A Practical Low-Cost NILM Device Based on Tiny Machine Learning

Abstract

This study addresses the growing importance of Non-Intrusive Load Monitoring (NILM) in enhancing energy efficiency in load consumption monitoring. The objective of this research is to develop an integrated system that utilizes NILM, combining TinyML and IoT technologies, for real-time monitoring and control of household devices. This approach leverages the efficiency of TinyML for on-device processing while enabling seamless connectivity and data management through IoT. We employed a Random Forest machine learning model alongside the ESP32 MCU to achieve this goal. Key findings indicate that the system can classify various load types with high accuracy and minimal latency, demonstrating effective performance in real-world conditions. The implications of this study suggest that NILM can significantly improve user engagement in energy management while offering a cost-effective solution for load consumption monitoring.