Incipient Fault Detection Method for Transmission Lines Based on Multi-Scale Neural Network

The detection of incipient faults in transmission lines is crucial for improving the reliability of power systems. Incipient faults often exhibit weak waveform characteristics, making them difficult to distinguish from disturbances such as load variations and fluctuations in renewable energy sources. To address this challenge, this paper proposes an incipient fault detection method based on a multi-scale neural network (MSNN). The proposed approach analyzes disturbance waveforms in the time-frequency domain and employs a multi-scale convolutional neural network to identify characteristic differences between faults and disturbances across different time scales, thereby improving detection accuracy and system robustness. The experimental results show that this method achieves 100% identification accuracy on the simulated dataset and 95.25% identification accuracy on the actual measurement dataset.