Professional Documents
Culture Documents
TM, p.p.r w prepared for presen!.!iw at the. S.PEmOE NIIIh symposium:. !mpr:ved 01 Rem..V h.ld r. TUIS.. Okl,h.m~ U.S.A.. 17-= APrjl 19W
This wet was selected @r pmsentalm by an WE Promm C.mmnme following ravlew of [nfon’mfl.n contained 1.. w AL4mct s.bmmed bY m. ~th.r(,). @.t.*s .f @ P.Per.
m presented, have ..[ bean revl?vied by me society of Paro!e.um tis[neers and are s.bled t. ..r:.ab. W fh. .ufhti,). Th. mqflala F Wnl~, w, M ~-rY ~.fr..!
MY FOWM.1 ow s.ci.w..f p.~.!..m W.,.-, It, Om_@-~. .r m.mb.m. W.,. pr-..d.d at sQE m-tl.s, ~ ~bi=t ~ PUb~C.tIO. r.v,ew by ~tor[.l Cmmit$eos 01 ‘e ‘ocia~
01 Pewleum Eng!.eaffi, ?erml~ton to COPY!! restrb!ed !. m abaraaof ..1 more IhaP ?~w.rde. lllus~!?ns may .oth. mpl~d. ~. *mm?h..ld ‘w~.in COWPICUOUS..m~ledEm.nt
d tier. and by whom the pa!aer 1s p!ese.ted. Wile tib.~~-a., SpG P~. B?. 83sw6, ~.hard,:.., F 75~33a33. U.S.A., TM~ 16=45 Sp~T
291
.
“Super MM’ computer in a few tilnutek per project models. Therefore, the study concentrated on comparing
(reservoir). the-oil production reported to DOE for TIP projtits to the
production predicted by the predictive models using the
The 1984 NPC study (1982-1984)2 included may industxy reservoir aad project psmmetera that were reported to DOE.
experts organized in EOR process specific committei% The accuracy of all of the EOR process models was
These committees tested and dined the computer predklive considered: Steam (SFPM), Miscible Gas (CQ2PM),
models (steam drive, micellar-polymer, and COZ miscible) Polymer (PFPM), In-s@ Combustion (ICPM), and
and the rerervoir data they were provided to estimate the Chemical (CFPM) (for both surfactant and alkaline
domestic E.OR potential. Additional predctive models for processes). TIP projects that had 3 or more years of
polymer, alkaline, and in-situ combustion were developed production information and reservoir parameters that were
by the NPC committees: Project development strategies, reasonable and internally consistent were used in the stady.
cost functions, expected recoveries, and the logic of the For those processes with sufficient number of projects and
models were developed, thoroughly reviewed, and tested by reasonable precision, statistics for accuracy were esdmated
leachng EOR .indmtry experts. The results oft@ “effort at a reasonable confidence level. The predictive models
were a suite of EOR. predictive models and a larger were modified to use the average wellhead oil prices for the
integrated system using reservoir dats and preprocessor period 1980 to 1991. The preprocessor programs of the
programs with tbcse models to analyze the domestic oil TORIS analysis system and defaults built into the DOE
resource. The lager system has become the DOE TORIS predictive models were used for missing data.
~nalY~i~ ~Y~tern,3 The” “iiidividual simplified EOR
predictive models were subsequently documented and LIMITATIONS OF THIS STUDY
published by DOE in 19T6.4-8 Although thk study had a significant amount of data on
projects, the quality of the data was a concern. Efforts were
Model refinements and testing have continued Assessing made to screen obvious errors and the information was
the relistility and accurncy of individual predictive models accepted unless inconsistencies could not be resolved using
by compa@g the predicted qesglts to..tield project rcsuhs engineering judgment and other information sources.
and”finite difference” shiudation has shown Mat the. models Some of the project areas planned at the beginning of the
are less th~ prefect but make fair to good estimates of oil TIP progr.m were not fully developed, and conversely,
~r0du~tion,9 Unfortunately, results of prior studk have some developed areas were greater than that planned or
been qualitative and have not estimated a cotildence limit ind]cated by the annual reports. Unfortunately, without
for the models because of limited information on EOR @ect correspondence with a knowledgeable representative
projects, ~ of the operator, these errors might not have been detected
and therefore might be included in this study. Additional
Another DOE program that was initiated about the same work is planned to identify errors and expand the project
time the TORfS analysis system” was conceived was the data
Terdmy Incentive Program (TIT) of 1978. Thk progmm
promoted the development of EOR projects and piIots Although significant amounts of data were available, they
through certification ‘ti]th the .DOE which. allowed were not sufilcient to fully assess the accuracy of the
operators to receive world oil prices for their oil. Projects predictive models except for the SFPM. The data is also
were certified based on OMB No, 38-ROW initial report weighted toward the early production years of the project.
forim submitted to the DOE. Each year after 1980, TIP The statistical remdta iadicate the combhd accurncy of the
operators have been requested to return a completed TIP data arrd the predictive models. Refinements of the
“Annual Report for Enhanced Oil Recovery Incentive project data will likely result in lager improvements in the
Program; form FE-748. This annual update mquesta the measured accuracy than a refinement of the models.
