finally some
conclusiones aboutthe accuracy of results are mentioned.

"Clculos con CFD del Flujo Alrededor de una Gndola de un Aerogenerador"

Vrela, J. Berceba!, D.
41 pp. 38 fig. 5 refs.

Resumen:
El Propsito de este trabajo es la identificacin de la influencia de la gndola de un aerogenerador MADE
AE-30 sobre el anemmetro de calibracin del emplazamiento situado en lo alto de la parte posterior de la
gndola, mediate simulaciones del flujo alrededor de sta, usando FLUENT, un cdigo de Fluidodinmica
Computacional (CFD), que es capaz de modelizar simulaciones fuidodinmicas laminares y turbulentas. Se
realizan diferentes simulaciones en 2D y 3D para poder estimar los efectos de la geometra compleja de la
gndola en el comportamiento del flujo.Tambin se presentan los valores de aceleracin y deceleracin del
flujo en la posicin del anemmetro de calibracin para diferentes condiciones del flujo. Finalmente se mencionan algunas conclusiones sobre la exactitud de los resultados.

Contents
1 Introduction

2 Methodology

3 Physical Models
3.1 Transport Equations
3.2 Trbulence models
3.3 Near-Wall treatment

5
5
7
8

4 Numerical Simulation

5 Results of the CFD calculations

5.1 2-D calculations
5.1.1 Different domains
5.1.2 Different velocities
5.2 3-D calculations
5.2.1 Different velocities
5.2.2 Different vertical inclination
5.2.3 Different horizontal inclination

10
10
10
20
27
27
29
34

39

Conclusions

List of Figures
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18

The nacelle
Enclosure for case A (2-D)
Enclosure for case B (2-D)
Enclosure for case C (2-D)
Velocity magnitude. Case A (2-D)
Velocity magnitude. Case B (2-D)
Velocity magnitude. Case C (2-D)
Turbulent viscosity ratio. Case A (2-D)
Turbulent viscosity ratio. Case B (2-D)
Turbulent viscosity ratio. Case C (2-D)
Turbulent kinetic energy. Case A (2-D)
Turbulent kinetic energy. Case B (2-D)
Turbulent kinetic energy. Case C (2-D)
Wind profiles with different domains (2-D)
Velocity magnitude. Vne = 4 m/s (2-D)
Velocity magnitude. Viniet = 8 m/s (2-D)
Velocity magnitude. Vini&t = 12 m/s (2-D)
Velocity magnitude. Viniet = 16 m/s (2-D)

4
11
12
12
13
13
14
14
15
15
16
16
17
19
20
21
21
22

19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38

Turbulent viscosity ratio. Vinet = 4 m/s (2-D)

Turbulent viscosity ratio. Viniet = 8 m/s (2-D)
Turbulent viscosity ratio. Viniet = 12 m/s (2-D)
Turbulent viscosity ratio. Vinlet = 16 m/s (2-D)
Turbulent kinetic energy. Viniet = 4 m/s (2-D)
Turbulent kinetic energy. Viniet = 8 m/s (2-D)
Turbulent kinetic energy. Vin\et = 12 m / s (2-D)
Turbulent kinetic energy. Vne = 16 m / s (2-D)
Contours of velocity difierence (2-D)
Comparison of the wind profiles between 2-D and 3-D cases
Velocity vectors. 10 of vertical inclination. (3-D)
Velocity vectors. 30 of vertical inclination. (3-D)
Wind profiles with different vertical inclination (3-D)
Contours of velocity difierence, 10of vertical inclination . .
Contours of velocity difierence, 30of vertical inclination . .
Velocity vectors. 10 of horizontal inclination. (3-D)
Velocity vectors. 30 of horizontal inclination. (3-D)
Wind profiles with different horizontal inclination (3-D). . .
Contours of velocity difference, 10of horizontal inclination
Contours of velocity difference, 30of horizontal inclination

22
23
23
24
24
25
25
26
28
. 30
31
31
32
. 33
. 33
35
35
. 37
. 38
. 38

Introduction

During September 1998 and March 1999 a power performance measurement according to standard IEC 61400-12 [1] was carried out on a
MADE E30 wind turbine at the 'A Capelada', a wind farm in Galicia,
Northwest of Spain. The orographic complexity of the terrain surrounding the wind turbine led to perform a calibration of the site, using the wind
turbine itself as second wind mast.
The purpose of this paper is to identify the influence of the wind turbine
nacelle on the site calibration anemometer placed on the upper back of the
nacelle by means of flow simulations around the nacelle using FLUENT,
a commercial Computational Fluid Dynamics code (CFD), which provides
modeling capabilities for the sknulation of wide range laminar and turbulent
fluid flow problems.
These simulations were performed at the Renewable Energy Department
of CIEMAT.

