Electricity is an essential energy source for both daily life and industrial activities. The electricity generated at power plants is transmitted at ultra-high or high voltage, then gradually reduced to lower voltages through power transformers before being delivered to customers. In the electricity distribution system, transformers play a critical role in both transmission and voltage transformation. Early detection of internal faults in transformers is important to prevent sudden power outages and maintain a stable energy supply. Dissolved Gas Analysis (DGA) is one of the most widely used diagnostic tools for detecting and evaluating faults in oil-filled transformers. This research focuses on classifying power transformer operating conditions, specifically distinguishing between normal and faulty states, through the analysis of DGA data. Due to the rarity of fault events, the dataset is highly imbalanced, posing challenges for traditional machine learning and deep learning methods. To address this issue, an anomaly detection approach using autoencoders is employed.
VUONG et al. (Wed,) studied this question.