Hy dr ocar bonat e r eser v oi r char act er i zat i on
const r ai ned t o 3D sei smi c at t r i but es
Aut hor s: Ol ivi er GUILLOU (Pet roleum engi neer , UNICAM P), Dr . Deni s Jose SCHIOZER (Pet r ol eum depar t ment , UNICAM P), Dr . Pi err e BIVER (Geomodell ing and Uncer t ai nt i es, Tot al SA) I nt r oduct i on The gr ow i ng di f f i cul t i es encount er ed i n pet r ol eum expl or at i on and pr oduct i on, such as decl i ni ng di scover i es, i ncr easi ng coast al di st ances and f i el d dept h, cr eat e a const ant need f or i nnovat i on. To i mpr ove t he knowl edge and domi nat e reser voi r s l ocat ed i n r emot e ar eas, new t ool s and met hodol ogi es must be devel oped. Wi t h t he st eady i ncr ease i n comput i ng pow er and t he bi r t h of new al gor i t hms, t hi s demand can be sat i sf i ed and pr oj ect r i sks can be r educed. Fr om new 3D sei smi c i nt egr at i on met hodol ogies devel oped and i nt egr at ed i nt o an opt i mi zat i on pr ocess based on f or w ar d model i ng, di f f er ent vali d r esul t s have been obt ai ned. In t hi s cont ext , sei smi c const r ai nt char act er i zat i on has show n an ef f ect i ve w ay t o i mpr ove t he sei smi c quali t y and t he r el evance of gener at ed model s (Bar ens et al , 2004). Thi s paper pr oposes t o compar e t he r esul t s of t hr ee r eser voi r char act er i zat i on met hodol ogi es on a f i el d devel opment r i sk anal ysi s. The case st udy i s r eal ized on a deep of f shor e West Af r i can t ur bi di t es w i t h a r el evant expl or at i on w ell s number and a good 3D sei smi c sur vey quali t y. Reser voi r model s gener at i on St ruct ura l, Archit ect ura l, Fa cies, Pet rophysica l modeling f or t hree met hods For al l st udy met hodol ogi es, t he model char act er i zat i on i s done t hr ough a cl assi cal pr ocedur e as show n i n Fi gur e 1. Fi r st , sei smi c at t r i but es ar e i nt egr at ed qual i t at ively i n t he def i ni t i on of st r uct ur al hor i zons as w el l as i n t he ar chi t ect ur al i nt er pr et at i on of t he r eser voi r . Then, r ock t ype and pet r ophysi cal var i abl es ar e di st r i but ed i nsi de t he w hole vol ume w i t h or wi t hout i nt egr at i on of sei smi c at t r i but es bef or e t he posi t i oni ng of t he w at er -oi l cont act and t r ansi t i on zone def i ni t i on (Ler at et al 2007). Complet e int egra t ion of 3D seismic a t t ribut es f or 2 met hodologies, pa rt ia l int egra t ion f or 1. Tw o met hodol ogi es benef i t of a compl et e i nt egr at i on of 3D sei smi c at t r i but es, acoust i c i mpedance (IP) and Poisson r at i on (PR). Thi s i nt egr at i on i s r eal i zed quant i t at i vel y by condi t i oni ng rock t ype and pet r ophysi cal model s t o r el evant Figure 1: Basic reservoir characterization methodology. sei smi c at t r i but es as pr esent ed by Bi ver et al (2009) usi ng r espect i vel y cr oss pl ot smoot hi ng met hods and geost at i st i cal cosi mul at i on. Fi r st , t he r ock t ypes need t o be def i ned accor di ng t o t hei r r eservoi r char act er i st i cs and t hei r sei smi c r esponses. The r ock t ype model condi t i oni ng i s, t hen, r eal i zed by def i ni t i on of a r ock t ype pr opor t i on cube used i n a Tr uncat ed Gaussi an (TG) algor i t hm t hat gener at es, f i nall y, a r ock t ype cube i n t he w hol e r eservoi r . The t hr ee char act er i zat i on met hodol ogi es compar ed i n t hi s ar t i cl e have di f f er ent sei smi c const r ai nt degr ees (Fi gur e 2, l