Due to the limitations of the limited area, lateral boundary perturbation is one of the primary perturbation methods for regional ensemble prediction. However, it remains unclear how to construct lateral boundary perturbations for the high-resolution regional ensemble prediction system of China Meteorological Administration (CMA) to improve forecast skill. This paper develops a mixed lateral boundary perturbation method using the perturbation field of the CMA global ensemble prediction and the lateral boundary field of the regional deterministic model, and also adjusts the perturbation magnitude through dynamic perturbation coefficients. The results show that no lateral boundary perturbation scheme can suppress the growth of perturbation energy at later forecast periods, leading to insufficient ensemble spreads. The mixed lateral boundary perturbation scheme can enhance the perturbation energy spectra at meso-α scale and large scale, and improve the spread-skill relationships and probabilistic forecast skills for isobaric elements and precipitation. Compared to the mixed lateral boundary perturbation scheme, the dynamic mixed lateral boundary perturbation scheme can enhance the spectral energy above 100 km and improve the spread-skill relationships and the probabilistic forecast skills for low-level variables and precipitation beyond 24 h. So, it has a good potential for operational application.