SPE-195103-MS

Simulation and Performance of Immiscible WAG Pilots in Mauddud

Reservoir Using Three Phase Relative Permeability with Hysteresis

Ali AL-Muftah, Yusuf BuAli, Ahmed Mahmoud, and Hamed AlGhadhban, Tatweer Petroleum

Copyright 2019, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Middle East Oil and Gas Show and Conference held in Manama, Bahrain, 18-21 March 2019.

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
The Bahrain Field, being the first oil discovery in the gulf region in 1932, is now in a mature stage of
development. Crestal gas injection in the Mauddud reservoir has continued to be the strongest driving
mechanism since 1938. Over the last five years, gas injection and fluid production rates have grown three
folds with expanded drilling, workovers, and high volume lift activities. However, there are significant
opportunities to increase oil production and optimize gas injection.
An Immiscible-Water-Alternating-Gas injection (IWAG) process was carried out on two composite
samples extracted from the Mauddud reservoir of the Bahrain Field. The resulting production and pressure
profiles were history matched by using hysteresis and three-phase relative permeability modeling options.
Representative relative permeability and capillary pressure curves with the associated hysteresis and three-
phase relative permeability parameters were obtained by history matching the experimental IWAG flood
results. The history match was carried out by generating the hysteresis parameters and relative permeability
curve sets. Experimental results, including two-phase water/gas flood steady state and unsteady state results,
were honored to the degree possible. In both composite samples, the IWAG process showed incremental
recovery compared to the base case water and gas injection cases. The incremental recovery obtained (above
10% PV) was largely due to the reduction of gas relative permeability during three-phase flow. A maximum
trapped gas saturation of 23% was used to history match the core-flood results.
A sector model of the Mauddud reservoir was run using the relative permeability and hysteresis model
parameters obtained from the history matching of the composite core-floods. A water and gas flood base case
was run and compared to the IWAG sequence. The IWAG process showed incremental recovery compared
to the base case water injection. In the up-dip pattern where the water saturation is low, IWAG recovers 3%
more than base case gas injection, while gas injection recovers 5% more than the IWAG sequence in the
down-dip pattern where water saturation is higher.
The objective of introducing the Immiscible Water Alternating Gas process (IWAG) in Mauddud was to
reduce gas production by controlling the mobility during the three-phase flow. Incremental oil, compared
with gas and water injection was also to be evaluated. Three IWAG pilots were introduced after an extensive
study on optimum locations. Two inverted 5-spot patterns and one line drive pattern were selected; each
pattern is around 40 acre spacing, targeting Mauddud B interval. The original Water Alternating Gas (WAG)
2 SPE-195103-MS

ratio was designed to be 1:3 (Water: Gas) and the WAG period was originally designed to be from three to six
months based on simulation work. WAG ratio and duration optimization were subject to performance. After
one year of cyclic injection, both inverted 5-spot patterns showed lack of response to the WAG cycles. In
one of the two latter patterns, the water cycles critically affected oil production. In the line drive pattern, the
WAG cycles initially showed a favorable response. After one year of injection, water and gas overcame oil
production, leading to higher oil decline and the termination of the pilot due to confinement and operational
issues.

Introduction
The Mauddud reservoir is one of the most productive reservoirs of the Bahrain Field, responsible for 60%
of the total oil production from the Bahrain Field. The Mauddud formation is a carbonate reservoir with oil-
wet characteristics (AL-Muftah et al, 2017). The Field commenced production in 1932. Crestal gas injection
was initiated in 1938 following a rapid pressure decline. Production continued for the next seventy years
with crestal gas injection which led to the development of a secondary gas cap. The increased producing
Gas to Oil Ratio (GOR) from the crestal gas injection has led to decreased field performance and has
strained surface handling capacities. It is estimated that there is at least 50-60% residual oil saturation in
the secondary gas cap. In addition to crestal gas injection, water flood patterns have been implemented to
recover oil and to maintain reservoir pressure. High water saturation regions have shown a water cut above
98%. An IWAG process has been considered for mobility control and to improve oil recovery.
In order to study the impact of Immiscible Water-Alternating-Gas Injection (IWAG) in the performance
of the Mauddud reservoir, an advanced core analysis study was carried out on two composite core samples
extracted from the Mauddud formation. The objective of the core study was to determine the performance
of IWAG injection and to identify representative relative permeability and capillary pressure drainage and
imbibition curves with appropriate three-phase hysteresis models and parameters.

Core Flood Experiments

Experimental Setup
Several core plugs at different depths were extracted from the Mauddud formation and were assembled
into two composite samples for WAG experimentation. The composites are formed by stacking the core
plugs end-to-end in the hydrostatic core holder. Each composite sample was weighted and vacuum pressure
saturated with formation brine. The samples were spun in a high speed centrifuge in an air- displacing-
brine configuration to reduce the water saturation to Swirr. The samples were aged using dead crude oil
in a hydrostatic core holder.

Sequence of Experiments
The composite samples were subjected to different injection sequences. Composite-1 was gas flooded and
subsequently the oil saturation was reduced to residual. Then, the sample was subjected to 15 cycles of
equivalent water and gas injection volumes of 0.1 PV until oil recovery ceased. Finally, the composite
sample was flushed with 2PVofbrine. Terminal conditions including oil saturations and recovery factors
were recorded for each injection sequence.
Composite-2 was subjected to a constant 0.24 cc/min water injection rate. After this primary water
flood, 15 WAG cycles were injected at an equivalent rate of0.1 PV. The WAG sequence was continued
untiloil recovery ceased. Finally, the composite sample was flushed with 2 PV brine. Then, the sample was
disassembled and the core plugs were loaded into an overburden centrifuge at 600 psi and 140°F. The core
plugs were spun at a capillary pressure equivalent of114 psi in a gas-displacing-liquid system. Similarly, the
SPE-195103-MS 3

oil saturations and recovery factors were recorded at the end of each sequence. Table 1 shows the injection
sequences and recovery from each sequence.

