Post-Contingency Cyber-Physical Self-Adaptive Resilient Control of Power Networks: A Drone-Assisted Edge Intelligence Approach

Abstract

The rapid growth of distributed energy resources (DERs) is reshaping distribution systems, increasing dependencies on information and communication technology while exposing extra fragile points during extreme events. Current overlooked communication infrastructure integrity will significantly affect post-contingency mitigation and recovery. This paper proposes a cyber-physical, self-adaptive resilient control framework of microgrids that leverages edge-intelligent drones to sustain real-time control of DERs. Drones serve dual roles as mobile sensing and communication nodes and as distributed control agents, which enable two-level scalable coordination: local DER-to-Drone links support aggregated sensing and actuation, while Drone-to-Drone networking enables aggregated peer-to-peer coordination. Distributed controller gains are tuned based on a multi-agent deep reinforcement learning (MADRL) algorithm, constrained by a heterogeneous delay-dependent analysis integrating cyber-network formation, wireless channel effects, and real-time control and also guiding reward design and safe online adaptation. All these elements enable iterative and progressive exploration of drone positioning and robust DER coordination without reliance on the distribution network layout. Such a cyber-physical design is comprehensively demonstrated through specific case studies.