A comparison of alternative occupant classification approaches for the modelling of window opening behaviour in office buildings

In the past 20 years, better representation of occupants’ window operation in building performance simulation has received great attention, and several useful window opening behaviour models have been developed. Beyond these, this paper describes the development of window opening behaviour models based on alternative occupant classification approaches, namely, modelling occupants’ window operation actions as a whole; modelling actions based on sub-groups (i.e. gender, floor level, etc.); and modelling window actions of groups based on the observed propensity to operate windows (tendency to leave open, closed, etc.). The paper examines the benefits of more specifically modelling occupants’ action versus modelling actions more generally, in terms of predication accuracy. A comparison between predictive performances reveals that modelling occupants’ behaviour based on their observed personal preference helps improve the accuracy of the model predictions, when compared with traditional approaches, but requires a greater degree of knowledge about personal preferences of a building’s occupants.