Abstract:Wind power holds a significant role in the transformation of China''s energy structure, yet its volatility also poses severe challenges. The wind speed data, forecasted by numerical models, serves as a crucial foundation for predicting wind power output and its efficient consumption. Therefore, it''s necessary to evaluate the forecasting effects of different models. In this study, we compare and analyze the wind speed prediction effects of four mainstream numerical models. We aim to thoroughly assess their prediction accuracy in different regions and under varying conditions during winter in China. This is done with the goal of providing a reference for wind speed prediction during China''s winter gale period. Relying on four numerical prediction models with different resolutions, initial fields, and assimilation schemes, and in conjunction with observation data from 131 stations across China, we investigate and analyze the characteristics of error distribution and the forecasting capability of the predicted wind speeds. Simultaneously, we focus on typical stations to analyze the characteristics of forecasting errors and the forecasting capability across different wind speed bands and regions. The results indicate that the forecast results of the ensemble forecast model are more scientifically sound under complex terrain conditions. The high-fraction single model excels at forecasting the volatility of wind speed in simple subsurface conditions. The forecast effect is superior during daytime compared to nighttime, and the forecast effect is optimal in plains wind speed.