DC Arc Fault Detection For Grid-Connected Large-Scale Photovoltaic Systems

Abstract

The detection and localization of dc arc faults is a major problem in large grid-connected photovoltaic systems. Hence, a time-domain technique based on the mathematical morphology called the decomposed open–close alternating sequence (DOCAS) is proposed in this article for the detection and localization of such fault conditions. The dc arc is usually sustained if sufficient fault current exists, and this phenomenon is captured by generating sustained random spikes at the output of the DOCAS algorithm that correlates to the rate of change in the dc arc current and voltage signals to detect incidences of dc arc faults when they occur. The proposed method further incorporates a technique for fault localization based on the increased effective resistance under dc arc fault conditions. Moreover, the effective fault resistance is further used as a noise filtering function for noise suppression. The method has been tested under different levels of low irradiances, fault location, and noise condition, and the results are presented to demonstrate the characteristics of the proposed technique.