Extreme Learning Machine for Short and Mid-term Electricity Spot Prices Forecasting

Abstract

In a deregulated electricity market, market participants define trading strategies and models to assist the decision-making process. Companies whose operation is heavily dependent on this market are increasingly adopting electricity price forecasting models to identify sales and purchase contracts with the best price. This paper intends to contribute to the improvement of the decision-making process for purchasing electricity in the Iberian Electricity Market. The purpose is to develop a forecasting model for electricity spot prices based on prices established on the derivatives markets. The model uses Artificial Neural Networks trained with the Extreme Learning Machine algorithm to determine the monthly average spot prices for the next six months and provides a tool for making trading decisions considering the risk of exposure to spot market volatility. The forecasting model was applied in two scenarios: pre-pandemic and pandemic. The results prove that its application can contribute to improving decision-making for trading electricity in the short/medium term. Experimental results considering both scenarios show that the proposed model can provide month-ahead forecasts with an RMSE up to 6.38 €/MWh.