配电网暂时过电压奇异值分解结合支持向量机的识别方法  

IDENTIFICATION FOR DISTRIBUTION NETWORK BASED ON TEMPORAL OVERVOLTAGE SINGULAR VALUE DECOMPOSITION AND SUPPORT VECTOR MACHINE

作　　者：付华[1] 赵天一 Fu Hua;Zhao Tianyi(Faculty of Electrical and Control Engineering, Liaoning Technical University, Huludao 125105, Liaoning, China)

机构地区：[1]辽宁工程技术大学电气与控制工程学院,辽宁葫芦岛125105

出　　处：《计算机应用与软件》2019年第4期230-235,310,共7页Computer Applications and Software

基　　金：国家自然科学基金项目(51274118);辽宁省重点实验室资助项目(LJZS003)

摘　　要：针对配电网暂时过电压的辨识分类问题,提出一种结合时频矩阵奇异值分解和多级支持向量机的配电网暂时过电压辨识方法。利用变分模态分解、Hilbert变换及带通滤波构造配电网暂时过电压零序电压波形的时频矩阵描述其时频特征。通过对时频矩阵进行奇异值分解,提取所获得波形奇异谱的分布参数作为特征向量,结合时域特征量输入多级支持向量机,对配电网暂时过电压进行自动辨识。通过仿真实验和测试,结果表明该识别方法具备训练时间短、识别率高和防干扰能力强的优势,可实现对配电网暂时过电压故障的有效辨识。For the identification and classification of temporary overvoltage in distribution networks, we proposed a method of temporary overvoltage identification for distribution network based on singular value decomposition of timefrequency matrix and multi-level support vector machine. The variational modal decomposition, Hilbert transform and band-pass filtering were adopted to construct the time-frequency matrix of temporary overvoltage zero-sequence voltage waveforms in the distribution network to describe the time-frequency characteristics. Through singular value decomposition of time-frequency matrix, the distribution parameters of waveform singular spectrum were extracted as eigenvectors. Combined with time-domain eigenvalues, multi-level support vector machine was input to identify the temporary overvoltage in distribution network automatically. Through simulation experiments and tests, the results show that the recognition method has the advantages of short training time, high recognition rate and strong anti-interference ability. It can realize the effective identification of temporary overvoltage faults in distribution networks.

关 键 词：配电网故障 暂时过电压 变分模态分解 奇异值分解 多级支持向量机 自动辨识 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术] TM93[电气工程—电力电子与电力传动]