Sei sulla pagina 1di 9

Developing New Fields Using Probabilistic

Reservoir Forecasting
C.S. Kabir and A. Chawathe
, ChevronTexaco, and S.D. Jenkins, A.J. Olayomi, C. Aigbe,
and D.B. Faparusi, Chevron Nigeria Ltd.

Summary the reservoir. In addition, the traditional tornado chart, used for
Limited and uncertain geologic and engineering data at the onset ranking the variables, does not provide any information pertaining
of any new field development are the bane of reservoir character- to statistical significance of the independent-variable effects on
ization and simulation. The problem stems from the uncertainty in the dependent variable.
various model-input variables, such as reservoir connectivity, fluid However, systematic approaches19 have emerged to account
viscosity, and endpoint saturations, to name a few. Given this for uncertainty associated with various input variables, on the basis
scenario, an ad hoc, one-factor-at-a-time approach to earth and of ED. For instance, Chewaroungroaj et al.5 demonstrated the use
flow-simulation modeling cannot possibly yield unbiased informa- of ED with a series of dimensionless variables to allow extension
tion for making objective business decisions. of the results to similar systems. On the other hand, Corre et al.7
This study presents three field cases in which both engineering used ED to integrate data from diverse sources, including seismic,
and earth-model variables were varied in a systematic way to in their effort to quantify uncertainty. Application of ED also has
assess reservoir performance by use of the experimental design been extended to history-matched reservoirs.8
(ED) approach. In ED, geologic (e.g., fluid contacts, net-to-gross pay thickness)
Results of the field cases show that well requirements (both and engineering (e.g., anisotropy, well count) variables are varied
producers and injectors) turned out to be fewer than anticipated. simultaneously, unlike the ad hoc, one-variable-at-a-time ap-
Equally important, one case study showed that laboratory mea- proach. The advantage of the ED approach is that it generates
surements could minimize uncertainty surrounding oil viscosity relatively unbiased p-10, p-50, and p-90 probabilistic estimates by
and endpoint saturations. At the same time, we learned that the capturing nonlinear interactions of variables. Full-factorial three-
preferred horizontal-well orientation was marginally superior to level designs, which account for nonlinearities, are currently rea-
vertical wells in light of high reservoir anisotropy. In another case, sonable with analytic simulations10 or when only a few (n<4)
stratigraphy, gas/oil contact (GOC), and aquifer strength became variables are involved. Realistically, with finite-difference simu-
the primary variables for the full-factorial design after the initial lations, one can afford to do only a subset of the simulations at a
screening. Here, we proved that the project could proceed because cost of some higher-order interactions between the variables. Here,
it met the minimum reserves criterion. Perhaps most importantly, we chose to rely on the statistical principle of sparsity; that is, the
all studies showed how to obtain unbiased information in far fewer main effects of a few variables cause most primary influences.
flow-simulation runs than one would do using an ad hoc approach. In this paper, we used a three-step procedure that involves
variable screening using a simple, linear design, followed by a
nonlinear analysis and the response surface approach. This meth-
Introduction
odology is outlined in detail in Friedmann et al.2 and is similar to
The availability of very limited data with large uncertainty pre- the one originally presented by Damsleth et al.1
sents significant challenges to any new field development. Stakes This paper demonstrates the use of ED for making rapid and
are high when deepwater prospects are evaluated. With advances unbiased business decisions and, therefore, differs from those
in 3D seismic, the reservoir surface may be mapped with a certain available in the literature. In particular, the paper stresses an op-
degree of confidence. However, a few exploratory wells cannot erating-company approach, vis-a` -vis developing new data-
provide detailed information about the reservoirs internal archi- gathering strategies and the like. We also discuss the limitations of
tecture, particularly with respect to flow barriers or baffles. In ED, especially in an earth-science context, wherein variable
short, we are confronted with large uncertainty in the reservoirs screening tends to be more qualitative than quantitative in nature.
flow and, sometimes, fluid properties.
Given limited and uncertain data, questions arise about how to Methodology
proceed with a field-development plan. Historically, we have used In essence, the proposed approach entails a three-step procedure
reservoir simulation and sensitivity analyses as tools for predicting using ED. First, we screen a large number of variables by using the
various scenarios, followed by economic analysis. But the ap- two-level (low and high) design of Plackett and Burman (PB).11
proach has been less than satisfactory because of the ad hoc nature This step is designed to identify the major variables, or heavy
of the exercise, meaning the need to change one variable at a time. hitters, influencing the dependent variable (recovery, for in-
In essence, this approach relies on setting one variable at the p-10 stance) by capturing only the linear effects. Rapid screening with
or p-90 level while keeping others at the p-50 level in a simulation a minimum number of flow simulations is the focal point here.
run. Subsequent ranking of the independent variables can conceiv- Second, the time-lapse recovery (or any other dependent vari-
ably be biased. Potentially, this bias stems from two sources. First, able) is used to rank the sensitivity of the major variables, identi-
simulations may not contain independent information in the result- fied in Step 1. The variables belonging to the 95% confidence
ant 2n+1 simulations, where n represents the number of variables. interval are selected for the three-level (low, mid, and high) D-
Second, the method relies on comparing solutions in which all but optimal or full-factorial design. Thereafter, flow simulations are
one variable are set at the p-50 level. Finally, when one uses p-50 run on the basis of this design. This step is designed to capture both
values in the conventional analysis, the p-50 outcome may be an linear and nonlinear effects in addition to interactions between
unrealistic expectation owing to nonlinear fluid-flow behavior in the variables.
Third, we generate a response surface with multivariate analy-
sis by fitting a polynomial, which then serves as a proxy to the
flow simulator. To fill in the information void generated by the few
Copyright 2004 Society of Petroleum Engineers
flow-simulation runs, Monte Carlo simulations are done with the
This paper (SPE 87643) was revised for publication from paper SPE 77564, first presented response surface to generate the cumulative distribution function
at the 2002 SPE Annual Technical Conference and Exhibition, San Antonio, Texas, 29
September2 October. Original manuscript received for review 13 February 2003. Revised
(CDF). Recovery factors corresponding to different probabilities
manuscript received 6 August 2003. Paper peer approved 12 December 2003. are then derived from the CDF.

