TheRoleofTurbulenceonWindEnergy:FromSingleBlade

toWindArray

Source: Oak Creek Energy

LucianoCastillo**,VictorMaldonado*,Sheilla TorresNieves* &

CharlesMeneveau
**TexasTechUniversity,Lubbock,TX,USA
RensselaerPolytechnicInstitute,Troy,NY,USA
TheJohnsHopkinsUniversity,Baltimore,MD,USA
Funding:

NSF-GK-12: 0742436

OUTLINE
Motivation
AirfoilStudies(S809)&Turbulence

RoughnessEffectsonBlades
EffectsofFreeStreamTurbulenceatVariousAngleofAttacks
CombineEffects
VGsversusNoVGs

AScaledDownWindArray:3x5
Energyentrainment
Fluxes&Energybalance
ImportanceofTurbulence

Conclusions

MainObjectives
1. To quantify aerodynamic performance (lift and drag) of an S809
wind turbine blade with and without additional levels of free stream
turbulence (FST) including the combine effect of surface roughness
with FST
2. To acquire the mean flow over the blade in order to gain insight
into turbulent flow scale mixing on stall behavior and its effect on
aerodynamic performance.
3. To investigate the effect of aerodynamic enhancing vortex generators
on post-stall performance of wind turbine blades.

4. Measure profiles of horizontally averaged momentum

fluxes & Mean K.E. Compare turbulent shear stress with
canopy (dispersive stress) mean velocity shear stress.
5. Understand the role of the fluxes of kinetic energy in the vertical
direction.

Motivation:EffectsofFreeStreamTurbulence
OnWindTurbineBlade

Many investigations have studied the aerodynamics of a wind turbine airfoil

subject to high levels of free-stream turbulence.
However, no studies have analyzed the effect of this condition on the mean
flow when surface roughness is present.

Windturbinesoperateinturbulentflowconditionsincludingwakeinduced
turbulence inwindfarms.

Theeffectofthelengthandtimescales ofthisturbulenceonbladeloads
mustbeunderstoodtoimproveaerodynamicperformanceandprevent
prematureturbinebreakdown.

Aerodynamic studies have shown that the addition of free-stream turbulence

(passive grid) to the flow over wind turbine blades delays flow separation.

The aim of this study is to analyze the influence of high levels of FST on the
flow around a smooth and rough surface airfoil.
Examine how wind turbine airfoils are affected by highly turbulent flow
under stall conditions.
Analyze interaction of turbulent length scales with wind turbine blades,
particularly at high angles of attack.
4

Motivation:
Designedforthis
niceflow

...butexperiencea
highlyturbulentflow

Wind power has become one of the

most promising alternatives of
renewable energy.

Unanswered Questions:
-

How are wind turbine farms

interacting with the highly
turbulent atmospheric boundary
layer?

How is the highly intermittent,

turbulent free-stream influencing
the performance, power extraction
and control of wind turbine rotors?

How to improve Capacity Factor

and thus the Profit?

How to model and mitigate the

uncertainties in wind plant
performance?

Horns Rev 1 owned by Vattenfall.

Photographer: Christian Steiness
Source: Oak Creek Energy

Leading edge
contamination due to
insect debris causes
production losses of 25%
Corten and Veldkamp (2001)

windpowerengineering.com

Wind turbine blade

under icing
conditions

Research Motivation: Turbulence Affects Important

Aspects of Wind Power Extraction
Efficiency
Blade pitch is difficult to control in a highly turbulent environment. Therefore,
efficiency is not always at its highest.
Power loss due to wakes of upstream turbines of about 15%.
Cost
Wind turbines are designed to last 20 years; however, breakdowns appear as
early as 7 years (cracked blades, broken gearboxes and generators).
Costly repairs that increase pay-back time and cost of energy (COE).
Environment
Wake extends for miles.
Christiansen & Hasager: Wake effects of large wind farms identified from satellite
synthetic
aperture
radar
(SAR)
(2005)
Effect on humidity, pollination and local weather is unknown. Wakes are highly
turbulent.

