Taking the Changzhao Reservoir Basin of the Cao’e River as the study area, rainfall and runoff were selected as the primary influencing factors to construct the longshort term memory network (LSTM) model. The hydrometeorological characteristics and runoff generation mechanism in the basin were analyzed using the LSTM model and compared with the simulation results of the Xin’anjiang model. The results indicate that the LSTM model and the Xin’anjiang model perform well in flood simulation in the Changzhao Reservoir Basin. The LSTM model has a higher qualification rate, and the relative errors of the LSTM model in simulation of average runoff depth and peak flow are smaller, demonstrating a higher accuracy of the LSTM model. Meanwhile, the Xin’anjiang model has a relatively stable coefficient of determination and smaller peak occurrence time difference. The LSTM model reduces the dependence of the model on human experience and can be used for realtime flood forecasting with high precision requirements. The Xin’anjiang model can explain the source of errors based on the physical process expressed by parameters for some emergencies, and is more suitable for analysis of complex scenarios such as extreme floods and explanation of physical processes.