Forecasting electricity consumption with structural time series models

Abstract

The time series of New Zealand's weekly electricity consumption is modelled by the structural time series models as proposed by Harvey ( 1984 ). These models are based on the traditional decomposition of a time series into trend, seasonal and irregular components and because they are in state space form use is made of the Kalman filter. The characteristics of the time series under study lead to a number of refinements of earlier models. Results of estimation and forecasting are presented.