Sei sulla pagina 1di 18

Petrel TIPS&TRICKS from SCM

Knowledge Worth Sharing

HistogramsandSGSModeling
Histogramsareuseddailyforinterpretation,qualitycontrol,andmodelinginPetrel.ThisTIPS&TRICKSdocument brieflydescribeshowhistogramsarebuiltandusedforQCbutprimarilyfocusesontheiruseinPetrophysical propertymodeling,specificallySequentialGaussianSimulation(SGS).PorosityfromtheGulfaksdemodata distributedwitheachPetrelreleasewillbeusedtodemonstratehistograminputtoSGS.Wewilldemonstratethe dramaticimpactthesehistogramshaveonpropertydistributioninthemodel.Probablythemostimportantpointto learnfromthispaperisthathistogramsforeachzoneandfaciesofthemodelmustbereviewedtoensurethatthe modelisreasonable,representsthedata,andmatchesthegeologicinterpretation. Figure:Porosityhistogramsforonezonefrom:initiallogs(topleft)andupscaledcells(topcenter),andfrom modelsbuiltusinghistograminputcontrol:modelusingupscaledcells(topright),modelusingreasonabledata analysis(bottomleft),modelusingmeanandstddev(bottomcenter),andmodelusingextremedataanalysis (bottomright).Notetheamountofvariationthatcanexistforthesamerockunit.

Petrel is a mark of Schlumberger

4801 Woodway Drive, Suite 150W Houston, TX 77056 www.scminc.com info@scminc.com


2011 SCM E&P Solutions, Inc.

HistogramDisplays
Petrelhastwomethodsforbuildinghistogramdisplays.TheSettingsDialoghistogramisquickandusedforQCofa specificobject.TheHistogramWindowisusedforcreatingfinaldisplays,comparingseveralobjectsonthesame histogramdisplay,andforconvertinghistogramstodistributionfunctions(data).

SettingsDialogHistogram
TheHistogramtabofanobjectsSettingsDialogissecondonlytotheStatisticstabasthemostusedtoolinPetrel. TheStatisticsandHistogramtabsarethefirstlineofattackforQCandshouldbecheckedfornearlyeveryobject thatisbuilt. ToviewtheSettingsdialogsHistogramdoubleclickontheobjectandgototheHistogramtab.Theavailable parameterswillvaryslightlydependinguponwhethertheobjectisanupscaledpropertyorjustanobject.Forall objectsyouhaveparametercontrolover: Numberofintervals(bins) Displayaspercentorcount(percentletsyoucompareobjectswithdifferentcounts) Logorlinear Minimumvaluetodisplay Maximumvaluetodisplay Printhistogramtoaprinteryouselect Copybitmap(allowscopyingtoclipboard,thenyoupastetoInputtabanddisplayinotherwindowsorto anexternaldocument)

Whenthehistogramrepresentsanupscaledpropertyyouhaveadditionalcontrolover: Whichzonethehistogramrepresents Whatvaluesaredisplayed:Log,Upscaledcells,allcells,oranycombination Useactivefilters Volumeweightthecount(usefulwhencellsofthemodeldiffersignificantlyinsize)

Youcannotcontrolthecolorsofthebarsonthistypeofhistogram. Figure:Histogramtabrelatedtoageneralwelllog(left)andapropertymodel(right).

