A new approach to statistical forecasting of daily peak power demand

Abstract

Load forecasting is essential for various operational and management purposes. Short-term load forecasting through stochastic time series analysis has been increasingly used, but most applications have been based on total power demand on a given generation system. In this study, this approach is further refined so that with routinely available data from a number of key substations the substation demand patterns are separately characterized and forecast before being finally combined to generate future system demands. This technique has been found to yield a very satisfactory level of forecasting accuracy in actual applications. © 1986.