2024安装在跟踪器上的双面光伏组件的风速和后部玻璃破裂报告（英文版）-DNV

DNV

WHENTRUSTMATTERS

WINDSPEEDANDREAR

GLASSBREAKAGEON

BIFACIALPVMODULES

MOUNTEDONTRACKERS

Correlationbetweenmid-levelwindspeedandrearglassbreakagesonnon-largeformatbifacialPVmodulesontrackersinasolarfarm

Authors:

DarrylWang

AndiHermawanEvanWoolard

WHITEPAPER

WindspeedandrearglassbreakageonbifacialPVmodulesmountedontrackers

CONTENTS

Abstract3

1.Introduction4

2.PVsystemsdesign5

3.Method5

4.Resultsanddiscussions6

5.Summary9

Acknowledgement9

References9

-2-

WindspeedandrearglassbreakageonbifacialPVmodulesmountedontrackers

ABSTRACT

Totheauthors’knowledge,thisisthefirstreportwithonsite-recordedevidencessuggestingthat

not-highbutmid-levelwindspeedisamajorinfluenceonoratriggerforlarge-scale(>15%)rearglassbreakagesonbifacialPVmodulesmountedontrackersinasolarfarmlocatedintheAsiaPacificregion.

Timeandspatialcorrelationswerefoundbetweenmid-levelwindspeedrecordsandrearglassbreakagesonnon-largeformatglass/glassbifacialPVmodulesmountedonsingle

portrait(1P)single-axistrackers.Fortimecorrelation,

intra-daymid-levelwindspeedclusterswerefoundto

consistentlyprecedemajoronsiteO&Meventsthatincludes

orcanberelatedtoglassbreakages.Goodspatialcorrelationwasalsoobserved,wheretheareaswithhighestglass

breakageconcentrationswerefoundclosetoaweather

stationthataccountsfor~75%ofallthemid-levelwindspeedrecords.Themaximumwindspeedis,however,wellbelow

thresholdsthatcanexertwindloadexceedingthetypical

moduledesignwindload,representinganengineering

missinglinkthatisconsistentwithotherworks.Theglass

crackpatternsfoundonsitearealsodiscussed.ItishopedthisworkcanhelpthePVR&Dcommunitytosharpenthefocus

towardsfindingthemissinglink.

-3-

WindspeedandrearglassbreakageonbifacialPVmodulesmountedontrackers

-4-

1.INTRODUCTION

GlassbreakageinbifacialPVmodulesinstalledinsingle-axistrackers-basedsolarfarmshasbeen

increasinglyreportedinrecentyears[1]-[6].Whileearlyattentionontrackermodulefailureswason

2Ptrackersduetotorsionalgalloping[1],since2020thereisagrowingbodyofreportsforrearglass

breakagesafflictingsolarfarmsinstalledwith1Ptrackerdesignsandlargeformat(≥2.5m2[4])bifacialglass/glassmodulessupportedonitslongedgesbyshortmounts[2]-[4](Figure1showsanexample).

Importantly,suchbreakagesoccurintheabsenceofhigh

windspeed(let'scallthis"thresholdwindspeed")eventsthatcouldexplainthemodulebreakagebasedonthemodule

manufacturer'sdesignwindloadonapprovedmodule

mountingdesignwithshort(usually~400mm)mountsonsuchtrackerdesigns[2]-[6].Thisrepresentsanexistingengineeringknowledgegapofanominousfailuremodethat,giventhattrackersystemsarealreadyamainstreamPVdesign[2],couldhavewideimplicationsintheglobalsolardecarbonizationdriveifleftunresolved.Whilemanyhypotheticalexplanationswereoffered[2]-[6],therehasbeenlittleclearevidencetofurthersteertheattention

towardsanyofthehypotheses.

Inthiswork,wereportanencounterwith>15%ofbifacialPVmodulessmallerthanlargemoduleformatson1Ptrackersafflictedwithrearglassbreakagesacrossthesolarfarm.

Moreover,mid-levelwindspeedlevelswerealsofoundto

correlatewellwithglassbreakagesbothspatiallywithinthesolarfarmandchronologicallywhencomparedtomajor

infieldO&Mevents(includeglassbreakages),whichsuggestthatglassbreakageswereinfluencedbymid-levelwind

speedeventsthatarewellunderthethresholdwindspeeds.Thisisthefirstreportwithsuchcorrelationbetweenonsiterecords.

