Multi-rate simulation based on andes

[xjlcode]

Dear Dr.Cui,
I want to write a multi-rate simulation program in the andes framework. Do you think this idea is feasible? Or do you have any suggestions?
Thanks.

[cuihantao]

It is definitely doable, but that sounds like a large project. What specific goals do you want to achieve?

[xjlcode]

This is just my preliminary idea.
I would like to run two subsystems with different step sizes in andes and write interface files to connect them together.

Multi-rate with different variables may be added later.

[cuihantao]

This is doable. You can create two systems objects and prepare input files accordingly. Your interface can control the time for data exchange and do the data exchange. This is known as the diakoptics, or the method of tearing. I’m not sure about different variables using multiple rates, but I’m happy to discuss if you provide more details. Get Outlook for iOS<https://aka.ms/o0ukef>

…

________________________________ From: xjlcode ***@***.***> Sent: Thursday, March 30, 2023 9:33:46 PM To: CURENT/andes ***@***.***> Cc: Hantao Cui ***@***.***>; Comment ***@***.***> Subject: Re: [CURENT/andes] Multi-rate simulation based on andes (Discussion #435) This is just my preliminary idea. I would like to run two subsystems with different step sizes in andes and write interface files to connect them together. Multi-rate with different variables may be added later. — Reply to this email directly, view it on GitHub<#435 (reply in thread)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ABSNZA7KYJWGLQW7UPXY3ODW6Y7AVANCNFSM6AAAAAAWLXDQP4>. You are receiving this because you commented.Message ID: ***@***.***>

[xjlcode]

Yes，jinning，I agree with your final statement. However, I have a few questions that I don't understand.
Why is not helpful for this multi-rate simulation with changing the initial value of bus voltage before PFlow is solved？
and the trouble of phase angle, I don't quite understand.

[xjlcode]

I have encountered a problem and do not know how to break through this interface technology.
If I directly change the voltage and phase angle of the node and run the "run" function again, the changes will be meaningless. Do I need to rewrite a new "run" function to start running from the changed voltage and phase angle?

[jinningwang]

It is because the variables are calculated from DAE.

If I directly change the voltage and phase angle of the node and run the "run" function again, the changes will be meaningless.

As Dr. Cui commented: "It is definitely doable, but that sounds like a large project. What specific goals do you want to achieve?"

Do I need to rewrite a new "run" function to start running from the changed voltage and phase angle?

[jinningwang]

Although you can change the initial values of voltage angles and magnitudes of the PV bus in TDS, these variables are then calculated by solving DAE.

Yes，jinning，I agree with your final statement. However, I have a few questions that I don't understand. Why is not helpful for this multi-rate simulation with changing the initial value of bus voltage before PFlow is solved？ and the trouble of phase angle, I don't quite understand.

[xjlcode]

oh, thanks, that I understand.
Then I have no idea about multirate simulation. Do you have some Unique viewpoints？

[jinningwang]

Hello @xjlcode

Does "two subsystems" mean separating one system into two systems in ANDES? If so, it will be two independent systems before interfacing them. In this way, I am curious about how to connect the two systems.

Regards,
Jinning

[xjlcode]

Yes, it will be two independent systems before interfacing them.
I am trying to make interface files to connect them. It is still in test.

Regards,
xjl

[jinningwang]

This can be a reference for further development related to multi-rate simulation. Welcome further discussion if interested in such implementation.

Overall idea

There have been some techniques to do the multi-rate simulation, such as DAE parallelism and note splitting.

A potential proof case to multi-rate simulation

When analyzing interconnected power systems, it is sometimes possible to simplify parts of the system by using Thevenin or Norton equivalents. This can be useful for specific scenarios, although it is not a true multi-rate simulation.

The model PLBVFU1 is a playback of voltage and frequency as a generator, and may be helpful for this proof case.

DAE in ANDES

In the code, the DAE is built as an attribute dae in the class system located in file system.py. DAE solving can be found in function run located in the file tds.py. To be more specific, the function run calls functions f_update and g_update.

For the DAE parallelism, I guess the implementation might need to restructure the DAE into several sub-DAEs. Then solve and interface them. However, this in-depth refactorization can be skipped in the proof stage.

@xjlcode Still, I would be happy to discuss potential implementation details if you clarify your desired results.

oh, thanks, that I understand. Then I have no idea about multirate simulation. Do you have some Unique viewpoints？

[xjlcode]

Sorry, jinning. I still don't understand “Although you can change the initial values of voltage angles and magnitudes of the PV bus in TDS, these variables are then calculated by solving DAE.”

For example, I change the initial values of voltage angles and magnitudes of the PV bus of Kundur system according to the results of IEEE14 system. Then, I run the TDS for about 3 times of Kundur on the basis of the initial values of IEEE14. When it finishes, I will run the TDS of IEEE14 on the basis of the voltage angles and magnitudes of Kundur.

Is this implementation infeasible with ANDES v1.8?

[jinningwang]

This is doable BEFORE solving PFlow in the Kundur system.

For example, I change the initial values of voltage angles and magnitudes of the PV bus of Kundur system according to the results of IEEE14 system.

This is doable.

Then, I run the TDS for about 3 times of Kundur on the basis of the initial values of IEEE14.

This is doable if create a NEW instance of 14-bus by changing the setting points before solving PFlow. I am afraid this is not doable if continue the TDS of 14-bus system, and we have discussed this as above.

When it finishes, I will run the TDS of IEEE14 on the basis of the voltage angles and magnitudes of Kundur.

[cuihantao]

@xjlcode We would better help you if you could, more clearly, define your problem and articulate how you plan to do it. I don't think I fully understand the multi-rate you described. Finishing one simulation and use its data as an input to the other doesn't seem multi-rate. My impression of multi-rate is that you have two parallel, ongoing simulations running at different step sizes and exchanging data at defined points.

Updating voltages or other algebraic variables will not take effect because most likely the provided voltages, combines with other variables, will not satisfy the DAE. All algebraic variables are immediately replaced by the solutions from the DAE in the subsequent step. If you need to enforce exact same voltages at the boundary buses, you will need a voltage source and might look into PLBVFU1 as Jinning proposed.