Engineering question: Will the system remain stable, and how close is it to a stability boundary?
Focus¶
Transient rotor-angle stability and critical clearing time (CCT) under stated fault/clearing rules. Finite-horizon simulation labels are operational, not asymptotic certificates.
Learning path¶
Stability labels and CCT — IEEE/CIGRE placement, three cases, CCT metrics
Surrogate models for stability tasks — classifier vs trajectory vs margin
Direct methods — equal-area / energy as classical check
Lab 5 — Stability and CCT (SMIB first)
ANDES as network reference — after SMIB CCT works; Kundur gallery as linked depth
Relation to foundations¶
Use the same SMIB reference from Lab 1. Trajectory accuracy alone does not imply correct CCT or safe decisions near the boundary. Network CCT (ANDES / Kundur) reuses the same SciML evaluation ideas on a larger plant — it does not replace Foundations.
Discussion: what Stability methods can and cannot claim¶
| Can claim (with evidence) | Cannot claim |
|---|---|
| CCT error and false-safe count against the reference simulator, under a stated fault, clearing rule, and study window | A stability certificate beyond the study window or the tested faults |
| Agreement with the equal-area criterion on the classical SMIB | That a learned margin is protection-grade near the boundary |
| Ranking of surrogates under the same stability criterion | That a better trajectory fit implies a better CCT |
Rule: report false-safe errors separately from average error. Near the boundary, one optimistic mistake matters more than many small ones.
Status¶
Stability narrative deepened (classification, surrogates, direct methods). Executable SMIB CCT notebooks come after Lab 1–2 tooling is stable. Network path: summarize-then-link to ANDES; pin envs/course_andes.yml.