Methodology

A site calibration is usually done with two meteomasts, one of them at

the turbine position, to obtain the correlation factors for the correction of
the power performance measurement. Because the turbulence was already
installed at the wind farm, the second anemometer was placed on a seprate
small mast on the top of the nacelle's back. The wind turbine was stopped
and yawing downwind during site calibration procedure.
The influence of the turbine nacelle on the site calibration anemometer are determined by means of CFD calculations simulating the air flow
around the MADE AE30 nacelle using real manufacturing geometry. The
CFD simulations identify the speed up or deceleration factors of the flow
at the anemometer position due to the nacelle in comparison to the undisturbed air flow.
Two different kind of simulations were accomplished by means of FLUENT;
first they were carried out 2-D calculations in order to estimate the amount
of space surrounding the nacelle necessary to achieve accuracy vales of
wind profiles while providing a number of cells as least as possible, trying to
decrease the computational effort. Later they were made 3-D calculations
to consider the effects of the complex geometry in the simulations.
A view of the nacelle is shown in figure 1 on page 4.
The dimensions of the nacelle are grossly 3 m of height, 2 m of depth
and 5 m of length. In all the simulations the nacelle's rear end was oriented
upwind, and the blades were not modeled. The speeds of interest in this

Figure 1: The nacelle

study ranges from 4 m/s to 16 m/s, therefore the Reynolds number is very
high and the effects of turbulence must be considered.
The adimensional numbers that characterize the flow are:
^ i ^ ~ 3.89
v

Re =

fe

-0.72

The height of the boundary layer is

H
- 3
m
T ~ 2.4 W
And the Kohnogorov scale is
H

~ 5.4 1(T5 m

Re

In the next section it will be shown how FLUENT solves the NavierStokes equations and how turbulence is modeled.

Physical Models

3.1 Transport Equations

FLUENT solves the Navier-Stokes equations for conservation of mass,
momentum and energy. Additiona! conservation equations for 'jfc' and 'e' are
solved when the flow is turbulent.
Mass conservation
The equation for conservation of mass or continuity equation can be
written as:
^
O
(i)
Momentum conservation
Conservation of momentum in each component 'i' is described by [2]:
^L

with

= 1,2,3

(2)

where rj is the stress tensor, given by:

The first term on the right is the volunae dilation and i is the molecular
viscosity.

Energy conservation
The energy equation can be written in terms of the sensible enthalpy h
as (neglecting the effects of viscous heating):

| W + V (PW) = V (AVT,+ (
K i.

vi i-

where A is the thermal conductivity.

The sensible enthalpy is defined as:

h= I

CpdT

JTref

Equations 1, 2, 4 contain v, p, T, p as unknown dependent variables, so one

further scalar equation is needed to make possible the determination of the
flow field. This additional relationship is provided by the equation of state,
which may be written generally as:
f(p,T,p)=0
where the functional form of the equation of state, depends on the nature
of the fluid.
Equation of State
The pressure valu used in the code is a gauge pressure, relative to the
reference pressure: p = pabs pref, henee the equation of state has the form:
(P+Pref)W
RT

FLUENT has an approximation denoted incompressible ideal gas law for

density. In this approximation the overpressure p is very small compared
with pref, henee its contribution to density is ignored, and pressure only
affeets density through the reference pressure:
P

_ PrefW
RT

This approximation implies that pressure fluctuations are ignored in the

equation of state, cutting out sound waves. In fact, it is assumed:
dp
c

In all the simulations it was assumed that the flow is incompressible and
isotherm, so the valu of p at the pressure of reference (1 atm) is 1.225 kg/m3.

