Research on Partial Discharge Signal Monitoring and AI-Based Intelligent Diagnosis of Distribution Equipment Based on Multi-in-One Sensors

This study presents a method for monitoring partial discharge (PD) signals and AI-based intelligent diagnosis of distribution equipment, using multi-in-one sensors. The aim is to improve the fault diagnosis accuracy and real-time performance of distribution systems. By integrating ultrasonic, current, and voltage sensors, the method overcomes the limitations of using a single sensor, enabling comprehensive capture of PD signals in distribution equipment. High-frequency signal acquisition technology is employed to achieve precise synchronization of the signals. The AI diagnostic model, based on deep learning, combines Convolutional Neural Networks (CNN) and Long Short-Term Memory networks (LSTM), which significantly enhances the recognition rate and diagnostic accuracy of partial discharge faults. Experimental results show that when using a test dataset of 5000 partial discharge signal samples, the CNN + LSTM fusion model achieves an accuracy rate of 92 %, a recall rate of 90 %, and an F1 score of $\mathbf{9 1 \%}$, which is significantly better than traditional machine learning methods. The multi-in-one sensor system demonstrates significant advantages in high-frequency signal acquisition and anti-interference capability, resulting in a $\mathbf{1 2. 1} \boldsymbol{\%}$ improvement in overall diagnostic accuracy and stable performance in real-world engineering environments. This method provides a new technical path for the fault diagnosis of electrical equipment and has broad prospects for engineering applications.