operator to tabulate monthly fluid injection and production
rates and indicate changes in key reservoir parameters used To generate the data input sets for the predictive models,
to describe a project. The information from these snnunl the pre-processcr portions of the TOR3S Analysia System
reports is maintained in the Enhanced !X1Recovery Project were used. These pro~ams assisted in validating the
Datnbase at the DOE Bartlesville Project Office (BPO).10 internal consistency of the data, assigned missing data from
Since completing form FE-748 is volunt.a.~, numerous correlations, and prepared the input tiles to the predictive
operators stopped reporting after the initial certification. models. TIis simplified the data preparation, but it also
The number of completed reports continues to decrease included assumptions on the development of the project
because many of the original projects have been terminated, that may not have been consistent with historical
completed, changed, were never started, or are not development. Specific steps were taken to rectify this
reporting. .Approximately 50 projects have provided 10 possible problem in one case for each process with mixed
years of production data, and roughly another 100-have results.
provided more than 3 ycam of data.
The models themselves are simple and often have built-in
SCOPE OF THIS..STUDY
assumptions that do not allow key characteristics of “a
The objective “of this ‘study was to use tie data on TIP projec; to be simulated. For example, a 5-spot well pattern
projects to assess the accuracy of the DOE predictive is assumed for the SFPM which is rarely used in practice. ,
292
. .
The use of a single value for key reservoir par~eters is submitted monthly totsl: instead of an average mnnthly
another limitation add potential source of error-choosing rste.)
one permeab@ value for a large field project has obvious
problems. The model results were compared to gross oil production
reported by the TLP project operators. It was assumed that
DEVELOPMENT OF METHODS FOR all produced oil for the steam projects was the result of
COMPARISON steaming operations although s. few operators reported
Tbe basic approach of this comparative study was to apply incremental oil production lower than the gross. oil
increasing levels of engiij+.ring judgment to improve upon production. The chemical floodlng, in-situ tnmbustion,
the project descriptions-taken from the databsse and fie and polymer projects were compared to prcdcted gross oif
inputs defaulted by the preprocessor progcams. The fmt production. The miscible projects required developing an
priority wss to model the appropriate target oil volumes. iddkional factor to account for mobile oif to model gross
Second, inject~on ‘well counts were “rnodlfied to oil production. Tertiary production from the miscible
approximate actual Conditions. Third, the model was model was also anafyzed.
prowded with actual fluid injection rates ralher than
Selected .$teamj’food Projects
accepting default rates determined by the models
themselves. A general approach was developed with a The comparative study began with 14 steamflood TIP
subset of steam projects “itndthen applied to the full set of projects of which 13 were in California and one in
stenm projects and the other process=. Wyoming. These projects were selected becnuse they
appe~d to have the most complete project descriptions.
Most of the data descr~bing the TIP projects originated In addition, a small group of projects was initially chosen
from the OIvfB No.. 38-R0445 ifiti.al report forms. The because of the ease of handling in determining the
forms were submitted to the DO.E, and the information was methodology that would be applied to evaluate the
entered into the DOE project database. In many cases, the predictive models. The study began with thermal projects
yearly up&te forms, FE-748, coritained dlffeient andlor because steamffooding probably has the greatest short-term
contradictory information from that found in the initial futnre potential, and the model has been debugged and used
reports. During preparation of the-project description, the extensively. The 14 steamflood projects chosen are shown
FE-748 data were considered current and generally more in Table 1.
sccurate when the data input file was being prepared. The
data retrieved from the DOE EOR Project Database were SFPM Development Runs
checked during the review process for accuracy, and upon
dkcovery of emois, Corr?.ktiorisw“efern”adeto “tie input data The approach employed in his compsmtive study wss to
sets. used in this study. The corrections included provide the model with the best project descriptions
omissions, data that could not be substantiated, and possible. Fhe cases were constructed, each hopefully more
common errors that were either transcription or accurate than the previous, and each employing @eater
typographical in nature. The changes were made to basic engineering judgment. The subsequent model results were
reservoir properties amdtorepmted production and injection then compared to the reported production andlor injection
data. (A common and easily detected error in the data. Case I used an approach similar to that used by NPC
production and injection dTta was that the. operators (rcf.ning to the 1984 National Petroleum Council Study).
Table 1
Steam Projects Used To Develop Methodology
. ..
Project No. FieId Formation Operator
SF019 Midway Sunset “” – Potter ““uNow
SF1322 .Cnspei’Creek South Tenslecp UNOCAL
SF026 .Midway Sunset Tularc Sand UNOCAL
SF030 Ouadshqre Sisquoc UNOCAL
SF091 Cat Canyon Sisquoc Chevron
SFIOO Midway Sunset Monarch Chevron
SF107 Midway Sunset Webster Chevron
SF158 NRdway Sunset Upper Miocene Snnta Fe
SF186 Midway Sunset Reef Ridge Ssnta Fe
SF242 McKittrick Tubwe Sand Chevron
SF345 Midway Sunset POttcr Sun/ARCCl
SF407 M]dway Sunset 7A Sand Santa Fe
SF414 Peso Creek Etchegoin Elf Aquitairie
SF.422 ~ Kern River Kcm River Chevron
293
4 COMPARISONOF ACTUAL RESULTS OF EOR FIELD PROJECTSTO SPE 277(T?”
CALCULATEDRESULTSOF EOR PREDICTIVEMODELS
That is, changes in the input data were made only to and resolving them became one of the most important
achieve a reasonable original-oil-in-place (OOIP) match and is.aueain the project descriptions.
