Identification and classification of incipient faults of power transformers are critical issues. There are several testing and diagnostic methods to resolve these issues. Dissolved gas analysis (DGA) is one of the best methods for detecting incipient faults in power transformers. High ambiguity in DGA data and appropriate feature selection are the major challenges of this method. To overcome these challenges, 15 significant features have been extracted from the data and four machine learning (ML) classifiers namely complex decision tree (CDT), weighted K-nearest neighbor (W-KNN), ensemble bagged trees (EBT), and artificial neural networks (ANN) have been trained. The models are tested on the three DGA datasets including the benchmark dataset, IEC TC-10. The classifiers identified faults with high accuracy on all datasets and it is validated with 96.5% average accuracy on the standard dataset IEC TC-10.
Machine Learning Based Incipient Fault Identification of Power Transformer Through Dissolved Gas Analysis
31.07.2024
353536 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Grey-Extension Method for Incipient Fault Forecasting of Oil-Immersed Power Transformer
Online Contents | 2004
|A New Method for Fault Detection and Identification of Incipient Faults in Power Transformers
Online Contents | 2008
|