This study classified the weather situations of heavy precipitation events in the Yangtze River Delta Region from September 2022 to September 2024 and evaluated the forecast performance of the numerical models CMA-MESO, CMA-GFS, CMA-TYM, CMA-SH9 and the ECMWF model under four main weather types. The research results show that in the 24 h precipitation forecasts, there is a high false alarm rate of forecast for light rain, while torrential rain and above are difficult to accurately forecast, thus the TS score is low. In the 3 h forecasts, the CMA-MESO model performs best for light precipitation. However, under cold shear and low-vortex shear types, heavy precipitation becomes harder to capture, and the model forecast performance weak. In terms of spatial feature evaluation, except the CMA-SH9 model, most models have northern systematic errors in the low-vortex shear and subtropical high with low trough types in the north-south direction, but the situation is the opposite for the typhoon body and periphery type. In the east-west direction, these models generally exhibit an eastern systematic bias in the typhoon body and periphery types, while in other weather types, most models have western systematic errors. For the evaluation of temporal characteristics, all the models have the highest accuracy in forecasting the start time of precipitation, followed by the forecast of end time, and the accuracy in forecasting the peak time is relatively low.