then rnn through the preprocessors. Normally, once the
preprocessor results were accepted, no changes were made C?S&I
to the data. Casea II -“V required deliberate and manual Case I represented the run based on project d+criptiom
intervention to “cha”ge the data either before or after the tnrgeted toward resolving O@P inconsistencies.. The
preprocessor programs to.systematically apply engineering aggregate calculated O“OIP volume was 2’/3 IyffvfFjlj
judgment to improve the project description. ” The compmed to 396 MMBbl reported by the project operators.
assumptions used in the five caies are described aa followx A. closer match could only be obtained if reservoir
parameters were ignored. The resulting aggregate ROIP
Cise I: Revise any Ybhunetric parameter, either a
calculated by SFPM was 243 MMBbl compared to a
reported value or a value supplied by the
pRPKXeSSOr programs, within reason .[normally reported 339 MMBbl. The calculated volumes were 69”aid
72% of the reported values for OOIP and ROIP,
*IO%) to generate 00~ volumes that agree witiln
respectively, Some of the projects did not report an OOIP
i3% of the reported OOIP values. If a reasonable
or ROIP, and the total calculatedheported (c/r) ratios reflect
change could not produce the desired effect, it was only those projects with complete data... Frve of the
assumed that the reported 00IP volume was an error
projects, after revisions wire made to tbeir”input data, still
and the calculated value based on volumetric was
had calculatedkeported O_OIPkranging..fmm 52 to 115%.
substituted. Remaining-oil-in-place (KOJ.P) or the
Table 2 summarizes the results for ezch project..
target oil was develgped by deductions from the
OOIP. The execution of thk group of projects by ..the SF!’M
Case II: Revise the calculated ROIP, if necessary, showed that their predicted oil recovery ranged from 27 to
by making whatever reasonable changes necessary to 492% of that reported. The aggregate predicted oil
the volumetric parameters to achieve agreement production of the 14 projects is 71% of the reported
within +370 of the reported values. Generally, production which is consistent “withthe under represented
porosity, net thickness, and oil saturation were ROIP.
changed. ROIP is the same @ue as the Initial.-.Oil- =.. ... . .
In-Place (IOIP) used in the .SWM—the volume of
oil-in-place at the start of the project. The target oil (IOIP or ROIP) volume was more important
to the SFPM than the OOIP, therefore, it waa critical to
Case IIT Revise the calculated ROIP, if necessai’y, obtain a good match between the reported ROLP and the
by ch~ging only. the drg@.a.geWea. to. achieve calculated IOW. In Case If, volumetric parameters were
agreement wit.tin i3c% of the reported values. Thk modified to obtain an ROIP rn”atchthat was M 7. of the
case assumed that the drainage area was the reported value. Thk assumed that the reported ROIP was
volumetric vnriable most likely to be inaccurate-all more accurate than .t!rerock and fluid property data. Whh
other volumetic pmameters were assumed to be reasonable changes to the input data, a total c/r ratio of
correct. 99.%for the ROIP of the projects wna obtained. The model
generated an aggregate oil production p/r ratio of 99% and
Case IV: Same as Case III except refine pattern the individual project ratios ranged from 23 to. 609%.
size, number of patteins,-imd pattern development Total water production and steam injection plr ratios for the
timing “to improve ‘the match of injection and
projects were 40 and 9690, respectively. The total ratios
production wells. for oil production and steam injection were excellent, but
Case V Same is Case fl except provide the model once again the individual projects exhibited a large
with actual steam injection rates rather than variance. The total water production plr ratio is not good
allowing the model to determine the rates. The fust but this parameter is typically one of the most difficult
yeaI average steam injection values_ were used .~ production values to report with any accuracy. The
wherever possible. predicted and repOfi!d Oil recOve~, and the Oil and water
production and steam injection p/r ratios for each project,
After several preliminary ruiis w“ith tbe project data, it are presented in Table 3. The projects were sorted by oil
became appnri?ntthat some of the volumetic parameter data production plr ratio. The shaded rows were those projects
and .00IWROP” vdumm reported for the projects were riot whose oil production plr ratios were tiOY. of a perfect
internally consistent. For most projects, there was match (p/r=l.O or IOU%).
