Semantic Segmentation-Based Intelligent Threshold-Free Feeder Detection Method for Single-Phase Ground Fault in Distribution Networks

Feeder detection for single-phase ground fault (SPGF) is challenging in a resonant grounded system due to the difference in feeder capacitance to ground and the influence of the arc suppression coil. This article uses semantic segmentation algorithms to implement feeder detection for SPGF in distribution networks. The proposed method overlays transient zero-sequence voltage (ZSV) derivatives and transient zero-sequence current (ZSC) waveforms on the same image. Then, a semantic segmentation algorithm is used to classify the pixel points of the image. The segmentation map output by the semantic segmentation algorithm contains category prediction results for each pixel in the input image. Detecting faulty feeders based on the number of pixels of different categories in the segmentation map can make the final decision-making process more transparent and easy to understand. The validity and adaptability of the proposed method have been confirmed through tests using both simulation and field data. The proposed method achieves an accuracy of over 95% on simulated data, even in the presence of noise interference and asynchronous sampling, etc. The proposed method, furthermore, achieves an accuracy of over 99% when applied to full-scale test data.