Article(id=1199644781044400370, tenantId=1146029695717560320, journalId=1149653034449285133, issueId=1199644777219191267, articleNumber=null, orderNo=null, doi=10.16790/j.cnki.1009-9239.im.2023.12.006, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=research-article, receivedDate=1676908800000, receivedDateStr=2023-02-21, revisedDate=1680537600000, revisedDateStr=2023-04-04, acceptedDate=null, acceptedDateStr=null, onlineDate=1763947204123, onlineDateStr=2025-11-24, pubDate=1703001600000, pubDateStr=2023-12-20, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1763947204123, onlineIssueDateStr=2025-11-24, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1763947204123, creator=13701087609, updateTime=1763947204123, updator=13701087609, issue=Issue{id=1199644777219191267, tenantId=1146029695717560320, journalId=1149653034449285133, year='2023', volume='56', issue='12', pageStart='1', pageEnd='135', issueExtLink='null', onlineDate='null', pubDate='null', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=null, createTime=1763947203211, creator=13701087609, updateTime=1766563728010, updator=13701087609, preIssue=null, nextIssue=null, ext={EN=IssueExt(id=1210619277716099586, tenantId=1146029695717560320, journalId=1149653034449285133, issueId=1199644777219191267, language=EN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=), CN=IssueExt(id=1210619277716099587, tenantId=1146029695717560320, journalId=1149653034449285133, issueId=1199644777219191267, language=CN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=)}, issueFiles=null}, startPage=43, endPage=53, ext={EN=ArticleExt(id=1199644781518356728, articleId=1199644781044400370, tenantId=1146029695717560320, journalId=1149653034449285133, language=EN, title=Defect recognition method of oil-paper insulation based on information fusion of PRPD spectrum and dissolved gas data, columnId=1199644778280350182, journalTitle=Insulating Materials, columnName=Advanced Electrical Materials for Large Capacity Offshore Wind Power Transmission, runingTitle=null, highlight=null, articleAbstract=

The transformer fault diagnosis technique based on a single detection method is difficult to identify the same type of defects of oil-paper insulation in detail, which cannot meet the requirements of power system on equipment operation reliability under the background of rapid development of deep offshore wind power. Therefore, an oil-paper insulation defect identification method based on information fusion of phase-resolved partial discharge (PRPD) spectrum and dissolved gas analysis (DGA) data was proposed. Six kinds of electrode models were designed and made to simulate the typical defects of surface discharge in transformers with different electric field inhomogeneity coefficients, and PRPD and DGA data were collected. Then convolutional neural network (CNN) and back propagation neural network (BPNN) were adopted to recognize the patterns of PRPD spectrum and DGA feature vector of six kinds of defects, respectively. Finally, the CNN-BPNN information fusion model based on D-S evidence theory was proposed to realize joint diagnosis based on PRPD spectrum and DGA data. The results show that the CNN-BPNN model based on the D-S evidence theory can effectively correct the wrong output of the single criterion model and reduce the uncertainty of the classification results. When the input dimensions of PRPD spectrum are 8×8, 16×16, and 32×32, the recognition accuracy of the model integrated with the DGA feature vector is 93.21%, 97.53%, and 99.17%, respectively, which is 4.81%, 2.78%, and 0.84% higher than that of PRPD single criterion model. The CNN-BPNN model can effectively integrate the electrical physical information and chemical product information of partial discharge, which not only improves the accuracy of defect identification, but also enhances the confidence of the output results, and reduces the data storage requirements, providing accurate, reliable, and lightweight defect identification methods for intelligent operation and maintenance of transformers.

, correspAuthors=null, authorNote=null, correspAuthorsNote=null, copyrightStatement=null, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=null, magXml=null, pdfUrl=null, pdf=null, pdfFileSize=null, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=null, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=null, mapNumber=null, authorCompany=null, fund=null, authors=null, authorsList=Yuanxiang ZHOU, Yongyin LI, Jianning CHEN, Zheng BAI), CN=ArticleExt(id=1199644787423936994, articleId=1199644781044400370, tenantId=1146029695717560320, journalId=1149653034449285133, language=CN, title=基于局部放电相位图谱和油中溶解气体信息融合的油纸绝缘缺陷识别方法, columnId=1199644778431345128, journalTitle=绝缘材料, columnName=面向大容量海上风电送出的先进电工材料, runingTitle=null, highlight=null, articleAbstract=

基于单一检测手段的变压器故障诊断方法难以对油纸绝缘的同一类型缺陷进行细化识别,无法满足深远海风电快速发展背景下电力系统对设备运行可靠性的要求。因此,本文提出了一种基于局部放电相位(PRPD)图谱和油中溶解气体分析(DGA)信息融合的油纸绝缘缺陷识别方法,设计并制作了6种电极模型,模拟变压器中不同电场不均匀系数的沿面放电典型缺陷,并采集其PRPD及DGA数据;分别采用卷积神经网络(CNN)和反向传播神经网络(BPNN)对6类缺陷的PRPD图谱和DGA特征向量进行模式识别;提出基于D-S证据理论的CNN-BPNN信息融合模型,实现基于PRPD图谱与DGA数据的联合诊断。结果表明:基于D-S证据理论的CNN-BPNN模型可有效纠正单一判据模型的错误输出,并降低分类结果的不确定度,当PRPD图谱输入维度为8×8、16×16、32×32时,融入DGA特征向量的模型识别准确率分别为93.21%、97.53%、99.17%,较PRPD图谱单一判据模型的识别准确率分别提升了4.81%、2.78%、0.84%,该模型可有效融合局部放电的电气物理信息和化学产物信息,既提高了缺陷识别准确率,又增强了输出结果的置信程度,且降低了数据存储要求,可为变压器智能运维提供精确、可靠、轻量的缺陷识别方法。

, correspAuthors=null, authorNote=null, correspAuthorsNote=null, copyrightStatement=null, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=rEru2TvKuZTBCwLhiKzdlg==, magXml=3kU8FAnmXjYy+IBdfFzmuA==, pdfUrl=null, pdf=vzXOYgvJPBIb2tpfkbIxnA==, pdfFileSize=7052025, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=UWoToh3BDnQoQy79IeASVw==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=CjVNtjQn2VGwFWvLxN/sig==, mapNumber=null, authorCompany=null, fund=null, authors=

周远翔(1966-),男(汉族),福建莆田人,教授,主要从事高电压与绝缘技术的研究;

李永印(1996-),男(汉族),河南信阳人,硕士生,主要从事油纸绝缘局部放电与人工智能应用研究。

, authorsList=周远翔, 李永印, 陈健宁, 白正)}, authors=[Author(id=1211311075539162101, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1211311076805840898, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, authorId=1211311075539162101, language=EN, stringName=Yuanxiang ZHOU, firstName=Yuanxiang, middleName=null, lastName=ZHOU, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1The Wind Solar Storage Division of State Key Laboratory of Control and Simulation of Power System and Generation Equipment, School of Electrical Engineering, Xinjiang University, Urumqi 830047, China
2State Key Laboratory of Control and Simulation of Power Systems and Generation Equipment, Department of Electrical Engineering, Tsinghua University, Beijing 100084, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1211311076919087112, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, authorId=1211311075539162101, language=CN, stringName=周远翔, firstName=远翔, middleName=null, lastName=周, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1新疆大学 电气工程学院 电力系统及大型发电设备安全控制和仿真国家重点实验室风光储分室, 新疆 乌鲁木齐 830047
2清华大学 电机工程与应用电子技术系 电力系统及发电设备控制和仿真国家重点实验室,北京 100084, bio={"content":"

周远翔(1966-),男(汉族),福建莆田人,教授,主要从事高电压与绝缘技术的研究;

"}, bioImg=null, bioContent=

周远翔(1966-),男(汉族),福建莆田人,教授,主要从事高电压与绝缘技术的研究;

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1211311075337835490, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, xref=1, ext=[AuthorCompanyExt(id=1211311075342029795, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, companyId=1211311075337835490, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1The Wind Solar Storage Division of State Key Laboratory of Control and Simulation of Power System and Generation Equipment, School of Electrical Engineering, Xinjiang University, Urumqi 830047, China), AuthorCompanyExt(id=1211311075350418404, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, companyId=1211311075337835490, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1新疆大学 电气工程学院 电力系统及大型发电设备安全控制和仿真国家重点实验室风光储分室, 新疆 乌鲁木齐 830047)]), AuthorCompany(id=1211311075446887402, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, xref=2, ext=[AuthorCompanyExt(id=1211311075451081706, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, companyId=1211311075446887402, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2State Key Laboratory of Control and Simulation of Power Systems and Generation Equipment, Department of Electrical Engineering, Tsinghua University, Beijing 100084, China), AuthorCompanyExt(id=1211311075459470315, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, companyId=1211311075446887402, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2清华大学 电机工程与应用电子技术系 电力系统及发电设备控制和仿真国家重点实验室,北京 100084)])]), Author(id=1211311076990390287, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1211311077091053593, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, authorId=1211311076990390287, language=EN, stringName=Yongyin LI, firstName=Yongyin, middleName=null, lastName=LI, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1The Wind Solar Storage Division of State Key Laboratory of Control and Simulation of Power System and Generation Equipment, School of Electrical Engineering, Xinjiang University, Urumqi 830047, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1211311077191716896, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, authorId=1211311076990390287, language=CN, stringName=李永印, firstName=永印, middleName=null, lastName=李, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1新疆大学 电气工程学院 电力系统及大型发电设备安全控制和仿真国家重点实验室风光储分室, 新疆 乌鲁木齐 830047, bio={"content":"

