{"title":"Ageing Condition Assessment of Transformer Insulation in Visual Domain Using DCNN","authors":"Aniket Vatsa, Ananda Shankar Hati, V. Khadkikar","doi":"10.1109/PEDES56012.2022.10080633","DOIUrl":null,"url":null,"abstract":"Identifying ageing characteristics is essential for mitigating the transformer's catastrophic failure. However, the traditional methods for assessing the effects of ageing, such as the degree of polarisation measurement of kraft paper, are destructive. Transformers' ageing characteristics can be extracted from dielectric response analysis using frequency domain spectroscopy's low-frequency band. However, measuring frequency domain spectroscopy (FDS) in low-frequency regions is time-consuming. This research developed a unique deep convolutional neural network (DCNN) based ageing state identification approach by converting FDS into a visual domain using Markov Transition Field (MTF) to ascertain the hidden ageing attributes. Thus enhancing the ageing features and automatic feature extraction is utilised for ageing state diagnosis. The proposed MTF-DCNN method has a 96.60% diagnosis accuracy across five ageing categories. It also provides an automatic feature extraction-based network for a detailed evaluation of the transformer's insulation health.","PeriodicalId":161541,"journal":{"name":"2022 IEEE International Conference on Power Electronics, Drives and Energy Systems (PEDES)","volume":"708 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2022-12-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2022 IEEE International Conference on Power Electronics, Drives and Energy Systems (PEDES)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/PEDES56012.2022.10080633","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
Identifying ageing characteristics is essential for mitigating the transformer's catastrophic failure. However, the traditional methods for assessing the effects of ageing, such as the degree of polarisation measurement of kraft paper, are destructive. Transformers' ageing characteristics can be extracted from dielectric response analysis using frequency domain spectroscopy's low-frequency band. However, measuring frequency domain spectroscopy (FDS) in low-frequency regions is time-consuming. This research developed a unique deep convolutional neural network (DCNN) based ageing state identification approach by converting FDS into a visual domain using Markov Transition Field (MTF) to ascertain the hidden ageing attributes. Thus enhancing the ageing features and automatic feature extraction is utilised for ageing state diagnosis. The proposed MTF-DCNN method has a 96.60% diagnosis accuracy across five ageing categories. It also provides an automatic feature extraction-based network for a detailed evaluation of the transformer's insulation health.