Skip to content

runners.md

Latest commit

 

History

History
62 lines (48 loc) · 3.33 KB

runners.md

File metadata and controls

62 lines (48 loc) · 3.33 KB
Raw

Runners

perc-studio > Runners

run and sweep execute a .dyr record against a fault scenario through a runner. There are two, behind one interface.

class PsseRunner(Protocol):
    def run(self, raw_case_path, dyr_text, scenario, *, block_name, case_id) -> SimulationResult: ...

get_runner(name) (src/psse/runner.py) is the single factory every command and sweep goes through. Every command, dev/validation/sweep.py and dev/validation/checks.py depend only on the protocol, so a third runner would work with all of them.

--runner What it is Needs
mock (default) An analytic stand-in: the load model's own simulate hook nothing
psse blackbox.runner.PsseSubprocessRunner: each case in its own PSS/E process on the black-box one-bus rig a PSS/E selection

Selecting PSS/E

--runner psse reads the selection from the environment:

Variable Meaning
PERC_STUDIO_PSSE_VERSION e.g. 35.6 or 36.7. Required.
PERC_STUDIO_PSSE_DIR Install folder, when not the version's standard path
PERC_STUDIO_PSSE_PYTHON The Python that runs psspy, when not found from PSSPYxx
PERC_STUDIO_PSSE_PSPY The PSSPYxx folder, when not the newest
PERC_STUDIO_PSSE_DLLS The model's DLL(s), separated by ;

or the selection saved by the app's Black-Box Validation tab. With no selection the runner raises NotImplementedError and the command exits 4. perc-studio info shows which of these variables are set; perc-studio psse check proves a selection works (see PSS/E and black box).

export PERC_STUDIO_PSSE_VERSION=35.6
export PERC_STUDIO_PSSE_DLLS="C:/models/PERC2_v35_v2p3.dll"
perc-studio run case.json --fault-duration-s 1.0 --fault-voltage-pu 0.5 --runner psse --out psse_result.json

What MockPsseRunner computes

It is not a PSS/E simulation. It finds which registered load model wrote the record (load_models.find_model_for_dyr, by the quoted model name in the .dyr text), parses the text back into a CON dict with that model's own parser - so it always runs on the exact artifact a real run would consume - and runs that model's analytic simulate hook (simulate_frequency for a frequency event). For PERC1 that is load_models.perc1_dll.simulation.simulate_perc1_response, the same simplified approximation the desktop app's live chart uses.

  • It honors fault_duration_s and fault_voltage_pu/fault_frequency_hz. It ignores fault_start_s (analytic simulators start the event at 0.5 s) and adds a warning to the result when another value was requested.
  • It evaluates all four Option-3 stages per direction (VDPA/DMA + VDPA1-3/DMA1-3, VDUA/TDUA + VDUA1-3/TDUA1-3) like the real DLL: every stage whose own threshold the event crosses counts down from the same onset, and the shortest dwell that completes within the event trips. See Background.
  • It needs the block's model to have both a parser and a simulator (models list's mock_run); a CLOD motor block exits 2. A frequency scenario also needs frequency_mock_run.

A mock pass is not a substitute for a PSS/E run: no network solve, no generator dynamics. Use it to check your parameterization logic and the envelope bookkeeping, then confirm with --runner psse or blackbox.