SPE 166452

Insights From History Matching and Forecasting Work for a Steeply-

Dipping, Faulted Volatile Oil Reservoir
Sandeep P Kaul, Chevron ETC, Anil Ambastha, Chevron Nigeria Ltd., Vincent Eme, Chevron Nigeria Ltd., and
Jefferson Creek, Chevron ETC

Copyright 2013, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Annual Technical Conference and Exhibition held in New Orleans, Louisiana, USA, 30 September–2 October 2013.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not been
reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers, its
officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is prohibited. Permission to
reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
History matching and subsequent forecasting work becomes a challenging task if limited supporting
production data is available and the reservoir is severely depleted. For an offshore, volatile-oil reservoir,
added to this challenge was an uncertainty in fluid PVT, where the data clearly suggested presence of
condensate, but with black-oil properties. The permeability distribution from logs was counterintuitive
to the production data from the wells. The reservoir had a structural relief in excess of 1000 ft., most
likely having API gradient, but both the API and the GOR data indicated that there were possible errors
in measurement. There was uncertainty associated with original oil-water contact also. The production
data showed the reservoir to follow primarily a classical solution gas-drive response, but simple material
balance analysis proved a weak aquifer effect as well.

The approach followed in simulation was the process of elimination. Pressure match was first achieved,
but questions remained about its robustness around the main sealing fault. GOR was targeted next and
several different condensates and one full compositional fluid model of a nearby reservoir were
unsuccessfully tested. For matching the historical gas production, a new high condensate yield fluid
PVT was used. The idea of another oil-water contact (OWC) was tested in the saddle of the reservoir to
account for most likely early water breakthrough in a well there. The secondary gas cap formation and
its effects were crucial in achieving satisfactory history match.

The confidence in the history match, as having captured the physics of the flow, led to forecasting
scenarios which were not possible with a black-oil model. Most of the data was found not to be
erroneous. What was needed was judicious data interpretation to achieve satisfactory history match. To
produce these kinds of depleted, faulted reservoirs further, a strategy to better manage the evolution of
secondary gas cap was of utmost importance.

Introduction
Reservoir M is the deeper producing reservoir of a field in offshore Nigeria which has been continuously
on production since Oct. 1980 (~33 yrs). The field is a complexly faulted, collapsed rollover anticline
and is saddle separated from one of the biggest fields in Chevron’s portfolio in Nigeria. Its proximity to
that big reservoir has resulted in the pressure effects being felt in shallow sands suggesting that the
regional aquifer is common to both the reservoirs. The structure of this reservoir is having a dip of
around 1000 ft. from crest to the spill point. The crestal well of this reservoir has produced for the
2 SPE 166452

largest period in the life of the reservoir. In all, nine fault blocks have been identified and developed.
Some faults are known to be leaky, resulting in fluid communication.

Around the end of 2011, the asset decided to evaluate waterflood (WF) opportunity for all mature fields.
Based on decline curve analysis, Reservoir M was identified as a possible WF candidate. The reservoir
was already pressure depleted by 67% from initial pressure and initially seemed to have no upside. As
per good reservoir management strategy, majority of fields are further developed, beyond primary
depletion, by waterflooding. Earth model was constructed and a history match undertaken to ascertain if
there was a value in further producing the reservoir under repressurization (WF).
At a high level, the objective of history-matching exercise is to better understand the reservoir behavior.
This is usually associated with taking a closer look at the drive mechanisms of a given reservoir.
Conventional wisdom does require performing a classical material balance (MBAL) study, as shown by
Dake1, Kabir et al.2, Pletcher3, Esor4, etc. which gives a quick insight into the drive energies of the
reservoir, but such course is seldom taken. Part of the problem is the number of wells, which then
becomes large, makeing the MBAL model itself unwieldy. Coupled with this is the large data
management problem which discourages the practitioner to use this method. For these huge cases
(multi-million cell models), there are large number of papers in literature highlighting the various
approaches and techniques that can be used to solve the problem such as that by Williams et al.5,
Ambastha et al.6, etc. to name a few. But the underlying basis of all these deterministic approaches is
that the simulation engineer is provided with very reliable data inputs. A detailed model then captures
the reservoir performance.

Also since history matching has a non-unique solution; stochastic approaches like Experimental Design
together with Assisted History Matching (AHM) techniques have been found to be very powerful tools.
A few prominent papers which show the application of these techniques are Hoffmann et al.7, Emanuel
and Milliken8, King et al.9, etc. These approaches try to capture complex reservoir performance by
changing input parameter ranges and going ahead with a suite of reservoir models to get to most
probable reservoir performance.

