x _ t¸/t

NREIjTP-442-5619 • UC Category: 261 • DE93010032

Defining the Turbine InflowWithin a Wind Environment

i!

L!

i RECEIVEDi

JUL 1._t_t

' OSTtNeil D. KelleyPrepared for theIEA Topical Meeting 24--WindConditions for Wind Turbine DesignApril 29-30, 1993Roskilde, Denmark

t

National RenewableEnergy Laboratory1617 Cole BoulevardGolden, Colorado 80401-3393Operated by Midwest Research Institutefor the U.S. Department of Energyunder Contract No. DE-AC02-83CH10093

...... .oo,oo.oMA tLHrr=par=d trader Task No. """"" o_,, _, .

June 1993

IDISTRtBUT1ON OF- Trl"_l,:.%_':.)('.'._.';_.JJ_.._,_E_,,('_"._C',I,..:,i.INLIMITED

NOTICE

NOTICE:Thisreportwaspreparedas an accountof work sponsoredbyan agencyof theUnitedStatesgovernment.Neitherthe UnitedStatesgovernmentnorany agencythereof,noranyof theiremployees,makesany warranty,expressor implied,orassumesany legalliabilityorresponsibilityforthe accuracy,completeness,orusefulnessofany information,apparatus,product,orprocessdisclosed,orrepresentsthatitsuse wouldnotinfringeprivatelyownedrights.Reference hereinto any specificcommercial product,process,or serviceby wade name, _ademark, manufacturer,or otherwisedoes not necessarilyconstituteor implyitsendorsement,recommendation,or favoringbytheUnitedStates governmentorany agencythereof.The viewsandopinionsofauthorsexpressedhereindo notnecessarilystate or reflectthose of theUnitedStates governmentorany agency thereof.

Printedinthe UnitedStatesof AmericaAvailablefrom:

NationalTechnicalInformationServiceU.S. Departmentof Commerce

5285 PortRoyal RoadSpringfield,VA22161

Price:MicroficheA01PdntedCopyAO3

Codes are used forpricingali publications.The code is determinedby the numberof pagesinth_ publication.Informationpertainingto the pricingcodescan be foundinthecurrentissueofthe followingp.blicationswhichare generallyavailablein mostlibraries:Energy ResearchAbstracts (ERA); GovernmentReports Announcements and Index (GRA and I): Scientific and TechnicalAbstract Reports (STAR);and publicationNTIS-PR-360availablefromNTIS attheabove address.

i! "_el_ Printedon recycledpaper

i i ,,i , jjjjl , ,,

IEA Annex XI

Expert Mecdng 24, Wind Conditions for Turbine DesignRiso, April 29-30, 1993

DEFINING THE NORMAL TURBINE INFLOWWITHIN A WIND PARK ENVIRONMENT

N. D. KelleyWind Technology DivisionNational Renewable Energy LaboratoryGolden, Colorado U.S.A.

Introduction

This brief paper discusses factors that must be considered when defining the "normal" (as opposed to"extreme") loading conditions seen in wind turbines operating within a wind park environment. Wedefine the "normal" conditions to include fatigue damage accumulation as a result of

• start/stop cycles• emergency shutdowns• the turbulence environment associated with site and turbine location.

We also interpret "extreme" loading conditions to include those events that can challenge the ,.

survivability of the turbine.

Loading Characteristics Responsible for Maximum Fatigue Damage Accumulation

Recent analyses of the loading events associated with turbine rotor blades constructed of compositematerials have shown that the bulk of the fatigue damage is associated with infrequent (or low-cycle),high-amplitude (peak-to-peak load) events. This is particularly true of blade materials that arecharacterized by a high S-N exponent.

We recently analyzed two sets of extensive root flapwise bending moment measurements taken from twostall-controlled, rigid-hub Micon 65 turbines and an NPS-100 teetered-hub turbine. The blades of theserotors were of similar length and weight. We used rainflow counting to determine the alternating (p-p)stress cycle distributions seen by these three machines over record lengths of 67.5 and 70.1 hours,respectively, in wind park environments. Figure 1 compares the results for the rigid and teetered rotors.The maximum cyclic stress reduction for the teetered rotor occurs near 15 kNm but asymptotically

approaches the rigid ones at both the high- and low-cycle extremes. Above a p-p value of 15 kNm, bothhub designs exhibit a decaying exponential distribution, as is indicated in the figure.