February 2004 SPE Reservoir Evaluation & Engineering 15


This methodology produces unbiased results if sufficient care is fects of independent variables on a dependent variable, such as
taken to correctly identify the range of all variables. recovery factor.
Effects of different variables on oil recovery may be best
Case Studies judged by viewing the Pareto charts, which, in essence, represent
Case 1. The first two reservoirs reported here are defined by the tornado charts reflecting the hierarchy of independent vari-
several anticline-type structures and truncated by small splay ables. On a Pareto chart, the x-axis represents the t-test values, and
faults. Although no cores were available, an earth model was con- the y-axis represents the variables. Details can be found in a sta-
structed by integrating well logs, cores from other analogous res- tistical text, such as Ref. 13. Figs. 1 through 3 represent the
ervoirs, and 3D seismic. Six rock types or facies were identified in time-dependent nature of these variables influencing recovery.
the log signatures by establishing the log/core relationship from Note that the negative value associated with a variable suggests
reservoirs above and below the subject reservoir. that its increase has an adverse effect on the dependent variable
To speed up the simulation effort, we constructed the flow- (oil recovery, in this case). The opposite is true for a positive value.
simulation model from the detailed 278,124-cell (864277) earth In other words, the charts indicate that increasing the well count is
model by upscaling it both areally and vertically. The resultant good for recovery, while increasing the residual oil decreases re-
upscaled model had 25,284 cells (432128). Local grid refine- covery. These observations are consistent with those one might
ment around the producers ensured solution quality in this 100-m expect intuitively.
grid model. As expected, the number of producers plays a dominant role
Our main objective was to find the number of producers and early in the life of the project. However, as the waterflood matures,
injectors and their locations, given the uncertainty of certain geo- residual-oil saturation plays an increasingly dominant role. In fact,
logic and engineering variables. These variables included number toward the end of the projects life, it becomes the most influential
of wells (producers and injectors), types of wells (horizontal, ver- variable. This outcome is not too surprising given the displacement
tical, or slanted), reservoir anisotropy (kV/kH ratio), horizontal per- process involved in the adverse viscosity-ratio situation. Two other
meability, heterogeneity in terms of the Dykstra-Parsons12 coeffi- variables, kV/kH ratio and well orientation, begin to influence the
cient (CDP), and residual-oil saturation (Sorwf). recovery with the floods maturity.
Step 1. We invoke the principle of sparsity and use the PB This screening study led us to consider just four variables for
design, which involved just eight flow-simulation runs for the the next phase of the study. These variables included number of
seven-variable matrix. Table 1 presents the variable matrix and the producers, residual-oil saturation, well orientation, and the kV/kH ratio.
recovery results of flow simulations in % original oil in place Step 2. With four major variables, we constructed the 16-run
(OOIP). Certain variables, such as well type and kV/kH ratio, are experiment matrix for the D-optimal design, as shown in Table 2.
represented by 1, 0, and 1, which reflect low, mid, and high cases, Recoveries are reported in %OOIP. Note that the results of the two
respectively. Note that the PB method requires a minimum of a tables differ somewhat owing to parameter interaction or curvature
multiple of four runs. Here, because we started with seven vari- in the three-level D-optimal design. Curvature is the difference
ables, eight runs sufficed. Of course, for fewer than four variables, between the value predicted by the response surface of a two-level
one can use the full-factorial design involving 2n simulation design when all parameters are maintained at the 0 level (p-50)
runs. The main idea of the PB design is to capture the linear ef- and that predicted by the three-level design.