李永印(1996-),男(汉族),河南信阳人,硕士生,主要从事油纸绝缘局部放电与人工智能应用研究。

"}, bioImg=null, bioContent=

李永印(1996-),男(汉族),河南信阳人,硕士生,主要从事油纸绝缘局部放电与人工智能应用研究。

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1211311075337835490, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, xref=1, ext=[AuthorCompanyExt(id=1211311075342029795, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, companyId=1211311075337835490, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1The Wind Solar Storage Division of State Key Laboratory of Control and Simulation of Power System and Generation Equipment, School of Electrical Engineering, Xinjiang University, Urumqi 830047, China), AuthorCompanyExt(id=1211311075350418404, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, companyId=1211311075337835490, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1新疆大学 电气工程学院 电力系统及大型发电设备安全控制和仿真国家重点实验室风光储分室, 新疆 乌鲁木齐 830047)])]), Author(id=1211311077267214373, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1211311077372071984, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, authorId=1211311077267214373, language=EN, stringName=Jianning CHEN, firstName=Jianning, middleName=null, lastName=CHEN, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2State Key Laboratory of Control and Simulation of Power Systems and Generation Equipment, Department of Electrical Engineering, Tsinghua University, Beijing 100084, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1211311077447569460, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, authorId=1211311077267214373, language=CN, stringName=陈健宁, firstName=健宁, middleName=null, lastName=陈, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2清华大学 电机工程与应用电子技术系 电力系统及发电设备控制和仿真国家重点实验室,北京 100084, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1211311075446887402, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, xref=2, ext=[AuthorCompanyExt(id=1211311075451081706, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, companyId=1211311075446887402, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2State Key Laboratory of Control and Simulation of Power Systems and Generation Equipment, Department of Electrical Engineering, Tsinghua University, Beijing 100084, China), AuthorCompanyExt(id=1211311075459470315, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, companyId=1211311075446887402, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2清华大学 电机工程与应用电子技术系 电力系统及发电设备控制和仿真国家重点实验室,北京 100084)])]), Author(id=1211311077518872634, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, orderNo=3, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1211311077623730242, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, authorId=1211311077518872634, language=EN, stringName=Zheng BAI, firstName=Zheng, middleName=null, lastName=BAI, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2State Key Laboratory of Control and Simulation of Power Systems and Generation Equipment, Department of Electrical Engineering, Tsinghua University, Beijing 100084, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1211311077816668231, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, authorId=1211311077518872634, language=CN, stringName=白正, firstName=正, middleName=null, lastName=白, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2清华大学 电机工程与应用电子技术系 电力系统及发电设备控制和仿真国家重点实验室,北京 100084, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1211311075446887402, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, xref=2, ext=[AuthorCompanyExt(id=1211311075451081706, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, companyId=1211311075446887402, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2State Key Laboratory of Control and Simulation of Power Systems and Generation Equipment, Department of Electrical Engineering, Tsinghua University, Beijing 100084, China), AuthorCompanyExt(id=1211311075459470315, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, companyId=1211311075446887402, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2清华大学 电机工程与应用电子技术系 电力系统及发电设备控制和仿真国家重点实验室,北京 100084)])])], keywords=[Keyword(id=1211311078030577754, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=EN, orderNo=1, keyword=surface discharge), Keyword(id=1211311078148018267, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=EN, orderNo=2, keyword=PRPD spectrum), Keyword(id=1211311078248681566, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=EN, orderNo=3, keyword=dissolved gas in oil), Keyword(id=1211311078408065125, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=EN, orderNo=4, keyword=neural network), Keyword(id=1211311078559060076, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=EN, orderNo=5, keyword=D-S evidence theory), Keyword(id=1211311078659723380, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=CN, orderNo=1, keyword=沿面放电), Keyword(id=1211311078751998072, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=CN, orderNo=2, keyword=PRPD图谱), Keyword(id=1211311078848467069, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=CN, orderNo=3, keyword=油中溶解气体), Keyword(id=1211311078936547455, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=CN, orderNo=4, keyword=神经网络), Keyword(id=1211311079049793669, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=CN, orderNo=5, keyword=D-S证据理论)], refs=[Reference(id=1211311083793551664, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2015, volume=68, issue=null, pageStart=164, pageEnd=181, url=null, language=null, rfNumber=[1], rfOrder=0, authorNames=RAYMONDW J K, ILLIASH A, BAKARA H A, journalName=Measurement, refType=null, unstructuredReference=RAYMONDW J K, ILLIASH A, BAKARA H A, et al. Partial discharge classifications: Review of recent progress[J]. Measurement,2015,68:164-181., articleTitle=Partial discharge classifications: Review of recent progress, refAbstract=null), Reference(id=1211311083869049138, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2021, volume=54, issue=6, pageStart=107, pageEnd=115, url=null, language=null, rfNumber=[2], rfOrder=1, authorNames=孙长海, 李天明, 陈百通, journalName=绝缘材料, refType=null, unstructuredReference=孙长海,李天明,陈百通,等.基于概率神经网络的油纸绝缘老化诊断技术研究[J].绝缘材料,2021,54(6):107-115., articleTitle=基于概率神经网络的油纸绝缘老化诊断技术研究, refAbstract=null), Reference(id=1211311083957129526, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2022, volume=48, issue=9, pageStart=3384, pageEnd=3393, url=null, language=null, rfNumber=[3], rfOrder=2, authorNames=李岩, 冯俊杰, 卢毓欣, journalName=高电压技术, refType=null, unstructuredReference=李岩,冯俊杰,卢毓欣,等.大容量远海风电柔性直流送出关键技术与展望[J].高电压技术,2022,48(9):3384-3393., articleTitle=大容量远海风电柔性直流送出关键技术与展望, refAbstract=null), Reference(id=1211311084028432698, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2021, volume=54, issue=7, pageStart=10, pageEnd=24, url=null, language=null, rfNumber=[4], rfOrder=3, authorNames=范路, 陆云才, 陶风波, journalName=绝缘材料, refType=null, unstructuredReference=范路,陆云才,陶风波,等.人工智能在局部放电检测中的应用(二):模式识别与状态评估[J].绝缘材料,2021,54(7):10-24., articleTitle=人工智能在局部放电检测中的应用(二):模式识别与状态评估, refAbstract=null), Reference(id=1211311084103930174, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2018, volume=12, issue=8, pageStart=1031, pageEnd=1038, url=null, language=null, rfNumber=[5], rfOrder=4, authorNames=BASHARANV, SILUVAIRAJW I M, VELAYUTHAMM R, journalName=IET Science, refType=null, unstructuredReference=BASHARANV, SILUVAIRAJW I M, VELAYUTHAMM R. Recognition of multiple partial discharge patterns by multi-class support vector machine using fractal image processing technique[J]. IET Science, Measurement & Technology,2018,12(8):1031-1038., articleTitle=Recognition of multiple partial discharge patterns by multi-class support vector machine using fractal image processing technique, refAbstract=null), Reference(id=1211311084200399169, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2022, volume=48, issue=3, pageStart=1113, pageEnd=1123, url=null, language=null, rfNumber=[6], rfOrder=5, authorNames=许辰航, 陈继明, 刘伟楠, journalName=高电压技术, refType=null, unstructuredReference=许辰航,陈继明,刘伟楠,等.基于深度残差网络的GIS局部放电PRPD谱图模式识别[J].高电压技术,2022,48(3):1113-1123., articleTitle=基于深度残差网络的GIS局部放电PRPD谱图模式识别, refAbstract=null), Reference(id=1211311084317839688, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2021, volume=40, issue=3, pageStart=72, pageEnd=80, url=null, language=null, rfNumber=[7], rfOrder=6, authorNames=张聪聪, 王刚, 高栋, journalName=电工电能新技术, refType=null, unstructuredReference=张聪聪,王刚,高栋,等.基于卷积网络的GIS局部放电缺陷诊断方法与应用[J].电工电能新技术,2021,40(3):72-80., articleTitle=基于卷积网络的GIS局部放电缺陷诊断方法与应用, refAbstract=null), Reference(id=1211311084405920074, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2022, volume=58, issue=4, pageStart=158, pageEnd=164, url=null, language=null, rfNumber=[8], rfOrder=7, authorNames=唐志国, 曹智, 何宁辉, journalName=高压电器, refType=null, unstructuredReference=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., articleTitle=卷积神经网络迁移学习在局部放电类型诊断中的应用, refAbstract=null), Reference(id=1211311084498194764, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2020, volume=56, issue=9, pageStart=20, pageEnd=25,32, url=null, language=null, rfNumber=[9], rfOrder=8, authorNames=杨为, 朱太云, 张国宝, journalName=高压电器, refType=null, unstructuredReference=杨为,朱太云,张国宝,等.电力物联网下基于卷积神经网络和迁移学习的GIS局部放电模式识别分类方法研究[J].高压电器,2020,56(9):20-25,32., articleTitle=电力物联网下基于卷积神经网络和迁移学习的GIS局部放电模式识别分类方法研究, refAbstract=null), Reference(id=1211311085735514446, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2020, volume=46, issue=8, pageStart=2897, pageEnd=2905, url=null, language=null, rfNumber=[10], rfOrder=9, authorNames=王涤, 马爱军, 归宇, journalName=高电压技术, refType=null, unstructuredReference=王涤,马爱军,归宇,等.基于P-CNN的局部放电绝缘故障融合诊断[J].高电压技术,2020,46(8):2897-2905., articleTitle=基于P-CNN的局部放电绝缘故障融合诊断, refAbstract=null), Reference(id=1211311085890703699, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2015, volume=41, issue=3, pageStart=947, pageEnd=955, url=null, language=null, rfNumber=[11], rfOrder=10, authorNames=黄亮, 唐炬, 凌超, journalName=高电压技术, refType=null, unstructuredReference=黄亮,唐炬,凌超,等.基于多特征信息融合技术的局部放电模式识别研究[J].高电压技术,2015,41(3):947-955., articleTitle=基于多特征信息融合技术的局部放电模式识别研究, refAbstract=null), Reference(id=1211311086062670164, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2022, volume=48, issue=12, pageStart=4925, pageEnd=4932, url=null, language=null, rfNumber=[12], rfOrder=11, authorNames=艾嘉伟, 牛海清, 陈泽铭, journalName=高电压技术, refType=null, unstructuredReference=艾嘉伟,牛海清,陈泽铭,等.基于神经网络的D-S多信息融合GIL局部放电联合识别方法[J].高电压技术,2022,48(12):4925-4932., articleTitle=基于神经网络的D-S多信息融合GIL局部放电联合识别方法, refAbstract=null), Reference(id=1211311086142361942, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2022, volume=55, issue=11, pageStart=109, pageEnd=117, url=null, language=null, rfNumber=[13], rfOrder=12, authorNames=耿伊雯, 芮逸凡, 范路, journalName=绝缘材料, refType=null, unstructuredReference=耿伊雯,芮逸凡,范路,等.基于D-S证据多源信息融合与固态光电倍增-UHF联合检测的GIS局部放电模式识别[J].绝缘材料,2022,55(11):109-117., articleTitle=基于D-S证据多源信息融合与固态光电倍增-UHF联合检测的GIS局部放电模式识别, refAbstract=null), Reference(id=1211311086230442329, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2022, volume=41, issue=5, pageStart=172, pageEnd=179, url=null, language=null, rfNumber=[14], rfOrder=13, authorNames=王磊, 张磊, 牛荣泽, journalName=电力工程技术, refType=null, unstructuredReference=王磊,张磊,牛荣泽,等.基于CNN+D-S证据理论的多维信息源局部放电故障识别[J].电力工程技术,2022,41(5):172-179., articleTitle=基于CNN+D-S证据理论的多维信息源局部放电故障识别, refAbstract=null), Reference(id=1211311086297551195, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2017, volume=50, issue=12, pageStart=43, pageEnd=47, url=null, language=null, rfNumber=[15], rfOrder=14, authorNames=魏振, 孙昭昌, 史蕾玚, journalName=绝缘材料, refType=null, unstructuredReference=魏振,孙昭昌,史蕾玚,等.复合电压比例对油纸绝缘沿面放电的影响[J].绝缘材料,2017,50(12):43-47., articleTitle=复合电压比例对油纸绝缘沿面放电的影响, refAbstract=null), Reference(id=1211311086364660061, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2019, volume=26, issue=6, pageStart=2019, pageEnd=2026, url=null, language=null, rfNumber=[16], rfOrder=15, authorNames=REFFASA, BEROUALA, MOULAIH, journalName=IEEE Transactions on Dielectrics and Electrical Insulation, refType=null, unstructuredReference=REFFASA, BEROUALA, MOULAIH. Comparison of creeping discharges propagating over pressboard immersed in olive oil, mineral oil and other natural and synthetic ester liquids under DC voltage[J]. IEEE Transactions on Dielectrics and Electrical Insulation,2019,26(6):2019-2026., articleTitle=Comparison of creeping discharges propagating over pressboard immersed in olive oil, mineral oil and other natural and synthetic ester liquids under DC voltage, refAbstract=null), Reference(id=1211311086440157536, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2022, volume=55, issue=3, pageStart=72, pageEnd=77, url=null, language=null, rfNumber=[17], rfOrder=16, authorNames=张君颖, 吴楠, 王野, journalName=绝缘材料, refType=null, unstructuredReference=张君颖,吴楠,王野,等.基于放电统计参数动态变化率的油浸芳纶绝缘纸局部放电发展过程研究[J].绝缘材料,2022,55(3):72-77., articleTitle=基于放电统计参数动态变化率的油浸芳纶绝缘纸局部放电发展过程研究, refAbstract=null), Reference(id=1211311086503072099, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2015, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[18], rfOrder=17, authorNames=International Electrotechnical Commission, journalName=null, refType=null, unstructuredReference=International Electrotechnical Commission. High-voltage test techniques-Partial discharge measurements:IEC 60270:2015[S]. Geneva:IEC,2015., articleTitle=null, refAbstract=null), Reference(id=1211311086574375270, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2015, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[19], rfOrder=18, authorNames=魏振, journalName=null, refType=null, unstructuredReference=魏振.复合电压下油纸绝缘局部放电发展过程及缺陷诊断方法[D].