基于残差卷积神经网的高压直流故障测距方法  

UHVDC Transmission Fault Location Based on Residual Convolutional Neural Network

作　　者：李冀 张学友 张俊杰 施雯 阮巍 戴剑丰 LI Ji;ZHANG Xueyou;ZHANG Junjie;SHI Wen;RUAN Wei;DAI Jianfeng(EHV Branch,State Grid Anhui Electric Power Co.,Ltd.,Hefei 211525,China;College of Automation&College of Artificial Intelligence,Nanjing University of Posts and Telecommunications,Nanjing 210023,China)

机构地区：[1]国网安徽省电力有限公司超高压分公司,安徽合肥211525 [2]南京邮电大学自动化学院、人工智能学院,江苏南京210023

出　　处：《哈尔滨理工大学学报》2024年第5期94-104,共11页Journal of Harbin University of Science and Technology

基　　金：国家自然科学基金(52207009)。

摘　　要：针对传统的特高压直流输电线路故障测距过度依赖线路本身衰减常数的问题,提出了一种基于残差卷积神经网的高压直流故障测距方法。利用残差卷积神经网络(Residual Convolution Neural Network,RCNN)的非线性拟合能力,不用计算线路衰减常数就可以实现故障准确定位。首先,使用小波变换提取不同频带的故障电压、电流,将其作为残差卷积神经网络的输入;然后,对输入数据进行故障特征学习,训练网络,更新参数,最终形成特高压直流输电线路故障测距模型,输出故障测距参数;最后,经过实验结果表明,基于残差卷积神经网的高压直流故障测距方法检测速度快、精度高,并且在过渡电阻接入下仍然具有很高的精度,验证了该方法的有效性。A method based on residual convolutional neural network for high-voltage direct current transmission fault location is proposed to address the issue of excessive reliance on the attenuation constant of the transmission line in traditional ultra-high voltage direct current fault distance measurement.Using the nonlinear fitting ability of residual convolution neural network(RCNN),the accurate fault location can be realized without calculating the line attenuation constant.Firstly,wavelet transform is used to extract fault voltage and current of different frequency bands,which are used as input of residual convolution neural network.Then,the input data are used to learn the fault characteristics,train the network,update the parameters,and form the fault ranging model of UHVDC transmission line and output the fault ranging parameters.Finally,experimental results demonstrate that the fault distance measurement method for high-voltage direct current based on residual convolutional neural network achieves fast detection speed and high accuracy.Moreover,it still maintains a high level of accuracy even when there is transitional resistance connected,validating the effectiveness of this method.

关 键 词：特高压直流输电 故障测距 衰减常数 残差卷积神经网络 小波变换 

分 类 号：TM723[电气工程—电力系统及自动化]