Majority of above history-matching techniques are centered on the modification of static parameters
such as porosity, permeability, transmissibilities, compartmentalization, etc. Challenge and complexity
in history matching comes from understanding complex reservoir behaviors which are not easily
apparent based on the available data. As a result, various possible reservoir behaviors may not get
investigated within the current normal practice of using available softwares and therefore, associated
drives may or may not make it under the microscope of the simulation engineer. Bartlett et al.10
describe the challenges associated with Atlantis, a reservoir in Gulf of Mexico, which started with being
simple but after development drilling and initial production performance showed extensive faulting,
baffles and presence of perched water. Availability of drilling, seismic and other additional reliable data
helped in a better reservoir development plan. The restricted aquifer support was supplemented by water
injection for better reservoir management. An improved understanding of reservoir connectivities was
also achieved. Although the reservoir was geologically more complex, there was no surprise in
identifying the reservoir drive mechanisms which determined the performance.

Surprises are often associated when there is interaction of multiple reservoir drive mechanisms in a
complex reservoir architecture aided by gravity. One of these one-off approaches is modeling gravity
segregation drive. Although it is possible to demonstrate it with the MBAL method, as done by
Ambastha and Aziz11, it can elude the engineer if sufficient attention is not paid. A full-field
compositional simulation study by Ypma12 was performed on Statfjord reservoir to know the role of
compositional effects during secondary, gravity-stable nitrogen injection. The overall conclusions are
similar to this study as gas condensates formed near the top of the reservoir and volatile oils got
SPE 166452 3

accumulated downdip of the structure. But what sets this work apart from Statfjord study is the critical
thinking associated with arriving at this gravity-segregation scenario, without the aid of any commercial
history-match tool or extensive phase-behavior experiments performed before simulation. We did not
have any fluid characterization report to begin with. Alternatively, commercial material balance
software, like MBAL™13 model, may help in identifying these complex drive interactions, but would
require advanced model setup using equation of state option and/or pore volume variation with depth.
Also, initially we did not make a concerted effort to identify this drive as all PVT models did not
generate a big enough secondary gas cap, the driving force behind gravity segregation drive in our
model. Instead, water from the main aquifer broke through in the wells producing a strong water drive.
The identification of three reservoir drives, which were present in this reservoir, became apparent in the
simulation model after improved PVT model was used. This is not to suggest that we did not realize the
huge dip making an impact, but all the initial results were either inherently inconsistent or did not go
along with what MBAL model was predicting.

Reservoir Model Description

The reservoir has five penetrations in all. Out of these five penetrations, three wells have been active
with one well, Well#1, producing for 28 years continuously before being shut down for high GOR in
2008. Well#4 is still producing in the field at a very low rate and Well#5 was shut down on account of
high BS&W after five years of production. Well#2 and Well#3 were not brought online because the
intent was to produce oil and these wells were suspected to produce gas and hence never perforated.

Earth Model
This reservoir represents reworked shoreface deposits located within the clastic nearshore depositional
environment. The stratigraphic section of the field is shown in Figure 1a with all the five penetrations.
To put this in a better perspective, the gross thickness map, Figure 1b, is also shown. It shows that
thickness is increasing from southwest, where it is the lowest, to northeast, where it is the highest. The
other main features of this earth model are highlighted in the following:

Reservoir M is a reworked shoreface deposit formed via the winnowing of deltaic deposit by
wave and current action.

Reworking is more prominent in NW as evidenced by serrated log character.

It consists of two stratigraphic lobes separated by ~5ft thick, shale interval.

Lower lobe is a thicker, overall coarsening upwards progradational sequence.

Upper Lobe is a thinner, overall fining upwards retrogradational sequence.

Taking into account relevant G&G uncertainties, low, mid and high static models were constructed. The
original oil-in-place was an uncertainty and gave a wide range to probabilistic volumetric estimates. The
original static fluid contacts were by far the biggest uncertainty. The final contact ranges are given in the
Table 1. The main points which need to be highlighted are:

OGOC uncertainty was approximately 69 ft.

OOWC uncertainty was approximately 621 ft. Production data suggests OOWC may be closer to
the maximum closing contour.