北京:华北电力大学,2015., articleTitle=复合电压下油纸绝缘局部放电发展过程及缺陷诊断方法, refAbstract=null), Reference(id=1211311086641484137, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2019, volume=43, issue=6, pageStart=2219, pageEnd=2226, url=null, language=null, rfNumber=[20], rfOrder=19, authorNames=万晓琪, 宋辉, 罗林根, journalName=电网技术, refType=null, unstructuredReference=万晓琪,宋辉,罗林根,等.卷积神经网络在局部放电图像模式识别中的应用[J].电网技术,2019,43(6):2219-2226., articleTitle=卷积神经网络在局部放电图像模式识别中的应用, refAbstract=null), Reference(id=1211311086700204396, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2022, volume=48, issue=5, pageStart=1705, pageEnd=1715, url=null, language=null, rfNumber=[21], rfOrder=20, authorNames=陈健宁, 周远翔, 白正, journalName=高电压技术, refType=null, unstructuredReference=陈健宁,周远翔,白正,等.基于多通道卷积神经网络的油纸绝缘局部放电模式识别方法[J].高电压技术,2022,48(5):1705-1715., articleTitle=基于多通道卷积神经网络的油纸绝缘局部放电模式识别方法, refAbstract=null), Reference(id=1211311086767313263, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2021, volume=54, issue=5, pageStart=10, pageEnd=20, url=null, language=null, rfNumber=[22], rfOrder=21, authorNames=陆云才, 范路, 陶风波, journalName=绝缘材料, refType=null, unstructuredReference=陆云才,范路,陶风波,等.人工智能在局部放电检测中的应用(一):去噪与故障定位[J].绝缘材料,2021,54(5):10-20., articleTitle=人工智能在局部放电检测中的应用(一):去噪与故障定位, refAbstract=null), Reference(id=1211311086838616433, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2018, volume=44, issue=11, pageStart=3625, pageEnd=3633, url=null, language=null, rfNumber=[23], rfOrder=22, authorNames=宋辉, 代杰杰, 张卫东, journalName=高电压技术, refType=null, unstructuredReference=宋辉,代杰杰,张卫东,等.复杂数据源下基于深度卷积网络的局部放电模式识别[J].高电压技术,2018,44(11):3625-3633., articleTitle=复杂数据源下基于深度卷积网络的局部放电模式识别, refAbstract=null), Reference(id=1211311086905725299, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2021, volume=54, issue=8, pageStart=8, pageEnd=17, url=null, language=null, rfNumber=[24], rfOrder=23, authorNames=郭金明, 田树军, 卓浩泽, journalName=绝缘材料, refType=null, unstructuredReference=郭金明,田树军,卓浩泽.绝缘油老化特性及诊断方法研究进展[J].绝缘材料,2021,54(8):8-17., articleTitle=绝缘油老化特性及诊断方法研究进展, refAbstract=null), Reference(id=1211311086985417077, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2017, volume=24, issue=2, pageStart=1239, pageEnd=1248, url=null, language=null, rfNumber=[25], rfOrder=24, authorNames=FAIZJ, SOLEIMANIM, journalName=IEEE Transactions on Dielectrics and Electrical Insulation, refType=null, unstructuredReference=FAIZJ, SOLEIMANIM. Dissolved gas analysis evaluation in electric power transformers using conventional methods a review[J]. IEEE Transactions on Dielectrics and Electrical Insulation,2017,24(2):1239-1248., articleTitle=Dissolved gas analysis evaluation in electric power transformers using conventional methods a review, refAbstract=null), Reference(id=1211311087077691767, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2021, volume=45, issue=8, pageStart=3324, pageEnd=3332, url=null, language=null, rfNumber=[26], rfOrder=25, authorNames=张育杰, 冯健, 李典阳, journalName=电网技术, refType=null, unstructuredReference=张育杰,冯健,李典阳,等.基于油色谱数据的变压器故障征兆新优选策略[J].电网技术,2021,45(8):3324-3332., articleTitle=基于油色谱数据的变压器故障征兆新优选策略, refAbstract=null), Reference(id=1211311087136412025, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2018, volume=465, issue=null, pageStart=1, pageEnd=20, url=null, language=null, rfNumber=[27], rfOrder=26, authorNames=DOUZASG, BACAOF, LASTF, journalName=Information Sciences, refType=null, unstructuredReference=DOUZASG, BACAOF, LASTF. Improving imbalanced learning through a heuristic oversampling method based on k-means and SMOTE[J]. Information Sciences,2018,465:1-20., articleTitle=Improving imbalanced learning through a heuristic oversampling method based on k-means and SMOTE, refAbstract=null), Reference(id=1211311087211909499, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2017, volume=60, issue=6, pageStart=84, pageEnd=90, url=null, language=null, rfNumber=[28], rfOrder=27, authorNames=KRIZHEVSKYA, SUTSKEVERI, HINTONG E, journalName=Communications of the ACM, refType=null, unstructuredReference=KRIZHEVSKYA, SUTSKEVERI, HINTONG E. Imagenet classification with deep convolutional neural networks[J]. Communications of the ACM,2017,60(6):84-90., articleTitle=Imagenet classification with deep convolutional neural networks, refAbstract=null), Reference(id=1211311087274824061, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=1986, volume=323, issue=6088, pageStart=533, pageEnd=536, url=null, language=null, rfNumber=[29], rfOrder=28, authorNames=RUMELHARTD E, HINTONG E, WILLIAMSR J, journalName=Nature, refType=null, unstructuredReference=RUMELHARTD E, HINTONG E, WILLIAMSR J. Learning representations by back-propagating errors[J]. Nature,1986,323(6088):533-536., articleTitle=Learning representations by back-propagating errors, refAbstract=null), Reference(id=1211311087337738623, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, doi=null, pmid=null, pmcid=null, year=2020, volume=14, issue=24, pageStart=5774, pageEnd=5781, url=null, language=null, rfNumber=[30], rfOrder=29, authorNames=XUY, LIY, WANGY, journalName=IET Generation, refType=null, unstructuredReference=XUY, LIY, WANGY, et al. Integrated decision-making method for power transformer fault diagnosis via rough set and DS evidence theories[J]. IET Generation, Transmission & Distribution,2020,14(24):5774-5781., articleTitle=Integrated decision-making method for power transformer fault diagnosis via rough set and DS evidence theories, refAbstract=null)], funds=[Fund(id=1211311083416064291, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, awardId=52037009, language=CN, fundingSource=国家自然科学基金重点项目(52037009), fundOrder=null, country=null), Fund(id=1211311083508338981, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, awardId=2021YFB2402102, language=CN, fundingSource=国家重点研发项目(2021YFB2402102), fundOrder=null, country=null), Fund(id=1211311083604807976, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, awardId=5108-202218280A-2-356-XG, language=CN, fundingSource=国家电网科技项目(5108-202218280A-2-356-XG), fundOrder=null, country=null)], companyList=[AuthorCompany(id=1211311075337835490, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, xref=1, ext=[AuthorCompanyExt(id=1211311075342029795, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, companyId=1211311075337835490, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1The Wind Solar Storage Division of State Key Laboratory of Control and Simulation of Power System and Generation Equipment, School of Electrical Engineering, Xinjiang University, Urumqi 830047, China), AuthorCompanyExt(id=1211311075350418404, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, companyId=1211311075337835490, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1新疆大学 电气工程学院 电力系统及大型发电设备安全控制和仿真国家重点实验室风光储分室, 新疆 乌鲁木齐 830047)]), AuthorCompany(id=1211311075446887402, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, xref=2, ext=[AuthorCompanyExt(id=1211311075451081706, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, companyId=1211311075446887402, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2State Key Laboratory of Control and Simulation of Power Systems and Generation Equipment, Department of Electrical Engineering, Tsinghua University, Beijing 100084, China), AuthorCompanyExt(id=1211311075459470315, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, companyId=1211311075446887402, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2清华大学 电机工程与应用电子技术系 电力系统及发电设备控制和仿真国家重点实验室,北京 100084)])], figs=[ArticleFig(id=1211311079242731666, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=EN, label=Fig.1, caption=Six defect models and their electric field distribution, figureFileSmall=yUK003dqyO9Kqx9HItOuVQ==, figureFileBig=p3W/HhLXLsgSIDhGa40pUA==, tableContent=null), ArticleFig(id=1211311079339200662, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=CN, label=图1, caption=6种缺陷模型及其电场分布, figureFileSmall=yUK003dqyO9Kqx9HItOuVQ==, figureFileBig=p3W/HhLXLsgSIDhGa40pUA==, tableContent=null), ArticleFig(id=1211311079477612701, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=EN, label=Fig.2, caption=The experiment platform of partial discharge, figureFileSmall=3jTx+UWq5Tk1LFJUj8Q9LQ==, figureFileBig=6ZIvbNLZCsyYysVfScg7Xw==, tableContent=null), ArticleFig(id=1211311079574081699, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=CN, label=图2, caption=局部放电试验平台, figureFileSmall=3jTx+UWq5Tk1LFJUj8Q9LQ==, figureFileBig=6ZIvbNLZCsyYysVfScg7Xw==, tableContent=null), ArticleFig(id=1211311079687327913, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=EN, label=Fig.3, caption=PRPD grayscale maps of six defect types, figureFileSmall=69+jaeOCSXucWPke1MgYKQ==, figureFileBig=XTEAr0M6/y/1pi05GJvlKw==, tableContent=null), ArticleFig(id=1211311079792185519, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=CN, label=图3, caption=6种缺陷类型的PRPD灰度图, figureFileSmall=69+jaeOCSXucWPke1MgYKQ==, figureFileBig=XTEAr0M6/y/1pi05GJvlKw==, tableContent=null), ArticleFig(id=1211311079867682996, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=EN, label=Fig.4, caption=PRPD grayscale map of defect 2 with different noise, figureFileSmall=xjAtph0vk9KMoX53WhE3wA==, figureFileBig=CdN2/Vh9kODDxtAsAtq4iA==, tableContent=null), ArticleFig(id=1211311079918014648, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=CN, label=图4, caption=添加不同噪声后缺陷2的PRPD灰度图, figureFileSmall=xjAtph0vk9KMoX53WhE3wA==, figureFileBig=CdN2/Vh9kODDxtAsAtq4iA==, tableContent=null), ArticleFig(id=1211311080001900731, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=EN, label=Fig.5, caption=The structure of convolutional neural network, figureFileSmall=Y6WUkdIIkIYR/eHuTxqBig==, figureFileBig=X06ReqQfon/9RSgl4RBdHg==, tableContent=null), ArticleFig(id=1211311080098369728, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=CN, label=图5, caption=卷积神经网络结构, figureFileSmall=Y6WUkdIIkIYR/eHuTxqBig==, figureFileBig=X06ReqQfon/9RSgl4RBdHg==, tableContent=null), ArticleFig(id=1211311081314717891, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=EN, label=Fig.6, caption=The flow chart for information fusion of PRPD and DGA, figureFileSmall=RE+sFh3haCc4ACXh8R6Nqg==, figureFileBig=SfA39M0GgCT+sFYpLUC+BA==, tableContent=null), ArticleFig(id=1211311081406992581, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=CN, label=图6, caption=PRPD和DGA信息融合流程图, figureFileSmall=RE+sFh3haCc4ACXh8R6Nqg==, figureFileBig=SfA39M0GgCT+sFYpLUC+BA==, tableContent=null), ArticleFig(id=1211311081524433098, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=EN, label=Fig.7, caption=Effect of different of convolution pooling layers on the recognition accuracy of convolutional neural network, figureFileSmall=Uo7BYV0Lf9RRB2n7M+8vTA==, figureFileBig=no5rMLWrsyk2B0oLi0R29g==, tableContent=null), ArticleFig(id=1211311081620902097, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=CN, label=图7, caption=不同卷积池化层数对卷积神经网络识别准确率的影响, figureFileSmall=Uo7BYV0Lf9RRB2n7M+8vTA==, figureFileBig=no5rMLWrsyk2B0oLi0R29g==, tableContent=null), ArticleFig(id=1211311081704788179, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=EN, label=Fig.8, caption=Effect of different fully connected layers on the recognition accuracy of convolutional neural network, figureFileSmall=0FC0EqUJTLwN8zaX6vLc+Q==, figureFileBig=3Mg2INl06Yc7HagY96zO/g==, tableContent=null), ArticleFig(id=1211311081780285654, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=CN, label=图8, caption=不同全连接层数对卷积神经网络识别准确率的影响, figureFileSmall=0FC0EqUJTLwN8zaX6vLc+Q==, figureFileBig=3Mg2INl06Yc7HagY96zO/g==, tableContent=null), ArticleFig(id=1211311081876754650, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=EN, label=Fig.9, caption=Effect of different input dimensions on the recognition accuracy of convolutional neural network, figureFileSmall=JDHB87OUjDaWbClEaw26Wg==, figureFileBig=L9/U9dwOh103Qfl67ZeAhg==, tableContent=null), ArticleFig(id=1211311081969029345, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=CN, label=图9, caption=不同输入维度对卷积神经网络识别准确率的影响, figureFileSmall=JDHB87OUjDaWbClEaw26Wg==, figureFileBig=L9/U9dwOh103Qfl67ZeAhg==, tableContent=null), ArticleFig(id=1211311082057109734, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=EN, label=Fig.10, caption=Effect of different hidden layers on the recognition accuracy of BPNN, figureFileSmall=Gv25cGLllE5ZfSYu2vRLlg==, figureFileBig=lVZt3dhcpZw0Prjorw9/ag==, tableContent=null), ArticleFig(id=1211311082161967336, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=CN, label=图10, caption=不同隐含层数对反向传播神经网络识别准确率的影响, figureFileSmall=Gv25cGLllE5ZfSYu2vRLlg==, figureFileBig=lVZt3dhcpZw0Prjorw9/ag==, tableContent=null), ArticleFig(id=1211311082271019247, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=EN, label=Fig.11, caption=Comparison on recognition accuracy of three different criterion models under different input dimensions of PRPD spectrum, figureFileSmall=Z1pLyFXRLsQaxUOaWs5ckg==, figureFileBig=raSKau7DVAkhAAGekT1gNg==, tableContent=null), ArticleFig(id=1211311082363293941, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=CN, label=图11, caption=不同PRPD图谱输入维度下三种不同判据模型的识别准确率对比, figureFileSmall=Z1pLyFXRLsQaxUOaWs5ckg==, figureFileBig=raSKau7DVAkhAAGekT1gNg==, tableContent=null), ArticleFig(id=1211311082476540153, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=EN, label=Tab.1, caption=A set of gases ratio of six defect models, figureFileSmall=null, figureFileBig=null, tableContent=
缺陷比值
C2H2/H2C2H4/CH4H2/TCHCH4/TCHCH4/D
缺陷10.8221.2360.8470.1120.119
缺陷21.5361.5250.5030.0810.083
缺陷31.0321.1630.7070.1100.113
缺陷40.9900.9770.7090.1350.139
缺陷50.8170.9730.8280.1440.150
缺陷60.7570.7570.9140.1510.158
), ArticleFig(id=1211311082585592058, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=CN, label=表1, caption=

