<p>Parameters are crucial to hydrological models, and different categories of parameters significantly impact model performance. There is currently a lack of detailed research systematically evaluating the impact of parameter types, physical properties, and optimization methods on the flood forecasting performance of hydrological models. In this study, the Liuxihe Model, a new generation distributed physical hydrological model for watershed flood prediction, was employed to simulate floods. An improved chaotic particle swarm optimization algorithm was employed to automatically optimize the model parameters, thereby evaluating the influence of each parameter category on the model's flood simulation performance. A comparison of the flood simulation effects of the model's initial parameters and the calibrated parameters revealed that those effects with calibrated parameters were significantly better than those with initial parameters were. The improved chaotic particle swarm optimization algorithm in this study improved the model's calibrated efficiency by up to 2 times, the Nash certainty coefficient increased by 26%, and the flood peak error decreased by 70%. The classification of model parameters and the improvement of the chaotic particle swarm optimization algorithm provide key parameter analyses for model calibration and offer parameter optimization methods for hydrological models, which are crucial for flood forecasting, flood control and disaster reduction.</p>