No additional data available to narrow the current band of uncertainty (No MDT availability and
low seismic data quality which needed re-processing. Amplitude extractions were not definitive).

Low and High case OGOC & OOWC scenarios used to generate probabilistic volumetric
estimates.
4 SPE 166452

Table 1: Summary of Original Contacts

Low Mid High

-6474’ss -6386’ss
OGOC (HKO in MAL-19) (Crest Of Structure)

-6859’ss -7480’ss
OOWC (LKO in MAL-28) (Max. Closing Contour)

Well#3 Well#4 Well#2 Well#5 Well#1

Figure 1a: Stratigraphic Section of Reservoir M

Well#1
Well#5
Well#3 Well#2

Well#4

Figure 1b: Gross Thickness Map

Booked volumes to date were based on a previous Simulation Study (OOWC sensitization done to
achieve result). Additionally, the static parameters uncertainty analysis, done on the original oil-in-place
(OOIP), suggested that OOWC was the single biggest hitter. This information was very important, as far
as dynamic modeling is concerned. It resulted in two different scenarios: first scenario having 3 different
OOWCs (Black Oil PVT used) and the second scenario with two different OOWCs (Volatile Oil PVT
used). At this stage, there was no consensus on which particular PVT model could be used since there
was no fluid characterization report available for the reservoir. PVT model choice caused the original
oil-in-place (OOIP) estimates of dynamic model to be different from that of the static model. The
SPE 166452 5

structure has the least impact and hence it was taken out of dynamic uncertainty analysis to reduce
scenarios for dynamic modeling.

Dynamic Model
Before upscaling, the fine-scale static model had cell size dimensions of 50 ft  50 ft  2 ft with total
grid blocks of 100  100  72 having 373,104 active cells. It was upscaled in X and Y directions only
to cell size of 100 ft  100 ft  2 ft with total blocks of 50  50  72 having 89,568 active cells.
For the purpose of identifying unique features of this reservoir, for discussion purposes and from CPU
run time stand point, the upscaled grid was divided into 5 different regions as per the following:

Region 1 – The saddle near Well#5 to account for OWC1

Region 2 – The fault block for Well#1.

Region 3 – The fault block for Well#4 to account for OWC2.

Region 4 – Aquifer Region (Deactivated).

Region 5 – Aquifer Region (Deactivated).

Region 4 and Region 5 were deactivated to speed up the run time. The coarse scale model with OWCs is
shown in Figure 2 and region numbers are shown in Figure 3.

Well#5 Well#1

Well#4

Figure 2: Upscaled Simulation Model showing OWCs

The model has an aquifer which is common to Region2 and Region3. Region1 (saddle region) is in
pressure communication with Region3 through the oil zone. The motivation for having an oil-water
contact in saddle was based on water cut production data for Well#5 which showed a value greater than
50% from first month of production. Also water showed up in Well#1 after Well#5 was shut down.
Completions in Well#5 also came to be of questionable integrity, adding to the list of uncertainties in the
reservoir.
6 SPE 166452

Another uncertainty was difference in pressures between Region1, Region2 with that of Region3. If the
single well test value for Region 3 was to be believed, then there was a pressure differential of ~600 psi.
A lot of conclusions cannot be derived from a single data point, but the fact remained that the fault was
sealing in hydrocarbon area. If the fault remained fully communicable, the gas movement became
uncontrollable between Region1 and Region 3. The data did not suggest pressure equalization and in the
presence of gas, sealing had to be present. Figure 3 also shows the possible fault communication
scenarios which were considered.

Well#5

Region 1 Well#1

Well#4

Region 2

Region 3

Figure 3: Upscaled Simulation Model showing Equilibrium Regions & Possible Communication
between Fault Blocks

Challenges Associated with Earth Modeling

The immediate challenge at the onset of history-match process was the inability to have Well#1 produce
at the required historical oil constraint. The basic history-matching exercise was dictated by it. With the
existing earth model parameters, the reservoir pore volume around Well#1 had to be increased to ensure
that historical oil rate constraint was met. The pore volume multiplier of 1.5 was used, which was
deemed to be unacceptable. A closer look at the problem revealed that the reason for this was associated
with the porosity, which showed average porosity values around Well#1 to be depressed. This is also
illustrated in Figure 1a where this information is displayed as light green color. The bright yellow color
as seen in the figure represents higher porosity values. The production data showed that this particular
well, with varying oil rate, had produced uninterrupted for 28 years. As the static information could not
be verified from seismic data, owing to its low quality, it was decided to change the geological setting in
the reservoir based on production data. Keeping the porosity-permeability transform constant, the
geological trend of the reservoir was revised so as to help meet the oil rate constraint during history
matching. The exact reason for going against the information provided by logs could not be arrived at.
The well in question was the first penetration in the reservoir and the fact that the logging tools could
SPE 166452 7

