Leveraging campus-scale Wi-Fi data for activity-based occupant modeling in urban energy applications

The widespread availability of open datasets in urban areas is transforming how urban energy systems are planned, simulated, and visualized. Urban energy models, however, require an understanding of urban dwellers, as their activities create the demands for energy in buildings. In this paper, we explore using campus-scale Wi-Fi data to identify typical occupant activity patterns as an input to an agent-based model of building occupants at the district scale. The data is taken from a Singaporean university's Wi-Fi network at high resolution. Each record comprises a timestamp, a device identifier, the location of the device within the campus, and the access point to which it is connected. The Wi-Fi dataset contains 120 different buildings on campus and 10,300 anonymized individual devices. Activities are then assigned to each location on campus according to the building use type. In order to test the methodology, the activity plans of 27,604 undergraduate students, 8,304 graduate students, and 12,018 employees were simulated over a workweek. The results show the model's ability to produce plausible activity plans but could be improved by implementing sampling rules and expanding the source dataset to include off-peak dates. Nevertheless, using such an agent-based modeling approach at the district scale appears to be a promising methodology to assess the impacts of different planning strategies on occupant behavior and district energy demand.