Wind turbine curtailment via ROSCO

[lorenzoschena]

Hello there,

I was wondering if it is possible to employ ROSCO for power curtailment purposes.

A first, naive, way to tune the controller to this end might be to modify the V_rated of the turbine model to be used? Is there any drawback that you can see in this approach?

If yes, can you please point me in the right direction?

Many thanks,

[dzalkind]

Hi,

Modifying v_rated in the tuning_yaml will not actually help you de-rate. v_rated is used to separate the range of wind speeds used in the automated tuning process. It's a common misunderstanding about the input we should probably clarify better.

I would change the rated_rotor_speed and/or rated_power. Together, those will determine the rated torque, and the ROSCO tuning process should figure out the proper gains and set points from that.

I hope this helps.

Best, Dan

[dzalkind]

I'm going to convert this to a discussion because it's not really an issue with the code, but perhaps others would like to contribute or learn more.

[lorenzoschena]

Hi,
In the context of the discussion above, I was wondering if it was possible to switch between different de-rated controllers during a simulation. From a quick look into the source code, the DLL appears to be loaded statically at the initialization of the simulation, making an "hack" in this direction quite challenging.

Does ROSCO by any chance support this change, maybe trough ZMQ? If not, do you have any suggestions in this sense?

Best,
Lorenzo

[dzalkind]

Hi Lorenzo,

Switching between different controller libraries in a single simulation is not possible at this time. There is a need for this in controller testing environments on deployed turbines, so it's on our radar. I think the change would need to happen on the OpenFAST side, or at least outside of ROSCO. I don't think it would be too difficult of a hack, but you will need to be careful about how the controllers initialize. There may be some updates needed on the ROSCO side to ensure that the library can be loaded in the middle of a simulation.

You may be interested in a new feature we've added to control the power output of a turbine. It's on the develop branch, and you can change the power rating (fraction) with a few different methods, including a ZeroMQ input: #375

I hope this helps.

Best, Dan

[lorenzoschena]

Hi Dan,

Thanks a lot for the quick response. Indeed, the new functionality of ROSCO seems to be exactly what I was looking for, I am trying it right away.

Is the WISDEM version changed with respect the main ROSCO requirements? Because I'm getting:

  File "/Users/lorenzoschena/miniconda3/envs/rosco-env-tmp/lib/python3.10/site-packages/rosco/toolbox/ofTools/case_gen/run_FAST.py", line 31, in <module>
    from wisdem.commonse.mpi_tools              import MPI
ModuleNotFoundError: No module named 'wisdem.commonse.mpi_tools'

Even though:

Name: wisdem
Version: 3.18.0
Summary: Wind-Plant Integrated System Design & Engineering Model
Home-page: https://github.com/WISDEM/WISDEM
Author: 
Author-email: NREL WISDEM Team <systems.engineering@nrel.gov>
License: Apache-2.0
Location: /User/miniconda3/envs/rosco-env-tmp/lib/python3.10/site-packages
Requires: dearpygui, jsonschema, marmot-agents, moorpy, numpy, openmdao, openpyxl, pandas, pydoe3, python-benedict, pyyaml, ruamel.yaml, scipy, simpy, sortedcontainers, statsmodels
Required-by: rosco

Best,
Lorenzo

[lorenzoschena]

Follow-up

leaning on Openmdao instead of wisdem in run_FAST works, e.g. substituted from wisdem.commonse.mpi_tools import MPI with from openmdao.utils.mpi import MPI in imports.

[dzalkind]

Ah, thank you for the reminder. That is the recommended fix.

We should do a hotfix into main. I opened an issue #399