have been of early 1980 era was the only justification which seemed plausible. This resulted in two
scenarios where, more than likely, either the logging tools were unable to pick up correct porosity or the
tool may have had a calibration problem. In conclusion, the production data overrode the information
that was provided by the logs. Figure 4 shows the changes to the values of porosity incorporated to the
earth model as represented through the Kh sum maps (keeping the same porosity-permeability
transform).

Well#1 Well#1

Figure 4: Upscaled Simulation Model showing OWCs

Challenges Associated with Data Measurement Errors

The producing GOR and the API data were also analyzed and are shown in Figures 5 and 6,
respectively. The data suggests that the Rsi = 770 scf/STB (initial GOR) remains constant for two years
for Well#1, before the reservoir goes below the bubble-point pressure and producing GOR starts to
increase. Initial GOR value, allowing for errors in measurement in the field, indicates that it is a black-
oil system as suggested by McCain14. Since gas measurements are the least accurate in the field, the
GOR measurement integrity also came to be questioned.

Well#4 Well#1

Figure 5: Reservoir GOR Data

Communication between fault blocks also needed revisiting. Well#4 was brought on production after
~20 years of reservoir life and still saw Rsi which was below the 1500 scf/STB line (black-oil limit).
8 SPE 166452

This reinforced the idea of fault being sealing in the hydrocarbon area. Initially, this motivated us to try
black-oil as the reservoir fluid. A similar parallel can be drawn when API data was analyzed and shown
in Figure 6. The API of the crestal well, Well#1, has remained constant at 42o API (volatile oil range),
whereas the other two wells have varying API. This is especially true for Well#4 which was brought into
production after ~20 years and produces at 32o API for the first 2 years and API data for Well#5 had a
range from 28o API to 46o API. The reservoir was clearly having volatile oil, but at the same time
exhibiting black-oil properties. As would have been customary under the circumstances, there were two
distinct lines of opinion; one which suggested that there were huge measurement errors in API data and
the other suggesting caution so as not to miss any vital information which the reservoir data was trying
to suggest.

Well#4 Well#1
API (Degree)

API (Degree)

Well#5
API (Degree)

Figure 6: Reservoir API Data

Identification of Reservoir Drive Mechanisms

The production data analysis was carried out with material balance (MBAL model) to understand the
various drive mechanisms and the effect of aquifer. As mentioned previously, the look at the pattern of
GOR production led us to believe that the reservoir was following the classical solution-gas drive. This
was the first drive of the reservoir and has been explained very well in Dake’s book1. The second drive
mechanism was the water influx and the results of this analysis are shown in Figure 7. The unique
feature of this MBAL model output, which mathematically calculates drive indices and shown in the
figure, is that the water influx energy is decreasing in time. Water influx usually follows a pattern of
SPE 166452 9

being ever increasing once breakthrough is achieved in the reservoir. This is a sign of a weak water
drive. Campbell15 plot also supported this hypothesis. Also, the best fit Havlena-Odeh16 plot, given the
error in the fit, suggested no upward revision in the original oil-in-place.
Apart from the fact that there was a dip of around ~1000 ft from the crest to the spill point in the
reservoir, impact of a weak water drive was not very clear at this point in time of the history-matching
exercise.

Figure 7: MBAL Model Output to Identify Drive Mechanisms

Challenges Associated with PVT Fluid Modeling During History Matching

This was the biggest simulation challenge. The data indicated borderline black-oil fluid, but in strict
theoretical sense, it was Volatile oil. With Well#1 as the benchmark, following scenarios were tried in
history-match exercise:

1. Black Oil with single bubble point (no variation with depth)
2. Black Oil with variable bubble point with depth
3. Black Oil with single bubble point and Kv/Kh Ratio = 0.3
4. Condensate (Analog Data) Option and Full Compositional Model
5. Final Condensate Model

