Abstract
Modern electricity networks depend on a communication layer for measurement, protection and control, and that communication layer draws its own power from the grid it supervises. This mutual dependence creates failure pathways that single-layer vulnerability analyses cannot capture, and it shifts continuously as topology, loading and control assignments evolve. This study develops and evaluates a self-updating framework that identifies the K entities whose simultaneous loss produces the largest interdependent damage. The framework couples a direct-current power-flow cascade simulator with a control-dependency propagation rule, and scores every node using an Interdependent Criticality Index that combines topological position, flow-based loading, interlayer support burden and dependency depth. Candidate sets are assembled by a lazy-greedy selection procedure, and a drift monitor based on Kendall's rank correlation and the Page-Hinkley test decides when the ranking must be revised. Revision is carried out incrementally over an affected set rather than by full recomputation. The framework was tested on four coupled systems built from the IEEE 14-, 39-, 118- and 300-bus test cases paired with synthetic communication topologies, across 500 operating snapshots per system and 1000 Monte Carlo replications. Measured against exhaustive ground truth, the framework achieved recall@5 between 0.85 and 0.94, compared with 0.38 to 0.52 for degree centrality and 0.66 to 0.80 for a static version of the same index. At K equal to 10 it produced 38.2 per cent demand not served on the coupled 118-bus system, exceeding betweenness-based selection by 10.9 percentage points, while incremental updating reduced update time from 41.6 s to 0.83 s.