Paper Contents
Abstract
This paper presents a real-time transformer fault detection system that leverages sensor fusion and machine learning to identify faults with high accuracy. The system integrates electrical measurementsincluding voltage, current, and temperaturewith mechanical vibration analysis using the MPU6050 sensor. A NodeMCU microcontroller interfaces with all sensors, and analog readings are routed through the MCP3008 ADC due to NodeMCUs single analog pin limitation. The collected data is transmitted to a Python backend, where a K-Nearest Neighbors (KNN) classifier is trained and applied to classify fault types such as overvoltage, undervoltage, short circuit, overheating, and abnormal vibration. Additionally, MATLAB Simulink is used to simulate vibration responses under various fault conditions, enhancing model robustness. The proposed system achieved 100% classification accuracy during testing and successfully identified faults within 23 seconds in a live hardware setup. The results confirm its feasibility as a low-cost, deployable solution for smart grid and substation monitoring.
Copyright
Copyright © 2025 Sonal Noel Raj R. This is an open access article distributed under the Creative Commons Attribution License.
