Abstract: The impact of model perturbation technique that is applied to medium and long-term ensemble forecast is not clear when applied to the storm-scale ensemble forecasting system. In order to explore the possible impacts of this technique on storm-scale ensemble forecasting, three perturbation schemes are designed to study a severe convection weather process in Changjiang-Huaihe River basin from 5-6 July 2013 using the Weather Research and Forecasting (WRF) model. These perturbation schemes include the multi-physics (MP) scheme, the Stochastically Perturbed Parameterization Tendencies (SPPT) scheme and the SPMP scheme. The SPMP scheme is a mixing of MP and SPPT. Results show the precipitation probability score of the MP scheme is better than that of the SPPT in the early period of integration and the effect of MP scheme in the upper atmosphere disturbance is more reasonable. The SPPT scheme mainly affects simulation in the middle and later periods of integration and performs well in the lower atmosphere and over the land, especially for the simulation of surface water vapor field. The SPMP can significantly improve the spread of forecast variables and reduce the root mean square error in the upper troposphere, and thus enhances the credibility of the ensemble members. The use of SPMP scheme can effectively compensate for the shortcomings that occur when only using the MP or SPPT scheme in different integration periods. The propagation of perturbation stream function in the SPMP is mainly dominated by MP in the horizontal direction. In the vertical direction, the pattern of SPMP in the lower layer is consistent with that of SPPT, but it is controlled by MP in the upper layer. The analysis of spectral energy shows the energy has a tendency to propagate from small scale to large scale with increasing integration time. And SPMP can effectively compensate for the energy dissipation shown in the other two schemes in different scales.