Abstract: In this paper, the prediction performance of the Chongqing Mesoscale Ensemble Prediction System is comprehensively analyzed by using the forecast data and corresponding observation data started at 8pm from June to August in 2017, Result show that: The root mean square error of the ensemble mean prediction of 500hPa geopotential height, 850hPa temperature and 2 m temperature is better than that of control prediction. The ensemble spread is significantly lower than the root mean square error of the ensemble mean. The Talagrand distribution all showed a "J" or "U" distribution. The Outlier score of 500hPa geopotential height and 850hPa temperature ranged from 0.26 to 0.32.The Outlier score of the 2 meter temperature is generally higher, totally ranged from 0.33 to 0.67. In the aspect of Precipitation forecast. The ensemble mean and the probability matching mean have obvious advantages over the control forecast. The TS score of light rain and moderate rain precipitation of ensemble mean is better than that of probability matching mean. The TS score of rainfall and rainstorm precipitation of probability matching mean is better than that of ensemble mean. The talagrand distribution of 0～24 hours and 24～48 hours accumulated precipitation show a certain "U" distribution, which is more obvious in 0～24 hours, and the other forecast time is more ideal. The Outlier score decreases with the extension of the forecast time, the highest 0.31, the lowest 0.20. In the aspect of precipitation probability forecast, the Brier score and AROC of each precipitation magnitude are impressive.