Contingency-Aware Diagnosis and Mitigation of Seismic Multi-Element Outages on a Public 10,000-Bus Grid
Year: 2026
Type: preprint
Abstract: High-impact low-probability earthquakes can damage substations and trip multiple lines simultaneously. N-1 screening cannot cover the resulting multi-element outages, while seismic-grid studies often rely on small cases or non-public data. We build a reproducible public benchmark on ACTIVSg10k (10,000 buses, 12,706 branches), USGS ShakeMap scenarios, and HAZUS fragility. States are scored by Δ = lost load (MW) + 10⁴·overload (p.u.²); equivalently Δ = 100δ with δ dimensionless. Among 8,837 converged N-1 cases the worst overload is 6.319 p.u.²; among the 1,220 converged top-50 N-2 pairs the worst is 14.897 p.u.² (2.36×). Of 200 random N-2 samples, 92 converge and none overload. A greedy baseline seeded by N-1/N-2 candidate sets reproduces the worst pair in two actions; this is a low-cost diagnostic protocol, not an independent discovery claim. One reconnection at bus index 7231 clears the scored overload: branch 9096 remains open but has zero individual N-1 penalty. A lossless DC-OPF absorbs the N-2 with zero load shed; its L1 deviation equals base-loss reconciliation and is not a contingency cost. The tested N-3 families do not exceed N-2. Under San Andreas M7.8 and Cascadia M9.0, 30-step boundary-priority switching reduces Δ by 20.4% and 15.6%, respectively; median-fragility sensitivity gives 30.5% and 46.3%. These are single-realization results. The scenario states are fully islanded, so their reductions are invariant to the overload weight. All artifacts carry SHA-256 data hashes and SHA-256 audit digests.
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7236651184