Allfiguresinthispublicationarerepresentativeillustrationsoftheactualencounter,andsomedetailsareintentionallykeptambiguouswithwideranges,buttheoverallpresentation

herecanstillconveyasimilarstoryoftheactualencounter.

FIGURE1

Atypical1Psingle-axistrackerdesignwithframedbifacialdouble-glass144half-cut-cellsPVmodules.Imageistakenfrom[7].

WindspeedandrearglassbreakageonbifacialPVmodulesmountedontrackers

-5-

2.PVSYSTEMSDESIGN

MODULE

Reputablebrand,

144half-cutc-Sicells

bifacial,2mmglassfront

andback,3junctionboxes,

30-35mmframes.

SOLARFARM

10-50MWp,2-7yearsold,locatedinAPAC.

Thesolarfarmdetailsareasfollows:

MODULEAREA

MOUNTINGSTRUCTURE

Between2.0-2.3m2

IPsingle-axistracker,

shortmountattached

tothemid-sectionof

modulelong-sideframes.

Sitelayoutisapproximatelyrectangular,withtwometeorologicalstationseachcontainingananemometer(windspeedsensor;~3mhigh)recordingperminutewindspeeddatasinceCOD.

3.METHOD

Twodocuments,compiledbythesolarfarmownerandrecordedontwo

differentdates,countingmoduleswithglasscracksandtheirlocationsin

thesolarfarmwerereviewed.ThelatestdocumentisdatedwithinQ2-Q32023,whiletheearlierdocumentisdatedwithin12monthsoftheformer.MonthlyO&MrecordswerereviewedsincetheplantCOD.Meteorologicaldata,includingwindspeedatperminutebasis,fromthetwoweather

stationsinthesolarfarmwerealsoreviewed.Acomprehensiveassessmentwasconductedacrossthevariousdata.Asitevisitwasalsodonetoassessthemoduleswithglasscracks.Assessmentswerealsodoneonmodule

productionbatchesandglasssuppliers/batchesbutnoclearcorrelationtotheglassbreakageswereuncovered.Thesubsequentfocusisoncrackpatternsandwindspeeddata.

WindspeedandrearglassbreakageonbifacialPVmodulesmountedontrackers

-6-

4.RESULTSANDDISCUSSIONS

A.Crackpatterns

Fromonsiteinspections,glasscrackswereonlyfoundon

therearglassofthebifacialmodules.Thecrackpatternsaresimilarthoseinheat-strengthenedglassesshowninFigure2part(b).Thisisconsistentwiththeheat-strengthenedglassesusedinthemodulesasshowninthemoduledatasheet.

AnnealedHeat-strengthenedFullytempered

FIGURE2

Crackpatternsforannealed,heat-strengthenedandfullytemperedglasses[8]

FIGURE3

Typicalcrackpatternsfoundintherearglassesofthecrackedmodules.

Theredlinesarecrackpatternsthatmostlyruninthedirectionofthelong

sideframe,whiletheorangelinesarecrackpatternsthatrunsclosetothe

short-sideframes,withsomeeventuallyturntowardsthecenterofthemodule.

FRONTGLASS

ENCAPSULANT

2

3

REARGLASS

1

4

FIGURE4

Compressive(yellowarrows)andtensilestresses(redarrows)exertedoneachlayerofthelaminateinalaminatebow.Attheglass/encapsulantinterfaces

(3&4),thereareinterfacialcounter-forces,whilethecompressiveandtensilestressesattheair-sidesurfaces(1&2)arefree-standing.

Mostoftherearglasscrackswerealsofoundtoruninthe

directionparalleltothemodulelong-sideandendinginto

oneofthe3junctionboxes(Figure3).Thelong-sidecrack

routeandrearglasscracksisconsistentwithanonsite

observationthatmostofthemoduleshavemildbutvisuallyapparentdownwardbowinginthecenteralongthedirectionoftheshortside(notethattrackermountsareonmid-

sectionsofthelongsidemoduleframes).ThesecrackroutesinFigure3areconsistentwiththepresenceofdrillholesintherearglassatthe3junctionboxlocationsthatcanactasstressconcentratorspots,thepropensityfortherearglass

(vsfrontglass)thatisundertensilestresstobreakfirstwhenthemodulelaminateisinthedownwardbowingstate

(Figure4),andtheexpectationthatglasscrackswouldpropagateinthemodulelongsidedirectionwhenthebowingisalongitsshortside(Figure4).