Table 1—Injection Sequences and Oil Recoveries from Composite-1 and Composite-2

Composite -1, Porosity=0.337, PV=91.40 cm3

Oil Recovery

Soi Sof Oil Recovery-Fraction Pore-Space Oil Recovery-Remaining Oil-in-Place

Gas Flood 0.883 0.481 0.402 0.455

IWAG,0.1PV 0.481 0.383 0.098 0.204

2PV Brine Flush 0.383 0.381 0.002 0.005

Composite-2, Porosity=0.361, PV=93.31cm 3

Oil Recovery

Soi Sof Oil Recovery-Fraction Pore-Space Oil Recovery-Remaining Oil-in-Place

Water Flood 0.899 0.612 0.287 0.319

IWAG, 0.1PV 0.612 0.452 0.160 0.261

2PV Brine Flush 0.452 0.431 0.021 0.046

Composite – 2 Gas - liquid centrifuge 0.431 0.251 0.180 0.418

Core Experimental Results

Composite-1: After the unsteady-state gas flood experiment, the oil saturation was lowered by an amount
that was roughly proportional to the amount ofgas injected with no fixed residual. A total of552 PV of gas
was injected in the gas flood sequence giving a remaining oil saturation of0.481 PV and oil recovery of
45.5%. The following IWAG sequence was run at a rate calculated to be equivalent to about 1 ft/day. This
rate corresponds to a lab rate of 0.0812 cc/min. The IWAG sequence recovered an additional 9.8% PV with
an oil recovery of 20.4%. The WAG flood results show that gas trapping may be the operative mechanism
in providing additional recovery. The final flush with 2 PV brine resulted in a negligible incremental
oil recovery (<1%). This indicates that there may be incremental oil recovery from an IWAG sequence
following gas injection; however, it would be small (9.8% PV) since the gas injection resulted in a high
recovery (40.2% PV) previously.
Composite-2: The initial water flood sequence was carried out by injecting 1.1 PV of water at 0.24 cm3/
min, which resulted in a remaining oil saturation of 0.612 PV. The water flood oil recovery was 31.9%. Water
breakthrough occurred almost immediately after injection of 0.21 PV of water. Figure 1 shows the recovery
efficiency and the Water to Oil Ratio (WOR) from the initial water injection. This figure shows that a large
fraction ofthe oil was recovered after water breakthrough. A 50% water cut was achieved after injecting
0.34 PV and achieved an oil recovery efficiency of 29%. Further injection of water, after breakthrough,
gradually increases the WOR with additional oil being recovered. The water front must have left a higher
saturation of oil due to the oil-wet characteristics of the samples.
4 SPE-195103-MS

Figure 1—Primary water Injection of Composite-2. Water breakthrough occurs after injection
of less than 0.21 PV of water. Further water injection recovers more oil but at a higher WOR.

Further injection of water may provide incremental oil but at a much higher WOR. A 99% water cut was
reached at the end of the water flood. Following the water flood, 0.1 PV of water and gas were injected in
the IWAG sequence. This rate corresponds to a 1 ft/day advance rate and a lab rate of 0.0831 cc/min. The
IWAG recovered 26.1% of the remaining oil-in-place after injection of 1.3 PV of water and 1.4 PV of gas.
This represents a 16% PV reduction in the remaining oil saturation. Following the IWAG sequence, 2 PV
of water was injected recovering 5% of the remaining oil-in-place. The final gas-liquid centrifuge resulted
in the highest recovery of 41.8%. This indicates that IWAG recovery may be lower following a water flood
compared to gas injection following a water flood.

Core-Flood Simulation and History Matching

Core-Flood Simulation Setup
The simulation model of the core floods is a 2D model with length and cross-sections of the composite core
samples. The model properties are assigned the reported values from the core flood experiment. Table 2
summarizes the model properties.

Table 2—Core Simulation Model Properties

Model Properties Composite-1 Composite-2

Simulation model 2D 2D

Model length 24.91 cm 23.76 cm

Cross-sectional area 11.3 cm2 11.19 cm2

Model Dimensions NX=30,NY=1,NZ=1 NX=30,NY=1,NZ=1

Pore volume 91.4 cm 3 93.31 cm3

Porosity 0.337 0.361

Effective permeability to oil at

12.3 md 17.9 md
Swi

Initial Water Saturation 0.117 0.101

The fluid properties are according to the experimentally reported values. Table 3 shows a summary of
the fluid properties.
SPE-195103-MS 5

Table 3—Core Simulation Model Fluid Properties

Fluid Properties at 140 °F Viscosity, cp Density, g/cc

Crude Oil 5.64 0.882

Synthetic Formation Brine 0.474 0.992

Nitrogen 0.0192 --

The Eclipse simulator was used to history match the production and pressure profiles from the IWAG
core-flood experiment. Three-phase relative permeability and hysteresis modeling options were used. The
key mechanisms that affect the flow behavior during IWAG displacement include:

• Gas trapping

• Reduction in gas mobility during three-phase flow

• Reduction in water mobility during three-phase flow

• Reduction of residual oil saturation

The above mechanisms are modeled by including hysteresis in the gas and water phase coupled with the
STONE1 three-phase oil relative permeability interpolation method.