ef t ). Fr om a det er mi ni st i c met hodol ogy based on sedi ment ar y concept s, sei smi c at t r i but es ar e gr adual l y i nt egr at ed in t he r ock t ype pr opor t i on cube and geol ogi st i nt er pr et at i on becomes l ess i nf l uent i al . Secondl y, t he pet r ophysi cal model i s def i ned per r ock t ype. Tw o model i ng met hods ar e compar ed: a non sei smi c condi t i oni ng met hod and a condi t i oni ng one (Fi gur e 2, r i ght ). The non condi t i oni ng met hod i s a cl assi cal Sequent i al Gaussi an Si mul at i on (SGS). St at i st i cal di st r i but i ons as w ell as a var i ogr am l aw ar e i mposed t o t he al gori t hm t hat gener at es t he pet r ophysi cal vari able. The respect of w el l obser vat i ons i s guar ant ied by a kr i gi ng met hod i nt egr at ed i n t he al gori t hm. The condi t i oni ng met hod i s a Sequent i al Gaussi an Cosi mul at i on (SGC). As w el l as t he non condi t i oni ng one, i t r espect s var i able di st r i but i ons, var i ogr am l aw and w ell obser vat i ons. It al so r espect s t he obser ved cor r el at i on coef f i ci ent bet w een t he simul at ed var i abl e and a secondar y one and makes t he use of a r el evant sei smi c at t r i but e t o const r ai nt t he pet r ophysi cal model possi bl e.
Full uncert a int y st udy on model (mult iple rea liza t ions) Rock t ypes and pet r ophysi cal model def i ni t i on i s i nt egr at ed i n an uncer t ai nt y quant i f i cat i on st udy. The l at t er eval uat es how geol ogi cal uncer t ai nt i es impact i ni t i al hydr ocar bon vol umes as w el l as dynami c f l ow behavi or s. The model i ng pr ocess i s real i zed sever al t imes t hr ough a spat i al M ont e Car l o appr oach. Uncer t ai n var i ables ar e: Reser voi r ar chi t ect ur e (sedi ment ar y envi r onment i nt erpr et at i on) Rock t ype pr opor t i ons Figure 2: Different geological and petrophysical modeling, from pure geological concepts to strong seismic constraint. Pet r ophysi cal var i abl es di st r i but i on Wat er / oi l Cont act posi t i on PVT model Eva lua t ion of t he volume dist ribut ion a ccor ding t o t he init ia l uncert a int ies The pur el y geol ogi c case is def i ned accor di ng t o a val i dat ed met hodol ogy, hence, i t i s consi der ed as a r ef er ence i n t hi s compar at i ve st udy. The i ni t i al vol ume di st r i but i on i s per f ect l y cent er ed on t he base case model . The ot her compar i son cr i t er i a ar e consi der ed as r ef er ences as well . A per f ect cent er i ng of t he vol ume di st r i but i on can be obser ved on bot h t he w eakl y and st r ongl y const r ai ned cases as w el l. It can be obser ved t hat t he uncer t ai nt y on i ni t i al vol ume (di st ribut i on s st andar d devi at i on) i ncr eases w i t h t he degr ee of sei smi c i nt egr at i on i n t he model (Fi gur e 3). Tabl e 1 pr ompt s t he obser ved r el at ive var i at i on bet w een a const r ai nt case and t he pur el y geol ogi c case consi der ed as a r ef er ence. Rel at i ve var i at i on i s comput ed f or t he aver age and st andar d devi at i on of each vol ume di st r i but i on. Table 1: Relative differences between volume distributions according to characterization methodology Q50 St d. dev Pur el y geol ogi c - - Weakl y const r ai ned 0% +0,6% St r ongl y const r ai ned + 5,5% +33,1%
