Coordinated networked microgrids can help restore neighboring distribution grids after a major blackout and promise to significantly enhance the day-to-day reliability of the power grid. As anticipated by the U.S. DOE, R&D for networked microgrids will lead to the next wave of smart-grid technology, which will help achieve the vision of a highly resilient grid. Three main challenges, however, have prevented NMs from serving as dependable resources and thus prohibited their wide adoption: 1) Lack of understanding of NM dynamics under frequent changes in status, ubiquitous uncertainties, fast ramping, low inertia, and non-synchronism; 2) Big data but limited and unscalable analytics, as current technologies are unable to handle the massive volume of dynamic data needed for real-time decision making, and the need to ensure data privacy inevitably renders model-based system analyses and control impractical; and 3) Bottlenecks in the cyber-infrastructure due to delays, congestions, failures, cyberattacks etc., and the ever-increasing pace of functional/structural changes can plague microgrid cyber-networks.
To address the above challenges, this project aims to develop and demonstrate AI-Grid: AI-enabled, provably resilient networked microgrids. Using a programmable platform that integrates deep learning, reachability analysis, formal control, and high-assurance software architectures, AI-Grid will be deployed at three highly influential networked microgrids, where it will demonstrate its capabilities to modernize/decarbonize America’s power sector.