ROSCO Walkthrough Issues

[sautee]

Hello,

I'm trying to run the ROSCO walkthrough locally - and I came across a few issues.

1. Tuning the controller requires optional params to be defined
2. Plotting PI gains throw a dimension exception error
3. When set up locally, loading the OpenFAST turbine model throws an error

- Tune Controller Error

# Tune controller 
controller      = ROSCO_controller.Controller(controller_params)
controller.tune_controller(turbine)

# Necessary parameters
---> 66         self.U_pc = list_check(controller_params['U_pc'], return_bool=False)
     67         self.zeta_pc = list_check(controller_params['zeta_pc'], return_bool=False)
     68         self.omega_pc = list_check(controller_params['omega_pc'], return_bool=False)

KeyError: 'U_pc'

This section threw a number of missing key errors. The NREL 5MW YAML file has optional control parameters not defined in it (like 'U_pc'). But it looks like the controller.py file does not have handlers set if the key isn't present in the YAML and throws a missing key error. Errors were thrown for -

• U_pc
• max_pitch
• ss_vsgain
• WS_GS_n and others

Once I defined all the params required in controller.py, this cell executed successfully.

- Plotting PI gains throw a dimension exception error. It looks like the numpy ndarrays have different dimensions.

# Plot gain schedule
fig, axs = plt.subplots(1, 2, sharey=True)
axs[0].plot(controller.v[len(controller.vs_gain_schedule.Kp):], controller.pc_gain_schedule.Kp)
axs[0].set_xlabel('Wind Speed')
axs[0].set_ylabel('Proportional Gain')
axs[0].grid('True')

axs[1].plot(controller.v[len(controller.vs_gain_schedule.Ki):], controller.pc_gain_schedule.Ki)
axs[1].set_xlabel('Wind Speed')
axs[1].set_ylabel('Integral Gain')
axs[1].grid('True')

ValueError                                Traceback (most recent call last)
[<ipython-input-112-8158ec24a967>](https://localhost:8080/#) in <module>()
      1 # Plot gain schedule
      2 fig, axs = plt.subplots(1, 2, sharey=True)
----> 3 axs[0].plot(controller.v[len(controller.vs_gain_schedule.Kp):], controller.pc_gain_schedule.Kp)
      4 axs[0].set_xlabel('Wind Speed')
      5 axs[0].set_ylabel('Proportional Gain')

[/usr/local/lib/python3.7/dist-packages/matplotlib/axes/_base.py](https://localhost:8080/#) in _plot_args(self, tup, kwargs)
    340 
    341         if x.shape[0] != y.shape[0]:
--> 342             raise ValueError(f"x and y must have same first dimension, but "
    343                              f"have shapes {x.shape} and {y.shape}")
    344         if x.ndim > 2 or y.ndim > 2:

ValueError: x and y must have same first dimension, but have shapes (31,) and (30,)

• Error reading OpenFAST model: I used a local Anaconda installation to download and run ROSCO. However, now I come across a new error. when trying to load the NREL 5MW OpenFAST model.

# Load turbine data from openfast model
turbine = ROSCO_turbine.Turbine(turbine_params)
turbine.load_from_fast(path_params['FAST_InputFile'],path_params['FAST_directory'],dev_branch=True,rot_source='txt',txt_filename=path_params['rotor_performance_filename'])

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-20-671a7bf61fbd> in <module>
      1 # Load turbine data from openfast model
      2 turbine = ROSCO_turbine.Turbine(turbine_params)
----> 3 turbine.load_from_fast(path_params['FAST_InputFile'],path_params['FAST_directory'],dev_branch=True,rot_source='txt',txt_filename=path_params['rotor_performance_filename'])

/srv/conda/envs/notebook/lib/python3.7/site-packages/ROSCO_toolbox/turbine.py in load_from_fast(self, FAST_InputFile, FAST_directory, FAST_ver, dev_branch, rot_source, txt_filename)
    180         fast.FAST_InputFile = FAST_InputFile
    181         fast.FAST_directory = FAST_directory
--> 182         fast.execute()
    183 
    184         if txt_filename:

/srv/conda/envs/notebook/lib/python3.7/site-packages/ROSCO_toolbox/ofTools/fast_io/FAST_reader.py in execute(self)
    396     def execute(self):
    397 
--> 398         self.read_MainInput()
    399         self.read_ElastoDyn()
    400         self.read_ElastoDynBlade()

/srv/conda/envs/notebook/lib/python3.7/site-packages/ROSCO_toolbox/ofTools/fast_io/FAST_reader.py in read_MainInput(self)
    482         self.fst_vt['Fst']['DT_Out'] = float_read(f.readline().split()[0])
    483         self.fst_vt['Fst']['TStart'] = float_read(f.readline().split()[0])
--> 484         self.fst_vt['Fst']['OutFileFmt'] = int(f.readline().split()[0])
    485         self.fst_vt['Fst']['TabDelim'] = bool_read(f.readline().split()[0])
    486         self.fst_vt['Fst']['OutFmt'] = f.readline().split()[0][1:-1]

ValueError: invalid literal for int() with base 10: '"NRELOffshrBsline5MW_Onshore_ServoDyn.dat"'

Help!
Thanks.

