Flux and level prediction based on an wavelet neural network flood model

This paper uses wavelet neural networks for flood prediction. It presents an flood prediction model and give an rapid algorithm. The water flux and level are used as input and output variables in the prediction model. The analysis of time-frequency characteristic of wavelet transformation is given. The prediction precision is improved by combining low frequency feature vector with high frequency ones. High frequencies of signals, which are middle or low numbers, are decomposed into small scales in wavelet space in flood flux and level analyses, and low frequencies of signals, which are large numbers, are decomposed into big scales. The model developed in this paper provided a new procedure for flood prediction. The experiment shows that the application of wavelet neural networks in flood prediction can give more accurate results.