Data-Driven Insights of Thermal Comfort Behavior in Large Populations

Presenter: Kunind Sharma

Research Category: Engineering and Technology
Student Type: Graduate
Additional Authors: Kane Michael
PI: Asst. Prof. Michael B. Kane
Award Winner Category: Engineering and Technology, COE Solution Awards

Thermal comfort drives ~40% of residential energy use, with the thermostat as the primary point of interaction. Yet, studies modeling thermostat use behavior have been derived from relatively small data and surveys. This project developed a population-scale data-driven approach to validating and identifying these mental models using Ecobee’s “Donate your Data” Dataset of ~75,000 households from all over the world. Within this dataset, single occupancy buildings are used to study individual dynamics of discomfort (i.e., thresholds and delays to thermostat changes) resulting in automation overrides and wasted energy. Time of day, week, and year; indoor and outdoor temperatures and humidity; and location are found to be drivers of thermostat use behavior. Guided by these insights of various behavior patterns for operating a thermostat, analytical behavior models are developed to improve thermostat control algorithms, demand response value, and user satisfaction.