Stochastic occupancy profiles for Urban Building Energy Modelling (UBEM): a case study in northwestern Italy

Occupancy and occupant-related inputs in Building Performance Simulation (BPS) conceal an intrinsic stochastic and heterogeneous nature. Thus, they need extensive data to be modelled and assessed when the goal is their realistic representation. These factors significantly contribute to the challenge of accurate building modelling. Currently, the most common approach to model occupants in BPS relies on the adoption of a single static and deterministic set of user schedules, suggested by standards or codes (e.g., ASHRAE 90.1). When dealing with Urban Building Energy Modelling (UBEM) this translates into the assignment of the same identical schedules to all buildings of the same use type and, therefore, in homogeneous yet unrealistic load curve predictions for the modelled area. While this approach may be suitable for aggregated time resolution and large spatial scales, various urban applications (e.g., demand-response management, district heating and cooling systems dimensioning, renewable energy community design, etc.), may require a more detailed and realistic representation of how people move and act in the urban environment.

In this context, the present study leverages open-source data to create stochastic occupancy profiles tailored for northwestern Italy. The last available Italian Time Use Survey (TUS) data are clustered to identify recurring occupancy patterns across a wide range of urban spaces which are then used to inform a stochastic model using first-order, time-inhomogeneous Markov Chains to generate multiple individual urban occupancy profiles. Initially characterized by multi-location patterns, these profiles can be disaggregated by occupation place and statistically assigned to buildings based on their use type for UBEM simulations. To test their applicability and potential, the obtained occupancy profiles are integrated into the energy simulation of a neighbourhood in Milan using Urban Modeling Interface (umi). Finally, the differences in energy outcomes between simulations using the new stochastic occupancy profiles and those using static and deterministic ASHRAE profiles are investigated to quantify the impact of varying occupancy models on simulation results.

The presented methodology enhances the characterization of building occupancy for both BPS and UBEM. In UBEM applications, it allows for more varied and realistic schedules per building based on use type and population characteristics. Moreover, the methodology provides insights into people’s presence across a wide range of building use types and urban spaces (e.g., parks, transportation facilities, etc.) allowing the definition of occupancy patterns suitable for mixed-use district simulations while paving the way for future integration of mobility models in UBEM framework.