The Research Of The Runoff Forecasting Model Based On Intelligent Algorithm To Optimize The BP Neural Network

Hybrid predicted model based on intelligent algorithm optimizing neural network has become a hot problem of research in recent years and has been widely used in energy,engineering,hydrology and other different fields.Runoff as one of the important factors in the analysis of hydrological sequences,the accurate and reasonable prediction for the national economic development and water conservancy construction has the vital significance.This paper,we use MEA,CS and BA algorithm as well as improved MBA,MeanBA,MCBA and MeanMCBA algorithm to optimize the BP neural network,and then apply them to the runoff series prediction research.First,we use mind evolutionary algorithm(MEA),cuckoo search algorithm(CS)and bat algorithm(BA)to optimize the BP neural network weights and threshold parameters,then according to the daily runoff time series with strong volatility,we establish three hybrid forecasting models based on wavelet de-noise(WDN).Empirical analysis shows that WDN-BABPNN and WDN-CSBPNN models are suitable for daily runoff time series prediction.Secondly,for average monthly runoff series with strong seasonal,we adjust to it first and then establish MEABP-BPNN,CSBP-BPNN and BABP-BPNN three hybrid models,respectively.Empirical result shows that the X12-BABP-BPNN model has highest prediction accuracy in the month runoff series.Finally,based on the BA,MBA,MeanBA and MCBA algorithm,we propose a new MeanMCBA algorithm and establish a new hybrid forecasting model with separating linear and nonlinear,then use it to study the average annual runoff.Through analysis and comparison we obtain that the prediction accuracy of the proposed model is higher than other models.In conclusion,the intelligent algorithm optimized BPNN model is superior to the single BPNN model,and optimization ability of the improved MBA,MeanBA,MCBA,MeanMCBA algorithm is better than the no improved BA algorithm.