Fault Diagnosis for Oil-Immersed Power Transformers and On-Load Tap Changers Using Ensemble Learning

In power transformers, insulating oil is primarily used to insulate and cool windings and iron cores. When a transformer fails, its high temperature, discharge, and arc phenomenon will cause gas to be generated in the insulating oil. The condition inside the transformer can be analyzed and possible fault types can be diagnosed by analyzing the dissolved gas in the insulating oil. This study proposes a novel method for fault diagnosis based on dissolved gas analysis. The proposed method implements an intelligent fault diagnosis model using an ensemble learning method, which can be used for individual fault assessment of power transformers and on-load tap changers. In the design of the input data of the proposed fault diagnosis model, the percentages of five gases (H<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf>, CH4, C2H6, C2H4, C2H2) are selected. Finally, the normal and fault data from a power company and IEC TC10 document are used to train the proposed model. Subsequently, actual data from a technology company are used for testing. The test results verify the feasibility of the proposed fault diagnosis technology. The accuracy of transformer fault diagnosis exceeds 71 % (seven diagnostic types) and more than 82 % (three diagnostic types); the accuracy of on-load tap changer fault diagnosis achieves 75 % (three diagnostic types).