ExperimentalSetup
The Corrsin Wind Tunnel Facility
Test section: 1.22 (m) width x 0.91 (m) height x 10 (m) length
Background turbulence intensity < 0.1%

ExperimentalSetup:ActiveGrid

Producesfreestream turbulence,Tu 6%
Eachshaftindependentlycontrolled
Randomrotationalspeedofwinglets
Located5.5mupstreamoftheblade

FlowParameters
ReynoldsNumber,Rec
ActiveGridFSTIntensity,Tu

6.14%

IntegralLengthScale,L

0.321m

MeasurementsbyKangetal.2003

AsproposedbyMydlarski andWarhaft (1996)

1.82x105

ExperimentalSetup:WindTurbineBlade
S809WindTurbineBladeModel

BasedontheNRELS809airfoil

Manufacturedusingaribandspartechnique

0.3

S809Airfoil

0.2

2Dblade,b =1.22m,c =0.25m

21%chordthickness

36staticpressureportsaty/b =0.5

-0.1

Tripwire(D=1.6 mm)4mmfromleadingedge

-0.2

0.1

z/c

-0.3
0

0.1

0.2

0.3

0.4

0.5

x/c

WindTurbineBladeMountedintheCorrsin WindTunnel

0.6

0.7

0.8

0.9

Results:AerodynamicPerformance
Thepressuredistributionsandvelocitydeficitofthewakewasmeasuredwith
a pitottubeinordertocomputetheliftcoefficientandtotaldragoftheblade.
TestConditions:Rec ~ 1.82x105 (U = 10 m/s) , Tu = 6.14%, L = 0.321 m
ImportanceofFreeStreamTurbulence(FST)
1.4

0.3

(a)

1.2

(b)

No FST
FST

No FST
FST

0.24
1
0.18

Cl

Cd

0.8
0.6

0.12

0.4
0.06
0.2
0

0.04

0.08

0.12

0.16

Cd

0.2

0.24

0.28

12

16

20

Freestreamturbulenceresultsinalowerdragcoefficientforagivenliftcoefficient,
particularlyatmoderatetohigh(poststall)anglesofattack.

Results:AerodynamicPerformance
TestConditions:Rec ~ 1.82x105 (U = 10 m/s) , Tu = 6.14%, L = 0.321 m
PressureDistributions,CP
-3.6

-3.6

(a)
CP, U
CP, L

-3

Tu 0%

-2.4

C P, U
C P, L

Turbulence

CP

CP

-1.8

-1.2

-1.2

-0.6

-0.6

0.6

0.6

0.2

1.2

0.4

x/c

(a)

0.6

0.8

1.2

0.2

0.4

28

x/c

0.6

0.8

(b)

No FST
FST

No FST
FST

24

Turbulenceresultsinasignificantly
greateraerodynamicefficiency,
i.e.lifttodragratio(L/D) formost
operatinganglesofattack.Thisis
duetothemixingandhigher
momentumofeddies inthe
turbulentboundarylayer.

20

Cl

L/D

0.8

0.6

0.4

16
12
8

0.2

Thepressuredistributionsindicatethat
theflowhascompletelyseparatedforthe
casewithoutturbulenceat =18,
howeverwithturbulence,flowseparation
isdelayed stillproducingsignificantlift
asshownintheplotofCl vs below.

Tu 6.14%

-2.4

No Turbulence

-1.8

1.2

(b)

-3

12

16

20

12

16

20

Results:WakeVelocityDeficit
(Resultsweremeasuredwithapitottube,alsomeasuredwithahotwireprobe similarresultsobtained)

=0

60

80

No FST: Cd = 0.0264

40

No FST: Cd = 0.041
FST: Cd = 0.0436

Probe Height (mm)

40

20

Above Blade TE
Below Blade TE

-20

20
0

Above Blade TE
Below Blade TE

-20
-40

-40

-60

-60
-1

-0.8

-0.6

-0.4

-0.2

-80
-1.6

-1.4

u - U (m/s) Velocity Deficit

80

100

FST: Cd = 0.0358

Probe Height (mm)

20
Above Blade TE
Below Blade TE

-40

-1

-0.8

-0.6

-0.2

-0.2

FST: Cd = 0.0421

-40

-1.2

-0.4

20

-20

-1.4

-0.6

40

-20

-60
-1.6

-0.8

No FST: Cd = 0.0694

60

40

-1

=12

80

No FST: Cd = 0.0456

60

-1.2

u - U (m/s) Velocity Deficit

=8

100

Probe Height (mm)

Probe Height (mm)

