CEID Wind Energy Seminar Feb 12 Ashvin - LUTLarge Eddy Simulation (LES) ─ Large Eddy Si l...
Transcript of CEID Wind Energy Seminar Feb 12 Ashvin - LUTLarge Eddy Simulation (LES) ─ Large Eddy Si l...
Large Eddy Simulation forLarge Eddy Simulation for Atmospheric Boundary Layer (ABL)
Ashvinkumar Chaudhari
Centre of Computational Engineering and Integrated Design (CEID)Lappeenranta University of Technology (LUT)Lappeenranta University of Technology (LUT)P.O.Box 20, FI‐53851 Lappeenranta, Finland
01 Feb 2012 www.lut.fi/ceid
IntroductionIntroduction
Th d l t f bl i ti l l f i d− The development of renewable energies, particularly energy from wind,water, and solar power, is a wide public interest nowadays.
− Renewable energy has an important role to play in reducing CarbonDioxide (CO2) emissions in environment.
− Increasing the share of renewable energy in the energy balance systemg gy gy yenhances sustainability.
− It also helps to reduce the other energy sources, such as oil and naturalgas.g
− Numerical simulation of wind flow in Atmospheric Boundary Layer (ABL) is agreat interest for engineers to help in the modeling of wind energy issuesgreat interest for engineers to help in the modeling of wind energy issues.
01 Feb 2012 2www.lut.fi/ceid
Atmospheric Boundary Layer (ABL)Atmospheric Boundary Layer (ABL)
− What is ABL?What is ABL?− Lowest region of the atmosphere, directly affected by the Earth’s surface
− The ABL plays an important role in many fields, including air pollution and the dispersal of pollutants, aeronautical, meteorology, weather forecasting, and climate studies, etc.
− Very high-Re (~107) turbulent flow, over rough wall, strongly affected by buoyancy forces
Free Atmosphere
Turbulent Eddies
Free Atmosphere
Atmospheric
Courtesy of Marcelo Chamecki
Boundary Layer : 1‐2 km
01 Feb 2012 www.lut.fi/ceid
3
CFD in environmental flowsCFD in environmental flows
M d li f i d fl i i t i f t i t t i t f i d− Modeling of a wind flow in environment is of great interest in terms of windenergy applications, such as it helps us to locate, control, and optimize thewind farms, as well as power stations, airports, etc.
− Several theoretical, experimental and numerical studies have been reportedabout modelling of a wind flow in complex terrains contain hills, forest, lake,etcetc.
− Numerical calculation of a wind flow using computational fluid dynamics(CFD) has been becoming a popular technique(CFD) has been becoming a popular technique.
− It can provide significant cost benefits for optimizing engineering designsolutions related to environmental concernssolutions related to environmental concerns.
01 Feb 2012 www.lut.fi/ceid 4
Turbulence modelsTurbulence models
− Most of the research currently taking place in the field of CFD concerns the study of turbulent flows. Almost any naturally occurring flow is turbulent, and hence it is important to be able to model turbulent flows accurately.
− Different approaches to make turbulence computationally tractable: Reynolds-Averaging Navier-Stokes Equations (RANS)
Gi di ti f th l it d th l l f• Gives a prediction of the mean velocity and the mean level of turbulent quantities
Direct Numerical Simulations (DNS) It t ll f th l t l f th t b l t ti B t thi• It captures all of the relevant scales of the turbulent motion. But this approach is extremely expensive,
• Computational cost requires for DNS is proportional to (Turbulent Reynolds number)
3RetReynolds number)
Large Eddy Simulations (LES) Detached Eddy Simulations (DES)
01 Feb 2012 www.lut.fi/ceid 5
Large Eddy Simulation (LES)Large Eddy Simulation (LES)
L Edd Si l ti (LES) i d l t i l t th t b l t fl i─ Large Eddy Simulation (LES) is model to simulate the turbulent flows inCFD where the smaller eddies are filtered and are modeled using a sub-gridscale (SGS) models, while the larger energy carrying eddies are simulated.
─ It was initially proposed in 1963 by Joseph Smagorinsky to simulateatmospheric air currents.
─ Why LES ??─ Some applications need explicit computation of accurate unsteady fields.
− Bluff body aerodynamics where the flow is governed by large turbulent scales− Bluff body aerodynamics, where the flow is governed by large turbulent scales− Aerodynamically generated noise (sound)− Atmospheric boundary layer (ABL)− Mixing
Figure courtesy of Ansys Fluent
Mixing− Combustion − Examples:
01 Feb 2012 www.lut.fi/ceid 6
Unsteady flowUnsteady flow
− Turbulent flow over a complex terrain contains complex flow characteristics suchTurbulent flow over a complex terrain contains complex flow characteristics suchas, separation, reattachment, aerodynamic instabilities, etc. Therefore, anadvance approach should be applied.Due to inherent unsteady phenomena of this flow it is difficult to model by− Due to inherent unsteady phenomena of this flow, it is difficult to model byRANS approach. Thus, time dependent computations such as DNS or LES arerequired.R l ti t RANS LES b t ti ll i i i b t 1000− Relative to RANS, LES can be computationally expensive, requiring about 1000times greater computational resources, however, it yields fidelity solutions forflow configurations where RANS fails.
