Artificial Intelligence Fault Detection Method in Resonant Grounding System Based on Waveform Characteristics

Single-phase grounding fault occurs frequently in distribution network, and its fault characteristics are weak and difficult to extract effectively. In this paper, the waveforms of zero-sequence voltage derivative and current are analyzed, and the waveforms after half cycle of fault are superimposed on an image. Secondly, three kinds of convolutional neural networks are constructed to achieve adaptive fault feature extraction, and then an artificial intelligence fault detection method in resonant grounding system is proposed. Finally, the simulation model is used to obtain fault samples to train and test the convolutional neural networks. The results show that under the conditions of different fault locations, fault phase angle and fault grounding impedance, the method based on waveform characteristics and convolutional neural network can identify the faulty line quickly and accurately, and has the ability to resist interference.