insufficient inf6”rmati6n in the files to determine whether
the volumetric parameters, the hydrocarbon volumes, or Cases JII &IV
both, were incorrect. Besides estimates of” OOIP,
Case Itf matches the ROIY but assumes that all volumetric
comparing reported ROIP by reducing the OOIP by the
data were correct except the drainage area. There were two
cumulative production identified additional inconsistencies
exceptions, SF022 and SF026. There was sufficient
of tbe data in many casea.. These types of problems
information in the project files to modify the pattern size
hecnme commonplace and repetitive throughout thk study,
and number of patterns when the data was reviewed. The
resulting aggregats ROIP ch ratio for all projects W*
294
SPE 27763 JAMES F. PAUTZ, ROBERT A. WELCH AND CHANDRA M. NAUTIYAL 5
Table 2
Summary of OOIP and ROIP Revisions
.,
-“””””’’OOIP”””””
‘“ ‘“’”””:“ ““ “ROIP
c/rl cfrl
Calc Reported ratio Calc Reported ratio
-. .: ~,g9g ..: .. .. ~2,41
SF019 ~ “_.1626 ... ...16.40 ,,.11:60 “ :“: ‘ 1.070
SF022 ..-”4L102 ,58:00 0.690 33.73 55.00 0.613
SF026 ““14..64 14.70 0.996 ., 14.26 13.98 1.020
SF0312 3.9.08.. .. 62.40 0.625 ..,38.26 50.00 0.765
SF091 8.43 11,00 ._ 0.766:. ~ 6.69 10.26 0:652 .. .
sF*oo .. ..-
12.30 12.30.. -.= 1.000 11.80. io.oo 1.180
SF1137 36.45. 37.00 ., 0.985 .35.11 33,60 ‘1.045 :’
SF158 . ..5.55 .,.7.,20 0.771 4.70 7.05 0.666””. .’ .”” ““.
SF186 -:.. 1.30 .-.. ..1.31. ~ 0.992 . -1.11 da da
SF242 8.02 =8.00. ....1.003 7.29 ....7.00 < 1.041
SF345 ,82.3~ Ii/az . . !@ 72.65 nlaz nla2
SF407 4.10”. .4.15 ~Q.988 4.00 .= 3.s4 1.131.
sb4i4 4.47 3.90, . . 1.146 ...4.$8 3.80 .1.152
SF422 8236”.” ..160:00 0..516 70.56 .133.00 0531
“Table 3
Case” 11 Predicted Production and Injection for 14 Projects
““”
Fiedi&ediRk~Ortad (plr)
Oil Production, MBbl Production Injection
ModeI Reported Oil Water Steam “‘
.706..
SF414 164 “-0232 ‘0..173 0.366. “ .“- .“ ““”””” : ~
SF030 “ .3,227 7,167 0.450 “0:641 1.168
SF242 1,557 2,934. 0.531 0.159. 0.485
SFO19 1,010 1,541 0.655 1.696 2.423
SF186 .
SF407 1,433 701. ., 2.044 5.929 9.744
SF158 1,752 . 736; 2.381 1.756. 0.966
SF026 2,131 433 4.921 2.471 4.527
,,689. --,...
SF022 ., .8,741 5.175 0.312 4.552
SF091 1,263”. “___“207 :.: 6.093 1.9~i 1.662
total =. ---. --.70,168.. ,. 70,951
P/r = 0.989
mean = . . ..2: ,.-. .:=. ._ .=....,,1.946 1.363 2.222
,..
Shaded ama is + 307. of a perfect match. “-””
295
6 COMPARISON.OF ACTUAL RESULTS OF EOR FIELDPROJECTS TO SPE27763
CALCULATEDRESULTSOF”EORPREDICTIVEMODELS
107%, and the aggregate oil production p/r ratio was 79%, complete data for the fust year, so the best average of early
indicating that on the whole, modifying only the dm.inage injedmdata war used. The aggregate total production p/r
area Cld not improve the results. ratio decreased from 99..to 837., and Ore range of the
individual project p/r ratios, 3 to 338%, was an
For Case “IV, the project description andlor pattern improvement when compared to Case II (23-”609%)...Stezun
development timing for seven projects wis changed. injection p/r ratios for all projects also decreased, from 222
Projects SF022 and SF026, which were modified in Case to 110%, and once again, the range improved from Case
IR by changing the pattern size and the number of patterns, II’s 37-974~o to 87-218%.. ..Mean water. production was
were further modified. SF022 wai fully developed in the 81% of “thereported volume which is comparable to Case
first 2 “yeah of operation and SF026 in tbe filst year. II. The individual project values are presented in Table 4.
Projects SF030,” S“F158, &d SF407. were changed to
develop all their patterns in the first year. These changes The best re.$uftsilom all five cases revealed seven projects
were made because the yearly well count dam for these whose oil production p/r ratio agreed within zE3070of a
projects in~icated.they were fully developed early in their perfect match. Three projects, SF107, SE345, and SF422
project life and their well counts remained constant over appead in fOU of”the five cases. Based on ROfP volume,
time. SF414S development occurred over the fust 3 years. these projects ranked lst, 2nd, and 5th in terms of size
SF107 was. unique in that its formatirki dip angle was 530.. Based on thimean p/r vhes, the SFPM overpredlcted oil
The ~ea was adjusted from 104 surface acres to 173 acres production (105-1!34’%)and steam injection (110-288%) f6r
to compensate for the formation dlp angle. The resulting every sensitivity case perfotied. Mean water recovery was
aggregate ROIF c/r ratio for. Case ~ titer .rnaking these overpredlcted for Case I-IV (136-178%) and only
changes ww 102~R, and the aggregate oil production p/r underpred]cted in Case V (81 To). This compares to
ratio was 917.. These changes had little effect on tbe underpredicting the aggregated total oil production in all
agg~gate results, and in many instances hurt the accuracy CIWS(79-99%). No other trends emerg~, when the model
of individual projects. results were ““sorted ag”~nit any key reservoir property.
