基于CBAM-FCN的高压输电线路发展性故障识别方法  

An Evolved Faults Identification Method of HV Transmission Lines Based on CBAM-FCN

作　　者：刘志远 于晓军 罗美玲 林泽暄 郝治国[2] 张宇博 杨松浩 LIU Zhiyuan;YU Xiaojun;LUO Meiling;LIN Zexuan;HAO Zhiguo;ZHANG Yubo;YANG Songhao(State Grid Ningxia Electric Power Co.,Ltd.,Yinchuan 750001,Ningxia,China;Xi’an Jiaotong University,Xi’an 710049,Shaanxi,China)

机构地区：[1]国网宁夏电力有限公司,宁夏银川750001 [2]西安交通大学,陕西西安710049

出　　处：《电网与清洁能源》2022年第9期25-33,44,共10页Power System and Clean Energy

基　　金：国家电网有限公司科技项目(5229NX19008P)。

摘　　要：高压输电线路的发展性故障往往受运行状态和故障发展过程的影响,给故障甄别和保护决策造成困难。为了准确识别高压输电线路的发展性故障,保证继电保护装置动作的正确性,将全卷积神经网络(fully convolutional network,FCN)与卷积注意力模块(convolutional block attention module,CBAM)相结合,提出一种基于CBAM-FCN的发展性故障识别方法,通过在传统全卷积网络中引入CBAM模块,使神经网络能够聚焦于故障波形的突变、幅值等重要特征,忽略无关干扰。此外,所提方法能够输出表征故障状态变化的一维时序序列,实现对输电线路发展性故障的全过程识别。最后大量仿真验证了所提方法的抗噪性能和泛化能力,并通过可视化技术展示了网络模型的可解释性。Evolved faults of HV transmission lines often involve multiple operating states and faults development process,which cause difficulties in fault identification decisionmaking of relay protection.In order to accurately identify the evolved fault in HV transmission lines and ensure the action correctness of the relay protection device,this paper proposes an evolved fault recognition method based on CBAM-FCN by combining full Fully Convolution Network with Convolution Block Attention Module.In the paper,CBAM is introduced into the traditional FCN to focus on important features like mutation and amplitude of fault waveforms,and ignore irrelevant interference,which greatly improves the identification accuracy of complex faults.In addition,the proposed method can output one-dimensional series representing the change of the fault state,and realize the whole process identification of evolved faults in transmission lines.Finally,a large number of simulation results have verified the anti-noise and generalization ability of the proposed method,which has also demonstrated its interpretability through visualization technology.

关 键 词：输电线路故障识别 发展性故障 故障识别方法 全卷积神经网络 卷积注意力模块 

分 类 号：TM773[电气工程—电力系统及自动化]