Fig. 1Pareto chart showing recovery after 5 years. Fig. 2Pareto chart showing recovery after 15 years.

16 February 2004 SPE Reservoir Evaluation & Engineering


nomials are shown below; note that only two variables are con-
sidered because of the marginal impact of the other two.
RF@5YRS = 3.59 + 15.01*Nwells 6.13*SORWF
2.68*Nwells**2 + Error
RF@10YRS = 2.59 + 17.91*Nwells 19.24*SORWF
3.29*Nwells**2 + Error
RF@15YRS = 10.89 + 15.65*Nwells 29.35*SORWF
3.01*Nwells**2 + Error
RF@25YRS = 29.88 + 1.69*Nwells 41.97*SORWF + Error
Using these polynomials, we used Monte Carlo simulations
with 1,000 realizations to generate the cumulative probability dis-
tributions of the dependent variable, oil recovery. Figs. 7 through
Fig. 3Pareto chart showing recovery after 25 years.
9 represent these distributions. The probability density function
(PDF)-derived p-50 recovery is shown in each graph with dashed
Following the flow-simulation runs, we developed the Pareto arrows. Note that this p-50 recovery is not necessarily the same as
charts, which are shown in Figs. 4 through 6. As expected, the that obtained from a simulation run with all variables set at the
well count dominates, not just as a linear (L) term or a primary p-50 level. This comment is even more appropriate in the context
variable but also as a quadratic (Q) term or a secondary variable. of p-10 and p-90 recoveries.
Herein lies the value of this step, which is designed to capture the These results show that most of the oil is recovered in the first
nonlinear or secondary effects (ignored during initial PB screen- 10 years. Thereafter, incremental oil diminishes with increasing
ing) in addition to linear effects. water cut, as expected in an adverse mobility-ratio situation. In
Figs. 4 through 6 show a trend of variables that is reminiscent fact, given the range of input variables, the p-50 recovery factor is
of what we observed earlier in Figs. 1 through 3; that is, the well only about 23% of OOIP, less than half of that projected for the
count dominates early in the projects life. However, with the Case 2 reservoir, as discussed later. Of course, the oil viscosity in
waterfloods maturity, the residual-oil saturation plays an increas- that reservoir is less than 1 cp, whereas in Case 1, it is approxi-
ingly important role and eventually surpasses the well count as the mately 5 cp.
major variable. The statistical significance of the ranking is indi- During this phase of the study, oil viscosity, derived from the
cated by the p0.05 vertical line, which implies that there is only oil sample collected at the exploratory well, was estimated to be
a 5% chance that these results could be a fluke. In other words, all 5.2 cp at the bubblepoint. Because of possible contamination of
variables to the right of this vertical line are considered significant. this oil owing to the presence of oil-based, petrofree mud, viscosity
To put it another way, we say that the well count, both linear and was thought to be a variable with a range of uncertainty. Therefore,
quadratic, and Sorwf are significant in Figs. 4 through 6. we made another set of D-optimal design experiments by simply
Step 3. Only 16 flow-simulation runs out of a possible 81 (34) adding viscosity as the fifth variable. Table 3 presents the range of
were actually made. The first column of Table 2 identifies those variables selected for this set of experiments. Also included here
runs. To fill in the information void, we generated polynomials are other assumptions that were imbedded in the simulation runs,
using multivariate nonlinear regression. These polynomials, devel- consistent with those used before.
oped at various points in the projects life, represent response Following exactly the same procedure discussed before, we
surfaces, which serve as a proxy to the flow simulator. The poly- generated probabilistic oil-recovery curves with number of pro-

February 2004 SPE Reservoir Evaluation & Engineering 17


Fig. 5Pareto chart showing relative contributions of variables
Fig. 4Pareto chart showing relative contributions of variables to recovery factor at 15 years.
to recovery factor at 5 years.

