Harnessing Building Thermal Storage with Time-Varying Emission Signals: An Emission-Responsive Path to Lower Carbon

Abstract

Building electrification is central to decarbonizing cold-climate regions, yet it concentrates electricity demand in winter hours when solar generation is minimal and marginal grid emissions are high. This dissertation evaluates whether buildings can function as distributed thermal storage resources that shift electrified heating demand away from carbon-intensive hours without compromising indoor comfort, and identifies the physical design conditions under which this is possible. A multi-scale framework links four stages: fit-for-purpose EnergyPlus calibration of HouseZero®, a high-performance retrofitted building; real-world field validation of a transparent, rule-based control strategy driven by forecast marginal emissions; stock-scale scenario analysis of the Suffolk County, Massachusetts residential and commercial building stock using ResStock and ComStock; and machine-learning analysis of the design determinants of comfort-constrained demand-response capability. Threshold screening identified a 95th-percentile forecast marginal-emissions trigger as the preferred operating point. Field experiments, including an extreme-cold stress test at -7.5 °C, showed that HouseZero® repeatedly reduced heat-pump load by approximately 95% while median zone-air temperature dropped by no more than 0.5 °C, yielding measurable rebound-adjusted emissions reductions. At the urban scale, the modeled standard heat-pump pathway with electric backup reduced site energy but increased grid-side marginal emissions by roughly 37% and created a winter-dominant electric peak, whereas a high-performance geothermal-and-envelope pathway reduced marginal emissions by about 10% relative to the current baseline. Demand response reduced electricity-related emissions in every pathway, yet comfort-constrained screening revealed that the deployable participant pool is far smaller than the technical potential, expanding substantially with added thermal mass. Machine-learning analysis showed that physical design variables explain approximately half of the weighted variation in response capability, with thermal mass emerging as the most robust design lever, alongside envelope exposure, airtightness, and other envelope characteristics. The dissertation concludes that buildings can serve as grid-interactive thermal batteries, but their usable grid value depends on integrating electrification, envelope performance, thermal mass, and rebound-aware control into a cohesive system rather than as independent interventions.