Skip to content

Latest commit

 

History

History
146 lines (107 loc) · 7.85 KB

model_perc2_dll.md

File metadata and controls

146 lines (107 loc) · 7.85 KB

PERC2 DLL v2.3: validation results

PERC2 v2.3 is ISO New England's second user-defined model (PERC2_v34, v35 and v36 builds of PERC2_v3x_v2p3.dll). It turns out to be the PERC1 DLL with six CONs appended: a reactive-power lead-lag (TAQ, TBQ) and momentary cessation's own reconnect band and ramps (VLRCMC, VURCMC, RAMPAMC, RAMPUAMC). See the model overview for how it compares with the others and the model reference for how the app writes it.

PSS/E versions validated 35.6.4 with PERC2_v35_v2p3.dll and 36.7.1 with PERC2_v36_v2p3.dll, one plan on both
Needs the DLL for your PSS/E version, loaded as a user-defined model
Record USRLOD 'PERCBL' 12 IT 2 179 33 250 0, two ICONs, then 179 CONs
PSS/E runs 285 cases on 35.6.4 (89 of them on the PERC1 DLL, for comparison) and 194 on 36.7.1
Automatic checks 351 of 351 pass
Reference the vendor document and its sample record

The PSS/E runs behind these counts are in the authors' PSS/E validation library, which is not published.

How it was validated

One 285-case plan ran on both builds: the record, every documented CON on its own, ramps against the load factor, momentary cessation, out-of-range values, a one-machine network, the records the app exports, and the same scenarios on the PERC1 DLL. The app's simulator was then compared with PSS/E on all 79 single-dip cases, 31 of them with momentary cessation: it reproduces PSS/E's first cease, ramp start and ramp length within one sample.

Results

Question Answer
What is v2.3? The PERC1 DLL plus six CONs. Every PERC1 DLL behavior holds on it, and the 79 cases that use no v2.3-only CON give the PERC1 DLL's exact trace outside the voltage-reconnect ramps.
Does the document's record work? Its record format (173 CONs) loads without a message but is wrong: a load that ceases never comes back. The vendor sample's header (179) is right.
Does the DLL do what the document says? Mostly. Eight statements are contradicted (below).
Defects The v36 build stops PSS/E 36.7 when a high-voltage momentary cessation ramp ends, and returns not-a-number for any non-zero TA or TB. The v35 build does neither.
Is v36 otherwise the same? Yes. Outside those two defects the builds agree within 0.001 on V, P, Q and FracOn on every case.

Where the DLL departs from the document

Document PSS/E
The record has 173 CONs 179. 173 loads silently and breaks momentary cessation
ICON(M) = 1 means 50 Hz Ignored. The DLL always measures frequency against 60 Hz
TA/TB are bypassed by setting them to 0 TA with TB = 0 returns not-a-number: both must be 0. On v36, any non-zero value does
RAMPA is multiplied by the load factor Right, and so is RAMPUA, which is not documented
TBLK waits for the voltage between VLBLK and VUBLK It waits for VLRCMC < V < VURCMC
The ramp after a cessation is RAMPA or RAMPUA It is RAMPAMC or RAMPUAMC
Option 2 uses CONs J+160 to J+163 Also FCEASE, FRECON and the stage trips, which take it over
The reactive lead-lag is J+166 and J+167 Those are the reactive washout. The lead-lag is J+173 and J+174

Records PSS/E accepts that change the result

A 173-CON record: the load never comes back after a cessation

Both DLLs register PERCBL: whichever is loaded first runs every record

The header's CON count is the only thing that tells the two DLLs' records apart, which is how the app routes them. A record with no ICON line switches momentary cessation on from unset memory. With no DLL loaded, PSS/E prints a "not accessible" message once and then simulates the load as zero.

Ramps and the load factor

Ramps and the load factor

RAMPA and RAMPUA are stretched by the load factor (LFM): a 2 s ramp is 1.6 s at LFM 0.8. RUFRN1, ROFRN1, RAMPAMC and RAMPUAMC are not. A PERC1 DLL record moved to v2.3 unchanged therefore reconnects its voltage ramps 1.25 times faster at the usual LFM of 0.8. The app divides both ramps by LFM so a preset keeps its seconds.

Momentary cessation

Momentary cessation cycle

The load ceases at once when the filtered voltage is strictly below VLBLK (or above VUBLK), with no delay, and returns after the voltage has been inside VLRCMC to VURCMC for TBLK + TDELY. It ramps over RAMPAMC (or RAMPUAMC) to 1 - FCEASE + FRECON × FCEASE, so with FRECON below 1 every cessation loses part of the load for good.

A stage trip takes over the cessation

The stage timers keep running during a cessation. A stage that trips takes the cessation over: the load then waits for the stage's reconnect window and TDRA. The vendor's sample with momentary cessation on comes back after about 10 s, not the 0.05 s the cessation's own settings suggest.

The same record on the PERC1 DLL and on v2.3

After a cessation the PERC1 DLL returns once the voltage is above VLBLK; v2.3 waits for VLRCMC. A recovery to 0.79 pu reconnects on one and stays ceased on the other.

The two builds

The v35 and v36 builds give the same trace

The v36 build stops PSS/E

At the end of every high-voltage momentary-cessation ramp the v36 build stops the whole PSS/E 36.7 process (Fortran runtime error, exit code 66). A study loses every result of the run, not just this load, and the error appears only on the console. The low-voltage side never crashes.

The P lead-lag on each build

On the v36 build any non-zero TA or TB returns not-a-number from the first time step. The reactive lead-lag (TAQ/TBQ) works on both builds when TBQ is above 0.

Frequency base

ICON(M) = 1 (50 Hz) does nothing

In a 50 Hz case the DLL trips on over-frequency with no disturbance, with ICON(M) at 0 or

  1. It cannot be used in a 50 Hz system unless its frequency settings are written for 60 Hz × (f / 50).

Out-of-range values

Out-of-range values are used as given

Only LFM below 0.001 is replaced (by 0.8). A FCEASE above 1 makes the load export power; a negative FCEASE makes a trip raise the load to 150 %; a negative ramp runs FracOn below zero and never stops; a reconnect window that excludes 1.0 pu never reconnects. None of these produces a message.

What the app does with this

  • Records are written with the 179-CON header and routed on import by the header's CON count (173 reads as the PERC1 DLL, 179 as this model).
  • Presets carry over from the PERC1 DLL's bundles CON for CON, with RAMPA and RAMPUA divided by LFM, VRPA capped at 0.95 pu (the five presets with 1.0 would never reconnect), momentary cessation off, and the four new cessation CONs at the vendor sample's values.
  • The editor flags every hazard above: out-of-range FCEASE, FRECON and ramps; TA/TB and TAQ/TBQ that return not-a-number, and any TA/TB for PSS/E 36.7; ICON(M) set; the high-voltage crash with momentary cessation on; the permanent partial loss and the stage takeover; reconnect windows that exclude 1.0 pu; and cycling.

Limits of this validation

  • The two builds were run on one PSS/E version each (35.6.4 and 36.7.1). The PSS/E 34 build was not run.
  • Dips to exactly 0 pu are excluded from the simulator comparison, because the test setup swells the bus to about 1.7 pu for 0.1 s after them.
  • Out-of-range behavior was measured at the values listed, not swept.
  • Results are for the builds tested; a later build of either DLL may behave differently.