Energy Disaggregation with NILM on a Raspberry Pi with Smart-Metering Extension

Non-intrusive load monitoring (NILM) is emerging as a crucial technique for providing detailed and effective energy feedback, thereby facilitating the development of low-cost and low-energy management systems in the residential sector. However, achieving acceptable disaggregation results requires sampling frequencies that exceed the 15-minute intervals of commercial smart meters. State-of-the-art approaches require a measurement frequency of at least 1Hz. While technically possible, measurement devices that can provide measurements at such intervals increase the cost of the overall system. In addition, consumers may raise privacy concerns regarding these systems. To address these issues, we propose a low-cost single-device smart meter that provides direct feedback based on a local processing of the user's data. The proposed system leverages the Raspberry Pi and the YoMoPie Monitor to provide an efficient, compact, and accurate system. The system's performance was tested in a laboratory setting under two different scenarios, and promising results were obtained considering the disaggregation performance and computational complexity.