Comput at i on of synt het i c sei smi c r esponse for each model gener at ed Genera t ion of synt het ic seismic a t t ribut e The comput at i on of t he sei smi c r esponse i s i nt r oduced i n t he mul t i pl e reali zat i on pr ocess. It gi ves t he synt het i cs sei smi c at t r i but es (IP, PR) cor r espondi ng t o each pet r ophysi cal model gener at ed. To do t hi s, a pet r oel ast i c model (PEM ) i s i nt egr at ed i nt o t he model i ng pr ocess. In or der t o eval uat e t he synt het i c sei smi c qual i t y, an aut omat i c compar i son must be devel oped. It i s pr oposed t o si mpl i f y t he measur e of t he sei smi c r esponse qual i t y f r om t hr ee t o onl y one di mensi on, gi vi ng a scor e t o each gener at ed model . Thi s scor e not onl y al l ow s a r el at i ve compar i son of t he gener at ed model s of all cases but al so t he abi l i t y of each char act er i zat i on met hodol ogy t o r espect t he sei smi c i nver si on. Thi s scor e i s t hen used t o sel ect cer t ai n r epr esent at i ve geol ogi cal models. Comput a t ion of a seismic score ba sed on simila rit y w it h a ct ua l a t t ribut es f or ea ch model Figure 3: Oil volume distribution for three characterization methodologies a) Pur el y geol ogi c b) Weakl y const raint c) St r ongl y const r aint The sei smi c scor e t o be def i ned must r espect t w o condi t i ons: f i r st , i t has t o consi der onl y t he qual i t at i ve cont ent of t he dat a w hi l e r espect i ng exi st i ng cont r ast s, secondl y, i t must be si mpl e enough t o be comput ed at each reali zat i on and i nt er pr et ed r api dly. A scor i ng al gor i t hm i s devel oped i n or der t o cl assi f y model s accor di ng t o t he simil ar i t y of t hei r synt het i c r esponse wi t h t he act ual sei smi c. To do so, t he at t r i but e cumul at ed densi t y f unct i on i s used (Fi gur e 4). The cumul at ed densi t y f unct i on i s used t o keep onl y t he qual i t at i ve cont ent of t he dat a. Indeed, i f synt het i c and act ual at t r i but es may have di f f er ent var i at i on t hr eshol ds, t hei r cor r espondi ng cumul at ed densi t y var i es syst emat i cal ly bet w een 0 and 1. Know i ng t he cumul at ed densi t y of synt het i c an act ual at t r i but es f or each cel l , t hei r di f f er ence can be comput ed. Fi nal l y, t he quali t y of a r eal i zat i on i s obt ai ned aver agi ng t he cumul at ed densi t y di f f er ence on al l cel l s. Scorc = 1 n | cJ( u) cJ( u i ) | n =1
u: act ual sei smi c at t r i but e u: synt het i c sei smi c at t r i but e Cla ssif ica t ion of ea ch met hod a ccording t o it s seismic response Af t er mul t i ple r eal i zat i ons, t he sei smi c scor es ar e cl assi f i ed i n ascendi ng or der t o be anal yzed. The char act er i zat i on met hodol ogi es ar e w el l di f f er ent i at ed w hi ch al l ows f or easi l y compar ison among t hem. The scor i ng al gor i t hm cl assi f i ed t he best model s w i t h t he l ow est scor e. Consi der i ng t hi s, t he t hr ee met hodol ogi es can be cl assi f i ed: St r ongl y const r ai ned met hodol ogy Weakl y const r ai ned met hodol ogy Pur el y geol ogi c met hodol ogy Bot h t he met hodol ogi es combi ni ng sei smi c dat a and geol ogy obt ai n t he best scor es. Thi s l eads t o t he concl usi on t hat dat a combi nat i on has a si gni f i cant i nf l uence on r eser voi r char act er i zat i on. Bet t er i ng t he sei smi c r esponse qual i t y means t hat accur acy on oi l vol ume est imat i on i ncr eases as w ell . Fi nall y, t he scor es obt ai ned i n t he pur el y geol ogi c case conf i r m, once mor e, t he i mpor t ance of i nt egr at i ng sei smi c dat a i n model char act er i zat i on (Fi gur e 5).