Whilethecrackpatternscanbeexplainedbymodulebowanddrillholesinnon-temperedglass,itisnotconclusivethattheseareenoughtoactasthetriggersoftheglass

breakages.

WindspeedandrearglassbreakageonbifacialPVmodulesmountedontrackers

-7-

B.WindspeedrecordsvsO&Mrecords

1

2

FIGURE5

Arepresentativerectangularsketchofthesolarfarmareawitha

representativespatialheatmap(inblue)ofcrackedmodulesacrossthesite.Theyellowsquaresrepresentthetwoonsitemeteorologicalstations1and2(let’scallthem‘ST1’and‘ST2’).

Shareofwindspeedrecords>10m/s

80%

70%

60%

50%

40%

30%

20%

10%

0%

ST1ST2

FIGURE6

Shareofwindspeedrecordcumulativecountsexceeding10m/sinthetwometeorologiclstationsinFigure5sinceCOD.

Fromdesktopanalysisofthelatest(inQ2-Q32023)modulecrackcount,thebreakagecountis>15%ofallmodules.

Figure5showsthesolarfarmsitewithaspatialheatmapofthebreakagesandlocationoftwometeorologicalstations

(ST1)and(ST2),eachwithananemometer.Thisheatmap

showsthatwhiletheglassbreakageisdistributedoverthe

site,thereareselectedareaswithhigherconcentration

(especiallyrightside).Theearliermodulecrackcounthas

10-15%breakagerateandshowsaspatialdistributionsimilarasFigure5butislessdense.

Fromtheper-minutewindspeedrecordssinceCOD,a

cumulativecountofmid-levelwindspeedlevels(>10m/s)(Figures6and7)showsthat(1)themid-levelwindspeed

countismuchlessthan0.01%oftotal,whichappearsto

showmid-levelwindspeedsarenotcommonon-site,(2)

ST2accountsfor~3xmorecountsthanST1(Figure6)whichappearstocorrelatewellspatiallywiththeglassbreakageheatmapinFigure5.

Itisalsonotedthatthemaximumwindspeedisunder

15m/s(Figure7).Basedoninternalcalculations,itis

~2.5xlowerthanthethresholdwindspeedneededtoexert

1600Padesignwindloadonthemodule/trackershortmountdesign,basedonthemodulemanufacturer’sinstallation

manual.Thisisconsistentwithfindingselsewhereofbifacialmodulecrackingintheabsenceofnotablewindspeed

events[2]-[6].

Figure7showsatimeplotofthemid-levelwindspeed

recordssinceCOD,whichrevealsthatmid-levelwindeventsoccuronveryfewdays,withmid-levelwindcluster(≥3countsaday)daysoccurringonlyonafewselectdays(veryrare).

Interestingly,fromthemonthlyO&Mrecordsandother

documentations,wefoundfourO&Mactivities(yellow

overlaysinFigure7),eitherdirectlyorlikelyrelatedtomoduleglassbreakages,thatappearstobeconsistentlyprecededbytherarelyoccurringaforementionedmid-levelwindcluster

days(seenumbersatoptheyellowoverlaysinFigure7).

1.Modulerearglasscracks,followedbyrepeatedinspections(overmultipleconsecutivemonths)

2.Insulationresistance(Riso*)issuesfrominverters

3.Moduleglassbreakagecountsdocumentation(earlier)

4.Moduleglassbreakagecountsdocumentation(latest)

*RisoissuesininverterscanbecausedbymorningdewselectricallyconnectingtheinternalPVmodulecircuitrytogroundthroughglasscracks.

2

1

4

3

Windspeed[m/s]

1413.5 1312.5 1211.5 1110.5 109.59

ST1ST2Days[A.U.]

FIGURE7

Mid-levelwindspeed(>10m/s)recordsmeasuredintheonsitemeteorologicalstations1(ST1)and2(ST2)fromFigure5;theyellowboxoverlaysshowwhenandhowlongarethefourmajorO&Mevents.Thex-axisfullrangecanbeupto7years.