For brevity, this paper carries detailed discussions about scenarios #1, #4 and #5 only. The most
important point to be noted was that for all the three black oil models, the historical oil production
constraint was only honored with OOIP which were more than 20% higher than that proposed by
MBAL model. Also, all these black oil models had to have three OWCs to match water breakthrough
at the wells and were repeatedly giving strong water drive signature. In-house program was used to
generate the relevant black-oil PVT data. The base PVT with which simulation was attempted, to begin
with, is shown in Table 2.
10 SPE 166452

Table 2: Summary of PVT Data

Reservoir Properties
Initial Pressure 2,949 psia
o
Reservoir Temperature 218 F
Datum 6775 ft TVDSS

Oil Properties
Oil API Gravity 42 Degree
Initial Solution GOR 770 scf/STB
Bubble Point Pressure 2,500 psia

Black Oil with Single Bubble Point (No Variation with Depth)
This was the first attempt in history matching. Initial GOR of 770 scf/STB was chosen in fault block of
Well#1. The simulation result of this attempt is shown in Figure 8. The GOR match for the whole
history period, especially 1996-2002, for Well#1 was very poor. The gas production, as shown, tracks
historical values early in the life of the well, but once the pressure goes below the bubble point, gas
production is unable to keep up and follows a flat profile. This is also seen in the GOR figure. The
pressure match is shown in Figure 9.

Figure 8: Single Bubble Point Simulation Results for Well#1

SPE 166452 11

Well#1 Well#4

Figure 9: Single Bubble Point Static Pressure Match for Well#1 and Well#4

Condensate (Analog Data) Option and Full Compositional Model

In CHEARS®, Chevron’s in-house reservoir simulator, condensate option was activated and an Analog
PVT (Condensate Yield, RL, of 35 STB/MMscf, API = 35o) was used. This PVT model was attempted to
see whether any further improvement in GOR could be possible. A 4-component full compositional
model was also attempted to see whether that would help. Figure 10 shows the results of static well
pressure for both these Analog condensate and full compositional models. The condensate model gave a
good pressure match, but the full compositional model died after it ran for the first 5 years of history.
GOR match (shown in Figure 11 for condensate model only) was somewhat better than the black-oil
model. The biggest reservation for using these two models was the OOIP, which was too optimistic than
the runs with the black-oil PVT. Both these options were eventually abandoned.

Well#1
Full Compositional (4 Comp.)
History
Condensate Option

Figure 10: Condensate Option and Full Compositional Static Pressure Match for Well#1
12 SPE 166452

Figure 11: Condensate Option Simulation Results for Well#1

Final Condensate Model

Since the GOR match had improved using analog condensate model, it was decided to modify this
existing PVT so that simulation could produce more gas. The basis of change was API of 66o (volatile
oil) measured at the facility. The modified PVT was created out of analog data (RL of 35 STB/MMscf,
API = 35o) and was modified to RL of 75 STB/MMscf at 3000 psi dew-point pressure, API = 62o.

To understand what was happening in the reservoir, it is important to turn to the basics of PVT analysis.
Figure 12 shows the phase envelope of various reservoir fluids. Theoretically, volatile oils have the
largest phase envelope. What this means is that a condensate model may fit into a volatile oil envelope,
but the possibility of black oil fitting into volatile oil envelope was very slim. This is because the
envelope for black oil is very constricted in comparison to volatile oil. This was the major reason why
black-oil models were failing to produce the required gas rate or match GOR.
SPE 166452 13

CPCP® Manual
17 is dictated by its composition
The character of a fluid type

7000
Reservoir Temperature Critical Points

6000
Volatile I

5000
Condensate Volatile II
Pressure (psia)

4000

3000

2000 Wet Gas

Black Oil
1000

Dry Gas
0
-200 -100 0 100 200 300 400 500 600 700 800
Reservoir Temperature o
F

Figure 12: Typical Phase Envelopes

What made the modified condensate model (extrapolated envelope) to work was the gradient of the
liquid lines which would have been similar to the actual volatile oil envelope, had a fluid
characterization report been available. This is depicted in Figure 13.

Actual Extrapolated
Envelope if Envelope
PVT data Critical  Point
were
available Volatile  Oil
Pressure

%  Liquid
50%
30%

10%

Temperature

Figure 13: Extrapolated Phase Envelope

14 SPE 166452

As a result of this, the history-match results obtained were very satisfactory as shown in Figure 14. The
key message which needs highlighting is that all this was achieved without the use of any permeability
or PV multiplier.

