Abstract
Non-Intrusive Load Monitoring (NILM) has been an active area of research for the past two decades, with the goal of inferring the consumption of individual electrical devices in a building from mains meter data. Despite substantial progress, the potential of facilitating disaggregation by incorporating additional sensor data has received limited attention. In this paper, we investigate the benefits of integrating water consumption data into eventless NILM models and compare its impact with that of reactive power. Our experiments show that appliances directly associated with water usage benefit from the inclusion of water data; however, the overall gains from reactive power are larger. The best performance is achieved by combining reactive power and water data as additional input channels alongside active power, resulting in reductions of 28% in mean absolute error and 63% in signal aggregate error in the geometric mean across all circuits.
Justus Breyer, M.Sc.
Researcher
Paul Sonnenschein
Prof. Dr. Muhammad Hamad Alizai
LUMS, Lahore, Pakistan
