Neural forecasting of the day-ahead hourly power curve of a photovoltaic plant

In this paper we propose a study to identify the best ANN configuration in terms of number of neurons, number of layers, training-set size, in order to perform the day-ahead energy production forecast for a Photo-Voltaic (PV) plant. This set up is applied to a novel hybrid method (PHANN Physic Hybrid Artificial Neural Network) in order to enhance the energy day-ahead forecast combining both the deterministic Clear Sky Solar Radiation Algorithm (CSRM) and the stochastic ANN method. This hybrid method has been tested on different PV plants in Italy. Finally, different weather conditions have been taken into account to test the robustness of the algorithm as well as the effects of particular events on the forecasting output.