Abstract:

To guarantee dependability and operating efficiency, power transformers must have improved fault detection. Many fault detection methods rely on conventional dissolved gas analysis (DGA) techniques. Recently, a strong statistical approach called Linear Discriminant Analysis (LDA) has been developed to improve fault classification. By emphasizing differences across classes and decreasing variance within them, LDA improves the separation of fault types. Hydrogen, Methane, Ethane, Ethylene, and Acetylene concentrations are used to categorize transformer defects in this research, which utilizes LDA. A total of 83.64% accuracy was achieved during training and 81.1% accuracy during testing using the MATLAB classification learner tool and the gathered datasets. The competitive performance of LDA is shown by a comparison with different DGA approaches. The results show that LDA can be used to diagnose transformer faults, and there's room for improvement by creating hybrid models that use machine learning to make predictions with more precision..

Description:

.