3.2

Turbulence models

In this problem the concept of boundary layer, turbulent flow and wall
effects will play a dominant role. The turbulent model used in all the simulations was the Standard fe-e model.
Proposed by Jones and Launder [3], it is a two-equation model based on an
isotropic eddy-viscosity concept, derived from the Reynolds-Average NavierStokes equations, based on the assumption that the flow is fully turbulent
and the effects of molecular viscosity are negligible. It is a semi-empirical
model and provides a reasonably accuracy for a wide range of turbulent
flows.
In the Reynolds averaging, all the solution variables in the original NavierStokes equations are decomposed into the mean (ensemble-averaged or timeaveraged) and fluctuating components. For an scalar <j>:
cf) = (f) -f (jj where:
_

rt+6t

ot Jt-st
where St is a time scale larger than the turbulent fluctuations and smaller
than the time scale we want to resolve. Substituting the expressions of
this form for the flow variables into the instantaneous continuity and momentum equations 1, 2; the "Reynolds averaged" Navier-Stokes equations
can be written (dropping the bars from Reynolds averaged mean quantities
for all primitive variables except averages involving products of fluctuating
quantities) as:
^

0
pVjV1)

(5)
with

i=

The effects of turbulence are represented by the "Reynolds stresses": pv^v'j,

this term arise from the non-linear convective trra in the un-averaged equations, and need to be modeled in order to cise equation 2. In the Standard
fe-e model, Reynolds stresses are modeled using the boussinesq hypothesis:

-pon = IH ^

+ e) - 3 [Pk + /t^TJ

Now ieff = i + fif, where /x is the turbulent viscosity and is computed

using the turbulent kinetic energy 'fe' and its rate of dissipation 'e'
from:
fe2
C

And k and e are obtained from the solutions of their "modeled" transport
equations:
^(pk) + V-(pkv)=V

^~) Vfc] + Gfc + Gb - pe

(7)

~(pe) + V-(pev) = V

where G^ and G& is the generation of k due to turbulent stress and buoyancy:
Tr 9Vn \

3.3 Near-Wall treatment

The near-wall modeling significantly impacts the fidelity of the numrica!
solution, due to walls are the main source of mean vorticity and turbulence.
The near-wall regin can be subdivided into three layers:
1. The viscous sublayer, it is the inner regin, where the flow is almost
laminar ke.
2. The fully turbulent layer, it is the outer regin, and where turbulent
plays the main role.
3. The buffer layer, between both, where the effect of molecular viscosity
and turbulent are equally important.
There are two approaches to model the near-wall regin:
Near-Wall modeling
The viscosity-affected regin is solved, including the viscous sublayer.
In the near-wall model the viscosity-affected regin is solved all the way
to the viscous sublayer. The whole domain is subdivided into two zones:
a viscosity affected regin and a fully turbulent regin; the limit between
them is established considering the turbulent Reynolds number:
Rey =

pVky

where y is the normal distance from the wall to the cell center.
When Rey > 200 the cell belongs to the fully turbulent regin, and the

turbulence model is employed.

When Rey < 200 the cell belongs to the viscosity affected near-wall regin,
and the one-equation model of Wolfstein is employed, where the turbulent
viscosity nt and the e field are computed from:
M

= pC
__ A; 3 / 4

where ZM and le are length scales, and they are function of the turbulent
Reynolds number.
Wall-function approach
Where the viscosity-affected layers (the viscous sublayer and the buffer
layer) are not solved, instead semi-empirical functions are used to link the
viscosity affected and the fully turbulent regions. The standard wall-function
in the code is based on the proposa! of Launder an Spaiding, and has been
most widely used for industria! flows.
The law of the wa for mean velocity yields:
U* = ~ln(Ey*)
where k is the Von Karman's constant and E is an emprica! constant. The
logarithmic law for mean velocity is valid for y* > 30 60, while in the code
is used when y* > 11.225. When y* < 11.225 at the wall-adjacent cells, they
apply the laminar stress-strain relationship:

Numerical Simulation

In order to conduct a CFD analysis it is necessary to crate a suitable

domain around the nacelle, where it will be generated the mesh on the
outside of the nacelle, therefore it will be created a brick around the nacelle
to represent the flow domain.
The scheme selected in the solver was initially of first order, and when it was
achieved a converged solution it was substituted by a second order scheme
in all the equations, and was run again until it reached the new converged
solution. In all the simulations the velocity is very low compared with the
Mach number, so it could be assumed that the flow is incompressible and
isotherm, henee it is not necessary to solve the equation of energy.

