Photovoltaic energy production forecast using support vector regression

Forecasting models for photovoltaic energy production are important tools for managing energy flows. The aim of this study was to accurately predict the energy production of a PV plant in Italy, using a methodology based on support vector machines. The model uses historical data of solar irradiance, environmental temperature and past energy production to predict the PV energy production for the next day with an interval of 15 min. The technique used is based on ν-SVR, a support vector regression model where you can choose the number of support vectors. The forecasts of energy production obtained with the proposed methodology are very accurate, with the R2 coefficient exceeding 90 %. The quality of the predicted values strongly depends on the goodness of the weather forecast, and the R2 value decreases if the predictions of irradiance and temperature are not very accurate.