Abstract
The wave of grid modernization and climate change is rapidly changing the landscape of residential energy demands. For example, hotter summers suggest increased use of A/C units, use of electric vehicles implies increased energy demands and use of rooftop solar indicates local generation. A central question thus is to understand how energy is consumed at granular social, spatial, and temporal resolutions. Such an understanding can lead to better solutions to demand-response events, study the diffusion process of solar adoption, predict household-level energy use, or analyze weather impacts. In order to answer these social impact questions, several ‘Modeling & Simulation’ solutions are appearing in the literature at a noteworthy rate. However, we observe some critical problems that still need to be addressed, especially in the areas of data quality, robust and scalable energy modeling infrastructure, and effective analysis tools for complex behavior simulations. Due to these drawbacks, many public policies and social impact questions requiring detailed data and knowledge of the domain remain unexplored. I address these research gaps in my dissertation to facilitate large-scale analytics, personalized (or detailed) energy policy recommendations and solve social impact questions.
First, I resolve the data & infrastructure problem by generating a digital twin of residential disaggregated energy use time series for U.S. households. In order to generate this large data (approx. 30TB), I have designed a scalable and extensible big-data pipeline infrastructure using a microservices-oriented architecture. To ensure the quality of the digital twin, this thesis contributes by proposing novel validation metrics for the household-level energy time series. In the second part of the dissertation, I propose the use of machine learning techniques and agent-based models for solving fairness and sustainability questions in residential energy in two topics: (a) fairness in residential dynamic pricing; (b) comparison of solar adoption models in rural and urban areas.
