Center for Research in Energy Systems Transformation (CREST)

In the built environment, the paradigm is shifting from providing uniform environment for all occupants to accommodating individual preferences through decentralised comfort devices enabled by IoT (Internet of Things) and ubiquitous computing. The effective implementation of localised and personalised comfort-enhancing and energy-saving strategies depends critically on the ability to detect the occupant presence in the immediate local environment. In this paper, estimating individual user presence in a shared office-space room using plug-load meters monitoring the power consumption of users' desktop computers and PIR (passive infrared) sensors at every user's desk is investigated. By extracting informative features from the data obtained and simple k-means clustering analysis, best-case presence accuracies of 89-99% and absence accuracies of 87-96% are achieved as validated by comparing with the ground-truth data for 4 different users.