Dynamic Functional Connectivity Graph for Assessing Cascading Events in Power System

During power system cascading disturbances it becomes crucial to quickly identify the vulnerable transmission interconnections. Understanding the impact of a triggering outage on these critical interconnections is important for enhancing situational awareness and taking targeted control actions. This paper proposes a new machine learning (ML) based graph theoretic approach for learning the dynamic functional connectivity (DFC) between power system buses with respect to their vulnerability to cascading failures (CF). The learnt DFC graph is then used to characterise vulnerable regions of the power-system using complex network theory based indices. A key feature of the proposed DFC graph is that it takes into account detailed power system dynamics and the action of protection devices when deriving the DFC, going beyond a static representation of the power system graph based on electrical admittances. Multiple operational scenarios for load and renewable generation are also considered when doing so. The proposed algorithm is validated for a dynamic model of the IEEE-10 machine 39 bus system with Type IV wind generation.