<![CDATA[The global transition toward decarbonization has led to a greater integration of renewable energy sources (RES) into power systems, facilitating the widespread adoption of direct current (DC) microgrids. DC microgrids are particularly compatible with modern power systems because they support solar photovoltaic systems, batteries, and electronic loads. Despite these advantages, high levels of intermittent RES introduce challenges related to power balance, voltage stability, and reliable operation. Artificial Intelligence (AI) has emerged as a critical tool, enabling advanced forecasting and intelligent energy management systems (EMS) to address these issues. This comprehensive review examines state-of-the-art AI-based methods for DC microgrids, analyzing a wide range of studies from simulation-based models to real-world experimental pilots. It starts with an overview of the system architecture and operational challenges, followed by a novel taxonomy of AI approaches. The review critically compares machine learning for forecasting and reinforcement learning for real-time control, highlighting their respective performance in handling uncertainty. AI-driven EMS strategies, especially reinforcement learning for optimal scheduling, are detailed. The symbiotic relationship between accurate forecasting and robust EMS is explored, along with challenges such as data dependency and model explain ability, for which emerging solutions, such as federated learning and explainable AI (XAI), are discussed. The paper concludes by outlining future research directions, such as federated learning and standardized benchmarks. It underscores this review's key contribution by providing an integrated framework that bridges the gap between AI-driven forecasting and control for resilient and efficient DC microgrid operation.]]>
Copyrights © 2025
