The heavy rainstorm process in Zhengzhou on July 20,2021 caused huge losses to people "s lives and property. However, the precipitation of this rainstorm is extremely strong, and the uncertainty of numerical prediction is great. In order to improve the ability of CMA numerical model to describe the uncertainty of extreme rainfall intensity forecast of such rainstorm, the 3 km horizontal resolution convective scale numerical forecast system (CMA-Meso 3km) is used to carry out the convective scale ensemble forecast experiment by designing the ensemble assimilation observation disturbance scheme and the cloud analysis radar reflectivity filtering threshold adjustment scheme. The uncertainty characteristics of extreme rainfall intensity forecast of convective scale ensemble forecast are evaluated and compared with CMA global ensemble forecast and regional ensemble forecast. The results show that: 1) there is still a certain deviation in the forecast of extreme precipitation based on the convective scale ensemble forecast with 3km horizontal resolution, but the extreme rainfall intensity values of different ensemble forecast members have good spread, and the extreme rainfall intensity values of individual members are close to the actual 624.1mm, which better represents the forecast uncertainty of extreme precipitation. 2) By comparing the extreme rainfall intensity, dispersion and cumulative precipitation neighborhood fraction skill scores of CMA ensemble forecasts with different resolutions, it is shown that the extreme rainfall intensity and dispersion are closely related to the horizontal resolution of the model. The higher the resolution, the higher the probability forecast skill of extreme rainfall intensity indicating that the convective scale ensemble forecast can better describe the uncertainty and extremeness of extreme rainfall intensity forecast. 3) The convective-scale ensemble forecast that simultaneously disturbs the conventional observation data and radar data can affect the model water vapor field and circulation situation in a short forecast period, which promotes the development of precipitation forecast ensemble dispersion in a short forecast period and effectively improves the probability forecast ability of extreme rainfall intensity. In general, the convective-scale ensemble forecast by disturbing conventional observation data and radar data is helpful to improve the probability forecast skill of extreme rainfall intensity.