Petrel is a mark of Schlumberger

4801 Woodway Drive, Suite 150W Houston, TX 77056 www.scminc.com info@scminc.com


2011 SCM E&P Solutions, Inc.

HistogramWindow
TheHistogramWindowallowsoneorseveralobjectstobedisplayedatthesametime.Thisdisplayhasmore versatilitythantheDialogHistogramandisusedmoreforanalysisandfinaldisplaythanforQC. TocreateaHistogramWindowclickontheWindowbuttonatthetopofthePetrelwindowandselectNew histogramwindow.Resizethewindowandthencheckthebuttoninfrontoftheobjectorobjectsyouwantto displayinthewindow.Theparametercontrolsforthiswindowarefoundintwoplaces,someareaccessedthrough theiconstotherightofthedisplayareaandothersusingtheViewportSettingsdialogforthehistogram.Thisdialog iconorbygoingtotheWindowstabofthePetrelexplorerHistogram canbegottentobyclickingonthe windowHistogramsubfolderanddoubleclickingonthesubfolder.Fromthisdialogyoucontrolmostofthe parametersyoucontrolledintheSettingsdialogHistogram.Inadditionyoucanlinkmultiplehistogramwindows togetherinacoordinategroupsothatachangetoonesdisplayformchangesall.Youcanalsolinkthemtoavisual groupsothatwhatisdisplayed(turnedonoroff)inonememberofthegroupisadjustedsimilarlyinallother membersofthatgroup. Figure:DoubleclickontheHistogramsubfolderintheWindowstabofthePetrelexploer(left)todisplaythedialog forcontrollingtheparametersoftheHistogramwindow.

Petrel is a mark of Schlumberger

4801 Woodway Drive, Suite 150W Houston, TX 77056 www.scminc.com info@scminc.com


2011 SCM E&P Solutions, Inc.

Iconstotherightofthedisplayareacontrolavarietyofthingsthatareeasilyturnedonandofftotest.Afewof interestare: Selectandeditoraddpointstoadisplayedfunction Selectandeditlineofadisplayedfunction Displayhistogramasbars Displayhistogramasaline

Showcdf(cumulativedistributionfrequency)curve Displayaspercent Volumeweighting Displayhistogramforallcellsinselectedproperty Displayhistogramforupscaledcellsinselectedproperty

Displayhistogramforlogvaluesassociatedwithupscaledcellsinselectedproperty RangeselectpointsalongtheXaxis(worksbutdoesntseemedtobeusedforanything) Createnewdistributionfunction Smoothfunctionwhenoneisdisplayed Usesplinetoconnectpointsofadisplayedfunction

Oneofthemostfrequentlyaskedquestionsis:Howdoyouchangethecolorsofthebarsonthehistogram?Thisis doneinthreeplaces. Property:DoubleclickonthePropertiesfoldercontainingthatpropertyandgototheStyletab.Alterthe parametersintheHistogramareaofthatdialog. Welllogs:DoubleclickontheWellsfolderandgototheHistogramtab.AlterparametersundertheHistogramarea ofthisdialog. Objects:DoubleclickontheobjectandgototheStyletab.AlterparametersundertheHistogramareaofthisdialog.

Petrel is a mark of Schlumberger

4801 Woodway Drive, Suite 150W Houston, TX 77056 www.scminc.com info@scminc.com


2011 SCM E&P Solutions, Inc.

Figure:Alterhistogrambarcolorsandotherdataspecificparametersusingoneofthesethreedialogs:Properties folderSettings(left),WellsfolderSettings(center),andObjectSettings(right).

Petrel is a mark of Schlumberger

4801 Woodway Drive, Suite 150W Houston, TX 77056 www.scminc.com info@scminc.com


2011 SCM E&P Solutions, Inc.

Figure:Histogramwindowwithapropertyanditsthreeassociatedhistogramsdisplayed.Thecdfforeachisalso displayedscaledtothehistograms.

SubsetLogsforHistogramDisplay
Welllogsusuallystartoutrepresentingtheentirewellorsignificantportionsofthewell.However,sometimesitis desirabletodisplayoranalyzeonlyasmallportionofthelog.Forinstance,youmightwanttoanalyzetheportionof thelogcorrespondingtoonlyonezoneortoonlyonefaciesinonezone.Histogramscreatedforsubsetsofthe entirelogareoftenconvertedtodistributionfunctions(discussedbelow)sothattheycanbebotheditedandused asinputtotheSGSmodelingprocess(discussedbelow).Thelogcalculatorisusedtooutputpartsoflogsthat correspondtopartsofotherlogs.Typicalexamplesmightbe: Outputporositywherezonelogis2 Outputporositywherefaciesis3andzonelogis2 Outputporositywherethenetnonnetlogis1

Eachoftheseoperationswouldcreateanewporosityloghavingvaluesonlyintheportionofthewellwheretheif statementwastrue.ThesenewlogscouldthenbedisplayedinaHistogramwindowandthedatacapturedasa distributionfunction.