Mesh considerations
The approach of an standard wall-function was used in all the cases
considered, this was taken into account when meshing the volume, so the grid
is very fine neax the surface of the nacelle, and more coarse as it approaches
the boundaries. Specifically the mesh near the surface was meshed using
the model of boundaxy layer provided by GAMBIT (the mesh generator of
FLUENT), so it could be achieved a valu of y+ very cise to 20, enough to
have an accurate simulation according to [5].
The grid was generated using the pave model of GAMBIT, with triangular cells in the cases of 2-D and with tetrahedral cells in the 3-D cases.
This unstructured mesh was selected because it is the simplest to adapt to
the complex geometry, while providing accurate results.

Results of the CFD calculations

5.1 2-D calculations

In these simulations the boundaxy conditions were:
Velocity inlet, for the boundaxy in front of the nacelle's reax end, with
a constant valu in each simulation.
Pressure-Outlet, for the boundaxy in front of the nacelle, with a reference pressure of 101325 Pa.
Wall boundaxy, for all the walls comprising the geometry of the nacelle.
o Symmetry, for the top and bottom of the enclosure considered.
For all the simulations the turbulent parameters were:
Intensity of turbulence, I = 12%. This valu was estimated analyzing
the chaxacteristics of the wind typical of the emplacement.
@ Turbulent viscosity ratio, /t//i = 10. This input valu is enough for
the code to consider that turbulence is fully developed.
5.1.1

Different domains

First they were carried out three simulations in other to consider how
the amount of space surrounding the nacelle affected the wind profiles and
how much CPU time was needed to achieve a converged solution.
Three cases with different grid surrounding the nacelle were studied:
1. Case A. In the first case the size of the grid is 13 m x 15 m, with 25725
cells.
10

2. Case B. In the second case the size of the grid is 19 m x 15 m, with

36551 cells.
3. Case C. In the third case the size of the grid is 25 m x 15 m, with
45821 cells.
For all the cases the velocity at the inlet was 12 m/s. The problems to be
considered and the position of the anemometer are shown schematically in
figures on page 11.
The contours of velocity, kinetic energy and turbulent viscosity ratio for

Grid

Figure 2: Enclosure for case A (2-D).

all the cases are shown in figures on page 13

11

Grid

Figure 3: Enclosure for case B (2-D).

Grid

Figure 4: Enclosure for case C (2-D).

12

... .2.58e+O1
2.33e+01

2.07e+01

1.81e+01

1.55etO1

'.

'; . 1.29&+01

1.03B+01

7.75ertO

5.17&f00
| ;2.58e+O0

LjO.OOe+00

Contours of Velocity Magnitude (m/s)

Figure S: Vtacly magnitude. Gasc Al

2.4Se+01
.2.20B+01
1-96etO1
1.71&+01
1.47e+0i

g.79e+oo

7.35e+00

4.90e+00

. 2.4Se+00

!0.O0e+O0

Contours of Velocity Magnttude (m/s)

Figure 6: Velaaby magnafcude. Gase B (2-D)

13

:2.176+01
1.936+01
1.696+01
. 1.45e+01
:

1.21e+01

-9.67e+00
7.25e+00

4.836+00

2.42&+00

LjO.OOe+OO 'Contours of Velocity Magnitude (m/s)

Figure 7: Vbfacity magnitude. Qase C (2-DJ

._ ,9.576+03
8.61 e+03
7.65e+03
6.70S+03
5.74e+03
'

'4.786+03
!

3.836+03
2.87e+03
1.91e+03

.. ; 9.57e+02
LJi.24e-01

-- -

Contours of Turbulent Viscosity Ratio

Figure 8: Ttabiulenfc vscxjsity ratio. Case A (2-D)

14

2.05e+O3
,'i.03e+03
I

1.33e-01

Contours of Turbulent Viscosity Ratio

Figure 9: Ttaibuenifc visaasilty ratio. Qaisc B flZ-D]

1.07

1.1.

"i '

0-

"--

Contours of Turbulent Visoosity Ratio

Figure 10: Ttonbufeoii vLacosity natio. Case C (2-D)

15

6.72e+00
6.05e+00
5.37e+00
4.70e+00

- 4.03e+00
'. 3.366+00
' 2.69e+00
2.02e+00
1.346+00
6.72e-01

E J 1.97604
Contours of Turbulent Kinetic Energy (k) (m2/s2)

Figure 11: Tunbullent kinetic ensngyj.

