Title:Event-Triggered Islanding in Inverter-Based Grids

Abstract:The decentralization of modern power systems challenges the hierarchical structure of the electric grid and necessitates automated schemes to manage adverse conditions. This work proposes an adaptive isolation methodology that can divide a grid into autonomous islands, ensuring stable and economical operation amid deliberate (e.g., cyberattacks) or unintentional abnormal events. The adaptive isolation logic is event-triggered to prevent false positives, enhance detection accuracy, and reduce computational overhead. A measurement-based stable kernel representation (SKR) triggering mechanism initially inspects distributed generation controllers for abnormal behavior. The SKR then alerts a machine learning (ML) ensemble classifier to assess whether the system behavior remains within acceptable operational limits. The event-triggered adaptive isolation framework is evaluated using the IEEE RTS-24 and 118-bus systems. Simulation results demonstrate that the proposed framework detects anomalous behavior with 100% accuracy in real-time, i.e., within 22 msec. Supply-adequate partitions are identified outperforming traditional islanding detection and formation techniques while minimizing operating costs.