Using Machine Learning to Target Retrofits in Commercial Buildings under Alternative Climate Change Scenarios

Buildings account for 40% of the energy consumption and 31% of the CO2 emissions in the United States. While deep efficiency goals can be set for new construction, energy retrofits of existing buildings provide another effective means to reduce building consumption and carbon footprint. A key step in retrofit planning is to predict the effect of various potential retrofits on energy consumption. Decisionmakers currently look to simulation-based tools for detailed

assessments of a large range of retrofit options. However, simulations often require detailed data inputs, high expertise, and extensive computational power, presenting challenges when considering portfolios of buildings or evaluating large-scale policy proposals. Data-driven

methods offer an alternative approach to retrofit analysis based on statistical modeling of measured energy use in existing buildings, and are generally easier to implement, faster to run, and require less building systems expertise. If implemented properly, these methods could be applied to portfolio-wide retrofit plans and generate more realistic estimates due to the use of real-world data. However, current applications of data-driven approaches focus heavily on evaluating past retrofits, and predominantly report effects either very specifically (e.g., for a single building), or very generally (e.g., the average effect of a large set of buildings). These two extremes limit the ability to generalize these retrospective estimates to other buildings, thus

providing little decision support for future retrofits. This thesis uses data from a substantial portfolio of federal buildings and demonstrates a data-driven approach to generalize the heterogeneous treatment effect of past retrofits to future potential savings, assisting retrofit

planning by targeting buildings with high predicted savings. This dissertation also extends this model to estimate retrofit effects under alternative climate change mitigation scenarios, providing valuable information for long-term retrofit planning in the face of climate change.