Goals¶
Create reference labels and CCT brackets on classical SMIB (Lab 1 tooling), then evaluate a learned surrogate or forward PINN near the boundary. Report signed and absolute CCT errors, false-safe decisions, and one case where trajectory ranking and CCT ranking disagree.
Prerequisites¶
Prefer also Lab 2
Chapters: CCT, surrogate stability, optional direct methods
Reference CCT (no ML yet)¶
Using simulate_smib / angle_stable:
Fix prefault state and fault inception
Sweep or bisect clearing time until the operational label flips
Record bracket width and tolerance
Optional: if the case matches equal-area assumptions, compare search CCT to the classical estimate (direct methods)
Learned CCT (after Lab 2)¶
Replace the inner trajectory call with a surrogate; keep the same stability rule. Compare to the reference CCT.
Deliverables¶
Reference CCT table for ≥3 operating / fault settings
Signed CCT errors for the learned method
False-safe count on a boundary-focused set
One disagreement case write-up
Materials (launch paths)¶
Assessed SMIB path: use Lab 1 tooling locally until the dedicated SMIB CCT notebook is added; then Binder/Colab badges will match Labs 1 and 3
ANDES network extension is local only (not on Binder / Colab)
Optional network extension (after this lab)¶
Assessed Lab 5 stays on classical SMIB. When SMIB CCT works, the preferred network follow-on is Kundur via ANDES (not a sixth assessed lab):
Notebook:
lab05_network_cct_andes.ipynb— one-bus CCT sweep + signed error / false-safe metrics (local /envs/course_andes.ymlonly)Helper:
sciml_ps.andes_cct(requiresenvs/course_andes.yml, andes==2.0.0)Bridge page: ANDES as network reference
Official full map (link only): Critical Clearing Time Mapping
Status¶
Narrative and protocol ready; full SMIB surrogate CCT notebook follows Lab 2. Network extension notebook + ANDES pin (2.0.0) are in place.