Medium And Long-term Runoff Forecast In The Ganjiang River Basin

Medium and long-term runoff forecasting is one of the important links in water dispatching,and it is a key technical support for the implementation of the most stringent water use control objectives for water management systems and the development of water distribution plans for river basins.The Ganjiang River Basin is the largest river basin in Jiangxi Province.In recent years,there have been problems such as shortage of water resources and intensification of supply and demand conflicts in the basin.In order to implement the water dispatching in the Ganjiang River Basin and provide more accurate water regime forecast for the annual water volume scheduling of the Ganjiang River Basin,this paper carries out a study on the medium and long-term runoff forecasting model of the Ganjiang River Basin.First,based on the analysis of the runoff characteristics of the Ganjiang River Basin,the correlation coefficient method,stepwise regression method,and principal component analysis method were used to select the future of Ganjiang River Basin from the two conventional factors of rainfall and runoff and 130 atmospheric circulation indices.The most influential factors of influent water were used as predictors of runoff forecasting in the Ganjiang River Basin,and three different predictive factor optimization schemes were compiled.Secondly,based on the selected predictive factors,multiple regression models,BP artificial neural network models,and Elman were combined respectively.The neural network model and PSO-SVR model are used to build a medium and long-term runoff prediction model based on the physical cause analysis in the Ganjiang River Basin.The forecasting effects of different models with different forecasting factors are compared and analyzed.Finally,taking into account the limitations of the single forecasting model,the Coupling the advantages of multimodel medium and long-term runoff forecasting coupled model,and based on the coupled multi-models,the residual error correction based forecasting model result correction method was proposed,which effectively improved the accuracy of forecasting model.The research mainly has the following results:(1)Through comparative analysis of multivariate regression model,BP artificial neural network model,Elman neural network model,and PSO-SVR model in three different factor optimization schemes,the results show that the forecasting factors based on principal component analysis can be better the description of the trend of future runoff can be used as a factor selection method for medium and long-term runoff forecasting in the Ganjiang River Basin.(2)By comparing and analyzing the forecast results of different models under the same factor preference scheme,the results show that the PSO-SVR model is better than the three factor optimization schemes relative to multiple regression models,BP artificial neural network models,and Elman neural network models.The forecast effect is better.(3)The multi-model coupled forecasting technology based on residual correction is higher than 80% in the prediction rate of all three factors.It can meet the requirements of the "Code for Forecasting Hydrological Information" and can be used for medium and long-term runoff forecast in the Ganjiang River Basin.The screening method of medium and long-term runoff forecasting factors,multimodel coupling and residual error correction method proposed in this paper constitute a medium and long-term runoff forecasting technical system that supports medium and long-term runoff forecasting in the Ganjiang River Basin.The research results are used to carry out the water dispatch in the Ganjiang River Basin.The preparation of the plan provides strong technical support.It is of great and far-reaching practical significance to solve the problems such as the increasing water demand in the Ganjiang River Basin and the intensified contradiction between water supply and demand.