| Autor: |
M. Tomašević, M. Đorđević, Mileta Žarković, Zlatan Stojković, V. Shiljkut |
| Rok vydání: |
2021 |
| Předmět: |
|
| Zdroj: |
The 12th Mediterranean Conference on Power Generation, Transmission, Distribution and Energy Conversion (MEDPOWER 2020). |
| DOI: |
10.1049/icp.2021.1204 |
| Popis: |
This paper presents intelligent algorithms applied to power transformer monitoring data. Algorithms illustrate how to use big data obtained during the exploitation of power transformers in order to create optimal maintenance plan. The related database encompasses values obtained by measurements of polarization index, dielectric loss factor, idle current, short-circuit impedance, etc. The first algorithm applies unsupervised machine learning (UML) in order to classify these data and assigns power transformers to the groups with similar properties and probability of failure. The second algorithm uses artificial neural networks (ANN), as a part of supervised machine learning (SML), to assess the exploitation age of the power transformer based on the history of monitoring. The calculated exploitation age is compared then with the actual age and premature aging of the power transformer can be indicated. Both algorithms are trained on the data related to the power transformers installed in the electric power plants and the results are presented in the paper. The results of algorithms application are used to improve the maintenance of power transformers in Public Enterprise Electric Power Industry of Serbia. The current practice of periodical, Time Based Maintenance can be replaced with contemporary approaches, like Condition Based Maintenance (CBM) and Risk Based Maintenance (RBM). The results of engineering application demonstrate the effectiveness and superiority of the machine learning and improve the diagnostics and maintenance plan for the power transformers. |
| Databáze: |
OpenAIRE |
| Externí odkaz: |
|