The final earth model consisted of 215,280 cells (2305218),


ducers as the variable, as shown in Fig. 10. Similarly, Fig. 11 was with areal cell sizes of 5050 m and average thickness of 2 ft.
generated with well orientation as the variable. In both figures, These grid dimensions were selected to effectively model the res-
vertical lines indicate the probability, from 1% to 99%. Here, the ervoir and to maintain a maximum layer thickness of 3 ft in the
effect of the number of producers is clearly demonstrated. The fine-scale model. Of the 215,280 cells, 172,203 cells were active in
benefits of the second producer in early life, both in terms of flow simulations. This is designated as Model II. To preserve
recovery and cash flow, are obvious. However, the same is not true geologic nuances, neither model was upscaled.
when we compare the performance of the second and third wells; Similar to the previous case, we wished to optimize the number
thus, we reach the conclusion that two producers suffice. Interest- of producers and injectors and their locations, given ranges of
ingly enough, if the oil viscosity is treated as a constant, the un- uncertainty for certain geologic and engineering variables. These
certainty spread remains about the same in early life but reduces variables include horizontal-well length, reservoir anisotropy (kV/
quite a bit at late times, unlike those observed here. Clearly, the kH ratio), horizontal permeability, injector completion, and pro-
message is that reduction in the uncertainty band for a dependent ducer and injector counts. The simulation model uses a 100-m grid
variable, such as recovery, must include determination of oil vis- with local grid refinement around producers. Sensitivity to cell size
cosity and residual-oil saturation in the laboratory. Note that oil also was investigated using an alternative geologic model with a
pressure/volume/temperature (PVT) data were unavailable at the 50-m grid.
initiation of study. Unlike the first case, the first step (PB design calculations) was
As Fig. 11 shows, the horizontal wells provide marginal supe- unnecessary because both fluid viscosity and endpoint saturations,
riority over conventional wells because of high reservoir anisot- important variables likely to have a large impact on recovery, were
ropy. Perhaps an explanation is in order. Separate 1D vertical flow well defined. Results of the D-optimal design strongly support
simulations showed that the effective or intercell vertical perme- eliminating the PB phase of the study.
ability was one to two orders of magnitude lower than those of the The variable matrix and uncertainty ranges used in this study
assigned cell values. Such ultralow effective vertical permeability are presented in Table 4. Beyond this matrix, we assumed that
does not provide the right environment for optimal performance of production and injection rates for a well could not exceed 3,000
horizontal wells. Nonetheless, horizontal wells were favored be- STB/D and 10,000 STB/D, respectively, for maintaining comple-
cause of completion considerations in this unconsolidated formation. tion integrity in this unconsolidated formation. We also assumed
that a single geostatistical realization represents the reservoir.
Case 2. A quick-look faulted earth model was initially constructed Note that well location is not explicit in this experimental ma-
for this reservoir. This model contains approximately 57,000 cells trix. Numerous preliminary runs were made beforehand to opti-
(1162718), with areal cell dimensions of 100100 m and aver- mize locations. Within the scope of desired higher structural el-
age thickness of 2 ft. This model was made from a single strati- evation for producers and correspondingly lower elevation for in-
graphic grid encompassing the entire sand, without using addi- jectors, we searched for sweet spots and reservoir connectivity as
tional internal stratigraphic control. It was used as a screening tool
to quickly establish optimal methods for property distribution, ini-
tial estimates of OOIP, and reserves for different development
scenarios. This quick-look model is called Model I.

Fig. 6Pareto chart showing relative contributions of variables


to recovery factor at 25 years. Fig. 7Probability distribution of oil recovery at 5 years.

18 February 2004 SPE Reservoir Evaluation & Engineering


Fig. 9Probability distribution of oil recovery at 25 years.
Fig. 8Probability distribution of oil recovery at 15 years.

