Modeling and forecasting of electricity spot-prices: Computational intelligence vs classical econometrics

In European countries, the last decade has been characterized by a deregulation of power production and electricity became a commodity exchanged in proper markets. This resulted in an increasing interest of the scientific community on electricity exchanges for modeling both market activity and price process. This paper analyzes electricity spot-prices of the Italian Power Exchange (IPEX) and proposes three different methods to model prices time series: a discrete-time univariate econometric model (ARMA-GARCH) and two computational-intelligence techniques (Neural Network and Support Vector Machine). Price series exhibit a strong daily seasonality, addressed by analyzing separately a series for each of the 24 hours. One-day ahead forecasts of hourly prices have been considered so to compare the prediction performances of three different methods, with respect to the canonical benchmark model based on the random walk hypothesis. Results point out that Support Vector Machine methodology gives better forecasting accuracy for price time series, closely followed by the econometric technique