OpenFast vs. FastFarm loads for identical inflow

Dear Community,

I am new to the forum and very happy to meet you. I’m Fred, currently pursuing a Master’s degree in Aerospace Engineering and was drawn into a fascination with wind turbines through a course in aeroelasticity.

To the question: What can cause loads to deviate from OpenFast to FastFarm environment for identical initial conditions and inflow?

I am performing simulations with a 3 turbine layout in FastFarm. Recently, i wanted to quickly cross check some simulations with a simplified OpenFast model.
Initial conditions for both OpenFast and FastFarm are identical, the same inflow (created with the Kaimal spectrum in Turbsim - turbulence box set up in accordance with Farm dimensions) is applied and the damage equivalent loads are calculated over a 10 minute period.
I expected the loads from the OpenFast simulation to be identical to the loads from the first wind turbine in FastFarm, yet I observe substantial deviations of 3-4% for blade, shaft and tower top bending DELs, and very large deviations for the tower base moment DELs. It seems like FastFarm tends to underestimate the loads compared to OpenFast (I have read in Shaler et al 2019 that this could be an indication of improper spatiotemporal discretisation of my FastFarm setup)

Has this been observed before? What could be the root cause of this?

Is it due to smoothening of the incoming wind caused by interpolation in the low or high resolution domain in FastFarm, hence improper spatiotemporal discretisation? Yet i would expect, that if no wakes are presented (as I am just checking this for the first turbine), the time series of the 2D-turbulence grid should be imprinted equally onto the rotor disc in OpenFast and FastFarm. Or is the time series imprinted onto the “ambient wind & array effects” model and then propagated and possibly interpolated according to the FastFarm adjustments of the low resolution domain?

I appreciate your help and am thankful for any input to the matter.

Dear Fred,

This sounds like a problem associated with interpolation of the ambient flow field. If the high-resolution domain does not align with the grid in TurbSim (in X, Y, and Z), then the TurbSim data will be interpolated to the high-resolution domain grid points. This interpolation will smooth out the ambient turbulence a bit, resulting in lower fatigue loads. For this reason, the Modeling Guidance section of the FAST.Farm documentation ( Modeling Guidance — OpenFAST v3.2.0 documentation) recommends to set up the TurbSim grids and domains in FAST.Farm so that the grid points are aligned. And Python scripts are provided to aid in the inflow generation. We are in the process of improving those scripts to make them user friendly.

Best regards,

1 Like

Dear Jason,
thank you very much for your quick response.

If i may take the opportunity to ask an additional question:
The user guide suggests to align dX and dT through dX=V_advect*dT.

If i plan to perform an uncertainty analysis over my input parameter space, does best practice then require to always adapt the FastFarm discretisation in x-direction accordingly to changing wind speeds, in Order to provide best comparability of different input scenarios?

Thanks a lot, and I await with excitement the updated modelling guidelines.


Dear @Fred.Segnitz,

Yes, your understanding is correct. For optimal usage of FAST.Farm, it is ideal to change the discretization of the low- and high-resolution domains when changing the ambient mean wind speed.

Best regards,

1 Like