Wind forecasting is an important support for intelligent grid prediction data. Improving the accuracy of wind forecasting provides the core guarantee for wind energy and weather forecasting. Based on a comprehensive assessment of hourly 10 m wind forecasting capabilities of CMA-MESO 3 km (China Meteorological Administration Mesoscale Model at 3 km resolution) during the flooding season of 2023 in the Shanxi region, we conduct objective correction experiments based on regional and temporal differences in forecast effect, with a focus on improving the plan to address the differences in wind speed forecasting for different intensities. Objective correction of zonal wind (U) and meridional wind (V) components is carried out by applying an adaptive Kalman filtering scheme, and the correction results are also analysed in detail. The results show that: (1) The forecast errors of wind speed and wind direction occur clearly with a characteristic of diurnal variation, with one peak occurring during 18:00-20:00. Wind speed with positive forecast errors is mainly located in Xinding Basin, Taiyuan Basin, and southwest Shanxi. (2) The forecast errors of U and V (components of wind) are positively correlated with the forecast values. It is necessary to consider the temporal variation characteristics of the error in predicting wind speeds with different intensities, in order to avoid insufficient or excessive correction. (3) The correction of Kalman filtering (KM) is small and unstable, with the revised RMSE reduced by less than 6% and accuracy improved by less than 2%. (4) CBKM (Classification-based Kalman filtering method) based on dynamic classification improvement breaks the bottleneck of KM. Systematic errors are more accurately estimated and effectively corrected by CBKM. The diurnal variation characteristics of wind speed in different regions are reproduced better, and the forecast accuracy of wind direction and wind speed is improved by 8.29% and 7.92% respectively. ME tends to zero, RMSE has been slashed by 32.8%, and the correction rate of peak time is 83.49%. The forecasting capability of CMA-MESO 3 km 10 m wind is systematically evaluated to enhance objective understanding. We evaluate the spatiotemporal distribution characteristics of forecast errors, and an objective correction scheme of 10 m wind adapted to the regional characteristics of Shanxi is established. Through the above-mentioned work, we promote the application of domestic numerical model forecasting products in local refined forecasting services and provide a reference for further development of wind power forecasting services.