Artificial neural network for probabilistic forecasting of the output power of photovoltaic systems

摘要

The article discusses the use of a probabilistic artificial neural network to estimate the electric energy production of a solar photovoltaic system using measured data from real systems made available in an energy forecast competition. The precision of the series generated by the neural networks will be evaluated by the pinball loss function, metric used for the evaluation of the probabilistic prediction of quantiles. The results show that the modeled network can realize a good forecast of the electric power production for the majority of analyzed months.