REVI-twin: an integrated AI-driven methodology for creating digital twin of residential electric vehicle infrastructure

Integrating electric vehicles (EVs) into homes and the electrical grid creates complex dynamics that traditional planning tools struggle to address. Accurately estimating residential EV charging demand typically requires resource-intensive agent-based simulations reliant on substantial input data, limiting scalability. We present REVI-Twin, an AI-driven digital twin of residential EV infrastructure that scales without computationally-intensive simulations. It encompasses: ( i ) household-level EV ownership; ( i i ) user behavior and charging preferences; ( i i i ) hourly power consumption; and ( i v ) planned trips. Our framework performs two tasks: ( i ) predicts EV adoption using transfer learning, semi-supervised learning, and Bayesian optimization; ( i i ) synthesizes hourly consumption with active-learning multi-output Gaussian processes from < 1% of data. We also release a comprehensive hourly integrated residential energy dataset. Our case study indicates that each 1% of battery adoption reduces Virginia’s net imports by ~ 0.06% daily and ~ 0.08% during peak hours. REVI-Twin assists policymakers and planners in analyzing adoption and infrastructure needs for resilient electrification.