Short-term load forecasting using wavelet networks

Abstract

This paper describes an application of a type of neural network, called the wavelet network, to short-term load forecasting. Based on the wavelet transform theory, the recently developed wavelet network is proposed as an alternative to the feed-fonvard neural networks for approximating arbitrary nonlinear functions and for solving classification problems. Performance results are given for the daily peak and total load forecasting using data from a real power system. A comparison among results from the wavelet network, the radial-basis feedforward (RBF) neural network and back-propagation (BP) neural network shows that the wavelet network is encouraging.