History Matching of Core-Flood Experiment

History matching was carried out with over fifteen variable parameters. The parameters describe the
endpoints and curvatures of the relative permeability and capillary pressure curves, and parameters used to
control hysteresis in the gas and water phases. The relative permeability curves are generated using a LET-
type model. This model provides higher flexibility and is capable of capturing different behaviors across
saturation ranges compared to Corey's model. It is characterized by an "S-shape" with three parameters
(L, E, T) to control the low-end (L), middle elevation (E), and top-end (T) portion of the curve. The LET
correlation for oil and water relative permeability is given by:

(1)

(2)

where Swn is the normalized saturation, kro is the maximum of the oil relative permeability curve, and krw is
the maximum of the water relative permeability curve. A similar curve was used to generate gas-oil relative
permeability curves. Capillary pressure curves were generated using a Brooks-Corey type correlation.
The history matching process was affected by inaccuracies in the reported experimental measurements.
The incremental oil recovery from IWAG is characterized by erratic behaviors as the oil droplets get hung
up in the tubing and gradually accumulate in the receiving tube as a larger drop. This affects the quality
of the interpretation. In all cases, gas production was not measured. Incremental oil and water productions
were requested and incorporated to gauge the quality of the match.
The assumptions and considerations made during the history match are as follows:

• A maximum trapped gas saturation of 23% is assumed in all cases.

• There is no spontaneous imbibition of water. Samples are assumed to be partially wetting.

• The WAGHYSTR Eclipse keyword is used to model hysteresis in the gas-phase during three-phase
flow. The input gas imbibition curve is ignored when using this keyword.
6 SPE-195103-MS

• Hysteresis in the water phase is modeled using Killough's model.

• Capillary pressure hysteresis is modeled using Killough's model.

• Since oil phase relative permeability is dependent on water and gas phase saturations, the three-
phase oil relative permeability is generated using Stone's interpolation model.
It should be noted that the selected hysteresis and three-phase modeling parameters significantly affect
the results obtained. The best result was obtained when using the three-phase model (WAGHYSTR) to
model gas phase hysteresis and Killough's option to model hysteresis in the water and oil phases, along with
the STONE1 model to calculate oil relative permeability from oil-water and oil-gas relative permeability
curves. These modeling options were selected and fixed during the history matching process.

Description of Hysteresis Models

Hysteresis in gas relative permeability. Most two-phase hysteresis models available in the simulator
(Killough, 1976; Carlson, 1981) tend to overestimate the gas relative permeability after a water flood.
A recent model developed by Larsen and Skauge (Larsen and Skauge, 1999) correctly models the non-
reversibility of the gas relative permeability scanning curves. This model is specified in Eclipse using the
WAGHYSTR keyword. The model as implemented in Eclipse requires specification of the drainage gas
relative permeability curve. The imbibition curve is generated from the drainage curve whenever the gas
saturation increases and when the water saturation is greater than the connate water saturation. Subsequent
drainage and imbibition curves are reduced by a factor discussed below. The WAGHYSTR keyword requires
specification of:
Land's Gas Trapping Parameter (C): Trapped gas saturation is given by a gas trapping model proposed
by Land (1968). The input imbibition curve is ignored and is instead calculated analytically. The imbibition
curve is drawn connecting the trapped gas saturation to the reversal point on the drainage gas relative
permeability curve. The trapped gas saturation is calculated from the Land's parameter using the following
relation:

(3)

where Sgm is the maximum gas saturation at the point of reversal on the gas drainage curve and Sgcr is the
critical gas saturation. This relation implies that the trapped gas saturation is dependent on the saturation
history approaching the maximum trapped gas saturation defined by Land's parameter. Sgcr is assumed to be
zero and the maximum trapped gas saturation was fixed at 23%. This implies a Land's parameter C = 3.12,
which is kept constant during history matching. The production and pressure profiles are very sensitive to
this parameter.
Reduction in Krg During 3-Phase Flow (a): This option allows reduction of gas relative permeability
during three-phase flow. It is observed that Krg is significantly reduced during three-phase flow and two-
phase hysteresis models (Killough, 1976; Carlson, 1981) overestimate hysteresis in gas phase (Spiteri and
Juanes, 2004). The value can be zero where there is no reduction in the gas relative permeability to higher
numbers indicating significant reduction in the gas relative permeability. If α = 0, the input drainage curve
is followed with no hysteresis. The effect of this parameter is observed whenever the water saturation in the
model is higher than the connate water saturation. This parameter was crucial in matching the production
and pressure profiles, and also in controlling the stability ofthe model.
Residual Oil Saturation Modification Parameter (a): This option allows reduction of the residual oil
saturation in the presence of trapped gas according to:
(4)
SPE-195103-MS 7

Where Sor is the final residual oil saturation and Sor,min is the minimum residual oil saturation to water and
the residual oil saturation to gas.
It is documented that for most intermediate-wet to oil-wet conditions, the trapped gas saturation has less
impact in reducing the residual oil saturation. Areasonablematchwas obtained with this assumption. Data
reported from early experiments indicate low values of the modification parameter (a) in oil-wet samples,
indicating minor impact of trapped gas saturation in reducing the residual oil saturation in oil-wet systems
(Kralik et al, 2000).
Gas Phase Hysteresis Water Saturation Threshold: This parameter sets a threshold water saturation
above the connate water saturation (Krw = 0) where the secondary and subsequent drainage curve will
be reduced by the (α) parameter described above. When water saturation is below the threshold plus the
connate water saturation, gas phase hysteresis will follow the input drainage curve. For history matching,
it is assumed that this threshold value is 0.001, which is the default in Eclipse. This implies that the gas
drainage curve is reduced by the (a) parameter when water saturation is increased above the connate water
saturation.
The gas drainage curve was experimentally obtained from the unsteady state gas flood experiment of the
Composite-1 core. Since the unsteady-state method covers limited gas saturation ranges, the oil-gas drainage
relative permeability curve was extended until the residual oil saturation at the sample's Swc is reached.
Figure 2 shows the experimentally obtained gas relative permeability curves. The results from steady-state
core flood experiments on core plugs from the Mauddud B formation is also displayed for comparison. The
Krg and krog curves are generated analytically using the LET model to match the experimental data. The pre-
IWAG gas flood sequence was perfectly matched using these drainage curves; hence, they were not altered
during history matching.

