Opened 5 years ago
Last modified 3 years ago
#5960 assigned defect
Large scale electrical models
| Reported by: | Owned by: | Karim Adbdelhak | |
|---|---|---|---|
| Priority: | high | Milestone: | |
| Component: | Initialization | Version: | v1.14.1 |
| Keywords: | Solver | Cc: |
Description
Dear staff member,
I am working on large scale electrical models and I am having lots of problems with the solvers. Especially when adding inductances and capacitors to the model, I get a lot of errors with the linear and non-linear systems that can't be solved. Should I use any particular non-linear solver? And send the model? I have already tried the different ones and limiting the tearing, but no results.
Thank you
Attachments (2)
Change History (11)
comment:1 by , 5 years ago
follow-up: 3 comment:2 by , 5 years ago
Thank you for your quick response.
I attached a screen capture of the main error I get. Should I send the model by mail also?
comment:3 by , 5 years ago
Replying to lluc@…:
Thank you for your quick response.
I attached a screen capture of the main error I get. Should I send the model by mail also?
If there are no confidentiality issues, you should preferably attach it to this ticket. Otherwise you can send it to me francesco.casella@….
Please make sure all required libraries are included to run the model.
comment:5 by , 5 years ago
I attach the file. In the library LSPVPP you can find different models and components such as a voltage source converter and a photovoltaic power plant. NGen1MW model is the one that I can't simulate because of adding inductances in the LV-MV transformers. As you can see in NGen1MW_NO_IND it is exactly the same model with the same parameters but without the inductances in the transformers LV-MV. I would like to know how to debug it to be able to build larger models. I have already tried different methods for the nonlinear solvers and setting initialization parameters but I haven't succeeded.
Thank you for your time.
comment:6 by , 5 years ago
| Milestone: | Future → 1.16.0 |
|---|---|
| Owner: | changed from to |
| Status: | new → assigned |
I briefly discussed this with Karim today. There could be some problems of missing index reduction. If you look at notification with the list of states, you can see, e.g.,that these three variables were picked as states
T1.inductor2.inductor[1].i T1.inductor2.inductor[2].i T1.inductor2.inductor[3].i
I'm not sure about your application, but in most three-phase models, there are constraints due to the connection topology so that the currents in the three windings of a machine sum to zero (e.g. because of a star connection). If this is not recognized by the backend, then sooner or later some singular equations will be found, which seems to be the case here. See, e.g. #5452, which was eventually fixed by means of the ASCC algorithm by Karim.
Can you comment on whether those three currents can actually be independent state variables, or if they are somehow constrained?
comment:8 by , 4 years ago
| Milestone: | 1.17.0 → 1.18.0 |
|---|
Retargeted to 1.18.0 because of 1.17.0 timed release.
We really need some concrete examples that demonstrate the problems you are running into, otherwise the description of your issues is too vague for us to help you. Can you provide some in this ticket?