Nwtc seminar overview of the impact of turbulence on turbine dynamics and their simulation, september 21, 2011

NREL is a national laboratory of the U.S. Department of Energy, Office of Energy Efficiency and Renewable Energy, operated by the Alliance for Sustainable Energy, LLC.

The Atmospheric Dynamics Associated with Turbine Dynamic Response and Their Simulation

NWTC Seminar

Neil D. Kelley

September 21, 2011

Innovation for Our Energy Future


Outline

2

• What we learned from the first seminar

• Role of atmospheric buoyancy and stability on turbine dynamic response

• What atmospheric process is responsible for the observed turbine response?

• Identification of a critical stability range, the atmospheric dynamics associated with it, and the resulting turbine response

• Extending our research to the Great Plains turbine operating environment

• The development of the TurbSim simulation code

• Conclusions and recommendations


What We Learned From the First Lecture

3

• The ingestion of coherent turbulent structures by the turbine rotors was responsible for the large, damaging fatigue loads observed but not modeled by the SNLWIND-3D simulation code

• The greatest turbine dynamic response occurs in a narrow, weakly stable range of the turbine layer Richardson number stability parameter: +0.01 ≤ RiTL < +0.05 can also contain significant levels of coherent turbulent kinetic energy Ecoh.

RiTL

-0.3 -0.2 -0.1 0.0 0.1 0.2

Hub

Pea

k E

coh

(m2 s-2

)

20

30

40

50

60

6 8 10 12 14 16 18 20 22 24 26

kNm

FBM DEL

Peak Response @ RiTL = +0.02


Significant Turbulence Activity Occurs at RiTL = +0.02 Maximum Turbine Dynamic Response

4

RiTL

-0.2 -0.1 0.0 0.1 0.2

w'T'

(o K-

ms-1

)

0

2

4

6

8

10

12

u'w' (m2s

-2)

-5

-4

-3

-2

-1

0

w'T'u'w'

RiTL

-0.2 -0.1 0.0 0.1 0.2

Posi

tive

Peak

w' E

coh (m

3 s-3)

-300

-200

-100

0

100

200

300

400

Negative Peak w' Ecoh (m

3s-3)

-600

-400

-200

0

200

400

RiTL

-0.2 -0.1 0.0 0.1 0.2

E coh (m

2 s-2)

2

3

4

5

cohEσ

Intense downward momentum flux u′w′ (shear stress)

Rapid decrease in mean buoyancy, w′ T′

Rapid suppression of peak vertical fluxes of Ecoh

Rapid suppression of vertical velocity fluctuations, σw


Conclusion . . .

5

Atmospheric Buoyancy Is a Major Player as Revealed by RiTL!


Remembering this demonstration from the first lecture . . .

Time

An example of dynamic instability

The right combination of vertical temperature stratification and wind speed shear (Ri ) can

produce an vertical oscillatory or resonant response in the wind field

Heig

ht

warm air

cold air


Define Buoyancy Length Scale, Lb

7

Lb = σw /Nbuoy where Nbuoy = buoyancy or Brunt-Väisälä frequency = and σw = vertical velocity standard deviation

Relationship between Ri and Nbuoy

2 2/ ( / )buoyRi N u z= ∂ ∂

Lb is a measure of the maximum buoyant displacement of air parcels and is the wavelength of the dominant instability mode.

1/2[( / )( / )]g T T z∂ ∂


Scaling Lb

8

If we scale Lb with respect to the turbine rotor dimensions by normalizing it by its diameter or

Lb /D then Lb /D = 1 equates the dominant wavelength of the flow instability is the same as the rotor diameter. We will see that helps us explain the physics behind the highest turbine dynamic response in the critical +0.01 ≤ RiTL < +0.05 range


Lb /D is an indicator of the degree turbulent eddies are being damped by negative buoyancy as seen on the ART Turbine

9

Peak

Fla

pwis

e St

ress

Cyc

le (k

Nm)

0

100

200

300

400

500

600

Turb

line

laye

r Ri

0.001

0.01

0.1

1

TL Ri vs TL Lb/D

Turbine layer lb/D0.1 1 10

Hub

Peak

CTK

E (m

2 /s2 )

1

10

100

Turbine layer Ri0.0001 0.001 0.01 0.1 1

Turb

ine

Laye

r lb

/D

0.1

1

10 moderate buoyancy damping

Buoyancy Damping Limits Coherent Structure

Size & Intensity and Reduces Induced Stress

Cycle Magnitude

Lb= buoyancy length scale, D = rotor diameter

/b w buoyL Nσ=

Turbine layer Lb /D

zero to very low buoyancy damping

critical RiTL

range

highest buoyancy damping


Influence of Increasing Stability and its Damping Effects on Turbulent Eddies . . .

