An Integrated AI-Driven Methodology for Synthesizing a Digital Similar of Residential PV Adoption

Residential rooftop solar energy adoption is crucial for achieving energy independence amid rising energy prices. However, factors contributing to residential solar adoption and its impact on adoption spread, as well as the spatio-temporal variations in energy generation, are not well understood due to the lack of photovoltaic (PV) data at a finer resolution (e.g., household, hourly levels). In this paper, we discuss a novel data-driven methodology to generate a highly granular, residential-scale realistic dataset for residential rooftop solar adoption. The data-driven approach has two main components: ( i ) integrated machine learning (ML) models to identify PV adopters and ( ii ) explainable artificial intelligence (AI) techniques to extract insights about key features and their interactions as learned by the trained ML models. The resulting synthetic dataset is validated against real-world data and can serve for modeling various downstream tasks. Our explainable AI-informed cluster analysis uncovered distinct patterns of solar adoption. For example, in Washington State, 85.3% of solar adopters cluster in a single group, indicating a shared motivation, with income being a significant contributor. In contrast, the Commonwealth of Virginia and the state of Louisiana exhibit a wider range of solar adopter profiles, which presents challenge to the development of specific policies in those states. Furthermore, we conducted two case studies using the synthetic dataset to demonstrate its value: ( i ) generating household-level hourly energy profiles for adopters through an analytical model and ( ii ) exploring the spread of PV adoption through a policy-based study that considers individual influences, peer effects, and spatial elements using a simulation-based technology adoption model. The policy-based case study in Virginia demonstrated that the 30% federal solar investment tax credit significantly increased rooftop solar adoption, particularly in low- and moderate-income communities. The outcome of this work has major implications for various stakeholders, guiding policy decisions and enhancing regulatory frameworks for distributed energy resources.