Figure 4: Seismic score computation is based on comparing actual and synthetic seismic attribute cumulative density function
Repr esent at i ve model sel ect i on for dynami c si mul at i on Cla ssica l select ion: Sa mpling of init ia l volume dist ribut ion (represent a t ive uncert a int ies) Tw o modes of r epr esent at i ve model select i on ar e st udi ed. The f i r st i s a r egul ar sampli ng of t he i ni t i al oi l vol ume di st r i but i on. 21 r epr esent at i ve model s are ext r act ed (one ever y f i f t h per cent il e). The dynami c si mul at i on r esul t anal ysi s of t hose model s al l ow s eval uat i ng t he i mpact of t he char act er i zat i on met hods on pr evi si on f or ecast uncer t ai nt i es w i t hout consi der i ng t hei r r espect i ve sei smi c r esponses. Sma rt model select ion: Best seismic response select ion (represent a t ive uncert a int ies + seismic qua lit y crit eria ) The second mode of sel ect i ng r epr esent at i ve model s i s based on sei smi c scor e. In t hi s case, t he 21 best sei smi c scor es ar e sel ect ed t o r epr esent t he i ni t ial oi l vol ume uncer t ai nt y. In t hi s second case, t he model s l ess compat i bl e w i t h sei smi c ar e not consi der ed. Thi s sel ect i on mode shows how t he select ed models ar e l ocali zed i n t he gl obal i ni t i al oi l vol ume di st r i but i on. A r egul ar r epar t i t i on of t he sel ect ed model s means t hat t he sei smi c r esponse does not al l ow a r educt i on of uncer t ai nt y on i ni t i al oi l vol ume. On t he cont r ar y, a concent r at i on of t he sel ect ed model s i n t he same regi on of t he i ni t i al oi l vol ume di st r i but i on t est i f i es t hat t he select ed i ni t i al oi l vol ume has a bet t er sei smi c r esponse and hence, a bet t er abi l i t y t o r epr esent uncer t ai nt i es. For al l char act er i zat i on met hodol ogi es t he f i r st sel ect i on mode does not af f ect t he di st r i but i on par amet er s (aver age, di sper si on). The second sel ect i on mode i s t hen analyzed f or each char act er i zat i on met hodol ogy. In bot h, t he pur el y geol ogi cal and t he w eakl y const r ai ned cases, a sli ght decr ease can be obser ved on t he sel ect ed model s di st r i but i on mean af t er sei smi c sel ect i on. In t he st r ongl y const r ai ned case, t hi s mean i s not af f ect ed. On t he ot her hand, t he di st r i but i on di sper si on i s not si gni f i cant l y af f ect ed i n t he pur el y geol ogi cal case w hen i n t he ot her t w o i t i s r educed by al most 50%. Thi s observat i on show s t he abi l i t y of t he pr oposed sei smi c const r ai nt t o r educe t he i ni t i al vol ume uncer t ai nt y and, as a consequence, t o i ncr ease t he accur acy of t he r epr esent at i ve model s (Fi gure 6). Figure 5: Seismic score distribution. Best models have lower scores and are on left side of the figure. Table 2: Relative differences on mean and standard deviation between original volume distribution and seismic selected models Am An Purely Geologic - 5,4 % 0 % Weakly Constrained - 2,5 % - 46,8 % Strongly Constrained 0 % - 47,9 %
Dynami c si mul at i on The f l ow si mul at i on i nt ends t o show and compar e t he pr oduct i on f or ecast s f or t he t hr ee st udi es met hodol ogi es. In t hi s st udy, onl y geol ogi cal uncer t aint i es ar e consi der ed. Uncer t ai n par amet er s ar e ext r act ed f or dynami c si mul at i on accor di ng t o t he t w o sel ect i on modes f or mer l y descr i bed and i nt egr at ed i n a dynami c mul t i pl e si mul at i on. Those geol ogi cal par amet er s ar e t he ones t hat best i nf l uence i ni t i al oil vol ume and pr oduct i on f or ecast s. Ot her par amet er s such as f l ui d and pr oduct i on st r at egy ar e det er mi ni st i c accor di ng t o a best guess case. Net t o gr oss r at i o ( NI0) Net sand por osi t y ( P0R0) Net w at er sat ur at i on ( Sw I ) Net per meabil i t y ( PERHX, PERHZ) Result s a na lysis f or ea ch met hodology Figure 6: Selection criterion has an influence on representative volume distribution. Seismic criterion allows for the reduction of dispersion by selecting only models with best A gl obal obser vat i on i s done; f or all t he met hodol ogi es, sei smi c based sel ect i on mode (Fi gure 7, t op l i ne) has a l ow er di sper si on t han t he cl assi c sel ect i on mode based on di st r i but i on sampl i ng (Figur e 7, bot t om l i ne). Thi s obser vat i on i s par t i cul ar l y appar ent f or t he t w o combi ned char act er i zat i ons. For t he sei smi c based sel ect i on mode, t he second obser vat i on t hat i s done i s t hat i n w hi ch bot h t he pur el y geol ogi cal and t he w eakl y const r ai ned cases have simil ar di sper si ons and eval uat e i n t he same pr oduct i on f or ecast s envel ope w hen t he st r ongl y const r ai ned case has a hi gher pr oduct i on uncer t ai nt y w i t h l ow er pr oduct i on f or ecast s. A hi gher het er ogenei t y i n t hose model s is supposed t o be mai nl y r esponsi bl e f or t hi s behavi or . Fi nal ly, t he dynami c r esul t s don t demonst r at e t he super i or i t y of any met hodol ogy nei t her combi ni ng sei smi c dat a nor usi ng excl usi vel y geol ogi cal concept s t o r educe uncer t ai nt i es. Also, condi t i oni ng t o sei smi c i nt r oduces mor e het er ogenei t y i n t he models. The l at t er coul d be bet t er anal yzed i n t he pr esence of pr oduct i on dat a and hi st or y mat chi ng (Fi gur e 7).