Following the generation of Pareto charts, we developed re-


the drivers for preferred well locations. Both static reservoir de-
sponse surfaces at every 5-year interval, using the multivariate
scription and dynamic flow-simulation results formed the basis for
regression analysis discussed earlier. Only 25 of the total 243
location optimization. We observed that minimal spatial changes
(35) flow simulations were actually made. This is how the ED
were needed to maximize oil recovery. To summarize, optimal
approach reduces the large number of simulation runs to a man-
well placement was searched in the background before initiating
ageable few.
the ED work.
We generated the PDF and CDF after Monte Carlo simulations.
Figs. 12 and 13 represent the time-dependent nature of five
Figs. 14 and 15 present the charts showing the recovery factors
variables on recovery. Clearly, the number of producers plays a
corresponding to p-10, p-50, and p-90. Note that the steepness of
dominant role in determining recovery from this reservoir. Note
the elongated S-curve does not change because we are not adding
that the number of producers occurs as both linear (L; primary) and
new information to the project in these flow simulations. However,
quadratic (Q; secondary) terms in the response surface equation,
with the projects maturity and an update of reservoir models,
beyond 5 years of production. Well length becomes marginally
uncertainty can be reduced in an absolute sense. Recovery factor is
important toward the end of the projects life. All other variables
not the only dependent variable that can be optimized. For in-
fall outside the 92% confidence interval. The reason that well
stance, Fig. 16 shows that the oil rate can be viewed the same way.
length becomes a minor contributor stems from setting the with-
After generating results from the quick-look earth model,
drawal rate at a constant value of 3,000 STB/D, owing to comple-
Model I, we verified the solutions by making simulation runs for
tion/operational considerations. In other words, the increasing pro-
the p-50 cases using Model II. Note that Model II has approxi-
ductivity index with increasing length is not taken advantage of in
mately four times as many cells as Model I. Here, the benefits of
early life because of the imposed rate constraint.
employing a fourth producer and a second injector were investi-
These charts certainly confirm that uncertainties associated
gated. Because oil recoveries of the two models are very similar,
with horizontal-well length, kV/kH ratio, horizontal permeability,
as shown in Table 5, we did not see a compelling need for invok-
and injector completion have a minor impact on oil recovery.
ing the ED approach for Model II.
These results are not entirely surprising because displacement is
Questions arise about why the fourth producer does not make a
occurring in a high-permeability, favorable mobility-ratio situation
larger positive impact on recovery (Table 5). Examining time-
in this reservoir, resulting in recovery of approximately 50%
lapse saturation profiles suggests that the location of the fourth
OOIP. We emphasize that the presence of lateral barriers, if any,
producer (in the fairway of the flood front) causes its suboptimal
may change the outcome of actual response in the field. Lateral
barriers, arising from partially sealing faults, can potentially im-
pede flow, resulting in less-than-optimal recovery, as projected in
this study. The number and distribution of flow barriers are very
difficult to ascertain before development drilling occurs. Transient
testing has the potential for identifying the distance and strength of
these partially sealing barriers that go undetected in 3D seismic.

Fig. 10Probabilistic reservoir forecasting: effect of well count.

February 2004 SPE Reservoir Evaluation & Engineering 19


We relied on reasonable aquifer support based on the perfor-
mance of an analog in the proximity of this reservoir, but the lack
Fig. 11Probabilistic reservoir forecasting: effect of well orientation. of long-term data could not confirm this assumption. The compe-
tency of the shale barriers (stratification) was inferred from the
performance. Then again, that is the only logical location for a well logs. Admittedly, this factor, even if quantified, cannot be
fourth producer, given the anticipated reservoir heterogeneity. controlled, but the intention was to mitigate its effect through a
As the water-saturation profiles indicate, sweep initially occurs sound distributed reservoir-pressure measurement with forma-
from south to north, before the flood fronts move east and west. tion testers.
This pattern of fluid movement is a direct consequence of the These factors were carried forth into our ED approach. As with
reservoir permeability trend and injector location. A second injec- other cases, we performed the well-optimization studies indepen-
tor does help recover incremental oil; however, the incremental dently. The results of the PB analysis confirmed our suspicions, as
gain is not large and may lie within experimental error for forward shown in Fig. 17 (Step 1). GOC (which was expected to directly
simulations without any production history. impact the reserves) and stratification (shale competency) showed
up as significant variables after 25 years of production. Because
Case 3. To develop or not to developthat was the question for we sought to establish the ultimate reserves, this analysis was not
the Case 3 reservoir. This borderline reservoir needed a minimum done in a time-lapse mode, unlike the previous cases. Fig. 17
of 40 million STB reserves to be economically viable. Primary indicates that higher stratification results in lower recovery. Satu-
uncertainties in the reservoir included the location of the GOC, ration maps generated from the simulation results confirmed that
aquifer strength, residual-oil saturation, and the competency of stratification enhances aquifer encroachment. Consequently, many
shale layers, which could act either as vertical baffles or barriers wells watered out prematurely, thereby reducing recovery. This
to flow. finding prompted us to perform further sensitivities on the strength
The uncertainty in the GOC stems from our inability to resolve of aquifer support. Although horizontal wells performed better
the gas crossover limit at 7,674 ft, as observed in the well logs. than vertical wells in general, GOC and stratification dominated
Because the actual GOC was never penetrated, the possibility of the overall uncertainty in the system.
changes in lithology (sand to shale) or fluid type contributed to this Based on the results of the PB analysis, a full-factorial design
uncertainty. The lowest known gas was estimated in one of the was run on two significant variablesGOC and stratification
wells at 7,686 ft, whereas the highest known oil was spotted (Step 2). This was a three-level design resulting in nine (32) runs.
at 7,780 ft. Using the results of the full-factorial design, a quadratic response
To compound the GOC uncertainty issue, the solution gas/oil surface with one cross term was generated, as shown in Fig. 18.
ratio (GOR) in the field was fairly high at 900 scf/STB. An inde- In Fig. 18, open circles represent the results of the nine full-
pendent PVT measurement from a different region of the reservoir factorial simulations.
had indicated that the system should be undersaturated, yet expe- This quadratic response surface was subjected to 1,000 Monte
riences show that a slight change in the estimation of the solution Carlo simulations, and a probabilistic recovery CDF was generated
GOR could result in a saturated system. The water/oil contact was (Step 3). The CDF indicated p-10 and p-90 recovery values of 29.6
well established at 7,866 ft. Given these factors, the overall un- and 57.1 million STB, respectively. The CDF also ascertained a
certainty in the GOC could reduce the height of the oil column by p-50 recovery of 43.3 million STB, which gave us the initial green
up to 50%. Consequently, the other question being asked was light to proceed. However, given the marginal p-50 reserves above
whether the reservoir should be developed with horizontal or ver- the economic threshold value of 40 million STB, the project got
tical wells. stalled because it could not compete with other lucrative projects

