Avaliação da capacidade preditiva de modelos ARIMA e VAR-VEC: o caso da demanda por energia elétrica no Rio Grande do Sul

Abstract

This paper presents the modelling of electricity demand in State of Rio Grande do Sul for the three main consumer sectors: residential, commercial and industrial, through the autoregressive vector model, complemented by the error correction vector model. In this approach, we also considered information regarding energy tariff, GDP, appliances and electrical material and equipment prices. The predictive capacity of all fitted models was compared to the Box-Jenkins framework, specifically, with the autoregressive integrated moving average (ARIMA) models. All models were fitted using data from 1971 to 2010, and their validation were performed from 2011 up to 2017. In general, for all three consumer sections, the best predictive capacity was returned by the ARIMA models. Nevertheless, the other models performed better on one-step-ahead predictions.