Professional Documents
Culture Documents
910
diciembre, 1999
J. Vrela
D. Bercebal
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.
Editorial CIEMAT
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
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
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
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
~ 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
with
= 1,2,3
(2)
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-
h= I
CpdT
JTref
_ PrefW
RT
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=
-pon = IH ^
+ e) - 3 [Pk + /t^TJ
And k and e are obtained from the solutions of their "modeled" transport
equations:
^(pk) + V-(pkv)=V
(7)
~(pe) + V-(pev) = V
where G^ and G& is the generation of k due to turbulent stress and buoyancy:
Tr 9Vn \
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
= 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
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.
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
Grid
11
Grid
Grid
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
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
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
._ ,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
-- -
14
2.05e+O3
,'i.03e+03
I
1.33e-01
1.07
1.1.
"i '
0-
"--
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)
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
Lj1.35e-04
Contours of Turbulent Kinetic Energy (k) (m2/s2)
16
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
Iterations
413
342
302
18
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)
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
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)
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)
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
22
, 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
. 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
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
(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)
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)
. ...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
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)
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
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
V a (m/s)
4.0480
8.0955
12.1429
16.1902
Vdiff(%)
1.20
1.19
1.19
1.19
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
30.8
27.7
24.7
21.6
18.6
.15.5
12.4
:9.4
6.3
;3.3
((j-g) uoxyexipm
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
... . 50.00
40.00
30.00
20.00
10.00
.
. 0.00
;-lo.oo
-20.00
-30.00
:-4o.oo
l'
1.00
33:
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
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
.. ..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)
Va(m/s)
12.1429
12.172
12.026
v diff (%)
1.19
1.44
0.22
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
.__ 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
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