Hybrid wavelet and chaos theory for runoff forecasting

This paper introduced a method of decomposing non-stationary runoff time series. By wavelet decomposing, the runoff time series is decomposed into stationary time series and stochastic time series, and AR(n) model be imposed for forecasting stationary time series. By studying chaos characteristic of stochastic time series, this paper put forward a nonlinear chaos dynamics-forecasting model to dispose runoff time series with high-embedded dimension. It can effectively decrease the Lyapunov exponential sum in added dimensions of reconstruction set when the dimensions of reconstructed space are increased. Finally, the forecasting result is reconstructed based on wavelet theory. The forecasting result of original runoff time series is achieved. The method is high precision and feasible through example test.