HYBRID MODEL TESTING OF FLOATING WIND …...2 Outline 1. Context : Wave tank hybrid testing of FWT 2. Specifications of actuators setpoint 3. Test bench presentation 4. System identification

HYBRID MODEL TESTING OF FLOATING WIND TURBINES: TEST BENCH FOR SYSTEM

IDENTIFICATION AND PERFORMANCE ASSESSMENT

Vincent Arnal, Jean-Christophe Gilloteaux, Félicien Bonnefoy, Sandrine Aubrun

LHEEA, Ecole Centrale de Nantes, France

OMAE 2019-96374 Vincent Arnal - 12/06/2019 1

2

Outline

1. Context : Wave tank hybrid testing of FWT

2. Specifications of actuators setpoint

3. Test bench presentation

4. System identification and performances

OMAE 2019-96374 Vincent Arnal - 12/06/2019

3

Wave tank testing of floating wind turbines


Drag diskGeometry

scaled rotorThrust matched

scaled rotor

100% physical

MeanThrustforce

Unsteadyaerodynamicforce

Test of Control strategies

Actuatorresponsetime

Scalability to large rotor (10-15 MW)

Porous Disk + + - - - - -

Geometry scaled rotor + - - ++ -

Thrust Match rotor ++ - - ++ -

Wind

Reynolds

Wave

Froude Hybrid with actuators

4

Wave tank hybrid testing of FWT : principle

Physic → Numeric

Positions and velocities and accelerations Platform and tower

Physic ← Numeric

Rotor Force (Aero, Gyro, Inertia...)

Full scale :Simulation of a wind turbine :Wind + Aero + Servo + Elastic

Model Scale:Waves + Platform + Tower + Moorings+ Actuators

Motions

RNA Force

©SINTEF Ocean (Thys, 2018)


Numerical tool : Realistic rotor

loads Real-time Similar to fully

coupled models

Actuators :Accuracy Steady state

performance High bandwidth

Physical Numerical

5

Wave tank hybrid testing of FWT : actuators


SINTEF Ocean 6 cables(Chabaud, 2016);(Sauder et al., 2016);...

SOFTWIND Project

Focus on control law influence during wave tank testing campaign

On-land

IH Cantabria 6 on-board fans(Urbán and Guanche, 2019); (Battistella et al., 2018)

On-board

Bandwidth ??

𝟏𝑷 𝟑𝑷

𝑇𝑤𝑟

Blades

6

Specifications of actuators setpoints: frequencies


0 0,04 –0,05 [Hz]Turbulence 0,3

𝑿𝟑

𝑿𝟏 & 𝑿𝟐

𝑿𝟒 & 𝑿𝟓

Wave-induced hub velocity

𝑿𝟔

0,2 1-20,6

Tower shadow

𝑿𝟑 TLP,𝑋4& 𝑋5 𝑇𝐿𝑃

𝑿𝒊 = natural frequency of Platform DoF 𝑖

Highest frequency captured ?

Main questions :• Relative contributions of the different frequencies of

interest?• Required dynamics of the actuators ?• Mean and Max forces ?

7

Specifications of actuators setpoints : methodology


OC4 Semisubmersible OC3 Hywind Spar Triple Spar

NREL-5MW NREL-5MW DTU-10MW

Selected floating wind turbines : Numerical model :OpenFAST v1.0 with:• Rigid blades• Flexible tower (modal)• Active controlerFree access input files

Load cases considered :Norm Design Load Cases of type 1.X (power production) Severe waves Normal Turbulence Model Wind-wave misalignment

8

Specifications of actuators setpoints : outcome


𝐹𝑥Wave frequencies ++Tower bending mode and 3P : +

• Q: Relative contributions of the different frequencies of interest?

𝑂𝑡ℎ𝑒𝑟 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡𝑠 ∶ Spread on the different frequencies