10

As the air flow becomes more stable; i.e., the Ri increases due to the air cooling near the ground which • increases the temperature gradient (∆T/∆z) across the turbine

rotor disk

• the rising air contained in the turbulent eddies then sees more resistance due to negative buoyancy

• the largest turbulent eddies are affected first limiting their maximum size.


Definition of Stability Classes

11

Stability Class Designation Range

Moderate to Weakly Unstable Class, STC02 -1 < RiTL ≤ 0

Weakly Stable Critical Range, CRR +0.01 ≤ RiTL < +0.05

Weakly Stable High Range Critical, CRRH +0.05 ≤ RiTL < +0.10

Moderately Stable Range, STC04 +0.10 ≤ RiTL < +0.25

Very Stable Range, STC05 +0.25 ≤ RiTL < +1.0

Based on Turbine Layer Gradient Richardson Number, RiTL


Mean and Peak Root Flapwise Bending Fatigue Load Distributions by Stability Class Comparisons

12

Peak FBM Load

Stability class

STC02 CRR CRRH STC04

100

200

300

400

500

600

700

Root FBM DEL Distributions

50

100

150

200

250

300

NWTC ART Turbine Response

NREL Rotor Root FBM 3-Blade Mean DELs

kN

m 4

8

12

16

20

24

NREL Rotor Peak Root FBM

Stability Class


8

12

16

20

24

28

32

36

San Gorgonio Micon 65 Turbine Response



San Gorgonio Micon 65 NWTC ART

kNm

Stability class

Means

Peaks


Lb /D Relationships to Turbine Response DELs

13

Peak Response

Micon 65s NWTC ART

Peak Response

CRR CRRH STC02 STC04 CRR CRRH STC02 STC04


If we now look at the observed distributions of Lb /D by stability class

14

San Gor NREL Rotor

L b /D

0.0

0.5

1.0

1.5

2.0

2.5

not defined

ART Turbine

Stability class


L b /D

0.0

0.5

1.0

1.5

2.0

2.5

not defined

STC02 CRR CRRH STC04 Stability class

San Gorgonio Micon 65

NWTC ART L b /

D

1.0

1.0

The largest turbine dynamic loads that occur in the CRR stability class correspond to Lb /D = 1 or a buoyant wavelength of the same approximate dimension as the turbine rotor diameter!

not defined

not defined


The Buoyancy Damping Present Influences the Nature of the Transient Loads Seen on Wind Turbines

Turbine layer Ri0.0001 0.001 0.01 0.1 1

Turb

ine

Laye

r lb

/D

0.1

1

10moderate buoyancy damping

Ri =+0.034 Ri = +0.007

Upwind arrayinflow CTKE

m2 /s

2

0

20

40

60

80

100

120

0

20

40

60

80

100

120rotor top (58m)rotor hub (37m)rotor left (37m)rotor right (37m)rotor bottom (15m)

IMU velocity components

0 2 4 6 8 10 12

mm

/s

-20

-10

0

10

20

-20

-10

0

10

20

Time (s)

492 494 496 498 500 502 504

vertical (Z)side-to-side (Y)fore-aft (X)

zero-meanroot flapbendingmoment

kNm

-400

-300

-200

-100

0

100

200

300

400

-400

-300

-200

-100

0

100

200

300

400

Blade 1Blade 2

Ri = +0.015

critical buoyancy damping

high buoyancy damping

Micon 65/13 Micon 65/13

ART

Innovation for Our Energy Future 16

What Atmospheric Process is Responsible ?


Atmospheric Dynamics Associated with CRR Stability Range

17

Atmospheric Conditions . . .

