To address the issues of low efficiency and insufficient accuracy in manual calibration of runoff and peak time for multi-parameter urban hydrological models, this study proposed an automatic parameter calibration method based on the improved particle swarm optimization (PSO) algorithm. The method introduces Logistic mapping for particle initialization and Lévy flight for position updating within the PSO framework to avoid local optima. Additionally, considering the characteristics of urban runoff generation and concentration processes, a weighted multi-objective fitness function incorporating overall fitting, peak flow, and peak time was constructed to enhance the model’s ability to capture key hydrological features. The proposed method was implemented in Python and coupled with a mechanistic model (storm water management model, SWMM). Using field monitoring data from a test site, ten key hydrological parameters were calibrated, and the performance of fitness functions with different weight assignments was compared. The results demonstrate that the weighted multi-objective fitness function is more advantageous for urban drainage system emergency management, particularly in improving the simulation accuracy of peak flow and peak time. When applied to a real drainage system in Jiujiang City, the method achieved peak flow and peak time errors of 0.56% and -6.82%, respectively, confirming its feasibility and accuracy.