Figure 2—Gas-Oil (Left) and Oil-Water (Right) Relative Permeability Curves Sets From Experimental Measurements

Hysteresis in Water Relative Permeability. The water phase relative permeability curves are inferred from
the partial unsteady-state water flood experiment conducted on the Composite-2 sample. A total of 1.1
PV of water was injected resulting in a remaining oil saturation of 61.2%. Oil-water relative permeability
was calculated from results of the unsteady-state experiment using the JBN method; however, since the
results are from a partial-flood experiment, the end-points of the relative permeability curve could not be
determined explicitly.
The relative permeability from the unsteady-state experiment is shown in Figure 2. Steady state
experiments conducted on core plugs from the Mauddud B formation were also available and are shown
in Figure 2. These core plugs have effective oil permeability ranging from 6-15 mD. The history matching
water imbibition and oil-water drainage curves are inferred from these experimental results. During
8 SPE-195103-MS

numerical simulation, corresponding curves were generated using the LET model. It is important to extend
the water saturation range beyond 1-Sorw to Sw = 1 to account for water saturation ranges in the reservoir.
Killough's hysteresis model was used to model the reduction of water relative permeability in three-phase
flow. Unlike the gas phase hysteresis treatment using the WAGHYSTR keyword, there is no equivalent
to Land's parameter to describe the trapped water saturation during the water drainage process. Killough's
model requires an imbibition curve going from Swc to Sw = 1 and a drainage curve beginning from a certain
trapped water saturation going to Sw = 1. Since there is no experimental result to infer the water drainage
curve, the curvature and trapped water saturation endpoints were made history matching parameters. It
should be noted that the WAGHYSTR keyword is also able to generate water phase hysteresis. In this case,
the hysteresis curves are generated by interpolating between the provided two-phase curve and reduced
three-phase imbibition curves.
Oil Relative Permeability. Hysteresis in the oil phase is modeled using Killough's model. The oil-water
and oil-gas two-phase drainage and imbibition curves need to be provided. In addition, since the oil relative
permeability is a function of water and gas phase saturations, the three-phase oil relative permeability is
generated using a modified version of the STONE1 three-phase relative permeability model. This option is
selected by specifying the STONE1 keyword in Eclipse. This allows reduction of the residual oil saturation
in three- phase flow conditions. In addition, the dependence of residual oil saturation on the trapped gas
saturation can be represented as described in Equation 4. The STONE1 algorithm uses the minimum of the
Sorw and Sorg to calculate the residual oil saturation in three-phase flow. Oil recovery was sensitive to the
residual oil saturation values.
The two-phase oil drainage and imbibition curves were inferred from experimental results when available
and generated using the LET model during history matching whenever experimental results were missing.
The oil-gas drainage curve is obtained from the experimental results of the Composite-1 gas flood
experiment.
The oil-water curve is inferred from the water flood experiment on the Composite-2 sample. Figure
3 shows a comparison of the two-phase krow and krog curves obtained from oil-water and oil-gas flood
experiments. As shown in Figure 3, the individual core plug Krog curves fall slightly below the Krow curves.
This is also seen from the experimental Krog curves obtained from Composite-1 and the experimental Krow
(JBN) curves obtained from Composite-2. It is noticed that the Krow curve from Composite-2 declines faster
as water saturation increases (Sw > 40%).

Figure 3—Krow and Krog Curves from Composite-1 and Composite-2 Samples. Results from a
steady state experiment using core-plugs from the Mauddud formation are shown for comparison.
SPE-195103-MS 9

Discussion of Lab Simulation Results

Table 4 and Figure 4 show the parameters and relative permeability and capillary pressure curves used
to history match the core flood experiment. The history matching was carried out by parameterizing the
unknown curve models and hysteresis parameters described in the previous sections. It was necessary to
have closed loops in all the curves sets for the hysteresis effects to function properly. The history matching
results were highly dependent on the hysteresis and three-phase models.

Table 4—Core Flood History Matching Parameters

Relative permeability curves L E T

Krg, D 0.98 6.3 3.2

Krg,I 3.5 -- --

Krw,D 2 1 0.6

Krw,I 2.5 0.6 3

Krow,D 2.5 2 3

Krow,I 8 4 2

Krog,D 7 -- --

Krog,I 2.5 -- --

Residual Oil saturations

Sorg, Residual oil saturation to gas 0.15

Sorw, Residual oil saturation to water. 0.2

Hysteresis and Three-phase oil relative permeability options

Water phase hysteresis Killough's model

Gas phase hysteresis WAGHYSTR options

C =3.12 a = 1.2 Sgt, trapped gas saturation = 23%

Three-phase oil relative permeability STONE1

Figure 4—Saturation path showing the injection sequences. Gas injection was followed
by IWAG in Composite-1 and water injection was followed by IWAG in Composite-2
10 SPE-195103-MS

As stated above, the unsteady-state gas flood experiment was followed by a WAG injection sequence in
Composite-1. Similarly, Composite-2 was partially water flooded before the WAG sequence was started.
The evolution of the fluid saturations was monitored at the injection wells. Figure 5 shows the saturation
path traversed during the primary flood gas and water flood and WAG sequences in the Composite-1 and
Composite-2 models, respectively.

Figure 5—Core flood history matching relative permeability and capillary pressure curves
Composite-1               Composite-2

The gas relative permeability scanning curves are calculated with the three-phase hysteresis model
provided by the WAGHYSTR keyword. The scanning curves generated from this model are shown in
Figures 6 and 7. The scanning curves show that the drainage and imbibition curves are irreversible. The
subsequent scanning curves are reduced significantly by the Krog reduction factor (α) in the presence of
water compared to drainage process in two-phase oil-gas displacement. As the water saturation increases,
SPE-195103-MS 11

gas mobility will be further reduced. Reduction of gas relative permeability and gas trapping with increasing
gas saturation results in increased oil recovery.

