Classification of fault pattern in string level monitoring of utility-scale PV

The construction of utility-scale PV plants has rapidly increased in recent years due to a rising demand for clean energy and a severe decrease in system costs. To remain competitive, PV asset owners and managers are under a lot of pressure to optimize PV plant operation and reduce energy losses. Daily automated post-processing of monitoring data is crucial for swift fault detection resulting in targeted prescriptive measures. The objective of this master thesis is to detect and diagnose the most common PV operation faults at string, combiner box and inverter levels and derive generally applicable fault patterns. This information is then used to develop required measures for future fault prevention. A large volume of monitoring data from a sample of nine different utility-scale PV plants is used to study the various fault patterns. The evaluated cases include faults in data logging, structural losses, module damage, shading and soiling conditions, module degradation, and inverter failures. Performance as well as weather data is visualized, producing characteristic (deformed) curves for each loss type. Deviations from the normal pattern signal early warnings. In the long run, the objective will be to facilitate automated error detection with a standardized method which relies on typical patterns.7