Transformer fault diagnosis strategy considering parameter optimization of oversampler and classifier

被引：0
|
作者：
Li L. [1 ]
Wang T. [2 ]
He J. [1 ]
Niu J. [1 ]
Liang Y. [1 ]
Miao S. [2 ]
机构：
[1] Electric Power Research Institute, State Grid Ningxia Electric Power Co.，Ltd., Yinchuan
[2] School of Electrical and Electronic Engineering, Huazhong University of Science and Technology, Wuhan
来源：
Dianli Zidonghua Shebei/Electric Power Automation Equipment | 2023年 / 43卷 / 01期
关键词：
fault diagnosis; HS-MHHO algo⁃ rithm; oversample; power transformers; SMOTE-NND; unbalanced samples;
D O I：
10.16081/j.epae.202206011
中图分类号：
学科分类号：
摘要：
The imbalance of transformer fault samples makes the accuracy of fault diagnosis and classifica⁃ tion low，and it is easy to weaken the classification effect of a few types of fault samples. Therefore，the oversampling method is used to realize the equalization of fault samples，and a transformer fault diagnosis strategy considering the parameter optimization of oversampler and classifier is proposed. Firstly，the overall structure of transformer fault diagnosis model is built，and the implementation process of fault diagnosis is described. On this basis，the algorithm implementation of three main links of oversampler，classifier and parameter optimizer in the diagnosis model is proposed. For the oversampler，an improved synthetic minority oversampling technique based on nearest neighbor distribution（SMOTE-NND） algorithm is proposed to realize the equalization of fault samples. For the classifier，the hierarchical directed acyclic graph support vector machine（HDAG-SVM） algorithm is used to realize the multi-label classification of fault samples. For the parameter optimizer，a double-layer parameter optimization method is proposed. The upper layer uses the hierarchical search（HS） algorithm to optimize the oversampling ratio，and the lower layer uses modified harris hawks optimization（MHHO） algorithm to optimize the parameters of support vector machine. Finally，an example is given to analyze the proposed strategy. The results show that the proposed strategy can syn⁃ thesize a few fault samples with higher quality and realize the accurate classification of fault samples. © 2023 Electric Power Automation Equipment Press. All rights reserved.
引用
收藏
页码：209 / 217
页数：8
相关论文
共 23 条
  • [1] (2015)
  • [2] Mineral oil-impregnated electrical equipment in serviceguide to the interpretation of dissolved and free gases analysis:IEC 60599-2007, (2007)
  • [3] DUVAL M., A review of faults detectable by gas-in-oil analysis in transformers[J], IEEE Electrical Insulation Magazine, 18, 3, pp. 8-17, (2002)
  • [4] DUKARM J., Improving the reliability of transformer gas-in-oil diagnosis［J］, IEEE Electrical Insulation Magazine, 21, 4, pp. 21-27, (2005)
  • [5] ZHAO Lingling, WANG Qunjing, CHEN Quan, Et al., Fault diagnosis of transformer based on BP neural network optimized by IBBOA［J］, Advanced Technology of Electrical Engineering and Energy, 40, 9, pp. 39-46, (2021)
  • [6] ZHAO Wenqing, YAN Hai, ZHOU Zhendong, Et al., Fault diag⁃ nosis of transformer based on residual BP neural network［J］, Electric Power Automation Equipment, 40, 2, pp. 143-148, (2020)
  • [7] YUAN Jinsha, ZHANG Liwei, WANG Yu, Et al., Study of transformers fault diagnosis based on extreme learning machine ［J］, Electrical Measurement & Instrumentation, 50, 12, pp. 21-26, (2013)
  • [8] WU Jiekang, QIN Weimei, LIANG Haohao, Et al., Transformer fault identification method based on self-adaptive extreme learning machine［J］, Electric Power Automation Equipment, 39, 10, pp. 181-186, (2019)
  • [9] HUANG Xinbo, MA Yutao, ZHU Yongcan, Transformer fault diagnosis method based on information fusion and M-RVM［J］, Electric Power Automation Equipment, 40, 12, pp. 218-225, (2020)
  • [10] YIN Jinliang, LIU Lingling, CS-RVM and its application in fault diagnosis of power transformers［J］, Electric Power Auto⁃ mation Equipment, 34, 5, pp. 111-115, (2014)
123