Using synthetic population data for prospective modeling of occupant behavior during design

Abstract

This paper addresses the challenge of incorporating occupant behavior into building performance simulation models used during the design process-that is, before the actual occupants are known. It proposes the use of synthetic population data, an approach that is novel in building performance modeling although common in urban planning and public health. A simpler approach embodied in the ASHRAE Fundamentals volume is to report standard distributions of values for behavioral variables, assuming that parameters vary independently of one another when in fact many co-vary or are interdependent. An alternative approach calibrates models of occupant behavior against actual occupants in specific existing buildings, but this raises questions of transferability. Needed is a database of "generic" occupants that designers can use prospectively during the design process. This paper documents a process of combining disparate field studies of commercial buildings into a larger occupant behavior database and generating a statistically similar synthetic data set that can be shared without compromising confidentiality requirements associated with field studies. The synthetic data set successfully incorporates much of the covariance structure of the underlying field data and supports multivariate modeling. Its scope and structure necessarily serve the needs of the associated modeling framework. Cooperative and systematic sharing of data by field researchers is crucial for building large enough data sets to serve as a behaviorally-robust basis for building design.

An ontology to represent synthetic building occupant characteristics and behavior

“…Data on the lighting condition usually varies from the amount of natural and artificial lights to the amount of glare, also, from the two identifiable sources. Controls over lighting fixtures often include the adjustment of window blinds, illuminance intensity on task light, and overhead light [36]. The rise of machine learning and its application also contributes to the development of better lighting controls.…”

“…IAQ is also often related to occupant's lifestyles and activities in the indoor environment. These include pet ownership, smoking, the presence of plants, fragrances, and insecticides [36]. CO2 is usually used as a proxy of IAQ in the built environment to determine ventilation needs.…”

“…Several occupant archetypes were developed to understand behavior to reduce electricity consumption and maintain comfort [42], [65]. Personality traits and subjective values also determine occupant capacity to share controls [5], [36].…”

Investigating the Impact of Actual and Modeled Occupant Behavior Information Input to Building Performance Simulation

“…This, on one hand, is inevitable as building simulation models were usually used during the design process, and therefore the actual occupant behaviors are unknown. Andrews et al [36] generated synthetic population data with the combination of three data sets from different sources in field surveys and stated that the combined data inherits characteristics of all the field data sets which satisfied the needs of the associated modeling framework. However, on the other hand, using actual occupant behavior information of an existing building in a BEM may help optimize operation and control for the building, and yield more thorough and holistic understanding of building occupants for future building design.…”

Generating synthetic occupants for use in building performance simulation

“…Finally, methods for generating synthetic populations based on empirical data at certain geographical locations have been developed in fields of demography, urban planning and transportation. Despite various applications, there are, however, no found studies or applications of synthetic occupant population applied to building peeerformance simulation as suggested in (Andrews et al, 2016). There are several comprehensive reviews on various methods for population modeling for applications in other fields (Barthelemy & Toint, 2013;Chapuis & Taillandier, 2019;E.…”