LF X_5 WF Twr & 3P > 3P

f_low [Hz] 5,0E-03 2,5E-02 4,0E-02 3,3E-01 7,0E-01

f_high [Hz] 2,5E-02 5,0E-02 3,3E-01 7,0E-01 1,5E+00

9

Inflow, Aero, Servo, Structure

OpenFAST code

On boardcontroler

Actuator

ESC

Actuator command

Host PC

Communication

Mast


Balance

Motion Capture system

Accelerometer Data acquisition

Data acquisition

Wave tank Dry

Floater

TEST BENCH : Overview

Envisaged system in the wave tank

10

Inflow, Aero, Servo, Structure

OpenFAST code

On boardcontroler

Actuator

ESC

Actuator command

Host PC

Communication

Mast

6 DoF Hexapod


Balance

Motion Capture system

Accelerometer Data acquisition

Data acquisition

TEST BENCH : Overview

11OMAE 2019-96374 Vincent Arnal -

12/06/2019

TEST BENCH : Validation of the methodology

Overall validated methodology1. Realistic FWT motions 2. Motions reproduced by Hexapod3. Motions capture4. Force computed in real-time by integrated

numerical code5. Actuator commanded to reproduce this force

Success Indicators :• The real-time computed force corresponds

to the load case we are reproducing.• The actuator reproduces with a sufficient

accuracy the setpoint force.

𝐻𝑠[m]

𝑇𝑝[s]

𝑈𝑤[m/s]

Mean thrust

[kN]

LC1 3 5 11.4 680

LC2 4 5 18 320

LC3 6 10 11.4 680

LC4 7 10 18 320

LC5 7 17 11.4 660

LC6 8 17 18 320

OC3 Hywind Spar – 5MW scale 1:30• Rigid blades and tower• Active controler

Name of FWT OC4_Semi OC3_Spar TripleSpar_10MW

Representative

picture

[14] [17] [15]

Wind Turbine NREL-5MW [18] NREL-5MW [18] DTU-10MW [19]

12

STATIC CALIBRATION ACTUATOR

ESC Turbine

Thrust [N]

Du

tyR

atio

[%]


Throttle or Duty ratio 𝛽[%]

Thrust 𝑇[N]

𝑻 = 𝑲 𝜷 ∗ 𝜷


12/06/2019

TEST BENCH : Performance / static calibration

3P

waves

Load CaseMeanThrust (N)

Error (%)

LC1 25 3LC2 12.5 5LC3 25 5

LC4 12 10

LC5 24 4LC6 12 5

Scale 1: 30

Results OK but higher bandwidth preferred

Rising steps


12/06/2019

𝑻𝒂𝒄𝒕 𝒔

𝜷 𝒔= 𝑯 𝒔 ≈

𝑲 (𝑻𝒎𝒆𝒂𝒏)

𝟏+𝝉 𝑻𝒎𝒆𝒂𝒏 𝒔𝒆−𝑻𝒅𝒆𝒍𝒂𝒚𝒔𝑻𝒎𝒆𝒂𝒏 [N] a working point

TEST BENCH : Identification

Pure delay

White noise


12/06/2019

0

0.05

0.1

0.15

0.2

0.25

0.3

0 20 40 60τ

[s]

Thrust [N]

Time characteristic as a function of mean thrust

𝑻 𝒔



𝟏+𝝉 𝑻𝒎𝒆𝒂𝒏 𝒔𝒆−𝑻𝒅𝒆𝒍𝒂𝒚𝒔

𝑻𝒎𝒆𝒂𝒏 [N] a working point

TEST BENCH : Identification

White noise

16

TEST BENCH: Dynamic calibration

ESC Turbine

Throttle or Duty ratio 𝛽[%]

Thrust 𝑇[N]




𝟏+𝝉 𝑻𝒎𝒆𝒂𝒏 𝒔𝒆−𝑻𝒅𝒆𝒍𝒂𝒚𝒔𝑻𝒎𝒆𝒂𝒏 [N] a working point

Identification :

𝜷𝒅𝒚𝒏 𝒔 = 𝑻𝒔𝒆𝒕 𝒔 𝑯−𝟏 𝒔

Inverse dynamics

Static Dynamic


≈ 𝑻𝒔𝒆𝒕 𝒔𝟏

𝑲 𝑻𝒎𝒆𝒂𝒏𝟏 + 𝝉 𝑻𝒎𝒆𝒂𝒏 𝒔 𝒆𝑻𝒅𝒆𝒍𝒂𝒚𝒔 Not

implemented


𝑻𝒔𝒆𝒕 𝒔≈ 𝟏 ?

