Flood forecasting research based on the chaotic RBF neural network model
Jianlong Zhang, Xie Jiancang, Han Yuping, Yu Shen · Journal of Northwest A&F University · 2009
【Objective】 The study was to establish a better flow of the flood forecasting model.【Method】 At present,most of the forecasting models of the great flood peak flow numerical prediction are not ideal.In the chaos of the flood systems on the basis of identification,forecasting model was estaldished based on chaos theory and RBF neural network to measure flood sequence of space reconstruction by training samples,and the network structure was determined by using MATLAB 7.0 toolbox.【Result】 The RBF forecast model was used by Fenhe Shitan Hydrometric Station in 2004 to measure the largest flood forecasts,and the results showed the pass rate,with an average relative error,correlation coefficient(R),root mean square error(RMSE) and Nash-Sutcliffe coefficient(NSC) were 100%,4.69%,0.979 3,4.226 0 and 0.955 2,and those of the traditional Volterra adaptive prediction model were 93.75%,8.97%,0.954 0,10.263 2 and 0.735 8.RBF model can have better predication results and has made better numerical prediction of large flow flood peak.【Conclusion】 To build predictive models based on Chaos Theory and the RBF neural network can be a new attempt in improving flood forecasting accuracy,which has some reference value on flood forecasting in the future.