Previsão e índice de incerteza da tarifa de energia elétrica brasileira

Abstract

Energy plays a crucial role in the economy, influencing various social and economic aspects. Due to its importance, the field of Electricity Price Forecasting was developed, aiming to select models for predicting this variable. The objective of this study is to compare the predictive efficiency of Brazilian electricity tariffs using univariate and multivariate time series models, as well the construction and consideration of a tariff uncertainty index, through which variations arising from moments of instability in the electricity market, caused by water scarcity and regulatory policies, can be visualized. Specifically, it compares forecasts obtained through autoregressive and moving average models (ARMA), vector autoregressive models (VAR), and factor-augmented vector autoregressive models (FAVAR). Among the main results, the simple AR(1) model stands out with the best out-of-sample forecasting indicators.