Figure 7: Seismic criterion has an influence on models dynamic behaviors, reducing dispersion on production forecasts. Strongly constrained models give lower production forecasts; production history is necessary to know which characterization methodology is most accurate. Ri sk Anal ysi s The dynami c r esul t s ar e used i n or der t o assess t he pr oj ect r i sk. As i t has been demonst r at ed f or mer l y, t he t hr ee char act er i zat i on met hodol ogies gi ve si mi l ar st at i c vol ume eval uat i on but have di f f er ent quali t y i n t er ms of sei smi c r esponse and r espect t o sedi ment ar y concept . For t hi s r eason, t he r i sk anal ysi s must be done w hi l e car ef ull y consi der ing t hose di f f er ent obser vat i ons (Fi gur e 8). Risk of t he project considering ea ch met hodology sepa ra t ely Fi r st , i ndi vi dual r i sks ar e eval uat ed. For t he t hr ee met hodol ogi es, di f f er ent pr oduct i on f or ecast behavi or s ar e obser ved. Consi der i ng each one i ndi vi dual l y, t he pr oj ect r i sk can be under est i mat ed by not consi der i ng t he r esul t s of t he ot her t w o. On Figur e 8, lef t col umns, di f f er ent met hodol ogi es pr edi ct di f f er ent r i sks. Consi deri ng onl y one met hodol ogy as t he val i d one cannot be done as each one gi ves di f f erent r i sk est i mat i on. Risk of t he project considering a ll met hodologies t oget her The val i di t y of t he t hr ee met hods i mposes t o consi der t hem t oget her . It i s pr oposed t o combi ne all r esul t s i n t he same r i sk anal ysi s. In t hi s case, t he r esul t i s a maxi mum r i sk eval uat i on t hat consi der s an equal w ei ght f or each met hodol ogy (Fi gur e 8, t op r i ght ). But act ual l y, t he pr esent ed met hodol ogi es have di f f er ent qual i t ies and shoul d not be consi der ed w i t h equal wei ght . The quest i on ar i ses: How t o w ei ght t he di f f er ent met hodol ogi es accor di ng t o t hei r r espect i ve quali t i es? The quest i on i s not answ er ed i n t hi s paper , but i t seems t hat t he met hodol ogy weight shoul d quant i f y bot h t he r espect of sedi ment ar y concept s and t he sei smi c r esponse qual i t y. As show n on Fi gur e 8, bot t om r ight , t hi s w eight i ng met hod per mi t s t o under l i ne t he r i sk envel ope of t he pr oj ect as a r esul t of t he super posi t i on of al l w ei ght combi nat i ons (each dashed l i nes r epr esent a w ei ght combi nat i on). It i s t hen possi bl e t o obt ai n t he mi ni mum and maxi mum ri sk cur ves, consi der i ng t he li mi t s of t hi s envel ope. Consi der i ng t he exi st ence of an i deal w eight f or each char act er i zat i on met hodol ogy, t he r eal ri sk cur ve of t hi s pr oj ect exi st s and i s i ncl uded i n t he ri sk envel ope.
