Te

ch

nis

che

Un

ive

rsitä

t M

ün

ch

en

Win

d E

ne

rgy In

stitu

te

Numerical and Experimental Study of Wake Redirection Techniques in a Boundary Layer

Wind Tunnel

J Wang, S Foley, E M Nanos, F Campagnolo, C L Bottasso, A Zanotti, A Croce

Technische Universität München, Wind Energy Institute

Wake Conference 2017 May 31th 2017, Visby

Outline 1. Motivation

2. Numerical Model

3. Experimental Setup

1. PIV measurement 2. Hot-wire measurement 3. Steady Inflow map

4. Computational Setup

1. Mesh setup & convergence study 2. Boundary condition

5. Results and Analysis 1. Cyclic pitch control 2. Yaw misalignment control

6. Conclusion and Outlook

Motivation Wind farm wake redirection control

Yaw and/or cyclic pitch to deflect wake

Set-point control to optimize: • Wind farm power production • Wind turbine loading

Active load alleviation in wake-interference conditions

Motivation Development of numerical model

Wind tunnel (scaled) testing Scaled wind farm simulation

Fullscaling

A digital copy of the scaled wind farm facility were developed:

• Data Validated by experimental data

• Efficient and Robust

• Ability to simulate the most critical aerodynamic behavior of scaled wind farm

• Ability to simulation state-of-art wind farm control strategies

Numerical Model CFD + Aero-servo-elastic solver

• The simulation model is developed within

SOWFA (Fleming et al., EWEA 2013).

• Solver: standard incompressible solver in

the OpenFOAM repository.

• Actuator line method is embedded in a

large-eddy simulation (LES) environment,

coupled with the aero-servo-elastic

simulator FAST.

• Lagrangian scale-dependent dynamic SGS

model is used for LES modeling.

• Immersed boundary method is used to

model nacelle and tower effects on the flow.

• Low diffusive differencing scheme is

employed within the near wake.

Aero

-ela

sti

c s

olv

er

(FA

ST

)

1. A

ero

Dyn is u

sed t

o c

om

pute

both

the a

ero

dynam

ic a

nd d

ynam

ic r

esponse o

f th

e s

yste

m

2. Pit

ch-Torq

ue w

ind t

urb

ine c

ontr

ol is

enable

d

Simulation input:

1. Domain boundary conditions

2. Wind turbines Operating points

3. Wind farm control inputs

4. Outputs control

CFD solver (OpenFOAM)

1. Initialization

2. Pimple control

3. Blade points location & velocities are provided to FAST

4. FAST-computed body forces are properly smeared at blade

points locations

5. Iterate until convergence criteria is satisfied.

Simulation output

1. Flow quantities (i.e. u, p, TI, etc.)

2. Integral rotor quantities (i.e. Power, thrust, etc.)

3. Higher order quantities

Numerical Model High resolution convection differencing scheme - Gamma, Ref. Jasak et al.

Experimental setup Scaled wind farm facility

Scaled wind farm section Precursor section

3.84 m

19.2 m

16 m

Experimental setup PIV measurement

Experimental setup Hot-wire measurement

Hot-wire probe Hot-wire probe

Experimental setup Steady inflow map

• LiDARs are instruments equipped with a

steering laser beam that res rapid

pulses.

• LiDARs data were obtained from the

collaborative efforts of ForWind-

Oldenburg, TUM, Technical University of

Denmark and POLIMI

• Inflow map clearly shows the lack of

uniformity of the inflow, probably due

to the presence of a discrete number of

fans.

• Inflow was calibrated by LiDARs data

gathered 3D (3.3 m) upstream of the 1st

wind turbine.

LiDAR equipment

Inflow velocity profile (facing upstream) as measured by

two scanning LiDARs

Computational Setup Mesh setup & Convergence study

Inflow

Fig.1 Views of the computational domain. bottom: cross-

section, top: lateral view

• Zone 1: background mesh

(𝚫𝒙 = 𝚫𝒚 = 𝚫𝒛 = 0.073D ).

