Prediction of indoor clothing insulation levels: A deep learning approach

Abstract

Clothing insulation is a key variable in the prediction of occupant thermal comfort. Consequently, the aim of the current study was to develop predictive models that forecast clothing insulation levels of building occupants. Using field measurements, we investigated the influence of outdoor environment factors and mode of transport on clothing insulation levels of university students. Our results showed that both the mode of transport and weather variables influenced the clothing insulation levels of the students. We then developed a deep neural network model that forecasts mean daily clothing insulation levels using outdoor air temperature at 6 am, dew point temperature at 6 am, gender, season and mode of transport in the based on the collected data from 1316 questionnaire surveys. In addition, we revealed that outdoor environment factors had stronger associations with clothing insulation levels than indoor environment elements. The developed deep neural network model indicated a high R² value of 0.90. In comparison to the deep neural network model, a developed linear model using the same data indicated a lower R² value of 0.698, which implies that the proposed deep neural network model provides an efficient method to forecast clothing insulation levels.