王子凌,汪科,柴卫健,李业欣,邹国平,安斯光
WANG Ziling,WANG Ke,CHAI Weijian,LI Yexin,ZOU Guoping,AN Siguang

• 1:国网浙江省电力有限公司电力科学研究院
• 2:中国计量大学机电工程学院

摘要(Abstract)：

油中溶解气体在线监测装置对变压器运行状态的准确监测是保证其稳定运行的关键，监测装置若出现故障会导致在线监测数据异常，严重影响在线监测的效果。针对这一问题，提出了一种基于PCA（主成分分析）和Fast-MCD（快速最小协方差行列式）的变压器油中溶解气体异常数据识别方法。首先，利用PCA算法对油中溶解气体时间序列进行降维，有效消除了冗余特征信息。然后，结合稳健统计理论，对降维后的数据进行Fast-MCD稳健分析，实现异常值识别。最后，分析识别出的异常值，判断故障来源。算例研究结果表明，该异常识别方法能够有效识别出油中溶解气体在线监测装置故障，性能优于其他常规方法，准确率高达99.1%。
To detect the operational status of transformers by online monitoring of dissolved gases in oil is critical for ensuring its stable operation. However, faults in monitoring devices can result in abnormal data, severely impacting the effect of online monitoring. To address this issue, a method for identifying abnormal data in dissolved gas monitoring based on principal component analysis(PCA) and fast minimum covariance determinant(Fast-MCD) is proposed. First, the PCA is used to reduce the dimensionality of time-series data for dissolved gas in oil, effectively eliminating redundant features. Next, robust statistical theory is applied to the reduced data using Fast-MCD to identify anomalies. Finally, the identified anomalies are analyzed to determine the source of the fault. Case studies demonstrate that the proposed method effectively identifies faults in online dissolved gas monitoring devices, outperforming conventional methods with an accuracy of 99.1%.

关键词(KeyWords)： 异常检测;油中溶解气体分析;Fast-MCD;主成分分析
anomaly detection;DGA;Fast-MCD;PCA

Abstract:

Keywords:

基金项目(Foundation): 国网浙江省电力有限公司科技项目（B311DS230008）

作者(Author): 王子凌,汪科,柴卫健,李业欣,邹国平,安斯光
WANG Ziling,WANG Ke,CHAI Weijian,LI Yexin,ZOU Guoping,AN Siguang

