Prediction of Inflow at Three Gorges Dam in Yangtze River with Wavelet Network Model

A wavelet network model is developed to predict the inflow of Three Gorges dam in Yangtze River, China. The model makes use of the multi-resolution analysis of wavelet analysis and the nonlinear capability of artificial neural network. The short and long term input runoff of Three Gorges dam, such as annual mean discharge, seasonal mean discharge of 10 days period, daily mean discharge and annual maximum flood peak discharge, have been predicted with the wavelet network model (WNM). At the same time a kind of threshold auto-regressive model (TAR) has also applied for those predictions. The comparison of WNM with TAR has been executed. The results show that the accuracy of model predictions with WNM is generally better than that with TAR. The suggested wavelet network model is functional and feasible for runoff prediction.