Self-Learning Framework for Personalised Thermal Comfort Model

This paper presents a novel self-learning framework for building personalised thermal comfort model. The framework is built with the understanding that each occupant has a unique thermal comfort preference. Current thermal comfort models focus on analysing average data for groups of people in different types of building, rather than considering individual thermal preference. We argue that building a personal level comfort model using learning algorithms may provide the basis to represent personalised dynamic thermal demands. By bringing more personal interest and data, the ground-up personalised model may help us better understand the internal links of personal factors from psychology, physiology and behavioural aspects. Furthermore, we developed an Smart Thermal Comfort (STC) environment sensors and mobile application to efficiently collect distributed personal data and make it open-sourced for other researchers to use. The aim of this paper is to rethink current comfort studies to standardize the methods in modelling personalised thermal comfort. By summarising the past five years' papers on personal thermal comfort model, this paper critically evaluates the methods used for personal data collection and learning algorithms. Finally, we conclude an Personal Thermal Comfort (PTC) framework including distributed personal measurement tools and machine learning algorithm for personalised thermal comfort study.