Abstract: By using station observations and 12~36h precipitation forecast of EC, EC revised field (ECR), CMA_3KM and SWC_3KM models, based on TS score, SAL verification and other indicators, the effects of multi-model precipitation forecast in 2022 flood season in Sichuan Province were experimented and compared. The results show that: (1) The forecast of rainy days in SWC_3KM was the closest to the actual distribution, and the EC model had the most obvious wet deviation about the rainy days, especially in the western Sichuan Plateau and Panxi region. The EC model was better than other models in predicting heavy rain days. (2) Bias score showed that EC model predicted less heavy rain days than observation, and the other models were mainly wet. The ECR field got a higher TS score of rainstorm magnitude in each month. (3) Comparison of individual scores showed that ECR had the most stable prediction effect, the number of highest TS score in precipitation process was the most, and the SWC model was the least. (4) The precipitation intensity and the location and trend of rain band of ECR were the closest to the actual situation for the case forecast, while the EC model forecast was weak. The SWC_3KM model had a high reference value for the location of heavy rainfall rain belt in the northwestern part of the Basin and the northern part of Liangshan Prefecture. There was a wide range of heavy rain false alarm rate of CMA_3KM and SWC_3KM in high altitude areas.