Intelligent Transformer Fault Identification Using GALO-FIS and Refined Ratio Techniques

Transformer fault identification using Dissolved Gas Analysis (DGA) is crucial for maintaining the consistency and effectiveness of electrical grids. The Rogers' four-ratios technique (RFRT) and IEC 60599 ratio code technique (IECRCT) are well-established DGA-based techniques for detecting transformer faults. However, their diagnostic accuracy is less accurate, and the fact that ratio techniques cannot tell the difference between some produced gases makes them less helpful in detecting different fault types. Therefore, this study introduces an optimized detection platform, which integrates a genetic algorithm with learning-based optimization for a fuzzy inference system (GALO-FIS). The proposed approach has been evaluated using a dataset of 705 collected samples. The results demonstrate that the newly established ratio boundaries and modified rules enhance the diagnostic accuracy of GALO-FIS RFRT and GALO-FIS IECRCT. The proposed GALO-FIS RFRT enhances accuracy from 45.86% to 92.57% compared to traditional RFRT, while GALO-FIS IECRCT achieves 91.58% accuracy, significantly improving upon the 40.23% of conventional IECRCT. These results suggest that the proposed GALO-FIS platform provides a more robust and precise tool for transformer fault detection. The recommended technique is extensively modelled using MATLAB/Simulink software and verified using DGA data collected in real-world scenarios.