Study On Flood Forecasting Model And Correction Method Suitable For Small And Medium-sized Rivers In Hilly Area

Accurate and timely flood forecasting is an important non-engineering measure for alleviating flood disaster.Is located in the complex terrain hill region of middle and small rivers,the floods,sudden strong,the runoff time is short,harm,at the same time affected by climate change and human activities in recent years,the impact of the change of underlying surface,the medium and small rivers flood trends to increase the number of the frequency,and brought new challenges for flood forecasting,in addition,in the process of flood forecast error is difficult to avoid and how to reduce the error has been an important work of small and medium-sized rivers flood forecasting.In this paper,taking Tunxi Basin as an example,combined with the actual flood prevention and control of small and medium-sized rivers for the information of flood peak location,a parameter calibration method of Xin'anjiang River model is proposed,which takes the mean of the qualified rate of flood peak error and the qualified rate of peak time error as the objective function,and the calculation results of Xin 'anjiang River model and BP neural network model are coupled.Finally,the error autoregressive method,feedback method and K-nearest neighbor correction method based on flood feature database are used to correct the errors.The main research contents and conclusions are as follows:(1)With runoff depth percent of pass,the qualified rate of peak,peak time now weight percent of pass,deterministic coefficient respectively(1:2:2:1)the choice of the objective function for the parameter model of xin an river in rate on a regular basis and Jian Yan Qi average deterministic coefficients above 0.8,and compared with the traditional weight than for(1:1:1:1)the rate of method,the method of flood peak and peak time now qualified rate 100%,97.56% respectively,were higher than traditional rate method,on the flood peak error and time error now has certain advantages,and is more conform to the requirements of the small and medium-sized rivers flood forecasting.(2)Useing arithmetic average method,combination xin an river model and BP neural network model calculation results,the inspection period,the average runoff depth error after combination,average peak error,the average peak time now is respectively 0.59%,4.29%,1.63 hours,not only make up for the disadvantage in the runoff depth forecast,xin an river model and simulating flood peak,peak time now also got a certain level of ascension;(3)The flood feature database was established according to different types of historical floods,and the k-nearest neighbor correction method based on the flood feature database was proposed.The correction results of error autoregressive method and feedback method were compared,and the respective advantages of the three correction methods and their applicable basin range were determined.The results show that the K-nearest neighbor correction method based on flood feature database has the advantage of correcting the flood peak error and the peak occurrence time,which is more in line with the flood prediction requirements of small and medium-sized rivers.