应用LSTM-RNN的特高压直流输电系统继电保护故障检测方法  

Research on fault detection method of UHV DC transmission system relay protection based on LSTM-RNN

作　　者：张学友 石永建 李冀 郭振宇 戴剑丰 ZHANG Xueyou;SHI Yongjian;LI Ji;GUO Zhenyu;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]南京邮电大学自动学学院、人工智能学院,江苏南京210023

出　　处：《中国测试》2025年第3期177-184,共8页China Measurement & Test

摘　　要：为解决传统特高压直流保护对高阻故障检测准确率不高、故障检测时间过长以及故障选极不完善的问题,提出基于长短时记忆(long short term memory,LSTM)循环神经网络(recurrent neural network,RNN)的特高压直流输电线路继电保护故障检测方法。首先,基于快速傅里叶变换分析特高压直流输电系统暂态故障特征,使用相模变换和小波变换提取出故障特征量作为输入数据。其次,将输入数据输入到LSTM-RNN中进行前向传播,对系统故障特征进行深度学习,同时使用反向传播方式更新网络参数,将深层的特征量输入到Softmax分类器中进行分类,把故障识别分成区外故障、母线故障和线路故障,故障分类为正极故障、负极故障和双极故障,并输出识别结果。最后,在PSCAD/EMTDC仿真条件下,搭建特高压直流输电模型。验证结果表明:所提的方法在特高压直流输电线路继电保护的故障检测、故障选极上具有更好的效果,相比于人工神经网络、卷积神经网络、支持向量机,故障识别准确率分别提升4.71%、6.57%、9.32%。To solve the problems of low accuracy,long fault detection time,and incomplete fault pole selection in traditional ultra-high voltage DC protection for high resistance faults,a relay protection fault detection method for ultra-high voltage DC transmission lines has been proposed based on long short term memory(LSTM)recurrent neural network(RNN).Firstly,transient fault characteristics of ultra-high voltage direct current transmission systems are analyzed based on fast Fourier transform,and fault feature quantities are extracted as input data using phase modulus transform and wavelet transform.Secondly,the input data is input into LSTM-RNN for forward propagation to deeply learn the system fault features.At the same time,backpropagation is used to update network parameters,and the deep feature quantities are input into Softmax classifier for classification.Fault identification is divided into external faults,busbar faults,and line faults.Faults are classified into positive faults,negative faults,and bipolar faults,and recognition results are output.Finally,under PSCAD/EMTDC simulation conditions,a validation method was used to build an ultra-high voltage direct current transmission model.The validation results showed that the proposed method has better performance in fault detection and fault pole selection of ultra-high voltage direct current transmission line relay protection.Compared to artificial neural networks,convolutional neural networks,and support vector machines,the fault recognition accuracy has been improved by 4.71%,6.57%,and 9.32%,respectively.

关 键 词：LSTM-RNN 特高压直流输电线路 继电保护 快速傅里叶变换 故障识别 

分 类 号：TB9[一般工业技术—计量学] TM73[机械工程—测试计量技术及仪器]