Day-ahead electricity demand forecasting with nonparametric functional models

Efficient modeling and forecasting for the electricity demand is an important issue in competitive electricity market. In most electricity markets the daily demand is determined the day before the delivery by means of (semi-)hourly auctions for the following day. Therefore, adequate and reliable day-ahead demand forecasts are very important. In this paper, the forecasting performances of parametric and non parametric models based on the functional approach are compared with those of other standard models, namely univariate AR models, univariate kernel-based nonparametric models and multivariate AR models. Empirical results refer to the next-day demand forecasts for the Italian (IPEX) and British (APX Power UK) electricity markets. Predictive performances are first evaluated by means of descriptive indicators and then through a test to assess the significance of the differences. The analyses suggest that the multivariate approach leads to better results than the univariate one and that, within the multivariate framework, functional nonparametric models are the most accurate, with VAR being a competitive model.