<![CDATA[This study aims to develop an intelligent fault prediction model for a 25 MVA power transformer using the Adaptive Neuro-Fuzzy Inference System (ANFIS), to improve classification accuracy and ensure selective, reliable protection decisions in power systems. The research is grounded in the limitations of traditional differential relay protection, which struggles to distinguish between internal and external faults during transient conditions. ANFIS combines fuzzy logic’s ability to handle uncertainty with the adaptive learning of neural networks, making it a suitable tool for non-linear fault data analysis. A simulation was carried out in MATLAB/Simulink. Current signals from CTs on both primary and secondary sides served as input features. The model was trained using 270 data samples and tested with 30 samples. Two membership functions generalized bell-shaped (Gbell) and triangular (Tri) were evaluated. RMSE was used as the performance metric. The ANFIS model with Gbell MF yielded a lower RMSE (0.0116) compared to Tri MF (0.0445), indicating better prediction accuracy and stability. The system consistently identified internal faults (output 1), and external faults (output 0) based on a 0.5 decision threshold. The findings validate the potential of ANFIS for integration into digital relay systems, enhancing real-time transformer protection through adaptive learning.]]>
Copyrights © 2026
