LAUSR

Abstract It has been widely acknowledged that occupants are one of the major contributors to account for the performance gap between predicted vs. actual building energy consumption. So far, occupancy has been represented as an average value model (deterministic), Markov chain model, survival model, or data mining model. Recent studies by Ahn et al. and Ahn & Park [1, 2] suggest a concept — the “random walk” occupancy model — that in random-walk driven buildings, occupancy is unpredictable. In this study, the authors investigate whether such predictability of occupancy is influenced by temporal and spatial measurement resolutions. For this purpose, occupancy data were recorded in six rooms in a library building for 16 days. Normalized Cumulative Periodograms and the Bartlett test were used to examine the predictability of occupancy. It was found that the occupancy of a small group (in this study, occupants numbering 18 or fewer) is unpredictable and can be regarded as a random-walk driven space. In contrast, when the number of occupants is equal to or more than a certain number (in this study, 27 people) either in a single large space or aggregated over multiple spaces (up to the whole-building level), occupancy becomes predictable. Additionally, it is shown that the predictability of the time-series occupancy data is not significantly influenced by varying the temporal resolution (sampling interval) between 5 and 60 min, but is dominantly driven by the number of occupants.