Combining KNN Algorithm with Traditional DGA Techniques for Transformer Defect Identification and Analysis

This paper explores the application of KNN for identifying and analyzing dissolved gases in power transformer (PT) oil. We want to make a strong diagnostic model that can correctly classify different kinds of PT defects by combining traditional dissolved gas analysis (DGA) methods with the KNN algorithm. To do this, traditional techniques like Doernenburg ratio techniques (DRT), International Electrotechnical Commission ratio techniques (IECRT), and Duval triangle techniques (DTT) are used to prepare the DGA data. These ratios are then used as input features for the KNN model. The main objective of this study is to evaluate the effectiveness of the KNN-based model in diagnosing PT defects and to compare its performance with traditional techniques. Key metrics such as accuracy, F1 score, specificity, precision, and recall will be used to evaluate the model’s effectiveness. Furthermore, the study will explore the impact of different input features on model performance to identify the optimal feature set for accurate defect diagnosis.