Artificial intelligence and machine learning for smart grids: from foundational paradigms to emerging technologies with digital twin and large language model-driven intelligence
Yaser Mike Banad, Safura Sharifi, Zahra Rezaei
Abstract
• Comprehensive review of AI/ML applications in smart grids and power systems. • Bibliometric analysis of 123 studies maps forecasting, control, and security clusters. • Examines frontier paradigms: Digital Twins, Federated Learning, LLMs, and GenAI. • Identifies gaps in interoperability, privacy, scalability, and adversarial robustness. • Proposes hybrid DT–LLM frameworks for resilient, sustainable energy intelligence. The evolution of modern power systems into smart grids is increasingly powered by Artificial Intelligence (AI) and Machine Learning (ML), which provide effective solutions for managing renewable intermittency, dynamic demand, and cybersecurity challenges. This paper presents a comprehensive review of AI/ML applications in smart grids, tracing their development from foundational paradigms to cutting-edge technologies such as Federated Learning (FL), Generative AI (GenAI), Large Language Models (LLMs), the Artificial Intelligence of Things (AIoT), and Digital Twin (DT)-driven intelligence. Enabling infrastructures, including IoT, 5G, edge–cloud ecosystems, and ML-based smart sensors, are discussed alongside advanced approaches such as multi-agent systems. Key applications explored include load forecasting, predictive maintenance, anomaly and cyber-attack detection, demand-side management, and electric vehicle integration. Special emphasis is placed on Digital Twin and LLM architectures, which enable real-time cyber-physical replicas and context-aware reasoning, thus improving predictive analytics, resilience, and autonomous decision-making. Despite notable advancements, challenges remain in interoperability, data privacy, computational scalability, adversarial robustness, and ethical constraints. By synthesizing these insights, the study highlights the transformative role of AI in creating resilient, sustainable, and intelligent energy systems, and outlines future research trajectories toward standardized DT frameworks, active learning paradigms, and LLM-driven energy intelligence.