Computationally efficient upscaling of Markovian occupancy models for urban-level building simulations

Despite recent progress in the field of urban building energy modelling (UBEM), integration of occupant behaviour models is still an area in need of development. Although abundant for building energy modelling (BEM), stochastic occupancy models tailored for urban-level simulations are lacking, as well as systematic analyses of accuracy versus complexity. Occupancy models aimed for BEM typically introduce a heavy computational load, but a small reduction in accuracy might allow a large increase in speed. This paper proposes methods for scaling up stochastic Markov-chain (MC) occupancy models previously developed for BEM to fit UBEM, and compares them in terms of accuracy and computational efficiency. For the scales normally involved in UBEM (tens to hundreds of occupants per zone and hourly time resolution), a model based on multinomial sampling is proposed that is several hundred times faster than the original MC model and only slightly overestimates the variability in the occupancy patterns.