Figure courtesy of Ansys Fluent
DNS
3D, unsteadyRANS
2D or 3D,steady or unsteady
LES
3D, unsteady
01 Feb 2012 www.lut.fi/ceid 7
or unsteady
Test case:LES f Ch l flLES for Channel flow (Re_tau=180)
f d d l Information on Geometry and Grid resolution in LES
− Lx= 2πδ
− Ly= 2δLy= 2δ
− Lz= πδ ,where δ is the boundary layer depth (=1).
− Mesh resolution: − Coarse: (60 x 61 X 60)=219 600
− Fine: (100 X 101 X 100)= 10 10 000
− Flow time=21.78 s, time step= 0.0015 sec
M t l (1999) h f d th DNS f f ll Moser et al. (1999) have performed the DNS of fullydeveloped channel flow at Re_tau=180, 395, 590.
Domain: 4πδ X 2δ X (4/3) πδMesh: (128 x 129 x 128)=2113536
08 Dec 2011 8www.lut.fi/ceid
Stream wise velocityStream-wise velocity
u‐ instantaneousu instantaneousu‐ instantaneous
u‐mean u‐ RMS
08 Dec 2011 www.lut.fi/ceid 9
Mean Velocity profile in wall-unit y p(Re_tau=180)
Finer MeshGrid: 60 X 61 X 60 Grid: 100 X 101 X 100
− In fine mesh case, LES has a better agreement in mean velocity with DNS.
08 Dec 2011 www.lut.fi/ceid 10
RMS velocity and Reynolds stressRMS velocity and Reynolds stress
l li d i h− Values are normalized with ut.
− Only resolved Reynolds stress
− Reasonably good agreements with DNS again.
08 Dec 2011 www.lut.fi/ceid 11
LES for air flow over 2D hillLES for air flow over 2D-hill
St d d hill t i i d t l i t i d t b− Standard hill geometry in wind tunnel experiments carried out by Khurshudyan et al. (1981).
H i h f Hill (H) 0 117− Height of Hill (H) =0.117 m − Length of Hill= a=3H− The shape of hill given by following parametrical formula:p g y g p
01 Feb 2012 www.lut.fi/ceid 12
CFD model and Grid resolutionsCFD model and Grid resolutionsFull Domain
Streamwise (X) Vertical (Y) Spanwise (Z)
Domain length 45 64 H (3xh + 20H) 10 H 6HDomain length 45.64 H (3xh + 20H) 10 H 6H
Total nodes 300 141 46
Mesh Resolution 0.153H 0.0043H‐0.251H 0.133H
Total number of hexahedral nodes=1945800
01 Feb 2012 www.lut.fi/ceid 13
Boundary conditionsBoundary conditions
− Inflow condition: Velocity given by logarithmic flow profile,Inflow condition: Velocity given by logarithmic flow profile,
hzU
hzzzku
Uinf
0* ),/ln(
Wh Th h i ht b th d R h (1 57E 04)
.0,0
WV
− Where, z= The height above the ground, z0= Roughness (1.57E-04),− K= Von Karman constant (=0.4), u*= friction velocity (=0.178 m/s),− Uinf= Free stream velocity (=4 m/s), h = Boundary layer depth (=1.17 m)
− Inflow turbulence: Turbulent intensity 12 % (was given to generatethe velocity fluctuations at inflow)
− Outlet=OutflowOutlet Outflow− Span-wise direction=Periodic boundary condition− Upper boundary= Symmetry− Lower boundary-= No-slip (Smooth Wall)Lower boundary-= No-slip (Smooth Wall)
01 Feb 2012 14www.lut.fi/ceid
Numerical method and flow statisticsNumerical method and flow statistics
FVM b d i l d ANSYS Fl t 13 0 ith th S i k Lill− FVM based commercial code ANSYS Fluent 13.0 with the Smagorinsky-Lilly sub-grid-scale model has been used.
Re nolds n mber based on hill height and free stream elocit Re =3040− Reynolds number based on hill height and free stream velocity, ReH=3040.− Frictional Reynolds number (Retau)=1187
− Pressure-velocity coupling scheme: PISO− Second order implicit solver − Discretization scheme:
− Pressure: standard− Momentum: Bounded central differencing
− Total averaging time= 37.5 sec (i.e. more than 10 x Lx)g g ( x)
− Results are time and space averaged in span-wise direction.
