Data-driven estimation of expected photovoltaic generation

Photovoltaic (PV) systems' monitoring and performance assessment are relevant tools to ensure its correct operation. The standard methodology consists in determining irradiance in the plane of the array and convert it into PV power, considering system efficiency. This work proposes an effective data-driven approach based on past PV generation of the system being assessed and local measurements of solar radiation. Using local global horizontal irradiance (GHI) measurements, model performance degrades with the difference in tilt between the PV module and the pyranometer; when radiation is measured in the plane of the array (in this case for a horizontal module) the data driven method is more accurate than the standard approach. The method was also tested for a small-scale residential PV system context, replacing the local irradiation measurements by satellite GHI estimates. Although errors increase significantly, the data-driven method outperforms the standard approach for all tilts.