An Event-Based Approach for Non-Intrusive Load Monitoring in Industrial Power Systems

Non-Intrusive Load Monitoring (NILM) techniques aim to identify the operating states of individual electrical loads by analyzing aggregate measurements of voltage and current. While NILM has shown promising results in residential contexts, its application in industrial settings presents unique challenges due to the complexity of load profiles, higher power levels, and increased noise. This paper introduces an event-based NILM system designed specifically for industrial environments. The system operates on the edge and combines high-resolution signal acquisition, event detection through RMS analysis, and load classification using features extracted around switching transients. A lightweight training strategy is adopted, based on single-load activation scenarios, to ensure practical deployment without requiring extensive labeled data. The proposed method is validated using real measurements collected from diverse industrial machines under realistic operating conditions.