Multivariate predictive window blind control models for intelligent building façade systems

This paper presents results from a window blind usage field study that was conducted in California, USA. In this study, the measurements of physical environmental conditions were cross-linked with participants’ window blind controlling preferences (n=83). A total of seven predictive window blind control multivariate logistic models were derived. As hypothesized, the probability of a window blind closing event increased as the magnitude of physical environmental and confounding factors increased (p < .01). The main predictors were window/ background luminance level and vertical solar radiation at the window. The confounding factors included MRT, direct solar penetration, and participants’ self-reported sensitivity to brightness.  The results showed that the models correctly predict between 84 – 89 % of the observed window blind control behavior. This research extends the knowledge of how and why building occupants manually control window blinds in private offices, and provides results that can be directly implemented in energy simulation programs.