Narrow, weakly stable Ri range of +0.01 to +0.05 High mean shear stress across rotor disk layer, u* Highest values of coherent turbulent kinetic

energy (Ecoh)

Buoyancy length scale, Lb ≅ rotor disk diameter Evidence of significant increase in buoyancy damping for Ri > +0.05 (CRRH range)

Conclusion: Some form of shear instability or instabilities present

• Kelvin-Helmholtz Instability (KHI)? • internal gravity wave instability (GWI)?

RiTL

-0.3 -0.2 -0.1 0.0 0.1 0.2

Hu

b l

oc

al

u*

(ms

-1)

1.6

1.8

2.0

2.2

2.4

2.6

2.8

8 10 12 14 16

FBM DEL(kNm)

NREL rotor

Micon 65/13 Dynamic Response

CRR CRRH


Comparisons of KHI and GWI Shear Instabilities

18

Kelvin-Helmholtz Instability (KHI) attributes that impact turbine dynamics . . .

Turbulent perturbations in the flow initiate KHI that extract energy out of the mean flow to create intense three-dimensional coherent turbulent filling the shear layer

Majority of turbulent Ecoh is contained within the shear layer and decays rapidly away from it

Formations of coherent turbulent structures or “patches” have a typical life span of 15-30 minutes which includes the initial rollup, rollover or breaking, and then decay.

The life cycle of an individual KH billow or turbulent patch is the order of tens of seconds

Internal Gravity Waves (GWI) have different attributes . . . GWI turbulent structures are much thinner that those associated with KHI

KE propagates away both upwards and downward from the shear layer in which it was formed

GWI turbulent structures grow much slower than KHI and can persist for many hours

May form at the conclusion of a KHI formation life cycle

CONCLUSION: KHI is the dominant atmospheric instability that affects the dynamic response of turbines


KHI as revealed in a cloud formation

19

Growing billows

Initial Shear Layer

Roll Over & Breakdown (Turbulence

Phase)

Source: Adapted from “Kelvin-Helmholtz Clouds” (DI00152) by Terry Robinson, copyright University Corporation for Atmospheric Research

flow direction 2-D vortex

braid


KHI Dynamics

20

• The initial stages of KH billows growing within a stably stratified atmospheric layer from a turbulent perturbation can be analyzed as an eigenvalue problem

• The billow can be represented as a spectral superposition of temporally growing oscillatory velocity normal modes of the form exp[ik(x-ct)] where k is the wavenumber (or equivalently the wavelength) ,

c is the complex phase speed c = cr + ici t is time. cr is the speed of the perturbation in the direction of the background flow kci is its rate of growth; i.e., growing modes are characterized by ci > 0.

• A necessary but not sufficient condition for KH billows or perturbations to grow

within the shear layer is Ri < +0.25


If we have the vertical profiles of wind speed and temperature which include a turbine rotor disk

21

U(z)

Hei

ght (

z)

T(z)

U = tanh(z)T = tanh(z)

→

→

↑

}Drotor

= shear layer depth Ri


Relationships between Fastest Growing K-H Modes, Buoyancy Length Scale, and Turbine Dynamic Response

22

K-H Mode Fastest Growth Rate

Adapted from: Rosenthal, A.J.; Lindzen, R.S. Journal of the Atmospheric Sciences (40:3) Ri

Turbine layer Ri

Disk-Diameter Normalized Buoyancy Length Scale, Lb /D Micon 65/13 Turbine Dynamic Responses (DEL)

CRR CRRH STCO4

max turbine dynamic response


Relationship Between KH Billow and Turbine Structural Vibratory Modes

23

Variation of Range of K-H Mode Growth Rate With Shear Layer Depth in Rotor Diameters

Adapted from Hogg and Ivey. Journal of Fluid Mechanics (477)

K-H Mode Growth Rates

Adapted from: Rosenthal, A.J.; Lindzen, R.S. Journal of the Atmospheric Sciences (40:3)

fastest growing K-H modes

• widest range of K-H mode wavelengths matches up with the most structural modes • narrower range of K-H mode wavelengths matches up with fewer structural modes

slowest growing K-H modes

• narrowest range of K-H mode wavelengths matches up with the fewest structural modes

fastest growing K-H modes occur at Ri = +0.02 and the shear layer depth = buoyancy length scale Lb /D =1; the rotor diameter!

