Rahman Azis Prasojo, Muhammad Akmal A. Putra, Hafiz Furqoni Wisam Azizi, Muhammad Fahmi Hakim, Indra Kurniawan, Suwarno
This study presents a novel approach for interpreting fault identifications in power transformers, leveraging a multi method approaches integrated with various conventional Dissolved Gas Analysis (DGA) techniques. Power transformers, crucial for electric distribution and transmission systems, rely heavily on their insulation systems, susceptible to ageing under thermal and electrical stressors. The conventional DGA methods—IEC Ratio Method, Rogers' Ratio Method, Duval Pentagon Method, and Duval Triangle Method—have limitations in fault detection accuracy, often leading to errors in fault interpretation. Addressing these limitations, this research introduces a combined technique incorporating a Naïve Bayes model that utilizes the majority vote and scoring index methodologies from DGA. The model was validated using a dataset containing 343 real transformer failure instances. The experimental findings demonstrate the superiority of the Multi-Method Naïve Bayes model over the individual and other multi-method DGA techniques. It achieves a significant accuracy improvement, registering an 85.7% success rate in fault detection. Moreover, the study explores the feasibility of an interface application using the Multi-Method model for assisting asset management in interpreting DGA data. This proposed model provides a more reliable method for integrating various fault detection techniques, mitigating conventional method limitations and reducing errors in fault interpretation. © 2024, School of Electrical Engineering and Informatics. All rights reserved.
Department of Electrical Engineering, Politeknik Negeri Malang, Malang, Indonesia; UIT JBM, PLN Indonesia, Sidoarjo, Indonesia; School of Electrical Engineering and Informatics, Institut Teknologi Bandung, Bandung, Indonesia