This paper uses the GFS (Global Forecast System) global numerical weather forecast product of NCEP (National Centers for Environmental Prediction) as the initial field of the model forecast, and uses the regional mesoscale forecast system CMA-MESO (China Meteorological Administration Mesoscale Model) (formerly GRAPES_MESO) version 5.1 to analyze a rainstorm process that happened in Sichuan Basin on September 3-5, 2021, Five groups of experiments were designed with three different resolutions (1km、3km、10km) and two cloud microphysical parameterization schemes (WSM6、Thompson) for numerical simulation,The results show that: (1) The simulated rain belt in the test is basically consistent with the actual situation, but there are differences between the time, area and intensity of heavy rainfall. With the increase of precipitation threshold, TS score decreases while Bias amplitude increases, which lead the rate of false report and missed report also increases. (2) There is no significant difference in the simulation results of water vapor flux between the same resolution cumulus parameterization scheme and different microphysical schemes; There are strong updraft in the simulated rainstorm regions of the five groups of experiments, and the simulation intensity increased with the increase of resolution. (3) The simulation results of liquid particles with different cloud microphysics schemes at the 1km resolution are similar, but the results of solid particles are obviously different. (4) When cumulus parameterization scheme is added to the 3km resolution, there is a large deviation for the simulation results of heavy precipitation center. On the whole, the experimental simulation results of the precipitation process show that under the condition of high resolution, the saturation adjustment scheme of Thompson scheme is slightly better than that of WSM6 scheme, and the 1km_thompson scheme is more accurate in depicting rain belts, and the precipitation simulation is the best.