Study On Annual Runoff Forecasting Model Of Stone River Reservoir

With the development of society and human progress,people pay more and more attention to the safety of water resources.At the same time,they have higher requirements for efficient utilization and rational allocation of water resources.It has practical significance and practical value to develop mid long term runoff forecasting,It can not only play an important role in reservoir operation,drought control,water supply,power generation and irrigation,but also effectively alleviate the contradiction between water use in different departments,and maximize the comprehensive benefits of limited water resources.The paper takes the Shitouhe reservoir of Shaanxi Province as the object of study,According to the 1954~2014 series of measured data of hydrological stations and rainfall stations in the basin,Based on the analysis of the interannual and annual variation characteristics of reservoir inflow runoff and watershed precipitation,The antecedent runoff,antecedent precipitation and precipitation over the same period were selected as predictors,Multivariate regression annual runoff forecasting model,BP Network Annual Runoff Forecasting Model and support vector machine(SVM)annual runoff forecasting model are established respectively,Taking model accuracy as the standard,the annual runoff forecasting model of Shitouhe reservoir is preliminarily determined.The main research results are as follows:(1)The average inflow of Shitouhe Reservoir is 13.8m3/s,the average annual runoff is4.36x108m3,the annual runoff coefficient of variation Cv=0.41,The interannual variation of runoff is large;The runoff distribution is very uneven during the period of time,The nonuniformity coefficient is 0.32,mainly from April to October,the concentration is 0.48,the corresponding focus in July.(2)in 1990,the annual runoff of the 1954-2014 year was mutated.After the mutation,the runoff decreased obviously.However,the precipitation has not been checked out.At the same time,the double cumulative curve of precipitation and runoff in the basin shows a straight line.Based on this,the annual runoff forecasting model of the reservoir is derived from the measured rainfall and runoff data after 1990.(3)according to the two variation periods of annual runoff,6a and 12 a after 1990,theannual runoff of 1990~2007 is chosen as the calibration period of the model,and 2008~2014is used as the verification period of the model.The main predictors of the current runoff are the annual precipitation,the previous year runoff and the previous year's precipitation.(4)The calibration period qualified rate of ARIMA(3,1,4)prediction model is 41.18%,the validation period qualified rate is 57.14%,the accuracy of this model is lower than that of C,cannot be used to forecast.(5)The calibration period qualified rate of rainfall runoff linear forecasting model is66.67%,the accuracy of grade C;the validation period rate is only 42.86%,lower than the precision grade C,this model cannot be used to forecast.The calibration period qualified rate of the multiple regression model is 66.67%,the validation period qualified rate is 71.42%,precision grade respectively C,B,can be used for reference to forecast.(6)The calibration period qualified rate of BP(2)annual runoff forecast model is94.44%,the accuracy level is A,the validation period rate is 57.14%,the accuracy level below C,this model cannot be used to forecast.The calibration period qualified rate of BP(3)annual runoff prediction model is 94.44%,the validation period rate is 71.42%,the accuracy levels are Grade A and Grade B,which can be used for operational forecasting.(7)The calibration period qualified rate of SVM(2)annual runoff forecast model is61.11%,the accuracy level is C,the validation period rate is only 57.14%,lower than the precision grade C,the model can be used to forecast;The calibration period qualified rate of SVM(3)annual runoff forecast model is 83.33%,the accuracy level is B,the validation period rate is 71.42%,the accuracy level is B,this model can be used to forecast.(8)comprehensive analysis of each model,and recommend BP(3)annual runoff forecasting model as the forecasting model of Shitouhe reservoir.