Research And Prediction Of Runoff Data Based On Combination Model Method

Water is an indispensable and precious natural resource for mankind to sustain life and develop economy.The development and utilization of water resources provides the necessary basic material guarantees for the progress of human society and the development of the national economy,and the amount of runoff is closely related to the development and utilization of water resources.The forecast results of runoff have very important reference values for formulating strategies such as flood control,drought relief,and reservoir optimal dispatch in the area.However,due to various intricate and complex factors,the research on runoff prediction has become a hot and difficult topic for scholars at home and abroad.In this paper,the WD-PSO-RBFN-Elman combination model is proposed based on wavelet decomposition(WD),particle swarm optimization(PSO),radial basis function network(RBFN),and Elman neural network.The combined model was applied to the prediction of the monthly average runoff data at the two stations of the Shiyang River(Zamoji Temple and Huangyang Reservoir).In view of the nonlinearity and instability of monthly average runoff data,this paper first preprocesses the original sequence and uses WD method to decompose the original sequence into low frequency sequence(A3)and high frequency sequence(D1,D2,D3).Among them,the A3 sequence reflects the main fluctuations of the monthly average runoff of the Zamu Temple and Huangyang reservoir,and is seasonally adjusted(SAM)based on the seasonal characteristics of the A3 sequence,and then is approximated by the PSO optimized RBF network;The frequency sequences D1,D2,D3 represent different details of the monthly average runoff data for the Zamu Temple and Huangyang reservoir,respectively,and Elman fits for the D1,D2,D3 sequences,respectively.Then the PSO-RBFN approximation sequence is linearly weighted with the Elman fitting high-frequency output,and the final prediction result of WD-PSO-RBFN-Elman is obtained.In order to verify the effectiveness of the new combination model,this paper compares the model with other three groups of models,such as RBFN network,WD-SAM-RBFN model,WD-PSO-RBFN hybrid model,etc.Simulation results show that the proposed model The WD-PSO-RBFN-Elman model has a better effect on the prediction of runoff data.