CCS335 _ Neural Networks and Deep Learning Laboratory_Lab Complete Record
Development of a Pseudo-Spectral 3D Navier Stokes Solver for Wind Turbine Applications
1. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Development of a Pseudo-Spectral 3D Navier
Stokes Solver for Wind Turbine Wakes
Emre Barlas
Technical University of Denmark
s110988@student.dtu.dk
May 26, 2014
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 1/57
2. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Overview
1 Introduction
2 Fourier & Chebyshev SM
3 3D Navier-Stokes
Time & Spatial Discretization
4 Comp. Domain & W.T. Representation
Computational Domain
Wind Turbine Representation Methods
5 Validation
AD-NR
AD-R
6 Simulations
7 Conclusions & Future Work
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 2/57
3. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Introduction
∂ρ
∂t
+ · (ρu) = 0 + ρ
∂u
∂t
+ u · u = − p + µ 2
u + f
Approximate with;
• Finite Difference FD
• Finite Volume FVM
• Finite Element FEM
• Spectral Methods FUNCTIONAL
• Spectral Element Methods SEM
• Vortex/Particle VP
etc.
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 3/57
4. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
∂ρ
∂t
+ · (ρu) = 0 + ρ
∂u
∂t
+ u · u = − p + µ 2
u + f
Approximate with;
• Finite Difference DISCRETE
• Finite Volume DISCRETE
• Finite Element FUNCTIONAL
• Spectral Methods FUNCTIONAL
• Spectral Element Methods HYBRID
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 4/57
5. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Approximate with;
• Finite Difference DISCRETE LOCAL
• Finite Volume DISCRETE LOCAL
• Finite Element FUNCTIONAL LOCAL
• Spectral Methods FUNCTIONAL GLOBAL
• Spectral Element Methods HYBRID
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 5/57
6. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
∂ρ
∂t
+ · (ρu) = 0 + ρ
∂u
∂t
+ u · u = − p + µ 2
u + f
Approximate with;
• Finite Difference FD
• Finite Volume FVM
• Finite Element FEM
• Spectral Methods GLOBAL+FUNCTIONAL
• Spectral Element Methods (maybe later)
etc.
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 6/57
7. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Polynomial Approximation
Replacing;
u(x) = uN(x) =
N
k=0
ˆuk φk
results with a residual/error;
Lu(x) = f (x) ⇒ RN(x) = LuN(x) − f (x) = 0
minimize the residual/error;
(RN)w :=
Ω
RN(x)ω(x)dx = 0,
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 7/57
8. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Spectral Collocation
RN(xk) = LuN(xk) − f (xk), 1 ≤ k ≤ N − 1
uN(x0) = g−, uN(xN) = g+
Plug this; uN(x) =
N
j=0
uN(xj)hj(x) into the above equations results;
N
j=0
[Lhj(xk)]uN(xj) = f (xk), 1 ≤ k ≤ N − 1
N
j=0
[hj(x0)]uN(xj) = g−,
N
j=0
[hj(xN)]uN(xj) = g+
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 8/57
9. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Spectral Collocation
um
N (xk) =
N
j=0
dm
kj uN(xj), where dm
kj = hm
j (xk)
The matrix Dm = (dm
kj )k,j=0,...,N
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 9/57
10. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Fourier Approximation & Differentiation
The approximation of a real, integrable periodic function with
truncated Fourier series;
uK (x) =
K
k=−K
ˆukeikx
via orthogonality;
2π
0
eikx
e−ilx
dx =
2π if k = l
0 if k = l
coefficients;
ˆuk =
1
2π
2π
0
u(x)e−ikx
dx, k = 0, ±1, ±2, . . .
