REGC_A (REGCA1) network non-convergence for close-in faults

asked 2023-01-04 12:53:56 -0500

nflodin
1 ●1 ●2 ●2

updated 2023-01-04 12:58:53 -0500

Hello everyone,

I am performing a renewable impact study, and as part of this study, we must simulate faults nearby the proposed wind farm. The wind farm uses the following WECC generic models, and we're observing numerical instability (network non-convergence) when the fault is applied for the entire duration of the fault, even when significant fault impedance is added:

REGCA1 REECA1 WTDTA1 WTPTA1 WTARA1 REPCTA1 WTTQA1

The faults we are simulating do not completely isolate the wind farm from the rest of the network. Despite this, we are having the network non-convergence issues. We've tried several approaches to fix this problem but nothing seems to work. These include:

Switching to the REGC_B (REGCBU1 in PSS/E) model, which supposedly has better numerical stability under weak grid conditions
Changing XSORCE of the wind machines... Tried 9999, 0.9, 0.05, none of these achieve convergence
Adding fault impedance. However, so much fault impedance has to be added that the faulted bus only drops in some cases to 0.45 p.u. voltage.

Does anyone have experience with diagnosing network non-convergence/numerical instability issues with the 2nd generation generic renewable models? I'm sure there's a way to achieve convergence with these models, and to truly simulate a "bolted" fault near (one or two buses away from) the new wind facility, but we haven't found a good solution so far.

Thanks for any and all help.

answered 2023-01-04 16:51:17 -0500

jconto
2701 ●10 ●30 ●41

Consider the following:

- Keep Xsorce at 9999.
 - Lower the integration step size to 0.001
- Lower the acceleration factor to 0.8 or lower (in simulation parameter setting)
- Tune the acceleration factor in REGCA1 to 0.7 or lower.

From PTI on model REGCAU1:

“To aid convergence we added a factor called Accel, which is a factor used only in that in the model to decide how much (fraction) of the newly calculated voltage is used in the calculation of current injection.” “The default value of Accel should be set to 1.0. We have tested with other values of Accel and some of our users have also used other values such as 0.8 etc.” From a dynamics expert (developer of WECC REGC_A model)

"I always set the acceleration factor to 0.7" The Accelerator Factor of the REGCAU1 model is NOT an actual control setting or a physical parameter but a computational variable that users can modify to help the PSS/e convergence of the model REGCAU1 in a dynamic simulation. Based on our experience acceleration factor=0.7 is a reasonable value to use. Lower values can be used if it lead to better dynamic run performances.

edit flag offensive delete link

Comments

I have determined that the non-convergence is in large part being caused by the low voltage active current management functionality, specifically the Lvpnt0 parameter, which was set to 0.01.

nflodin ( 2023-01-04 17:56:58 -0500 )edit

Raising Lvpnt0 and Lvpnt1 to 0.4 and 0.8, respectively (default values for these) and adding some fault impedance achieves convergence...

nflodin ( 2023-01-04 17:58:15 -0500 )edit

The only location I've been able to find discussion of numerical instability related to Lvpnt0 is here (https://www.nerc.com/comm/PC/LoadModelingTaskForceDL/Reliability_Guideline_-_DER_Modeling_Parameters_-_2017-04-04_-_FINAL_DRAFT.pdf). Do you know of anywhere else this is discussed?

nflodin ( 2023-01-04 17:58:44 -0500 )edit

Some info here on LVPL related to network convergence: https://www.nrel.gov/docs/fy16osti/64822.pdf. We have been experiencing issues with LVPL on models that have xsource set ~= 1.0 instead of >= 9999.0 as suggested in the PSSE PAGV2 manual.

jtkiv ( 2023-02-13 05:48:28 -0500 )edit

add a comment

REGC_A (REGCA1) network non-convergence for close-in faults

1 answer

Comments

Your Answer

Question Tools

Stats

Related questions