Online short-term forecasting of photovoltaic energy production

Short-term forecasting of the energy production is one of the key issues in smart homes that tend to achieve efficient balance among the energy production, storage and consumption. In this paper, we first perform an analysis of the features to be used by the most promising short-term forecast model: artificial neural networks. We determine the best performing offline model and then propose an online model that is very close to the offline model in terms of prediction accuracy. The evaluation is performed on a real world data and the resulting system is part of a proof-of-concept application for microgrid management.