17

TEST BENCH: Dynamic calibration performance

OC3 Hywind Spar – 5MW at scale 1:30Waves conditions : Severe Sea State, 𝐻𝑠 = 8.6𝑚 𝑇𝑝 = 13𝑠;

Wind conditions : Normal Turbulence Model, 𝑈𝑤 = 18𝑚. 𝑠−1; 𝑇𝐼 =14.6%;

3P

Waves


Representative

picture

[14] [17] [15]


30 ms delay Highly encouraging resultsOMAE 2019-96374 Vincent Arnal - 12/06/2019

18

Conclusion

Parametrical study with numerical simulations 1. Identification and quantification of relative importance of frequencies of interest. It is important to have 3P and 1st tower mode correctly reproduced

Test Bench identification2. Validation of communication protocol and overall methodology Real-time execution, motion and force observers design

3. Identification of actuator transfer function, delays, ... Good match between white noise and response to steps

4. Improvement of actuator command The inverse dynamics / dynamic calibration methodology gave significantly betterresults compared to static only



12/06/2019

Next steps

• Wave tank tests with 1 actuator in 09/2019 on 1:40 multipurpose platform(Blue Growth Farm project) and in 10/2019 on a 1:40 10MW SPAR with 1 actuator

• Calibration and identification of a more advanced emulation system• Wave tank tests with advanced emulation system in ~ 05/2020 on a 1:40

10MW SPAR Focus on control law influence

• Flexible blades modelling• Improve quality of dynamic calibration• Use braking of the turbine to improve decelerating performances

THANK YOU FOR YOUR ATTENTION


21

Specifications of actuators setpoints: background


𝐹𝑠𝑒𝑡 =

𝐹𝑎𝑒𝑟𝑜 𝑥𝐹𝑎𝑒𝑟𝑜 𝑦𝐹𝑎𝑒𝑟𝑜 𝑧0

𝑀𝑎𝑒𝑟𝑜 𝑦

𝑀𝑎𝑒𝑟𝑜 𝑧

+

000

𝑀𝑥 𝐿𝑆𝑆

00

+

0000

𝑀𝑔𝑦𝑟𝑜 𝑦

𝑀𝑔𝑦𝑟𝑜 𝑧

+

𝐹𝐼𝑛𝑒𝑟𝑡𝑖𝑎 𝑓𝑙𝑒𝑥 𝑥

𝐹𝐼𝑛𝑒𝑟𝑡𝑖𝑎 𝑓𝑙𝑒𝑥 𝑦

𝐹𝐼𝑛𝑒𝑟𝑡𝑖𝑎 𝑓𝑙𝑒𝑥 𝑧

𝑀𝐼𝑛𝑒𝑟𝑡𝑖𝑎 𝑓𝑙𝑒𝑥 𝑥

𝑀𝐼𝑛𝑒𝑟𝑡𝑖𝑎 𝑓𝑙𝑒𝑥 𝑦

𝑀𝐼𝑛𝑒𝑟𝑡𝑖𝑎 𝑓𝑙𝑒𝑥 𝑧

Aerodynamics Free rotor speed Inertia Rotor -rotating

Inertia Rotor - flexible

In the followinganalyses

Rotor Nacelle Assembly

RNA Physical model Numerical model

Gravity

Inertial | 6 DOF platform

Inertial | flexible tower

Inertial | flexible blades

Inertial | rotating blades (gyroscopic)

Aerodynamics

Free rotor speed

22



• Q: Required dynamics of the actuators ?

𝐿𝑜𝑤 𝑝𝑎𝑠𝑠 𝑓𝑖𝑙𝑡𝑒𝑟 ∶𝐹𝑟𝑜𝑡𝑜𝑟 𝑎𝑐𝑡𝑢𝑎𝑡𝑒𝑑, 𝑖 𝑠

𝐹𝑟𝑜𝑡𝑜𝑟 𝑠𝑒𝑡𝑝𝑜𝑖𝑛𝑡, 𝑖 𝑠=

1

1 +1

2𝜋𝑓𝑐𝑠

𝜖𝑅𝑌 < 5% 𝑓𝑐 > 0.3 𝐻𝑧 Full scale1.8 Hz scale 1:30

𝜖 ሷ𝑋𝑁𝑎𝑐< 5% 𝑓𝑐 > 0.5 𝐻𝑧 Full scale

2.7 Hz scale 1:30

23



• Max

24

TEST BENCH : Coupling of numerical model

• Wind Inflow• Aerodynamics• Servo• Structure

OpenFAST code

On boardcontroler


Methodology and time stepping ?

OC3 Hywind Spar – 5MW Waves conditions : Severe Sea State, 𝐻𝑠 = 8.6𝑚𝑇𝑝 = 13𝑠;

Wind conditions : Normal Turbulence Model, 𝑈𝑤 =18𝑚. 𝑠−1; 𝑇𝐼 = 14.6%;


Representative

picture

[14] [17] [15]



Motions RAO of the 3 floaters, for 2 incident wave directions. (a)Velocity RAO of the different floaters at origin point (b) Velocity and acceleration RAO at hub height for the different floaters

OC4 Semisubmersible OC3 Hywind Spar Triple Spar

NREL-5MW NREL-5MW DTU-10MW

Freewheel

26OMAE 2019-96374 Vincent Arnal - 12/06/2019


12/06/2019

Inverse dynamics ?