Figure 14: History Match Simulation Results using Modified PVT Data for Well#1

Identification of Third Reservoir Drive

It is worthwhile to pause here and go into the details of how the reservoir was behaving. This was the
main reason for the robustness of the history match process. Gravity segregation played an important
part in the overall performance of the reservoir. Figure 15 illustrates the effect of this drive.
SPE 166452 15

Well#5 Well#1

Figure 15: Gravity Segregation Effect in Reservoir M

The saddle was the place where all the action was taking place as shown in Figure 15. The crestal well,
Well#1, was continuously stripping the reservoir of all the lighter components. This resulted in the
formation of secondary gas cap, formed in the structurally higher parts of the reservoir and acted as a
piston drive (similar to gravity-stable gas injection). This drove both the oil and water closer to Well#1.
But the water underrode the oil in the saddle and the gas overrode that same oil. This led to a churning
effect of the entire three-phase system in the saddle area. As a result, the API gravity went from being a
very light oil of 47o API to heavy fraction oil of 28o API (see Figure 6 for API data). This gave us
confidence in the history-match model as having captured the physics of fluid flow. This scenario was
not possible with the black-oil or analog condensate PVT. Again, for both black-oil and condensate
models, the main aquifer was active instead of the aquifer in the saddle, resulting in a very strong water
drive and producing inadequate history match.

End of History Saturation and Best Case Prediction Scenario

As expected for a solution gas-drive reservoir with significant gravity segregation, an analysis of all the
layers of the simulation model revealed the oil saturation to be concentrated on the downdip portion of
the structure (see Figure 16). It was just a matter of putting a well in this area which fetched additional
reserves. Gas lift had to be used to sustain production in the new development well. To have the best
case prediction scenario, it became apparent that the evolution of the secondary gas cap had to be in a
controlled manner. Also, the main aquifer was not the main drive in the reservoir, which was in line with
the MBAL model predicting a weak water drive. Thus the water drive energy came from the fault
leakage / water accumulation in the saddle. Out of the many prediction scenarios being evaluated, none
gave a better result than the one in which Well#5, a producer, was converted to injector and was
responsible for slowing the expansion of the secondary gas cap. This yielded an incremental reserve of
8 MMSTBO, while maintaining the same original oil-in-place as all the previous studies.
16 SPE 166452

Well#5 Well#1

Well#4

Figure 16: End of History Saturations

Conclusions and Recommendations

As a result of this study, the following conclusions and recommendations were arrived at:
1. In a brown field development, with good judgment, most forms of data were honored and to the
maximum possible extent. It was a given that there were errors in data measurement, but these
errors could not be consistently wrong for a long period of time. The drastic variation of API
held a lot of information in the end and proved this point.
2. Generally, two drive mechanisms can be present in any reservoir and are more common.
Sometimes these drives can be induced to efficiently produce the reservoir. Alternatively, these
drive mechanisms maybe apparent in commercial tools, like MBAL™, model which have a
theoretical basis of showing them. What becomes more challenging is to identify the drives that
are outside the general scope of these commercial softwares. This requires more critical thinking
and forces one to study and understand the impact they create on reservoir performance.
3. There was a very subtle difference between fault leakage and presence of small aquifer. In this
simulation study, we could not arrive at a definitive conclusion early on. The only way the idea
got implemented eventually in the simulation model was by incorporating oil-water contact in
the saddle area. All these challenges encountered broke the myth that only large reservoirs are
the ones which are most complicated.
4. Repressurization during water flood in this reservoir would yield good results only when each
drive of the reservoir was identified and steps taken to mitigate the downside associated with
each one of them. For this reservoir, controlled expansion of secondary gas cap was the only
viable option, knowing that the reservoir is severely pressure-depleted.
5. Team approach for such simulation studies is a given, but what makes it successful is the
iterative process amongst various disciplines during history matching. This was seen in this study
when use of production data was favored over well log data.
SPE 166452 17

Nomenclature

GOR = Gas Oil Ratio

kv/kh = Vertical to Horizontal Permeability Ratio
MBAL = Material Balance
OWC = Oil-Water Contact
OOWC = Original Oil-Water Contact
OGOC = Original Gas-Oil Contact
OOIP = Original Oil-in-Place
PV = Pore Volume, ft3
Rsi = Initial Gas Oil Ratio, SCF/STB
RL = Condensate Yield, STB/MMSCF
WF = Waterflood

Acknowledgement

The authors wish to thank Chevron Nigeria Limited for permission to publish this paper. They also
recognize the valuable contributions by Ricardo Combellas, Aigul Tyshkanbayeva and Aluba Oragwu
during revamping the static model and providing quality reservoir performance data.

