Bug with multi-dimensional Cp - Ct table

[RRiva]

Dear developers,

I'm trying to make a turbine model with a multi-dimensional Cp - Ct table.

The table is saved in csv format and looks like

      power_demand    ws        power  thrust_coefficient
0                5   3.0    98.705963            0.816387
1                5   3.1   108.909235            0.816387
2                5   3.2   119.792482            0.816387

Then, I define the model as

from floris.turbine_library import TurbineInterface

iea3_4 = TurbineInterface.from_turbine_dict(
    {
        "turbine_type": "IEA 3.4 MW OpenFAST",
        "rotor_diameter": 130.0,
        "hub_height": 110.0,
        "TSR": 8.16,
        "multi_dimensional_cp_ct": True,
        "power_thrust_table": {
            "ref_air_density": 1.225,
            "ref_tilt": 5.0,
            "cosine_loss_exponent_yaw": 1.88,
            "cosine_loss_exponent_tilt": 1.88,
            "power_thrust_data_file": "./LUT_IEA3_4MW_4floris.csv",
        },
    }
)

This code crashes in power_thrust_table.validator, with:

ValueError: power_thrust_data_file must be defined if multi_dimensional_cp_ct is True.

As far as I can see, the table is correctly read in _initialize_multidim_power_thrust_table, hence it seems to be a problem with the validator.

Furthermore, according to the documentation:

• power_thrust_data_file should be 1 level up, but that produces the same error, with the difference that the table is not read.
• The name and content of the columns should be different, but that also appears to be wrong.

My example is coherent with floris/turbine_library/iea_15MW_multi_dim_cp_ct.yaml.

Best regards,
Riccardo

[misi9170]

Hi @RRiva ,

Thank you for raising this issue. After looking into this, I believe the issue comes from how we handle the "multidimensional" indexing variables (in this case, the power_demand). The problem, as you said, comes from the validator; this line triggers the error. The bug comes in slightly upstream of that, I think in this line. What happens here is that the indexing variable is set to be the pandas index; but in your case with a single indexing variable power_demand, df.set_index(index_col.tolist()) returns an index that is of type int; whereas in our tests and examples, there are always at least two indexing variables (e.g. Tp and Hs) and so df.set_index(index_col.tolist()) returns an index of type tuple. Now, this line only checks for tuple types, so is skipped over in your example since the index is an int, causing the error to be raised.

This is indeed a bug: we should be able to handle "single" indexing variable cases such as yours. I'm going to add this issue to the work to do in #996 , which I plan to get back to working on later in the year.

In the meantime, I think you can work around this issue by adding another channel to your multidimensional table that always takes the same value, e.g. placeholder

power_demand,placeholder,ws,power,thrust_coefficient
0,5,0,3.0,98.705963,0.816387
1,5,0,3.1,108.909235,0.816387
2,5,0,3.2,119.792482,0.816387

Here is the modified csv file as an attachment: LUT_IEA3_4MW_4floris.csv

That's a bit of a hacky solution for now, but I hope will let you move forward with your work.

Best,
Misha

[RRiva]

Thanks! This works 🙂 now I'll have to figure out how to evaluate it 😅

[RRiva]

Calling iea3_4.plot_power_curve() works, but how to get 1 power curve for an arbitrary row in the multi-dimensional Cp - Ct table? From the documentation it appears that the way to do it is to define a wind farm with only 1 turbine.

import floris
import os

conf = floris.FlorisModel.get_defaults()
conf["name"] = "Simulation"
conf["description"] = "Basic model"
conf["flow_field"]["reference_wind_height"] = iea3_4.turbine.hub_height
conf["flow_field"]["wind_shear"] = 0.2
conf["flow_field"]["multidim_conditions"] = {}
conf["flow_field"]["multidim_conditions"]["power_demand"] = 2  # Set power demand here.
conf["flow_field"]["multidim_conditions"]["placeholder"] = 0
conf["farm"]["turbine_type"] = ["iea3_4_pywake_ccblade.yaml"]
conf["farm"]["turbine_library_path"] = os.path.dirname(__file__)    

conf["farm"]["layout_x"] = np.array([0.0])
conf["farm"]["layout_y"] = np.array([0.0])
fmodel = floris.FlorisModel(conf)

Unfortunately, fmodel.run() crashes in Farm.initialize(). Even if that will work, I still don't know how to set a different row of the table per findex and turbine.

