UC Berkeley

Nearly 60% of global energy consumption in buildings is used for space heating and cooling to provide occupant comfort. Yet, a large portion of occupants are dissatisfied with the buildings’ thermal environment. There are many reasons for thermal dissatisfaction in buildings, but a fundamental cause is the current practice of delivering uniform thermal conditions based on universal rules, without accounting for individual differences in comfort requirements. To address these issues, a growing body of research has emerged to better reflect individual’ comfort requirements. This dissertation contributes to this research by providing the following primary innovations: 1) Internet-connected personal comfort system (PCS) and 2) personal comfort models that can help to deliver personalized comfort experiences in occupied spaces. In particular, I developed and field-tested the new capabilities of PCS (data reporting, wireless connectivity) that could support individualized learning and coordinated controls with other building systems. I also proposed a new framework for thermal comfort modeling – personal comfort models that can predict individuals’ thermal comfort, instead of the average response of a large population, using Internet of Things and machine learning. As a practical use case, I developed a set of personal comfort models using the PCS field study data to demonstrate how the proposed framework can be implemented. The results showed that personal comfort models produced superior accuracy over conventional comfort models (PMV, adaptive) and that PCS heating and cooling control behavior was a strong predictor of individuals’ thermal preference and could be used as an individualized comfort feedback for HVAC controls. The results of this dissertation showed a synergistic effect between PCS and personal comfort models that could enable occupant-centric comfort management in buildings.