Figure 6—Gas relative permeability scanning curves obtained using the three-phase hysteresis model (WAGHYSTR)

Figure 7 shows the scanning curve generated using Killough's model for the water phase hysteresis. The
history matched oil cumulative and pressure for each composite core are shown in Figure 8.

Figure 7—Water relative permeabilitycurves generated from input imbibition

and drainage curves using Killough's two-phase hysteresis model.

Figure 8—Cumulative oil production and pressure history matching results of the core composite data
12 SPE-195103-MS

IWAG - Sector Model Simulation

Relative permeability and three-phase hysteresis parameters determined by history matching the production
profiles of the core flood experiments were used to run similar injection sequences on a sector model of the
Mauddud reservoir. The objective of the sector model simulation was to observe if the recovery benefits of
the three-phase hysteresis models and parameters observed in the core-scale model simulation continue to
play a role in a complex geological model.

Sector Model Description

The Mauddud reservoir is a carbonate reservoir characterized by complex geology with multiple faults.
The reservoir contains a thin oil rim with an overlying gas cap. The reservoir produced through continued
crestal gas injection since its discovery in the 1930's. The gas injection formed a secondary gas cap which
eventually reached the producing wells in 1965. The geometry of the model is an anticline trending in the
North-South direction.
The sector model which was cut from the Bahrain Field model has 21 layers with thicknesses that ranges
from 0.5 – 15 ft. Permeability ranges from 10 – 1200 md and porosity ranges from 0.20 – 0.34 (average
~28%). Figure 9 shows histogram ofthe model permeability and porosities. The Kv/kh ratio is assumed to be
1. The maps show that the crestal region has higher porosity than the oil rim region indicating better rock
quality in the crest than in the oil rim.
The continuous gas injection has recovered most ofthe oil from the crestal region, leaving small remaining
oil saturations, especially in the Mauddud A interval. The down-dip pattern has high water saturation and
a thin oil rim region.

Figure 9—Porosity and permeability distribution of the sector model

The model has a pore volume of 103 million reservoir barrel. The original gas-in-place is 12.2 BSCF, of
which ~60% is in the up-dip pattern. The model is characterized by at least five (5) rock types, however,
for this model only one rock type was considered. The hysteresis phenomenon is highly dependent on the
endpoints of the relative permeability curves which, in this case, are assumed to be the endpoints from
the core flood simulation results. Changing the endpoints will have a significant impact on the properties
obtained. The relative permeability curves are similar to those described in the core flood experiment,
however a lower connate water saturation of 5% has been assumed and the relative permeability and
capillary pressure curves have been adjusted to accommodate this low water saturation.
The initial conditions in the sector model represent the reservoir fluid and pressure distribution as of
January 2013. The sector model properties are summarized in Table 5.
SPE-195103-MS 13

Table 5—Properties of the IWAG Sector Model.

Model Properties

Top of Reservoir Depth, ft 1,805 ft

Thickness of Reservoir, ft 100 ft

Discretization I=150,J=17,K=21

Grid Dimensions Along X,Y:50-130 ft: Along Z:0.5-15 ft

OOIP 32 MMSTB

Gas-in-place 7.4BSCF

The PVT properties were obtained from a recent PVT test done on a recent samples taken in the Mauddud
reservoir.
Relative permeability and capillary pressure curves estimated from history matching the core flood results
were used. The curves were adjusted slightly to accommodate an assumed connate water saturation of 5%
for the sector model.

IWAG Patterns
Two inverted 5-spot patterns were considered to test the feasibility of the IWAG injection sequence. The
patterns are located in the up-dip gas-rich crestal region and in the down-dip region which has thin oil rim
underlain by a high water saturation zone. The model patterns are a little more than 40 acres with dimensions
of 1,600 by 1,130 ft. The producer wells are 1,000 ft away from the injection wells. To avoid possible
interference between the patterns, injection rates are adjusted when considering individual pattern floods.
Figure 10 shows the fluid saturation distribution in the model. The fluid saturations in the patterns are
also indicated. This figure shows that the continuous gas injection has left large gas saturation in the up-
dip pattern (Pattern-1). The down-dip pattern (Pattern-2) has a thin oil rim and higher water saturation. The
gas saturation in Pattern-1 is 30% compared to 3% in the down-dip Pattern-2. The water saturation in up-
dip Pattern-1 is 20% compared to 43% in down-dip Pattern-2. The oil saturations are 50% and 54% in
Pattern-1 and Pattern-2, respectively.

Figure 10—Fluid distribution and location of pattern injection wells. The crestal region
has higher gas saturation and the down-dip pattern has higher water saturation.

The following injection sequences were considered to test the performance of IWAG in the Mauddud
sector model:
1. A base case water injection
2. A base case gas injection
14 SPE-195103-MS

3. An IWAG sequence in both patterns

The well constraints and targets are provided in Table 6. In all injection sequences, the two crestal gas
injectors are injecting gas at 4,000 MSCF/D to keep the reservoir pressure at approximately 950-1000 psi.
The pattern injection wells are set to inject gas and water for the base case gas and water injection sequences
and alternate during the IWAG sequences. The producer wells are constrained by a flowing Bottom Hole
Pressure (BHP) of 700 psi and the injection wells are limited by a maximum injection BHP of 1,500 psi.

Table 6—Well Target Rates and Constraints

Wells Parameters Constrains/Targets Units

Injector I-1 Gas Inj.TargetRate 4,000 MSCF/D

Injector I-2 Gas Inj. Target Rate 4,000 MSCF/D

Up-dip/Down-dip Pattern Water

Target Rate/Well 1,000 RESV/D(1000 STB/D)
Injectors

Up-dip/Down-dip Pattern Gas

Target Rate/Well 3,000 RESV/D (~960 MSCF/D)
Injectors

All Injector Constraint Maximum BHP 1,500 Psi

Producer Constraints Minimum BHP 700 Psi

Hysteresis and oil relative permeability three-phase modeling options were similar to the ones used for the
core flood simulation. It should be noted that there are several options available for modeling hysteresis and
oil phase relative permeability during three-phase flow in Eclipse. All the different options and combinations
of these options will result in different solutions with varying recovery from an IWAG flood. The models
and options selected during the core flood history match gave the best results that matched the experimental
data. The options used in the model are repeated here and are shown in Table 7.

