Comparisons of machine learning methods for electricity regional reference price forecasting

Abstract

Effective and reliable electricity price forecast is essential for market participants in setting up appropriate risk management plans in an electricity market. In this paper, we investigate two state-of-the-art statistical learning based machine learning techniques for electricity regional reference price forecasting, namely support vector machine (SVM) and relevance vector machine (RVM). The study results achieved show that, the RVM outperforms the SVM in both forecasting accuracy and computational cost. © 2009 Springer Berlin Heidelberg.