Assessing Adaptive Thermal Comfort Using Artificial Neural Networks in Naturally-Ventilated Buildings

This paper presents a method for predicting occupants’ indoor thermal sensation in naturally-ventilated environments, based on real thermal sensation samples, using a GA-BP neural network model. This method improves the traditional back propagation neural network by incorporating an integrated genetic algorithm into the BP neutral network which aims to optimise the connection weight or threshold of the parameters in the input layer of the GA-BP neutral network model, which represent the factors affecting adaptive thermal comfort. The model has been tested by comparing the results with the actual thermal sensation votes of occupants in field studies carried out in a residential building in the Yangtze River Region in China. The results indicate that the maximum deviation is 3.5%.