Abstract: On the basis of improving the Xin′anjiang Hydrological Forecasting Model, an error correction model for flood forecasting based on the Long Short-Term Memory (LSTM) network was constructed by using the error between the forecasted value and the measured value. Taking 50 representative floods in the Lanhe River Basin of Shanxi Province as the research object, and the forecast results of the improved Xin′anjiang Model were corrected in real time. The research results indicate that the forecast error correction model based on the LSTM algorithm can effectively improve the accuracy of flood forecasting in the Lanhe River Basin, providing valuable technical support for local flood prevention, disaster mitigation, and water resource management.