Petrel is a mark of Schlumberger

4801 Woodway Drive, Suite 150W Houston, TX 77056 www.scminc.com info@scminc.com


2011 SCM E&P Solutions, Inc.

Figure:Welllogcalculatorusedtosubsettheporositylogbasedonnetflag,facies,andzone.

Petrel is a mark of Schlumberger

4801 Woodway Drive, Suite 150W Houston, TX 77056 www.scminc.com info@scminc.com


2011 SCM E&P Solutions, Inc.

Figure:Wellsectionshowingtheoriginalporositylog,thelogsusedtosubsetporosityandthefinalporositylog (left)andthehistogramforthenewlog(right).

DistributionFunctionfromaHistogram
AdistributionfunctionisaseriesofXYpairs.WhenthefunctionisbuiltfromahistogramtheX representsthevalueatthecenterofahistogrambinandtheYisthecount(numberofvalues)inthatbin. Onceafunctionhasbeencreateditcanbedisplayed,edited,exported,orusedasdataforproperty modeling.
Figure:Histogram(left),thedistributionfunctionbuiltfromit(center),andtheXYvaluesdefiningthefunction (right).

Petrel is a mark of Schlumberger

4801 Woodway Drive, Suite 150W Houston, TX 77056 www.scminc.com info@scminc.com


2011 SCM E&P Solutions, Inc.

CreateDistributionFunction
UsethefollowingstepstocreateaDistributionFunction. 1. DisplayonedatasetonaHistogramwindow.(Ifyoudisplaymorethanonedatasetyoucannotgeneratea functionrepresentingthedata.Youwillbeallowedtoonlygenerateafunctionasaconstantvalueorasa normaldistributionbyspecifyingmeanandstandarddeviation.) 2. SetthenumberofbinstoyourlikingusingtheSettingsdialogforthehistogramwindow,reachedusingthe iconandthenalteringtheNumberofintervalsortheIncrementparameter.

3. CreatetheDistributionFunctionby: a. ClickingontheCreatenewdistributionfunction icon.ACreatedistributionfunctiondialogwill appear b. Enteringthedesiredname c. UncheckingtheOverwritelastparameter(Thisdefaultstocheckedwhich2timesoutof3willcause youtowipeoutthepreviousfunctionyoubuiltandoftenwipeourmanyofthepreviousfunctions beforeyoucatchit,aproblemthatshouldbecorrectedinacomingreleasebutmaybenot.) icon d. SelectingtheFitdistributiontoActivehistogramcolumns e. ClickingExecutetobuildthefunction 4. Notethatthefunctionisdisplayedonthehistogram 5. GototheInputtabandseethatthefunctionhasbeenplacedatthebottomandhasnotoverwrittenanalready existingfunction. Figure:Createdistributionfunctiondialog(left),Histogramwindowwithhistogramandgeneratedfunction displayed(center),andInputtabshowingtheresultingfunctionaftermanuallymovingitintoafolder (right).

Petrel is a mark of Schlumberger

4801 Woodway Drive, Suite 150W Houston, TX 77056 www.scminc.com info@scminc.com


2011 SCM E&P Solutions, Inc.

EditDistributionFunction
OftenthereasonforcreatingadistributionfunctionfromlogsalreadyinPetrelistoaltertheformofthefunctionto betterrepresenttheinterpretation.Forexample,theremaybetoolittledatatoproperlydescribetheporosity distributionofaparticularfaciesorzone.Inthatcaseyouwoulddisplayahistogramofthedata,createthe distributionfunction,andtheneditthefunctionuntilitmatchesyourinterpretationofhowthedistributionreallyis intherocks.AlthoughitispossibletoeditafunctionusingitsSettingsdialogFunctiontab,itisbesttoeditusing aHistogramwindowsothatabackgroundhistogramcanbedisplayedforreferencewhileediting. Figure:HistogramfromZoneNess1porositylog(left),Distributionfunctiondisplayedonsamehistogram(center), andeditedhistogramwithsamehistograminbackground(right.

