Day-Ahead Electricity Price Forecasting Using an Intelligent Algorithm Based on Feature Selection

Restructuring in power systems has caused electricity price forecasting became one of the most important challenges facing electricity market participants. The precise electricity price forecasting helps both consumers and producers to make better decision in order to maximize their benefit. In this paper, the historical data of electricity price and energy consumption are utilized for prediction of electricity price for the next 24 hours. The proposed intelligent algorithm is realized through three main steps: 1- preprocessing step, 2- feature selection and 3- forecasting step. At first, the price signal is decomposed to different modes by using Empirical Wavelet Transform (EWT). Afterward, in the second step, the feature selection method based on mutual information is applied on input data to improve the performance of forecasting engine. In the third step, for day-ahead hourly electricity price forecasting, the Support Vector Regression (SVR) is trained by selected features. The performance of the proposed algorithm is evaluated using real data of two electricity markets i.e. Pennsylvania New-Jersey Maryland (PJM) and Operador del Mercado Ibérico de Energía-Polo Español (OMIE).