6种缺陷模型的一组气体比值

, figureFileSmall=null, figureFileBig=null, tableContent=
缺陷比值
C2H2/H2C2H4/CH4H2/TCHCH4/TCHCH4/D
缺陷10.8221.2360.8470.1120.119
缺陷21.5361.5250.5030.0810.083
缺陷31.0321.1630.7070.1100.113
缺陷40.9900.9770.7090.1350.139
缺陷50.8170.9730.8280.1440.150
缺陷60.7570.7570.9140.1510.158
), ArticleFig(id=1211311082703032575, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=EN, label=Tab.2, caption=Parameter configuration of BPNN with different hidden layers, figureFileSmall=null, figureFileBig=null, tableContent=
网络输入层神经元/个隐含层神经元/个输出层神经元/个
第一层第二层第三层
N1560006
N25603006
N356030126
), ArticleFig(id=1211311082791112963, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=CN, label=表2, caption=

不同隐含层数的反向传播神经网络参数配置

, figureFileSmall=null, figureFileBig=null, tableContent=
网络输入层神经元/个隐含层神经元/个输出层神经元/个
第一层第二层第三层
N1560006
N25603006
N356030126
), ArticleFig(id=1211311082895970569, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=EN, label=Tab.3, caption=A set of BPA of defect 2 before and after D-S evidence fusion, figureFileSmall=null, figureFileBig=null, tableContent=
BPAPRPD图谱DGA数据D-S融合
m(A1)0.158 80.145 50.085 0
m(A2)0.379 50.511 30.714 1
m(A3)0.445 00.116 40.190 6
m(A4)00.060 10
m(A5)000
m(A6)000
m(θ)0.016 70.166 70.010 3
), ArticleFig(id=1211311082992439565, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=CN, label=表3, caption=

一组缺陷2的D-S证据融合前后的BPA

, figureFileSmall=null, figureFileBig=null, tableContent=
BPAPRPD图谱DGA数据D-S融合
m(A1)0.158 80.145 50.085 0
m(A2)0.379 50.511 30.714 1
m(A3)0.445 00.116 40.190 6
m(A4)00.060 10
m(A5)000
m(A6)000
m(θ)0.016 70.166 70.010 3
), ArticleFig(id=1211311083109880080, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=EN, label=Tab.4, caption=The recognition accuracy of different neural network fusion models, figureFileSmall=null, figureFileBig=null, tableContent=
数据类型模型识别准确率/%模型大小/MB
PRPD图谱+DGA数据CNN-BPNN99.170.69
Vgg16-BPNN99.01128
ResNet18-BPNN98.3042.74
DenseNet121-BPNN98.1827.94
), ArticleFig(id=1211311083214737688, tenantId=1146029695717560320, journalId=1149653034449285133, articleId=1199644781044400370, language=CN, label=表4, caption=