Fig. 12Relative contributions of variables to the recovery Fig. 13Relative contributions of variables to the recovery
factor at 5 years, Case 2. factor at 19 years.

20 February 2004 SPE Reservoir Evaluation & Engineering


Fig. 15Probability distribution of oil recovery at 19 years.
Fig. 14Probability distribution of oil recovery at 5 years.

at the time of recent budget allocation. This project will get a a given OOIP. Of course, major changes in reservoir connectivity
favorable management response when the risks are mitigated by may have a significant impact on a development plan.
codevelopment of gas reserves in this field. In the first two case studies, we presumed a single geostatistical
We point out that risk mitigation is an important element in the realization to be valid for answering the questions that we posed
proposed approach by gathering additional data through new seis- (i.e., how many producers and injectors were needed for field
mic acquisition, new drill, and the like. Continuous updating of the development). Experiences suggest that different geostatistical re-
model and recomputation of CDF must follow for reassessment of alizations do not necessarily yield different solutions when busi-
the projects economic value at a given point in time. ness decisions are made on the basis of recovery. This point was
made when the two geologic models, quick-look and detailed,
Discussion yielded very similar recoveries in Case 2. However, Case 3 indi-
Although we studied the uncertainty of various variables, the issue cates that major geologic uncertainties, such as stratigraphy, will
of optimal well location was not explored explicitly. Rather, we impact recovery. But we caution that such variables are extremely
relied on a deterministic approach for ascertaining well locations, difficult to quantify in continuous probabilistic terms, thereby
independent of the ED approach. We recognize that there are ways questioning the estimation of low, mid, and high identities.
to optimize well locations en route to maximizing an objective Some geological variables, such as structure, are even more
function, such as net present value.14,15 Gu yagu ler and Horne14 difficult to quantify because of their inherent spatial nature. Avail-
extended the well-optimization strategy with a hybrid genetic al- able algorithms can produce a CDF of structures (and/or geologic
gorithm (GA)-polytope method. Unfortunately, the current well- models), which may be ranked on the basis of OOIP. However, we
optimization strategies1416 are not well suited for multiple geo- perceive this approach to be of limited value because a true un-
logical realizations and complex well combinations. However, certainty measure is incomplete without incorporating a develop-
given the size of the prospects and the degree of heterogeneity, we ment plan, which leads us back to the chicken-and-egg situation.
think that the ED approach used here suffices. Experiences under- We also think that the presence of yet-to-be-established flow
score the need to find ways to incorporate the development plan barriers will have the most impact on reservoir fluid flow and,
early in the uncertainty-estimation process. However, this notion therefore, on recoveries. To improve the odds, we recommend
sometimes leads to the chicken-and-egg situation; that is, a devel- updating these relatively inexpensive studies as additional data are
opment plan is based on a specific geologic model, but what is a gathered with the drilling of development wells. In this way, one
p-10 (or p-50, or p-90) geologic model without a development can minimize uncertainty in forecasting the performance of
plan? One recent study17 suggests that the well count, a measure of new fields.
development plan, does not change with rock properties alone for Finally, there is the issue of uncertain variables vs. engineer-
ing sensitivities. Some variables can be controlled, such as the well
count. On the other hand, permeability cannot be controlled and is
considered to be a pure uncertainty. In an ideal scenario, we rec-
ommend that the variables be split into two separate designsone
for the uncontrollable variables, and the other for the control-
lable variety. In general, experience indicates that engineering
variables (controllable) tend to dominate the reservoir performance
and, therefore, if coupled with the uncontrollable subsurface
variables, might overshadow the pure subsurface uncertainty.

