A Semi-Supervised Low-Complexity Appliance Identification and Monitoring System for Residential Houses

Non-intrusive load monitoring (NILM) has emerged as a vital element of modern energy monitoring and management approaches. Artificial intelligence methodologies have led to the creation of numerous solutions over the last twenty years. However, a significant challenge remains the need for a large labeled dataset, which interferes with the development of accurate and simple NILM solutions. Moreover, integrating consumer feedback is essential for recognizing and including new appliances in the system. This study introduces a semi-supervised, real-time NILM system aimed at overcoming these issues, with consumer feedback being crucial for improving system accuracy. Practical applications demonstrate the efficacy of this method, emphasizing its ability to reduce energy use and improve household energy management. The paper finishes with a discourse on practical applications and prospective development potential for the suggested system.