Skip to content

background.md

Latest commit

 

History

History
36 lines (29 loc) · 2.45 KB

background.md

File metadata and controls

36 lines (29 loc) · 2.45 KB
Raw

Background: Option 3

perc-studio > Background

An open engineering question used to be: does the compiled PERC1 DLL always evaluate Option 3's four-stage trip table, even for a facility this app only ever drives via Option 1? That question is resolved (2026-09-24): yes, it does, confirmed by direct PSS/E v35.6.4 simulation against the real compiled PERC2_V35.dll. Write-up: PERC1_MODEL.md. The raw evidence is in the PSS/E validation library, which is not published.

That validation ran through a standalone psspy script. The app's own runner (get_runner("psse"), see Runners) now runs the same fixtures on real PSS/E once a selection exists. The two purpose-built .dyr fixtures are checked in as a permanent regression: ercot_nogrr282_vendor_default_stages.dyr and ercot_nogrr282_full_curve_stages.dyr. src/dev/validation/checks.py still operationalizes the question:

  • MULTISTAGE_UNDERVOLTAGE_PROBE/MULTISTAGE_OVERVOLTAGE_PROBE are the two discriminating scenarios.
  • evaluate_multistage_probe(current_results, full_curve_results) reproduces the decision logic, given the two fixtures' results against both probes.
  • tests/test_multistage_trip_regression.py::test_multistage_stages_are_confirmed_live asserts the confirmed outcome (stages_live is True). It is skipped unless a PSS/E install is selected (PERC_STUDIO_PSSE_VERSION, and PERC_STUDIO_PSSE_DLLS for the DLL).

Why the sweep example passes now

tests/fixtures/sweep_example.json sweeps ercot_nogrr282 and spp_hill against a 0.6 s / 0.75 pu sag and a 0.6 s / 0.85 pu ride-through probe. It exits 0 (passed_cases: 4, failed_cases: 0): 0.85 pu is above the required floor at 0.6 s so nothing trips, and 0.75 pu is below the 0.80 pu / 0.50 s tier both standards publish, which the mock's stage-2 dwell timer (VDPA1=0.80, DMA1=0.50, populated from that same tier) completes inside the 0.6 s window.

Before 2026-09-24 this sweep exited 5: the exported .dyr only drove stage 1 (VDPA=0.90 pu, DMA=2.00 s, which had not reached its dwell by 0.6 s) and the mock only evaluated that one stage, so the mock and the envelope disagreed. That was a real finding about a real gap, and is how the multi-stage behavior was found.