Zone 2: 1st level refinement

(𝚫𝒙 = 𝚫𝒚 = 𝚫𝒛 = 0.036D ).

Zone 3: 2nd level refinement.

(𝚫𝒙 = 𝚫𝒚 = 𝚫𝒛 = 0.009D ).

30 million cells

• The computational mesh successfully

passed mesh independency study.

• Blue shadow square indicates where

low diffusive differencing scheme is

imposed.

Computational Setup Boundary condition

• Steady not-uniform inflow conditions

(LiDARs data).

• Cyclic pitch control (CyPC) uses:

𝛉𝒊(𝒕) = 𝛉𝟎 + 𝛉𝒄 cos(𝝍𝒊(𝒕) + 𝜸)

where 𝛉𝒊(𝒕) is instant blade pitch, 𝛉𝟎 the

collective pitch, 𝝍𝒊(𝒕) the azimuth angle,

𝜸 the phase angle.

• CyPC pitch amplitude 𝛉𝒄 is 5.3 deg. Two

values used for 𝜸: 52 and 270 deg.

• Yaw misalignment (YM) angle: 20 deg.

G1 Model

Rated power @ 6 [m/s] 46 [W]

Rated rotor speed 850 [RPM]

Inflow speed @ hub height 5.97 [m/s]

Turbulence Intensity 2 %

Fine pitch angle 0.41 [deg]

Torque controller Look-up table

Results and analysis Integral rotor quantities

• Reasonable agreements between simulation and experiment are achieved in terms of both power and thrust.

Comparison of rotor power and thrust between simulation and experiment for the baseline case

CyPC and YM

Epsilon = 0.22 m Epsilon = 0.31 m

Results and analysis Wake analysis for CyPC

• From left to right: baseline

condition; CyPC with phase shift

of 52; CyPC with phase shift of

270 deg.

• Results show reasonable

agreement in terms of wake

shape and average velocities.

• Potential explanations:

Accuracy of multi-airfoil

table & airfoil interpolation

technique

Tip-root loss modelling

Uncertainty of measurement,

such as inflow map

calibration.

Normalized time-averaged streamwise on yz-plane at 0.56D and

6D downstream of the wind turbine

Results and analysis Wake analysis for CyPC

Normalized time-averaged streamwise along hub-height horizontal

line at 0.56D and 6D downstream of the wind turbine

• From left to right: baseline

condition; CyPC with phase shift

of 52; CyPC with phase shift of

270 deg.

• Results show reasonable

agreement in terms of wake

shape and average velocities.

• Potential explanations:

Accuracy of multi-airfoil

table & airfoil interpolation

technique

Tip-root loss modelling

Uncertainty of measurement,

such as inflow map

calibration

Results and analysis Wake analysis for YM

From left to right, hub-height horizontal line time-averaged

streamwise velocity, TI and Reynolds stress component at 4D

• Gaussian width is tuned for 20 deg YM in order to better match the power output and the wake profiles.

• TI and Reynolds shear stress are well predicted.

• Effects of nacelle and tower significantly influence the downstream wake behavior.

• Tip loss model and dynamic stall are not included.

YM = 0 deg

YM = 20 deg

Conclusions and Outlook

1. LES tool were developed to model scaled wind farm facility.

2. Power and thrust are well predicted by LES framework.

3. Velocity profiles correlate well with the experiments for baseline and CyPC simulations, either for near and far wake.

4. By tuning Gaussian width for ALM, good agreement in terms of velocity, TI and Reynolds stress were achieved for large yaw angle conditions.

5. The LES framework is validated: possible to extensively simulate wake redirection techniques.

Acknowledgment

This project was partly funded by the EU Horizon 2020 research and innovation program under the Marie Sk lodowska-Curie grant agreement No. 642108.

The authors wish to thank G. Campanardi and D. Grassi from the Politecnico di Milano for their contribution to the PIV measurements.

Thank you for your attention! Any questions?