Verification of stochastic behavioural models of occupants' interactions with windows in residential buildings

Valentina Fabi, Rune Korsholm Andersen, Stefano Corgnati

    Research output: Contribution to journalJournal articleResearchpeer-review


    Realistic characterisation of occupants' window opening behaviour is crucial for reliable prediction of building performance by means of building energy performance simulations. Window opening behaviour has been investigated by several researchers, leading to a variety of logistic regression models expressing the probability with which actions will be performed. But only very few attempts have been made to investigate the reliability of the models. In this paper, data from a measurement campaign in 15 apartments was used to estimate the predictive accuracy of four sets of models of window opening. Initially three models from literature were investigated by comparison of predicted probabilities and the actual measured state of the windows. Data from one of the papers was reanalysed to create new models, based on measurements from single dwellings. These models were used to predict window transition probabilities using data from the field survey. The output was then compared to the measured transitions. Results showed that the models which most accurately predicted both the state of the window (open or closed) and the number of actions on windows had certain characteristics in common: A positive correlation between the probability of opening and CO2 concentration and illumination values and a negative correlation with sun hours and illumination level for closing windows.
    Original languageEnglish
    JournalBuilding and Environment
    Issue number1
    Pages (from-to)371-383
    Publication statusPublished - 2015


    • Behavioural verification
    • Occupant behaviour
    • Statistical modelling
    • Window opening


    Dive into the research topics of 'Verification of stochastic behavioural models of occupants' interactions with windows in residential buildings'. Together they form a unique fingerprint.

    Cite this