摘要
受故障信号微弱、配电网存在噪声干扰等因素的影响,高阻接地故障情况下行波波头提取和检测困难,导致基于行波信号的高阻故障检测方法可靠性不高。针对上述问题,提出一种基于行波全景故障特征自辨识的高阻接地故障检测方法。首先,借助行波全景波形对高阻接地故障与正常暂态扰动电压行波信号的时-频差异性进行分析;然后,搭建卷积注意力模块-卷积神经网络(convolutional block attention module-convolution neural network,CBAM-CNN)模型,使其较传统的卷积神经网络(conrolution neral network,CNN)模型更具抗干扰能力,将行波全景波形以灰度图形式输入卷积神经网络,实现对多维故障特征的提取与利用;最后,在PSCAD上搭建10 kV配电网模型进行各种故障条件下的仿真分析。结果表明:所提方法能够可靠检测高阻接地故障,抗噪性能良好,且不受故障位置、过渡电阻、初相角的影响,大大提高了基于行波信号的高阻接地故障检测方法的可靠性与灵敏性。
Influenced by weak fault signal and noise interference in distribution network,it is difficult to extract and detect the downlink wavefront in the case of high impedance fault(HIF),which results in the low reliability of the traveling-wave-based HIF detection method.To solve the above problems,a novel HIF detection method based on traveling wave full waveform fault feature self-identification is proposed.Firstly,the time-frequency difference between high-resistance ground fault and normal transient disturbance voltage traveling waveform is analyzed with the help of traveling waveform panorama waveform;Then,the CBAM-CNN model is built to make it more anti-interference than the traditional CNN model,and the traveling wave full waveform is input into the convolution network in the form of gray image to obtain the feature representation with more anti-interference ability,so as to realize the extraction and utilization of multi-dimensional fault features.Finally,a 10 kV distribution network model is built on PSCAD for simulation analysis under various fault conditions.The results show that the proposed method can reliably detect the HIF with good anti-noise performance,and is not affected by fault location,transition resistance,inception angle,which greatly improves the reliability and sensitivity of the HIF detection method based on traveling wave signals.
作者
胡一鸣
史鸿飞
张玉龙
王钰清
刘嘉庆
袁军
邓丰
HU Yiming;SHI Hongfei;ZHANG Yulong;WANG Yuqing;LIU Jiaqing;YUAN Jun;DENG Feng(College of Electrical and Information Engineering,Changsha University of Science and Technology,Changsha 410114,China)
出处
《供用电》
2023年第4期39-46,54,共9页
Distribution & Utilization
基金
2022年国家级大学生创新训练计划项目(S202210536019)。
关键词
配电网
高阻接地故障
行波全景波形
卷积神经网络
注意力机制
distribution network
high impedance fault(HIF)
traveling wave full waveform
convolution neural network
attention mechanism
作者简介
胡一鸣(2003-),女,本科在读,研究方向为电力系统保护与控制;史鸿飞(1999-),男,硕士研究生,研究方向为电力系统保护与控制;张玉龙(2002-),男,本科在读,研究方向为电力系统保护与控制;通信作者:邓丰(1983-),女,博士,副教授,研究方向为电力系统微机保护、故障行波保护及故障定位。