The region of the curves of Figure 1 above p-p values of 15 kNm is the greatest concern from theviewpoint of fatigue damage accumulation, lt is therefore very important, when defining the "normal"loading environment for a wino park, to identify such low-cycle events and include them in any dynamicsimulations.

Impact of Coherent Turbulent Structures

The consequences of a turbine rotor blade encountering a coherent or organized patch of turbulence aredemonstrated in Figures 2 and 3. Figure 2 plots the root edgewise bending moments from Blade No. 1 oftwo, side-by-side, Micon 65 turbines. One of the turbines had blades based on the NREL (SERI) thinairfoil family and the other with refurbished, original-equipment AeroStar blades. Also presented are theroot flapwise bending moments of the three blades on the AeroStar-equipped turbine. The plots indicatethat the AeroStar rotor reacted differently to an excitation that lasted approximately 3-4 seconds.

The corresponding hub-height turbulent inflow characteristics are presented in Figure 3. Figure 3ashows the strong correlation of the instantaneous u'w' and v'w' shear or Reynolds stresses within the

period of enhanced blade response. The estimated hub-elevation vorticity components (o)i) and helicity(uioai) time series are plotted in Figure 3b. These parameters indicate the existence of a vortical structurethat could be responsible for the shear stress pattern in Figure 3a. The evidence implies a one-to-onecorrespondence between the enhanced cyclic activity on Blade No.1 of the AeroStar rotor and thepresence of a vortical stTucture in the inflow.

We examined the inflow conditions associated with the largest observed root flapwise tension peak andone of the larger compression (negative) ones. Figure 4a plots a five-second record of the flapwise loadsfrom each of the three blades on the NREL rotor. The largest excursion occurs on Blade No. 3. The

remaining two blades also are affected but to lesser degrees perhaps indicating that a coherent structureis convecting through the rotor disk. The enhanced loading extends to a period of about 2 seconds. Thecorresponding hub-height estimated vorticity/helicity record also lasts about the same amount of time.When the later is aligned with the Blade No. 3 peak of Figure 4a, the plot of Figure 4b results. Similarly,Figure 5a documents the peak compression load experienced by Blade No 2. Again, the corresponding

vorticity/helicity record is superimposed in Figure 5b, but no time alignment has been performed. The ..plots of Figures 4 and 5 do seem to support the hypothesis that organized, vortical structures are indeedconnected with large induced blade stresses.

We looked more closely at the population of inflow conditions connected with the 25 largest flapwisestress cycles seen on either of the two Micon 65 rotors. In applying damage theory, it is assumed that therotor materials nave an infinite memory. This requires that ali cyclles must be closed to assess the

corresponding damage. When the entire 67.5-hour record was rainflow counted as a single time series, asignificant number of large-amplitude cycles were found that spanned the original ten-minute records.Figure 6 plots the largest maximum and minimum pairing found in the data set from the NREL-equippedrotor. The period between the minimum (compression) and positive (tension) peaks was 43 hours oralmost three days. From a meteorological point of view, the conditions associated with each of the peakscan be considered independent events. Thus, we can look at the melteorological conditions associated

with the population of the peak tension and compressional stress events independently also. We did thatand found the following:

• Ali of the tension peaks associated with the largest stress cycles occurred during slightly stable flowsemerging from a deep canyon southwest of the wind park.

• The common time of occurrence was 22 h local standard time.

• The nogative (compression) peaks of these cycles occurred during slightly more stable conditionscentered near 04 h.

2

The identification of these high-stress loading events, and the inflow conditions associated with them,has raised the following questions that will need to be answered in the near future:

• How often can these large peaks be expected to occur?

• What are the appropriate statistical models to describe the distributions of such events?

• What are the turbulence conditions associated with them?

• Can such inflow conditions be simulated for use by dynamic simulation codes such as ADAMS ®(Automated Dynamic Analysis of Mechanical Systems) ?