Variation of Range of K-H Mode Wavenumbers (Wavelengths) with Growth Rate

Corresponds to maximum turbine dynamic response!

Mode wavenumbers (2π/λ) →


Summary of What We Have Found . . .

24

• Buoyancy length scale Lb coincides closely with the fastest growing (most unstable) K-H mode.

• If Lb /D = 1, then it is a good predictor of significant turbine dynamic response.

• If Lb = D then the equivalent wavelength of the fastest growing or unstable mode is approximately the same as the rotor diameter.

• This allows for the coupling of the coherent K.E. contained in the higher order K-H modes into the corresponding turbine structural modes.


Source: R. Banta, NOAA/ESRL

Horizontal distance (km)

Heig

ht (k

m)

NOAA HRDL LIDAR Observation of Wave Motions in Stable Boundary Layer

in Southern Kansas 22:34 LST

Generation of Coherent Turbulence by Atmospheric Wave Motions in Nocturnal Stable Boundary Layer

low-level jet organized

turbulent air motions from waves

height of wind speed maximum

high vertical shear region

SCHEMATIC OF COHERENT TURBULENCE GENERATION

waves

25


Role of Shear Layer Instability and Turbine Loads

Intense vertical shear within rotor disk layer provides the catalyst for developing atmospheric wave motions

Breaking KHI atmospheric wave motions produce bursts of coherent turbulence

Transient loads are induced when turbine rotors encounter coherent turbulent regions

26


LIDAR Observations of Coherent Turbulent Structures In Stable Nocturnal Boundary Layer

Buoyancy & vertical shear play a major role in shaping the impact of coherent turbulent structures within shear layers in the stable boundary layer and the subsequent impact on wind turbine components

Hei

ght

Time

wind turbines

Coherent turbulent structures observed in stable boundary layer by NOAA/ESRL HRDL Lidar in southeast Colorado during NREL/NOAA Lamar Low-Level Jet Project, September 2003.

wind speed profile

intense vertical shear

bursts of coherent turbulent energy

27


Details of Coherent Turbulence Episode at Great Plains Site Using LIDAR, SODAR, & Tower Measurements

28

wind turbines

LIDAR wind profile

SODAR wind profile

SODAR σw

profile

tower measured Ecoh

strong vertical motions


Turbulent Buoyancy and Coherent Events

29

coherent structures

negative buoyant damping

buoyancy (w′ T′ ) Coh TKE (Ecoh)


How Does a Wind Turbine Respond to a KH Billow?

30

• Peter Sullivan and Ned Patton of NCAR/MMM created an non-dimensionalized LES simulation of the life cycle of a stationary KH billow turbulence structure for a shear layer Ri of +0.05.

• Marshall Buhl of the NWTC interfaced the NCAR KH billow simulation into the NREL aerodynamics routine AeroDyn that is used to drive the NWTC turbine dynamics simulations (FAST and MSC.ADAMS).

• The NCAR KH billow simulation was used to excite the dynamic simulation of the 1.5 MW WindPACT virtual turbine using the MSC.ADAMS multi-body code.

• We examined the turbine dynamic response to the KH billow simulation.


The Dominant Atmospheric Dynamics Associated with the Critical Stability (CRR) Range

31

• Is weakly stable with the RiTL covering the range of +0.01 to +0.05

• Vertical shear exists

• The majority of the hub-height mean wind speeds are at or below rated, however the highest mean wind speeds also occur within this range

• In the San Gorgonio wind farm a persistent downward buoyancy flux exists that damps the largest turbulent eddy sizes

• The highest values of Ecoh occur within this range suggesting the frequent appearance of coherent turbulent structures

• The buoyancy length scale is the same as the diameter of the Micon and ART turbine rotor disks at least 50% of the time; i.e. Lb = Drotor

Atmospheric Flow Characteristics Associated with the CRR Stability Classification . . .


