It was found from monitored data from eight dwellings in a case study building in Quebec City (Canada) that there are clear differences in the window opening behavior between different households. This paper aims to develop from data a probabilistic window opening model that accounts for occupant behavior. Logit regression is employed to predict the state (opened/closed) of windows according to indoor and outdoor temperatures environmental and temporal parameters. To replicate the diversity of behavior, normal distribution functions applied to the logit regression coefficients are used so that simulated occupants respond differently to environmental stimuli. It was found that the model offers good prediction for the monitoring by only using the outdoor and indoor temperatures as predictors. The proposed methodology was tested by simulating 10,000 times a full validation year of the case study building and comparing the results with measured data. The agreement was good. The model overestimated slightly the total frequency of window opening in the dwellings and the number of window changes-of-state. A vast range of window opening behavior was generated by the model, showing its ability to reproduce both the aggregated window opening behavior and the diversity of behaviors of the case study building.
Probabilistic window opening model considering occupant behavior diversity: A data-driven case study of Canadian residential buildings
Article de Jean Rouleau et Louis Gosselin