Chaotic Time Series In Hydrological Forecasting

At present, the comparable conventional methods in hydrological forecast are mostly based on linear system , Their common shortcomings lies in emphasizing study of mathematical problems too much or simply reflecting changeable situation utmostly, overlooking dynamic evolutional process of hydrological phenomenon accompanied with time.Based on discussing the conventional methods ,In this paper, the chaotic nonlinear theory was applied into hydrology forecast, having forecasted heihe daily river inflow series, the author discovered evolutional process of hydrological phenomenon accompanied with time of heihe daily river inflow time series and made a comparable successful prediction according to phase space reconstruction and using "Matlab" software. The main achievements are as follows:(1) By G-P saturation correlation dimension method , Heihe daily river inflow data series was considered a chaotic series ,which was produced in nonlinear definite system.(2) Time delay ten of Heihe daily river inflow series was chosen by using autocorrelation function method.(3) Reconstruct dimension nine of Heihe daily river inflow series was obtained by graf of saturation correlation dimension method .(4) Based on phase space reconstruction, the conclusion was got .it can be seen that weighting and fitting method of approximation superior to method of least squares in both theory and practical calculation during calculated method by forecasting the process of Heihe daily river inflow series.(5) By local linear forecast model of chaotic series method based on reconstructed state,the rational rate was 61% and the predictive trend was comparably conherent, especially with relatively high wave. The forecast of chaotic time series is not only to make a prediction but also to verify and state space parameters, in order to study and improve the calculated method.