Analysis Of Long-term Regional Flood Forecasting In Daxinganling Area Based On Big Data

In this paper,the flood peak series of Heilongjiang,Ganhe and Huma rivers in Greater Khingan Range are selected as the research object,and the natural factors such as flood characteristics and meteorology are combed.Based on the constructed flood cause database,the traditional and modern long-term prediction methods are used to analyze the evolution law of flood in this region,explore the meta driving mode,and establish the long-term flood prediction model in this region.The main contents are as follows:(1)Due to the special geographical location of Greater Khingan Range,the influence of polar vortex and polar movement is obvious.The formation of flood in summer is coupled with the formation of runoff caused by snow melting in winter.The annual maximum peak discharge(the annual maximum peak water level is selected for the water level station)of the designated hydrological stations in the north and south of the Greater Khingan Range is selected as the analysis series,and the driving law of flood in the Greater Khingan Range is obtained by SPSS software and traditional methods.(2)The period and trend of the annual maximum flood peak series of each station are analyzed by using the corresponding mutation detection methods such as Kendall and EMD.Combined with the background analysis method,the meta driving mode and relevant background mode of flood in this area are explored,and the prediction model is established for mutation detection.(3)Using MATLAB software,using the modern long-term flood prediction method and support vector machine(SVM)and other methods to carry out numericalquantitative prediction of the annual maximum flood peak of each station,establish the long-term flood prediction model of each station.The results show that the SVM model has high simulation accuracy.The results of the traditional and modern long-term flood forecasting model are integrated forecast,which is equal weight set of the model forecast results.The prediction results of each model are evaluated.The prediction results of each model are evaluated,and the peak discharge of each station from 2013 to2014 is selected to verify the model.