Renmin Zhujiang (Jan 2022)
Research on Monthly Runoff Forecast in Beijiang River Basin Based on Multi-model Ensemble Method
Abstract
The accuracy of the monthly runoff forecast plays a fairly important role in aspects such as optimal allocation of water resources,flood control and drought relief in a basin,water dispatching,and power generation optimization of reservoir groups.The commonly used methods for the monthly runoff forecast mainly include water balance models,mathematical statistics models,and artificial neural networks.Studies have shown that any single model cannot achieve the optimal monthly runoff forecast.Therefore,the multi-model ensemble method provides an effective way to eliminate model uncertainty and improve the accuracy of the monthly runoff forecast.Specifically,the research takes Pingshi,Lishi,Hengshi,and Shijiao stations in the Beijiang River Basin as the research object to analyze and compare the effects of the seasonal auto-regressive (SAR) model,two-parameter monthly water balance (TPMWB) model,and artificial neural network (ANN) model.Then,the multi-model ensemble method for the above-mentioned stations is proposed on the basis of the Bayesian model averaging (BMA) method.The research results reveal that compared with any of the three models,the multi-model ensemble method has significantly improved the accuracy of the monthly runoff forecast with a higher determination coefficient (DC) and a lower mean absolute percentage error (MAPE),and thus it can provide better support for decisions in dispatching in the basin.
