Abstract

At least 65% of the existing residential building stock will still be in use by 2050, thus retrofitting existing buildings will be critical to reduce energy consumption. Prioritizing these retrofits typically requires thorough evaluation of the envelope’s thermal performance, and the traditional methods to undergo such evaluation (e.g. energy audits) can be cost prohibitive, especially if it aims to cover hundreds or thousands of buildings. To this end, this study presents a novel data-driven approach to investigate the thermal performance of existing buildings using data collected from smart thermostats. The study focused on more than 60,000 houses across North America and relied on real-time indoor and outdoor temperature measurements at 5-minute intervals over a period of four years. Two grey-box modelling approaches namely, least-squares fitting of 1) decay curves, and 2) numerically integrated thermal energy balance equations were used to estimate a thermal time constant for each house. This time constant represented the time it takes for a house to achieve a new thermal equilibrium in response to changes in its internal and external thermal conditions. The resulting time constant values from both models were used to estimate lower and upper bound effective R-values for the entire envelope of each house. These results were also analysed with respect to ASHRAE climate zones, building-age, building-style, and floor-area. Finally, a classification model was developed to identify the time constant range for houses based on their attributes. The classification model indicated that floor area and ASHRAE climate zone were the most influential factors on time constant values obtained using both methods. By using a large sample size covering thousands of buildings nationwide, results of this research can be used to prioritize retrofits for existing buildings and can provide inputs for urban-scale energy simulations.