The Need for Stochastic 3-D Turbulence Simulation

Events such as those described above are impossible to simulate using only the longitudinal (u)component of the wind even if it is stochastic. A full-vector or three-dimensional simulation is requiredthat includes not only spatial coherence but local cross-axis correlation as well. The latter is needed tosimulate the dynamic shearing stress fields observed in actual flows, lt is not clear which fluid dynamicsparameter or parameters are the best indicators of the coherent turbulence that is responsible for theincreased dynamic loading of wind turbines. We hope to answer this question in the near future withfurther analysis and field measurements.

We have used the computational kernel of the SNLWIND Code developed by Paul Veers of Sandia ..National Laboratories [1] as a basis for achieving a full-vector simulation. The objective of this workhas been to develop the ability to simulate a statistically relevant ten-minute record of the three-dimensional wind field found in and near a large wind farm. This simulation also includes, at least

statistically, the temporal and spatial variations of coherent structures embedded in the more randominflows found in these locations.

The modeling of the turbulent inflows upwind, downwind, and within a large wind farm was based on 'extensive measurements taken at a large wind farm in San Gorgonio Pass in southern California [2,3,4].The methodology used to expand the SNLWIND Code to a full-vector simulation included:

• Identifying suitable homogeneous terrain spectral models for each of the three wind components

(u,v,w) for use as references

• Developing empirical "target" velocity spectral scaling based on the measurements taken in SanGorgonio

• Developing empirical relationships for the vertical coherence of the longitudinal (u) and crosswind(v) wind components

• Deriving empirical relationships between the normalized cross-axis correlations (rij) and boundarylayer scaling parameters

We used the homogeneous (smooth) terrain models of Olesen, Larsen, and I-Iajstrup [5] as target spectrareferences. Local spectral scaling was accomplished by applying empirically derived ratios to theappropriate homogeneous terrain model, lt was necessary to include up to three spectral peaks todescribe the observed 10-minute spectral variation distributions. For example, to describe the crosswind

(v) spectra downwind of the wind farm under unstable flows three peaks were required: Sv(n ) = SL(n) +SH(n) + Swake(n). Here, the total spectrum is composed of low-frequency, high-frequency, and turbinewake contributions. The predicted crosswind (v) spectra for a mean hub-height wind speed of 12 ms-1are plotted in Figure 7 for representative unstable, near-neutral, and stable conditions.

The spatial coherence was introduced using an exponential decay model using empirically derivedcoherence decrements based on the measurements upwind and downwind of the San Gorgonio windfarm. lt was found that these decrements were monotonic functions of the hub-height mean wind speed.The observed decrements and corresponding linear regressions are plotted in Figure 8 for the horizontal

wind components. Empirical normalized cross-axis covariances (rij) scaled with boundary layerparameters are used to crossfeed the wind components to simulate the observed shear stresses. For

example, a reasonable facsimile of the shear stress means and distributions may be accomplished bycrossfeeding the u component with only the v and w velocities per the relationship

u'(t) = u'(t) + ruv v'(t) + 2ruw w'(t).

When simulation locations are downwind and within the wind farm, ali three wind components must be

crossfed to achieve the reasonable replicas of the observed shear stress distributions.

Conclusions

There is strong evidence that coherent turbulent structures ingested by turbine rotors are responsible forthe largest peak stresses seen in the fiapwise and edgewise bending moments. These interactionsproduce a coherent (phase specific) response in the rotor and other structural components because ofmultiple structural modes being excited simultaneously. The most damaging tension stresses occurduring boundary layer conditions most likely to support atmospheric wave motions and enhanced turbinewakes. The expanded version of the Veers SNLWIND Code (SNLWIND-3D) provides a much morerealistic simulation of the turbulent inflow seen by turbines installed in various locations within a wind

farm in complex terrain.

Acknowledgement

This work has been supported by the U.S. Department of Energy under contract DE-AC02-83CH10093.

References

1. Veers, P., "Three-Dimensional Wind Simulation," SAND88-0152, Sandia NationalLaboratories, March 1988.

2. Kelley, N.D., "An Initial Look at the Dynamics of the Microscale Flow Field within a LargeWind Farm in Respo_se to Variations in the Natural Inflow," SERI TP-257-3591, Solar EnergyResearch Institute, October 1989.

, ,111

!

3. Kelley, N.D. and A.D. Wright, "A Comparison of Predicted and Observed Turbulent WindFields Present in Natural and Internal Wind Park Environment," NREL TP-257-4508, National

Renewable Energy Laboratory, October 1991.