Major project description changes, based on re-
C&S&Y interpretation of the data, helped some projects and hurt
This case was a refinement of CaSe If by supplying the others. Most of the projects did not differentiate between
model with actual steam injection rites rather than letting total and incremental oil production or no incremental
the model determine them. In most cases, the average production was cited. For those projects that dld report
(bcwepd, barrels of cold water equivalent per day) frst year separate incremental production, very little was gained
steam rates. were used. Several of the projectr did not have when the model results were compared to the incremental
Table 4
Case V Predicted Production and Injection for 14 Projects
-.
Predicted/Reported (p/r)
Oil Production, MBbl Production Injection
Model Reported Oil Water Steam
SF407 20 701 0.029. 0.894 1..297.
SF019 .123 1,541 0.080 0.676 0.869
SF030. ~ ..- 2,439 ““.. 7,167 0.340 0.581 1.022:
SF414 36.6 ~ ‘=706 0-518 0.291 1.001
SF345 13.741 23.067 0.596 0.745 1.185..
SF026 ’288 “’”433 0.665 ““”0.535 .0.922
. ..”.
296
,.
.
297
. .
Table 5
Confidence Limits for Cases I-V for 14 Projects
~.-
. . . ... .
Percent of total Percent of Mean sI)~o Confidence Level
Case Reported Volume Total reported oil Proj. Prod. FaRs between
. OOIP, ROIP, Production, Production, LIrnlts of Predicted, %
% % plr, % plr
....+..... . .. .
1 69 _.” 12 71 ““. ” 1,50. 52t0174 ““”
99 . .. .. .99
II n/a 1.95 -. 41 to 162
m ““”””n/a 107 79 1.37 57 to 172
Iv n/a 102 91. 1.83 43 to 132
v n/a 99 83 .1.05 65 to 381
.. ..—
mini–mum of 3 years production, insufficient data for a selecting the initial 14 projects. Case V for both sample
project description w-ere”avaiiable, the project was never groups was an attempt to tine tune the project descriptions,
implemented, or tlere were uncorrectable inconsistencies ii so the poorer total or aggregated oil production was
the data. unexpected as was the broader confidence interval when
compared to Caae II. The confidence interval improved 6’%
Based on the previous results, only “two model runs were for Case V with the addition of 32 projextta. The mesn oil
performed cm this larger group. They are analogous to production p/r of 90 to 1277. for the 46 project sample
C&i II “imidV, matching ROfF and supplying actual steam indicates that the SFPM slightly overpredicts oil
injection rates, respective y. The results of Case I were production if the default steam injectiori”rates are used and
included to report the OOIP and ROIP volume ratios after slightly underpredicta if steai injection rates ze used with
attempting to match the OOIP. The results are the model on average. Using the combined statistics of
summarized in Table 6. 1.27 p/r overprdlction and a range of 69 to 208% at 50%
confidence, the results indicate that we are 5070 confident
Comparing the small sample results in Table 5 .m. the that the actual til production is between 54 and 163Y. of
larger sample i: Table 6 after ROIP consistency was the predktad oil production for a specific project. When
addressed, the Case E results show a significant reduction tictud steam injection rates we used, we are 5070 confident
in the total p/r (81 from “9970),but the mean oil production that the actual oil production is between 89 and 419% Of
ratio ““im”prFved(127 from 195%). ‘The SO% confidence the predicted oil production for a specific project. The
inteival increased sli~htly, and the limits each shifted about under-prediction causes thk second confidence Iimita to
30 pciints higher. N.ormaIly, tb: confidence internal sound significantly worse than the frost, but in application
narrows when more elements (projects) are added, but the they are only slightly worse.
14 project subset was evidently a high-graded sainple.
Case V, where actaal first year average jiearn ratea were The designers of the predictive models have indicated that
input to the model, also generated worse statistics the aggregate oil production is more accurate than the
compared to the smaller group, when comparing the total prediction for a specific project. The accuracy of the
production ratio and the aggregated total oil production. aggWg~e Oilproduction (9990 for Case D and 83’%for Caae
These trends are. consistent. with the method used for V) for ~e initial 14 projects was excellent, but tbe
Table 6
Confidence .Llmits on 46 Steam Projects
Percent of Totaf Percent of” Mean 50% Confidence Level
Case jkfrOrted volume Total Reported Oil Prod. Falls Between
OOIP, ROIP, Production, Production, Limits of Predicted, %
% % plr, % plr
85 .:: ..-. da
I 90 nia n/a
g da “99 “ 81 1.27 69 to 208
v n(a .99. ,67 0.90 ..:. .80 to 377
Aggregat+ nJa II/a da 0.78 10> to 160 at 75% Confidence
298
.,,
299
10 COMPARISONOF ACTUAL RESULTS OF EOR FIELDPROJECTSTO SPE 27763
CALCULATEDRESULTSOF EORPREDICTIVEMODELS
Table 7
Confidence Limits for Case I-V for Gas Displacement Projects
Percent of Total Percent of Mean 50% Confidence Level
Case Reported Volume Total Reported oil Prod. Falls Between
ROIP, Production, Production, Limits of Predicted, q.