ThestepsusedtoeditaDistributionFunctionare: 1. OpenaHistogramwindow 2. Displayabackgroundhistogramfromdatatouseasreferencewhileediting(usuallythisisthehistogram fromwhichthefunctionwasderived) 3. Displaythepreviouslycreateddistributionfunctionthatistobeedited. 4. ClickontheSelectandedit/addpoints icon 5. Editthepoints Movetheexistingpointstodesiredlocations Touchapointandhitthedeletekeytoremoveit(oftentherearemorepointsthanneeded) Touchalineandmovethecursortocreateanewpoint

HistogramControlforSGSModeling
SequentialGaussianSimulationisacommonlyusedalgorithmforbuildingpropertymodelsofcontinuousvariables suchasporosity.Thisalgorithmusesasinputahistogramofthedatasdistribution.Itassumestrendshavebeen removed,thedataarenormallydistributed,andthatthedistributionhasamean0andstandarddeviation1.The Petrophysicalmodelingprocesshasseveralwaystoinputhistogramdata:upscaledcellsrepresentingthelogdata, DataAnalysisprocessresults,meanandstandarddeviationvaluesrepresentinganormaldistribution,anda DistributionfunctionfromtheInputtab.Regardlessofwhichmethodisusedasinputtheprogrameventually appliesaNormalScoretransformation(derivedfromthehistogram)tothedatatoconvertthedatafromitsinput formtoanormaldistributionwithmeanzeroandstandarddeviation1.Thenormalizeddataarethenmodeled usingparametersyouspecifiedfortheSGSalgorithm.Thenewmodelsnormalizedcellvaluesarethenback transformedintoporosityunitswithadistributionmatchingtheinputhistogram.

Petrel is a mark of Schlumberger

4801 Woodway Drive, Suite 150W Houston, TX 77056 www.scminc.com info@scminc.com


2011 SCM E&P Solutions, Inc.

10

Figure:HistogramportionofprocessingflowduringSGSmodeling.Bluerepresentsinputinoriginaldataunits,light redrepresentsinternalNormalScoreprocessing,greenrepresentsSGSmodelinginmean0standarddeviation1 units,andyellowrepresentstheoutputmodelinoriginaldataunits.

Porosity data

Porosity histogram

Define Normal Score Transform

Normal Score Transformation

SGS modeling

Porosity model

NormalScoreTransformationandBackTransformation
Thenormalscoretransformationinvolvesfoursteps: 1. Buildacumulativedistributionfunction(inputcdf)fortheinputhistogram. 2. Buildacumulativedistributionfunction(normalcdf)foramean0standarddeviation1histogram 3. Mapeachinputdatavalueontheinputcdfbynotingitspercentileposition(Yaxis)onthecdfcurve 4. Extractanormalizedvalueforeachinputdatavalueby:locatingitscdfpercentilepositiononthenormal cdfcurvesYaxis,goingtothatlocationonthecurve,andreadingoffthenormalizedunitsvaluefromthe Xaxis. Theprocessisdoneforeachupscaledcellinthepropertymodel.Onceinnormalizedspace,theSGSalgorithmis appliedusingallotherparametersyouspecified.Thiscreatesamodelwithcellshavingamean0andstandard deviation1(ifnospecialtrendswereapplied).Thesemodeledcellvaluesarethenbacktransformedbywalking throughthefourstepsaboveinreverseorder.Theresultingvaluesarenowinporosityunits.Anyothertransforms (outputtruncation,log,trend,etc.)thatyouappliedtotheinputdataarereverseappliedatthispointtocreatethe finalporositymodel.Notethatthesameinputcdfcurveisusedinbothtransformdirections.Thismeansthatthe originalhistogramsformwillbereproducedintheoutputmodelunlessspecialtrendsorothercontrolswere appliedduringmodelingthatforcedtheresultingmodeltodeviatefromanormaldistribution.Eventhenthe resultswillbesimilartotheinputhistogramwithminordistortions.

