Yaw optimization with disabled turbines

[MiguelMarante]

Hello everyone,

I want to implement yaw optimization for a time-series, taking into account that different turbines are not operational at different time records.

Below is an example for a time-series with two records: the mid-stream turbine is disabled in the first record, and the upstream turbine is disabled in the second. The results show that this information is ignored in the optimization procedure.

Is there a workaround to ensure these turbines are completely excluded from the optimization?

Note: I am using Floris v4.1 in this example.

Best regards,
Miguel Marante

CODE:

import numpy as np

from floris import FlorisModel
from floris.optimization.yaw_optimization.yaw_optimizer_sr import YawOptimizationSR

# Create floris object
yaml_file = 'inputs/emgauss.yaml'
fmodel = FlorisModel(yaml_file)

# Set wind conditions
fmodel.set(wind_directions=[270.]*2, wind_speeds=[8.]*2, turbulence_intensities=[.06]*2)

# Disable turbines
fmodel.set_operation_model('mixed')
fmodel.set(disable_turbines=np.array([[False, True, False], [True, False, False]]))

# Run optimization and print results
yaw_opt = YawOptimizationSR(fmodel)
df_opt = yaw_opt.optimize()
print(df_opt)

[misi9170]

Hi @MiguelMarante ,

Thank you for raising this and sharing your example code. This is a bug, to do with the power_setpoints (which are set to near zero when turbines are disabled) not being correctly applied on a copy of the FlorisModel created within the yaw optimization routine.

I'll try to work on a bugfix in the coming days. In the meantime, you may be able to create a workaround by using a combination of the turbine_weights and minimum_yaw_angle and maximum_yaw_angle keyword arguments; however, you'll then have to run a single yaw optimization per time step, which will be clunky and slow.

Again, thanks for bringing this to our attention. I hope we'll be able to provide a suitable fix reasonably quickly.

Misha

[MiguelMarante]

Hi @misi9170,

Thank you for the quick response and for looking into this. Do you think this fix will come in the next release? And if so, do you have a prevision for when that will be?

Best regards,
Miguel Pereira

[misi9170]

Hi Miguel,

I'm working on a fix now and hoping to have something working by the end of the day. That will be a new pull request, and will hopefully be merged into the develop branch in the coming days. Then, depending on the status of a couple of other issues/pull requests, we are hoping to get a new patch release of FLORIS (4.2.1) done by the end of next week or possibly sooner.

You'll hopefully be able to try things out anywhere in that process once I open the PR---either on the PR branch, once it's in the develop branch, or once it makes it into main and v4.2.1 is released.

I'll update this comment with progress on that.

Cheers,
Misha

---

UPDATE: I have now opened a fix for this in #1027