% plr, % plr
Tertiarytarget 27 0.287 294t0 1,818 “““’
:- .. . .
-92 180 2.188 33 to 138
III 92 182 2.303 32 to 131
~9.m 368.
N 92 145 1.947
v Tefi”@-target 42 0.669 , 212 to 1,493
underpredicts tertiary oil production both for individual production and to the reported gross oil production when f.
projects (677.) and in aggregate (42%). The range of p/r information was generated. When compting incremental
for Case I was 0“.t0”151% and 1.5 to 300%. for Csse V tertiary production, the model results mrderpredlct
indicating that precision was lost with thk refinement in aggW+@d oil projection by 73% when the CC?2.PMdefault
the data while accuracy was gained. Sorw Is us~ ~d 58~0 when literature values are used for
SOrw. The mean p/r tertiary production for Case If at 69%
Case II, which also used tbe default value of Sorw but relative to the aggregate implies that smaller projects are
adjusted fo’to allow for mobile oil, generated 180% of the more accurately predicted than large projects. For Cases II,
aggregated reported production and 219% mean oil III, and IV, each employing mobile oil, the total aggregated
production for individual projects. The predicted to reported oil production p/r ratio was 180, 182, and 145%,
(p/r) ratios of production for the individual projects ranged respectively. The result again indicate a slight bias toward
from 12 to 754% for Case If and the actual production falls a better project match for larger projects for gross oil
between 33”aad 138% of the predictti production. Case III production.
improved upon Case If by adjusting the iIIjector_well CO.~nt
to closely match the actual injection well count., in the polymer Flood., (PFPM) Displacement Process
project description. Taken directly from the Project
Database with only minor changes, @e If projected 3,320 The Polymer Flood study began with 48 potential projects
injection wells, or 39% of the total 8,466 injetion wells identified from the DOE Project files. Twenty-seven of
for all projects. from 19S 1 to 1992. With manual these were in the Project Database, and 17 of those projects
intervention to the input data set, the Case III model nm had stilciint resefi6ii Ziid/orproduction data to be included
projected 9,021, “or 107% .of the actual injection well in thk study. One project, SF270, was eliminated because
count. The effect’ on oil production, however, was of problems with &ta input accuracy leaving a total of 16.
minimal. (This was an unexpected result bul the model The quality of the OOIP and ROIP raw data for” the
optimizes iirjection rates so that predicted productions are polymer projectr was quite good and required much less
similar from a given area and relatively insensitive to modification “than the thermal or miscible project data.
changes in pattern size and well counts unless the inject Since there was Iittfe”tertiary production information, only
rate is specified.) Total predicted oil production was 182’% the polymer augmented waterflood option was analyzed.
of the total reported, and the range of plr was 12 to 7567..
TMs Fetinernent dld not have an significant affect on either With only minor changes to the project descriptions, the
a precision or accuracy: Oni final refinement to the project volumetric ROIP total volume of 429 MMBbl, which
descriptions, Case IV, provided actual COZ injection fitcs. represented 13.of the 16 project$; was 105.% of the total
Actm4 average f@t year carbon dioxide injectkui rat& W.me reported ROIP. The injection well count bafance was also
quite good. The TfP operators reported 1,383 ictive
supplied to the model, and thk impmved the total produced
injectors for all projects from 1981 to 19.92as compared to
oil pfr ratio to 14570. Unfortunately, the match for
the models projected 1,187 injectors. Surprisingly, with
individual projects dld not improve as reflected in the
this favorable matching, the total oil production predcted
increase in the confidence band compared to Case II and HI.
by the model was 75 MMBbl or 3639.” of aggregated oil
A summary of the results is presented in Table 7.
production reported (Case II).. The range of predictedto
reported (@) production was also large varying .frOm9. to
The SFPM production was compared to gross oil
production because all production was assumed to have 1,035%.
been a d]rect result of steaming operations; i.e; the oil was
unproducible without steam. The miscible flood model Rk.finihg the project descriptions by using actual
runs have been compxed to reported incremental teI’d~ pOlymer/water injection rates. from .fiist yew”.ive.rages
resulted in ameven greater overprediction of oil production
300
.,.
SPE 27763 JAMES F. PAUTZ, ROBERT A. WELCH AND CHANDRA M. NAUTIYAL 11
Table 8
Results for Polymer Flood, Chemical and In-situ Combustion Projects
Number Percent of Percent of Mean Range of Mean
Case of Reported Volume Total Reported Oil OiI Production
Projects ROIP, Production, Production, pfr, %
% plr, % plr
Polymern. . 16 ..,. ~ ,io5...=.- . , 363. ,
3.191 9 to 1,03s ‘“ ‘“
Polymer V .16. 105.. 442 4.347 86 to 1,077
SUrfacmnt 8 lofl 117 ‘1.627 71 to 294
Alkaline 100 1.171 3 to 894
1;. .
