Question about VSG with current saturation

[xiaoheimai95]

Dear Dr. Cui and Jinning,
After I modify the code of REGCV1 so that it can meet the current limitation issue (code is shown below), I test it with ieee39 and kundur cases. And it always raise the problem of initialization and convergence failed. Could you help me to fix this problem?

The error report of kundur case as an example:

Suspect initialization issue! Simulation may crash!
       Name        | Var. Value | Eqn. Mismatch
-------------------+------------+--------------
 PIId_xi REGCV1 1  | 0          | 0.654        
 PIId_xi REGCV1 2  | 0          | 0.630        
 PIId_xi REGCV1 3  | 0          | 0.630        
 PIId_xi REGCV1 4  | 0          | 0.630        
 PIIq_xi REGCV1 1  | 0          | -0.099       
 PIIq_xi REGCV1 2  | 0          | -0.205       
 PIIq_xi REGCV1 3  | 0          | -0.209       
 PIIq_xi REGCV1 4  | 0          | -0.095       
 Iref REGCV1 1     | 10.800     | 7.350        
 Iref REGCV1 2     | 10.800     | 7.362        
 Iref REGCV1 3     | 10.800     | 7.376        
 Iref REGCV1 4     | 10.800     | 7.080        
 Idref1_y REGCV1 1 | 0          | 7.268        
 Idref1_y REGCV1 2 | 0          | 7.000        
 Idref1_y REGCV1 3 | 0          | 7.000        
 Idref1_y REGCV1 4 | 0          | 7.000        
 Iqref1_y REGCV1 1 | 0          | -1.095       
 Iqref1_y REGCV1 2 | 0          | -2.280       
 Iqref1_y REGCV1 3 | 0          | -2.324       
 Iqref1_y REGCV1 4 | 0          | -1.061       
 udref REGCV1 1    | 1.024      | 13.058       
 udref REGCV1 2    | 1.051      | 12.549       
 udref REGCV1 3    | 1.052      | 12.548       
 udref REGCV1 4    | 1.024      | 12.576       
 uqref REGCV1 1    | 0.162      | -2.132       
 uqref REGCV1 2    | 0.156      | -4.260       
 uqref REGCV1 3    | 0.156      | -4.338       
 uqref REGCV1 4    | 0.156      | -2.065       
Initialization failed!!

Jacobian matrix is singular.
Suspect diagonal elements: [101, 102, 103, 104]
NaN found in solution. Convergence is not likely

[jinningwang]

It seems that the initialization error is caused by the initialization equations. Take a look at this #172.

Here are some other tips that might be helpful:

1. It would be better to create a new model rather than revise the existing model. For example, you can have your revised model named VSGOuterPIModel2 and VSGInnerPIModel2 or any other meaningful name. Then, manufacturing a new model named REGCV3 with the revised PI model. In this way, you can easily compare your model with the existing REGCV1.
2. During the model development, it can be easier to include only one device of the developed model in the system.
3. After a failed TDS.init(), you can print out the related variables to see if they have a reasonable or expected value.

[xiaoheimai95]

Hi Jinning, when I change the name into REGCV3, it report that <REGCV3> is not an existing model..
What should I do to fix it?

[jinningwang]

Check this, https://docs.andes.app/en/latest/modeling/examples.html#finalize

[xiaoheimai95]

Thx, Jinning!

[cuihantao]

Here's a command to output the data used for initialization:

andes -v 10 run ieee14_linetrip.xlsx -r tds --init &> init_debug.txt

The text file logs the initialization sequence, the inputs and outputs of the initialization, etc.

[xiaoheimai95]

I'm sure it is installed cause when I test other case without REGCV3, it can run smoothly.

[jinningwang]

As Dr. Cui commented below, you might have more than one copy installed. You can run andes in terminal to get the version information to check which version are you using.

[xiaoheimai95]

I will recheck later. And to be on the safe side, is it possible to print out the init_debug.txt message via python code?

[jinningwang]

If the message does not involve sensitive information of yourself, it's ok to post it here for discussion.

[cuihantao]

The messages in init_debug.txt are actually the stderr logs. There's currently no API to export the messages via a Python API.

[cuihantao]

Likely you have more than one copies of ANDES installed in different environments. The one working is installed in the development mode, but the one not working is installed as a site package. Get Outlook for iOS<https://aka.ms/o0ukef>

…

________________________________ From: xiaoheimai95 ***@***.***> Sent: Tuesday, April 18, 2023 10:17:51 AM To: CURENT/andes ***@***.***> Cc: Hantao Cui ***@***.***>; Comment ***@***.***> Subject: Re: [CURENT/andes] Question about VSG with current saturation (Discussion #441) I'm sure it is installed cause when I test other case without REGCV3, it can run smoothly. — Reply to this email directly, view it on GitHub<#441 (reply in thread)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ABSNZA3WUU2CFNSSD5JXVR3XB2WB7ANCNFSM6AAAAAAW76XPNI>. You are receiving this because you commented.Message ID: ***@***.***>

[xiaoheimai95]

Dear Dr. Cui and Jinning,
The init_debug.txt is shown. Would you please give me some advice?
init_debug.txt

[jinningwang]

1. Var. Value is not used to compared with Eqn. Mismatch. Check this https://docs.andes.app/en/latest/modeling/variables.html?highlight=equation#value-and-equation-strings
2. AliasAlgeb is not used for equations but final output. Check the init message and take a look on the sequence of the variables assignment, also take a look on this, https://docs.andes.app/en/latest/modeling/_generated/andes.core.var.AliasAlgeb.html?highlight=aliasalgeb

[xiaoheimai95]

So I have already set Iref And I've checked the formula for correctness.

[jinningwang]

e_str might be wrong, check other models.

[cuihantao]

See https://docs.andes.app/en/latest/modeling/_generated/andes.core.var.Algeb.html#andes-core-var-algeb for e_str.

[xiaoheimai95]

Thank you very much!