Advancing thermal comfort prediction: Improvement of the Berkeley local thermal sensation model and overall thermal sensation prediction by machine learning

The in-vehicle driving environment is typically complex and dynamic. To enhance the comfort of both drivers and passengers, accurately predicting the thermal sensation within the vehicle is crucial. This lays the foundation for future research aimed at optimizing air conditioning control systems. The present study involved 108 male participants from China, who were studied through subjective thermal sensation assessments. The research explored the discrepancies between the thermal sensation ratings reported by participants and the predictions made by the Berkeley thermal sensation model. By incorporating data from human thermal neutrality tests, the skin temperature set point in the model was modified, and the coefficient representing human thermal sensitivity was refined using the Particle Swarm Optimization (PSO) algorithm. This led to the development of a model capable of accurately predicting the local thermal sensation of Chinese male participants. Furthermore, through machine learning techniques, the relationship between local and overall thermal sensation was established. Among the four machine learning algorithms evaluated, Support Vector Regression (SVR) demonstrated the highest effectiveness, achieving excellent accuracy.