Wind Turbine Simulations Using CPU/GPU Heterogeneous Computing

In this study, a heterogeneous solution framework using both CPUs and GPUs was used to numerically simulate flow over the National Renewable Energy Laboratory (NREL) Phase IV horizontal-axis wind turbine. An in-house line-based unstructured flow solver implemented on CPUs was coupled to an in-house structured solver implemented on GPUs via a lightweight Python-based framework within an overset mesh system. First, computations were conducted for an isolated rotor at three different wind speeds of 7 m/s, 10 m/s, and 20 m/s, and subsequently full wind turbine simulations that included the nacelle and the tower. The entire system was used to understand the blade-tower interference on both upwind and downwind configurations, and the predictions were compared with the experimental data in terms of blade airloads. The effects of the laminar-turbulent transition were also investigated on a blade using the two-equation transition model coupled with Spalart-Allmaras turbulence model, whose inclusion resulted in a more accurate torque prediction. The downwind tower interaction was much more severe than the upwind interaction on the blade owing to its blade-wake interaction. Finally, a normal wind profile model was used to simulate the freestream wind shear during the wind turbine operation in an atmospheric boundary layer. Even a small variation in the wind speed resulted in a high level of unsteadiness in the blade airloads, which could generate vibratory loads on the wind turbine.