A comparison of the forecast performance of double seasonal ARIMA and double seasonal ARFIMA models of electricity load demand

Abstract

A half-hourly electricity load demand of Malaysia for 6 months, from 1 January 2010 to 30 June 2010 is used in this study with the purpose of improving the accuracy of short term electricity load demand forecasts. The results of the identification step show that the load data have daily and weekly seasonal periods. Thus, the aim of this paper is to develop a forecasting model by studying the long-term characteristics of time series based on double seasonal ARFIMA model. The best order of the model which is based on fractional referencing is determined and the results are compared with the established seasonal ARIMA model. Using the mean absolute percentage error (MAPE) as the forecast accuracy measure, the current study shows that double seasonal ARFIMA model performs better than double seasonal ARIMA in forecasting electricity load demand in Malaysia where the mean absolute percentage error of the forecast value is reduced by about 0.04%.