Concl usi on Di f f er ent met hodol ogi es can be used t o char act er i ze hydr ocar bon r eser voi r s accor di ng t o avai l abl e dat a. Resul t qual i t y is st r ongl y dependent on t he i nt egr at ed dat a sour ce. It has been show n t hat w i t h i nt egr at i on of sei smi c at t r i but es, t he sei smi c r esponse qual i t y i s i mpr oved. In t he case of st r ong geol ogi cal concept i nt egr at i on, t he same obser vat i on can be done on geol ogi cal model accur acy. The combi nat i on met hodol ogi es pr esent ed i n t hi s paper ar e good al t er nat i ves t o make a bet t er use of i nf or mat i on i n t he char act er i zat i on pr ocess, t he conser vat i on of geol ogi cal concept s and a st r ongl y i mpr oved sei smi c r esponse. It al so al l ow s a smar t repr esent at i ve model sel ect i on based on sei smi c qual i t y eval uat i on. The select i on cr i t er i a r educed st at i c and dynami c di sper si ons on f or ecast s. Unf or t unat el y, t he l ack of pr oduct i on dat a does not per mi t t he select i on of mor e accur at e char act er i zat i on met hodol ogy. The r i sk analysi s present ed her e w as based on t he t hr ee met hodol ogi es and a ri sk envel ope has been def i ned. In or der t o bet t er l ocal i ze t he case st udy i n t hi s envel ope, a w eight must be def i ned f or each met hodol ogy. It i s pr oposed t o est abl i sh t hi s w eight on bot h, t he r espect of sedi ment ar y concept s and t he sei smi c r esponse qual i t y of each met hodol ogy. Nomencl at ur e CDF: Cumul at ed densi t y f unct i on IP: Acoust i c i mpedance NTG: Net t o gr oss r at i o Figure 8: Risk analysis (cdf of the total production). FOPT FOPT FOPT FOPT FOPT PDF: Pr obabi l i t y densi t y f unct i on PEM : Pet r oel ast i c model PERM X: Per meabil i t y i n X di r ect i on PERM Z: Per meabil i t y i n Z di r ect i on PORO:Por osi t y PR: Poi sson r at i o Q50: SGC: Sequent i al Gaussi an cosi mul at i on SGS: Sequent i al Gaussi an si mul at i on SWL: Li mi t w at er sat ur at i on TGS: Tr uncat ed Gaussi an si mul at i on Am: Aver age vari at i on bet w een or i gi nal vol ume di st r i but ion and sel ect ed model s An: St andar d devi at i on var i at i on bet w een or i gi nal vol ume di st r i but i on and sel ect ed model s Bi bl i ogr aphy Bar ens, L. (2004). Reser voi r f aci es pr edi ct i on f r om geost at i st i cal i nver t ed sei smi c dat a. Abu Dhabi Int er nat i onal Pet r ol eum Exi bi t ionand Conf er ence. Abu Dhabi : SPE. Beucher t -Dar r i cau, H. (2006). M odeli ng compl ex r eser voi r s wi t h mul t i pl e condi t i onal t echni ques: A pr act i cal appr oach t o r eser voi r Char act er i zat i on. In T. C. Cobur n, St ochast i c model i ng and geost at ist i cs; Pr incipl es, met hods and case st udi es, volume II (pp. 289-299). Bi ver , P. (2008). Li t ho-t ype model i ng usi ng sof t pr obabi l i t ies f r om sei smi c at t r i but es and ot her sour ces of i nf or mat i on. Int er nat i onal geost at ist i cs congr ess . Bi ver , P. (2002). Uncer t ai nt i es i n f aci es pr opor t i on est i mat i on i i. appl i cat i on t o geost at i st i cal si mul at i on of f aci es and assesment of vol umet r i c uncer t ai nt i es. M at h. Geol. , pp. 701-712. Hass, A. (2002). Uncer t ai nt i es i n f aci es pr opor t i on est i mat i on i . t heor i cal f r amew or k: t he di r i chl et di st r i but i on. M at h. Geol . Ler at , O. (2007). Const r uci t on of a geol ogi cal model const r ai ned by hi gh r esol ut i on 3D sei smi c dat a - Appl i cat i on t o t he gi r assol f i el d, of f shor e angol a. Annual Techni cal Conf er ence and Exhibi t ion. Anahei m: SPE.
SRS documentation of Virtual Classroom System , SRS documentation of Personal Identity Management ,SRS documentation of EMentoring for women system , SRS Documentation of Employee Performance Management SRS Documentation of Online Ticketing