[dzalkind]

Hi sautee,

Which branch/version of ROSCO are you using? Please try on the latest main branch and I will do the same.

The inputs you point out are filled with defaults if not provided in the yaml, but this feature was only added in recent months.

Best, Dan

[sautee]

Hi @dzalkind

I'm using the package downloaded with Anaconda/conda-forge - which seems to be v2.4.0. I used the instructions under https://rosco.readthedocs.io/en/latest/source/install.html.

I just noticed that the latest version at conda-forge is 2.5.1. Somehow conda install -y ROSCO gets v2.4.

I'll try to build v2.5.1 locally or force conda to pull v2.5.1.

To note, I was able to resolve most issues I encountered earlier. I just have the dimension exception error when trying to plot PI gains. I'll install v2.5 and let you know if this persists.

Thanks for the help so far!

[sautee]

Hi @dzalkind

I updated my local environment to ROSCO v2.5.1.

• I can confirm that the 'optional' tuning parameters are not really optional. The controller.py has additional optional params in v2.5.1, and they throws new missing key errors. Moreover, 2.5.1 requires you to setup 'filter params' as well, which if not setup in the YAML file, throw errors as well.

File ~\anaconda3\envs\rosco-env\lib\site-packages\ROSCO_toolbox\controller.py:125, in Controller.__init__(self, controller_params)
    122     self.F_LPFDamping = 0.0
    124 # Filter parameters
--> 125 self.f_we_cornerfreq        = controller_params['filter_params']['f_we_cornerfreq']
    126 self.f_fl_highpassfreq      = controller_params['filter_params']['f_fl_highpassfreq']
    127 self.f_ss_cornerfreq        = controller_params['filter_params']['f_ss_cornerfreq']

KeyError: 'filter_params'

• The issue with array dimensions persists in 2.5.1. The wind array has one extra element vs the gain array.

File ~\anaconda3\envs\rosco-env\lib\site-packages\matplotlib\axes\_base.py:498, in _process_plot_var_args._plot_args(self, tup, kwargs, return_kwargs)
    495     self.axes.yaxis.update_units(y)
    497 if x.shape[0] != y.shape[0]:
--> 498     raise ValueError(f"x and y must have same first dimension, but "
    499                      f"have shapes {x.shape} and {y.shape}")
    500 if x.ndim > 2 or y.ndim > 2:
    501     raise ValueError(f"x and y can be no greater than 2D, but have "
    502                      f"shapes {x.shape} and {y.shape}")

ValueError: x and y must have same first dimension, but have shapes (31,) and (30,)

Hope this helps.

[dzalkind]

Hi,

I made some updates to the jupyter notebook here: #155 to handle this new default parameter feature.

Generally, I'd recommend using the example scripts because they are tested during each commit and pull request. The examples cover everything in this notebook and more, and this notebook isn't part of the continuous integration, so we may take it out if we can't easily support it going forward.

I hope this helps.

Best, Dan

[sautee]

Thanks @dzalkind.

All previous errors are resolved. To note, I had to update my local wisdem package using conda update wisdem.

When running the 1-DOF simulation, I get the error -

FileNotFoundError: Could not find module 'C:\Users\andre\rosco-test\ROSCO\build\libdiscon.dylib' (or one of its dependencies). Try using the full path with constructor syntax.

Is there a build step I missed somewhere?

[sautee]

Update - I found the libdiscon.dll file under the site-packages folder of the conda environment. Once I pointed the notebook path appropriately, the 1DOF simulation ran successfully. Note - the default file name in the notebook libdiscon.dylib is for Mac environments.

I'm now at the OpenFAST simulation step. Here I get the error AssertionError: File, ../Test_Cases/IEA-15-240-RWT-UMaineSemi/IEA-15-240-RWT-UMaineSemi.outb, does not exists.

How do I generate an outb file? (I couldn't find info in readthedocs)

[dzalkind]

You can generate an outb file using openfast. An example can be found here. You may need to edit some paths to work on your system, like the default dynamic library (.dll). OpenFAST is good about error messages.

[sautee]

Thanks @dzalkind ! I'll give it a go and let you know how it goes.

[sautee]

When running example_06.py, I run into the following error.

Writing new controller parameter file parameter file: C:\Users\andre\rosco-test\Examples\DISCON.IN.
Using IEA-15-240-RWT-UMaineSemi.fst to run OpenFAST simulation
Running OpenFAST simulation for IEA-15-240-RWT-UMaineSemi.fst through the ROSCO toolbox...
Traceback (most recent call last):
  File "example_06.py", line 87, in <module>
    run_openfast(
  File "C:\Users\andre\anaconda3\envs\rosco-env\lib\site-packages\ROSCO_toolbox\utilities.py", line 525, in run_openfast
    subprocess.run([fastcall, os.path.join(fastfile)], check=True, cwd=cwd)
  File "C:\Users\andre\anaconda3\envs\rosco-env\lib\subprocess.py", line 493, in run
    with Popen(*popenargs, **kwargs) as process:
  File "C:\Users\andre\anaconda3\envs\rosco-env\lib\subprocess.py", line 858, in __init__
    self._execute_child(args, executable, preexec_fn, close_fds,
  File "C:\Users\andre\anaconda3\envs\rosco-env\lib\subprocess.py", line 1311, in _execute_child
    hp, ht, pid, tid = _winapi.CreateProcess(executable, args,
FileNotFoundError: [WinError 2] The system cannot find the file specified

It isn't clear to me which file it cannot find. Do I have to install OpenFAST separately to run this simulation? (I haven't installed OpenFAST yet).