WindspeedandrearglassbreakageonbifacialPVmodulesmountedontrackers

-8-

Alltheabovewindspeedandglasscrackobservations

showthatthemoduleglassbreakagescorrelatewellboth

spatiallyandtemporallywithmid-levelwindspeedevents,

andsuggestthatthesemid-levelwindspeedeventseither

haveamajorinfluenceoristhetriggerforthemodule

glassbreakages.Moreover,withthemaximumwindspeed

recordedbeingmuchlowerthanthethresholdwindspeedneededtoexceedthemoduledesignwindload,theexistingwind-loadengineeringdesignunderstandingof1Psingle-

axistrackers(includingwithshortmounts)maybeinsufficienttomitigateagainstinfieldmodulefailuresmodes[2]-[6],

whichrepresentsamissinglink.Thisreportalsoshows

bifacialglassbreakagescanalsohappentonon-large-formatmodules.

Severalhypotheseshavebeenraisedinpreviousworks

[2]-[6]toexplainforthisunexpectedfailureintheabsenceofhighwindspeeds.Increasedmodulesizeswithshorter

trackersupportbeamshavemadethemoduleglassesevolvefromaloadsupportedbythetrackerstoaloadbearing

componentinstead[2]-[4],makingexistingtestingregimesnolongersuitabletotestforinfieldfailuremodes.

Thefailurecouldalsobeduetoasymmetricwind-loads

inmodules-on-trackersystemspresentinfield[4]-[6],

ormultiplestresseshappeningallatonceduringinfield

operations[9],somethingwhichstaticloading(SML)testsonPVmoduleswouldnothavesufficientlycovered.Therearealsoreportsofdynamicmechanicalload(DML)tests

replicatingtheglassbreakagepatternsobservedinfield

[4]thatweremissedbySMLtests.Itishopedthisreport

canhelpthePVR&Dcommunitytofurthersteerthefocustowardsfindingthemissinglinkintheexistingengineeringunderstandingofthismatter.

WindspeedandrearglassbreakageonbifacialPVmodulesmountedontrackers

-9-

5.SUMMARY

Wereport,forthefirsttime,evidencessuggestingthat

not-highbutmid-levelwindspeedisamajorinfluenceon

oratriggerforlarge-scale(>15%)rearglassbreakageson

bifacialPVmodulesontrackersinasolarfarmlocatedintheAPACregion.Timeandspatialcorrelationsinthesolarfarmwerefoundbetweenmid-levelwindspeedrecordsandrearglassbreakagesonnon-large-formatglass/glassbifacialPV

modulesmountedonsingleportrait(1P)single-axistrackers.Themaximumwindspeedis,however,wellbelowthresholdsthatcanexertwindloadexceedingthetypicalmoduledesignwindload,representinganengineeringmissinglinkthatwerealsoreportedinotherworks.Theglasscrackpatternsfoundonsitearealsodiscussedanditscausesareexcludedasthetrigger.ItishopedthisworkcanhelpthePVR&Dcommunitytosharpenthefocustowardsfindingthemissinglink.

ACKNOWLEDGEMENT

TheauthorswouldliketoacknowledgeMichaelNiu,RachelOh,andLowZhengHuaforvaluabletechnicaldiscussions,site

inspectionandoverallprojectcontributions.

REFERENCES

[1]D.Valentín,C.Valero,M.Egusquiza,A.Presas,Failureinvestigationofasolartrackerduetowind-inducedtorsionalgalloping,EngineeringFailureAnalysis,135,106137,2022

[2]T.Weber,“Glassbreakage–agrowingphenomenoninlarge-scalePV”,inPVMagazineWebinar,20November2023

[3]C.Sillerud,“CurrentissuesinPVmodulereliabilitytesting”,in50thIEEEPhotovoltaicSpecialistConference,2023,Plenary.

[4]H.Hieslmair,“Stressconcentrators”,DNV,2023,

/article/stress-concentrators-246984

[5]D.Chang,“Modulewindloadresistance:Standardsvs.reality”,inPVMagazineWebinar,30August2021

[6]S.vanPelt,“Modulewindloadresistance:Standardsvs.reality”,inPVMagazineWebinar,30August2021[7]

/monoline-plus-solar-tracker/

[8]P.C.Louter,“Adhesivelybondedreinforcedgla