Towards non-intrusive and high accuracy prediction of personal thermal comfort using a few sensitive physiological parameters

Abstract

The existing thermal comfort models that pursue high personal thermal comfort prediction accuracy inevitably cause disturbances because they require data measured from multiple parts of the human body. To address this issue, we proposed the personal thermal comfort prediction method, which realized a high level of prediction performance using no more than 3 physiological parameters (the skin temperatures of the wrist, the neck and the temperature of the point 2 mm above the wrist) by means of an artificial neural network (ANN). This method compares the performance results of the models with different combinations of the measured parameters and determines the optimal personal comfort model (PCM). Human subject experiments were conducted under different ambient conditions, during which physiological parameters and thermal sensation surveys were recorded. The mean skin temperature was also recorded, and its predictive performance was compared with that of the optimal PCM. The results show that the average prediction accuracy of our PCMs reaches 89.2%. The accuracy of the optimal PCM using the individualized combinations of the 3 physiological parameters as input variables is close to or even superior to that of the traditional models using mean skin temperature. The method presented in this paper provides accurate predictions for personal thermal sensation and only requires no more than two monitoring sites.