IEA15MW - Changing from direct drive to geared turbine

[Carsten.Hansen]

My colleague Lisa and I are working on a 15MW turbine model and are facing some issues. For a better understanding and learning, we would like to make a thread on the topic (mainly how to adapt a model and switch from direct drive to a geared turbine).

-–

We want to do some load sensitivity studies for a generic model of a 15MW offshore wind turbine with a monopile foundation. As a baseline we started with the IEA-15-240 Model and the instabilities started when switching to a geared drive train. When we introduce a gear ratio of 50, we run into implications and instabilities with the controller. It is not fully clear to us which parameters are influenced by the generator speed and/or rotor speed (which is unchanged for now) and where the control instabilities may arise from.

We are using:

• Openfast v412
• Rosco v297

Steps taken so far:

1. Baseline was the IEA-15-240 Model ( IEA-15-240-RWT/OpenFAST/IEA-15-240-RWT-Monopile at master · IEAWindSystems/IEA-15-240-RWT · GitHub )
2. Set the gear box ratio on 50 in elastodyn and decreased the generator inertia (see list of values below).
3. Changed the following parameters in the DISCON file (see list of changes below). All speed related values were multiplied by 50 and torque related values divided with 50. Also tried to optimize with the rosco toolbox.

Status:

• Simulating a wind ramp with the original IEA-15-240 model shows only small indications of instabilities on speed and power but large instabilities in the adjusted model with the gearbox
• Simulating the adjusted model with the gearbox with turbulent wind is failing due to instabilities in the model.

Images of the wind ramp follow.

Default model:

Model with gearbox: (I can only add one image here. So it will follow in a next post)

#####

Changes made:

• Elastodyn
  • GBRatio changed from 1 to 50
  • GenIner changed from 1836784 to 735 (divided by 50^2)

DISCON file

• FILTERS

  • F_LPFCornerFreq (multiplied with 50)
  • F_VSRefSpdCornerFreq(multiplied with 50)

• BLADE PITCH CONTROL

  • PC_GS_KP (divided by 50)
  • PC_GS_KI (divided by 50)
  • PC_RefSpd (multiplied with 50)

• VS TORQUE CONTROL

  • VS_ArSatTq (divided by 50)
  • VS_MaxRat (divided by 50)
  • VS_MaxTq (divided by 50)
  • VS_MinOMSpd (multiplied with 50)
  • VS_Rgn2K (divided by 50^3 - showed best results)
  • VS_RtTq (divided by 50)
  • VS_RefSpd (multiplied with 50)
  • VS_KP (tuned with Rosco toolbox, roughly divided by 30)
  • VS_KI (tuned with Rosco toolbox, stayed the same)

• FIXED PITCH REGION 3 TORQUE CONTROL

  • VS_FBP_Omega (multiplied with 50)
  • VS_FBP_Tau (multiplied with 50)

• WIND SPEED ESTIMATOR

  • WE_GearboxRatio (obviouslyset to 50)

Could someone please help and indicate, where the instabilities may arise from?

[Carsten.Hansen]

This is the result of the geared turbine:

[Daniel.Zalkind]

Hi Carsten,

I would see what ROSCO Toolbox does independently, compared to your manual changes; it should change these gains consistently, similar to yours.

It looks like your torque control gains are too high. You can reduce omega_vs in the tuning yaml (ROSCO toolbox tuning).

I hope this helps.

Best, Dan

[Carsten.Hansen]

Hello Daniel, thanks for you great advise. We made a new run and put some trust in the tuning process. There were some parameters added as new lines and some were removed. We took all the changes from the tuning process and the match was perfect. Basically the lines were identically and we had to check on the last digits too see that it was not exactly the same (and a issue in our code).

So great process on the gear box and the rosco toolbox. Amazing.

Our next step was to adjust other parameters to the model. Like a increased rotor speed, different geometry for tower and blades. Slightly different gear box ratio. Now the pitch controller was acting too fast in the step wind, causing failing simulations. After reducing the omega_pc, we could adjust the response.

Looking good so far:

But something is going wrong for turbulent simulations, we don’t see in the steady simulations. Plot will follow.

Any suggestions what is causing this and how to get rid of this?

We tried different steps of: increase zeta_vs and reducing omega_vs, but this ramp up is seen in all turbulent simulations and for all wind speeds.

Best regards, Carsten

[Carsten.Hansen]

Plot of turbulent responses: