Data-driven control and operation of cyber-physical microgrid systems

Abstract

By interconnecting distributed energy resources (DERs) and local loads, microgrids have become a promising paradigm for the future smart grid. On the other hand, due to smaller grid size and higher penetration of renewable energy sources (RESs), the microgrid has a lower system inertia compared to the conventional grid, which may bring more challenges for stable and effective operation of microgrids. This thesis studies the control strategy for flexible and resilient operation of microgrids. First, a data-driven optimal secondary control is developed in islanded AC microgrids based on deep reinforcement learning (DRL). Next, a decentralized and economic frequency control is proposed in a networked-microgrid (NMG) system. Further, a data-driven gain-scheduling approach is designed for distributed secondary controllers, which enhances the stability of time-delayed microgrids. Finally, a learning-based cyber-attack tolerance method is proposed to support secondary control in microgrids.