Research On Medium And Long-term Runoff Forecast Of Three Gorges Reservoir Based On Ensemble Empirical Mode Decomposition And Artificial Neuron Network

With the extreme climate events and human action,hydrological runoff series is becoming more and more complex.Under the operation condition of reservoir,improving the runoff forecast accuracy has become the current focus.It can help the reservoir make full use of water resources according to the hydrologic information.Therefore,in this research,an artificial neural network(ANN)model coupled with the ensemble empirical mode decomposition(EEMD)is presented for forecasting medium and long-term runoff time series,in order to improve the prediction accuracy and provide decision support.This research mainly includes the following aspects:(1)The changes of runoff are quantified according to statistical analysis of the runoff data.Adopted moving average method,Mann-Kendall method and Spearman method to detect the trend of annual runoff series.The results indicate that the annual runoff showed obvious decreasing trend.Adopted three kinds of mutation detection methods,Pettitt mutation point method,cumulative deviation method and Worsley likelihood ratio method to detect the change of the annual runoff series.The results show that three methods have shown runoff mutation point in 1968.In addition,runoff mainly concentrated in the flood season.The flood season runoff takes on a declining trend,and the dry season runoff takes on a increasing trend obviously.(2)EEMD-ANN model,artificial neural network(ANN)and autoregressive model(AR)are presented for forecasting annual runoff time series.The stable runoff series and the original runoff series are decomposed into a finite and often small number of intrinsic mode functions(IMFs)and a residual series using EEMD technique for attaining deeper insight into the data characteristics.Then all IMF components and residue are predicted through appropriate ANN models,respectively.The forecast results of the modeled IMFs and residual series are summed to formulate an ensemble forecast for the stable runoff series and original annual runoff series.The results show that EEMD-ANN model can effectively enhance forecasting accuracy and the proposed EEMD-ANN model can attain significant improvement over ANN approach.The stationary runoff series obtained by mutation detection method should be used to build annual prediction model.(3)EEMD-ANN model and artificial neural network(ANN)are presented for forecasting monthly runoff time series.The results show that EEMD-ANN model has better prediction accuracy,especially at the flood season.The annual runoff and monthly runoff have the same plentiful or short level.It proves that research on medium and long-term inflow forecasting based on EEMD-ANN model is meaningful.