Atmospheric Stability-Based Meso–Microscale Coupled Wind Field Simulation Method for Wind Farms in Ultra-High Altitude Complex Terrain

Accurate wind field simulation is crucial for precise wind resource assessment (WRA) and wind power development. In the ultra-high-altitude complex terrain regions of the Qinghai–Tibet Plateau, large-scale local circulations and complex wind characteristics induced by valley channeling and atmospheric stratification variations pose severe challenges to wind field simulation methods. In this paper, LiDAR observation experiments were conducted to analyze the wind characteristics in the ultra-high-altitude complex terrain of the Qinghai–Tibet Plateau. An Atmospheric Stability-based Meso–Microscale Coupling (AS-MMC) wind field simulation method was established and validated using measured data. The results indicate that, under the combined effects of local dynamic and thermal mechanisms, the wind speed in this region exhibits significant diurnal variations and typical day–night alternating mountain–valley wind circulation features. Low-altitude cold air jets and pronounced negative wind shear phenomena are generated under certain conditions at night. Compared to the Direct CFD Extrapolation (CFD-DE) method and the Non-Stability Meso–Microscale Coupling (NS-MMC) method, the AS-MMC method utilizes the atmospheric stability classes extracted from WRF results as features. By adjusting the inflow boundary conditions of the microscale CFD, it indirectly reflects the impact of macroscopic atmospheric motions on the surface layer flow field. Under unstable atmospheric conditions, the AS-MMC method achieves higher accuracy across all heights. However, when atmospheric stability conditions are not differentiated, the AS-MMC method exhibits better performance only at low altitudes. At a height of 50 m, its Root Mean Square Error (RMSE) is reduced to 2.50 m/s, which is 13.8% and 11.7% lower than those of the CFD-DE method and the NS-MMC method, respectively. The AS-MMC method improves the accuracy of refined WRA in ultra-high-altitude complex terrain, and the reduction in prediction error at a height of 50 m is statistically significant at the 95% confidence level, providing technical support for precise WRA and micro-siting in such regions.