Eletrical energy supply for Rio Grande do Sul, Brazil, using forecast combination of weighted eigenvalues

Abstract

The technique of forecast combination (FC) can achieve results superior to those generated by individual forecasting techniques. The purpose of this research is to present a method of FC based on the principal component analysis (PCA) applied to forecast values originating from individual forecasting models such as ARIMA, ARFIMA and SARIMA and their variants. From the PCA, it is possible to weigh up the values from each individual model in order to obtain a linear combination which represents all characteristics of valid models for each supplier. As an example, the FC method proposed was applied in the industrial sector of the three largest electric power suppliers in the State of Rio Grande do Sul, Brazil. The proposed method proved to be very useful since it presented better results than the individual models.