MATLAB Answers

How do I simulate a microgrid to which is applied reinforcement learning using Simulink and Simscape?

15 views (last 30 days)
I have to simuate using Simulink and Simscape a microgrid to which is applied reinforcment learning. I am having troubles on finding a starting point to this since it is a new field and I don't find some examples on this topic.
I am trying to take inspiration from the following paper: but I don't have to simulate this, I am just using it as a guideline to follow.
I have found some videos made by Mathworks on how to build a microgrid, but they use some blocks the have build and that are not on the blockset, and moreover I don't know hot to apply reinforcment learning to the microgrid.
Can somebody help me understand how to operate and/or link me some material that can help me understand how to do this simulation using Simulink and Simscape?
Thanks in advance.


Sign in to comment.

Accepted Answer

Tony Lennon
Tony Lennon on 20 Apr 2020
I'll pass this along to some people who work in reinforcment learning. In the meantime, may I suggest that you look at these resources?
Use Simulink and Simscape™ models to represent an environment. Specify the observation, action, and reward signals within the model.


Flavio Clarizia
Flavio Clarizia on 23 Apr 2020
Thanks for answering. Isn't there any practical tutorial or example of application of reinforcement learning for energy management?
To understand better how to start my work I am looking at this tutorial by Mathworks:
but it does not explain a lot on how to design a microgrid, since it focuses more on how to use Simscape, which probably is the reason for the video.
And then, even if for starting I use a similar model of this, I have no idea on how to apply to it reinforcement learning.
Do you have any good source or suggestion to give me?
Thanks again.
Emmanouil Tzorakoleftherakis
Hi Flavio,
We do not have an example specific to microgrid as of now. Reinforcement Learning Toolbox is still a new product and there is a lot of active work happening, including new example development. This is why I suggested the resources above, because there is a standard high-level recipe you follow every time you apply RL to a new problem. Sure some specifics like reward signal design and hyperparameters change across problems, but the template you follow is the same (environment->reward->policy+agent->training). The examples link above would be a more hands-on way to see how to set up a problem for training.

Sign in to comment.

More Answers (0)

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!