Abstract: Global climate change and urbanization have led to an increase in the frequency and intensity of extreme rainfall events， resulting in significant alterations in urban hydrological and climatic conditions. In this context， refined simulation of urban drainage systems is essential for enhancing management and emergency response capabilities. The traditional calibration methods rely heavily on subjective experience and are inefficient. This study addressed the challenge of low efficiency in automatic calibration techniques by proposing a novel approach based on an elitist preservation genetic algorithm and a parallel computing framework for the automatic calibration of the Storm Water Management Model （SWMM）. It utilized the SWMM-API library for interaction with SWMM， enabling the reading and execution of SWMM input files （INP） and the parsing of output results （OUT）. The elitist preservation genetic algorithm from the open-source genetic and evolutionary algorithm toolbox Geatpy was employed to enhance the global exploration capability and convergence speed of the algorithm. To overcome the limitations of Python's Global Interpreter Lock （GIL） for CPU multi-core parallel computing， the study utilized Python's built-in multiprocessing module to create a multiprocessing pool， thereby accelerated the computation of the optimization algorithm. The proposed method was tested on a residential community drainage network model， demonstrating that the automatic calibration program significantly reduced the time required for SWMM model calibration. The parallel framework achieved a 40.3% increase in auto calibration speed compared to the non-parallel framework while maintaining an equivalent model accuracy. The study showed that the parallel framework can search more extensively in the parameter space without compromising parameter optimization quality. The study successfully developed a reliable parameter calibration framework by coupling the elitist preservation genetic algorithm with the SWMM model， reduced the computational time for model parameter automatic calibration， demonstrating its efficiency in addressing large-scale and complex hydrological simulation problems. Future work should focus on multi-scenario simulations for different rainfall intensities， frequencies， and patterns， developing dynamic parameter calibration algorithms to meet the dynamic needs of actual urban rainwater management systems. Additionally， the optimal number of parallel processes under different hardware configurations needs to be determined experimentally， and the potential of GPU acceleration for further enhancing the optimization effects need to be explored.