基于CNN-SVM的特高压三端混合直流线路故障区域识别方法  

Fault Region Identification Method of UHV Three-terminal Hybrid DC Line Based on CNN-SVM

作　　者：周前华 陈仕龙[1] 邓健 毕贵红[1] 魏荣智 ZHOU Qianhua;CHEN Shilong;DENG Jian;BI Guihong;WEI Rongzhi(Faculty of Electric Power Engineering,Kunming University of Science and Technology,Kunming 650500,China)

机构地区：[1]昆明理工大学电力工程学院,云南昆明650500

出　　处：《电力科学与工程》2024年第4期21-30,共10页Electric Power Science and Engineering

基　　金：国家自然科学基金资助项目(52067009)。

摘　　要：提出一种基于卷积神经网络–支持向量机(Convolutional neural network-support vector machine,CNN-SVM)的特高压三端混合直流线路故障区域识别方法。首先,对昆北侧、龙门侧的直流线路边界和柳北侧T区边界的频率特性进行分析,发现不同故障区域的故障特征存在一定差异。然后,使用经验小波变换提取故障特征,将其作为CNN-SVM的输入量,故障区域作为输出量,构建并训练CNN-SVM模型;将由测量点得到的故障特征量输入到训练完成的CNN-SVM模型中,进行故障区域识别。最后,搭建昆柳龙仿真模型,进行故障仿真实验验证。结果表明,该方法的故障区域识别率高,且可耐受300Ω的过渡电阻。A fault region identification method based on the convolutional neural network-support vector machine(CNN-SVM)for UHV three-terminal hybrid DC lines is proposed.Firstly,the frequency characteristics of the boundary of DC lines on the north side of Kunming and the north side of Longmen and the boundary of T zone on the north side of Liu are analyzed,and differences of fault characteristics are observed across different fault regions.Then,the empirical wavelet transform is used to extract the fault feature as the input of CNN-SVM and the fault region as the output,and the CNN-SVM model is constructed and trained The fault features obtained from the measurement points are input into the trained CNN-SVM model to identify the fault region.Finally,the simulation model of Kunliulong is built to verify the fault simulation.The results show that the method has high recognition rate of fault region and can withstand 300Ωtransition resistance.

关 键 词：特高压三端混合直流 频率特性 卷积神经网络 支持向量机 故障区域识别 

分 类 号：TM773[电气工程—电力系统及自动化]