Considerations in coupling LES of the atmosphere to CFD around wind turbines

We explore the issue of coupling the inflow from Large Eddy Simulation (LES) of the Atmospheric Boundary Layer (ABL) with CFD around the wind turbine. We perform LES of the daytime ABL using a spectral algorithm. However a finite volume algorithm is necessary to simulate the flow around the wind turbine. In the first part of this paper, we analyze the role of algorithm in the inaccurate predictions of the mean shear in the surface layer of high Reynolds number flows. Brasseur and Wei¹ have proposed a solution to this problem in the R - ReLES parameter space. We perform the same simulation using two algorithms with different numerical dissipation characteristics, viz., a spectral algorithm and a finite volume algorithm. We repeat this procedure for two incompressible high Reynolds number flows: channel flow and a neutral atmospheric boundary layer. The increased dissipation in the finite volume algorithm compared to a spectral algorithm acts as an effective filter cutoff at a lower wavenumber than the grid can represent. We find that the finite volume algorithm experiences a lower resolved stress, increased variance in the streamwise velocity and lower variance in the vertical velocity in the surface layer. These combine together to affect the turbulence structure in the surface layer that then affects the whole boundary layer. We use the expertise gained from the above analysis to design an LES of the flow around a hypothetical wind turbine. The finite volume algorithm used for this purpose is built using the OpenFOAM² framework. This work is a step in the direction of our ultimate aim: a fully resolved CFD of the flow around a wind turbine in day time ABL using hybrid URANS/LES. There has been no systematic study of the propogation of imposed ABL turbulence through such domains. We carry out a series of tests to explore the effect of grid resolution, boundary conditions and a novel algorithm on the propogation of the imposed turbulence structures through the domain. The profiles of plane averaged resolved velocities and correlations from the finite volume simulations compare well with those from the spectral algorithm. However the finite volume algorithm causes a lower resolved stress near the surface as expected.