Improving the precision of DC arc detection in high absolute humidity environment

Abstract

Direct current (DC) arcs are among the primary causes of failures in DC systems, including photovoltaic and battery systems. DC arcs are difficult to detect because their noise characteristics are affected by discharge conditions, with humidity being a significant factor. However, most studies have been conducted under fixed humidity conditions. Thus, the mechanism by which humidity affects arc detection remains unclear. This uncertainty hampers reliable detection of DC arcs across a wide humidity range. This study provides a comprehensive investigation into the impact of absolute humidity on DC arc detection, covering arc noise behavior, feature extraction, and classification performance. The experiments reveal that the time-frequency distribution of arc noise consists of ascending and descending stages. Specifically, the amplitude of arc noise was found to be negatively correlated with absolute humidity, while the peak time was negatively correlated with both temperature and current. Furthermore, higher absolute humidity was found to lower the signal-to-noise ratio of arc features and increase misclassification risk, particularly during the noise descending stage. To address this, a concept of detection sweet zone was proposed to define a detection interval range under humid conditions. The sweet zone of a 9.8 A arc shortens from 2.49  s at 90 % relative humidity and 20 °C to 1.23  s at 80 % relative humidity and 45 °C. The detection failure rate increases proportionally as detection interval exceeds the sweet zone, ranging from 1.5 % to 60.1 %. Therefore, determining detection intervals based on the sweet zone can improve detection precision under high absolute humidity.