Power Output Differences Between FLORIS and OpenFAST at Yaw Angle 22°

[Xavier-shenxinghui]

Hi everyone,

I’ve been working with OpenFAST to verify results against FLORIS and have come across an misallignment in power outputs. I performed a yaw angle sweep from -30° to +30° in FLORIS with an incoming wind speed of 12 m/s, and found that the turbine reaches full power (5 MW) when the yaw angle is set to 22°.

However, when I perform a similar setup in OpenFAST: with all platform dynamics (surge, sway, heave, roll, pitch, yaw) disabled, the incoming wind speed set to 12 m/s (steady wind type), and the final nacelle yaw angle fixed at 22°, the power output I get is around 4.3 MW.

I’m curious to know what might be causing this difference. Is it something related to the way FLORIS handles yaw misalignment or perhaps an issue with my OpenFAST setup? I might have overlooked something or made a simple mistake, so any insights or suggestions would be greatly appreciated!

Thank you very much for your suggestion in advance!

PS: Please forgive any basic errors I might have made in my setup.

The following is my python code:

from floris import TimeSeries
import numpy as np
import matplotlib.pyplot as plt
from floris.core import rotor_velocity
from floris import FlorisModel

fmodel = FlorisModel("gch.yaml")
fmodel.set_operation_model("cosine-loss")
fmodel.set(layout_x=[0.0], layout_y=[0.0])
fmodel.set(
wind_data=TimeSeries(
wind_speeds=np.ones(100) * 12.0,
wind_directions=np.ones(100) * 270.0,
turbulence_intensities=0.06
)
)
fmodel.reset_operation()

yaw_angles = np.linspace(-30, 30, 100)
fmodel.set(yaw_angles=yaw_angles.reshape(-1,1))
fmodel.run()

powers = fmodel.get_turbine_powers()/1000

My result gained from Floris

My result gained from OpenFAST

[misi9170]

Hi @Xavier-shenxinghui ,

Thanks for the question. First of all, I think your python code looks great!

As you are probably aware, FLORIS uses a "cosine exponent" to model power loss due to yaw. In particular, the cosine_loss_exponent_yaw value determines how quickly power drops off due to a yaw misalignment. The default for the cosine_loss_exponent_yaw is 1.88 based on our experience with various wind tunnel and field experiments (however, notably, this is not fit directly to OpenFAST data). Looking at your OpenFAST simulations, it looks like the OpenFAST model of the NREL 5MW might drop below 5MW when the yaw misalignment is approximately 17 degrees. You can probably generate a similar result in FLORIS if you increase the value of cosine_loss_exponent_yaw on the NREL 5MW model to achieve a closer match (I find that setting it at around 3.0 achieves full power at around 17 degrees of yaw misalignment). I will point out that a value of 3.0 is an exact match for the "basic" actuator disk model with a geometric yaw modification, but is higher than what has been reported in field studies.

However, I'm not myself sure how OpenFAST models power loss due to yaw misalignment and how realistic/well-verified that is. I'm tagging @dzalkind, @andrew-platt, and @jjonkman here as they may be able to answer that better.

One last point so that you're aware: FLORIS has wind shear by default, which means that the rotor average wind speed value will be lower than the hub height wind speed value. If you'd like to ensure that 12 m/s is the rotor average wind speed, you can provide wind_shear=0 in your call to fmodel.set().

Cheers,
Misha

[jjonkman]

Dear @Xavier-shenxinghui and @misi9170,

Our TORQUE 2024 paper explains improvements we made in OpenFAST to avoid having the cosine_loss_exponent_Yaw be around 3.0, which it was in older versions: https://docs.nrel.gov/docs/fy24osti/88162.pdf. These improvements are included in OpenFAST as of v4.0 and newer when BEM_Mod = 2, Skew_Mod = 1, and SkewMomCorr = TRUE in AeroDyn, which we now generally recommend. I'm not sure if you are using these AeroDyn settings or not in your results.

Best regards,

[Xavier-shenxinghui]

Dear @misi9170, @jjonkman, and all,

Thank you very much for your helpful responses and detailed explanations!

I’ve tested cosine_loss_exponent_yaw = 3.0 in FLORIS along with wind_shear = 0, and I observed a noticeable power drop when the nacelle yaw angle exceeds around 17°.

On the OpenFAST side, I found that I do not have BEM_Mod or SkewMomCorr in the AeroDyn input file, likely because I am using OpenFAST v3.4. I set Skew_Mod = 1 and fixed the final yaw angle at 17°. (I gained a generated power of approximately 4.7MW)

By adjusting the cosine_loss_exponent_yaw and the input file, I achieved better alignment between the results from the two models. Based on your feedback and the information you kindly shared, I think the mismatch is due to the difference in cosine_loss_exponent_yaw settings due to the difference in versions — I'm using an older version of OpenFAST and the latest version of FLORIS.

I will change the cosine_loss_exponent_yaw back to 1.88, as it is based on field studies, as mentioned. I truly appreciate the insights and support you've provided!

Best regards,