Canadian residential occupancy schedules from four time-use cycles (2005–2022) and a hindcast-tested projection to 2025

Occupancy schedules for building performance simulation are usually fitted to one survey year and treated as fixed. This study builds household presence, activity and metabolic schedules for Canada from four General Social Survey time-use cycles (2005–2022) linked to the Census, and projects the household population to 2025 with a conditional variational autoencoder and a cluster-based drift model. In a hindcast against the held-out 2021 Census, the projection beat carrying the 2016 population forward on only 7 of 24 variables, missing the pre-set pass rule, so the 2025 cohort is used as a scenario, not a forecast. Households were at home 15.6–16.2 h per weekday in 2005–2015 and 17.9 h in 2022. Against a standard apartment schedule, the survey-based schedules raised simulated heating by 6–16% and lowered cooling by 1–9% in three Montreal house neighbourhoods, and changed both by under 7% in three apartment neighbourhoods.HighlightsFour Canadian time-use cycles (2005–2022) harmonized into household schedules.Presence was flat from 2005 to 2015 and rose in weekday daytime in 2022.A hindcast against the held-out 2021 Census tests the 2025 projection.Survey schedules raise house heating 6–16% and cut cooling 1–9% vs a default.Occupancy changes the size of the cooling peak, not its timing.