Artificial Intelligence–Enabled Resilience Enhancement of Electric Vehicle Integrated Microgrids

Abstract

Microgrids are beginning to include electric vehicles (EVs), which may provide many benefits and challenges for the stakeholders involved in microgrid operation. This report describes how the combination of artificial intelligence (AI) with microgrid operation through the use of predictive analytics, constrained optimization (CO), and safe reinforcement learning may improve the resiliency of microgrids. The framework developed for the experimentation considers the charging behavior of EVs, variability of renewable resources, and communication limitations in order to motivate proactive decision-making and adaptive control of microgrid operation during disturbances. A multi-agent hierarchical architecture allows for decentralized control of decision-making and assures achievement of system-level objectives. High-fidelity simulation and hardware-in-the-loop testing show that AI-based coordination reduces the duration of outages, maintains critical loads, and uses the vehicle-to-grid (V2G) service efficiently compared to existing state-of-the-art approaches. The framework incorporates three issues at an explicit level: uncertainty quantification, privacy-preserving coordination for EVs, and safety filters for operational safety as a means to mitigate operational violations. Our findings demonstrate that there is sufficient user participation by EVs to utilize the AI-based coordination to convert mobile storage into reliable resiliency resources; thereby providing operator and policy makers with actionable methods to enhance distributed energy systems. We provide recommendations for both deployment and research to enable the rapid, safe implementation of the concepts developed here in the marketplace.