Modeling wind flow using O.F. Wind, an OpenFOAM based CFD tool: validation of Turbulence Intensity in a testing Suzlon Energy site Ltd. L.Casella 1, W.Langreder 1, A.Fischer 2, M.Ehlen 2, D.Skoutelakos 2 1 Suzlon Energy A/S, Wind and Site Competence Centre, Aarhus, DK 2 IB Fischer CFD+engineering GmbH, München, GE Presenting author: Livio Casella, PhD First Symposium on OpenFOAM in Wind Energy 2013 20-21 March 2013 - Oldenburg 1
Testing Site Coastal area consisting of a platau bounded by cliffs at west and a range of hills at east Six masts used for the validation 2
Testing Site Geometry (stl format): the domain is split in terrain and 2 refinement areas 3
Meshing overview Max Edge lenght Box resolution above the surface layer refinement Number of elements 200m 12m 16.9M 260m 16m 8.8M 400m 24m 2.4M 800m 48m 0.5M 4
Mesh Convergence analysis Fig. 3. Convergence grid analysis for SST closure scheme: results of TI at 80m AGL at different points inside the domain obtained using different grid resolution. 5
Simulation set up Two directional sectors of the wind rose, SSE and S have been calculated for the testing case, running three different angles using step of 10 degrees (from 140 to 190 degrees). Oveserved wind rose at M5 (2 years of data) Neutral stable atmospheric condition is considered Coriolis terms are neglected in the momentum equations Air density is considered constant at 1.225 (kg/m 3 ) Atmospheric flow is assumed to be incompressible 6
Turbulence closure schemes SST (Shear Stress Transport) with standard parameters [4] K-epsilon in two configurations k-e [7], k-e_mod [5] Table 1. Values of the model constant used in the k-epsilon scheme. c c 1 c 2 k k-e 0.09 1.44 1.92 1.00 1.11 k-e_mod 0.03 1.21 1.92 1.0 1.30 [4].http://www.cfd-online.com/Wiki/SST_k-omega_model [5] D.M. Hargreaves, N.G. Wright, On the use of the k e model in commercial CFD software to model the neutral atmospheric boundary layer, Journal of Wind Engineering and Industrial Aerodynamics, 95, 355-369, (2007) [7] Aspley&Castro. The Apsley and Castro limited-lenght scale k-e model revisited for improved performance in the atmospheric surface layer, Boundary Layer Met. 144, 199-215, (2012) 7
Two used definitions of TI in comparison TI u h 2 u' v' u h 2 (4/ 3)* TKE u h Eq.1 TI u u u ' 2 (2/3)* TKE u h Eq.2 [1] ( u u ) 2 N i N u' ; v' i1 i1 N ( v v ) N i 2 TI= wind power industry definition of turbulence TKE= CFD definition of turbulence [1] C. Abiven, J. M. L. M. Palma, O.Brady, Time Dependent CFD Analyses of Wind Quality in Complex Terrain, Proceeding EWEC 2009. 8
Results Fig. 4. Turbulence Intensity from Eq.1 and horizontal wind vector obtained from SST (left panel) and k-e_mod (right panel) for an inlet wind direction of 150 degrees 9
Results Turbulence Intensity from Eq.1 and horizontal wind vector obtained k-e_mod for an inlet wind direction of 190 degrees 10
shifting of wind direction toward south in the centre (M3) and in the north part of the site (M4), where high TI is predicted Fig. 5. Turbulence Intensity (upper panel) and wind direction (lower panel) observed at met-masts M2, M4 and M5 Wind Rose from M5 show higher frequency at SSE Wind Rose from M4 and M5 higher frequency at SSW (in agreement with the results at 190 degrees) 11 (a) (b) (c) Fig. 6. Wind roses observed at met-masts M5 (a), M3 (b) and M4 (c)
Quantyfing the score Fig. 6. BIA values of TI (in %, y ais) obtained at the six met-masts (x axis) SST_Eq.1 SST_Eq.2 ke_mod_ Eq.1 ke_mod_ Eq.2 ke_ Eq.1 ke_ Eq.2 WBIAS(%) 4.26-0.34 10.25 3.90 3.73-0.71 WBIAS WRMSE n l l S M ( fij tij tij /( i1 j1 j1 ( n l l S M 2 fij tij tij /( i1 j1 j1 f f ij ij ))/ n ))/ n WRMSE (%) 5.45 2.43 11.94 5.41 4.76 2.52 Conclusion (1) Eq.1 always overestimates the measured values for all the closure schemes. SST scheme, when using Eq.2, exhibits the best score ke_mod model: worst score for both WBIAS and WRMSE when using Eq.1. All the used closure schemes achieve better performance using Eq.2 instead of Eq.1 12
Model Error correction methods: are they convenient? Two correction methods from SST_Eq1 results, using i=m5 validation analysis, are presented: 1. BIA method WBIA WL i l l S M r S i fij tij tij /( fij)) toi to WBIAi j1 j1 2. Linear method l l M S R S fij tij / tij /( fij)) t oi to * WL i j1 j1 o=m1 o=m2 o=m3 o=m4 o=i=m5 o=m6 AE s 2.2% 21.6% 16.2% 5.3% 26.6% 23.4% AE r 9.9% 39.8% 8.6% 19.0% 0% 41.5% AE R 19.3% 38.1% 8.2% 25.2% 0% 39.5% Conclusion (2) Both methods do not improve the model score in all the cases except for o=m3 and, of course, o=m5. BIA method seems to work better compared to the linear one when predicting the met-masts M1 and M4 which are located in high turbulence zone. A linear regression analysis, using five masts to predict at M4, does not even improve the final accuracy (not show) 13
Thank You Livio Casella Wind & Site Engineer, PhD Email: livio.casella@suzlon.com Suzlon Wind Energy A/S,Bredskifte Allé 13 DK-8210 Aarhus V 14