电气工程学报 ›› 2022, Vol. 17 ›› Issue (1): 95-103.doi: 10.11985/2022.01.013

所属专题: 特邀专栏:电力电子化配电网关键设备和运行控制

• 特邀专栏: 电力电子化配电网关键设备和运行控制 • 上一篇    下一篇

扫码分享

基于KPCA与IHHO-LSSVM的电力变压器故障诊断方法研究*

宋立业1(), 范抑伶1(), 王燚增2()   

  1. 1.辽宁工程技术大学电气与控制工程学院 葫芦岛 125105
    2.国网冀北电力有限公司检修分公司 北京 102488
  • 收稿日期:2021-12-06 修回日期:2022-01-20 出版日期:2022-03-25 发布日期:2022-05-06
  • 通讯作者: 范抑伶 E-mail:372492761@qq.com;1147893017@qq.com;946118644@qq.com
  • 作者简介:宋立业,男,1972年生,博士,副教授,硕士研究生导师。主要研究方向为智能电网新技术,电力系统数字化监控技术,工业过程控制与优化等。E-mail: 372492761@qq.com;
    王燚增,男,1993年生,硕士,助理工程师。主要研究方向为状态检测与电气试验。E-mail: 946118644@qq.com
  • 基金资助:
    *辽宁省自然科学基金指导(2019-ZD-0039);辽宁省教育厅科学技术研究创新团队(LT2019007)

Research on Fault Diagnosis Method of Power Transformer Based on KPCA and IHHO-LSSVM

SONG Liye1(), FAN Yiling1(), WANG Yizeng2()   

  1. 1. Faculty of Electrical and Control Engineering, Liaoning Technnical University, Huludao 125105
    2. Maintenance Branch of State Grid Jibei Electric Power Co., Ltd., Beijing 102488
  • Received:2021-12-06 Revised:2022-01-20 Online:2022-03-25 Published:2022-05-06
  • Contact: FAN Yiling E-mail:372492761@qq.com;1147893017@qq.com;946118644@qq.com

摘要:

为提高电力变压器故障识别精确度,提出基于核主成分分析(Kernel principal component analysis,KPCA)与改进哈里斯鹰算法(Improved Harris hawk algorithm,IHHO)优化最小二乘支持向量机(Least square support vector machine,LSSVM)的电力变压器故障诊断方法。首先,利用核主成分分析对变压器原始故障数据进行预处理,去除冗余数据;其次,结合Sigmoid变形函数以及点对称策略改进传统哈里斯鹰算法(HHO),并与HHO和遗传算法(Genetic algorithm,GA)进行性能对比,证明求解精度和网络收敛速度有所提升;最后,采用IHHO对LSSVM的相关超参数进行优化求解,获取KPCA与IHHO-LSSVM相结合的变压器故障诊断模型。结果表明所提模型的诊断精度为95.6%,同其他故障诊断模型相比分别提高了8.9%、16.7%,证明了所提方法能够有效地提升变压器故障诊断性能。

关键词: 变压器, 故障诊断, 核主成分分析, 哈里斯鹰算法, 最小二乘支持向量机

Abstract:

In order to improve the accuracy of power transformer fault identification, a power transformer fault diagnosis method based on kernel principal component analysis(KPCA) and improved Harris hawk algorithm(IHHO) optimized least square support vector machine(LSSVM) is proposed. First, use nuclear principal component analysis to preprocess the original fault data of the transformer to remove redundant data. Secondly, combine the Sigmoid deformation function and the point symmetry strategy to improve the traditional Harris hawk optimization(HHO), and combine with HHO and genetic algorithm(GA). The performance comparison proves that the accuracy of the solution and the speed of network convergence have been improved. Finally, IHHO is used to optimize the related hyperparameters of LSSVM to obtain a transformer fault diagnosis model combining KPCA and IHHO-LSSVM. The results show that the diagnosis accuracy of the proposed model is 95.6%, which is 8.9% and 16.7% higher than other fault diagnosis models, respectively, which proves that the proposed method can effectively improve the performance of transformer fault diagnosis.

Key words: Transformer, fault diagnosis, LDA, HHO, LSSVM

中图分类号: