9 / 2025-03-18 13:16:59

GPU-Accelerated Lattice Boltzmann Simulations of Wind Turbine Aerodynamics and Wake Dynamics

LBM,GPU,wind farm flow,Diffuser augmented wind turbine

Modern supercomputers are generally equipped with accelerators such as a GPU (Graphics Processing Unit), which is a many-core processor with high computing performance and high bandwidth memory. The lattice Boltzmann method (LBM) is well suited for parallel computing on GPUs due to its simple algorithm and data access locality. The excellent scalability of the fully explicit time integration of the LBM enables large-scale CFD simulations on supercomputers. We are developing a LBM framework that enables fast and scalable simulations of turbulent and free surface flows using multiple GPUs [1]. The adaptive mesh refinement method implemented in the framework improves computational cost and accuracy by locally arranging high-resolution meshes. In this study, we propose large-scale LBM simulations of wind turbine aerodynamics and wake flows. Even in using AMR method, it is difficult to resolve the boundary layer of a turbine blade using LBM with a Cartesian mesh. Therefore, the wind turbine is modeled using an actuator line approach, in which the blades are represented as a set of markers arranged in a straight line. As an example of multi-GPU LBM simulations of wind turbines, the results of LES for the entire wind farm, which consists of 80 turbines, and aerodynamic analysis of a multi-rotor system for a diffuser turbine are shown in Figure 1.