A Stochastic Approach to Peak Power-Demand Forecasting in Electric Utility Systems

Abstract

A stochastic procedure is described for producing probabilistic forecasts of monthly peak system demands for up to three years ahead. The procedure is based on well-known concepts of prediction theory of stationary stochastic time series, which are developed further to predict those types of nonstationary series that can be reduced to a stationary series by a finite linear transformation. The procedure yields a technique for forecasting the evolving, nonstationary and seasonal peak power demands.