Summary
Single‐phase‐to‐ground faults (SFs) could be characterized by multiple aspects, like time domain and frequency domain. There is no unified classification standard for such faults. Previous studies usually focused on detection and feeder‐selection techniques for SFs. This paper presents the classification architecture and recognition method for SF in nonsolidly earthed distribution networks. On the basis of the waveform data recorded by remote fault indicators and other recording devices, steady and transient characteristics of SFs are analyzed, and the faults are defined and classified into multiple levels. In each level, different kinds of SFs are recognized by quantization of time domain features, frequency domain features, and etc. It is rigorously tested through field data and artificial test data. Test results show that the proposed method could effectively recognize different kinds of SFs.