Dynamic Equipment Prioritization and Automated Three-Phase Power Switching Using IoT and Machine Learning in Critical Infrastructure

Abstract

Reliable power supply is critical in hospitals, industrial plants, and data centers, where outages can result in life-threatening or costly disruptions. Most existing automatic transfer switch (ATS) systems rely on fixed rule-based logic, lacking predictive intelligence or equipment prioritization. This study proposes an IoT-enabled, machine learning–driven framework for dynamic equipment prioritization and automated three-phase power switching. Real-time data—including voltage, current, temperature, and vibration—are captured through IoT sensors and analyzed using Decision Trees, Random Forests, and Neural Networks to predict faults and assess equipment criticality. A modular object-oriented architecture allows easy integration of new data sources and algorithms, ensuring scalability. The decision engine allocates power from grid, generator, and solar sources based on equipment priority scores and live load conditions, guaranteeing uninterrupted supply to mission-critical assets. Experimental validation in healthcare and industrial testbeds achieved 98% accuracy in prioritization, 100% correctness in switching, and reduced downtime by 40% compared with conventional ATS systems. The proposed approach advances resilient power management by coupling predictive maintenance with context-aware switching, improving reliability, efficiency, and operational safety in mission-critical environments.