Abstract: The performance of the GRAPES Global Forecasting System (GRAPES_GFS) was diagnosed and evaluated by comparing its forecasts with the fifth-generation ECMWF global climate and atmospheric reanalysis data (ERA5), and the effects of convection on cloud microphysical processes such as condensation (sublimate) and evaporation were optimized. The research emphasizes: (1) Cloud macro and micro physical quantities, such as cloud cover, cloud water content, column cloud water content etc.; (2) precipitation and outgoing long-wave radiation (OLR); (3) the feedback tendency of different physical process for temperature and humidity. The results show that after the cloud scheme is improved, liquid-water content forecasted by the GRAPES_GFS has increased in tropical regions, and the vertical distribution of hydrometers becomes more reasonable. The cloud cover forecast is closer to the ERA5. Particularly, the model performance for high cloud forecast has been improved significantly, which subsequently improves the OLR forecast. Total forecast precipitation in tropical areas increases slightly, while large-scale precipitation increase in the tropical areas is more obvious, which alleviates the underestimation of large-scale precipitation in the tropical areas.