4. Kelley, N.D., "Full Vector (3-D) Inflow Simulation in Natural and Wind Farm Environments

Using an Expanded Version of the SNLWIND (Veers) Turbulence Code," NREL TP-442-5225,National Renewable Energy Laboratory, November 1992.

5. Olesen, H.R., S.E. Larsen, and J. I-{ejstrup, "Modeling Velocity Spectra in the Lower Part of thePlanetary Boundary Layer," Boundary-Layer Meteorology, Vol. 29, 1984.

10000 " . ..........+ i I i _ I r

', : II NREL rotor, rigid hub /

_ ___ ....

, i.... ?-- , & AeroStar rotor, rigid hub1000 ....... ,' 8- "_ _ " " • NPS100 rotor, teetered hub= lO

1O0 .... ',.... _,.... _, , li==- .... -_ 0 L t

alternating : : Nrigid=13xl 04exp (-Mflap/3.5)10 ' : ' ....cycles .........i I _ I I i i

I , i i i I I

per 1 .............. I .... q .... T ..... ',..... f_ - ..... j .... -1 ...... T .... r ....

hour ........., : ,I

Nteetered=725exp(-Mflap/46) ' ....N Ol ......• '_ .... I-" - - - "-1 .... m ..... r- ....I i I i

it I I i I I -- •I I t I I i

I l I I I I i i A I

001. .................................................. "i i I i I I I i i i

t Il _ I t t P t

i I I I _ I I I

: I I I Il I I i

0.001

0 10 20 30 40 50 60

p-p flapwise root bending moment (kNm)

Figure 1. Comparison of rigid- versus teetered-hub low-cycle alternating root flapwis_ bending loads.

ii ''111' ' " !II .... ' lr

i!, +t I I i_ I ,

30I

da wisee_..e_____ 2o ............ _- --_- --,_. --,-- -_ ........ _--root , .. , FII_ ,bending 1o _ =L '_: ; :_ ;-:- -- ".... " - ".... - J-'.,k :--moment ' : -.L/'/ ',ii:_-._" ' "- ' 'iii'.. "X-' "b " "

(kNm) o , ?i' ' 'I I _ I I

10

_;i !i; !,.30 "r--'-r .: _ _ m Bladel []

, ' , , i!+ _ ,/ , ,_I , , _ l:llaaez []AeroStar 2o -- -,..... ,..... _-p- ._ .- -/L_l_.... ,..... _ .... _.... +.... I. Blade3 iflapwise , _ , , +_,/I,: Rti / , ;: , ,root

B "_.1 I i ' rv ' : _ i I I +1. .'., i ' _ ' . _ ,"-- .'_.,---,--h= ,_in...n..n= ,. ......... . , , . . ,moment o -(kNm) ................

--10

1 2 3 ;! 4- 5 6 7_: 8 9 10 11

_._.-_:_:_._:__--_:_.-_-.-_-_:r.-_:_:::_. seconds

Figure 2. An example of a rotor encountering a coherent turbulent structure.

I .i ii6O

(A) +o ............ ;................... .......... I .........vw []

a v,dlnstantane°us 20 ........." "."""_-..-) :" ....."._",-Y__' ' i-._,'_'u'..̂.............i"-' , ._':" ,,'_-"..e_no._s o - ,-+ ,.. W :, .t , :_....,..,.'.i., ,stresses 'A A_ //.f_J :_ :_. ":_ii "'\/ """'V '/4., V ,1y /,_ •lamns_2 -20 - _/-\-Y-_-/4_- _- __............... Sl-- - _ ..... _ .......... _1 ......\/ 't^l/I ' • '_. ' _ .... _I '

"40 _d ..... _.___L_L_ ..__1 ..... :___'_- .... LL .... I..... 1 .... _LJ___L_.I I i I 1 i i I i i I

I I I . F I I I I ', I I

-60 I 1 4, _ I i l I, , 1 1 i t "

(B) 80 ....., , ' ', "'........ " '! ' " ' I' m, ._::.--..,[][]60 _j .... L .... L_..__l .... -__.,___ L .... ,.J .... +.... .t .... L--I {It) l i

