Faulty Feeder Detection in Distribution Systems with Resonant Grounding Based on Recognition of Data Distribution Images

Accurate detection and timely isolation of the faulty feeder help ensure the stable operation after the occurrence of a single-phase-to-ground (SPG) fault in the distribution systems with resonant grounding. This paper introduces a novel faulty-feeder detection method based on recognition of data distribution images, and both the detection accuracy and detection efficiency of faulty-feeder identification can be improved significantly. Since the faulty feeder can be distinguished from healthy feeders by comparing the distribution features of the sampled currents, the distribution features of currents are obtained by generating a novel data distribution image, and the feature comparison is conducted by using a convolutional neural networks (CNN). The data distribution image can be acquired quickly owing to the proposed image creation method, and the established CNN with simple structure can be easily implemented. The experimental results by using PSCAD simulation and RTDS test system verify that the proposed method has good performance on detection accuracy and detection efficiency under varied fault scenarios, thus demonstrating its considerable potential for applications.