DGA Based Ensemble Learning Approach for Power Transformer Fault Diagnosis

Abstract

The power transformer is one of the most ubiquitous and crucial parts of the energy infrastructure. The use of Dissolved Gas Analysis (DGA) to clarify transformer incipient faults via machine learning algorithms is an intriguing engineering strategy. In the interest of discovering more about the fault classification capacity and suitability of multiple Machine learning algorithms, this article makes use of a wide range of numerous and diverse DGA data sets. This research focuses on detecting faults in power transformers by analyzing gases that are dissolved in mineral oil insulation using Machine-Learning algorithms such as the K-nearest neighbors (KNN) classifier, Logistic Regression, Naive Bayes classifier, Decision Tree Classifier, and Ensemble learning algorithm. This research also addresses performance indicators and assesses multiple algorithms to validate the best class algorithms. In addition, a top-performing algorithm is chosen using a collection of effectiveness criteria. This method will be useful for condition monitoring engineers mostly in the diagnosis of transformer insulation, the implementation of monitoring devices for large transformer fleets, and the comprehension of the behavior of insulation oil over the course of years to prevent catastrophic failure.