' ' ' ' --'".......... ' ' ' I ' iEstimated +o -_<.-r .... ,--_---7--_ ...... _ .... ,..... ,.... _.... r--i • []' "_-- , .......... , .". ,........ I Helicityi Ivorticity 20 - _ .... [ ....__Lj "5"".... _-.:/-LY_:+:.-.5_-"_Y--._-2_-J . =components o ,..11,c_ -_,. .......... .,+__---. _;.......... " -7-_. : .. _-' I

I .... i ..... I i _;._ / i I'_ I ; I t l

trau,sec, -:;'o -, r ,.... :X---/' .... r- -'_..... II I I : , ._ _ I i _ ', I I-4o ,, : .... :-.-.xxTz., ,, 1+ ,, + , , /

--60 - _ .... _ .... ,- - " - -,-- -:- _ .... _ .... _-".... _.... _ .... _ .... _- - 1I i I I " i I I I i I I I

- 80 i ; i i " i i 1 .... 1 ', I 1 ,J"

1 2 3 + 5 6 7. 8 9 10 11........................................seconds

Figure 3. Turbulence characteristics at hub height associated with the response shown in Figure 2.

I

!

Figure 4a. Largest positive (tension) peak stress cycle observed for the NREL rotor.

._

I I I (J _i I i i: I 1 1 1 I '

40

: , , , , rr\ ....i I I i I ?" I I !

30 -;..... ..... :..... '.... _ r .... ,..... ..... ,.....

root "

/

flapwise :. ', : : : :. , ,, :, / Helicity-;..... ,..... ,..... ___ _-- _- -_.... ,- --,..... :..... _moment 20 _ _ , , _

(kNm) , , I ,1 I.I ,X_ ,/_ , , '.-_

"_? _ Z_)__ X___:!_,___ Ii'_o- ....,._ _,,,_,.,_0 _':;'_ ....'..........__-___.:Z..L_...':----__--

0 1 2 3 4 5seconds

Figure 4b. Hub-height vorticity/helicity estimates aligned with peak flapwise moment.

7

30

I

I

t

+ I

I I

-10 ,

0 1 2 3

seconds

Figure 5a. Large negative (compression) peak stress cycle observed for the NREL rotor.

Figure 5b. Hub-level vorticity/helicity estimates overlaid on the recoid containing the negative peak.

8

I

July 27, 1990 01:11:32.2, ._;;__, ,

,o --- ....,'.... ...........,30 ....

July 25, 1990 06:15:50.4L I

40 ............ -_...... , ...... 20 ...... ± - -

I I /jI I

30 ............. 5 ...... T ......I J

20 ............ _ ...... z___ 0 rI r i

10 __ + 10 ................... _- .......

o _ /%, .. /[ o sec( nds 1I I

L...... i_ _-_ ""N_ ..... "- ......-,o , ,, ii', V/;:. ;_J

i 4_,- _ I......o _ii_'_seconds 1

Figure 6. Largest root flapwise bending cycle for the Micon 65 with the NREL rotor.

un_table flow near-neutral flow stable flow' _........ '........ _........ I........ _"'": '........ ' ........ '........ 1........ ,,,,,. ,........ ,........ D........ ] ........ ,,J,.

iI

,J

'_'_ 0.1 0.1 0.1

E

?

c"

i b

1000 100 10 1 0.1 (.c) 1000 100 10 1 0.1 (.¢) 1000 I00 10 0.1 (.c)o.ot ....',"'", ',', ',,,,,1"":' ,', ',,,,,',""',' ,', :,,,,1'"'',' ,', :,,-l"' o.ol ....I"" ',' ,', ',,,.,1'"': ',' - :,,,.I'"' : ',' • :,,,,I"'", ' ,', ',.,,,I'" o.o1 ....',"" ',' ,,, n,,,',""',' .', ',,,,.I""', ' ,', : ,,,,',""',' ,', ',,,,.I".'o.oo1 o.ot o.1 1 lO (Hz) O.OOI 0.01 0.1 _ 10 (Hz) 0.001 O.Ot 0.1 1 tO (Hz)

i ,,,,

m smooth terrain reference iim upwind of wind farm I:_'_:_: downwind of wind farm

Figure 7. Comparison of predicted crosswind (v) velocity spectra for a 12 ms'l mean wind speed.

9

Figure 8. Observed vertical coherence decrements with linear regression lines shown.

10

,r,mF II !, ' ii .........