Intelligent Diagnosis of Transformer Faults Using DGA

Power transformers are the most crucial components of an electrical power system, and their reliable operation is essential for maintaining a continuous power supply. The conventional IEC three-ratio method has been used for the incipient faults detection in power transformers by analyzing dissolved gas concentrations based on DGA results. However, accurate fault classification based on the conventional IEC method can be found challenging when the measured gas ratio is slightly diverged from the crisp boundaries of ranges assigned by this method. The paper proposes a novel approach to overcome the drawbacks by integrating artificial neural networks (ANNs) and classifiers with DGA data. The effectiveness and robustness of the proposed approach are confirmed by the simulation results of both techniques, which show that the proposed method is able to provide high prediction and reliable diagnosis for the identification of transformer faults.