Chaotic Nonlinear Time Series Analysis In Hydrologic River Inflow System

Hydrological forecast is one of the most important parts of Hydrology research. It is the key issue in flood control,water resource usage and protection,hydraulic structures design,reservoir operation,and industry and agriculture production. The hydrologic processes have been described by regular methods of internal determinacy or external randomness or both of them for a long time. In fact,hydrology system is dominated by the objective factors,such as weather,geography and human activities,with combination of determinacy and randomness. The chaotic analysis method combines determinacy and randomness,which seems more adaptive to describe hydrologic time series than conventional hydrologic methods,and becomes more and more attractive recently. In this paper,the chaotic nonlinear time series method was applied to improve hydrologic inflow data analysis. Firstly,chaotic characteristics,i.e.,fractal dimension and maximal Lyapuonv exponent,and state space parameters,including time delay,reconstruct dimension and neighborhood radius,were calculated respectively. Fractal dimension was estimated by G-P saturation correlation dimension method,and maximal Lyapuonv exponent was calculated by two methods,namely,Rosenstein method and Kantz method. Time delay was chosen by using autocorrelation function method and mutual information method,while reconstruct dimension was obtained by G-P saturation correlation dimension method and false nearest neighbor percentage method. Furthermore,initial neighborhood radius was computed by the estimated noise level based on the G-P saturation correlation dimension method. Secondly,a noise reduction of the inflow time series was carried out by chaotic nonlinear local projection noise reduction method,and the effects on noise to chaotic characteristics and state reconstruction parameters were discussed. Finally,a local linear forecast model based on the reconstructed state,which is in fact a global nonlinear model,was developed to predict inflow and verify the selection of the reconstructed parameters and the validity of the noise reduction.