Ngerem, Elvis Onyedikachi and Covenant University, Theses (2021) TRANSFORMER FAILURE PREDICTION IN DISTRIBUTION POWER SYSTEM NETWORK USING ARTIFICIAL NEURAL NETWORK. Masters thesis, Covenant University Ota..
![[img]](http://eprints.covenantuniversity.edu.ng/style/images/fileicons/application_pdf.png) |
PDF
Download (146kB) |
Abstract
Distribution Transformers (DTs) are critical components of the power distribution network, and their reliability is heavily reliant on them, necessitating the need for a thorough and precise maintenance process focused on the prediction of impending transformer faults. Distribution transformers are subjected to both internal and external faults as a result of the electrical, mechanical and thermal stresses they are subjected to during their operation, resulting in irregularities in some of their cooling and insulating materials, such as oil and cellulose insulation degradation. These lead to overheating, partial discharge, or corona, causing a degradation of the oil insulation and the introduction of dissolved gasses such as hydrogens (H2), carbon monoxide (CO2), carbon dioxide (CO2), methane (CH4), acetylene (C2H2), ethane (C2H6), and ethylene (C2H4). In this project, an Artificial Neural Network (ANN) model in python programming for the diagnosis of incipient faults of distribution transformers, using the DGA (Dissolved Gas Analysis) concentrations, was developed. Also, four machine learning models are developed for fault prediction from dissolved gas analysis data in distribution transformers using interpretation results from Roger’s ratio, IEC basic ratio, and Dornenburg Ratio on the basis of the IEEE C57.104 standard. The models developed are K-nearest neighbors, linear support vector classifier, logistic regression, and multilayer perceptron. These models were trained, tested, and evaluated to determine the best performing model. The identified best-performing model was implemented on a web based interacting application interface. A transformer fault prediction algorithm trained with back-propagation resulted in an improved accuracy of 97.93%. This will aid maintenance engineers for a faster and accurate diagnostic decision on faulty distribution transformers as it plays a significant role in power supply to large energy customers.
| Item Type: | Thesis (Masters) |
|---|---|
| Uncontrolled Keywords: | Artificial Neural Network, Dissolved Gas Analysis, Distribution transformers, transformer failure, fault prediction |
| Subjects: | T Technology > TK Electrical engineering. Electronics Nuclear engineering |
| Divisions: | Faculty of Engineering, Science and Mathematics > School of Electronics and Computer Science |
| Depositing User: | Mrs Patricia Nwokealisi |
| Date Deposited: | 28 Oct 2021 13:51 |
| Last Modified: | 28 Oct 2021 13:51 |
| URI: | http://eprints.covenantuniversity.edu.ng/id/eprint/15419 |
Actions (login required)
 |
View Item |