Fig. 16Probability distribution of oil rate at 10 years.

February 2004 SPE Reservoir Evaluation & Engineering 21


Fig. 17PB analysis shows the importance of GOC and stra-
tigraphy, Case 3.

Conclusions Fig. 18Quadratic response surface for two variables (GOC


A systematic ED approach is presented, enabling one to develop a and stratification) based on the results of the full-factorial design.
reservoir by considering uncertainty in various input variables.
Three field examples illustrate the application of ED in developing 4. Venkataraman, R.: Application of the Method of Experimental Design
new fields in west Africa. All examples had variables that were to Quantify Uncertainty in Production Profiles, paper SPE 59422 pre-
known to various degrees of certainty. Specifically, we reached the sented at the 2000 SPE Asia Pacific Conference on Integrated Model-
following conclusions. ing for Asset Management, Yokohama, Japan, 2526 April.
1. ED is an unbiased, rapid approach to obtaining probabilistic 5. Chewaroungroaj, J., Varela, O.J., and Lake, L.W.: An Evaluation of
results. The approach allows identification of significant vari- Procedures to Estimate Uncertainty in Hydrocarbon Recovery Predic-
ables influencing the dependent variable (e.g., recovery). tions, paper SPE 59449 presented at the 2000 SPE Asia Pacific Con-
2. Engineering variables governing recovery in terms of decreas- ference on Integrated Modeling for Asset Management, Yokohama,
ing importance are well count (often the most dominant), oil Japan, 2526 April.
viscosity, residual-oil saturation, well orientation, and anisotro- 6. van Elk, J.F. et al.: Improved Uncertainty Management in Field De-
py. Fluid contacts and stratigraphy are the dominating variables velopment Studies Through the Application of the Experimental De-
from geological considerations. sign Method to the Multiple Realisations Approach, paper SPE 64462
3. From a recovery standpoint, reservoir heterogeneity (in terms of presented at the 2000 SPE Asia Pacific Oil and Gas Conference and
anisotropy and CDP) plays a weak role. However, major geo- Exhibition, Brisbane, Australia, 1618 October.
logic features, such as stratigraphy and fluid contacts, can se- 7. Corre, B. et al.: Integrated Uncertainty Assessment for Project Evalu-
verely impact a development strategy. ation and Risk Analysis, paper SPE 65205 presented at the 2000 SPE
European Petroleum Conference, Paris, 2425 October.
Nomenclature 8. Manceau, E. et al.: Combination of Experimental Design and Joint
Modeling Methods for Quantifying the Risk Associated With Deter-
CDP Dykstra-Parsons coefficient, dimensionless
ministic and Stochastic UncertaintiesAn Integrated Test Study, pa-
kH horizontal permeability, md
per SPE 71620 presented at the 2001 SPE Annual Technical Confer-
kV vertical permeability, md ence and Exhibition, New Orleans, 30 September3 October.
n number of variables, dimensionless 9. Friedmann, F., Chawathe , A., and Larue, D.K.: Assessing Uncertainty
p probability of occurrence in Channelized Reservoirs Using Experimental Designs, SPEREE
Sorwf residual-oil saturation after waterflood, fraction (August 2003) 264.
o oil viscosity, cp 10. Kabir, C.S., Ainley, C.M., and Brown, D.R.: An Analytic Simulator
for Rapid Forecasting Rate Behavior of Oil Wells, paper SPE 36725
presented at the 1996 SPE Annual Technical Conference and Exhibi-
Acknowledgments
tion, Denver, 69 October.
We are grateful to ChevronTexaco management and that of our 11. Plackett, R.L. and Burman, J.P.: The Design of Optimum Multifac-
partner for permission to present this work. We are indebted to torial Experiments, Biometrika, vol. XXXIII, University Press, Cam-
colleagues Neil Young (for building the earth model for the Case bridge, England (1943) 305.
2 reservoir) and Alan Bernath (for all-around assistance with the 12. Dykstra, H. and Parsons, R.L.: The Prediction of Oil Recovery by
Case 1 and Case 2 studies). One of us (CSK) learned a great deal Waterflood, Secondary Recovery of Oil in the United States, API
about ED from Francois Friedmann, who also contributed to cre- (1950) 160.
ating and analyzing ED for the first two case studies. 13. Montgomery, D.C.: Design and Analysis of Experiments, fifth edition,
John Wiley & Sons, New York City (June 2000).
References 14. Gu yagu ler, B. and Horne, R.N.: Uncertainty Assessment of Well
Placement Optimization, paper SPE 71625 presented at the 2001 SPE
1. Damsleth, E., Hage, A., and Volden, R.: Maximum Information at Annual Technical Conference and Exhibition, New Orleans, 30 Sep-
Minimum Cost: A North Sea Field Development Study With an Ex- tember3 October.
perimental Design, JPT (December 1992) 1350. 15. Gu yagu ler, B. et al.: Optimization of Well Placement in a Gulf of
2. Friedmann, F., Chawathe , A., and Larue, D: Uncertainty Assessment Mexico Waterflooding Project, SPEREE (June 2002) 229.
of Reservoir Performance Using Experimental Designs, paper CIM 16. Aanonsen, S.I. et al.: Optimizing Reservoir Performance Under Un-
2001-170 presented at the 2001 Canadian Intl. Petroleum Conference, certainty With Application to Well Location, paper SPE 30710 pre-
Calgary, 1214 June. sented at the 1995 SPE Annual Technical Conference and Exhibition,
3. Dejean, J.-P. and Blanc, G.: Managing Uncertainties on Production Dallas, 2225 October.
Predictions Using Integrated Statistical Methods, paper SPE 56696 17. Badru, O. and Kabir, C.S.: Well Placement Optimization in Field
presented at the 1999 SPE Annual Technical Conference and Exhibi- Development, paper SPE 84191 presented at the 2003 SPE Annual
tion, Houston, 36 October. Technical Conference and Exhibition, Denver, 58 October.