I guess that for the time being I'll resort to using a basic turbine model and replace power_thrust_table as needed.

[misi9170]

Hi @RRiva ,

Generally speaking, it's easier to work with FlorisModel.set(). Here is some code that I think will work for you:

from floris.turbine_library import TurbineInterface
from floris import FlorisModel

import numpy as np
import matplotlib.pyplot as plt

iea3_4_dict = {
    "turbine_type": "IEA 3.4 MW OpenFAST",
    "rotor_diameter": 130.0,
    "hub_height": 110.0,
    "TSR": 8.16,
    "multi_dimensional_cp_ct": True,
    "power_thrust_table": {
        "ref_air_density": 1.225,
        "ref_tilt": 5.0,
        "cosine_loss_exponent_yaw": 1.88,
        "cosine_loss_exponent_tilt": 1.88,
        "power_thrust_data_file": "./LUT_IEA3_4MW_4floris.csv",
    },
}

# TurbineInterface is likely to be deprecated, and is mostly used for
# visualizing turbine data rather than passing into a FlorisModel.
iea3_4 = TurbineInterface.from_turbine_dict(
   iea3_4_dict,
)

#print(iea3_4)

fmodel = FlorisModel("defaults")
fmodel.set(
    layout_x=np.array([0.0]),
    layout_y=np.array([0.0]),
    turbine_type=[iea3_4_dict], # use the turbine dictionary directly
    reference_wind_height=-1, # Uses the turbine's hub height
    wind_shear=0.2,
)

# This is clunky, but works
fmodel_dict = fmodel.core.as_dict()
fmodel_dict["flow_field"]["multidim_conditions"] = (2, 0) # (power_demand, placeholder)

fmodel = FlorisModel(fmodel_dict)
print(fmodel.core.flow_field.multidim_conditions) # Confirm it is (2,0)

# Generate power curve by setting a range of wind speeds
n_points = 100
wind_speeds = np.linspace(0, 30, n_points)
fmodel.set(
    wind_speeds=wind_speeds,
    wind_directions=0.0*np.ones(n_points),
    turbulence_intensities=0.06*np.ones(n_points),
)
fmodel.run() # Could also use fmodel.run_no_wake() here, since not interested in wakes
powers = fmodel.get_turbine_powers() / 1000.0  # Convert to kW
print(powers.shape) # (n_points, n_turbines) = (100, 1)

fig, ax = plt.subplots(figsize=(8, 6))
ax.plot(wind_speeds, powers, label="IEA 3.4 MW")
ax.grid()
ax.set_xlabel("Wind speed [m/s]")
ax.set_ylabel("Power [kW]")

plt.show()

I'll make a couple of notes here that should be addressed in the eventual fix:

• The TurbineInterface cannot itself be passed to the turbine_type argument of .set(). In fact, TurbineInterface has a pretty limited use case, so will likely be deprecated, see Bug report: TurbineInterface not covered in tests and examples #901 .
• Generally, we recommend using .set(); but as you've found, there is not currently a way of setting the multidim_condition using set(). This is being added in [BUGFIX] Improve handling of multidimensional turbine conditions #996

Misha

[RRiva]

Thanks again 🙂 I've subscribed to the issue and PR, and will update my code accordingly.

[misi9170]

Hi @RRiva, I've now merged #996 into the develop branch, which I believe should address your issues. Feel free to try that when you get a chance and let me know if that is now working as you expect.

[RRiva]

Thanks a lot Misha! I'll update my code when you'll release Floris 4.5, since it will take me several hours.

[misi9170]

Closed by #966 and now available as of v4.5. See also a further upcoming enhancement allowing users to specify a different multidim condition for each inflow condition, #1152