Sachant 𝑻 𝒔


𝑲 (𝑻𝟎)

𝟏+𝝉 𝑻𝟎 𝒔𝒆−𝑻𝒔

𝜷𝒂𝒅𝒂𝒑𝒕𝒆𝒅 𝒔 = 𝑻𝒔𝒆𝒕 𝒔 ∗ 𝑯 𝒔 −𝟏

Table 1 : Limits of the frequency bandwidths at full scale

Aerodynamic « engineering » models


BEM DBEMGDW

FVW

CFD

Physics

CPU Time

Hypotheses Data required Main limitations CPU Time

BladeElementMomentum

Blade elements + momentum (induction) theoryPotential FlowStationnary

𝑪𝒍 and 𝑪𝒅 of each airfoil section.Each empirical correction has a proper set of inputs.

No vorticity and quasi-static assumption dynamic inflow need to be modelledNo viscous effects dynamic stall need to be modelled2D blades effects only 𝑪𝒍 𝒂𝒏𝒅 𝑪𝒅 .

short

GeneralizedDynamicWake

Potential FlowDynamic inflow partly modelled

𝑪𝒍 and 𝑪𝒅 of each airfoil sectionEach empirical correction has a proper set of inputs.

No vorticity dynamic inflow approximatedNo viscous effects dynamic stall need to be modelledDon’t converge with high induction factor

short

DynamicBladeElementMomentum

Blade elements + momentum (induction) theoryPotential FlowDynamic inflow partly modelled

𝑪𝒍 and 𝑪𝒅 of each airfoil section.Each empirical correction has a proper set of inputs.

No vorticity dynamic inflow approximatedNo viscous effects dynamic stall need to be modelled2D blades effects only 𝑪𝒍 𝒂𝒏𝒅 𝑪𝒅 .

short

Free-Wake Vortex Model

Potential flowlifting line theory

𝐶𝑙 and 𝐶𝑑 of each airfoil sectionEach empirical correction has a proper set of inputs.

Dynamic inflow and proper calculation of modified induction due to blades –wake interactionDynamic stall and other viscous effects need to be modelled

Medium

Modelling assumptions


Summary of hypotheses

Module Hypotheses Limitations

Hydrodynamics

HydroDyn - Potential 1st and 2nd order- Morison elements

- Nonlinear excitation forces and wave kinematics (steep waves)- Complex viscous loads not well represented by Morison in severe sea states

Aerodynamics

AeroDyn v15

Blade Element Momentum Theory- ~ Dynamic Stall- ~engineering models for Hub and Tip loss, skewed wake induction factor calculation-~Tower influence

- Highly Unsteady Aero in severe sea states and turbulent wind- Glauert correction for induction below rated (induction factor ++)- In skewed wake: interaction with the near wake

Control

ServoDyn and associated .dll

- Collective Blade Pitch Control- Generator Torque Control- Fixed Nacelle Yaw

Relevant control strategies include :- Individual Blade Pitch control - Nacelle Yaw control (not really dynamic)- Control adapted to disturbance rejections with

additional inputsNo blade pitch actuator dynamic

Moorings

MoorDyn -Dynamic Lumped-mass model

- Approximated drag + added mass- no VIV- no wave-induced kinematics

Structural dynamics

ElastoDyn

- Tower flexible- Modal description fitted to testing specifications- Rigid blades

- No blade pitch actuators modelling

- Assumptions of small angles with correction for orthogonality for platform rotations and tower deflections.

Summary of hypotheses for FAST simulations run for the parametrical studies

Modelling assumptions


[1] Courbois A 2013 Étude expérimentale du comportement dynamique d ’ une éolienne offshore flottante soumise à l’action conjuguée de la houle et du vent.PhD Thesis, École Centrale de Nantes [2] Azcona J, Lemmer F, Matha D, Amann F and Botasso C L 2016 INNWIND Deliverable 4.2.4 : Results of wave tank tests

Documents

HYBRID MODEL TESTING OF FLOATING WIND …...2 Outline 1. Context : Wave tank hybrid testing of FWT 2. Specifications of actuators setpoint 3. Test bench presentation 4. System identification