Date of Award

8-22-2025

Date Published

September 2025

Degree Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Mechanical and Aerospace Engineering

Advisor(s)

Bing Dong

Subject Categories

Engineering | Mechanical Engineering

Abstract

It is estimated that about 40% of all the electricity generated in the United States is consumed by the building sector. This scale of energy use by buildings can be understood by the fact that the U.S. alone consumes approximately 20% of all the energy produced globally, despite having only about 4.23% of the world's population. Given the threats posed by climate change, it is necessary to curtail the energy demand of this sector. Additionally, global climate change has increased the frequency, intensity, and duration of extreme climate events like heat waves, cold waves, hurricanes, floods, and droughts. Given the growing risk of more frequent extreme climate events, it is essential to retrofit older, high energy-consuming buildings to improve their resilience. Apart from building resilience, it's crucial to consider the stochastic nature of Occupant Behavior (OB), especially under future weather conditions. OB significantly influences building energy consumption but is often overlooked in simulations. This study examines the changes in building energy consumption, occupant behavior, indoor air quality, and thermal comfort in residential dormitories before and after deep energy retrofits. Significant reductions in Total Volatile Organic Compound (TVOC) exposure were observed following the installation of Heat Recovery Ventilation (HRV) units. The frequency and duration of natural ventilation usage also decreased, particularly during late summer and transitional seasons, due to improved indoor thermal conditions. Post-retrofit indoor thermodynamic conditions consistently fell within the comfort boundaries as defined by ASHRAE Standard 55. Additionally, HVAC energy use dropped by nearly 80%, driven by better envelope insulation, significant infiltration reduction, and the adoption of Air Source Heat Pumps (ASHPs). To extend these insights to the entire State of NY, retrofit outcomes from the campus dormitory were scaled to residential dormitories across New York using a Python-based framework. This framework sampled relevant building models from the Model America database, a large-scale energy model archive developed by Oak Ridge National Laboratory, encompassing over 5.1 million New York State buildings. The first part of the framework identified dormitories using the information available in NYS Department of Education. The second part enabled automated modification of EnergyPlus models, adjusting parameters such as insulation levels and infiltration rates to reflect retrofit scenarios. Future climate simulations were based on RCP8.5 projections sourced from ORNL’s repository, which included 20 years of weather data—ten representing near-future conditions (2045–2054) and ten representing long-term futures (2085–2094). Comprehensive EnergyPlus simulations for the year 2094 showed that infiltration control remained the most impactful retrofit strategy, reducing energy use by 11.05%, while insulation upgrades achieved a modest reduction of 3.11%. Furthermore, by leveraging metadata from the 5.1 million NYS Building Energy Models (BEMs), the study identified building models closely matching the geometry and characteristics of the original dormitories. This allowed the integration of machine learning–based occupant behavior meta-models into the statewide energy simulation. These meta-models were developed using transfer learning from original models and were designed to generalize occupant behavior patterns across different genders and demographic groups. The neural network-based meta-model combined multiple trained networks into a unified model capable of recognizing complex occupant behavior patterns. Lastly, the research examined the compression cycles of ASHPs under future climate scenarios and confirmed their continued suitability for small- to mid-sized residential buildings (less than 10,000 ft²) in the future. The results highlight ASHPs as a viable, future-ready technology for decarbonizing space heating in New York State homes under long-term climate change. In short, this research explored the following three important questions: 1. How can AI-based method help to quantify OB and assist in retrofit decision making? 2. How will OB change with future climate changes and what are the impacts on retrofit decisions? 3. How can we scale our methods developed using the data from our university dorms of the entire New York State?

Access

Open Access

Download

Share

COinS