Personalized Thermal Comfort System Development: High Accuracy Skin Temperature to Thermal Sensation Prediction Model

Abstract With the increasing diversification of driving environments, the thermal conditions inside the occupant cabin have become more complex and variable. To accurately predict human thermal sensation, it is crucial to establish a thermal sensation prediction model that is closely related to human physiological responses. This study investigates human perceptual responses to different thermal environments. Through thermal sensation experiments in an environmental chamber and questionnaire surveys, the effects of physiological responses in various thermal environments on the subjective thermal sensation of Chinese male drivers were assessed and compared. The study focuses on the environmental warming process, collecting local skin temperature data from participants and their corresponding overall thermal sensation voting values. Four machine learning algorithms, artificial neural networks, support vector regression, random forests, and Gaussian process regression, were used to establish correlation models between local skin temperature and overall thermal sensation. After comparison, it was found that the support vector regression model performed the best, achieving a prediction accuracy with R2 > 0.9. The established model provides theoretical support for optimizing occupant cabin thermal environments and intelligent control.