Personalized Thermal Comfort Modeling And Its Application To Energy Efficient Control Of HVAC Systems

Heating Ventilation and Air Conditioning(HVAC)systems are designed to provide a comfortable indoor thermal environment for its occupants.Conventionally,Predicted Mean Vote(PMV)model is used to represent thermal comfort which is only an average model and it cannot reflect the individual differences in thermal sensation.Existing thermal comfort standards like PMV model use only an average model and recommend very tighter specifications for comfortable temperature.However,studies conducted around the world show that individuals have different thermal sensation of the same thermal environment and can feel comfortable in much broader range of temperature than that recommended by the existing thermal comfort standards.HVAC systems mechanically cool or heat buildings in order to maintain a thermally comfortable environment for building occupants.Achieving this requires a handsome amount of energy to be consumed by HVAC systems to maintain the large difference between uncomfortable outdoor temperature and comfortable indoor temperature.We designed an Android application to collect feedback from occupants about their thermal sensation of the indoor environment.Meanwhile,environmental(temperature and relative humidity)data is collected through sensors installed in the room.Feedback data and corresponding environmental data is also stored to a database on a central personal computer(PC).The central PC uses our proposed modeling algorithm to recommend operating settings for the HVAC system.Based on the basic concept of thermal comfort and the brief principle of HVAC system,in this study,we propose a personalized thermal comfort model based on a machine learning classification algorithm called Support Vector Classification(SVC).The feedback data-set along with corresponding temperature and relative humidity is used to train the personalized thermal comfort model.The model's prediction accuracy has been verified by comparison between the predicted thermal sensation based on the proposed model and actual thermal sensation feedback.Accuracy of the proposed thermal comfort model shows that the proposed model has a potential to save energy as well as capturing the individual differences in thermal comfort.Unlike existing thermal comfort models,our proposed model can not only provide a systematic way to capture the individual's thermal comfort requirements,but also automatically adapt to any drift in the thermal sensation in an online fashion.Theproposed model is updated every time there is some new feedback from the occupant.The proposed model is easily integrated into the existing HVAC control system with minimal intrusion.Energy consumption simulations are performed using temperature set-points suggested by the existing PMV model and our proposed SVC model.A comparison of energy simulation results shows that our proposed model has a great potential for energy saving and a significant contribution to the future of HVAC applications.