Physics‐Informed Artificial Intelligence Architectures for Carbon‐Neutral Integrated Energy Systems

The transition toward carbon‐neutral integrated energy systems requires intelligent architectures capable of managing renewable intermittency, multi‐energy sector coupling, storage coordination, hydrogen integration, and dynamic carbon accounting while preserving stability and reliability. Although artificial intelligence has improved renewable forecasting, load prediction, and optimal dispatch, many existing approaches remain purely data‐driven and lack embedded physical constraints, limiting robustness under extreme operating conditions and reducing extrapolation capability. Carbon‐aware optimization and hydrogen‐integrated scheduling also often remain disconnected from stability modeling and system dynamics, creating fragmentation across planning, control, and sustainability assessment. Based on a structured review of recent studies, this paper identifies five essential operational layers for physics‐informed intelligent energy management: real‐time observability, physics‐based stability modeling, constraint‐embedded learning, AI optimization and control, and carbon intelligence. It then proposes a unified multi‐layer physics‐informed AI architecture for carbon‐neutral integrated energy systems. The framework integrates system monitoring, physical feasibility, adaptive learning, reinforcement‐learning‐driven control, and carbon‐intensity intelligence within a coherent operational paradigm. By linking energy system physics, advanced machine learning, and sustainability‐driven optimization, this review provides a structured roadmap toward resilient, scalable, and carbon‐aware intelligent energy infrastructures.