Implementing a new variable that changes dynamically in simulation

[TimThacker]

I am attempting to implement a coordinating signal to a governor model in ANDES, where the coordinating signal changes dynamically during TDS.

Specifically, I am attempting to modify the governor speed reference in the IEESGO governor model from omega_ref to (omega_ref + uc) where uc is the coordinating signal.

I believe the implementation would require the creation of a new variable and some way to define that variable. I am struggling with both the variable implementation and understanding how to properly update this variable in TDS. Ideally the signal would be proportional to the power output of a renewable energy source (DFIG wind turbine, etc) however I would be happy with a simple proof of concept where at some time it changes value discretely.

Where should I begin?

[cuihantao]

Hi Tim,

Changing the speed reference of a turbine governor model depends on the specific model. For IEEGSO, looking at https://www.powerworld.com/WebHelp/Content/TransientModels_PDF/Generator/Governor/Governor%20IEESGO.pdf, we can tell that it provides the paux signal as an auxiliary input.

In the ANDES implementation of IEESGO, paux is inherited from the TGBase class. The definition of paux is found here:

https://github.com/cuihantao/andes/blob/901a4fb7cda336ab3adec2cd47b63902e0ce803c/andes/models/governor/tgbase.py#L108-L112

To update paux, follow this example at https://docs.andes.app/en/latest/_examples/ex8.html. Whenever you update paux0, make sure write to its v field in place. See https://docs.andes.app/en/latest/_examples/ex2.html.

The IEESGO model does not provide a way to update the speed reference, but I can point you to a feasible solution if paux is not what you are looking for.

[TimThacker]

Thank you for the quick response. Unfortunately, paux is not what I am looking for.

I am attempting to add a new variable to the first differential equation representing the IEESGO model. The addition of the coordinating signal uc can be seen in equation 6 here: https://arxiv.org/ftp/arxiv/papers/2009/2009.07890.pdf
In this application, a reduced version of the IEESGO governor is used, where K2=K3=0. I have already created and tested this reduced model in a fork of ANDES.

You note that changing paux is a way one can input from external programs such as reinforcement learning. I am attempting a similar task, but through a modification of the model - the modification representing a coordinating signal.

To properly implement this, would I begin by adding uc as an attribute in the TGBase class?

Thank you for your help, I have been working with MATLAB recently and I am not as familiar with Python, however Python has more opportunities for the application of reinforcement learning to power systems.

[cuihantao]

In the case where you need to modify the speed reference, change the line here:

https://github.com/cuihantao/andes/blob/901a4fb7cda336ab3adec2cd47b63902e0ce803c/andes/models/governor/ieesgo.py#L70

to u='ue * (uomega0 + omega - wref)',, where uomega0 is a ConstService initialized to 0. uomega0 is basically the uc, but I named it in a way to denote a ConstService that modifies omega.

uomega0 will need to be updated during your simulation loops.

[cuihantao]

You might add domega0 to TGBase, but in the future, turbine governor model will need to be updated individually to include this constant. I haven't seen a lot of cases where the speed reference is updated; more often, the power reference is updated.