Application of wavelets in power system load forecasting

Forecasting of electric load demand on power system using wavelet transform is presented in this work. It utilizes the periodicities of past load demand data. Load demand data is presented as an image of size of 7times24 for a week, from which the image of a year is obtained by stacking 52 weeks. Medium range forecasting has been performed using wavelet with autoregressive modeling and smoothing techniques. Various types of wavelets bases are applied to extract the data features to be used as priori knowledge for prediction instead of the actual utility data as may be in the case of majority of forecast models and used to forecast the demand. Inversion of the forecast coefficients leads to the actual forecast. As it is simple and efficient, can effectively be utilized by the power sector utilities for forecasting of electrical load demand occurring on them