Search

The goal of this work is to give a full review of how machine learning (ML) is used in thermal comfort studies, highlight the most recent techniques and findings, and lay out a plan for future research. Most of the researchers focus on developing models related to thermal comfort prediction. However, only a few works look at the current state of adaptive thermal comfort studies and the ways in which it could save energy. This study showed that using ML control schemas to make buildings more comfortable in terms of temperature could cut energy by more than 27%. Finally, this paper identifies the remaining difficulties in using ML in thermal comfort investigations, including data collection, thermal comfort indices, sample size, feature selection, model selection, and real-world application. [ABSTRACT FROM AUTHOR]

Monitoring the thermal comfort of building occupants is crucial for ensuring sustainable and efficient energy consumption in residential buildings. Existing studies have addressed the monitoring of thermal comfort through questionnaires and activities involving occupants. However, few studies have considered disabled people in the monitoring of thermal comfort, despite the potential for impairments to present thermal requirements that are significantly different from those of an occupant without a disability. Additionally, people with disabilities can experience difficulties in expressing their thermal comfort, which further complicates assessment and monitoring. To overcome this, we propose the development of a new learning model using a deep neural network. Our model can predict the indoor thermal comfort of differently abled people in real time to facilitate remote monitoring. We generated our real dataset using a new Internet of Things (IoT) architecture. Our architecture also includes a data collection scheme to ensure an efficient collection process, enabling the collection of targeted data before transferring them to cloud servers for further data analysis. Experimental results illustrate the reliability of our data collection scheme in gathering useful and targeted data, as well as the efficiency of our deep learning-based model, which achieved an accuracy of 94% and a precision and recall of 98% and 97%, respectively. • Studied thermal comfort sensations of disabled people based on Fanger's PMV model. • Focused on three types of disability: Physical, Intellectual, and Neurological. • Through IoT network, we collected a real dataset on thermal comfort in France. • Analysed thermal sensation variations between disabled and abled people. • Build a new prediction model of disabled people's thermal sensations using deep neural network. [ABSTRACT FROM AUTHOR]