A systematic review of personal thermal comfort models

Personal comfort models have shown to predict specific thermal comfort requirements more accurately than aggregate models, increasing occupant acceptability and associated energy benefits in both shared and singleoccupant built environment. Although advances in the field of personal thermal comfort models are undeniable, there is still a lack of thorough and critical reviews of the current state of research in this field, especially considering the details of the predictive modeling process involved. This study has systematically reviewed 37 papers from over 100 academic publications on personal comfort models from the last two decades, and examined: (1) the data collection approach and dataset size, (2) number and type of participants involved, (3) climate, seasons and type of building involved, (4) model input and output variables, (5) modeling algorithm used, (6) performance indicator used, and (7) model final application. The review has identified a lack of diversity in building types, climates zones, seasons and participants involved in developing personal comfort models. It has also highlighted a lack of a unified and systematic framework for modeling development and evaluation, which currently hinders comparisons between studies. With most of the studies using machine learning techniques, the review has pointed to the challenges of black box models in the field. Finally, the review has indicated that personal input features using physiological sensing technologies can be further explored, especially considering the rapid advances seen today in wearable sensor technologies.

Automated summary

Personal comfort models have been shown to predict specific thermal comfort requirements more accurately than aggregate models, but there is a lack of diversity in building types, climate zones, seasons and participants involved in developing these models. There is also a lack of a unified and systematic framework for modeling development and evaluation, highlighting the challenges of black box models in the field.