Machine Learning on Buildings Data for Future Energy Community Services
Résumé
The decentralization of energy generation and transmission demand for the Internet of Energy with real-time peer-to-peer energy exchange. The intermittent renewable energy provides a challenge of consumption flexibility at consumer side. Buildings are significant in this new context, since they are the biggest energy consumer worldwide and can contribute to local renewable energy production as well. Therefore, proactive building services will be required in near future in order to operate energy community. This paper deals with energy flexibility obtained through PV production forecast for a community. Decision tree technique is applied on historical hourly data of 3 years for week-ahead forecast of building consumption, photovoltaic production, and for fault detection diagnostic. Feature engineering and energy expertise used to obtain good forecasting performances are discussed. The significance of these technics for prospective energy community service is also discussed in the paper.
Origine : Fichiers produits par l'(les) auteur(s)