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 10/57
11. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Fourier Approximation & Differentiation
Nth order truncated Fourier series
PNu(x) =
N/2−1
k=−N/2
ˆukeikx
For 2π (not mapped) periodic, the collocation grid points;
xj =
2πj
N
, j = 0, . . . , N − 1
the coefficients (again via orthogonality);
ˆuk =
1
N
N−1
j=0
u(xj)e−ikxj
, k = −N/2, . . . , N/2 − 1
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 11/57
12. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Fourier Approximation & Differentiation
1D Burgers Equation;
∂u
∂t
= v
∂2u
∂x2
− u
∂u
∂x
The discretized version;
1
∆t
(ˆun+1
k − ˆun
k ) = −vk2
ˆun+1
k − ˆun
k ikˆun
k
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 12/57
13. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Fourier Approximation & Differentiation
Figure: 1D Burgers equation solved with increasing modes
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 13/57
14. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Chebyshev Approximation & Differentiation
Basically Substituted ’Cosine’ functions;
Tk = cos(kz) where x = cos(z)
Possible FFT usage,or DCT
Suitable for Non-Periodic B.C.
Common collocation points;
Gauss (Chebyshev zero points)
xi = cos (i + 1
2 )π
k , i = 0, . . . , k − 1
Gauss-Lobatto points(Chebyshev extreme points)
xi = cos iπ
k , i = 0, . . . , k
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 14/57
15. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Chebyshev Approximation & Differentiation
’Collocated’ on the Gauss-Lobatto points;
u(xi ) = uN(xi ) =
N
k=0
ˆukTk(xi ) =
N
k=0
ˆuk cos(
k π i
N
), i = 0, . . . , N
via orthogonality the coefficients are;
ˆuk =
2
ck N
N
i=0
1
ci
ui cos(
k π i
N
), k = 0, . . . , N
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 15/57
16. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Chebyshev Approximation & Differentiation
1D Advection-Diffusion;
∂u
∂t
+ u
∂u
∂x
= v
∂2u
∂x2
Discretized version;
3
2∆t
ut+1
−
2
∆t
ut
+
1
2∆t
ut−1
− 2ut
Dut
+ 2ut−1
Dut−1
= vD2
ut+1
+ RHSt+1
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 16/57
17. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Chebyshev Approximation & Differentiation
Figure: 1D Advection-Diffusion solutions with various Chebyshev modes
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 17/57
18. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Chebyshev Approximation & Differentiation
10
1
10
2
10
−6
10
−5
10
−4
10
−3
10
−2
10
−1
10
0
N
Error
Figure: Convergence of Chebyshev method for 1D advection-diffusion
equation
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 18/57
19. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Pause & Recap
Introduced & Assessed the methods → SM,FD,FEM,FVM
Picked the method → S.M. - Collocation
Manipulated the methods individually → Fourier & Chebyshev
Next → Navier Stokes Implementation
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 19/57
20. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Time & Spatial Discretization
Splitting Algorithm
Velocity Step
1
2∆t
3un+1/2
− 4un
+ un−1
−
1
Re
∆un+1/2
= 2hn
− hn−1
+ f
un+1/2
= gn+1/2
where h = u · u
Pressure Step
1
∆t
(un+1
− un+1/2
) + pn+1
= 0
· un+1
= 0
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 20/57
21. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Time & Spatial Discretization
Staggered Grid
Figure: Staggered grid - [Canuto et.al.,Springer,2007]
yj = cos
jπ
Ny
, j = 0, ......, Ny (GL)
yj+1
2
= cos (j +
1
2
)
π
Ny
, j = 0, ......, Ny − 1 (G)Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 21/57
22. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Time & Spatial Discretization
Velocity & Pressure Representation
u(x, y, z, t) =
Nx /2−1
kx =−Nx /2
Ny
m=0
Nz /2−1
kz =−Nz /2
ˆukx,m,kz
¯Tm(y)e2πi(kx x/Lx+kz z/Lz)
p(x, y, z, t) =
Nx /2−1
kx =−Nx /2
Ny−1
m=0
Nz /2−1
kz =−Nz /2
ˆpkx,m,kz
¯Tm(y)e2πi(kx x/Lx+kz z/Lz)
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 22/57
23. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Time & Spatial Discretization
Pseudo-Code
1.Pre-Process
Grid,Re,Diff. Matrices,Operators
2.Turbine Preparation
Nodes,Airfoils,Indexing
3.Solver
1: for Time Marching do
2: Non-Linear Terms(Pseudo-Spectrally)
3: for each mode in X do
4: for each mode in Z do
5: Solve the system for intermediate vel where the forces are fed in.Then
solve for new pressure via that find the divergence free velocity
6: end for
7: end for
8: Check Continuity
9: Transform to the real space & update the forces
10: end for
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 23/57
24. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Computational Domain
Computational Domain
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 24/57
25. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Wind Turbine Representation Methods
AD-NR
UNIFORM FORCE DISTRIBUTION & NO ROTATION
LESS TIME CONSUMING
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 25/57
26. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Wind Turbine Representation Methods
AD-R
The velocity components at the rotor disc are extracted from
the flow solver, Vx ,Vy ,Vz (for 3D Cartesian coordinates)
The inflow angle was calculated considering the angular
velocity of the turbine,Vθ (projected from Vy & Vz) and Vx
φ = tan−1 Vx
ωr + Vθ
The local twist and pitch is subtracted from inflow angle in
order to find the angle of attack.
