基于卷积神经网络的开关柜局部放电监测

Switch-Gear Partial Discharge Detection Based on Convolutional Neural Network

提出了一种通过人工智能和机器学习理论来实现配网设备运行状况监测的方法,并展示了一套基于卷积神经网络使用超声波和地电波数据的开关柜自动局部放电现象的监测系统。设计了具体的卷积神经网络模型,使其在经过离线训练之后,可以对开关柜实际运行时所产生的信号进行实时的分析与辨别,进而判断设备的运行状况。实验结果表明,在错误信号进行降噪处理的前提下,该系统依然可以准确地给出设备运行的确切状况,这充分证明了提出方法的有效性。

This paper presents a method based on artificial intelligence and machine learning theory to realize the monitoring the operation status of distribution network equipment, and it also displays a system which uses the ultrasonic and ground wave data and is based on convolutional neural network to monitor the partial discharge of the switch cabinet. A concrete convolutional neural network model is designed in the paper, which after the off-line training, can be used to analyze and identify some of the real-time signal generated by the switch cabinet in actual operation, and further judge the running status of equipment. The experimental results show that in the premise of noise reduction of the error signal, the system can accurately indicates the exact situation of equipment operation, which fully demonstrates the effectiveness of the proposed method.