Previsão de médio prazo do consumo de energia elétrica no Brasil: estimação via metodologia box & jenkins e regressão dinâmica

Abstract

The main point of this dissertation is to find a monthly forecasting model to the Brazilian of electric energy consumption. The methodology consists of the construction of a model using a buttom-up strategy. In other words, it was first adjusted a Box & Jenkins model; i.e., a simple univariate model that could give information about the behavior of the series. Then, a dynamic regression model was fitted which relates brazilian electric energy consumption to all possible explanatory variable. In this dissertation, it was also carried out a study of the Brazilian electricity sector evolution, emphasizing the changes occurred in the nineties, that originated the so called “brand new” institutional model of the Brazilian electricity sector. By mean of the Box & Jenkins method, a SARIMA model (0,1,0) X (1,0,0)12 was found, and the dynamic regression model shows that the consumption of electricity in Brazil is related, among others factors, to the level of the economic activities in the country. Finally, tests in and out of sample were made, with the objective of comparing the obtained models, and the monthly forecasts for the 2008 months were produced by the selected model.