Analysis of Automatic Fault Detection Methods for Commercially Operated PV Power Plants

Automatic fault detection (AFD) is a key technology to optimize the Operation and Maintenance of photovoltaic (PV) systems portfolios. Although several recent studies suggest methods and procedures for automatic fault detection, they lack a common basis for testing their effectiveness under real operating conditions. In this study, we have selected and implemented a list of methods suitable for the data available in the standard monitoring of commercial systems. The performance of these methods was tested using measurements collected over 46 months on 85 rooftop-type PV systems installed in Germany. The maintenance records from these systems were used to measure the accuracy of the fault detection alerts and to analyze the sensitivity and specificity of each method. The best performing was able to identify almost one third (29.6%) of the faults observed in PV systems with an average of 15.8% false alerts. The method with the least false alerts (11.7%) could identify 19% of the faults. The AFD methods tested show different sensitivity depending on the type of fault, which suggests that a combination of methods can increase the overall sensitivity.