FST: Cd = 0.0298

=4

60

-0.4

u - U (m/s) Velocity Deficit

-0.2

-60
-1.6

-1.4

-1.2

-1

-0.8

-0.6

-0.4

u - U (m/s) Velocity Deficit

Results:WakeVelocityDeficit
=14

100
80

160

No FST: Cd = 0.1926

No FST: Cd = 0.0578

140

FST: Cd = 0.0897

FST: Cd = 0.0433

120
100

Probe Height (mm)

Probe Height (mm)

60

=16

180

40
20
0
-20

80
60
40
20
0
-20
-40

-40

-60
-60
-1.6

-1.4

-1.2

-1

-0.8

-0.6

-0.4

-0.2

-80
-2.8

-2.4

u - U (m/s) Velocity Deficit

-2

-1.6

0.3
160

0.24

-0.4

No FST: Cd = 0.2573
FST: Cd = 0.1144

140
120

Sharpspikeindragdue
toflowseparationat
=16 and18

100

Probe Height (mm)

Cd

-0.8

=18

180

No FST
FST

0.18

-1.2

u - U (m/s) Velocity Deficit

0.12

80
60
40
20
0
-20

0.06

-40
-60

12

16

20

-80
-4

-3.6

-3.2

-2.8

-2.4

-2

-1.6

-1.2

u - U (m/s) Velocity Deficit

-0.8

-0.4

VortexGeneratorsStudy
Vortexgeneratorswereutilizedtoinvestigatetheirroleonimprovingpoststall
aerodynamicperformance ontheS809windturbineblade.Thisstudyhasimplications
onthepowerproductionofwindturbines inthepoststallflowregime
Dimensions
h 2.5 mm
l 5 mm
z 12.5 mm

Vortexgeneratorgeometry(Velte etal.)

Vortexgeneratorsmountedontheblade

Experiments
1. Thebladepressuredistributionswereacquiredforanglesofattackof18,20,and22
withoutfreestreamturbulenceatavelocity,U of10m/s(Re=182,000).
2. Themeanvelocityfieldsoverthebladesurfacewasacquiredutilizing2DPIVto
capturetheflowphysicsrelatedtoflowseparationandbehaviorofthewakeforthe
sameconditions.

Results:PressureDistributions
AllCases:Rec ~ 1.82x105 (U = 10 m/s) without turbulence

-4.2

=20

-3.6

-3.6

=18

-2.4
-1.8

Cp

NO-VG C P,U
NO-VG C P,L
VG-SC CP,U
VG-SC CP,L

-4.2

-1.2

-3

-0.6
0

-2.4

Cp

NO-VG C P,U
NO-VG C P,L
VG-SC CP,U
VG-SC CP,L

-3

= 18o

-4.8

= 20

0.6

-1.8

1.2

0.2

0.4

-1.2

x/c

0.6

0.8

= 22

-4.2

-0.6
-3

0.6

-2.4
-1.8

0.2

0.4

0.6

0.8

x/c
WithVortexgenerators,thebladeleadingedgesuction
peakissignificantlyincreased,indicatingmorelift
generation.Thiseffectdecreases withhigherpoststall
anglesofattackof20 and22.

Cp

1.2

-1.2
-0.6
0
0.6
1.2

0.2

0.4

x/c

NO-VG C P,U
NO-VG C P,L
VG-SC CP,U
VG-SC CP,L

=22

-3.6

0.6

0.8

Results:WakeVelocityDeficitandDrag
Thewakewasmeasuredwithapitottube1.5cbehindthebladewithandwithoutvortex
generatorsat18degreesangleofattack.
Conditions:U =10m/s(Re=182,000)and =18 Withoutturbulence
200

Thelifttodragratioisincreasedfrom1.22to6.14usingVGs!
No VGs: Cd = 0.257
VGs: Cd = 0.170

150

Height (mm)

WakeProfiles
AerodynamicsResults

100

No VGs

VGs

Cl

0.315

1.045

Cd

0.257

0.17

L/D

1.224

6.145

VGs
50

1.5c
-7

-6

-5

-4

Pitottube
-3

-2

u - U (m/s) Velocity Deficit

-1

Results:2DPIVMeanVelocityFields
Themeanflowofthesuctionsurfaceofthebladewasacquiredwith2DPIVforthesame
conditionswithoutandwithvortexgeneratorsat18degreesangleofattack.
Conditions:U =10m/s(Re=182,000)and =18 WithoutFST
BoundarylayerseparationismitigatedwithVGs,resultinginalowervelocitydeficitand
wakethickness.Thistranslatestohigherliftandlowerdrag,increasingthelifttodragratio.