Table 7—Hysteresis and Three-Phase Oil Relative Permeability Modeling Options

Hysteresis/Three-phase modeling options Model

Killough's Model
Water and oil hysteresis modeling
Given by EHYSTR option 7 in Eclipse

WAGHYSTR option.
With this option, the following parameters are used.
Gas phase hysteresis
C α 3-phase Threshold

3.12 1.9 2.5

Three-phase oil relative permeability Stone's First mode specified by STONE1 in Eclipse

Gas trapping during flow reversal (imbibition) is characterized by Land's gas trapping parameter (C). A
small Land's parameter will give the most trapped gas saturation and the highest oil recovery. As the Land
parameter increases, the trapped gas saturation will decrease and so does the recovery. We have assumed
a trapped gas saturation of 23% to estimate a Land's parameter of C = 3.12, assuming a connate water
saturation of 0.117. Gas trapping will reduce the overall gas mobility and reduce the GOR.
The gas phase hysteresis is controlled by the α-parameter in the WAGHYSTR option. This parameter
reduces the relative permeability ofgasinthepresenceofwater. During three-phase flow, the gas drainage
curve will not follow the input drainage curve during secondary drainage processes as suggested by Carlson's
and Killough's two-phase models. Instead, gas relative permeability is reduced by the factor α according to:
SPE-195103-MS 15

(5)

Where KrgDRA(Sg) is the gas relative permeability at the current gas saturation
KrgDRA(SgStart) is the input Krg at the start of secondary drainage process, and
Krgimb(SgStart) is the relative permeability at the start of the secondary drainage process.
The calculated drainage curve is also related to the ratio ofcurrent water saturation and the connate water
saturation which is raised to the (α) parameter. This implies that when using the WAGHYSTR option,
if water saturation is notatconnate saturation atthe beginning ofsecondary drainage, the input drainage
gas relative permeability curve will be reduced as defined by the above equation. If water saturation is
not increasing, i.e SwStart=Swc, the model is assumed to be in two-phase (gas-oil) and Carlson's model is
assumed to calculate the scanning curves from the input drainage curve and an imbibition curve which
is fully defined by Land's parameter. As a result, the input imbibition curve is ignored. The effect ofthe
(α) parameter can be delayed by increasing the connate water three-phase threshold parameter in the
WAGHYSTR option.
Since the initial water saturation in our sector model is greater than the connate water saturation we
assumed, we expect the gas mobility to be reduced during gas flood (primary drainage) before WAG
sequence is commenced. During WAG processes, subsequent drainage and imbibition cycles are reduced
as defined by the (α) parameter. It is important to carefully consider the impact of this parameter when
performing field studies as it greatly affects recovery. A history match ofa field IWAG performance will
help in estimating appropriate parameters for field application.

Pattern Flood Results

Base case water and gas injection simulation cases were made in both patterns. The well targets and
constraints are given in Table 8.

Table 8—Base Case Water and Gas Injection Well Constraints.

Wells Parameters Constrains Units

Injector I-1 Gas Inj. Target Rate 4,000 MSCF/D

Injector I-2 Gas Inj. Target Rate 4,000 MSCF/D

Up-dip/Down-dip Pattern Water

Target Rate/Well 1,000 RESV/D (1000 STB/D)
Injectors

Up-dip/Down-dip Pattern Gas

Target Rate/Well 3,000 RESV/D (~960 MSCF/D)
Injectors

All Injector Constraint Maximum BHP 1400 psi

Producer Constraints Minimum BHP 700 psi

The recovery factors of the sector model from the base case gas and water injection are shown in Figure
11. The recovery factor from the base case gas injection was 17% in Pattern-1 compared to 34% in Pattern-
2, while the recovery from the sector was 12%. The effect of hysteresis must have reduced the mobility
of gas in Pattern-2, resulting in the highest recovery. The high recovery from gas injection is attributed to
the mobility control of gas during the three-phase hysteresis in the presence of water and lower residual oil
saturation to gas (Sorg=0.15) compared to the residual oil saturation to water (Sorw=0.20).
16 SPE-195103-MS

Figure 11—Sector model performance of base case water and gas injection prediction runs.

The base case water injection recovery from Pattern-1 and Pattern-2 was around 12%. Additionally, due
to the low oil-water relative permeability at high water saturations, oil recovery was affectedby high water
cut and the producer wells liquid production constraints.

IWAG Pattern Injection

IWAG prediction scenarios were conducted in both patterns, Pattern-1 and Pattern-2. The producers and
injection well constraints are shown in Table 9. The WAG ratio and the WAG cycle are important factors that
affect the performance of an IWAG injection sequence. For base case consideration, we have assumed a half
year WAG cycle period and a 1:3 WAG ratio. Representative injection rates are given in reservoir barrels.

