Hydrological Prediction Of Chaohe Basin Based On SVM

The Miyun Reservoir shoulders the task of providing domestic and industrial water for Beijing,and the Chaohe Basin is one of the important water sources of Miyun Reservoir.Therefore,the prediction of runoff and rainfall in Chaohe Basin is of great significance.The main contents are as follows:The GWO-SVM based on chaotic time series was proposed for the runoff prediction.The method of phase space reconstruction and chaotic characteristic analysis was used in 2004-2009 Dage hydrological station daily runoff data.The data delay time ?d=7,the embedding dimension m=2 and the maximum Lyapunov value ?=0.0745.So,the data satisfied the chaotic characteristics and used the reconstructed data to establish the GWO-SVM model(R2=0.996,MSE=0.014,MAPE=4.493%),and models of the SVM(R2=0.899,MSE=0.061,MAPE=13.578%),PSO-SVM(R2=0.907,MSE=0.069,MAPE=11.526%)and GA-SVM(R2=0.978,MSE=0.053,MAPE=10.053%)were established for comparing.The results show that the GWO-SVM has the best prediction accuracy.Combined with the chaotic time series analysis method can achieve better prediction of daily runoff.The GWO-SVM combined the interpolation method was proposed for the rainfall prediction.The 2009-2017 daily rainfall data of the Hushiha rainfall station,Anchungoumen rainfall station,and the Xiahui rainfall station were taken as input variables,and the Gubeikou rainfall station daily rainfall data was as the output variable.The data was used to established models of the SVM(R2=0.912,MSE=0.324,MAPE=10.010%),the PSO-SVM(R2=0.962,MSE=0.158,MAPE=8.430%),the GA-SVM(R2=0.975,MSE=0.065,MAPE=5.940%)and the GWO-SVM(R2=0.985,MSE=0.012,MAPE=6.450%).Then,the interpolation method combined with the GWO-SVM was proposed to predict the Gubeikou rainfall,the MAPE of the spline-GWO-SVM was 6.40%and the MAPE of the IDW-GWO-SVM was 6.37%.The results show that the GWO-SVM model combined the interpolation method can predict rainfall effectively.