Simulated 1.5 MW WindPACT Turbine Response to Ingestion of NCAR LES KH Billow

32

m/s

15

16

17

18

UH

m/s

-2

-1

0

1

2

u'v'w'

m2 /s

2

-2

-1

0

1

2

u'w'u'v'v'w'

Time (s)

0 50 100 150 200 250

m2 /s

2

0.0

0.5

1.0

1.5

2.0

2.5

EEcoh

Simulated Hub-Height Inflow Wind & Turbulence

horizontal wind speed

u′, v′, & w′ components

Reynolds stresses

ET and Ecoh

billo

w b

reak

dow

n 1.5 MW WindPACT Turbine Inflow & Response

flapwise bending moment

continuous wavelet transform scalogram

discrete wavelet transform detail bands

reduction in root stresses from mixing out of vertical shear


WindPACT Turbine Response to Coherent Structures in Billow Breakdown

33

-30 -20 -10 0 10 20 30

Time Record 566

60

80

100

-30 -20 -10 0 10 20 30

Time Record 2500

60

80

100

-30 -20 -10 0 10 20 30

Time Record 4000

60

80

100

z (m

)

-30 -20 -10 0 10 20 30

Time Record 5000

60

80

100

-30 -20 -10 0 10 20 30

x (m)

Time Record 6000

60

80

100

t = 0 s

t = 91.5 s

t = 166.2 s

t = 217.7 s

t = 271.7 s

rotorplane

x (m)

z (m

) Billow breakdown beginning with

max 2-D structure

scalogram of continuous wavelet transform

detail bands of discrete wavelet transform

Root Flapwise Bending Load Spectral Stress Distribution

phase coherent responses


Conclusions from Simulation

34

1. The turbine aeroelastic response prior to the KH billow breaking down was dominated in the high blade cyclic root stresses as a result of the strong shear across the rotor

2. The greatest transient loading takes place during and after the billow breakdown when intense coherent turbulent structures are created in the flow

3. Wavelet analysis shows that these loading transients contain significant levels of coherent turbulent kinetic energy that induce phase coherent dynamic responses in the blade root bending loads

4. These results are consistent with our previous discussions and underscore the role of turbulence generated in a weakly stable boundary layer when the Ri = +0.05 which corresponds with the fastest growing KH instability modes.


Reasons for Extending Our Research to the Great Plains Operating Environment

35

• The greatest wind resource in U.S. resides in the western Great Plains

• Best winds often occur during the stable nighttime hours

• Nocturnal low-level jet streams occur frequently, particularly during warmer months

• Intense vertical wind shears occur beneath in the stratified flows beneath the jets

• We found that coherent turbulent motions often form within these stratified layers

• We have shown that coherent turbulence can induce significant loading on wind turbine rotors and structures


Typical Diurnal Variation in Wind Speed and Vertical Shear on the Great Plains

observed wind shear

design wind shear

Emick Ranch South of

Lamar, Colorado (Colorado Green Wind Farm)

36


LLJ Source: Bonner, W. D., December 1968, “Climatology of the low-level jet,” Monthly Weather Review, 96(12), 833-850.

Strong Correlation Between Wind Resource and Jet Bi-annual Frequency

The Nocturnal Low-Level Jet and Its Geographic Frequency

Eventual Potential Turbine Heights

p

4 6 8 10 12 14 16 18 20 22

Hei

ght (

m)

0

100

200

300

400

500

Typical Vertical Wind Profiles Associated With Low-Level Jets

Colorado Green Wind Farm, Colorado

Low-Level Jet

GE 1.5S Turbine Layer

10-min mean wind speed (m/s)

Colorado Green Wind Farm

37


Annual Diurnal Turbine Fault and Wind Shear Patterns Observed at Big Spring, Texas

0

100

200

300

400

500

600

700

12 AM 4 AM 8 AM 12 PM 4 PM 8 PM 12 AM

Time

Faul

t Tim

e (h

ours

)

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

Win

d Sh

ear E

xpon

ent

Fault TimeShear

Source: DNV Global Energy Concepts

Evidence of Possible Turbine Availability Relationship to Great Plains Low-Level Jet

peak low-level jet activity

38


Lamar Low-Level Jet Project

39

• Joint effort of DOE/NREL & Enron Wind (October 2001 through September 2003) • Collaboration with NOAA/ESRL • Project Objectives

