Comparative Performance Analysis of Navier-Stokes and Lattice Boltzmann Based Wind Solver Approaches: OpenFOAM vs. waLBerla wind

Abstract

The transition to renewable energy has placed wind energy at the forefront of sustainable power generation. Simulating individual wind turbines and entire wind farms with high fidelity is critical for optimising performance and understanding complex aerodynamic phenomena. However, achieving a balance between computational efficiency and detailed predictions of aerodynamic loads and wake behaviour remains a significant challenge, particularly in large-scale wind farm simulations. This study aims to alleviate these challenges by evaluating two state-of-the-art computational fluid dynamics (CFD) solvers designed for wind turbine flow simulations: the waLBerla wind solver (Schottenhamml et al., 2024), which employs the lattice Boltzmann method (LBM) with an actuator line approach, and a Navier-Stokes-based OpenFOAM solver utilising an actuator surface method (Park et al., 2025). Using the benchmark MEXICO wind turbine case, we conduct a comprehensive comparison focusing on computational efficiency and relative accuracy of key flow predictions. Previous work (Schottenhamml et al., 2024) suggests that LBM's inherent scalability and high parallelisation potential will provide waLBerla wind with an advantage in terms of computational efficiency, particularly when moving from CPU to GPU runs. Additionally, the actuator surface approach used in the OpenFOAM solver will provide greater fidelity in terms of localised flow phenomena compared to the actuator line in waLBerla wind. This comparative analysis serves as a precursor to the development of a high-fidelity digital twin for wind turbine operations, identifying the strengths and limitations of each approach in addressing the computational and predictive demands of large-scale simulations. By determining the optimal modelling approach, this work aims to advance the adoption of physics-based digital twins in the wind energy sector, bridging critical gaps in computational speed and predictive accuracy to enhance their real-world applicability