Transformer Protection and Fault Detection through Relay Automation and Machine Learning

Transformers play a crucial role in modern power systems by enabling efficient voltage transformation and energy distribution across transmission and distribution networks. Their continuous operation and protection are vital to maintain grid reliability and economic stability. However, conventional relay-based protection schemes depend on predetermined thresholds that cannot adapt to variations in load, harmonics, and fault dynamics, often resulting in false tripping or delayed isolation. To address these limitations, this study proposes an intelligent transformer protection framework that integrates relay automation with machine learning (ML) algorithms for real-time fault detection, classification, and isolation. The proposed model utilizes high-resolution current and voltage waveforms to extract transient features that are analyzed using Support Vector Machine (SVM) and Convolutional Neural Network (CNN) algorithms. These models accurately identify internal, external, and incipient faults while discriminating between inrush and non-fault conditions. The automation layer dynamically adjusts relay settings through IEC 61850-based communication protocols, ensuring rapid and adaptive response. Simulation results in MATLAB/Simulink show a fault classification accuracy exceeding 98%, with reduced detection latency and minimal false alarms. The research demonstrates that combining data-driven ML analytics with relay automation significantly enhances the precision, speed, and resilience of transformer protection systems. This hybrid approach provides a scalable foundation for next-generation smart grid protection and predictive maintenance frameworks.