Fault-Diagnosis Architecture for Large-Scale Photovoltaic Power Plants That Does Not Require Additional Sensors

Abstract

To reduce implementation, operation, and maintenance costs of large-scale photovoltaic (PV) power plants, we improved upon our previously developed fault-diagnosis architecture that does not require additional sensors. The key technique in this architecture is that, instead of using the irradiation and module temperature measured using pyranometers and thermometers, it uses their theoretical values calculated from the PV-module characteristics and measured dc voltage and current of a PV inverter. Failure and degradation under both maximum power point tracking and clipping of output power are presumed with this technique. This improved architecture is applied to monitoring systems of large-scale PV power plants. It is experimentally demonstrated that this architecture can determine the number of fault modules even for a failure rate of only less than 0.2%. It was also demonstrated that this architecture can classify failure and degradation modes, namely, “series-resistance increase,” “shunt-resistance decrease,” and “bypass diode on,” even for a loss of less than 5%–6%.