Atmospheric Stability Impacts on Wind

Turbine PerformanceJulie K. Lundquist, Ph.D.

Professor, Dept. of Atmospheric and Oceanic SciencesUniversity of Colorado at Boulder

Fellow, Renewable and Sustainable Energy InstituteWind Energy R&D Workshop at NCAR

May 11-12, 2010

Upcoming AMS events relevant to the renewable energy community Summer AMS Joint meeting, 2-6 Aug 2010, in

Keystone, CO: Boundary Layers and Turbulence Meeting Urban Symposium Agricultural and Forest Meteorology Meeting Several renewable-energy joint sessions

Short course on “Wind energy applications, supported by atmospheric boundary layer theory, observations and modeling” on Sunday 1 August, Keystone, CO

Atmospheric Stability Impacts on Wind

Turbine PerformanceJulie K. Lundquist, Ph.D.

Professor, Dept. of Atmospheric and Oceanic SciencesUniversity of Colorado at Boulder

Fellow, Renewable and Sustainable Energy InstituteWind Energy R&D Workshop at NCAR

May 11-12, 2010

Contributions gratefully acknowledged from

Lawrence Livermore National Laboratory: Sonia Wharton Research funding LDRD 06-ERD-026

Dept of Energy Wind and Hydropower Technology Office for research funding: Renewable Systems Interconnect Support “Tall Turbines”

Ibderdrola Renewables, Inc. for provision of this unique and extensive dataset Justin Sharp Mike Zulauf Jerry Crescenti

Widespread impression of wind farm underperformance

“20% by 2030” depends on sufficient capacity factor, not just installations

Impression that many US parks underperforming can undermine public perception, financing, etc.

With support from IRI, we investigate the role of atmospheric variability in one wind farm’s performance

We analyze data from an operating wind farm with modern tall turbines

Lundquist and Wharton, 2009, IEA Experts Meeting on SODAR and LIDAR;

Wharton, Lundquist, Sharp, Crescenti, and Zulauf, 2009, AGU Fall Meeting;

Wharton and Lundquist, 2010, DOE Technical report and in preparation for Wind Energy

Can atmospheric conditions explain wind turbine performance?

1. Background on atmospheric dynamics, wind turbine power curves

2. Overview of unique dataset

3. Approaches to quantifying atmospheric stability

4. Can stability/turbulence variations explain wind park “underperformance”?

Modern wind turbines span heights ~ 200m, penetrating a complex atmosphere

Siemens 3.0 MW turbine

Hub Heightsrangefrom 60-120m

49 m blade:Rotor diameter~ 100m

The diurnal cycle of atmospheric stability strongly influences winds in the turbine rotor disk

Wind Speed [ms-1]

Potential Temperature [K]

Hei

ght a

bove

su

rface

[m]

Hei

ght a

bove

su

rface

[m]

1800 LST 2200 LST0200 LST0800 LST

1800 LST 2200 LST0200 LST0800 LST

Poulos, Blumen,Fritts, Lundquist, et al.,2002

Radiosonde profiles demonstrate that the cooling of the surface overnight is accompanied by dramatic accelerations in the winds

This wind farm provides a unique and valuable dataset

Characteristics:

Presence of both marine and terrestrial BL over hilly terrain

Little directional wind shear

Strongly channeled flow

Large dataset:

On-site met towers + SODAR

Turbine power and nacelle wind speeds available

Four seasons of data; strong seasonality and diurnal signal

NIGHT

DAY

The data surpass those typically available at wind farms

Meteorological data:

2 met towers w/ cup anemometers (u, v) at 5 heights (30, 40, 50, 60, 80 m), 10 min. avgs; (T, p measurements unusable)

SODAR observations (u, v, w) for 19 heights (20 m to 200 m, 10 m resolution), 10 min. avgs.

Nearby research station with a sonic anemometer (u, v, w, ), 30 min. avgs.

