基于多方法融合和多级支持向量机的电力系统过电压辨识技术研究  

Research on Excessive Voltage Identification Technology Based on Multi-method Fusion and Multi-level Support Vector Machine

作　　者：喻宙[1] 张育善 赵庆杰[2] YU Zhou;ZHANG Yu-shan;ZHAO Qing-jie(Development of Quality Department,Zhengzhou Electric Power College,Zhengzhou450000,China;Office of Academic,Zhengzhou Electric Power College,Zhengzhou450000,China;不详)

机构地区：[1]郑州电力高等专科学校发展质量部,河南郑州450000 [2]郑州电力高等专科学校教务处,河南郑州450000

出　　处：《武汉职业技术学院学报》2021年第4期106-111,共6页Journal of Wuhan Polytechnic

基　　金：2021年度河南省高等学校重点科研项目“多点布局电池储能系统中虚拟同步发电机技术的应用研究”(项目编号:21B470010)。

摘　　要：提出了一种多方法融合和多级支持向量机的电力系统过电压辨识方案,对铁磁谐振、不对称短路及弧光接地等5种过电压进行辨识,通过多尺度时频矩阵奇异值分解、中性点运行方式、时域、频域和小波分析多角度构造出特征量,并将特征量输入多级支持向量机,进行过电压的辨识。对变电站实测的5种过电压信号辨识结果表明:所提取的特征量具有较高区分度,辨识准确率高。This paper proposes a multi-method fusion and multi-stage support vector machine for overvoltage identification protocol,5 kinds of overvoltage such as ferromagnetic resonance,asymmered short circuit,and arc grounding,by multi-scale time frequency matrix odd value decomposition,Neutral point operation,time domain,frequency domain,and wavelet analysis multi-angle construct the feature amount,and input the feature amount input multi-stage support vector machine,and the identification of overvoltage.The identification results of five kinds of overvoltage signals measured in substation show that the extracted characteristic measure has high discrimination and identification accuracy.

关 键 词：过电压 模式辨识 奇异值分解 小波变换 支持向量机 

分 类 号：F276.5[经济管理—企业管理]