Applicability assessment and simplified calculation method of thermal indices in RayMan

Abstract Many thermal indices have been proposed for quantitative descriptions of human thermal perception over the past century to meet the strong demand in applied science. Meanwhile, developing the RayMan model provides a great interactive platform to calculate newly developed thermal indices. In this paper, we first used hourly meteorological observation data from national stations in Beijing from 2019 to 2021 to find the thermal indices applicable to the comfort evaluation of Beijing among the RayMan outputs. Using the UTCI and PT output from the RayMan model to evaluate the comfort level in Beijing, the results show that most of the time in winter, dominated by cold thermal perception, hot thermal perception appeared at an increased time in summer, and comfortable level dominated in spring and fall, which is consistent with seasonal changes of thermal perception in Beijing. Second, using the particular observation data from Daxing station in 2022, it is verified that RayMan outputs UTCI are consistent with the results calculated by the globe-thermometer method under most weather conditions. Finally, based on feature engineering and multi-model optimal parameter comparison, we used multi-layer perceptron for regression algorithm to build the UTCI and PT calculation models with reproduction errors less than or equal to 1°C of more than 93% and 95%, respectively. These models can calculate the thermal indices at different locations or time scales in Beijing, using station or grid meteorological observation or forecast data as inputs to meet different demands in applied science.