In-sitn 100 := .1.273 15t0402 .“”
combustion
—. ,. ..-
301
12 COMPARISONOF ACTUAL RESULTS OF EOR FIELD PROJECTSTO SPE 27763
CALCULATEDRESULTSOF EOR PREDICTIVEMODELS
p/r of 127%. The limited results indicate that the The chemical model showed mixed results. The simulation
aggregated prcdcted oil production is .conservativi and. that of surfactant baaed projects had reasonable agreement with
the ICPM is not as imprecise ai the rdkaline, C02PM, or reported results. A rough estimate shows a mean
PFPM. The range of acc~acy at ~: 50% cOnfidegse @+ production of 160% of aitual and a 50% confidence tfratthe
was 28 to 149% foi p/r or the actual production is 70 to actual production falls between the predicted production and
360’?~6ftie predicted production. Although the number of half the predicted production. Aggregated oil production
samples was less than the lower limit for meaningful for 8 srrrfactant projects w= slightly optimistic at 117%
statistical analysis, the results indicate a reasonable basis indicating a reasonably accurate model for surfactafit
for reliance on the results of the model. projects. Simulation of alkaline-based injectant projects
showed that this portion of the chemical model is overly
CONCLUSIOfiS conservative for ag~egate. total oil production. EMrer the
The project “descriptions and production data in the DOE data on a couple of projects are wrong or the model is
EOR Floject database is exte.nsl.ve, but there is no! imprecise.
sufficient data to do complete statistics on most of the
predictive niodels. In addition, there is concern that the The in-situ combustion simulations showed reasonable
project described at the beginning of “TIP doe: not match results. The mean p~r waa 1277. for eight projects. A
the project actually implemented. Although numerous rough estimate of the precision is that actual production is
inaccuracies were screened from the data, there continues to between 66 imd 350% of the predicted production at .50%
be a need to refine project data. confidence.
Given tbe lirn~tations. of the data, the statisilc~l All the Yesults..are just” indications 6f “the quality of the
conclusions area measure of the combined accuracy of the predictive models because the project description and
TIP project data and the predictive models. The results of, production data were collected and submitted over a number
tfrk study indicate the DOE steamflood model does a good of years by numerous people and very likely contain errors.
job of predicting iggiegated oil ‘production for a random WMch projects= improperly described and which projects
selection of projtiia-in “the range of “80% of the actual oil are poorly simulated can not be determined without
production titb 75% confidence that the r&ulta are HO%. significant amorrrita“ofadditional work.
RECOMMENDATIONS
For an individual steam project, the mean predicted oil
prodrrction varied sigriifkairtly ranging from 90 to 127% of The completeness and quality of the informatiori ori”the
the actual oil production. with the. application of reasonable TfY prnjecta is reflected in the lack of mandatory reporting
engineering”judgment m the project desaiption. The range requirements in”this program. A base line oil production
of the 50.% confidence limit is roughly 70 to 2109. of the projection would make it easier to determine what
calculated oil production. These statistics can be used with production is enhanced and would be desired as part of the
steamflood model results to indicate the accuracy of tbe project description.
estimated oil production for an individual project.
Improving and expanding on this study has already begun
Tbe miscible gas model underpredlcta production in the by refining the project descriptions.””Information has been
tertiary mode when the C02PM default values for are used sent to selected operators to ccmfirm project description aid
aa well as when literature values are used for SOW when production information. A review of literature to expand
mobile oil is included in the project description, the mean beyond the TIP project portion of the DOE project database
oil production is roughly twice the actual production. is also being explored. Outside information was used in
There is a 50% confidence that the actual production falls the miscible gas sensitivity ran. using published residual
between 3.0 to 140% “of the. pretictezl gross oil productiim oil saturations. Additional information on projects fight
when default valuer are used for COZ injection rates.’Tbe confirm or refute the limitations of the data.
use of reported or actual gas injection rates dld npt help the
predictions made b.y._Lhe .miscib!e ._gas model. The trend of results indicate that some of the predlcfive
Unfortunately, there ue”” not .enougb TfP project: to “modelsare definitely inaccurate. The Polymer model and
estimate the accuracy of “the aggregate total oil production the alkaline portion of the CFPM need to be reviewed,
or tertiary production for a group of miscible projects. Because the gas displacement process is such an important
process, thk model and project data should be given special
The polymer model grossly ove~redicted oil production for consideration.
ti polymer augmented waterflood project. The mean
predicted oil production for the sample of TfP projicts was ACKNOWLEDGMENTS
3 to 4.5”times the reported oil production. The aggregated This work was spmrsored by the U. S. Department” of
production for 16 projects wa.”rorighly over’predicted by the Energy under cooperative agreement DE-FC22-83FE601.49
same level as the mean. The large vwiance did not allow as part of the BE2 project. The authors thank Raymond J.
confidence levels to be estimated, Heemstra and Rex D. Thomas (consultants) for their
assistance in modifying Znd running the computer modelx
Carolyn A. Sellers of NIPER for extracting the
302
.!,
inforniatitiri from tle DOE Project Databas.q Associated Symposium on Enhanced 0,1 Recovery, April 15-18,
Western Universities” Intern David A. Terhune for 1984.
conducting the statistical analysis; and Brian Keltch, Bill 17. Fox, M.J., Simlote, V.N., Stark, K.L., and Brinlee,
Llnville, and Min K, Tham of BDM Oklahoma and R. L.D., Raview of COZ Flood Springer “A” Sand, NE
Mike Ray of the DOE Bartlesville Project OffIce for their Purdy Unit, Gamin County, Oklahoma, Paper
critical reviews.