OEIBC

A p-D]

6.73e+00

6.056+00

5.38e+00

4.71e+00

4.046+00

,3.36e+00

2.696+00

,2.02e+00

1.3Se+00

' ' 6.73&I1

i

Lj1.35e-04
Contours of Turbulent Kinetic Energy (k) (m2/s2)

Figure 12: TtaiMeat kiiaetic eruengy. Case B (2-D)

16

Contours of Turbulent Kinetic Energy (k) (m2/s2)

Figure 13: Tuxbuilent kinetic energy. Oase C (2-D)

17

The wind profiles along the y-axis (perpendicular to the velocity inlet)
at the position of the anemometer are shown in figure on page 19 (the position of the anemometer in the y-axis is represented by a red line), where
L = 2.7 Tn is the height of the nacelle and the origin of H is taken at the
middle of the nacelle.
Erom these it can be stated that the lower is the domain surrounding the
nacelle, lower is the change in the velocity at the same point; this is clear,
because when the grid is shorter, the wind has less space to adapt to the
geometry of the nacelle.
And the differences can be very significant, specially at the middle of the
nacelle, reaching a dference of 7% between the cases A and C.
Another interesting point is that at positions not very cise to the nacelle, in particular near the anemometer, the difference is negligible, as it is
shown in the next table, so it can be inferred that the speed-up due to the
nacelle is 4.7% approximately.
Case
A
B
C

V a (m/s)
12.518
12.585
12.569

Vdiff(%)

4.3
4.9
4.7

Table 1: Velocity magnitude and difference in per cent at the anemometer

However, compaxing the number of iterations, it is clear that diminishing
the space surrounding the nacelle doesn't mean a significant improvement in
the computational effort, because it is needed a greater number of iterations.
Therefore in the next simulations it was used the same grid as in case C.
Case
A
B
C

Iterations
413
342
302

Table 2: Iterations needed to achieve a converged solution

18

Figure 14: Wind profiles with different domains (2-D).

19

5.1!.2

Diffenent veliocities

Tlhcn. teyi were perbrmedi anotlier- four ualculationia witU dfFcrcTit vclacities rangng franx 4 m/s io 16 m/\s. tu watdi liow the wind apeed obanged
at aaemometer position and cx>mpar.e laten wifih. thJe 3-D oalnuiations.
Thie uontouirs of veloctyj, kinetlic energy and turbufant vscosity; ratio WCTOJ

8.03B+00

6.42e+00
5.626+00
.

4.82e+O0
4.O1B+O0
3.218+00

'.

2.41 e+00
1.61e+00

8.03&O1

LJo.oos+oo
Contours of Velocity MagnHude (m/s)

Figure 15J Velocity magnitude. Vniet = 4 m / s (2-1

20

1.45e+Q1
1-298+01

1.13e+01
9.65e+00
8.056400

6.44e+oo
4.83e+00
3.Z2e+00
"; . ; 1.61&+0Q

lO.OOS+00

Contours of Vetocity Magnitude (m/s)

Figura 16: Veocity magnifcude. Viniet ~ 8 m/s (.2-DJ.

1.936+01
1.59S+01
1.45e+01
.

-11e+01
' 9.67tH0Q

CD*

7.256+00
4.83e+O0
!

J2.42e+00

SlJo.OOe+00
Contours of Velocity Magnitude (m/s)

Figure 17: Vdooity. magnitude. Viniet 12 mfa (2-D)

21

3.22e+01
2.90&+01

'

2.58e+01

2.2&3+01

: "

1.93e+01

1.61e+01
;

1.29S+01

9.67e+00

6.45S+00

i i 3.22e+00
t_jo.00e+00
Contours of Velocity Magnitude (m/s)

Figura 18: Vblbcty magnitjude. Vjrafe

16 m/s ('2-D).

.. ,3.76e+03
3.398+03
3.01 e+03
*

2.63e+03
- 2.26e+03
;

1.88e+03
1.51 e+03

"

"

*fSv"5

-.-

1.136+03
- . , . - ; _

:>

7.53e+02
:

3.77B+02

l_J5.08e-01

Conlours of Turbulent Viscosity Ratio

Figure 1S: Huvbuilent viscosity ratita.

22

4' m/|s (2-D].

