Real-time prediction model of passenger thermal comfort for intelligent cabin

This article establishes a joint model of intelligent cabin and human thermal sensation, following experimental verification. And 12000 sets of datasets are obtained through joint model simulation under 120 random working conditions. Subsequently, a real-time prediction model of passenger thermal comfort is established and is trained using the Catboost algorithm based on the datasets. Finally, under various working conditions, the calculation results of the real-time prediction model were compared with these of the joint model and steady-state prediction model separately. The results indicate that the real-time prediction model can predict the passenger thermal comfort under unstable thermal environment, while the steady-state prediction model cannot. And compared to the accurate value calculated by joint model, the average mean absolute error (MAE) and mean absolute percent error (MAPE) of the real-time prediction model is 0.025 and 1.58 %, respectively. This indicates that the realtime prediction model established in this article has good accuracy.