SPE/DOE 14938 .presented attbe SPE/DOE Fifth
.—..
.. . ., ., . ‘Symposium on Enhanced Oil Recovery, April 20-23,
1986.
REFERENCES 18. Moore, J.S. and C1ark, G.C., History Match of the
Maljamar C02Pilot Performance, Paper SPEiD0E :“ -“:”
1. National Petroleum Council, Enhanced Oil RecoveW
17323 presented at the SPWJ)OE Sixth Enhanced 011
December, 1976.
Recovery Symposium April 17-20, 1988.
2. National Petroleum Council, Enhanced Oil Recovery,
19. Flanders, W. A., Stanbeny, W.A., and Martinez, M.,
December, 1984. Review of C02 P.erforqnce. of the Hansford
3. Interstat5–O”il Compact Commission’ Project on
k4armaton Unit, Paper SPEiJ30E 17327 presented at
Advanced (Xl Recovery and the States. An Eva(@on.
the SPE/DOE” Sixth Enhanced Oil Recovery
of the Known Remaining Oil Resource. in the Stat@
Symposium April 17-20, 1988T
of Texas, November 1989.
20.. Brock, W.R. and Bryan, L.A., Summary Results of
4. Supporting Technology for Enhanced Oil Recovery - COZ EOR Field Tests, 1972-1987, Paper SPE 18977”
Stcwmflood Predictive Model, DOE Report DOEiBC-
presented at the SPE Joint Rocky Mountain
86/6/SP, December, 1986.
Regional/Low Permeability Reservoirs Sympoiium
5. Supporting Technology fo!.Enhancgd 0!! Recove~ -
and ExhMtion,M~ch6-8, 1989.
CL22 Miscible Flood Predictive MOd@ DOERepOfi
21. Pittaway, K.R. and Rosato, R.J., The Ford Gerafdine
DOWBC-86(12LSP, December, 1986. Unit COZ Hood.’ Update 1990, Paper W% 20118
6. Supporting Technology for Enhanced Oil Recovery -
presented at the 1990 Pernrian Basin 011 and Gas ‘
Polymer Flood Predictive ModeL QOE Report
Recovery Conference, March 8-9,1990.
DOE/BC.86/l.Q/SP, December, 1986.
22. Lee, K.H. and E1-Saleh, M.M., A Full Field
7. Supporting Technology for EnhancedOil Recovery -
Numerical Modeling Study for the Ford Geraldine
In-Situ Combustion Predictive Modil, DOE Repoti Unit COZ Flood, Paper SPE 20227 presented at the
DOEIBC.861.7LSP,December, 1986.
“SPJYDOE Seventh Symposium on Enhanced Oil
8. Supporting Technology for Enhanced Oil Recovec$ -
RecoveV, April 22-25, 1990.
Chemical Flood Predictive. Model, DOE Report
23. Enhanced Recovery Week: EOR Project Sourcebook,
DOE/BC-~6/11/SP, December, 198.6.
Pasha Publications Inc., Arlington, VA., 1986.
9. Thomas, R.D., EOR ProcessPredictive Screening
k40dek, DOE Report NlPER-47, Ma.rch 18,1985.
10. Pautz, J.F.,. Sellers, C.A, Nautical, C., and Allison,
E. Enhanced Oil Recover yProject Database. DOE
Report NIPER-583, March, 1992.
11. H1ll, M, and Vang, E., SYSTAT for the Macintosh,
Version 5.2, SYSTAT, Inc., Evanston, JL, 1992.
12. Levone, R.L., Statistics@r Management,: Fourth
Edition, Prentice-Hall, Englewood Cliffs, NJ, 1987.
13. Gom.aa, E. E., Correlations. for Predicting 011
Recovery of Steamflood, J. PeLTech., (Feb. 1980),
pp 325-332- ... . . ...-==. . .. : .. =
14. Botiigoyne, Jr., A.T.,” K]mbrell, C., and Goa, W.,
Pilot Oil Atlas for Louisi&z’ Final Report, JuIY,
1992, DOE Grant DE-FG07-891D12842.
15. Kumar, R. and E1beck, J., COZ F~~odi~.g a
Waterflooded Shallow Pennsylvanian Sand in
Oklahoma: ACase History, Paper SPEMOE 12668
presented at the SPE/DC)E,Fourth Symposium on
Enhanced CMRecovery, April 15-18, 1984.
16. Bolline. J.D.. A Full Field Model of the East Velrrm
West ‘Block” Sims Sand.–Uni.i liwrvoi.r, Paper
SPE/DOE 12720 preiinted attbe SPE/DOEFourtb
303