A study on the prediction of thermal comfort for residuals using LSTM deep learning model

Abstract

Recently, the demand for thermal comfort of occupants in the building is increasing, and various studies to improve comfort are being actively conducted, but the studies predicted using time-series data are insufficient. Therefore, this study used time-series data to control temperature changes in an indoor space through an LSTM model to measure and predict the occupant’s body temperature and humidity data, and temperature and humidity data of individual and common spaces. Actual thermal comfort values and predicted values were compared and analyzed through various indicators, and the performance of the model was evaluated. As a result, when the actual thermal comfort is expressed from -3 to +3 on a 7-point scale, the MSE is within the range of 0.13, MAE is within the range of 0.25 to 0.28, and RMSE is within the range of 0.36 to 0.37. Therefore, these results were analyzed that the LSTM model can predict the thermal comfort of occupants. © International Journal of Sustainable Building Technology and Urban Development.