不同神经网络融合模型的识别准确率

, figureFileSmall=null, figureFileBig=null, tableContent=
数据类型模型识别准确率/%模型大小/MB
PRPD图谱+DGA数据CNN-BPNN99.170.69
Vgg16-BPNN99.01128
ResNet18-BPNN98.3042.74
DenseNet121-BPNN98.1827.94
)], attaches=null, journal=Journal(id=1146437660891000859, delFlag=0, nameCn=绝缘材料, nameEn=Insulating Materials, nameHistory1=null, nameHistory2=null, issn=1009-9239, eissn=null, cn=45-1287/TM, coden=null, periodic=0, language=CN, oaType=是, ccby=CC BY-NC-ND, superviseOffice=null, ownerOffice=null, pubOffice=null, editorOffice=null, officeType=null, aims=null, clcCode=null, officeProv=null, officeCity=null, officeAddr=null, officeZip=null, officeEmail=null, officePhone=null, editDirector=null, officeDirector=null, officeDirectorPhone=null, officeStaffNum=null, officeEmpNum=null, coverPicUrl=To3JhdgsOj1Q7eh5WnUN1g==, journalPrice=null, startedYear=null, abbrevIsoEn=null, journalRemark=null, publicationField=null, createdTime=1751261638140, updatedTime=1761735740591, createdBy=18614031015, updatedBy=13701087609, firstLetterCn=I, firstLetterEn=I, subjectCode=Natural Sciences, subjectName=Natural Sciences, subjectCodeEn=Natural Sciences, subjectNameEn=null, picCn=To3JhdgsOj1Q7eh5WnUN1g==, picEn=O7zoVkJ9hJoi8iGUSQYTxg==, jcr=null, cjcr=null, exts=[JournalExt(id=1190369230903152828, language=CN, name=绝缘材料, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1761735740638, updatedTime=1761735740638, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://jyct.cbpt.cnki.net/EditorDN/index.aspx?t=1, submissionEditorUrl=https://jyct.cbpt.cnki.net/EditorDN/index.aspx?t=3, submissionReviewUrl=https://jyct.cbpt.cnki.net/EditorDN/index.aspx?t=2, submissionCeEditorUrl=, submissionAeEditorUrl=, option={"copyright":""}), JournalExt(id=1190369230945095869, language=EN, name=Insulating Materials, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1761735740648, updatedTime=1761735740648, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://jyct.cbpt.cnki.net/EditorDN/index.aspx?t=1, submissionEditorUrl=https://jyct.cbpt.cnki.net/EditorDN/index.aspx?t=3, submissionReviewUrl=https://jyct.cbpt.cnki.net/EditorDN/index.aspx?t=2, submissionCeEditorUrl=, submissionAeEditorUrl=, option={"copyright":""})], databaseList=null, tenantJournalId=1149653034449285133, websiteList=[Website(id=1189939819993166814, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1149653034449285133, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/jycl/CN, language=CN, createTime=1761633361099, createBy=18614031015, updateTime=1761633401425, updateBy=18614031015, name=绝缘材料-中文, tplId=1146099689490845704, title=绝缘材料, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1189940292275991527, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1189939819993166814, code=articleTextType, value=kx, createTime=1761633473700, updateTime=1761633473700, creator=18614031015, updator=18614031015), WebsiteProps(id=1189940292250825700, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1189939819993166814, code=banner, value=null, createTime=1761633473694, updateTime=1761633473694, creator=18614031015, updator=18614031015), WebsiteProps(id=1189940292301157354, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1189939819993166814, code=grayFlag, value=0, createTime=1761633473706, updateTime=1761633473706, creator=18614031015, updator=18614031015), WebsiteProps(id=1189940292238242787, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1189939819993166814, code=logo, value=https://castjournals.cast.org.cn/joweb/jycl/CN/file/pic?fileId=MyqZAHzZT6tMetr2hjDKLQ==, createTime=1761633473691, updateTime=1761633473691, creator=18614031015, updator=18614031015), WebsiteProps(id=1189940292322128876, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1189939819993166814, code=minRunFlag, value=0, createTime=1761633473711, updateTime=1761633473711, creator=18614031015, updator=18614031015), WebsiteProps(id=1189940292267602918, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1189939819993166814, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/jycl/CN/file/pic, createTime=1761633473698, updateTime=1761633473698, creator=18614031015, updator=18614031015), WebsiteProps(id=1189940292309545963, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1189939819993166814, code=silenceFlag, value=0, createTime=1761633473708, updateTime=1761633473708, creator=18614031015, updator=18614031015), WebsiteProps(id=1189940292259214309, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1189939819993166814, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1761633473696, updateTime=1761633473696, creator=18614031015, updator=18614031015), WebsiteProps(id=1189940292284380136, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1189939819993166814, code=themeColor, value=null, createTime=1761633473702, updateTime=1761633473702, creator=18614031015, updator=18614031015), WebsiteProps(id=1189940292292768745, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1189939819993166814, code=themeStyle, value=null, createTime=1761633473704, updateTime=1761633473704, creator=18614031015, updator=18614031015)]), Website(id=1189939820060275680, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1149653034449285133, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/jycl/EN, language=EN, createTime=1761633361115, createBy=18614031015, updateTime=1761633397117, updateBy=18614031015, name=绝缘材料-英文, tplId=1146101810881728533, title=Insulating Materials, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1189940323548722161, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1189939820060275680, code=articleTextType, value=kx, createTime=1761633481156, updateTime=1761633481156, creator=18614031015, updator=18614031015), WebsiteProps(id=1189940323519362030, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1189939820060275680, code=banner, value=null, createTime=1761633481149, updateTime=1761633481149, creator=18614031015, updator=18614031015), WebsiteProps(id=1189940323569693684, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1189939820060275680, code=grayFlag, value=0, createTime=1761633481161, updateTime=1761633481161, creator=18614031015, updator=18614031015), WebsiteProps(id=1189940323506779117, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1189939820060275680, code=logo, value=https://castjournals.cast.org.cn/joweb/jycl/EN/file/pic?fileId=MyqZAHzZT6tMetr2hjDKLQ==, createTime=1761633481146, updateTime=1761633481146, creator=18614031015, updator=18614031015), WebsiteProps(id=1189940323586470902, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1189939820060275680, code=minRunFlag, value=0, createTime=1761633481165, updateTime=1761633481165, creator=18614031015, updator=18614031015), WebsiteProps(id=1189940323540333552, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1189939820060275680, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/jycl/EN/file/pic, createTime=1761633481154, updateTime=1761633481154, creator=18614031015, updator=18614031015), WebsiteProps(id=1189940323578082293, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1189939820060275680, code=silenceFlag, value=0, createTime=1761633481163, updateTime=1761633481163, creator=18614031015, updator=18614031015), WebsiteProps(id=1189940323531944943, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1189939820060275680, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_en_623/, createTime=1761633481152, updateTime=1761633481152, creator=18614031015, updator=18614031015), WebsiteProps(id=1189940323557110770, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1189939820060275680, code=themeColor, value=null, createTime=1761633481158, updateTime=1761633481158, creator=18614031015, updator=18614031015), WebsiteProps(id=1189940323565499379, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1189939820060275680, code=themeStyle, value=null, createTime=1761633481160, updateTime=1761633481160, creator=18614031015, updator=18614031015)])], journalTitle=绝缘材料, weixinUrl=null, journalUrl=https://www.insulation.org.cn/, iacademicId=null, status=1, seqNo=null, journalTitleEn=Insulating Materials, journalPhotoCn=To3JhdgsOj1Q7eh5WnUN1g==, journalPhotoEn=O7zoVkJ9hJoi8iGUSQYTxg==, journalFirstLetter=I, journalRecommend=null, journalNew=null, journalCollection=null, jcrJf=null, cjcrJf=null, jcrJfStr=null, cjcrJfStr=null, submissionFirstDecision=null, sciSubjectClassification=null, casSubjectClassification=null, citeScore=null, totalCitationFrequency=null, icpCode=null, psCode=null, advertisingLicenseCode=null, copyrightInformation=null, country=null, option=, provinceCode=null, provinceName=null, collectFlag=false), detailUrlCn=https://castjournals.cast.org.cn/joweb/jycl/CN/10.16790/j.cnki.1009-9239.im.2023.12.006, detailUrlEn=https://castjournals.cast.org.cn/joweb/jycl/EN/10.16790/j.cnki.1009-9239.im.2023.12.006, pdfUrlCn=https://castjournals.cast.org.cn/joweb/jycl/CN/PDF/10.16790/j.cnki.1009-9239.im.2023.12.006, pdfUrlEn=https://castjournals.cast.org.cn/joweb/jycl/EN/PDF/10.16790/j.cnki.1009-9239.im.2023.12.006, aliStartDate=null, aliEndDate=null, collectionFlag=false, citedCount=null, citedUrl=null, reference=null)
收藏切换
基于局部放电相位图谱和油中溶解气体信息融合的油纸绝缘缺陷识别方法
收藏切换
PDF下载
周远翔 1, 2 , 李永印 1 , 陈健宁 2 , 白正 2
绝缘材料 | 面向大容量海上风电送出的先进电工材料 2023,56(12): 43-53
收起
收藏切换
绝缘材料 | 面向大容量海上风电送出的先进电工材料 2023, 56(12): 43-53
基于局部放电相位图谱和油中溶解气体信息融合的油纸绝缘缺陷识别方法
全屏
周远翔1, 2, 李永印1, 陈健宁2, 白正2
作者信息
  • 1新疆大学 电气工程学院 电力系统及大型发电设备安全控制和仿真国家重点实验室风光储分室, 新疆 乌鲁木齐 830047
  • 2清华大学 电机工程与应用电子技术系 电力系统及发电设备控制和仿真国家重点实验室,北京 100084
  • 周远翔(1966-),男(汉族),福建莆田人,教授,主要从事高电压与绝缘技术的研究;

    李永印(1996-),男(汉族),河南信阳人,硕士生,主要从事油纸绝缘局部放电与人工智能应用研究。

Defect recognition method of oil-paper insulation based on information fusion of PRPD spectrum and dissolved gas data
Yuanxiang ZHOU1, 2, Yongyin LI1, Jianning CHEN2, Zheng BAI2
Affiliations
  • 1The Wind Solar Storage Division of State Key Laboratory of Control and Simulation of Power System and Generation Equipment, School of Electrical Engineering, Xinjiang University, Urumqi 830047, China
  • 2State Key Laboratory of Control and Simulation of Power Systems and Generation Equipment, Department of Electrical Engineering, Tsinghua University, Beijing 100084, China
出版时间: 2023-12-20 doi: 10.16790/j.cnki.1009-9239.im.2023.12.006
文章导航
收藏切换

基于单一检测手段的变压器故障诊断方法难以对油纸绝缘的同一类型缺陷进行细化识别,无法满足深远海风电快速发展背景下电力系统对设备运行可靠性的要求。因此,本文提出了一种基于局部放电相位(PRPD)图谱和油中溶解气体分析(DGA)信息融合的油纸绝缘缺陷识别方法,设计并制作了6种电极模型,模拟变压器中不同电场不均匀系数的沿面放电典型缺陷,并采集其PRPD及DGA数据;分别采用卷积神经网络(CNN)和反向传播神经网络(BPNN)对6类缺陷的PRPD图谱和DGA特征向量进行模式识别;提出基于D-S证据理论的CNN-BPNN信息融合模型,实现基于PRPD图谱与DGA数据的联合诊断。结果表明:基于D-S证据理论的CNN-BPNN模型可有效纠正单一判据模型的错误输出,并降低分类结果的不确定度,当PRPD图谱输入维度为8×8、16×16、32×32时,融入DGA特征向量的模型识别准确率分别为93.21%、97.53%、99.17%,较PRPD图谱单一判据模型的识别准确率分别提升了4.81%、2.78%、0.84%,该模型可有效融合局部放电的电气物理信息和化学产物信息,既提高了缺陷识别准确率,又增强了输出结果的置信程度,且降低了数据存储要求,可为变压器智能运维提供精确、可靠、轻量的缺陷识别方法。

沿面放电  /  PRPD图谱  /  油中溶解气体  /  神经网络  /  D-S证据理论

The transformer fault diagnosis technique based on a single detection method is difficult to identify the same type of defects of oil-paper insulation in detail, which cannot meet the requirements of power system on equipment operation reliability under the background of rapid development of deep offshore wind power. Therefore, an oil-paper insulation defect identification method based on information fusion of phase-resolved partial discharge (PRPD) spectrum and dissolved gas analysis (DGA) data was proposed. Six kinds of electrode models were designed and made to simulate the typical defects of surface discharge in transformers with different electric field inhomogeneity coefficients, and PRPD and DGA data were collected. Then convolutional neural network (CNN) and back propagation neural network (BPNN) were adopted to recognize the patterns of PRPD spectrum and DGA feature vector of six kinds of defects, respectively. Finally, the CNN-BPNN information fusion model based on D-S evidence theory was proposed to realize joint diagnosis based on PRPD spectrum and DGA data. The results show that the CNN-BPNN model based on the D-S evidence theory can effectively correct the wrong output of the single criterion model and reduce the uncertainty of the classification results. When the input dimensions of PRPD spectrum are 8×8, 16×16, and 32×32, the recognition accuracy of the model integrated with the DGA feature vector is 93.21%, 97.53%, and 99.17%, respectively, which is 4.81%, 2.78%, and 0.84% higher than that of PRPD single criterion model. The CNN-BPNN model can effectively integrate the electrical physical information and chemical product information of partial discharge, which not only improves the accuracy of defect identification, but also enhances the confidence of the output results, and reduces the data storage requirements, providing accurate, reliable, and lightweight defect identification methods for intelligent operation and maintenance of transformers.