Table 9—Pattern IWAG Sequence Well Constraints and Targets

Wells Parameters Constrains/Targets Units

Injector I-1 Gas Inj.Target Rate 1000 MSCF/D

Injector I-2 Gas Inj. Target Rate 1000 MSCF/D

Up-dip/Down-dip Pattern Water

Target Rate/Well 1,000 RESV/D (1000 STB/D)
Injectors

Up-dip/Down-dip Pattern Gas

Target Rate/Well 3,000 RESV/D (~1000MSCF/D)
Injectors

Injection Constraints Maximum BHP 1400 psi

Max Gas Rate 2,500 MSCF/D

Production Constraints Max Liquid Target Rate 3,000 STB/D

Minimum BHP 700 psi

Injectivity in the down-dip pattern was restricted in some of the injection wells due to the high pressure.
The pressure increases during IWAG due to the effects of hysteresis and increasing trapped gas saturation.
It should be noted that since this is a closed boundary model with fixed volume, any additional injection
to the sector model tends to increase the reservoir pressure in the patterns. In any case, the average sector
model pressure was kept around 950 psi.
The performance of the IWAG prediction cases is shown in Figure 12 and 13. These figures show the
GOR, water cut, oil production rates, and recovery factor from each pattern. The results of the IWAG cases
are compared to results from the base case gas and water injection sequences. The results show that an
IWAG sequence in the up-dip region recovers 10% more oil than the water flood case and 3% more oil than
the gas flood case in Pattern-1 (Figure 12). The IWAG sequence significantly reduced the GOR from 32
MSCF/STB in the gas injection base case to 8 MSCF/STB.
SPE-195103-MS 17

Figure 12—IWAG Sequence and Gas Injection Performance in the Up-Dip and Down-Dip Patterns.

Figure 13—IWAG Sequence and Water Injection Performance in the Up-Dip and Down-Dip Patterns.

In the up-dip Pattern-1, the IWAG sequence was able to reduce the mobility of gas and displace more oil
by gas trapping mechanism (Figure 12). The recovery from the IWAG sequence was similar to the recovery
from the gas injection sequence soon after injection was started, and increases gradually as water saturation
increases during the IWAG sequence. This is due to the hysteresis and three-phase effects are realized as
water saturation increases.
In down-dip Pattern-2, the base case gas injection performed better than the IWAG sequence (Figure
12). This is due to the lower residual oil saturation to gas compared to residual oil to water, and the reduced
oil relative permeability in regions with high water saturation. The IWAG sequence shows a higher water
mobility resulting in a water cut above 90% restricting mobility of oil in Pattern-2. This indicates that
continued gas injection will perform better in high water saturation regions compared to the IWAG sequence.
This was also observed in the core flood results.
Similarly, performance of the IWAG sequence was compared to the performance of the base case water
injection. The results are shown in Figure 13. In both patterns, the IWAG sequence performed better than
the base case water injection. The water cut of the base case increased and stayed above 90% soon after
injection was started. The IWAG sequence reduced the water cut from 97% to 85% in Pattern-1 and from
99% to 93% in Pattern-2.
Figure 14 compares recovery from the base case water and gas injections simulation runs with the IWAG
prediction run. Pattern-1 recovery from base case gas injection was 18% compared to 21% from the IWAG
sequence. The water injection recovery was 11% in Pattern-1. Pattern-2 recovery factor from the IWAG
18 SPE-195103-MS

injection was 30% compared to 13% from the base case water injection, and the recovery factor from base
case gas injection was 35%.

Figure 14—Comparison of Recovery Efficiency of the IWAG Injection

Sequence and the base case Gas and Water Injection Sequences.

In general, the pattern flood results show that IWAG performs better in low water saturation regions
compared to gas flood cases. On the other hand, gas injection performs better in high water saturation
regions compared to IWAG injection. Water flood performs poorly compared to the gas and IWAG injection
sequences.

IWAG Pilot Evaluations

This project is to evaluate Immiscible Water Alternating Gas (IWAG) recovery method in the oil rim of
Mauddud reservoir. The IWAG process is considered for mobility control and to improve oil recovery. The
key mechanisms that affect the flow behavior during IWAG displacement include gas trapping, reduction in
gas mobility and to a lesser extent to reduce residual oil saturation. Therefore, and to produce from Mauddud
B with reasonable GOR, IWAG was proposed for gas management. The previous sections outlined the basis
of the simulation study which suggested that IWAG gives an incremental recovery over the gas injection
in a mid-dip area where oil rim has low water with high gas saturation. Hence, three pilot locations were
selected in different part of the Bahrain Field. The patterns location which are designed to cover east and
west flanks for Mauddud reservoir, which will help in pilot expansion if decided.
The simulation study concluded that IWAG resulted in about 3% incremental recovery above the gas
injection in a mid dip pattern. It also show that IWAG injection resulted in lower GOR and water cut
compared to the base case gas and water injections. More optimization was performed during the Bahrain
Field pilot trials in terms of spacing, WAG, ratio and period.
Three pilot locations were selected as field trial. Two 40-acre inverted 5-spot patterns, each having
four vertical producers and injector. The third pilot was selected as 10-acre line drive with two horizontal
producers with one horizontal injector. They are reviewed from geological, petrophysical, production, and
mechanical integrity point of view. The criteria followed in the pilot selection are listed below:
1. Confinement - no bottom water influx from Nahr Umr or from above from Wara.
2. Preference to a well that is between gas cap and waterflood.
3. Wells located in same fault block.
4. Relatively mid-dip wells.
5. Producing from Mauddud.
6. Area with generally high GOR producers.
SPE-195103-MS 19

WAG ratio was selected to be 1:3 (water: gas). The WAG period was designed to be from three to six
months. WAG ratio and period are optimization parameters and had been changed over the project life
according to pattern performance. In the fourth quarter of 2014, the three IWAG patterns were online.
Starting with the gas injection cycle in all the patterns, different responses in each pattern were recognized.

Pattern 1: North-East Vertical Inverted 5-spot Pattern: Pattern A-X1

The gas cycle was initiated in June 2014. From October 2014 to February 2015, the total pore volume (PV)
of gas injected in Mauddud B was estimated to be 67%. The changeover to the water cycle commenced in
February 2015. Since then, only 7% PV of water was injected. The IWAG producers started to see more
water production associated with decrease in gas production. The gas reduction is mainly driven by stopping
the gas injection. There is no clear evidence that the IWAG cycles helped in recovering an oil incremental.
It is clear that the oil production shows higher decline after introducing the water cycle (Figure 15). The
GOR incline trend is still holding up even after the gas and water cycles. The reasons of not realizing the
incremental oil are due to confinement issues and operational challenges which was confirmed by PLT. The
Production Logging Tool (PLT) confirmed that most of the injected gas was in Mauddud A and only 6%
in the target zone of Mauddud B.