Turbine data:

Leading edge turbines: nacelle U and power, 10 min. avgs, 80m hubs

Vertical profile of

cup anemometers

Doppler Sound Detection and Ranging (SODAR)

'' vw θ

sonic anemometer

Wind speeds vary with seasons; summer winds exhibit strong wind shear

Wind speeds exhibit a strong daily cycle in spring and

summer

Seasonal variability in winds is reflected in turbine capacity factor: most power generated on summer/spring nights

In absence of temperature measurements, how can stability be assessed?

vw θ ′′

Obukhov length, L (off-site)

θv : virtual potential temperaturek : von Karman constantg : gravity

: sensible heat flux u* : friction velocity =

''

3*

v

v

wgk

uL

θ

θ

⋅⋅

⋅−=

(3) Turbulence kinetic energy, TKE

u' 2: variance of wind speed

)'''(5.0 222 wvuTKE ++=

4/122)''''( wvwu +

(2) Turbulence intensity, IU

σU : standard dev. of mean horz. wind speed (U) at 80 m

)(zUI U

Uσ

=

(1) Wind shear exponent, α

U : mean horz. wind speed at height z or zR

α)()(R

R zzUzU =

Estimates of stability from a typical cup anemometer fail to agree with more sophisticated measures

55%

Summer data

80m

80m40m to 120m

surface

Averaged rotor disk profiles reflect atmospheric stability

Stable Neutral Convectivez/L > 0.1 - 0.1 < z/L < 0.1 z/L < - 0.1

α > 0.2 0.1 < α < 0.2 α < 0.1

IU < 10% 10% < IU < 20% IU > 20%

TKE <0.6

0.6 < TKE < 1.0 TKE > 1.0

Summer

Stable conditions: high wind shear, low turbulence, and possible nocturnal low-level jets

Neutral conditions: minimal wind shear

Convective conditions have lowest wind speeds, very little wind shear in swept-area, and are highly turbulent.

stable

neutral

convective

A typical summer power curve based on equivalent wind speed still exhibits

significant variabilityCapacity factor, CF (%)

Pactual : actual power yield of the individual turbine

Prated : maximum power yield of the turbine as determined by the manufacturer

100×=rated

actual

PPCF

At 8 m s-1 the CF ranges from 35% to 70%!

Stratification of power curves reveal stability-related influences on power output

Lawrence Livermore National Laboratory

STABLENEUTRALCONVECTIVEMANUFACTURER

Even stronger variation seen in another leading-edge turbine

Lawrence Livermore National Laboratory

In fact, all leading edge turbines show that power generated is dependent on stability

Lawrence Livermore National Laboratory

In summary: Atmospheric stability, through the

mechanisms of turbulence and wind shear, governs the generation of power at these tall turbines.

Power varied by over 20% due to atmospheric stability.

“Underperformance” may be due to inaccurate assessments ofavailable power due to failure to account for variability of wind and turbulence across rotor disk due to atmospheric stability variations.

Power is a function of atmospheric stability

63% capacity (stable) vs. 41% (convective)

Wind farm “underperformance” can in part be explained due to incomplete

resource assessment Resource assessment instrumentation should be upgraded: SODAR stability parameters segregate wind farm data into stable, neutra

and convective periods in agreement with research-grade observations Cup anemometer data inaccurately estimate stability regimes SODAR performs poorly during precipitation, however – role for LIDAR?

Because of complex wind profile shapes, power curves should be a function of wind speed and turbulence over entire rotor disk (UequivTI) (as in Wagner et al., 2009)

Power output correlates well with atmospheric stability: Enhanced turbine performance during stable conditions Reduced turbine performance during convective conditions

Questions?

Lawrence Livermore National Laboratory

Julie K. LundquistUniversity of Colorado at Boulder

Julie.Lundquist@colorado.edu

Voice: 303/492-8932

http://atoc.colorado.edu/~jlundqui

Which velocity should be used to determine power curve?

Available measurements: Nacelle anemometers (cup

anemometers) in immediate wake Hub-height met mast up to 7 km

away from each turbine SODAR measurements/profiles up

to 5 km away from each turbine

Industry-standard power curves require met masts

Met masts cannot capture shear across turbine rotor disk

SODAR, which can capture shear, too far from turbines

Because hub-height often fails to indicate the true rotor wind speed, we calculate an “equivalent” wind

speed by integrating across rotor disk,

stable

Equivalent wind speed, UequivTI

A : rotor area, Ueff(z) : mean wind speed at height z, r : radius ofrotor area, H : hub-height

Ueff(z) calculated for each height within the rotor disk:

U(80m)

U(70m)

U(60m)

U(50m)

U(40m)

U(120m)

U(110m)

U(100m)

U(90m)

Rotor swept area

Ueff (z) = U(z)3(1+ 3IU2)3

accounting for the additional energy (turbulence) in the instantaneous wind speed(following Wagner et al. 2009)

dzzHzHrzUA

UrH

rHeffequivTI

2/1222 )2)((2∫+

−

−+−=

Hub-height wind speed often fails to represent momentum experienced by the

entire rotor disk

stable

neutral

convective

Hub-height winds are often maximum winds across the rotordisk!

