Disaggregating Building-Level Occupancy into Zone-Level Occupant Counts Using Sensor Fusion

This paper outlines a methodology for disaggregating building-level occupancy data into zone-level occupant counts using opportunistic data from Wi-Fi access points, motion detectors, and CO2 sensors via sensor fusion for a floor of an academic office building. The efficacy of different combinations of data for this purpose is explored and the occupant-count estimates from these different combinations of data are compared to one another. The impacts of different sensors, sensor grid densities, and their data on the occupant-count estimates are discussed. Historical CO2 data are analysed to determine if instances of under- or over-occupancy as estimated by this methodology are reflected in the measured CO2 trends where available. The results indicate that additional sensor data improves the disaggregation model’s ability to estimate which rooms are over-occupied, with the granularity of the Wi-Fi access point grid and inclusion of CO2 data causing the largest increase in purported accuracy.