A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Personal comfort model for automatic control of personal comfort systems
Personal comfort models could be used for the development of automatic controls for personal environmental comfort systems (PECS). These models often use indoor environment and physiological indicators as attributes for estimating the subjective response of occupants. Traditional indoor thermal environment research and standardization recommend 7-point scales for thermal comfort or thermal sensation estimation. However, many studies apply transformations to the response, thus oversimplifying the scales and generating controversy. The aim of this study is to determine the relevance of different indicators for the development of personal comfort models while investigating the implications and resulting model accuracy when using different thermal sensation scale discretization. Two simple machine learning algorithms, namely logistic regression and Naïve Bayes, were used in a multi-class setting to predict the overall thermal sensation of individual subjects when occupying a heated or cooled chair in steady state conditions. Multiple models were generated depending on the variables included in the feature set. Additionally, two response vectors were generated based on the thermal sensation vote, a three class and a seven class one, the latter being generated by further discretizing the hot and cold spectrum of thermal sensation. Both models performed better than a random guess at identifying thermal sensation classes and reached accuracies of up to 72% when predicting the overall thermal sensation of people using PECS. Including information of the PECS operation in the model, i.e. seat temperature, increased the prediction accuracy by up to 5%. The overall accuracy was higher when using three classes for the thermal response, as implementing seven classes led to a decrease of up to 21 percentage points. Nevertheless, the latter provided a finer adjustment without affecting the model’s ability to distinguish between the cold and hot spectrum, which may be an advantage for personal comfort systems that condition the ...
Personal comfort model for automatic control of personal comfort systems
Personal comfort models could be used for the development of automatic controls for personal environmental comfort systems (PECS). These models often use indoor environment and physiological indicators as attributes for estimating the subjective response of occupants. Traditional indoor thermal environment research and standardization recommend 7-point scales for thermal comfort or thermal sensation estimation. However, many studies apply transformations to the response, thus oversimplifying the scales and generating controversy. The aim of this study is to determine the relevance of different indicators for the development of personal comfort models while investigating the implications and resulting model accuracy when using different thermal sensation scale discretization. Two simple machine learning algorithms, namely logistic regression and Naïve Bayes, were used in a multi-class setting to predict the overall thermal sensation of individual subjects when occupying a heated or cooled chair in steady state conditions. Multiple models were generated depending on the variables included in the feature set. Additionally, two response vectors were generated based on the thermal sensation vote, a three class and a seven class one, the latter being generated by further discretizing the hot and cold spectrum of thermal sensation. Both models performed better than a random guess at identifying thermal sensation classes and reached accuracies of up to 72% when predicting the overall thermal sensation of people using PECS. Including information of the PECS operation in the model, i.e. seat temperature, increased the prediction accuracy by up to 5%. The overall accuracy was higher when using three classes for the thermal response, as implementing seven classes led to a decrease of up to 21 percentage points. Nevertheless, the latter provided a finer adjustment without affecting the model’s ability to distinguish between the cold and hot spectrum, which may be an advantage for personal comfort systems that condition the ...
Personal comfort model for automatic control of personal comfort systems
Bogatu, Dragos-Ioan (author) / Kazanci, Ongun B. (author) / Watanabe, Futa (author) / Kaneko, Yosuke (author) / Olesen, Bjarne W. (author) / Itard, Laure / Hensen-Centnerová, Lada / Boerstra, Atze / Bluyssen, Philomena / Hensen, Jan
2022-01-01
Bogatu , D-I , Kazanci , O B , Watanabe , F , Kaneko , Y & Olesen , B W 2022 , Personal comfort model for automatic control of personal comfort systems . in L Itard , L Hensen-Centnerová , A Boerstra , P Bluyssen , J Hensen , T Klein , M Loomans , P Pauwels , C Struck , M Tenpierik & B Geldermans (eds) , Proceedings of CLIMA 2022 Conference . TU Delft , pp. 41-48 , 14th REHVA HVAC World Congress , Rotterdam , Netherlands , 22/05/2022 . https://doi.org/10.34641/clima.2022.120
Article (Journal)
Electronic Resource
English
DDC:
690
Personal comfort systems: A review on comfort, energy, and economics
| Elsevier | 2020
Personal comfort systems: A review on comfort, energy, and economics
| Elsevier | 2020