Vulnerable node identification in active distribution networks under multi-scenario quasi-static time-series operation: A CVaR-based conditional-consequence screening framework

Expected-loss and composite-score rankings can underrepresent active distribution network (ADN) nodes whose severe consequences concentrate in stressed operating states. This paper develops a CVaR-based conditional-consequence screening framework for standardized node removals over weighted year-round quasi-static time-series (QSTS) operation. Each candidate is represented by a post-removal consequence distribution that combines energy not supplied, incremental voltage-violation severity, and relative branch-loading stress. CVaR, tail amplification, and resampling-based selection probability distinguish mean-driven from tail-dominated consequence profiles. In the modified IEEE 33-node feeder, nodes 29 and 28 rise from expected-loss ranks 8 and 9 to CVaR ranks 1 and 2 under the main grid-forming (GFM)-enabled scenario; the ordered Top-2 is preserved in the independent 2024 test with 90% Top-10 overlap. The 85-node feeder gives an 8/10 expected-loss/CVaR Top-10 overlap and promotes two strongly tail-amplified nodes into the CVaR Top-10. A current-domain diagnostic using a literature-based 255-A surrogate and ± 20% sensitivity further separates baseline-relative branch-loading stress from absolute current utilization: observed exceedances occur on solved main-grid branches, whereas the terminal GFM-supported-island pattern is associated with departures from branch-specific daily baselines. The resulting ranking is conditional on the standardized contingency, with event likelihood retained as a separate layer for downstream risk-aware applications.