Numerical Assessment of Steady-State RANS Models for Predicting Boundary-Layer Transition, Separation, and Wake Characteristics of Horizontal-Axis Wind Turbines

Reliable prediction of aerodynamic loads and near-wake structure of horizontal-axis wind turbines remains challenging, particularly under transitional and separated-flow regimes where steady eddy-viscosity RANS closures are still widely used in engineering practice. Yet, systematic steady-state benchmarking of transition-sensitive eddy-viscosity models remains limited. In this work, seven turbulence closures—Spalart–Allmaras, standard/RNG/realizable k–ε, k–ω SST, and the transition-sensitive k—kl–ω and γ—Reθ—are assessed against the MEXICO rotor experiments using exact blade geometry in a rotating subdomain with controlled mesh and near-wall treatment. Simulations are validated at inflow velocities of 10, 15, and 24 m s− using blade pressure distributions, sectional loads, thrust and power, and three-component near-wake velocities. The transition-sensitive k–kl—ω model provides the best overall agreement, maintaining thrust errors of 2.4–5.1% and power errors of 2.1–4.8% across the three operating conditions, at a computational cost of ≈120–172 h/CPU. Among fully turbulent closures, Spalart–Allmaras and realizable k–εe provide a competitive accuracy–cost compromise, whereas k–ω SST and RNG k–ε are inexpensive (≈18–36 h/CPU) but lose reliability under stall, with power errors reaching ≈46% and ≈28%, respectively. The γ–Reθ model shows the largest inconsistency in separated regimes, with power errors exceeding 50% at 24 m s−. These results quantify the accuracy–cost trade-off of common eddy-viscosity RANS models and provide practical guidance for steady wind-turbine CFD predictions.