Enhanced Circle Method for Discriminating Incipient Faults in Power Transformers Using Dissolved Gas Analysis

Power transformers are essential components in electrical power systems, ensuring the continuous transmission and distribution of electricity. Dissolved Gas Analysis (DGA) has become a widely accepted method for monitoring the condition of transformers and detecting incipient faults. Among the various DGA methods, graphical techniques have gained significant popularity in recent years. In this study, an Improved Circle Method is proposed as an enhanced extension of the Circle 1 and Circle 2 methods to address the limitations of traditional DGA methods in distinguishing overlapping fault types. The method employs a sigmoid activation function to control the convergence and divergence distances between fault samples, significantly improving the separation of electrical and thermal faults. By utilizing Gray Wolf Optimization (GWO), optimal angles are determined for the best fault separation. Experimental validation using the IEC TC10 database demonstrated that the Improved Circle Method achieves an overall accuracy of 79.49%, which can be considered an improvement over both Circle 1 and Circle 2 methods. The results confirm that the proposed method serves as a reliable alternative to other conventional graphical techniques.