01 Feb 2012 www.lut.fi/ceid 15
Mean velocity and turbulence RMS u‐ velocityand turbulence RMS u velocity
Mean u‐velocity
UV Reynolds stressUV Reynolds stress
Contours of mean u‐velocity
01 Feb 2012 16www.lut.fi/ceid
LES animationLES-animation− Instantaneous stream-wise velocity (mean + fluctuation)− This shows the flow separation and reattachment behind hill
01 Feb 2012 www.lut.fi/ceid 17
Comparison with measurement p[Khurshudyan et al. (1981)]
− Mean velocity profile at several locations before and after hillMean velocity profile at several locations before and after hill.− Results are normalized by free stream velocity, uinf =4.1 (m/s)
− Even with the different Reynolds number (ReH) the overall mean velocity
01 Feb 2012 www.lut.fi/ceid 18
Even with the different Reynolds number (ReH), the overall mean velocity profile has a quantitatively good agreement at several locations.
Reynolds stress ( uv/u2 )Reynolds stress (-uv/u2inf)
01 Feb 2012 www.lut.fi/ceid 19
Reynolds stress ( uv/u2 )Reynolds stress (-uv/u2inf)
− Not so good agreement in Reynolds shear stress (-UV) near wall region, after the hill
01 Feb 2012 www.lut.fi/ceid 20
Conclusions and future workConclusions and future work
Fl t b l t l t d ft th hill t d hi hl− Flow seems to be completely unsteady after the hill, as expected highlyturbulent flow occurs with a large recirculation bubble in wake region.
− Flow separation occurs after it passes the summit, a reverse flow can be seenbehind the hill and finally it reattach at X/H=5.5. The separation region in LESis found to be a bit smaller than measurement (X/H=6.5).
− Even with the completely different Reynolds number (ReH), overall meanvelocity profile has a quantitatively good agreement with the measured data aty p q y g gseveral locations.
− Still poor agreement with measurements for turbulence quantities near wall− Still poor agreement with measurements for turbulence quantities near wallregion, which need to be investigated.
01 Feb 2012 www.lut.fi/ceid 21
ReferencesReferences
ANSYS FLUENT 12 0 Th G id A il 2009− ANSYS FLUENT 12.0 Theory Guide, April 2009− ANSYS FLUENT 12.0 User’s Guide, April 2009− CFD-Wiki: www.cfd-online.com/Wiki/
Moser Kim and Mansour (1999): DNS of Turbulent Channel Flow up to Re tau=590− Moser, Kim and Mansour (1999): DNS of Turbulent Channel Flow up to Re_tau=590,Physics of Fluids, vol. 11, 943-945.
− Fredrik Carlsson, PPT slides: “LES With FLUENT“. LES, DES, hybrid LES/RANS andURANS, Chalmers University of Technology, Göteborg, Sweden, 2006.Marcelo Chamecki Lecture notes: ”Large Eddy Simulation Applications to Meteorology”− Marcelo Chamecki, Lecture notes: ”Large Eddy Simulation Applications to Meteorology”,(2010).
− Davidson L., (2009): Large Eddy Simulations: How to evalute resolution, Int. Jou. Of Heatand Fluid Flow, vol.-30, 1016-1025.
− Allen, T. and Brown, A.R., (2002): Large-eddy simulation of turbulent separated flow overrough hills. Boundary-Layer Meteorol. 102, 177-198.
− Kim, J. J., Baik, J. J., and H. Y. Chun, (2001): Two-dimensional numerical modeling of flowand dispersion in the presence of hill and buildings Journal of Wind Engineering andand dispersion in the presence of hill and buildings. Journal of Wind Engineering andIndustrial Aerodynamics. 89, 947-966.
01 Feb 2011 www.lut.fi/ceid 22
ReferencesReferences
C t I P A l D D (1997) Fl d di i t h i− Castro, I. P., Apsley, D. D., (1997): Flow and dispersion over topography: a comparisonbetween numerical and laboratory data for two-dimensional flows. AtmosphericEnvironment. 31, 839-850.
− Khurshudyan, L. H., Snyder, W. H. and Nekrasov 1. V. (1981): Flow and dispersion ofpollutants over two-dimensional hills. United States Environmental Protection AgencyReport EPA-600/4-8 I-067.
− Tamura, T., Cao, S., Okuno, A., (2007): LES study of turbulent boundary layer over asmooth and a rough 2D hill model. Flow, Turbulence and Combustion, 79, 405-432.
01 Feb 2012 www.lut.fi/ceid 23
− Thank you for your attentions !!!y y
− Comments are appreciated !!!− Comments are appreciated !!!
01 Feb 2012 24www.lut.fi/ceid