AI-enabled operation of power grids under high penetration of distributed energy resources: a survey

The rapid integration of distributed energy resources (DERs), including photovoltaics, battery storage, electric vehicles, and flexible loads, is reshaping power grid dynamics in ways that conventional control architectures cannot fully manage. This paper surveys AI methods for real-time power grid operation across four domains: stability and real-time control, fault management and resilience, operational forecasting and energy management, and DER coordination. A taxonomy of relevant methodologies, supervised learning, deep learning, reinforcement learning, multi-agent systems, and explainable AI, is presented with a structured review of reported performance. A central contribution is a systematic comparison of AI deployment requirements in isolated versus interconnected grids, showing that grid context is a first-order design variable affecting control authority, stability margins, and validation requirements. Five cross-cutting deployment barriers, safety certification, explainability, cybersecurity, standardisation, and data accessibility, are assessed against current technology readiness levels. Emerging paradigms including large language models, foundation models for grid copilot applications, and AI agents for autonomous EMS/SCADA integration are reviewed as the frontier of this transformation. Research priorities are identified, with emphasis on isolated-grid validation, safe reinforcement learning, and hybrid AI–conventional control integration.