Data-Driven Decision Making for Energy Efficiency Improvement in Buildings

Buildings are the single largest user of energy in U.S. and can account for more than half of greenhouse gas (GHG) emissions in large cities. With concerns related to natural resource depletion, population growth, and climate change, many policymakers at all levels of the government have set GHG emission reduction targets. Nonetheless, these targets cannot be met without supporting energy efficiency in buildings as the building sector has one of the highest potentials for delivering significant and cost-effective GHG emission reduction. By benefiting from the growth in the city-scale energy benchmarking programs, this dissertation is proposing a framework to investigate the energy performance of buildings and how data-driven approaches can be used to motivate the building sector toward more sustainable use of energy. To do so, the first part of the dissertation discovers the trends and patterns in energy and water consumption in residential and commercial buildings across the U.S. over three decades. The main goal of this part is to leverage machine learning to cluster US states based on their similarities in terms of per capita water and energy consumption over time. The next part investigates the city-scale building energy benchmarking programs and data along with reviewing the challenges and opportunities associated with them by reviewing energy benchmarking programs nationwide. In this part, both qualitative and quantitative analysis of the benchmarking programs are performed. The next part of the dissertation proposes a framework to integrate the energy benchmarking data with other urban data categories and extracting the important attributes affecting building energy use. Toward this goal, the capability of various machine learning techniques, both to model energy use and also to interpret the results were evaluated. Finally, the last part investigates the feasibility of using benchmarking datasets rather than confined surveys in developing models for grading buildings on energy performance.