Previsão multi-passos a frente do preço de energia elétrica de curto prazo no mercado brasileiro

Abstract

Electricity price forecasting is an important issue to all Market participants in order to decide bidding strategies and to establish bilateral contracts, maximizing their profits and minimizing their risks. Energy price typically exhibits seasonality, high volatility and spikes. Also, energy price is influenced by many factors such as power demand, weather, and fuel price. This work proposes a new hybrid approach for short-term energy price prediction. This approach combines auto-regressive integrated moving average (ARIMA) and neural network (ANN) models in a cascaded structure and uses explanatory variables. A two step procedure is applied. In the first step, the selected explanatory variables are predicted. In the second one, the energy prices are forecasted by using the explanatory variables prediction. The proposed model considers a multi-step ahead price prediction (12 weeks-ahead) and is applied to Brazilian market, which adopts a cost-based centralized dispatch with unique characteristics of price behavior. The results show good ability to predict spikes and satisfactory accuracy according to error measures and tail loss test when compared with traditional techniques. Additionally, is proposed a classifier model consisting of ANN and decision trees in order to explain the rules of price formation and, together with the predictor model, acting as an attractive tool to mitigate the risks of energy trading.