, 7.25B+03
6.53e*03
S.80e+03
5.08e+O3

4.35e+03
3.03e+03
2.90&+O3
2.1Be+03
1.4Se+03
' 7.26e+02

LJ2.07e-01
Contours of Turbulent Viscosity Ratio

Figure 20: Turbulent visoositly ratiio. Vnlet = 8 m/s (2-D).

. 1.07e+04
9.67e+03
8.603+03
7.S2S+03
6.45&+O3
. 5.37&+O3
4.30B+03
3.22e+03
2.1Se+03

--

. :1.07e+O3
I li.12e-01
Contours of Turbulent Viscosity Ratio

Figure 21: Ibrbulent vjisoosity ratiiu. Vini^ = 12 m,/s (2-Dj

1.41e+O4
1.27e+04
1.136+04
9.87e+03
8.46e+03
7.0SS+03
5.64e+03
4.23e+)3
2.82e+03
;i.41e+03

Contours of Turbulent Viscosity Ratio

Figure 22: Tuiitxulent viscrositly rat'io. VniBt => 1G m/s

(2-D).

5.54B-01
4.85e-01
4.156-01
;3.46e-01
2.77e-01
2.086-01
1.38e-01
;6.93e-02

LJ4.49e-O5
Contours of Turbulent Kinetio Energy (k) (m2/s2)

Figure 23: TlirbudeQt Icinetiio encrgy:. Vlnia = 4 m(\s. (2-D]

24

,2.816+00
2-53B+00
255B+O0
1.97e+00
1.69&+00
1.40e+O0

:i.12e+00
B.43&-01
5.62e-01

2.81e-01

LJ7.23&I5
Contours of Turbulent Kinetic Energy (k) (m2/s2)

Figure 243 Tuibudent ki'netio enengy;. Vniet ^ 8 ms (2-DJ.

. ...6.46e+00
5.81 e+00
5.17e+00
4.526+00
3.87e+00

2.58S+00
1.94e+00
1.29e+00
;

; 6.46e-01

i !
LJi.01e-04

Contours o Turbulent Kinetic Energy (k) (m2/s2)

Figure 25: Tlurhulent lci'nelilc en:ergy. Viniet = 12 m,/s (2-D]j

25

.. 1-I6e+Oi
1.04e+01
9.28e+00
8.126+00
6.956+00

5.806+00
' 4.64e+00
3.48S+00
..ri.'

'?;-.:.-..*

i>

i,

2.32e+00
; ,1.16e+00
LJi.29e-04
Contours of Turbulent Kinetic Energy (k) (m2/s2)

Figura 28 Turbwlcnt kinatic enengy

26

16 m/s (2-D).

And the difference between the velocity at any point and the velocity
inlet have all the same appearance, and it is shown on page 28.
The next table shows the velocity and the velocity difference in % at the
anemometer position.
V; n iet(m/s)

4
8
12
16

Va(m/s) Vdiff(%)
4.1883
8.3781
12.5688
16.7591

4.70
4.73
4.74
4.74

Table 3: Velocities at the anemometer in the 2-D cases

Therefore the change in velocity is insignificant, ant it can be stated that
the velocity inlet doesn't impact the speed-up at the anemometer, which
remains constant at 4.7% for all the cases.

5.2

3-D calculations

5.2.1

Different velocities

They were studied again the four cases analyzed in 2-D, to determine
how the complex geometry of the nacelle affects the wind profiles. The
enclosure that contains the nacelle is 20m x 12m x lOm (20 m along the
wind dixection, 12 m of height, and 10 m of width), these dimensions were
chosen taking into account the results of the 2-D calculations with different
domains. Also it was considered the symmetry plae of the nacelle, henee
it was only necessary to mesh 5 m of width.
The grid was again meshed with GAMBIT, refining the mesh near the
surface, producing a total of 305804 tetrahedral cells. The boundary conditions were the same as in the 2-D cases, but with the condition of symmetry
in the plae of symmetry of the nacelle:
a Velocity inlet, for the boundary in front of the nacelle's rear end, with
a constant valu in each simulation.
Pressure-Outlet, for the boundary in front of the nacelle, with a reference pressure of 101325 Pa.
Wall boundary, for all the waJls comprising the geometry of the nacelle.
a Symmetry, for the top, bottom of the enclosure considered and for the
plae of symmetry of the nacelle.

