Outlier detection rules for fault detection in solar photovoltaic arrays

Abstract

Solar photovoltaic (PV) arrays are unique power sources that may have uncleared fault current when utilizing conventional overcurrent protection devices. To monitor the PV operation and detect these unnoticed faults, outlier detection rules have been proposed for fault detection based on instantaneous PV string current. This paper discusses three rules in detail: 3-Sigma rule, Hampel identifier, and Boxplot rule. Unlike other methods, the proposed methods do not require weather measurement or efforts in model training. Our experimental results show that Hampel identifier and Boxplot rule may be recommended for PV fault detection. Furthermore, the proposed models become more reliable as the number of PV measurements increases. The developed methods may be integrated with PV monitoring system for real-time operation.