Академический Документы
Профессиональный Документы
Культура Документы
i
i
x
u
Momentum conservation
Reynolds stress tensor evaluated from a one-equation closure scheme.
i
( )
0 ' ' = +
(
(
|
|
\
|
i j i
i
j
j
i
j i j
i j
F u u
x
u
x
u
x x
P
x
u u
|
|
\
|
=
i
j
j
i
t j i
x
u
x
u
u u ' ' T T
L k
2 / 1
=
where
|
|
\
|
=
=
(
(
|
|
\
|
+ =
T T
j
j
i
j
j
i
T k
T
T
j k
T
j
k
j
j
L k
x
U
x
U
x
U
P
k
L
C
with
x
k
x
P
x
k
U
2 / 1
2
,
d L
T
* 5478 . 0 * 41 . 0 =
Where d is the distance to the nearest wall
<
=
=
|
|
\
|
+ =
=
16 , 0 , 085 , 0
16 , 0 ,
) 2231 , 0 )( 1 (
) 2341 , 0 )( 1912 , 0 (
96 , 1
,
1 1 1
2
0
2 3
if
if
if if
if if
m
m T
R si
R si
R R
R R
S
heigth z where
z l l
l S L
meteodyn WT
UrbaWind
meteodyn WT
Mesh : Cartesian structured
horizontale resolution : 3 m
verticale resolution : 0.5 m
Mesh points :1.5 M
Mesh generation duration : 15 minutes
Simulation duration : 24 minutes with1 CPU by direction Simulation duration : 24 minutes with1 CPU by direction
UrbaWind
Mesh : Cartesian unstructured
horizontale resolution around result points : 0.5 m
verticale resolution around result points : 0.1 m
Mesh points :2.2 M
Mesh generation duration : 18 minutes
Simulation duration : 90 minutes with1 CPU by direction Simulation duration : 90 minutes with1 CPU by direction
meteodyn WT and UrbaWind : Speed up factor at h=2 and 5 m (270)
Speeding up and slowing down are stronger for UW than for WT.
Grids close to the ground are differents
meteodyn WT : Speed up factor (239)
The speed-up profil at the mast n1 is well predicted
The speed-up profil at the mast n2 is over predicted f or h<2 m
meteodyn WT : Turbulence (239)
RANS models have some difficulties to well predict the turbulence
peak at the mast 2
Correct at the three others locations
UrbaWind : Speed up factor (239)
The speed-up profil at the masts n1 and n2 are well predicted
UrbaWind : Turbulence (239)
UW predicts better the turbulence peak at the mast 2
UrbaWind : Error at h=5 m => 7% best of the one equation model
Meteodyn WT : Error at h=5 m => 11 % Not so bad for a structured mesh
Few % between the best k- and the best 1 equation model
Lack of comparison on turbulence energy errors
From Ris DTU
Further comparisons :
How long to carry out the computations
How to make the perfect mesh according the convergence and the
accuracy?
Meshing and computing time (order of magnitude)
Further works :
Explanation of the gap between meteodyn WT and UrbaWind
Mesh type grid points Meshing time
Computing time
per direction
Error on speed-up
k-l (WT) Structured 1.5 M 15 min 24 min 11%
k-l (UW) Unstructured 2.2 M 18 min 90 min 7%
k- Unstructured up to 10 M fews hours 5 h 5-10%
LES Unstructured 1 M fews hours 24 h Unknown
The first point of the grid closest to the
ground
=0.5 m, =0.2 m, and =0.1 m
Modification of the wind shear (mast 2)
meteodyn WT and UrbaWind give
roughly the same velocity profiles
15
20
25
30
Z
(
m
)
DZ= 0.5 m
DZ=0.2 m
DZ=0.1 m
Riso
roughly the same velocity profiles
Discrepancies with RANS should be more a question of grid than a
question of models when speed-up is evaluated according the
European plate form challenge
V/Vref (5m) V/Vref (2m) (I-Iref)/Vref (2m)
Riso 1.25 0.92 0.16
WT - =0.5 m 1.15 1.10 0.10
WT - =0.2 m 1.18 1.05 0.13
WT - =0.1 m 1.20 1.06 0.13
0
5
10
0.80 0.90 1.00 1.10 1.20 1.30
V/Vref
Riso
MERCI BEAUCOUP