Petrel is a mark of Schlumberger

4801 Woodway Drive, Suite 150W Houston, TX 77056 www.scminc.com info@scminc.com


2011 SCM E&P Solutions, Inc.

11

Figure:Normalscoretransformofupscaledcellvaluesfromporosityunitstomean0standarddeviation1units (left),SGSmodelinginnormalizedunits(leftmiddle),Normalscoretransformofallcellvaluesbackfromnormalized unitstoporosityunits(rightmiddle),andfinalmodelinporosityunits(right).

HistogramfromUpscaledLogs
Ifyouhavealargenumberofwellswithlogdatawhichcreateahistogram(s)youfeelisrepresentativeofthe propertysdistributionthenyoucanusetheupscaledvaluesfromthosewellstodefinethehistograminputforSGS modeling.WhendefiningthePetrophysicalmodelingparametersyouwillusetheDistributiontabtopointtothe upscaledlogs.Inthistabyouwillwanttocheckseveralparameterssincetheyhaveanaffectontheinput histogramoroutputvalues.Theseparametersinclude: Islogarithmic:Checkingthiswouldremovealogtrend(commonforpermeabilitydata)fromthedata.This removalisdonebeforetheupscaledcellvaluesareusedtobuildahistogram. Outputdatarange:Thecellvaluesofthefinalmodelwillbecheckedagainstthesenumbersandclippedto themiftheyareviolated.YouusuallywanttoclickontheEstimatebuttontomakethesematchtheup scaledcellvalues. Distributionmethod:Standardmeansyouareusingthispropertysdataandhistogramtocontrolits distribution.Bivariatemeansyouareusinganotherpropertyscellvaluestoguidethispropertys distribution. Distribution:Fromupscaledlogsmeansyouwanttousethehistogrambuiltfromtheupscaledcellsas inputtoSGSmodeling.

[Note:Evenifdataaresufficienttodefineaninputhistogramfromupscaledcells,youmaywishtouseData AnalysisratherthantheDistributiontabtoinputthehistogram.Youwoulddothiswhenlogform,skewness,XYZ trends,etc.needtoberemovedfromthedataandthatfunctionalityisnotavailableintheDistributiontab.]

Petrel is a mark of Schlumberger

4801 Woodway Drive, Suite 150W Houston, TX 77056 www.scminc.com info@scminc.com


2011 SCM E&P Solutions, Inc.

12

Figure:DistributionparameterssettousetheupscaledcellvaluestobuildthehistograminputforSGS.

Figure:ZoneNess1Porosityhistogramjustfortheupscaledcells(left)andforallcells(right).Notehowthe propertydistributionfortheentiremodelmatchestheoriginalinputdatasdistribution.

HistogramfromDataAnalysis
Ifyouknowyoudonothaveenoughwelldatatodefinehistogramsthatcorrectlyrepresentthepropertys distributionthenyoushouldusetheDataAnalysisprocesstoeditthehistogramtomatchyourinterpretationofthe distributionandusethatforSGSmodeling.Dataanalysisallowstrendstoberemovedfromthedataandproperty distributiontrendstobeputintothedata(viathehistogram).AlthoughDataAnalysishasmanyparts,this discussionfocusesonlyoneditingtheNormalScorehistogram. ToedittheNormalScoreinputhistogram: 1. GototheDataAnalysisprocessandselecttheupscaledpropertyyouaremodeling,theassociatedfacies model(ifused),thezone,andthefacies(ifused). 2. GototheTransformationstabandhighlightNormalscoreintherightselectionarea. 3. ClickontheDefinecurveradiobuttonandyouwillseetheNormalScorecurvedefinedasthreepoints.

Petrel is a mark of Schlumberger

4801 Woodway Drive, Suite 150W Houston, TX 77056 www.scminc.com info@scminc.com


2011 SCM E&P Solutions, Inc.

13

4. ClickontheFitthedistributioncurvetothetransformedpropertyvalues thecurvetoadistributionfunctionthatfitsthehistogram. 5. Usethecursortoeditthecurvebymoving,adding,anddeletingpoints. 6. ClickApplytosavethechangesandmoveontothenextfaciesorzone.

