Posters

Abstract

Computational Fluid Dynamics (CFD) is currently considered as a relevant approach to obtain high fidelity predictions of wind turbine flows that can then be used to improve engineering tools for the design and control of wind farms. Many approaches are however present to address such flow high-fidelity prediction. In the following, two of such CFD methods are used to predict wind turbine wakes and compared. To do so, the retained configurations are wind turbines in a wind tunnel simulated by use of the Lattice-Boltzmann Method (LBM) and a Navier-Stokes resolving method both approaches making use of the so-called the Large Eddy Simulation (LES) formalism. First, LBM takes good advantage of High-Performance Computing (HPC) and is known to have low dissipative properties, yet this method needs further validation for wind turbine computations. Note that in this LBM framework, the wind turbine is modelled by construction with an Actuator Line method (ALM) implemented in WaLBerla along with a proper sub-grid scale model and law of the walls. Second, since conventional (i.e. finite volume or finite element based) LES is known to accurately predict flows around complex rotating geometries - aircraft turbines and compressors - at an affordable cost, it is also used in the following as a reference. In this last framework, the full geometry of the wind turbine is computed with the chimera method taking care of the rotating parts. Law of the walls and sub-grid scale model are applied to all flow boundary layers. note also that a static mesh refinement procedure developed to adapt the mesh to the flow as well as artificial incompressibility are applied for efficient return time. Two configurations of wind turbines in a wind channel are simulated as test benches. One consists in a single wind turbine under a uniform inlet velocity profile with low turbulence intensity. The other one consists in two in line wind turbines under a similar inflow as for the previous configuration. Comparisons are conducted between both numerical methods and experimental measurements for the velocity field along horizontal lines at various diameters behind the wind turbines. Results show that the coupled LBM-ALM approach predicts well both first order and second order statistics of the flow and their evolution in the wind turbine wake as long as the proper bodies of influence of the wind turbine are present in the simulation. In particular, the transition between the near and far wakes is well captured. In addition, relevant structures of the flow such as the rotation of the tower wake up to the horizontal plane are well-recovered. When compared to the conventional LES solver, both methods compare well with the experiment. Yet and thanks to its high efficiency, specific incompressible formalism and actuator type modeling, LBM yields lower computational costs for the addressed wind turbine models.