Flood Forecasting Simulation Based On Error Correction And Uncertainty Analysis

Floods have caused numerous natural disasters throughout history.Researchers are learning more about flooding.However,the simulation,prediction,correction and uncertainty analysis of flood are still the focus of hydrologists.This paper takes the Dalai-Fuyu-Xiadaiji River as the research object.Firstly,the runoff of Dalai and Fuyu hydrological stations is predicted by the least squares support vector machine model.Secondly,this paper preprocessed the initial prediction error and corrected the prediction error by using the 4D Copula joint distribution function.The corrected error was fed back to the runoff value and used as the boundary input of the MIKE HYDRO River HD model.Finally,the MIKE HYDRO River HD model was used to simulate the flood in the study section,and the uncertainty analysis of the related evaluation indexes of the simulation results was carried out.The main research contents and results of this paper are as follows:(1)In this paper,the LS-SVM model is used to make rolling forecast of runoff of Dalai and Fuyu hydrological stations,and the related evaluation indexes of runoff forecast value of two hydrological stations are calculated.The NSE of Fuyu Hydrological Station with the forecast period of 1,2,3 and 4 days was 0.88,0.80,0.73 and 0.71,respectively.RMSE values were 93.23,120.2,140.19 and 146.37,respectively.The NSE values of runoff prediction in Dalai Hydrological Station were 1.00,0.99,0.98 and 0.97,respectively.RMSE values were 39.96,60.95,83.31 and 109.91,respectively.(2)In this paper,the initial prediction error is corrected by the joint distribution function of preprocessing and 4D Copula.Compared with the initial predicted value,the NSE of Fuyu Hydrological Station after pretreatment and joint correction was increased by 8%,5%,5%and 16%,and the RMSE decreased by 39.37%,12.25%,10.38%and 32.6%,respectively.After the error of Dalai Hydrological Station is corrected by Copula,the NSE value and RMSE value change little,but the NSE value is above 0.97.The martingale model is used to make statistical analysis of the combined corrected error series,and the hypothesis test of unbiased and time-series independence shows that the combined corrected forecast value improves the bias of the initial forecast value,and the revised forecast value is more close to the measured runoff.(3)In this paper,the MIKE HYDRO River HD model is used to simulate the flood evolution in Dalai-Fuyu-Xiadaiji River section.After NSE value calculation and comparison of flood simulation results from 2014 to 2018,In this paper,the bed roughness of the research reach is set as 0.016,the initial water level at Dalai Hydrological Station is 125.15 m with initial flow of 202m~3/s,the initial water level at Fuyu Hydrological Station is 128.9 m with initial flow of 115 m~3/s,and the initial water level at Xiidaiji Hydrological Station is 120.86 m with initial flow of 368 m~3/s.The NSE values of simulated floods from 2014 to 2018 were 0.84,0.95,0.8 and 0.88,respectively.(4)Since the boundary input is uncertain,so is the flow of the simulated flood.In order to analyze the scope of uncertainty,this paper randomly simulated 100 floods,and evaluated and calculated the accuracy and reliability of the 100 simulated floods from different angles.The final results are as follows:In terms of accuracy evaluation,the maximum and minimum flood peak relative errors of 100 simulated floods during flood season are 7.9%and 7.4%,which are in line with Rep≤20%and meet the standard requirements.The NSE values varied from 0.782 to 0.907,and the NSE values of the other floods were all greater than 0.85 except for the one simulated flood which was 0.782.The overall effect of the simulated flood was good.The reference coefficient Be is between-0.074 and 0.087,and its absolute value approaches 0.In terms of reliability evaluation,this paper calculates and compares the interval dispersion of simulated flood in flood season when the confidence is 90%,95%and 99%.The results show that under the three confidence conditions,the mean dispersion DI of the simulated interval of 100 floods is 0.06,0.072 and 0.094,respectively.The interval dispersion of simulated flood is small.The numerical stability of the model simulation is very high.The uncertainty of simulation results is low and the results are practical.In the evaluation of interval coverage rate,the interval coverage rate of 100 simulated floods under the three confidence conditions is 5.2%,6.5%and 7.8%,respectively.The interval coverage rate is small,because interval dispersion and interval coverage rate are usually contradictory to each other.Therefore,the unit interval discretization coefficient is used to evaluate the rationality of the model simulation.By calculation,the PUCI values of 100 simulated floods under three confidence degrees are 2.53,1.597 and 0.936,respectively.The absolute value of PUCI is close to 0,and the simulated flood has strong rationality.