icon.Thisessentiallyconverts

Notethatyoucanuseoriginallogvaluesasthebackgroundhistogramratherthanupscaledcellsbyclickingonthe Uselogs icon.Thismaygiveyoumoreinsightintothedistribution.Sometimesafaciesorzonewillhavenolog data.Gotoazonethathasasimilardistribution,copythattransformation,andpasteitintotheonethatismissing data( ).Youcanthenusethetransformasitisoralterittofityourinterpretation. OnceallhistogramsareeditedyouarereadyforSGSmodeling.IntheSGSdialog,clickontheUsethe transformationsmadeindataanalysis iconjustabovetheVariogramtabtoforceSGStousetheedited histogramsfromDataAnalysis.ThisalsocausesallparametersexceptSeedtobeclearedofftheDistributiontab. Figure:DataAnalysisdialogforonezoneshowingtheNormalScoreasitfirstappears(left)andaftercheckingthe Definecurveradiobuttion(right).

Petrel is a mark of Schlumberger

4801 Woodway Drive, Suite 150W Houston, TX 77056 www.scminc.com info@scminc.com


2011 SCM E&P Solutions, Inc.

14

Figure:DataAnalysisdialogforonezoneshowingtheNormalscorecurvefittothedatashistogram(left)andafter editing(right).

Figure:ZoneNess1Porosityhistogramjustfortheupscaledcells(left)andforallmodeledcells(right).Notehow thepropertydistributionfortheentiremodelmatchestheeditedDataAnalysiscurveinthefigureabove.

Petrel is a mark of Schlumberger

4801 Woodway Drive, Suite 150W Houston, TX 77056 www.scminc.com info@scminc.com


2011 SCM E&P Solutions, Inc.

15

HistogramfromNormalDistribution
Ifyouhavelittleornologdataforaparticularpropertyandwantahistogrambasedonstatisticsderivedfrom neighboringfieldsortheliterature,thenyoucanentertheMeanandStandarddeviationyouwantthepropertyto haveforSGSmodeling.TospecifythisstyleofdistributionchecktheradiobuttoninfrontofNormaldistributionand entertheMeanandStd.IfthereissomedatayoucanpresstheEstimatebuttontoinsertthemeanandstandard deviationofthatdata. Figure:TheDistributiontaboftheSGSdialogwiththemeanandstandarddeviationspecified.

Petrel is a mark of Schlumberger

4801 Woodway Drive, Suite 150W Houston, TX 77056 www.scminc.com info@scminc.com


2011 SCM E&P Solutions, Inc.

16

Figure:ZoneNess1Porosityhistogramjustfortheupscaledcells(left)andforallmodeledcells(right).Notehow themodeledhistogrammatchesthedefinednormaldistribution.

HistogramfromDistributionFunction(InputTab)
Ifyouhavetoolittlelogdatainthemodelareabuthaveotherwellsoutsidethemodelareathat,ifcombinedwould createhistogramsrepresentativeofthepropertysdistribution,thenyoucancreateahistogramusingallwelldata, extractthedistributionfunction,andusethatfunctionforSGSmodeling.Wehavealreadydiscussedhowtosubset alog,buildahistogramofthatdata,andextractadistributionfunctionfromthehistogram.Giventhatyouhavea distributionfunctionreadytosubmittotheSGSalgorithm,allyouhavetodoischecktheradiobuttoninfrontof Generaldistributionandusethebluearrowtoinsertthenameofthedistribution.NotethattheLogandOutput datarangeparametersarestillavailableforadjustmentusingthisapproach.

Petrel is a mark of Schlumberger

4801 Woodway Drive, Suite 150W Houston, TX 77056 www.scminc.com info@scminc.com


2011 SCM E&P Solutions, Inc.

17

Figure:TheDistributiontaboftheSGSdialogwithaDistributionfunctionspecified.

Figure:ZoneNess1Porosityhistogramjustfortheupscaledcells(left),thedistributionfunctioninputtoSGS (center),andthehistogramforallmodeledcells(right).Notehowthemodeledhistogrammatchesthedistribution function.

Petrel is a mark of Schlumberger

4801 Woodway Drive, Suite 150W Houston, TX 77056 www.scminc.com info@scminc.com


2011 SCM E&P Solutions, Inc.

18

Potrebbero piacerti anche