References
1. Dake, L. P.: The Fundamentals of Reservoir Engineering, 72-98, Elsevier Publishing Company,
Amsterdam (1994).
2. Kabir, C. S., Al-Khayat, N. I., and Choudhary, M. K.: “Lessons Learned From Energy Models:
Iraq’s South Rumaila Case Study”, paper SPE 105131 presented at the 2007 Annual Middle East
Petroleum and Gas Conference, Dubai, Apr. 14-19.
3. Pletcher, J. L.: “Improvements to Reservoir Material Balance Methods”, paper SPE 75354
presented at the 2000 SPE Annual Technical Conference and Exhibition, Dallas, Texas, Oct. 1-4.
4. Esor, E., Dresda, S., and Monico, C.: “Use of Material Balance to Enhance 3D Reservoir
Simulation: A Case Study”, paper SPE 90362 presented at the 2004 SPE Annual Technical
Conference and Exhibition, Houston, Texas, Sep. 26-29.
5. Williams, M. A., Keating, J. F., and Barghouty, M. F.: “The Stratigraphic Method: A Structured
Approach to History-Matching Complex Simulation Models”, paper SPE 38014 presented at the
1997 SPE Reservoir Simulation Symposium, Dallas, Texas, June 8-11.
6. Ambastha, A. K., Al-Matar, D., Ma, E., and Kasischke, B.: “Full-Field Parallel Simulation
model: A Unique Tool for Reservoir Management of the Greater Burgan Oil Field”, paper SPE
102281 presented at the 2006 SPE Annual technical Conference and Exhibition, San Antonio,
Texas, Sep. 24-27.
7. Hoffman, B. T., Caers, J. K., Wen, Xian-Huan, and Strebelle, S.: “A Practical Data-Integration
Approach to History Matching: Application to a Deepwater Reservoir”, paper SPE 95557
presented at the 2005 SPE Annual Technical Conference and Exhibition, Dallas, Texas, Oct. 9-
12.
8. Emanuel, A. S., and Milliken, W. J.: “History Matching Finite Difference Models with 3D
Streamlines”, paper SPE 49000 presented at the 1998 SPE Annual Technical Conference and
Exhibition, New Orleans, Louisiana, Sep. 27-30.
9. King, G. R., Jones, M., Tankersley, T., Flodin, E., Jenkins, S., Zhumagulova, A., Eaton, W.,
Bateman, P., Laidlaw, C., Fitzmorris, R., Ma, X., and Dagistanova, K.: “Use of Brown-Field
Experimental Design Methods for Post-Processing Conventional History Match Results”, paper
SPE 159341 presented at the 2012 SPE Annual Technical Conference and Exhibition, San
18 SPE 166452

Antonio, Texas, Oct. 8-10.

10. Bartlett, R. M., Singh, P. K., Howie, J. M., Humphrey, K., Kurtz, J. A., and Croft, M. K.:
“Unlocking Atlantis: An Integrated Approach to Reservoir Management of a Giant Subsalt
Field”, paper OTC 20397 presented at the 2010 Offshore Technology Conference, Houston,
Texas, May 3-6.
11. Ambastha, A. K., and Aziz, K.: “Material Balance Calculations for Solution-Gas drive
Reservoirs with Gravity Segregation”, paper SPE 16959 presented at the 1987 SPE Annual
Technical Conference and Exhibition, Dallas, Texas, Sep 27-30.
12. Ypma, J. G. J.: “Compositional Effects in Gravity-Dominated Nitrogen Displacements” SPERE
(August 1988), 867.
13. MBAL™, Integrated Petroleum Modeling (IPM) Toolkit, Petroleum Experts Limited, Edinburgh,
Scotland.
14. McCain, W.D. Jr.: The Properties of Petroleum Fluids, 2nd Ed. PennWell (1990).
15. Campbell, R. A. and Campbell, J. M., Sr.: Mineral Property Economics, Vol. 3: Petroleum
Property Evaluation, Campbell Petroleum Series, Norman, OK (1978) 26.
16. Havlena, D. and Odeh, A. S.: “The Material Balance as an Equation of a Straight Line”, JPT
(August 1963) 896.
17. Chevron Phase Calculation Program (CPCP), Chevron propriety Phase-behavior software.