A Study of Building Thermal Dynamics from Large Data Sets: An Application for Residential Smart Thermostats

This thesis focuses on identifying the Thermal Time Constant (TTC), a thermal performance indicator related the building’s effective thermal insulation, airtightness, and thermal storage capacity. Using data from over 15,000 smart thermostats, data mining is applied to identify patterns in the short-term transient thermal response of Canadian and American dwellings. The data used consist of contextual information (i.e. metadata) and one year of measurements recorded at 5-minute intervals (i.e. indoor air temperature, outdoor air temperature, and Heating, Ventilation and Air-Conditioning (HVAC) equipment run times). The TTC is captured from the data by tracking the indoor temperature response of the free-running dwelling over a specific time period, and by also assuming this response can be accurately described by the characteristic exponential decay (or growth) of a first-order resistance-capacitance thermal model. Consequently, the results show significant differences between estimated TTC values for the summer and winter months across ASHRAE climate zones 1 through 7. In winter, the mean TTC related to these climate zones ranges from 7 to 47 hours. In contrast, the summer mean values vary between a lower and narrower range of 6 to 19 hours which can presumably be attributed to occupants opening the windows, and thus effectively reducing their dwelling’s overall thermal resistance. Towards the larger objectives of thermal resilience, energy savings and grid reliability, the estimated TTC values can be used in the residential sector to quickly identify buildings eligible for building enclosure retrofits or to rapidly generate a simple model to inform thermal load estimation and management