Research on stage-discharge relationship model based on random forest algorithm

Hydrological simulation and prediction is a vital aspect of the hydrological change research. Accurate prediction of hydrological factors such as stage and discharge is essential for water resources planning, reservoir dispatching and operation, shipping management and flood control. River discharge forecasting during flood season is an important issue in water resources planning and management. To improve the calibration accuracy and stability of the stage-discharge relationship model, the feasibility of integrated algorithm in the study of stage-discharge relationship is explored. A random forest algorithm based on neural network is proposed by using the framework of integrated algorithm. First, Levenberg-Marquardt (LM) algorithm is used to optimize the weight updating process of Back propagation (BP) neural network and improve the convergence speed of the model. Second, the LM-BP algorithm is used as a decision tree to build a random forest algorithm. The model is tested with the hydrological data of Hongqi Station in Dadu River in flood season. Based on the mean absolute error, mean square error and mean absolute percentage error of the performance indicators, the results for the classical model, BP neural network model, LM-BP neural network model and optimized algorithm model are evaluated. The evaluation results show that the optimized algorithm model (Mae = 3.13 m3/s MSE = 19.28 m3/s MAPE = 1.8%) is superior to other algorithm models, and the integrated algorithm model has high accuracy and good stability in flood season flow forecasting.