AI-powered vehicle-to-home energy management for grid outage response: A pathway to policy-ready energy resilience

The energy transition requires robust resilience strategies to address increasing risks from extreme weather events and grid instabilities. Electric vehicles (EVs), through vehicle-to-home (V2H) functionality, provide a promising solution to enhance household-level power reliability during outages. This study develops an Artificial Intelligence (AI)-driven energy management framework that dynamically coordinates rooftop photovoltaic (PV) generation, stationary energy storage systems (ESS), and V2H-enabled EVs using a Double Deep Q-Network (DDQN) algorithm, with the explicit goal of ensuring that household energy demand is continuously met even during outages. A probabilistic approach is employed to generate outage profiles, which are integrated with real-world household consumption and generation data. The novelty of this work lies in integrating DDQN with a resilience-oriented strategy, allowing the system to make robust decisions under both normal and extreme operating conditions while ensuring continuous demand satisfaction through adaptive and autonomous scheduling. Simulation results show that the proposed strategy achieved 100 % PV self-consumption at multiple time steps, consistently maintained zero energy-not-supplied (ENS) during outages, and reduced execution time to 2.5 s compared with 41.6 s for the deterministic baseline. These findings highlight the potential of AI-based energy management systems to support policy-aligned, cost-efficient, and resilient operation of future residential energy systems.