Machine learning for intelligent microgrid operation: fundamentals, challenges, and advanced techniques
International Journal of Electrical Power and Energy Systems, cilt.182, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Derleme
- Cilt numarası: 182
- Basım Tarihi: 2026
- Doi Numarası: 10.1016/j.ijepes.2026.112243
- Dergi Adı: International Journal of Electrical Power and Energy Systems
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Compendex, INSPEC, Directory of Open Access Journals, Academic Search Ultimate (EBSCO), Engineering Source (EBSCO)
- Anahtar Kelimeler: Cyber-Physical Resilience, Federated Learning, Graph Neural Networks, Machine Learning, Microgrids, Reinforcement Learning, Renewable Energy Integration, Seamless Mode Transitions
- Van Yüzüncü Yıl Üniversitesi Adresli: Evet
Özet
Microgrids (MGs) have emerged as a key enabling technology for next-generation power systems due to their ability to integrate distributed energy resources (DERs), enhance resilience, and support sustainable and flexible energy management. However, the increasing penetration of stochastic renewable generation, bidirectional power flow, and complex operational requirements particularly during seamless transitions between grid-connected and islanded modes introduce significant challenges in real-time control, optimization, and stability. These challenges necessitate advanced data-driven intelligence beyond conventional model-based approaches. This paper presents a comprehensive and systematic review that bridges fundamental MG concepts with advanced machine learning (ML) methodologies to address operational, control, forecasting, and resilience challenges. First, a detailed overview of MG architecture, components (DERs, energy storage systems, loads, and point of common coupling), and operational modes is provided, with particular emphasis on mode transition mechanisms, including islanding detection, synchronization, and grid-forming control requirements. Next, a structured taxonomy of ML paradigms is introduced, covering supervised learning, unsupervised learning, reinforcement learning (RL), deep learning (DL) architectures, and emerging approaches such as graph neural networks (GNNs) and federated learning (FL), highlighting their respective roles in MG environments. The paper further establishes an explicit mapping between MG operational challenges and ML techniques, demonstrating how supervised learning enables accurate forecasting and fault classification, reinforcement learning supports adaptive energy management and control optimization, and deep learning methods enhance time-series prediction and anomaly detection. In addition, ML-based solutions for unit commitment, economic dispatch, energy storage optimization, demand response, fault detection, and cyber-attack mitigation are critically analyzed and compared with traditional optimization methods such as MILP and dynamic programming. Special attention is given to resilience and cybersecurity, including false data injection detection, self-healing strategies, and adversarial robustness. Finally, the review identifies key research gaps and future directions, including explainable AI (XAI) for trustworthy decision-making, physics-informed and hybrid ML models, digital twin integration, privacy-preserving distributed learning, and real-time deployment constraints in large-scale MGs.