Research Of Flood Routing And Runoff Prediction In Xitiaoxi Basin

With the rapid development of the era and the urbanization advancement speeding up unceasingly,the flood control and disaster mitigation in China is facing new difficulties and challenges,the combination of non-engineering measures and engineering measures to change the management mode of thinking has produced results,further to do a good job of flood risk management of multiple regions is still the cause of the flood control and disaster mitigation.However,the change of river runoff has a direct impact on the whole hydrological system,so it is of great theoretical and practical significance to carry out the research of runoff evolution rule analysis and prediction for the prevention and control of flood disaster.In this paper,the detailed simulation of flood inundation evolution in Xitiaoxi Basin was carried out,the runoff evolution law was analyzed,and the artificial intelligence model for runoff prediction was constructed.The prominent study contents and results are as follows:The basic theory and method of elaborate simulation of flood inundation evolution are studied.Based on high-precision DEM and DLG data,a two-dimensional coupling hydrodynamic model of flood inundation evolution in Xitiao River Basin is established.The control equation for one dimensional form of the model is the Saint Venant equations,the two-dimensional form under the control equations for planar unsteady flow equations,the two-dimensional model was connected with the method of lateral effective coupling,the model reflected the spread evolution process of the flood from insede to outside the river,thus accurately simulating the Xitiao River Basin flood risk occurrence.The fractal theory was implied to make quantitative description of runoff time series correlation,sequence by R/S analysis method to calculate the runoff of Hurst index,further calculate the fractal dimension,the runoff series fractal features,explore the previous observations on the output of the latter sequence observed value,using years completely adjustment coefficient and apportion entropy to analyze the runoff distribution inhomogeneity,get the development trend of runoff distribution of years.According to the fractal characteristics of monthly average runoff,the monthly runoff time series prediction model of random forest-hydrometeorological factors series was constructed,using the theory of random forests of two random strategy,bagged predictor selection method and the feature subspace method,explaining variable importance measure to determine the model predictor of the final output,using the trained model to forecast 60 monthly average runoff of representative hydrological station;Finally,the RF model is compared with other models,such as ARMA model,metabolic fractal interpolation prediction model and FAR model,to analyze the advantages of the RF model.