DOI: 10.19585/j.zjdl.202503011

参考文献(References)：

• [1]郎雪淞，唐红，赵君娇，等.载气发生器式油色谱在线监测装置技术及应用[J].东北电力技术，2019,40(6):4-6.LANG Xuesong,TANG Hong,ZHAO Junjiao,et al.Technology and application of carrier gas generator oil chromatography on-line monitoring device[J]. Northeast Electric Power Technology,2019,40(6):4-6.
• [2]苏鑫.电力变压器油色谱在线监测装置的故障识别[D].福州：福建工程学院，2019.SU Xin.Fault identification of on-line oil chromatographic monitoring device for power transformer[D].Fuzhou:Fujian University of Technology,2019.
• [3]张正东，宋博，甘强.主变在线监测装置缺陷导致轻瓦斯报警事件的分析[J].浙江电力，2023,42(5):120-124.ZHANG Zhengdong,SONG Bo,GAN Qiang.Analysis of a light gas alarm triggered by a fault of an online monitoring device of the main transformer[J]. Zhejiang Electric Power,2023,42(5):120-124.
• [4] FARIA H,COSTA J S,OLIVAS J L M. A review of monitoring methods for predictive maintenance of electric power transformers based on dissolved gas analysis[J].Renewable and Sustainable Energy Reviews,2015,46:201-209.
• [5]荣智海.基于油色谱监测数据的变压器差异化评价及分层诊断方法[D].北京：华北电力大学，2020.RONG Zhihai.Differential evaluation and hierarchical diagnosis method of transformer based on dissolved gas monitoring[D]. Beijing:North China Electric Power University,2020.
• [6]李红雷，何清，钱之银.变压器油中特征气体分析诊断及检测技术[M].北京：中国电力出版社，2020.
• [7] WANI S A,GUPTA D,FAROOQUE M U,et al.Multiple incipient fault classification approach for enhancing the accuracy of dissolved gas analysis(DGA)[J].IET Science,Measurement&Technology,2019,13(7):959-967.
• [8]高树国，王学磊，李庆民，等.基于MCD稳健统计分析的变压器油色谱异常值检测及分布特性[J].高电压技术，2014,40(11):3477-3482.GAO Shuguo,WANG Xuelei,LI Qingmin,et al.Outliers detection and distribution characteristics of the transformer DGA data based on MCD robust statistics[J].High Voltage Engineering,2014,40(11):3477-3482.
• [9]李振杰，卞朝晖，陈学民，等.变压器油色谱在线监测数据有效性评估[J].高压电器，2018,54(6):158-163.LI Zhenjie,BIAN Zhaohui,CHEN Xuemin,et al.Validity evaluation of online monitored chromatographic data of transformer oil[J].High Voltage Apparatus,2018,54(6):158-163.
• [10]荣智海，齐波，张鹏，等.基于核主成分分析的油色谱在线监测装置异常状态快速辨识[J].高电压技术，2019,45(10):3308-3316.RONG Zhihai,QI Bo,ZHANG Peng,et al.Fast identification of abnormal state based on kernel principal component analysis for oil chromatography analysis on-line monitoring device[J].High Voltage Engineering,2019,45(10):3308-3316.
• [11]黄云程，高阿娜，王翌，等.变压器油色谱在线监测数据甄别与清洗技术[J].电力科学与工程，2019,35(6):37-43.HUANG Yuncheng,GAO Ana,WANG Yi,et al.Screening and cleaning technology of transformer oil chromatographic on-line monitoring data[J].Electric Power Science and Engineering,2019,35(6):37-43.
• [12]刘云鹏，王权，许自强，等.基于多层架构的油中溶解气体数据清洗与异常识别方法研究[J].华北电力大学学报（自然科学版），2022,49(1):81-89.LIU Yunpeng,WANG Quan,XU Ziqiang,et al.Research on data cleaning and abnormal recognition method of dissolved gas in oil based on multi-layer architecture[J].Journal of North China Electric Power University(Natural Science Edition),2022,49(1):81-89.
• [13]沈晓东.基于小波模极大值变压器油色谱在线异常数据识别[J].新型工业化，2018,8(6):26-31.SHEN Xiaodong. Anomaly data identification based on wavelet modulus maximum transformer oil chromatography[J].The Journal of New Industrialization,2018,8(6):26-31.
• [14]时海涛，苏智剑.齿轮测量中异常值的判定及剔除系统设计[J].机械制造与自动化，2020,49(3):53-56.SHI Haitao,SU Zhijian.Design of judgment and elimination system for outliers in gear measurement[J].Machine Building&Automation,2020,49(3):53-56.
• [15]周建平，李聪，万书亭，等.基于优化型SVM的高压断路器故障诊断方法研究[J].浙江电力，2019,38(3):14-19.ZHOU Jianping,LI Cong,WAN Shuting,et al.Research on fault diagnosis of high voltage circuit breaker based on optimized SVM[J].Zhejiang Electric Power,2019,38(3):14-19.
• [16] PENG C,CHEN Y Y,KANG Z,et al.Robust principal component analysis:a factorization-based approach with linear complexity[J]. Information Sciences,2020,513:581-599.
• [17] ZHAO B T,DONG X,GUO Y C,et al.PCA dimensionality reduction method for image classification[J].Neural Processing Letters,2022,54(1):347-368.
• [18]刘伟鹏，张国钢，刘亚魁，等.基于主成分分析和支持向量机的高压断路器机械状态识别方法[J].高压电器，2020,56(9):267-272.LIU Weipeng,ZHANG Guogang,LIU Yakui,et al.Mechanical status identification of high voltage circuit breakers based on principal component analysis and support vector machines[J].High Voltage Apparatus,2020,56(9):267-272.
• [19] LI X,DENG S R,LI L F,et al.Outlier detection based on robust mahalanobis distance and its application[J]. Open Journal of Statistics,2019,9(1):15-26.
• [20] ROUSSEEUW P J,VAN DRIESSEN K. A fast algorithm for the minimum covariance determinant estimator[J].Technometrics,1999,41(3):212.
• [21]张倩茹.充分降维问题中基于MCD方法的稳健降维[D].昆明：云南财经大学，2021.ZHANG Qianru. Robust dimension reduction based on MCD method in sufficient dimension reduction[D].Kunming:Yunnan University of Finance and Economics,2021.
• [22]杨雅琴，徐鹏，吴细水.基于Fast-MCD的自适应建模探索轨道不平顺劣化[J].清华大学学报（自然科学版），2022,62(3):516-522.YANG Yaqin,XU Peng,WU Xishui.Adaptive modeling method based on the Fast-MCD to analyze railway track irregularity deterioration[J]. Journal of Tsinghua University(Science and Technology),2022,62(3):516-522.
• [23]李诗勇，薛静，吴冕之，等.基于粒子群优化的改进加权支持向量回归的变压器顶层油温预测[J].高压电器，2021,57(12):103-109.LI Shiyong,XUE Jing,WU Mianzhi,et al. Prediction of transformer top-oil temperature with the improved weighted support vector regression based on particle swarm optimization[J].High Voltage Apparatus,2021,57(12):103-109.
• [24]盛乐，李学锋，周波，等.油色谱在线监测装置运维情况分析[J].国网技术学院学报，2020,23(3):1-6.SHENG Le,LI Xuefeng,ZHOU Bo,et al.Analysis of operation and maintenance of online monitoring devices for oil chromatography[J].Journal of State Grid Technology College,2020,23(3):1-6.
• [25]童超，童军心，周梦垚，等.油色谱在线监测装置异常情况及特征分析[J].江西电力，2022,46(2):31-34.
• [26]刘云鹏，许自强，和家慧，等.基于条件式Wasserstein生成对抗网络的电力变压器故障样本增强技术[J].电网技术，2020,44(4):1505-1513.LIU Yunpeng,XU Ziqiang,HE Jiahui,et al.Data augmentation method for power transformer fault diagnosis based on conditional Wasserstein generative adversarial network[J].Power System Technology,2020,44(4):1505-1513.
• [27]唐志国，曹智，何宁辉.卷积神经网络迁移学习在局部放电类型诊断中的应用[J].高压电器，2022,58(4):158-164.TANG Zhiguo,CAO Zhi,HE Ninghui.Application of convolutional neural network transfer learning in partial discharge type diagnosis[J].High Voltage Apparatus,2022,58(4):158-164.