Dong Jiang Reservoir Storage Of Runoff Forecast

Medium and long-term hydrological forecasting accuracy is often not high.The research also has many difficulties.Hydrological changes exist uncertainty.This uncertainty for medium and long-term hydrological forecast of the forecast accuracy influences greatly.The hydrological system is a highly nonlinear system,and it is difficult to establish a mathematical model for accurate representation of runoff.Because the hydrologic system is very complex,so far,there is no general mathematical model to solve the problem of long term forecasting of hydrological series.In this paper,Dongjiang reservoir monthly runoff time series data is used as the research object to establish the long-term reservoir inflow runoff forecasting model.(1)In order to improve the stability and operation precision of the BP neural network model,the parameters of BP neural network with genetic algorithm is chose to optimize value,and the optimized parameter values is regarded as the weights and thresholds of BP neural network to the forecast East River reservoir inflow runoff.Due to the small sample size,the BP neural network is difficult to effectively improve the accuracy and the fitting degree of the runoff forecast.(2)Support vector machine(SVM)can better adapt to the capacity of small training samples,and the prediction problem is changed into a quadratic programming problem that can avoid local optimal and repeated test error defect of neural network.For Dongjiang reservoir inflow runoff forecasting,SVM has got better forecasting value.(3)Projection pursuit method is able to handle high dimensional problems very well.In this paper,the artificial fish swarm algorithm and projection pursuit algorithm combined together to establish hybrid intelligent runoff forecast model which can transform fitting function according to the order of the sequence and better deal with the runoff prediction problem.BP artificial neural network,SVM model and projection pursuit of artificial fish swarm algorithm were used for Dongjiang reservoir runoff forecast.And the BP artificial neural network on runoff prediction result is not too ideal for that BP artificial neural network model of the structure is difficult to determine,poor network stability and reservoir inflow runoff time series is short,wthat is more,many predictors is short.The error of the results of SVM model are all small than 20%.The SVM model can predict the small samples,nonlinear time series and so on.The projection pursuit model automatic selects prediction factor,high order function of runoff to predict which can effectively improve the prediction accuracy of the model...