Study on Optimization and Combination Strategy of Multiple Daily Runoff Prediction Models Coupled with Physical Mechanism and LSTM

Accurate prediction of runoff is an important foundation for optimizing water resource allocation and reservoir scheduling operations. However, due to its complex characteristics such as time-varying and non-stationary, accurate prediction of the runoff process is very difficult. In view of the ability of physical mechanism runoff forecasting models to simulate and describe the physical process of runoff and the powerful mapping ability of big data methods for highly nonlinear runoff time series, this paper proposed a runoff forecasting modeling method that couples 16 physical mechanism models with Long Short Term Memory network (LSTM). At the same time, to accommodate the comprehensive analysis and evaluation of multi-model forecasting performance, this paper also proposed a comprehensive evaluation metric for runoff forecasting considering the characteristics of group models, and constructed a set of runoff forecasting model selection and combination strategies adapted to the runoff characteristics. The results of the case study show that the method can effectively improve the forecast accuracy of multiple models. When evaluating the forecast accuracy of runoff during the calibration period, the average reduction of Root Mean Square Error (RMSE) is 39.62%, and the average increase of Nash-Sutcliffe coefficient R² is 7.49%. During the validation period, the average reduction of RMSE is 62.68% and the average increase of R² is 24.24%. Meanwhile, the method proposed in this article provides a standardized modeling process for its application in other watersheds. It can effectively improve the overall forecasting performance of a single forecasting model and has good practical value for promotion and application.