Toward intelligent and resilient microgrids: A survey of machine learning approaches for renewable energy integration

The integration of renewable energy and distributed generation has transformed microgrids into complex, adaptive, and data-driven systems. Managing these systems requires advanced forecasting and control strategies beyond traditional rule-based or model-driven approaches. Machine learning provides a data-centric framework to optimize microgrid operations, enabling power flow management, demand forecasting, anomaly detection, and resilience enhancement. Existing reviews remain fragmented, often focusing on single domains, while emerging approaches such as federated learning, physics-informed learning, and generative artificial intelligence (AI) are scarcely addressed. To fill this gap, this paper presents a comprehensive review of recent machine learning applications in microgrids, covering supervised, unsupervised, reinforcement, and deep learning techniques. The review surveys peer-reviewed research from 2019 to 2025, with a focus on key domains such as load forecasting, energy management, voltage/frequency regulation, cybersecurity, and adaptive protection. The review highlights key technical challenges, including data scarcity, generalization, cybersecurity, and explainability, and explores emerging directions such as federated learning, transfer learning, and physics-informed learning, discussing their potential for advancing microgrid intelligence and resilience. By synthesizing state-of-the-art developments and outlining future opportunities, this review aims to guide researchers and practitioners in building intelligent, secure, and adaptive microgrid systems.