22 February 2004 SPE Reservoir Evaluation & Engineering


degree in petroleum and natural gas engineering from Penn-
SI Metric Conversion Factors sylvania State U. Chawathe has served as a Technical Editor
and Review Chair for SPEREE. Steve Jenkins has worked on a
bbl 1.589 873 E01 m3 number of seismic imaging and reservoir-characterization proj-
cp 1.0* E03 Pas ects with Chevron USA, Caltex Pacific Indonesia, Chevron Ni-
ft 3.048* E01 m geria Ltd., and Tengizchevroil. His current work includes estima-
ft3 2.831 685 E02 m3 tion of probabilistic oil in place for Tengiz field, Republic of
Kazakhstan, and evaluation of exploration opportunities for
*Conversion factor is exact. Tengizchevroil. Jenkins holds a BA degree in geology from the
U. of Tennessee at Knoxville and an MA degree in geophysics
from Indiana U. Adebola Joseph Olayomi is a reservoir simula-
tion engineer at Chevron Nigeria Limited in Lagos, where he
Shah Kabir is a senior advisor with the Nigeria/Mid-Africa busi- has been employed since 1991; previously, he had stints in
ness unit at ChevronTexaco Overseas Petroleum, Bellaire, production and completion engineering. Before joining Chev-
Texas. He has published extensively in the areas of well testing, ron, he worked as a wellsite operations engineer for three
fluid and heat flows in wellbores, and reservoir engineering; his years with Shell Petroleum. Olayomi holds a BS degree (honors)
latest effort includes the SPE book Fluid Flow and Heat Transfer in petroleum engineering from the U. of Ibadan, Nigeria. Chris
in Wellbores. He holds an MS degree in chemical engineering Aigbe is a reservoir engineer with Chevron Nigeria Ltd., where
from the U. of Calgary. Kabir has served on many SPE commit- he is responsible for the conduct of flow-simulation studies of
tees, including the editorial review committees of SPEPF, various black-oil and gas/condensate reservoirs. Previously, he
SPEREE, and SPEJ; currently, he serves as a Review Chair for was a lecturer at Auchi Polytechnic, Nigeria, as well as a field-
SPEREE. He has received commendation as an outstanding based production engineer and a petroleum engineer with
Technical Editor five times for two different journals. In 2002, he Chevron. He has extensive experience in well testing, well/
received the SPE Western North America Regions Service reservoir problem diagnosis and remediation, workover plan-
Award. Adwait Chawathe is a senior reservoir engineer at ning and execution, production optimization, and new-field
ChevronTexaco working on the Wafra field in Kuwait. Before development. Aigbe holds an MS degree in chemical engi-
joining Chevron in 1997, he worked for 2 years at the New neering from the U. of Benin, Nigeria. Biographical information
Mexico Petroleum Recovery Research Center. He holds a PhD for D.B. Faparusi was not available.

February 2004 SPE Reservoir Evaluation & Engineering 23

Potrebbero piacerti anche