In order to enhance the prediction accuracy and applicability of the Storm Water Management Model (SWMM) in urban flood management, this study conducted a case study at the East Campus of Huzhou University, systematically performing sensitivity analysis and automated calibration of the model parameters. First, the improved Morris screening method was employed to conduct sensitivity analysis on numerous parameters within the SWMM model, identifying five key sensitive parameters as the focus for subsequent calibration efforts. Subsequently, this study integrated the BP neural network with the Harris Hawk Optimization (HHO) algorithm to construct an efficient parameter calibration model. After optimizing the neural network weights and thresholds using the HHO algorithm, the model successfully established a nonlinear mapping relationship between the inputs (water depth and flow sequence) and the outputs (sensitive parameter combinations). The experimental results showed that the trained BP neural network model could accurately and efficiently deduce the optimal SWMM parameter combinations. When the deduced parameter combinations were re-entered into the SWMM model for simulation verification, the simulation results were found to have a high degree of consistency with the measured data. Nash-Sutcliffe efficiency coefficients all maintained above 0.82, significantly superior to traditional manual calibration methods. This research result verifies the feasibility and superiority of using BP neural networks in the calibration of SWMM parameters.