• Process and analyze 1 year record of multi-level turbulence data from 120-m tower on high plains of southeast Colorado

• Assess potential impact of KH instability associated with LLJs on turbine response dynamics

• Collaborate with NOAA using HRDL LIDAR to measure LLJ characteristics

• Use data to incorporate low-level jet profiles, turbulence characteristics, and coherent turbulent structures into TurbSim turbulence simulation code


Observation Systems Used

SODAR (Scintec MFAS)

Mean wind profiles up to 500 m

LIDAR (NOAA HRDL)

Vertical profiles and turbulence spatial structure

Direct turbulence measurements from Enron Wind 120-m tower (sonic anemometers)

High-resolution turbulence & vertical thermal structure

REMOTE SENSING

40

Innovation for Our Energy Future 41

The TurbSim Stochastic Simulator


Development of TurbSim Simulator

42

• The SNLWIND-3D simulator was used as the basis for TurbSim

• Site-specific turbulence spectral models based on the direct measurements made at the NWTC and the Lamar (Colo. Green) Site were added

• Low-level jet wind and direction profiles were included in the Great Plains Low-Level Jet Model (GP_LLJ) from SODAR profiles and corresponding tower data

• Direct measurements of the properties of coherent turbulent structures were made from the data collected at the three San Gorgonio wind farm sites (upwind of Row 1, Row 37, and downwind of Row 41), the NWTC, and the Colorado Green (Lamar) Site.

• Fully 3-D coherent structures based on LES and DNS KH billow simulations by NCAR and Colorado Research Associates (CoRA) respectively have been incorporated into each of the site-specific TurbSim spectral models whose stochastic intensity and temporal distributions are scaled by the direct measurements.


TurbSim Spectral Models Scaling

43

• Each of the site-specific turbulence models utilize diabatic scaling based on the following the observed turbine layer Richardson number stability ranges and mean wind speeds in 2 m/s increments between 3 and 28 m/s. STC01 : RiTL ≤ -1 STC02: -1 < RiTL ≤ 0 STC03: 0 < RiTL ≤ +0.10 STC04: +0.10 < RiTL ≤ +0.25 STC05: +0.25 < RiTL ≤ +1

• Power spectral models incorporating up to two peaks were derived for the u, v, and w turbulent wind

components for each specific site

• Mean hub-height Reynolds stresses were scaled with Ri and other parameters depending the choice of which depends on the particular site.

• Spatial coherence models were derived and scaled for each specific site

• Parametric random distributions were employed when the scaled residuals were not Gaussian distributed

• Low-level jet wind speed and direction profiles are scaled with stability and other boundary layer parameters in the GP_LLJ spectral model

• Coherent structure attributes are scaled with stability and other boundary layer parameters based on the specific site.


Example of Power Spectral Models Variations with Stability

44

NWTC

GP_LLJ

Peak turbulent energies occur at lower frequencies (longer wavelengths) at the NWTC as compared with the Great Plains Site


Spatial Coherence U-Component Models

45


Example of Adding Coherent Structures to Background Flow

46


Example of simulated NWTC inflow wind field with and without added coherent structures

47

without coherent structures added

with coherent structures added

more black color visible; structures more intense


NREL TurbSim Stochastic Inflow Simulator Documentation

48


Conclusions and Recommendations

49

• We believe the TurbSim site-specific spectral models provide the turbine designer with a range of realistic emulations of full-field turbulent inflows that turbines will likely encounter; i.e., downwind of very complex terrain and beneath low-level jets in the Great Plains.

• We believe that designers should pay particular attention to testing turbine designs with stabilities in the critical Richardson number range (CRR) because of the challenging turbulence conditions associated with it.

• At some point in the design process, we strongly suggest that designers use the TurbSim site-specific inflow models with a multi-body dynamics code in order to fully assess the impact of turbulence generated with the CRR stability range.

• We also suggest that for a more complete picture of the role of turbulence in the dynamic response of wind turbines we encourage our colleagues to examine our full report on this subject that will be available within the next few weeks.

Technology

Nwtc seminar overview of the impact of turbulence on turbine dynamics and their simulation, september 21, 2011