surface discharge  /  PRPD spectrum  /  dissolved gas in oil  /  neural network  /  D-S evidence theory
周远翔, 李永印, 陈健宁, 白正. 基于局部放电相位图谱和油中溶解气体信息融合的油纸绝缘缺陷识别方法. 绝缘材料, 2023 , 56 (12) : 43 -53 . DOI: 10.16790/j.cnki.1009-9239.im.2023.12.006
Yuanxiang ZHOU, Yongyin LI, Jianning CHEN, Zheng BAI. Defect recognition method of oil-paper insulation based on information fusion of PRPD spectrum and dissolved gas data[J]. Insulating Materials, 2023 , 56 (12) : 43 -53 . DOI: 10.16790/j.cnki.1009-9239.im.2023.12.006
随着深远海风电的持续开发,海上换流站面临气候恶劣、巡检不便的问题,对变压器等电力设备的安全可靠运行提出了新的挑战。沿面放电因其发展速度快、绝缘损伤大,被认为是造成高压设备损坏的重要原因之一[1]。不同类型的沿面缺陷劣化机理不同,其对应的局部放电特性也不相同,开展变压器局部放电模式识别对其绝缘状态评估及海上风电安全可靠送出具有重要意义[2-3]
目前国内外学者已对单一信息源的局部放电相位(phase-resolved partial discharge,PRPD)图谱模式识别开展了广泛的研究,主要包括两类方法:一类是支持向量机(support vector machine,SVM)和概率神经网络(probabilistic neural network,PNN)等传统机器学习方法[4],此类模式识别方法需要依赖领域内的专家经验设计、处理、提取、组合某一方面或几方面的统计特征,存在一定的主观性,无法反映局部放电的全部特征,泛化性能较差,因此难以得到令人满意的分类效果[5];另一类是深度学习算法[6-7]和网络迁移学习方法[8-9],该类方法优于传统的参数统计方法,但由于模型的层数较深,训练所需数据量较大,一方面容易出现过拟合,另一方面对数据存储要求高,为边缘侧计算带来了挑战。
近年来,随着人工智能和传感技术的发展,采用信息融合的方法对高压设备绝缘缺陷进行模式识别成为未来电力系统智能化运维的发展趋势。王涤等[10]和黄亮等[11]分别提取了局部放电不同特征信息并输入到神经网络中进行识别,再利用Dempster-Shafer(D-S)证据理论融合不同分类器的分类结果,结果表明融合后的识别准确率高于只提取单类特征分类器的准确率,但其本质上还是对同一信息源的利用;艾嘉伟等[12]、耿伊雯等[13]和王磊等[14]等利用不同传感器进行联合诊断,充分综合了传统特高频(ultra-high frequency,UHF)、光电等测量方法的优势,结果表明多信息融合有利于提高电气设备局部放电模式识别的准确率。以上研究选取的缺陷类型主要有气隙、沿面、悬浮等,而不同类型缺陷的PRPD图谱特征具有显著差异,通过传统的统计参数提取方法和简单的机器学习方法即可达到较高的分类准确率,但是其对同一类型缺陷的细分程度不足,在实际工程中难以根据测量与分类结果为缺陷尺寸识别、分级评估及故障定位提供有效指导,且需在现有的在线监测系统中部署额外的传感器。而油中溶解气体分析(dissolved gas analysis,DGA)作为油纸绝缘局部放电常用的化学检测法,不受电磁环境噪声的影响,利用电检测法与化学检测法对缺陷进行融合识别的研究鲜有报道。
本文综合局部放电常用的电检测法和化学检测法的优点,引入脉冲电流法和气相色谱法两种方法进行融合判别,提出了基于PRPD图谱和DGA数据信息融合的油纸绝缘缺陷识别方法。首先搭建局部放电及油中溶解气体分析试验平台,设计6种电极模型,模拟变压器中不同电场不均匀系数的沿面局部放电典型缺陷,并采集PRPD图谱和DGA数据;然后分别使用卷积神经网络(convolutional neural network,CNN)及反向传播神经网络(back propagation neural network,BPNN)将两类信息源的数据进行分类;最后利用D-S证据理论将识别结果融合,得到最终的缺陷类别,并与PRPD、DGA单一判据模型及视觉几何群网络(Vgg)、残差网络(ResNet)、密集连接卷积网络(DenseNet)等深度卷积融合模型的识别结果进行对比,验证本文所提的信息融合模型的有效性。
试验绝缘油样品选用克拉玛依KI25X型绝缘油,其处理流程为:将变压器油抽入40℃真空滤油机中进行过滤、干燥和除气,充分去除油中颗粒物、水分等杂质,过滤后的变压器油满足DL/T 1096—2018中的相关规定,随后将其装入经无水乙醇、去离子水充分洗净的干燥烧杯中,并放入真空干燥箱中干燥48 h以上,采用SF-5型微量水分测定仪测量水分,确保油中微水含量符合GB/T 7595—2017的要求。试验绝缘纸板选用厚度为1 mm的魏德曼纸板,油浸纸板的制作流程如下:首先将绝缘纸板统一放在温度为105℃、压强为100 Pa的真空干燥箱中干燥48 h;然后在80℃环境下对绝缘纸板进行真空浸油,充分浸渍后自然冷却至室温;最后再转入真空干燥箱中干燥48 h以上。在进行正式试验前,对油浸纸板进行理化测试,结果满足IEC 60641.2:2004的各项指标要求。
本文设计并制作了6种电极模型,对变压器中不同电场分布的沿面局部放电缺陷进行模拟,6种缺陷模型及其电场分布仿真如图1所示。其中缺陷1为IEC 60243.1:2013推荐的柱板电极放电模型,柱电极的高度和直径均为25 mm;缺陷2为球板电极放电模型,球电极直径为25 mm,其曲率半径等于球的半径,即12.5 mm;缺陷1、2对应变压器绕组之间、绕组对铁心或外壳等稍不均匀电场环境下的沿面爬电,电场不均匀系数分别为2.45、2.20;缺陷3、4是曲率半径分别为4 mm、0.7 mm的电极模型,对应变压器中的金属尖端缺陷,电场不均匀系数分别为2.42、3.72;缺陷5、6是曲率半径分别为0.1 mm、0.005 mm的电极模型,对应变压器绕组塑性形变后产生的锋利金属突出物缺陷,电场不均匀系数分别为9.17、33.2[15-17]。6种缺陷模型的高压电极与地电极之间为1 mm厚的方形油浸纸板,在强垂直电场的作用下,既存在油纸界面上平行分量产生的沿面流注放电,又存在沿电极边缘垂直分量引起的油隙放电,6类缺陷电场最大处均位于高压电极、油浸纸板和油的结合处。
为获取6种缺陷下油纸绝缘局部放电的PRPD图谱,搭建试验平台如图2所示,该试验平台满足IEC 60270:2015的局部放电测试标准[18]。试验电路包括三部分:电源、试样回路和信号采集。电源主要包括AFG 3011C型可编程信号发生器和Trek model(50 kV/12 mA)型高压功率放大器,可编程信号发生器负责产生电压波形,高压功率放大器将产生信号按照固定增益1∶5 000放大后加到样品上。试样回路中待测样品与耦合电容、测量阻抗直接并联,测量阻抗与耦合电容串联将脉冲电流信号转为电压信号。信号采集通过MPD600型局部放电测试仪从测量阻抗中采集局部放电信号,并将其传输至PC端进行数据分析与处理。数字示波器用来监视高压功率放大器的出口电压。试验回路中所有设备引线端口做平滑处理,并清空周围金属器件。
在每次调整电路或更换缺陷模型后,需要对空载电路进行工频升压测试,保证回路升压至35 kV时无放电发生,且测试系统噪声小于8 pC。在正式试验前,需要进行预试验测量6类缺陷的局部放电起始电压与击穿电压,最后确定6类缺陷交流试验电压有效值分别为:缺陷1:19 kV±0.5 kV;缺陷2:20 kV±0.5 kV;缺陷3:18 kV±0.5 kV;缺陷4:17 kV±0.5 kV;缺陷5:16 kV±0.5 kV;缺陷6:15 kV±0.5 kV。
为获取6类缺陷下油纸绝缘在局部放电过程中的DGA数据,使用GC 2002型气相色谱仪对油中溶解气体进行测量,该仪器的灵敏度能充分满足对油中≤1 µL/L的H2和≤0.1 µL/L的C2H2等烃类气体的检测需求。每次采集完PRPD图谱后,将试样静置20 min[19],保证产生的气体充分溶解于变压器油中,取样及脱气方法参照GB/T 7252—2001的规定,取油样的部位在油盒的中部。
油纸绝缘局部放电是一个间歇性的过程,在试验过程中,放电较为剧烈时,10 pC以上放电量的放电频次可达每秒几万次,而放电较为轻微时,放电仅为每秒几十次。为扩充PRPD图谱样本,使用滑动窗口法[20]对原始数据进行剪裁,其过程为:假设某次试验共录制放电数据点数为100 N,每张PRPD图谱数据量为N,需要生成n张图谱(n>100)。滑动数据剪裁从第一个放电点开始,每N个放电点为一张图谱,下一张图谱的起点为上一个数据起点加上[99/(n-1)]N个放电点。例如,第一组数据为0~N个放电点,第二组数据为[99/(n-1)]N~[99/(n-1)+1]N个放电点,以此类推。
将图谱灰度化一方面是为了降低数据量,另一方面是相比于原始PRPD图谱,灰度图具有信息集中、特征更为直观的优点。灰度化后的PRPD图谱更标准、规范,便于卷积神经网络对其进行特征提取。根据2.1.1得到的数据绘制不同放电次数的PRPD灰度图。由于放电幅值、试验电压以及放电间歇性等方面的差异,不同缺陷局部放电的数据差距较大。为提高输入图谱的规范性,使模型对不同放电量级的数据均进行有效地识别,绘制同一缺陷不同放电次数的PRPD图谱时均以统一值作为图谱的纵坐标最大值。
按照文献[21]的方法得到带参考电压波形的6类缺陷的PRPD灰度图,部分图谱如图3所示。由图3可知,图谱具有典型的相位分布,工频相位正、负半周上放电幅值都在工频电压幅值附近。总体上,随着缺陷电场不均匀系数的增大,局部放电的幅值逐渐减小,放电也越集中。其中,柱板电极下放电图谱形状为兔耳型,且其最大放电量要略高于曲率半径为12.5 mm的缺陷模型。5种不同曲率半径的缺陷模型图谱大致为三角形,以电压峰值为中心展开分布。电极曲率半径的减小意味着与纸板垂直的纵向电场分量将得到加强,放电更加剧烈,油浸纸板的劣化将更加严重,甚至会导致绝缘纸板碳化并击穿。但是,不同电极结构下局部放电的放电量-相位分布特征相似度较高,图谱均为对称山丘状,在实际工程中难以利用其统计参数对不同电场不均匀系数的缺陷进行准确识别和有效预警。
实际工程中,变压器所处的电磁环境复杂,噪声来源多[22]。变电站常见的噪声包括设备刚启动时的电磁噪声、变压器冷却风机和油泵等运行时引起的机械噪声等。为使图谱数据符合实际工况,进而提高模型的泛化性能,在数据集中间等概率地对PRPD图谱人工添加相位固定型噪声、区间型噪声和白噪声[23],添加不同噪声后缺陷2的PRPD灰度图如图4所示。
变压器油中含有约3 000种化学成分,这些化合物按照其结构可分为环烷烃、链烷烃和芳香烃3类[24],这些烃类化合物在电、热条件下会发生分解。对于油中溶解气体的分析,国内外学者建立了Doernenberg比值、IEC比值、Rogers比值等气体比值故障诊断方法[25],这些传统方法使用不同的气体比值作为诊断依据,在实际工程中得到广泛应用,但目前只能区分高能放电、低能放电、过热等故障,对缺陷的类型及故障严重程度无法准确评估。
不同气体比值丰富了缺陷特征的筛选范围,但也增加了特征之间的冗余性[26],所以本文选择5种差异性较大的气体比值作为样本特征,6种缺陷放电后其中一次采样的数据如表1所示,其中总烃(TCH)为CH4、C2H2、C2H4、C2H6含量之和,D为CH4、C2H2、C2H4含量之和。
