基于卷积神经网络的电网电缆局部放电自动化诊断系统
Automatic Diagnosis System for Partial Discharge of Power Grid Cables Based on Convolutional Neural Network
岳国荣1
作者信息
- 1. 国网北京市电力公司 昌平供电公司,北京 102200
- 折叠
摘要
针对当前电缆局部放电诊断过程中利用信号难以匹配特定放电故障等问题,设计基于卷积神经网络的电网电缆局部放电自动化诊断系统.采集电网电缆局部放电信号,将信号转换为二维图像,将放电信号二维图像作为卷积神经网络的训练样本,训练卷积神经网络;将采集测试样本,基于训练好的卷积神经网络实现电网电缆局部放电自动化诊断.实验结果显示,该系统在卷积核数量为100的条件下,基于GAF时间序列转换图像特征可获取较为满意的诊断结果,降低研究对象局部放电故障时间,提升研究对象运行稳定性.
Abstract
Design an automated diagnosis system for partial discharge of power grid cables based on convolutional neural networks to address issues such as difficulty in matching specific discharge faults with signals in the current process of cable partial discharge diagnosis.Collect partial discharge signals from power grid cables,convert the sig-nals into two-dimensional images,use the two-dimensional images of discharge signals as training samples for convo-lutional neural networks,and train the convolutional neural network.Collect test samples and use a trained convolu-tional neural network to achieve automated diagnosis of partial discharge in power grid cables.The experimental re-sults show that under the condition of 100 convolutional kernels,the system can obtain satisfactory diagnostic results based on GAF time series transformation of image features,reduce the time of partial discharge faults in the research object,and improve the operational stability of the research object.
关键词
卷积神经网络/电网电缆/局部放电/自动化诊断/格拉姆角场/脉冲电压信号Key words
convolutional neural network/power grid cables/partial discharge/automated diagnosis/Gram angle field(GAF)/pulse voltage signal引用本文复制引用
出版年
2024
NETL
NSTL
万方数据