VGs

Results:2DPIVMeanvVelocity
Themeanflowofthesuctionsurfaceofthebladewasacquiredwith2DPIVforthesame
conditionswithoutandwithvortexgeneratorsat18degreesangleofattack.
Conditions:U =10m/s(Re=182,000)and =18 WithoutFST
Thereisaregionoflownegativevmeanvelocity nearthebladesurface
astheboundarylayerseparates thisregionisslightlyreduced withVGs.

Highpositive
vcomponent

Highnegativevcomponent

PIV Measurements: Smooth, Rough & FST

2-D Particle Image Velocimetry

Double pulse Nd:YAG laser (120
mJ/pulse)
time between pulses of 100 s
FOV: 16 cm x 16 cm
Measurements captured entire
airfoil (upstream leading edge to
0.6c; 0.48c to 60 mm past trailing
edge)
3,000 samples at 7.25 Hz

Experimental Parameters

L approximation using measurements by Kang et al. 2003

19

c = 0.25 m

Streamwise Mean Velocity Contours: = 0

Smooth and Rough Surface

Baseline

Surface Roughness

Free-stream Turbulence

Roughness + FST

Low-speed region appears when roughness is present at the wall.

Free-stream turbulence weakens the wake.
Combined effect of free-stream turbulence and surface roughness further increases the wake-Dominant effect of surface roughness.

Streamwise Mean Velocity Contours: = 16

Smooth and Rough Surface
x/c 0.60

Baseline
x/c 0.68

Free-stream Turbulence

x/c 0.44

Surface Roughness
x/c 0.40

Roughness + FST

Combination of surface roughness and free-stream turbulence is advancing

separation.
Effect of surface roughness is dominant.

Streamwise Reynolds Stress Contours: = 16

Smooth and Rough Surface

Baseline

Free-stream Turbulence

Surface Roughness

Roughness + FST

Dominant effect of surface roughness on separation.

Wall-normal Reynolds Stress Contours: = 16

Smooth and Rough Surface

Baseline

Surface Roughness

Free-stream Turbulence
Roughness + FST
Dominant effect of surface roughness on separation.

Reynolds Stresses: ZPG

B. Brzek, S. Torres-Nieves, J. Lebron, R.B. Cal, C. Meneveau, and L. Castillo, Effects of freestream turbulence on rough surface turbulent boundary layers, J. Fluid Mech., 635, 207-243,
24
2009.

Reynolds Shear Stress Contours: = 16

Smooth and Rough Surface

Baseline

Surface Roughness

Free-stream Turbulence

Roughness + FST

Concluding Remarks

Individual effect of free-stream turbulence is to delay

separation.

Surface roughness results in earlier separation.

Combination of free-stream turbulence and surface roughness is

advancing separation.

When the flow over the wind turbine blade is mostly stalled (i.e., =
16), the non-trivial interactions among the different length scales
result in complex flow dynamics.
Highly non-linear interactions were observed in Reynolds shear
stress.

Unresolved questions

How energy is entrained in an array?

What is the importance of turbulence in arrays and wake-wake

interaction?
26

WindArray:ScaleddownexperimentsandRoleof
Turbulence

Experimental study of the horizontally averaged

flow structure in a model wind-turbine array
boundary layer
Lebrn, J Meneveau, C., and Castillo, L., Experimental study of the horizontally averaged
flow structure in loaded turbines array, to be submitted at the Journal of Wind Energy,
(2012).

Objectives
Showstudiesinawindarrayof3x5 scaled850timesfromfull
scaleturbine.
Measureprofiles ofhorizontallyaveragedmomentumfluxes&
MeanK.E.
Compareturbulentshearstresswithcanopy(dispersivestress)
meanvelocityshearstress.
Understandtheroleofthefluxesofkineticenergyinthevertical
direction.

Toshowthatwemustuseasystemofsystemsapproachindealing
withwindfarmunderperformanceissue.

