Personalised Thermal Comfort Model for Automatic Control of a Newly Developed Personalised Environmental Control System (PECS)

Abstract

Personalised Environmental Control Systems (PECS) are devices that cater to the individual needs by providing micro-climate heating, cooling, and ventilation. However, to ensure comfort, energy savings, and productivity, a comfort model based automatic control is required. For its development, thermal preference, physiological information, and data on the surrounding indoor climate were gathered from 24 subjects when using a newly developed PECS with heating, cooling, and ventilation functions. Since PECS should ensure a high level of comfort while providing energy savings through background temperature relaxation, multiple steady-state ambient temperature settings ranging from 18 to 28 °C were tested. The data were clustered according to the subject’s self-assessed general thermal preference, namely neutral, warmer, and colder. Machine learning was used to generate a cluster-based personalised comfort model using environmental, physiological, and behavioural indicators. The prediction performance of the models was 11 to 18 percent points higher than that of current group comfort models, predicted mean vote (PMV), which is independent of occupant similarities. The advantage of the personalised approach was the increased performance of the thermal comfort prediction at no expense of occupant sensitive information. Although reliant on estimates of physiological indicators, the models’ performance may be increased using real-time data acquisition.