| 引用本文: | 贾亚飞,朱永利,高佳程,袁博.基于样本加权FCM聚类的未知类别局部放电信号识别[J].电力自动化设备,2018,(12): |
| JIA Yafei,ZHU Yongli,GAO Jiacheng,YUAN Bo.Recognition of unknown partial discharge signals based on sample-weighted FCM clustering[J].Electric Power Automation Equipment,2018,(12): |
|
| 摘要: |
| 针对电力变压器待识别局部放电信号中可能存在不属于已知类别的未知样本的问题，提出了一种基于样本加权模糊C均值(FCM)聚类的未知类别局部放电信号识别方法。对已知类别的局部放电信号进行FCM聚类，确定各已知类的聚类中心；分别计算已知类别和待识别局部放电信号的样本权值，并根据Otsu准则确定样本权值的自适应阈值；将各待识别局部放电信号的权值与确定的阈值进行比较，判断其是否属于已知类别；采用支持向量机(SVM)对得到的属于已知类别的待识别局部放电信号进行分类，对未知类别样本进行人为分析判断。采用所提方法对实验室条件下的放电信号进行分析，实验结果表明，所提方法可以有效地区分待识别局部放电信号中的未知类别样本。 |
| 关键词: 电力变压器 局部放电 模式识别 未知样本 样本加权 FCM聚类 Otsu准则 支持向量机 |
| DOI：10.16081/j.issn.1006-6047.2018.12.016 |
| 分类号:TM41;TM835 |
| 基金项目:国家自然科学基金资助项目(51677072)；中央高校基本科研业务费专项资金资助项目(2016XS101) |
|
| Recognition of unknown partial discharge signals based on sample-weighted FCM clustering |
|
JIA Yafei1,2, ZHU Yongli1, GAO Jiacheng1, YUAN Bo3
|
|
1.State Key Laboratory of Alternate Electrical Power System with Renewable Energy Sources, North China Electric Power University, Baoding 071003 China;2.Xiongan New Area Electric Power Supply Company, State Grid Hebei Electric Power Supply Co.,Ltd.,Xiongan New Area 071800, China;3.Economic and Technology Research Institute, State Grid Hebei Electric Power Company, Shijiazhuang 050000, China
|
| Abstract: |
| In the power transformers' PD(Partial Discharge) signals to be identified, there may be samples belonging to unknown types, to solve this problem, a recognition method of unknown PD signals based on sample-weighted FCM(Fuzzy C-Means) clustering is proposed. The PD signals of known types are clustered by FCM clustering to determine the clustering centers of known types. The PD signal sample weights of known types and the signals to be identified are calculated respectively, and the adaptive thresholds of sample weights are determined by Otsu criterion. Then, the weights of the PD signals to be identified are compared with the determined thresholds to judge whether they belong to known PD types. The PD signals which belong to known types are classified by SVM(Support Vector Machine) and the PD signals of unknown types are artificially analyzed and judged. The PD signals under laboratory conditions are analyzed by the proposed method and experimental results show that the proposed method can effectively recognize the PD signals with unknown types from the PD signals to be identified. |
| Key words: power transformers partial discharge pattern recognition unknown sample sample-weighted FCM clustering Otsu criterion support vector machines |
