Predictive methods of electricity price: an application to the Italian electricity market

Abstract

Price forecasting is a crucial element for the members of the electricity markets and business decision making to maximize their profits. The electricity prices have an impact on the behavior of market participants, and thus, predicting prices for generation companies, and consumers is essential for both the short-term profits in the Day-Ahead, Intra-Day and Ancillary markets, and the long-term benefits in the future planning, investment, and risk management. Therefore, participants in the electricity market need to accurately and effectively predict the price signal to manage market risk. In this paper, different forecasting models have been compared, and the most promising ones have been employed to forecast the short term Italian electricity market clearing price for achieving forecasting accuracy. In particular, simulations are performed for four principal regression methods, including Support Vector Machine, Gaussian Processes Regression, Regression Trees, and Multi-Layer Perceptron. The performance of predicted models is compared through several performance metrics, including MAE, RMSE, R, and the total number of percentage error anomalies. The results indicate the SVM is the best choice for forecasting the electricity market price on the Italian case study.