27

p
c

p
C\

p
o

Figure 27: Canitours of velooiity] difference (2-D].

The velocity differences at the anemometer position in % were:

Vintet(m/s)
4
8
12
16

V a (m/s)
4.0480
8.0955
12.1429
16.1902

Vdiff(%)
1.20
1.19
1.19
1.19

Table 4: Velocities at the anemometer in the 3-D cases without inclination

Henee, again the velocity inlet didn't affect the velocity difference in per
cent, but there is a significant decrease in the speed-up valu, from 4.7% in
the 2-D cases to 1.2% in the 3-D's. This can be explained considering that
in the 2-D simulations the extrapolation to a tree dimensional problem is
considering an infinitely long nacelle in the new 3-D axis, and that could
produce a greater interaction with the wind resulting in a greater speed-up.
In the figure on page 30 it is shown how the wind profiles changed between
the 2-D and 3-D cases.
5.2.2

Different vertical inclination

It was analyzed how the vertical inclination of the wind impact the wind
speed at the anemometer. Six different cases were run, in tree of them the
vertical inclination was 10, while on the other tree it was 30 (always positive, oriented to the top of the enclosure). Also it was analyzed the influence
of the velocity, for this reason it was varied the velocity at the inlet, from
4 m/s to 16 m/s. The mesh used was the same as in the case before, changing the boundary condition of symmetry of the bottom to velocity inlet.
Taking a vertical section at the middle of the nacelle, on page 31 is represented the velocity vectors colored by velocity magnitude, for the cases with
= 16 m/s and with 10 and 30 degrees of vertical inclination.

29

Figure 28: Comparison of the wind profiles between 2-D and 3-D cases

30

_30.9
27.8

24.8

21.7

18.6

15.5

12.4

9.3
6.3
3.2
0.1

Figura 29: Velocity vectars. 10 of vertical nclination. (3-D)

30.8

27.7

24.7

21.6

18.6

.15.5
12.4
:9.4

6.3
;3.3

Figure 30: Vblocily. ver.tors. 30Q of vertical incMoaticui. (3-Q)j

The vekicity magnfcude albng the y-axs (perpendicular to th velacity

Inletj at thc positrn of thte anwtrDometer aire shown I figure on page 32 (th
posltlcm of. the aTiemomateiF in thc y-axis is represented hy a red Une].

((j-g) uoxyexipm

iAv saTjoid p x n ^ :jg

Velocity difference %

hfl contuuns uf: differeniue btetwaen thte velocitv at any point and thc
vek>.ciy inlet fon the cases with 10. and 30 degnees of vertical iacinatlan
were:
... 80.0
69.5
59.0
485
38.0
' 27.5
'' 1 7 . 0

: 6.5
-4.0
1-14.5
-25.0

Figure 32: Confluirrs of velooityj differen.ee. 10of vertica inclinatiort

... . 50.00
40.00
30.00
20.00
10.00
.

. 0.00

;-lo.oo

-20.00
-30.00

:-4o.oo

l'

1.00

Figure 33: Contours of veooityj diifl'eraniue, 30of vertioal in:clinatiioa

33:

And the resulting velocities at the anemometer were:

V n iet(m/s)

Degrees

Va(m/s)

4
4
4
8
8
8
16
16
16

0
10
30
0
10
30
0
10
30

4.0480
4.0430
4.1323
8.0955
8.0859
8.2639
16.1902
16.1716
16.5268

v diff (%)
1.20
1.07
3.31
1.19
1.07
3.29
1.19
1.07
3.29

Table 5: Velocities at the anemometer in the 3-D cases with vertical inclination
As in the other 2-D and 3-D cases, the variation of the velocity difference
in % with velocity inlet is negligible, but in contrast there are significant
changes with the vertical inclination of the wind. The velocity difference
changes in a complex way, decreasing with low inclination and later increasing when the inclination reaches 30.
5.2.3

Different horizontal inclination

There were also studied two different cases with different horizontal inclination of the wind. For these cases it was necessary to consider the total
volume in 3-D, because the wind inclination broke the symmetry of the
problem. So an enclosure analogous to the before cases, 20m x 12m x lOm
was meshed again, resulting in 333484 tetrahedral cells.
In the fixst case the inclination with respect to the horizontal plae was 10
degrees, while in the last was 30 degrees. On both cases the velocity magnitude was 12 m/s and there was no inclination in the vertical direction.
Taking an horizontal section at the middle of the nacelle, on page 35 it is
shown the velocity vectors colored by velocity magnitude, for the cases with
10 and 30 degrees of horizontal inclination.