The WTABL and the Momentum Theory Eqn.

averagedthrust
force

Momentum theory (time averaged + dispersive stress):

u xz
u xz
v
x

u xz
1 dp d

u'v'
xz
x
dx dy

xz

u ''v ''

xz

fx

xz

u '' u u
Horizontal average
of turbulent shear Reynolds stress

xz

Correlations between mean

velocity deviations from their
spatial mean dispersive stress
(Raupach et al. Appl Mech Rev 44, 1991)

The WTABL and the Mean Kinetic Energy

Eqn.
Multiplying the momentum by the mean velocity leads to the mechanical
energy describing the kinetic energy.

In the inner region, the following terms are dominant:

Kineticenergyflux

Dispersivefluxdue
tospatialaverage

Turbulentdissipation
dispersivedissipations

Whatistheroleofturbulentmomentum&KEfluxinenergy?

Productofthespatially
averagedvelocityandthe
averagedthrustforce

Experimental set up: Overview

Wind turbine models

Rotors - water jet cut + 3D print mold
DC motor
Improved proportions
Higher thrust
3 by 5 Array
sx = 5D and sz = 3D
Tip-speed ratio = 4
Rough plate
Emulate a rough flat terrain
Roughness made of steel chains
Separated 1.5 D (18 cm)
k = 1.5 cm

Experimental set up: Inflow

Strakes
Generate shear and turbulence
Iterative design
Laser Cutter

Experimental set up: Rough plate

Rough plate
Emulate a rough flat terrain
Roughness made of steel chains
Separated 1.5 D (18 cm)
k = 1.5 cm

Experimental set up: Wind turbine models

Wind turbine models

Rotors - water jet cut + 3D print mold
DC motor
Improved proportions
Higher thrust
3 by 5 Array
sx = 5D and sz = 3D
Tip-speed ratio = 4

Experimental set up: PIV

Particle Image Velocimetry

t setup to 80 s to 100s (faster in wake)
FOV 23 cm x 23 cm
Mirror
3,000 samples at 7 Hz
Laser sheet thickness 1.2 mm

Mirror

Wind turbine array power

H.S. Kang, C. Meneveau, Meas Sci Technol 21, 105206, (2010)

Target turbine power (0.23W)

~ 40% wake loss

Fully Developed?

Averaged over the duration of the experiments and over each of the 5 rows
of turbines.
Wake loss consistent with field experiments by Van Leuven (1992) and
Barthelmie et al. (2007)

PIV measurements locations

PIV fixed in x and y but traversed in z

Set up slide in x
Measurements w/o turbines along centerline
Measurements with turbines elsewhere

Distribution of mean streamwise velocity

Internal boundary
layer growth in x

Slowdown due to turbine

Faster wake recovery of
downstream turbines

Contour stretched in the z-direction

Distribution of vertical velocity

Wake rotation

Flow going around

turbine

Distribution of streamwise Reynolds stress

uu increases row after row

Distribution of vertical Reynolds stress

vv varies with row

Distribution of Reynolds shear stress

Almost no shear stress.

Limited contribution to the flux
due to Reynolds shear stress.

Little change between rows

Horizontally-averaged Reynolds normal stresses

Higher <vv>xz than that of Cal et al. JRSE (2010)

Horizontally-averaged Reynolds shear and dispersive stresses

44% shear stress

Fluxes of KE due to Reynolds shear and dispersive stresses

Dissipation of kinetic energy

nacelle wake effects

Contribution of the pressure gradient

Budget of KE Fluxes

All terms considered are of importance, including those associated with the
dispersive stress.
Budget is not balanced. Array is too small to be fully developed
Significant contribution of advection terms is expected

Conclusions
Impact of the fluxes of mean kinetic energy due to dispersive stresses (i.e. transport
and dissipation) on the overall budget is significant. This is consistent with LES
simulations by Calaf, Meneveau and Meyer, Phys. Fluids (2010).
As in Cal et al. JRSE (2010), flux due to Reynolds stress is of the same order as
the wind turbine power and larger than the flux due to mean vertical velocity.
Salient conclusion: The present study reveals that the vertical entrainment of
mean kinetic energy (i.e., dominant mechanism of energy exchange between
large WT arrays and the ABL), is dominated by both mean and turbulent
quantities.
Residual of the budget of mean kinetic energy fluxes is not zero. Array is too
small to be fully developed
Future research will include the calculation of the flux of mean kinetic energy due
to mean streamwise velocity.