综合智慧能源 ›› 2023, Vol. 45 ›› Issue (3): 9-16.doi: 10.3969/j.issn.2097-0706.2023.03.002

• 优化运行与控制 • 上一篇    下一篇

基于PCA与SSA-LightGBM的油浸式变压器故障诊断方法

肖宏磊1(), 留毅2, 夏红军1, 缪宇峰2, 俞啸玲1, 杨海琦3,*()   

  1. 1.国网浙江省电力有限公司杭州市余杭区供电公司,杭州 311100
    2.杭州电力设备制造有限公司余杭群力成气制造分公司,杭州 311100
    3.东北电力大学 机械工程学院,吉林 吉林 132012
  • 收稿日期:2022-06-10 修回日期:2022-10-10 出版日期:2023-03-25 发布日期:2023-03-30
  • 通讯作者: 杨海琦(1997),女,在读硕士研究生,从事输配电设备智能化、机电设备故障诊断的研究,xhaiqi0526@163.com
  • 作者简介:肖宏磊(1985),男,高级经济师,从事电力系统工程和自动化技术研究,18244891@qq.com
  • 基金资助:
    吉林省科技发展计划项目(2021050959RQ)

Oil-immersed transformer fault diagnosis method based on PCA and SSA-LightGBM

XIAO Honglei1(), LIU Yi2, XIA Hongjun1, MIAO Yufeng2, YU Xiaoling1, YANG Haiqi3,*()   

  1. 1. Hangzhou Yuhang Power Supply Company,State Grid Zhejiang Power Company Limited,Hangzhou 311100,China
    2. Yuhang Qunli Complete Sets Electricity Manufacturing Branch of Hangzhou Electric Power Equipment Manufacturing Company Limited,Hangzhou 311100,China
    3. School of Mechanical Engineering,Northeastern Electric Power University,Jilin 132012,China
  • Received:2022-06-10 Revised:2022-10-10 Online:2023-03-25 Published:2023-03-30
  • Supported by:
    Science and Technology Development Plan Project of Jilin Province(2021050959RQ)

摘要:

针对现阶段油浸式变压器故障复杂性导致诊断精度不高的问题,提出一种基于主元分析(PCA)与麻雀搜索算法-轻量级梯度提升机(SSA-LightGBM)的油浸式变压器故障诊断方法。采集油中溶解气体数据,结合无编码比值方法构建17维故障特征矩阵,并对特征矩阵进行标准化处理得到联合特征。利用主元分析法进行特征融合,消除变量之间的信息冗余,构造融合特征。构建基于SSA-LightGBM变压器诊断模型,并采用十折交叉验证法验证该模型的分类能力。试验结果表明:提出的模型平均故障诊断精度为93.6%,与GA-LightGBM和GWO-LightGBM故障诊断模型相比,诊断精度分别提高了8.1和5.7百分点,验证了该方法能够有效提高油浸式变压器的故障诊断性能。

关键词: 故障诊断, 变压器, 主元分析, 麻雀搜索算法, 轻量级梯度提升机, 麻雀搜索算法

Abstract:

In view of low diagnosis accuracy on oil-immersed transformer faults caused by the fault complexity at present,an oil-immersed transformer fault diagnosis method based on PCA and SSA-LightGBM is proposed.The data of dissolved gas in oil are collected,and a 17-dimensional fault feature matrix is constructed with codeless ratio method.The joint feature is obtained in the normalized matrix.Feature fusion is performed by using principal component analysis which can eliminate the redundant information between variables,and construct fusion features.The SSA-LightGBM transformer diagnosis model is constructed,and the ten-fold cross-validation method is used to verify the classification ability of the model.The experimental results show that the average fault diagnosis accuracy of the proposed model is 93.6%,which is 8.1 and 5.7 percentage points higher than that of GA-LightGBM and GWO-LightGBM fault diagnosis models,respectively, verifying that this method can effectively improve the fault diagnosis performance for transformers.

Key words: fault diagnosis, transformer, principal component analysis, sparrow search algorithm, LightGBM, sparrow search algorithm

中图分类号: