Ensemble methods for PV power production prediction

The present paper addresses the problem of the estimation of the day-ahead generated power (GP) of a photovoltaic plant using predicted regional solar radiation (SR). The contribution is twofold. First, it investigates the problems related to the design of day-ahead SR predictors and then studies their combination with the prediction data obtained from a meteorological service. Different setups of the time series model, the weather forecast service and their combinations using ensemble methods have been employed. Secondly, under a similar ensemble framework, the impact of the accuracy in the prediction of meteorological variables on the quality of the GP prediction is evaluated.