Trustworthy AI-Based Microgrids: Control, Cybersecurity, and Adversarial Robustness

The growing volatilization of artificial intelligence in the manner it operates microgrids has redefined the aspects of control, prediction, and management of energy, which can leverage a higher degree of efficiency and flexibility in a decentralized power system. But this shift, too, has brought with it some new vulnerabilities in the forms of dependence on data-based models, massive infrastructures of communication, and autonomous decision making. The review conducts the synthesis of research in the sport of intelligent control, and net-based security and robustness against adversaries by looking at high-quality microgrids in the era of artificial intelligence. This brings together a single conceptual framework to introduce a methodical approach to connecting AI control capabilities, cyber and adversarial threat surfaces, failure modes in the operational process, and mitigation measures at the levels of the microgrid system. The research critically analyzes reported performance improvements of AI-based controllers and reported cybersecurity vulnerabilities and adversarial risks using this framework. Synthesis has shown that even with the learning-based techniques, there is a significant improvement in situations where the nominal conditions are to be considered, but trust worthiness often suffers at cyberspace-sphysical boundaries where data credibility and communications dependability are lost. A comparative overview of the mitigation strategies reveals that no one defense can work, which highlights the significance of the layers and defense-in-depth strategies. This review takes the field a step further by explaining trustworthiness as a quantitative, multi-dimensional feature, which involves performance, security, and robustness. The study identifies the main research gaps, such as the absence of standardized adversarial standards and real-life testing, and presents the priorities of further research and policy as the conditions to facilitate the safe, resilient, and equitable implementation of AI-enabled microgrids.