基于小波包分解卷积神经网络的停运输电线路故障识别方法

王鑫明; 王祥宇; 贾晓卜; 张飞飞; 李少博; 胡永强

王鑫明,王祥宇,贾晓卜,张飞飞,李少博,胡永强.基于小波包分解卷积神经网络的停运输电线路故障识别方法[J].电测与仪表,2025,62(1):61-67.
Wang Xinming,Wang Xiangyu,Jia Xiaobo,Zhang Feifei,Li Shaobo,Hu Yongqiang.Fault tidentification method of outage transmission line based on convolutional neural network and wavelet packet decomposition[J].Electrical Measurement & Instrumentation,2025,62(1):61-67.

基于小波包分解卷积神经网络的停运输电线路故障识别方法

Fault tidentification method of outage transmission line based on convolutional neural network and wavelet packet decomposition

DOI：10.19753/j.issn1001-1390.2025.01.008

中文关键词: 同塔双回输电线路感应电压小波包分解时频分析卷积神经网络故障识别

英文关键词:double-circuit transmission lines on the same tower, induced voltage, wavelet packet decomposition, time-frequency analysis, convolutional neural network, fault identification

基金项目:国网河北省电力有限公司科技项目：基于空间感应电压特征的全过程接地安全智能监测技术研究（5204BB200028）

作者中文名	作者英文名	单位
王鑫明	Wang Xinming	国家电网河北省电力调度控制中心
王祥宇	Wang Xiangyu	华北电力大学电气与电子工程学院
贾晓卜	Jia Xiaobo	国家电网河北省电力调度控制中心
张飞飞	Zhang Feifei	国家电网河北省电力调度控制中心
李少博	Li Shaobo	国家电网河北省电力调度控制中心
胡永强	Hu Yongqiang	华北电力大学电气与电子工程学院

摘要点击次数: 326

中文摘要:

当输电线路处于热备用状态时，停运线路上仍可能发生短路故障，准确地判断停运线路的故障状态能有效地避免合闸到故障线路时对电力系统造成冲击并对故障的排除提供便利，因此有必要对停运输电线路进行故障识别。对于双回输电线路提出一种采用小波包分解生成的频谱图作为卷积神经网络（convolutional neural network, CNN）输入进行特征提取的停运线路故障识别方法。为减少人为提取特征产生的误差，首先对停运输电线路故障时三相电压暂态波形进行测量，采用小波包分解得到三相电压波形时频特性，最终通过CNN提取特征并进行故障分类。为验证该方法的故障识别效果，以河北省3条线路的实际数据为基础，在ATP-EMTP中建立500 kV同塔双回输电线路模型，为模拟现场各因素产生的误差在测得电压波形中加入10 dB高斯白噪声。结果表明，对热备用线路上故障状态识别准确率为99.98%，在一定程度上为停运线路的故障诊断及排除提供了参考。

英文摘要:

When the transmission line is in a hot standby state, a short circuit fault may still occur on the shutdown circuit. Accurate judgment of the fault state of the shutdown line can effectively avoid impact on power system and facilitate the failure troubleshooting. Therefore, it is necessary to identify the fault state of outage transmission line. For double-circuit transmission lines on the same tower, an outage line fault identification method using spectrum graph generated by wavelet packet decomposition as a convolutional neural network (CNN) input is proposed. To reduce the error of human extraction features, the faulty three-phase voltage waveform of the outage transmission line are measured, and wavelet packet decomposition is used to obtain three-phase voltage waveform spectrum maps, the features are extracted by CNN, meanwhile, CNN also undertakes the task of fault classification. The fault identification effect of this method is verified by simulation experiments. Based on the actual data of 3 lines in Hebei Province, 500 kV double-circuit transmission lines on the same tower models was established in ATP-EMTP to obtain the waveform of voltage, and 10 dB Gaussian white noise was added to the measured voltage waveform for simulating the error generated by various factors in the field. The results show that the identification accuracy of fault status on the hot standby line is 99.98%, which provides reference for fault diagnosis and troubleshooting of outage lines to a certain extent.

查看全文查看/发表评论下载PDF阅读器

关闭