Photovoltaic system fault identification methodology based on I-V characteristics analysis

In the photovoltaic field, regarding the importance of sustainability, monitoring systems are a paramount component for yield assessment. Yet in the industrial production, fault detection remains a manually handled issue. However, faults are responsible for significant power loss and sometimes, even dramatically damages such as fire hazard or material deterioration. This paper presents a methodology for fault detection in the photovoltaic systems regarding the different impacts of faults on the I-V curve. Indeed, fault classification is a crucial step for failure diagnosis. The proposed method mainly covers uniform dust faults, partial shading faults, short circuit faults, and aging. The proposed algorithm relies on electrical indicators that are extracted from single diode model and measured I-V curves and used as assessment parameters, the shape assessment parameter is recognized by training a neural network. This method is assessed through a simulation and validated using experimental I-V curves measurements in faulty cases.