[sautee]

FYI, I tried installing openfast in Windows, but it looks like the conda-forge openfast package is available only for MacOS and Linux environments.

[dzalkind]

I think you're right, you can only conda install openfast in mac/linux.

You can download executables from here, and insert the path to the executable in the example_06.py code.

[sautee]

Thanks @dzalkind. I downloaded the OpenFAST v3.2.0 'openfast_x64.exe' executable from the link, and used it in example_06.py like so

fastcall = os.path.join(this_dir,'openfast_x64.exe')
run_openfast(
  os.path.join(this_dir,path_params['FAST_directory']),
  fastcall=fastcall, 
  fastfile=path_params['FAST_InputFile'], 
  chdir=True
  )

It start to run the simulation, but it fails looking for the libdiscon.so file. Since I'm running in Windows, shouldn't it be looking for libdiscon.dll file instead? Here's the full error

Running InflowWind.
Running ServoDyn.
Running ServoDyn Interface for Bladed Controllers (using Intel Visual Fortran for Windows).

FAST_InitializeAll:SrvD_Init:BladedInterface_Init:The dynamic library
.\../../ROSCO/install/lib/libdiscon.so could not be loaded. Check that the file exists in the
specified location and that it is compiled for 64-bit applications.

 OpenFAST encountered an error during module initialization.
 Simulation error level: FATAL ERROR

 Aborting OpenFAST.

[sautee]

The error with OpenFAST is resolved. I asked here OpenFAST/openfast#1179.

You can change the path to the library in the ServoDyn input file. In your case, it is NRELOffshrBsline5MW_Onshore_ServoDyn.dat alongside your main OpenFAST input (.fst).

[sautee]

Yet another Example_06.py question! (this is becoming a long thread!)

The assert in example_06 fails for me.

# Now double Kp_float and check that it's passed through
Kp_float = -18
controller_params['Kp_float'] = Kp_float
controller_params['tune_Fl'] = False
controller      = ROSCO_controller.Controller(controller_params)
controller.tune_controller(turbine)
np.testing.assert_almost_equal(Kp_float,controller.Kp_float)

I get this error. I understand what it is, but I can't figure out why the Kp_float value does not go through. I removed the assert test block to run the example.

This is the error.

Loading wind turbine data for NREL's ROSCO tuning and simulation processeses
-----------------------------------------------------------------------------
-----------------------------------------------------------------------------
   Tuning a reference wind turbine controller using NREL's ROSCO toolbox
-----------------------------------------------------------------------------
Loading FAST model: NREL-5MW.fst
Loading rotor performace data from text file: C:\Users\andre\rosco-test\Test_Cases\NREL-5MW\Cp_Ct_Cq.NREL5MW.txt
Loading rotor performace data from text file: C:\Users\andre\rosco-test\Examples\../Tune_Cases\../Test_Cases/NREL-5MW/Cp_Ct_Cq.NREL5MW.txt
-----------------------------------------------------------------------------
   Tuning a reference wind turbine controller using NREL's ROSCO toolbox
-----------------------------------------------------------------------------
Traceback (most recent call last):
  File "example_06.py", line 61, in <module>
    np.testing.assert_almost_equal(Kp_float,controller.Kp_float)
  File "C:\Users\andre\anaconda3\envs\rosco-env\lib\site-packages\numpy\testing\_private\utils.py", line 599, in assert_almost_equal
    raise AssertionError(_build_err_msg())
AssertionError:
Arrays are not almost equal to 7 decimals
 ACTUAL: -18
 DESIRED: 0.0

[dzalkind]

It looks like you've changed the example to run the NREL-5MW and it's not a floating model. You can look at the code here for more information.

[sautee]

Thanks @dzalkind, clear.

I ran OpenFAST for the NREL 5.0 MW Baseline Wind Turbine (Onshore) for 60 seconds. In the simulation out file, the X, Y and Z wind velocities remain constant through the simulation - X at 8 m/s, Y and Z at 0 m/s.

Is there a way to generate changing wind speeds, or turbulent wind speeds in the simulation?

Here's what I get with example_08.py.

[sautee]

Think I figured it out.

• The simulation time is specified in the .fst file with a default of 60 seconds. This value can be changed as necessary
• The 'InFlowWind' file (in my case 'NRELOffshrBsline5MW_InflowWind.dat') determines simulated wind conditions - steady state, uniform, turbulent etc. In in the InFlowWind file, I changed the WindType to 2 and pointed the config to a time series wind field file. Example InFlowWind files are available here along with more detailed documentation here.