Occupancy determination based on time series of CO2 concentration, temperature and relative humidity

Occupancy receives considerable attention from building designers and managers. In spite of great importance, its determination is still challenging. Traditional methods have various limitations. Therefore, alternative approaches are developed. The goal of this study is a method which provides information about the total number of occupants inside intermittently occupied space. It is based on the continuous measurements of selected parameters of indoor air and multiclass classification. It was shown that the number of occupants may be successfully determined using time series of CO2 concentration (MCR = 1.6%, ME = 0.45 person), temperature (MCR = 1.6%, ME = 0.48 person) and relative humidity (MCR = 3.3%, ME = 0.79 person) individually. The data record applied for this purpose should include measurement results from 60 min long period. When using sensor array, this time may be shortened without performance loss. Particularly useful is the combination of temperature and humidity sensors. The analysis revealed that the choice of appropriate classifier is very important. Nonparametric, nonlinear, minimum distance classifier (k-NN) was very effective, while parametric classifier, utilizing linear discriminant functions (LDA) was unsuccessful. The presented method may be also used for determining the duration of occupancy periods.