Stable UequivTI > U80m

Neutral UequivTI = U80m

Convective UequivTI < U80m

Nacelle-based “equivalent” wind speed produces the most accurate power curve

Ongoing work: forecasting “ramping events” to predict sudden changes in power

100

0 Time (fraction of day) on ramp 1

100% power increase in 30 minutes

Forecasts capture the ramping event

Ramping events complicate integration of wind-generated power into power grids

Large ramping events at this farm have been identified in the data and are currently being modeled by CU-Boulder/NREL, LLNL, CSM, and UC-Berkeley: WRF: 1 km resolution mesoscale model

with increased vertical resolution in the lowest 200 m (CU-Boulder/NREL, LLNL)

PF.WRF: coupled subsurface-surface-atmosphere model (CSM)

WRF-LES: a higher-resolution, turbulence-resolving model (UC Berkeley, CU-Boulder/NREL)

U (m

s-1

) at 1

00 m

WRF-LES forecasted wind speed in comparison toactual wind speed. The two runs show that higher vertical resolution improves the forecast.

From Marjanovic, Chow, Lundquist, 2010

TWICS: Turbine Wake and Inflow Characterization Study Although large wind turbines are designed to

IEC standards, turbines regularly experience extreme wind inflow events outside of the limits defined by those standards: Need wind, turbulence, and stability

measurements across entire rotor disk

Downwind turbines experience wakes with decreased winds and increased turbulence Need detailed wake measurements along

with inflow meteorology to understand atmospheric effects on wakes and on downwind turbines

Background: Wuβow, Sitzki, & Hahn, 2007,CFD simulation using ANSYS FLUENT 6.3 LES

Characterizing turbine inflow and turbine wakes with Doppler LIDAR at a modern 2.3 MW turbine: couple

models (left) with observations (right)

Wuβow, Sitzki, & Hahn, 2007,CFD simulation using ANSYS FLUENT 6.3 LES

Project Plan: Deploy NOAA’s High Resolution Doppler Lidar at NREL’sNational Wind Technology Center (Fall/Winter 2010) to characterize inflow and wake from the Siemens 2.3 MW turbine; model with WRF and WRF-LES

Kelley et al., 2006: streamwisevelocity and velocity variance from HRDL

TWICS field plan (as of 4/2010)

135m met tower, 6+ sonics, T profiles

NOAA HRDL LIDAR

2 NREL SODAR

CU WindCube LIDAR

Two-week deployment during windy season

Plan subject to change

Mean wind direction (292)

Lundquist (CUB, NREL), Kelley (NREL), Banta/Pichugina (NOAA), Mirocha (LLNL)

Typical output data from the HRDL reveals rapid intermittent turbulence bursting events

HRDL observations enable high-time resolution characterization of atmospheric turbulencesuch as would be needed for anticipating turbine response (Banta et al., 2006; Pichugina et al., 2008)

Observations will be compared with nested WRF/WRF-LES simulations

D1: ∆x = 2430m

D2: ∆x = 810m

D3: ∆x = 270m

D4: ∆x = 90m

D5: ∆x = 30m

D6: ∆x = 10m

LES domains (D3+) use nonlinear backscatter with anisotropy model to capture stability effects

Challenges include complex topography and appropriate spin-up for LES turbulence

At a different site, we see LES winds differ from mesoscale winds in both timing and intensity due to fundamentally

different physics

Lawrence Livermore National Laboratory

D03, mesoscale, Δx = 1.33kmMYJ36 hrs

20 05 14 23 08Local Time

800

600

400

200

0Hei

ght a

bove

sur

face

(m)

800

600

400

200

0Hei

ght a

bove

sur

face

(m)

20 05 14 23 08Local Time

D06, LES, Δx = 49mTKE SFS closure36 hrs

Win

d S

peed

[m/s

]

10.0

5.0

0.0

LES shows increasedvariability in early LLJ development with quiescentperiod early and delayed onset of stronger winds

Meteorological insight can drive solutions to technical challenges inhibiting expansion of

domestic wind energy

New instrumentation can provide wind and turbulence profiles throughout turbine rotor disk Resource assessment instrumentation can be upgraded: Explain wind farm over/underperformance Increase accuracy of resource assessment

Coupling of innovative observations with high-resolution atmospheric modeling provides insight to turbine inflow conditions and wake behavior Important for downwind turbines Important for downwind microclimates, especially crops – upcoming

experiment with Iowa State University in June/July 2010