The adverse effects of typhoons on the power supply system are manifold. Accurate forecasting of typhoon wind and rain distribution can largely ensure the safe operation of the power grid, and make full use of wind resources for power generation as well. The precision and accuracy of typhoon wind and rain forecasts are contingent upon the numerical weather prediction model. In this paper, we employed the refined numerical forecast model of the Southern Power Grid to conduct a comprehensive assessment of the path, intensity, and impact of strong winds associated with Typhoon Koinu in 2023. The findings demonstrate that the Southern Power Grid refined numerical prediction model was reliable in forecasting typhoon path and intensity trends. The forecast of typhoon wind speed was systematically higher than the observation. However, the overall wind speed evolution trend closely aligned with the observed data, effectively capturing the trend of wind speed fluctuations. This suggests that the model has the potential for further refinement and optimization. The root-mean-square error and average error of the wind speed forecast were significantly reduced by revising the forecast wind speed through the random forest algorithm, thereby improving the forecast accuracy. Furthermore, wind power was forecast using the Light-GBM method, and the prediction results fit the actual situation well. This paper presents a preliminary discussion on the application and optimization of traditional numerical weather prediction models for new-type power systems.