Deep meta-learning for personal thermal comfort modeling in office buildings

Existing heating, ventilation, and air conditioning (HVAC) systems in office buildings typically operate with little feedback from occupants, leading to energy waste and low thermal comfort satisfaction. Since there is large thermal preference variance among groups of people, it is necessary to develop personal comfort models. With the development of Internet of things and machine learning technologies, personal thermal comfort modeling faces new opportunities. To be specific, based on the acquired human physiological data, thermal sensation feedback data, environmental data, and advanced machine learning algorithms, personal thermal comfort models with high accuracies can be developed. Although some progress has been made in existing works, the developed personal comfort models require a large number of data samples for training. To overcome the drawbacks, we develop a personal thermal comfort modeling method based on deep meta-learning, which can implement fast adaption to new environments using small data samples. Simulation results based on real-world data show the effectiveness of the proposed method, which is helpful for thermal comfort control and energy saving in office buildings.