34

,,22.6
,

20.3
18.1
15.8
13.6

,11.3
r
: i
i i 9.1
'

i 6.8
.4.6

:
!

i
!2-3

Fguna 34: Velocity, vecors. 10 of htonizonta incBnatJion. (3-D)

.. ..25.7
32
20.6
18.0
15.5
12.9
i '
:

10.3
7.8

52

'

21

\A
Figure 35: Velacltiy, vectora. 30 of Hunizanital inclination. (3-D)

The velocity magnitude along the y-axis (perpendicular to the velocity

inlet) at the position of the aaemometer are shown in figure on page 37 (the
position of the anemometer in the y-axis is represented by a red line).
The contours of difference between the velocity at any point and the
velocity inlet for the cases with 10 and 30 degrees of horizontal mclination
are shown on page 38. And the velocity magnitude at the anemometer
position is shown in the next table:
Degrees
0
10
30

Va(m/s)
12.1429
12.172
12.026

v diff (%)
1.19
1.44
0.22

Table 6: Velocities at the anemometer in the 3-D cases with horizontal

inclination
Erom these results we can see that when the inclination reaches high vales (30) the speed-up at the anemometer is lower, while at low inclination
(10) the speed-up is a little greater.

36

:gg 8JtiSt[
Velocity difference %
en

01

r_,

40.0
30.0
20.0
10.0
0.0
:-10.0
I
i-20.0
I-30.0
I -40.0
i -50.0

LJ-60.0

Fligume 37: Contours of vefcncity; differenca, Ii0aof hon*aantal ioalinafcion

.__ 40.0
33.0
26.0
19.0
12.0
5.0
i

. -2.0

:-9.o
-16.0

i - 23 - 0
LJ-30.0

Figure 38: Cant0urs ai veliacKy difercnae, 3flafi horiizonali nalnatian

38

Conclusions

The code FLUENT has been used to asses the speed-up due to the nacelle
in a variety of cases.
Comparing the 2-D and the 3-D cases, there is a clear diminution of the
speed-up due to the nacelle, from 4.7% in the 2-D's to 1.2% in the 3-D's.
This can be interpreted taking into account that when the simulation is in
2-D, in fact the extrapolation to 3-D assumes a nacelle which is infinitely
long in the other direction, and these could result in a raise in the speed-up,
due to a greater interaction with the nacelle.
Also it has been shown that this speed-up remains ahnost constant, for a
wide range of velocities (from 4 m/s to 16 m/s) in both 2-D and 3-D cases.
This is due to the approximation of incompressible and isotherm flow, which
allowed to not solve the equation of energy.
Greater importance has the inclination in the vertical plae, with a complex variation of the differences in %, decreasing at low inclination and increasing quickly at greater inclination. The inclination in the horizontal
plae resulted in a different behavior, increasing a little at low inclination
and decreasing quickly at high vales of inclination. Both behaviors could
be produced due to the influence of the complex geometry.
However, all these results must be considered as approximations within a
range of error, not exact vales, because there are a lot of uncertainties in the
simulation: influence of the enclosure selected, the number and type of cells,
the model of turbulence, the mesh quality near the boundary layer, ... and
of course the influence of the blades, not included in these simulations.
More simulations will have to be done to understand how the inclusin of the
equation of energy affects the speed-up at different velocities (not observed
in this paper), and another set of simulations will be necessary to understand
how the quality of the mesh near the surface of the nacelle affects the wind
profiles.

39

References
[1] IEC 61400-12: Wind Turbine Generator Systems. Part 12: Power Performance Testing, 1998.
[2] G. K. Bathelor. An Introduction to Fluid Dynamics. Cambridge Univ.
Press, England 1967.
[3] B. E. Launder and D.B. Spalding. Lectures in Mathematical Models of
Turbulence. Academic Press, London, England, 1972.
[4] V. Yakhot and S. A. Orszag. Renormalization Group Analysis of Turbulence: I. Basic Theory. J. Scientific Computing, 1986.
[5] FLUENT User's Guide. FLUENT Inc., 1998.

40