采用机器学习的变压器分层故障诊断Multi-level Fault Diagnosis of Power Transformer Based on Machine Learning
王子鉴,秦瑜瑞,李景丽
摘要(Abstract)：
油中溶解气体分析DGA(dissolved gas analysis)是进行变压器故障诊断的重要依据。针对传统变压器故障诊断未能有效利用特征气体与不同故障类型之间的相关性大小，对特征气体进行筛选，以致出现故障诊断结果不准确的问题，文章基于变压器分层故障诊断的方法 ，提出了利用卡方检验在每个故障诊断层中选取最优气体，剔除冗余气体，并采用不同机器学习分类器对所选择的最优气体进行分类的新方案。进行交叉验证后的结果表明，卡方检验能够有效提取特征气体，采用不同分类器进行分层故障诊断的效果优于用单个分类器对小类故障直接诊断。
关键词(KeyWords)： 油浸变压器;溶解气体分析;卡方检验;分层诊断;机器学习
基金项目(Foundation):
作者(Author): 王子鉴,秦瑜瑞,李景丽
DOI: 10.19635/j.cnki.csu-epsa.000849
参考文献(References)：
- [1]荣智海，齐波，李成榕，等(Rong Zhihai,Qi Bo,Li Chengrong,et al).面向变压器油中溶解气体分析的组合DBN诊断方法(Combined DBN diagnosis method for dissolved gas analysis of power transformer oil)[J].电网技术(Power System Technology),2019,43(10):3800-3807.
- [2]叶利振(Ye Lizhen).基于DGA的变压器绝缘油分析与故障诊断在澳门电网中的应用研究(Application of DGA Based Transformer Insulating Oil Analysis and Fault Diagnosis in Macao Network)[D].广州：华南理工大学(Guangzhou:South China University of Technology),2016.
- [3]蒲丽娟，刘念，刘航宇，等(Pu Lijuan,Liu Nian,Liu Hangyu,et al).基于油浸变压器故障数据的IEC三比值统计分析(Statistical analysis of IEC three-ratio method based on fault data of oil-immersed transformers)[J].陕西电力(Shaanxi Electric Power),2015,43(5):41-44,49.
- [4]蔡昆(Cai Kun).基于改良三比值法的变压器油色谱在线监测应用研究(Application Research on-Line Monitoring of Transformer Oil Chromatography Based on Improved Three Ratio Method)[D].广州：华南理工大学(Guangzhou:South China University of Technology),2018.
- [5]Rogers R R.IEEE and IEC codes to interpret incipient faults in transformers,using gas in oil analysis[J].IEEETrans on Electrical Insulation,1978,13(5):349-354.
- [6]朱瑞金，郝东光，胡石峰(Zhu Ruijin,Hao Dongguang,Hu Shifeng).小样本条件下基于卷积孪生网络的变压器故障诊断(Transformer fault diagnosis based on convolutional siamese network with small samples)[J].电力系统及其自动化学报(Proceedings of the CSU-EPSA),2021,33(1):64-69,84.
- [7]曲岳晗，赵洪山，马利波，等(Qiu Yuehan,Zhao Hongshan,Ma Libo,et al).多深度神经网络综合的电力变压器故障识别方法(Multi-depth neural network synthesis method for power transformer fault identification)[J].中国电机工程学报(Proceedings of the CSEE),2021,41(23):8223-8013.
- [8]李春茂，周妺末，刘亚婕，等(Li Chunmao,Zhou Meimo,Liu Yajie,et al).基于邻域粗糙集与多核支持向量机的变压器多级故障诊断(Multi-level fault diagnosis of transformer based on neighborhood rough set and Multiple kernel support vector machine)[J].高电压技术(High Voltage Engineering),2018,44(11):3474-3482.
- [9]张洋(Zhang Yang).支持向量机和相关向量机在变压器油溶解气体分析中的应用研究(Study on the Application of Support Vector Machine and Relevance Vector Machine in the DGA of Transformer Oil)[D].秦皇岛：燕山大学(Qinhuangdao:Yanshan University),2019.
- [10]胡青(Hu Qing).基于电力变压器故障特征气体分层特性的诊断与预测方法研究(Study on Power Transformer Fault Diagnosis and Prediction Based on Layered Characteristics of Fault Feature Gases)[D].重庆：重庆大学(Chongqing:Chongqing University),2010.
- [11]郑方丹，姜久春，陈坤龙，等(Zheng Fangdan,Jiang Jiuchun,Chen Kunlong,et al).基于数据统计特性的GS-SVM电池峰值功率预测模型(Peak power prediction model for batteries based on data statistical characteristic and GS-SVM)[J].电力自动化设备(Electric Power Automation Equipment),2017,37(9):56-61.
- [12]董镝，李恒真，易林，等(Dong Di,Li Hengzhen,Yi Lin,et al).基于形态小波和支持向量机的高压断路器状态监测和故障诊断(Condition monitoring and fault diagnosis of high voltage circuit breaker based on morphology wavelet and support vector machine)[J].电测与仪表(Electrical Measurement&Instrumentation),2017,54(4):92-99.
- [13]尹金良(Yin Jinliang).基于相关向量机的油浸式电力变压器故障诊断方法研究(Study on Oil-immersed Power Transformer Fault Diagnosis Based on Relevance Vector Machine)[D].北京：华北电力大学(Beijing:North China Electric University),2013.
- [14]Paul-Christian B,Jonah G,Vehtari A.Efficient leave-oneout cross-validation for Bayesian non-factorized normal and Student-t models[J].Computational Statistics,2020(prepublish).--------