表1可知,除柱板电极(缺陷1)外,随着缺陷曲率半径的减小,5种缺陷的气体比值都呈单调变化;曲率半径的减小意味着垂直纸板方向的放电加强,沿油纸界面上的流注放电将减弱,电场畸变更加严重,会加剧C-C键、C-H键断裂,促进变压器油分解为H2、CH4、C2H2等小分子,分解速度与分子量成反比,故C2H2/H2、C2H4/CH4的比值逐渐减小,而H2/TCH、CH4/TCH、CH4/D的比值逐渐增大。柱板电极缺陷的产气特点与5种不同曲率半径的缺陷模型略有差别,特征气体为H2和C2H2。综上,不同电场不均匀系数的缺陷在放电时产气特性具有一定差异,可作为电检测法的补充,提高缺陷模式识别的准确率。
为使不同缺陷下测得的DGA样本量与PRPD图谱相匹配,提高样本的多样性,避免模型过拟合,需要对DGA数据进行样本扩充。本文采用合成少数类过采样技术(synthetic minority over-sampling technique,SMOTE)[27]将DGA样本量扩充到与PRPD图谱一致。合成少数类过采样技术是在邻近的少数类样本之间进行随机线性插值合成新的少数类样本,从而实现不平衡数据集的均衡化。设样本集中少数类样本的集合为X={x1x2,…,xn},其中xn表示第n个少数类样本的特征向量。具体步骤如下:
(1)对于DGA数据中的每一个样本xn,以欧氏距离为标准计算它到同一类数据所有样本的距离,得到其k个近邻样本。
(2)随机选取这k个近邻样本中的l个样本(l<k),l的大小由采样比例决定。
(3)对于每一个随机选出的近邻样本xm,分别与原数据按照式(1)中随机线性插值的方法合成新样本。
xnew=xn+(xm-xn)δ
式(1)中:xnew为新的少数类样本;δ为一个处于0~1之间的随机数。
本文采用二维卷积神经网络[28]对PRPD图谱进行模式识别,其基本结构包括:输入层、卷积层、池化层、全连接层和输出层。卷积层将灰度化后的PRPD图谱进行特征信息提取并形成新的特征子图,其卷积核本质是一种过滤器,用于获取图片的纹理、边沿等特征。池化层又被称为下采样层,作用是降低输出参数量,用更高层次的特征表示图像。全连接层作为分类器,接收经卷积及池化后提取的特征,通过参数学习将其映射到样本标记空间,最后经Softmax层将输出映射为6类缺陷的概率分布。本文搭建的卷积神经网络结构如图5所示,其中包含1个输入层,2个卷积池化层、3个全连接层和1个输出层。
使用反向传播神经网络[29]对DGA数据进行模式识别,该网络主要由输入层、隐含层、输出层构成,输入层输入的是2.2节得到的5种气体比值构成的特征向量,隐含层用于学习输入特征至输出向量的非线性映射关系,输出层经Softmax激活函数将输出向量映射成6类缺陷的概率分布。
D-S证据理论最早应用于专家系统中,是一种处理不确定信息的理论,主要由概率分配函数、置信函数、似然函数等构成[30]。本文利用D-S证据理论将PRPD图谱和DGA数据的识别结果进行融合,结合多种测量手段联合识别的优势,从而实现比单一信息源更为准确的分类效果。具体融合步骤如下:
(1)建立辨识框架
D-S证据理论中,由多个互斥命题组成的全部集合称为辨识框架。对于本文的模式识别问题,识别框架Ф为6种缺陷类型,分别用A1A2A3A4A5A6表示,如式(2)所示,不确定度为θ
Φ=A1, A2, A3, A4, A5, A6
(2)设置基本概率分配(BPA)
识别框架Ф的幂集构成命题集合Ψ=2Ф={ø(空集),{A1},{A2},…,{A6},{A1A2},…,Ф},∀AΨ,若函数m满足式(3)条件,则m被称为基本概率分配(basic probability assignment,BPA),则m(A)是命题A的基本概率数,即支持A发生的概率。
m(ø)=0      m(A)0,1Aψm(A)=1
本文以脉冲电流法和气相色谱法测得的数据经模式识别后的输出值作为2个独立的证据,将其转换为满足D-S证据理论的BPA。识别框架上不同检测方法的识别结果BPA计算过程如式(4)~(5)所示。
mi(Aj)=αiμij, i1,2, j1,2,3,4,5,6
mi(θ)=1-j=16mi(Aj)
式(4)~(5)中:mi(Aj)为第i个证据中第j个类别的BPA;µij表示第i个证据输出第j类缺陷的隶属度,在本文中为第i个信息源分类模型经Softmax输出的概率值;αi为第i个证据源的可靠系数,在本文中为第i个信息源分类模型的识别准确率;mi(θ)表示第i个证据不确定度的BPA的大小。
(3)证据合成
证据合成实际上是多个证据作正交运算,对于PRPD图谱及DGA数据的BPA合成计算方法如式(6)所示。
m(Aj)=11-Km1(Aj)m2(Aj)j1,2,3,4,5,6K=jkm1(Aj)m2(Ak)j,k1,2,3,4,5,6
式(6)中,K为归一化因子。通过式(6)合成PRPD图谱及DGA数据的BPA,输出结果即为PRPD图谱和DGA数据信息融合得到的结论。
(4)决策
由步骤(3)计算D-S证据融合后的所有BPA值,通过以下3个规则判断其缺陷类型:
规则I:m(Amax1)=max{m(Aj), AjΦm(Amax1)为其输出BPA的最大值,即规则I将最大值的BPA作为输出。
规则II:m(Amax1)>m(θ),规则II说明最大的BPA值需大于不确定度θ的BPA值。
规则III:m(Amax1)-m(Amax2)>εm(Amax2)为融合后BPA的次大值,ε为阈值,本文取ε=0.4。规则III说明只有两类识别方法融合后的BPA差异足够大时才将其输出。
为避免D-S证据组合规则在处理高度冲突或完全冲突的证据时产生与常理相悖的结果,若出现不能同时满足规则I、II、III的情况时,仍将PRPD图谱的BPA作为输出,具体基于信息融合的局部放电模式识别算法流程如图6所示。
由试验获取6类缺陷的PRPD图谱和DGA数据经预处理后,每类缺陷得到2 700张PRPD图谱和2 700条DGA数据,其中80%数据用于训练,20%数据用于测试。
在服务器中搭建相应的神经网络进行模型训练和测试。在卷积神经网络训练过程中,使用交叉熵作为损失函数评估训练误差,采用随机梯度下降法对参数进行更新,训练次数为240次。每次迭代遍历训练集中所有图片,每个训练周期进行60次迭代,共进行4个周期。初始学习率设置为0.001,每经过一个周期下降50%。在反向传播神经网络训练过程中,中间隐含层的神经元激活函数设置为tanh,学习步长为0.001,迭代次数也为240次。本文采用的深度学习框架为Pytorch,编程语言为Python3.9。
对于PRPD图谱单一信息源的缺陷识别,为提高模型性能,需要对卷积池化层数、全连接层数和输入维度等参数进行寻优。
首先,分别建立卷积池化层数为1、2、3、4,全连接层数固定为1的卷积神经网络,研究不同卷积池化层数对卷积神经网络识别准确率的影响,结果如图7所示。由图7可知,随着卷积池化层数的增加,模型的识别准确率呈现先上升后下降的趋势;其中,以2层的卷积神经网络表现最佳。当卷积池化仅有1层时,网络参数规模较小,难以拟合PRPD图谱的深层特征;当卷积池化层数高于2层时,模型识别的过程中出现了过拟合现象,识别准确率出现下降的趋势。
保持2层卷积池化层不变,建立全连接层数分别为1、2、3、4的卷积神经网络,研究不同全连接层数对卷积神经网络识别准确率的影响,结果如图8所示。由图8可知,全连接层为3时,卷积神经网络的识别准确率最高。随着全连接层数的增加,训练后期的识别准确率趋于稳定。因此,后文统一采用卷积池化层为2层、全连接层为3层的卷积神经网络作为PRPD图谱的分类模型。其中,第1个卷积层共有4个卷积核,第2个卷积层共有16个卷积核,卷积核大小均为3×3;2个池化层均采用最大值池化;3个全连接层的单元数分别为96、24、6。
选取输入分辨率分别为8×8、16×16、32×32、64×64的PRPD图谱对卷积神经网络进行训练与测试,得到不同输入维度下识别准确率随迭代次数的变化如图9所示。由图9可知,输入维度为32×32、64×64时,PRPD图谱信息得到较好的保留,识别准确率达到98.2%以上。当输入维度为8×8时,PRPD图谱在压缩过程中信息损失较大,模型难以对图谱进行准确的特征提取。随着图谱分辨率增加,识别准确率有一定提升,但同时模型训练与测试所需的存储设备和计算资源要求也会成平方级地增加。因此,提高局部放电模式识别准确率、降低PRPD图谱存储分辨率对提高电力系统的运行可靠性、降低运维成本具有重要意义。
分别建立隐含层为1、2、3的反向传播神经网络,研究隐含层数对DGA数据识别准确率的影响,网络各层神经元数量如表2所示,识别结果如图10所示。
图10可知,三个网络的识别准确率介于82.35%~83.33%之间,其中N2网络表现最优。由于N1的隐含层数只有1层,神经元个数少,模型非线性程度低,无法对样本空间中不同缺陷类别数据进行有效区分,因此经过240次迭代后其识别准确率波动仍然较为明显;而N3网络参数规模较大,在本数据集中时较N2更易出现过拟合,模型识别准确率较低。因此,后文统一采用隐含层数为2层的反向传播神经网络作为DGA特征向量的分类模型。
由上文可知,在低输入维度下,基于PRPD图谱或DGA特征向量单一信息源的缺陷识别误诊率高,难以满足人工智能与传感技术日益发展的背景下电力系统对设备运行可靠性的要求,因此本文提出一种基于PRPD图谱和DGA数据信息融合的CNN-BPNN模型,两种模式识别方法融合可实现油纸绝缘沿面放电典型缺陷的电检测法与化学检测法的优势互补。
在PRPD图谱输入维度为32×32并融合DGA特征向量的条件下,选择一组缺陷2的D-S证据融合前后的BPA来说明所构建模型的纠错机制。通过D-S证据理论融合前后的概率计算结果如表3所示。由表3可知,根据DGA单一判据的分类结果是正确的,而根据PRPD图谱单一判据的分类结果出现错误。根据PRPD图谱单一判据模型被判为缺陷3的概率最大,为0.445 0,且判为缺陷2的概率也较大,为0.379 5。而根据DGA单一判据模型的识别结果判为缺陷2的概率最大,此时两种分类模型识别结果出现了分歧。通过D-S证据融合后,判为缺陷2的概率最大,为0.714 1,判为缺陷3的概率为0.190 6,输出结果满足决策的3个判决规则,同时不确定度降低至0.010 3。由此可知,当PRPD图谱单一判据模型在缺陷识别中出现误判时,加入DGA数据进行联合诊断能纠正基于PRPD图谱单一判据模型的错误,同时提高了信息融合模型的识别准确率和识别结果的置信程度。
为对比本文的信息融合方法在PRPD图谱不同输入维度下6类缺陷识别准确率的提升效果,将PRPD图谱在不同输入维度下的识别结果分别与DGA识别结果融合,不同PRPD图谱输入维度下三种基于不同判据的模型识别准确率如图11所示。由图11可知,D-S证据理论较好地融合了PRPD判据和DGA判据的优势,在不同PRPD图谱输入维度下,6类缺陷的识别准确率均有不同程度的提升。当PRPD图谱输入维度分别为8×8、16×16、32×32时,融入DGA特征向量的CNN-BPNN模型识别准确率为93.21%、97.53%、99.17%,较PRPD图谱单一判据的模型识别准确率分别提升了4.81%、2.78%、0.84%,同时节约了计算资源和存储空间,有利于在实际工程中应用。
为验证本文所提CNN-BPNN模型的优越性,将Vgg16、ResNet18、DenseNet121等深度卷积神经网络对PRPD图谱分类后,与DGA特征向量识别结果进行D-S证据融合,不同神经网络融合模型的识别准确率及模型大小如表4所示。由表4可知,CNN-BPNN模型的识别准确率较Vgg16-BPNN、ResNet18-BPNN、Dense Net121-BPNN等深度卷积神经网络融合模型分别提高了0.16%、0.87%、0.99%。由于6种缺陷的二维PRPD图谱具有一定规律性,特征较为显著,参数过多、结构复杂的深度卷积神经网络在训练过程中易出现过拟合,Vgg16-BPNN等深度卷积网络融合模型表现略逊于参数较少的CNN-BPNN模型。此外,CNN-BPNN模型更轻量,文件大小仅为0.69 MB,较深度卷积融合模型至少节约40倍的内存空间。
(1)当PRPD图谱或DGA特征向量的单一判据模型识别结果出现误判时,基于D-S证据理论的PRPD-DGA信息融合方法可有效纠正错误输出,降低分类结果的不确定度,进而增强了缺陷识别结果的置信程度。
(2)CNN-BPNN模型可有效融合局部放电的电气物理信息和化学产物信息,当PRPD图谱输入维度为8×8、16×16、32×32时,融入DGA特征向量的模型识别准确率分别为93.21%、97.53%、99.17%,较基于PRPD图谱单一判据的模型识别准确率分别提升了4.81%、2.78%、0.84%,既提高了缺陷识别准确率,又降低了数据存储要求。
(3)CNN-BPNN模型较Vgg16-BPNN、ResNet 18-BPNN、DenseNet121-BPNN等深度卷积神经网络融合模型识别准确率分别提高了0.16%、0.87%、0.99%,且网络参数大幅减少,抑制了模型的过拟合,并降低了模型的存储占用空间,更利于在电力系统智能化运维中应用。
  • 国家自然科学基金重点项目(52037009)
  • 国家重点研发项目(2021YFB2402102)
  • 国家电网科技项目(5108-202218280A-2-356-XG)
参考文献 引证文献
排序方式:
[1]
RAYMONDW J K, ILLIASH A, BAKARA H A, et al. Partial discharge classifications: Review of recent progress[J]. Measurement,2015,68:164-181.
[2]
孙长海,李天明,陈百通,等.基于概率神经网络的油纸绝缘老化诊断技术研究[J].绝缘材料,2021,54(6):107-115.
[3]
李岩,冯俊杰,卢毓欣,等.大容量远海风电柔性直流送出关键技术与展望[J].高电压技术,2022,48(9):3384-3393.
[4]
范路,陆云才,陶风波,等.人工智能在局部放电检测中的应用(二):模式识别与状态评估[J].绝缘材料,2021,54(7):10-24.
[5]
BASHARANV, SILUVAIRAJW I M, VELAYUTHAMM R. Recognition of multiple partial discharge patterns by multi-class support vector machine using fractal image processing technique[J]. IET Science, Measurement & Technology,2018,12(8):1031-1038.
[6]
许辰航,陈继明,刘伟楠,等.基于深度残差网络的GIS局部放电PRPD谱图模式识别[J].高电压技术,2022,48(3):1113-1123.
[7]
张聪聪,王刚,高栋,等.基于卷积网络的GIS局部放电缺陷诊断方法与应用[J].电工电能新技术,2021,40(3):72-80.
[8]
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.
[9]
杨为,朱太云,张国宝,等.电力物联网下基于卷积神经网络和迁移学习的GIS局部放电模式识别分类方法研究[J].高压电器,2020,56(9):20-25,32.
[10]
王涤,马爱军,归宇,等.基于P-CNN的局部放电绝缘故障融合诊断[J].高电压技术,2020,46(8):2897-2905.
[11]
黄亮,唐炬,凌超,等.基于多特征信息融合技术的局部放电模式识别研究[J].高电压技术,2015,41(3):947-955.
[12]
艾嘉伟,牛海清,陈泽铭,等.基于神经网络的D-S多信息融合GIL局部放电联合识别方法[J].高电压技术,2022,48(12):4925-4932.
[13]
耿伊雯,芮逸凡,范路,等.基于D-S证据多源信息融合与固态光电倍增-UHF联合检测的GIS局部放电模式识别[J].绝缘材料,2022,55(11):109-117.
[14]
王磊,张磊,牛荣泽,等.基于CNN+D-S证据理论的多维信息源局部放电故障识别[J].电力工程技术,2022,41(5):172-179.
[15]
魏振,孙昭昌,史蕾玚,等.复合电压比例对油纸绝缘沿面放电的影响[J].绝缘材料,2017,50(12):43-47.
[16]
REFFASA, BEROUALA, MOULAIH. Comparison of creeping discharges propagating over pressboard immersed in olive oil, mineral oil and other natural and synthetic ester liquids under DC voltage[J]. IEEE Transactions on Dielectrics and Electrical Insulation,2019,26(6):2019-2026.
[17]
张君颖,吴楠,王野,等.基于放电统计参数动态变化率的油浸芳纶绝缘纸局部放电发展过程研究[J].绝缘材料,2022,55(3):72-77.
[18]
International Electrotechnical Commission. High-voltage test techniques-Partial discharge measurements:IEC 60270:2015[S]. Geneva:IEC,2015.
[19]
魏振.复合电压下油纸绝缘局部放电发展过程及缺陷诊断方法[D].北京:华北电力大学,2015.
[20]
万晓琪,宋辉,罗林根,等.卷积神经网络在局部放电图像模式识别中的应用[J].电网技术,2019,43(6):2219-2226.
[21]
陈健宁,周远翔,白正,等.基于多通道卷积神经网络的油纸绝缘局部放电模式识别方法[J].高电压技术,2022,48(5):1705-1715.
[22]
陆云才,范路,陶风波,等.人工智能在局部放电检测中的应用(一):去噪与故障定位[J].绝缘材料,2021,54(5):10-20.
[23]
宋辉,代杰杰,张卫东,等.复杂数据源下基于深度卷积网络的局部放电模式识别[J].高电压技术,2018,44(11):3625-3633.
[24]
郭金明,田树军,卓浩泽.绝缘油老化特性及诊断方法研究进展[J].绝缘材料,2021,54(8):8-17.
[25]
FAIZJ, SOLEIMANIM. Dissolved gas analysis evaluation in electric power transformers using conventional methods a review[J]. IEEE Transactions on Dielectrics and Electrical Insulation,2017,24(2):1239-1248.
[26]
张育杰,冯健,李典阳,等.基于油色谱数据的变压器故障征兆新优选策略[J].电网技术,2021,45(8):3324-3332.
[27]
DOUZASG, BACAOF, LASTF. Improving imbalanced learning through a heuristic oversampling method based on k-means and SMOTE[J]. Information Sciences,2018,465:1-20.
[28]
KRIZHEVSKYA, SUTSKEVERI, HINTONG E. Imagenet classification with deep convolutional neural networks[J]. Communications of the ACM,2017,60(6):84-90.
[29]
RUMELHARTD E, HINTONG E, WILLIAMSR J. Learning representations by back-propagating errors[J]. Nature,1986,323(6088):533-536.
[30]
XUY, LIY, WANGY, et al. Integrated decision-making method for power transformer fault diagnosis via rough set and DS evidence theories[J]. IET Generation, Transmission & Distribution,2020,14(24):5774-5781.
2023年第56卷第12期
PDF下载
97
42
引用本文
BibTeX
文章信息
doi: 10.16790/j.cnki.1009-9239.im.2023.12.006
  • 接收时间:2023-02-21
  • 首发时间:2025-11-24
  • 出版时间:2023-12-20
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2023-02-21
  • 修回日期:2023-04-04
基金
国家自然科学基金重点项目(52037009)
国家重点研发项目(2021YFB2402102)
国家电网科技项目(5108-202218280A-2-356-XG)
作者信息
    1新疆大学 电气工程学院 电力系统及大型发电设备安全控制和仿真国家重点实验室风光储分室, 新疆 乌鲁木齐 830047
    2清华大学 电机工程与应用电子技术系 电力系统及发电设备控制和仿真国家重点实验室,北京 100084
参考文献
分享链接
https://castjournals.cast.org.cn/joweb/jycl/CN/10.16790/j.cnki.1009-9239.im.2023.12.006
分享至
全文二维码

扫描看全文

引用本文
BibTeX
本文的引用情况
2种不同金属材料的力学参数

Family
属数
Number of
genus
种数
Number of
species
占总种数比例
Percentage of
total species (%)

Genus
种数
Number of
species
占总种数比例
Percentage of total
species (%)
鹅膏菌科Amanitaceae 2 11 5.26 鹅膏菌属 Amanita 10 4.78
小菇科 Mycenaceae 2 12 5.74 丝盖伞属 Inocybe 5 2.39
多孔菌科 Polyporaceae 8 14 6.70 蜡蘑属 Laccaria 5 2.39
红菇科 Russulaceae 3 23 11.00 小皮伞属 Marasmius 6 2.87
小菇属 Mycena 11 5.26
光柄菇属 Pluteus 5 2.39
红菇属 Russula 17 8.13
栓菌属 Trametes 5 2.39
关闭全屏