Electricity prices forecasting by a hybrid evolutionary-adaptive methodology

With the restructuring of the electricity sector in recent years, and the increased variability and uncertainty associated with electricity market prices, it has become necessary to develop forecasting tools with enhanced capabilities to support the decisions of market players in a competitive environment. Hence, this paper proposes a new hybrid evolutionary-adaptive methodology for electricity prices forecasting in the short-term, i.e., between 24 and 168 h ahead, successfully combining mutual information, wavelet transform, evolutionary particle swarm optimization, and the adaptive neuro-fuzzy inference system. In order to determine the accuracy, competence and proficiency of the proposed methodology, results from real-world case studies using real data are presented, together with a thorough comparison considering the results obtained with previously reported forecasting tools. Not only is the accuracy an important factor, but also the computational burden is relevant in a comparative study. The results show that it is possible to reduce the uncertainty associated with electricity market prices prediction without using any exogenous data, just the historical values, thus requiring just a few seconds of computation time.