Abstract: Many parameters are involved in cloud microphysics schemes, and their constraint relationships are complex. As a result, choosing parameters and designing parameter perturbation combinations for ensemble forecasting remains unclear. This study, based on the Weather Research and Forecasting (WRF) v4.2 model, selects eight parameters from the WRF single-moment 6-class microphysics scheme (WSM6) to perform ensemble prediction experiments for parameter perturbations during a heavy rainfall event associated with the Meiyu front in 2020. The impacts of different parameters and perturbation range combinations on model performance are systematically assessed. Additionally, joint perturbations of sensitive parameters in both the cloud microphysics scheme and the boundary layer scheme are analyzed. The findings show that combined perturbations of four sensitive microphysics parameters give the best results, reducing the false alarm ratio in torrential rainfall forecasts and enhancing ensemble spread for humidity, zonal wind, low-level temperature, and mid-tropospheric meridional wind. The combined perturbations of the raindrop size distribution truncation parameter and the maximum cloud ice diameter demonstrate a synergistic effect, improving forecast skill. Conversely, combined perturbations involving graupel-related parameters (including truncation parameters and density) do not increase perturbation energy with additional parameters; instead, antagonistic parameter interactions limit forecast improvements. Perturbing three parameters in the WRF boundary layer scheme (MRF PBL Scheme)significantly enhances ensemble prediction of low-level humidity. Furthermore, combined perturbations outperform single-parameter perturbations. Joint perturbations across the cloud microphysics and planetary boundary layer schemes deliver the best overall performance. Ultimately, both parameter choice and perturbation range strongly affect ensemble forecast accuracy. Synergistic parameter combinations boost forecast skill, while antagonistic ones reduce it. Multiparameter perturbations that consider parameter dependencies and perturbation ranges are more effective for improving ensemble predictions, offering valuable guidance for future parameter selection.