Nonlinear Forecast Modeling Based on Wavelet Analysis

System modeling is a complex nonlinear procedure. Traditional methods are easy to get slow convergence and low efficiency. The nonlinear forecast modeling based on wavelet analysis is proposed and used to forecast the change trends of runoff. It inherits the multiresolution capability of wavelets analysis, and with the generalization of forecast modeling, the presented method is applied to annual runoff forecast of Yichang. The simulation performance is compared with that of autoregressive model. It is shown that wavelet analysis is able to find the latent detail information of time series and supply satisfying forecast results under different time scales.