α = φ − γ
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 26/57
27. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Wind Turbine Representation Methods
AD-R
Via the look up tables the lift and drag coefficients
(Cl (α, Re), Cd (α, Re)) are stored.
The forces acting on the rotor disc are found , by considering
an annular area of differential size dA = 2πrdr. The resulting
force per unit rotor area is;
dF
dA
=
ρV 2
rel
2
Bc
2πr
(Cl el + Cd ed )
FORCE DISTRIBUTION W.R.T. LOCAL CHARACTERISTICS &
WITH ROTATION
MORE TIME CONSUMING & STILL A DISC NOT ACL
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 27/57
29. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
AD-NR
ValidationAxialInductionFactor(a)
Disk Resolution (N)
20 40 60 80 100 120 140 160
0.28
0.3
0.32
0.34
Re 1000
Re 3000
Axial Induction Factor (a)
Ct&Cp
0 0.1 0.2 0.3 0.4 0.5
0
0.2
0.4
0.6
0.8
1
Ct-Theory
Cp-Theory
Computed
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 29/57
30. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
AD-NR
Brief Results
−6 0 4 9
0
1
V
Vo
−6 0 4 9
−0.2
0
P
x
D
Pressure
Velocity
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 30/57
31. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
AD-NR
Brief Results
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 31/57
32. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
AD-NR
Brief Results
V
V o
0.5 1
0
1
2
3
4
X
D =-2
y
D
0.5 1
X
D =0
y
D
0.5 1
X
D =1
y
D
0.5 1
X
D =2
y
D
0.5 1
X
D =4
y
D
0.5 1
0
1
2
3
4
X
D =6
y
D
0.5 1
X
D =10
y
D
0.5 1
X
D =14
y
D
0.5 1
X
D =18
y
D
0.5 1
X
D =20
y
D
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 32/57
33. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
AD-NR
Brief Results
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 33/57
34. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
AD-NR
Brief Results
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 34/57
35. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
AD-NR
Reynolds Number Effect
Figure: Contour of the streamwise velocity component at different
Reynolds Numbers; 500, 1000, 2500, 5000, 10000 (from top to bottom)
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 35/57
36. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
AD-R
Validation
Power Curve
Measurements
ACL−Troldborg
AD−R
Figure: Comparison of measured and computed power coefficient for the
Tjaereborg wind turbine
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 36/57
37. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
AD-R
Validation
0
0.1
0.2
r/R
AxialInduction
0 0.2 0.4 0.6 0.8 1
0
0.1
r/R
Circulation
Figure: Loading (1- V
Vo ) and Circulation ( Γ
RVo ) along the blade
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 37/57
38. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
AD-R
Validation
(3dvo)
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 38/57
39. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
AD-R
3D Vorticity Field
(vorfield)
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 39/57
40. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
AD-NR & AD-R Simulations
Simulations 1 & 2 & 3 & 4
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 40/57
41. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
AD-NR & AD-R COMPARISON
(adrcompnew)
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 41/57
42. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
AD-NR & AD-R COMPARISON
Figure: Contours of the streamwise velocity. Comparison of AD-NR
(Top), AD-R (Bottom)
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 42/57
43. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
AD-NR & AD-R COMPARISON
¯u
V o
0.6 1
X
D =15
y
D
0.6 1
X
D =12
y
D
0.6 1
X
D =10
y
D
0.6 1
X
D =7
y
D
0.6 1
0
1
2
3
4
X
D =4
y
D
0.6 1
X
D =3
y
D
0.6 1
X
D =2
y
D
0.6 1
X
D =1
y
D
0.6 1
X
D =0
y
D
0.6 1
0
1
2
3
4
X
D =-3
y
D
AD-NR
AD-R
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 43/57
44. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
AD-NR & AD-R COMPARISON
Figure: Contours of the streamwise turbulence intensity, σu
¯u . Comparison
of AD-NR (Top), AD-R (Bottom)
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 44/57
45. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
AD-NR & AD-R COMPARISON
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 45/57
46. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Reynolds Number Effect , AD-R
(Re500)
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 46/57
47. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Reynolds Number Effect , AD-R
Figure: Time averaged stream-wise velocity, ¯u
Vo ([∼]), at the middle
vertical plane. Comparison Re-500 (Top), Re-2000 (Bottom)
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 47/57
48. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Reynolds Number Effect , AD-R
Figure: Contours of stream-wise turbulence intensity,σu
¯u ,at the middle
vertical plane. Comparison Re-500 (Top), Re-2000 (Bottom)
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 48/57
49. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Reynolds Number Effect , AD-R
Figure: Contours of turbulence kinetic energy (u )2+(v )2+(w )2
2 at the
middle vertical plane. Comparison Re-500 (Top), Re-2000 (Bottom)
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 49/57
50. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Mixing Effect , 2 Turbines, Laminar&Disturbed
(2TurbineYESTurb)
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 50/57
51. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Mixing Effect , 2 Turbines, Laminar&Disturbed
Figure: Time averaged stream-wise velocity, ¯u
Vo [], at the middle vertical
plane. Comparison Perturbed Inflow (Top), Laminar Inflow (Bottom)
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 51/57
52. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Mixing Effect , 2 Turbines, Laminar&Disturbed
¯u
V o
0.6 1
X
D =17
y
D
0.6 1
X
D =14
y
D
0.6 1
X
D =12y
D
0.6 1
X
D =9
y
D
0.6 1
0
1
2
3
4
X
D =6
y
D
0.6 1
X
D
=3
y
D
0.6 1
X
D
=2
y
D
0.6 1
X
D
=1
y
D
0.6 1
X
D
=0
y
D
0.6 1
0
1
2
3
4
X
D
=-3
y
D
Turb-Inflow
Uni-Inflow
Figure: Time averaged stream-wise velocity, ¯u
Vo [], vertical profiles at
various downstream positions. Comparison Perturbed Inflow & LaminarEmre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 52/57
53. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Mixing Effect , 3 Turbines, Laminar&Disturbed
Figure: Time averaged stream-wise velocity, ¯u
Vo [], at the middle vertical
plane. Comparison Perturbed Inflow (Top), Laminar Inflow (Bottom)
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 53/57
54. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Staggered
(Staggered)
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 54/57
55. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Conclusions
Spectral methods are very convenient for such flows. Fast
codes can be developed
Chebyshev grid distribution was not very suitable for a case
where the boundaries are not of paramount interest
Under uniform inflow conditions both models perform
similarly, in terms of wake modelling apart from the near wake
region
Axis symmetric wake development were captured
TI-The tips are the regions where the highest turbulence
occurs. If the tower was modelled this might have been valid
for only ’upper-side’
Reynolds number’s role for wake development for laminar and
perturbed inflow are different
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 55/57
56. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
Future Work
In order to reach more realistic atmospheric Reynolds numbers
with reasonable grid points, it is required to implement a
turbulence model to this code.
The boundary layer inflow.
Taking energy equation into account in order to see the
atmospheric stability effect in the wind turbine/farm wakes.
Chebyshev grid issue.
Tip correction for more detailed loading investigations
ACL,sacrificing from computational time
Continuous and controlled turbulence should be provided to
the flow.
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 56/57
57. Introduction Fourier & Chebyshev SM 3D Navier-Stokes Comp. Domain & W.T. Representation Validation Simulations Concl
The End
QUESTIONS & REMARKS
Emre Barlas DTU-Wind Energy
Spectral 3D Navier Stokes Solver for WT Wakes 57/57