Modeling atmosperic boundary layer with standard linear models does not sufficiently reproduce wind conditions in complex terrain, especially on leeward sides of terrain slopes. More complex models, based on Reynolds averaged Navier-Stokes equations and two-equation k-? turbulence models for neutral conditions in atmospheric boundary layer, written in general curvilinear non-orthogonal co-ordinate system, have been evaluated. In order to quantify the differences and level of accuracy of different turbulence models, investigation has been performed using standard k-? model without additional production terms and k-? turbulence models with modified set of model coefficients. The sets of full conservation equations are numerically solved by computational fluid dynamics technique. Numerical calculations of turbulence models are compared to the reference experimental data of Askervein hill measurements.

In this paper we validate an improved finite volume approximation of Reynolds Averaged Navier-Stokes equations for simulation of wind flows in body-fitted grids generated by algebraic extrusion from digital terrain elevation data, proposed in N. Mirkov et. al. J. Comput. Phys. 287, 18-45(2015), [1]. The approach is based on second-order accurate finite volume method with collocated variable arrangement and pressure-velocity coupling trough SIMPLE algorithm. The main objective is the attenuation of spurious pressure field oscillations in regions with discontinuity in grid line slopes, as encountered in grids representing highly non-uniform terrains. Moreover, the approach relaxes the need for grid generation based on elliptic PDEs or grid smoothing by applying fixed point iterations (i.e. Gauss-Seidel) to initial grid node positions resulting from algebraic grid generators. Drawbacks of previous approaches which ignored treatment of finite volume grid cell cases with intersection point offset in non-orthogonality corrections are removed. Application to real-life wind farm project at Dobric (Srvljig, Serbia) is used to assess the effectiveness of the method. The results validate the view in which accurate discretization of governing equations play more important role than the choice of turbulence modelling closures. [Projekat Ministarstva nauke Republike Srbije, br. TR-33036]