Performance index of photovoltaic fields for diagnostic purposes

This paper deals with the issue to supervise the state of health of a PhotoVoltaic (PV) system by means of a cheap approach based only on the processing of the data stored into a standard monitoring system, thus no specific equipment is requested. The proposed solution is based on a linear predictor of an index strictly related to the producible energy, depending on the environmental conditions and on two coefficients, associated to the PV plant under test. The proposed approach is constituted by two steps: the definition of the predictor and the training needed to calculate two characteristic coefficients. The mismatch between the expected and the calculated predictor is a reliable marker of the health state of the PV system. The greater the mismatch, the worse its health. Finally, the methodology identifies the performance decrease, with a good confidence level about false alarm and identification delay.