Figure 15—Pattern A-X1 Performance

Pattern 2: East Flank Horizontal Line Drive Pattern: Pattern A-X2

Well A-X2 is a horizontal injector drilled in Mauddud B between two existing horizontal producers. The well
logs showed very good oil saturation and can be considered a producer. However, the well was completed
as an injector to evaluate IWAG recovery method in this 10-acre pattern.
The first gas cycle commenced in August 2014 for a month (Figure 16). During this month of gas
injection, the oil production in the pattern horizontal producers almost doubled. Changeover to water cycle
took place in September 2014, during which the incremental oil declined with stable oil production. More
20 SPE-195103-MS

water production was observed, which led to changeover to the next gas cycle. In February 2015, the second
gas cycle was initiated. During the second gas cycle, the oil production remained stable for two months,
however it started to decline after the increased in gas injection rate.

Figure 16—Pattern A-X2 Performance

Figure 17—Pattern A-X3 Performance

In April 2015, the second water cycle commenced with the aim to mask the increase in gas production.
From April 2015 to mid June 2015, the water cycle managed to reduce the gas production, however oil
SPE-195103-MS 21

production started to decline sharply. In August 2015, the third gas cycle was introduced to try to recover
the oil production. In November 2016, the WAG cycles was stopped to evaluate the cycles influence on the
pattern base decline. Due to operational issues during the changeover from gas to water and vice versa, it
was decided to stop the IWAG trials. In conclusion, it is observed that the pattern is responding to water and
gas injection in controlling the gas production during the water cycle and sustaining oil producing during
gas cycle.

Pattern 3: South-West Vertical Inverted 5-spot Pattern: Well A-X3

In this pattern, the gas injection cycle commenced in August 2014 and a total of 118% PV of gas was
injected during the first five (5) months. The decision of changing to the water injection cycle was taken
in late February 2015. Due to the late evidence of gas breakthrough in the pattern producers, the water
cycle was introduced after five (5) months, assuming it is an enough period for the gas to fill in the pattern.
Since February 2015, 13% PV of water was injected, which resulted in decreasing the gas production. It
is observed that the oil production continued to decline during the water cycle (Figure 18). Similar to the
first pattern, the GOR seems to have the same incline even after introducing the IWAG cycles. This is due
to confinement issues observed in the injector.

Conclusions
The following can be concluded from the core experiment, sector modeling and the field pilots in the Bahrain
Field Mauddud reservoir:
The following findings were concluded from simulation and history matching Mauddud core
experiments:

• The IWAG core flood experiment was successfully matched using the STONE1 three-phase oil
relative permeability model and hysteresis models.
• IWAG recovers more oil than primary gas injection and water injection.

• Gas flood provide the lowest residual oil saturation compared to a water flood. The IWAG sequence
recovers more oil following a water flood than when following a gas flood.
• Significant reduction of gas mobility is observed in the presence of water.

• Performance of the IWAG process is largely due to mobility control resulting from reduction of
Krg and Krw in three-phase flow.
• Trapped gas saturation is dependent on gas saturation and increases gradually to a maximum value
defined by Land's parameter (C).
• Three-phase gas hysteresis modeling coupled with three-phase oil relative permeability modeling
using the Stone1 correlation gave the best history match results.
• The residual oil saturation to water was higher than the residual oil saturation to gas.

The following findings were concluded from the sector simulation:

• IWAG recovery from the up-dip pattern (21%) was higher than the base case gas injection (18%)
and the base case water injection (11%).
• Gas injection performs better in the down-dip region (35%) compared to IWAG sequence (30%)
and water injection (12%).
• IWAG injection resulted in lower GOR and water cut compared to the base case gas and water
injections.
• IWAG performs poorly in the down-dip pattern where there is high water saturation. The low oil-
water relative permeability in this saturation ranges and high water mobility may have affected
22 SPE-195103-MS

effectiveness of the IWAG sequence. Gas injectivity was restricted by the rapid build-up of pressure
due to the IWAG process in this pattern.
• Water injection performs poorly in both patterns.

• Gas mobility reduction and the effect of gas trapping are the main reasons for the higher
performance of IWAG injection.
The following findings were concluded from the field trials:

• The IWAG objectives were met in reducing the GOR during the water cycle and sustain oil
production during the gas cycle.
• The sustainably of the oil incremental of the inverted 5-spot patterns is not as what was estimated
by simulation due to vertical confinement issues as observed in the production logs.
• The line drive pattern performance proof that IWAG cycles can control the gas and water phases’
mobility.
• Introducing gas in Mauddud B resulted in very good oil incremental.

• Introducing water injection to layered oil-wet reservoir is not recommended to control the phase
flow, since the water tends to flow easier in such system.
• Due to the challenges faced during the field trials, which included well integrity, reservoir
heterogeneity and changeover operational issues, the three IWAG pilot patterns were terminated
for further review.

References
1. AL-Muftah, A.E., Hameed, M., Parakh, S., Abdul-Latif, A. and Al-Anaisi, I.: "An Integrated
Surface, Wellbore, and Reservoir Gas Management Approach for the Mature Mauddud Reservoir
in Bahrain Field." SPE 183797 (March 2017).
2. Carlson, F.M.: "Simulation of Relative Permeability Hysteresis to the Nonwetting Phase," SPE
10157 (October 1981).
3. Land, C.S.: "Calculation of imbibition relative permeability for two- and three-phase flow from
rock properties." SPEJ, Trans., AIME243, 149–156 (June 1968).
4. Larsen, J. A. and Skauge, A.: "Simulation of the Immiscible WAG Process Using Cycle-
Dependent Three-Phase Relative Permeabilities," SPE 56475 (October 1999).
5. Killough, J.E.: "Reservoir simulation with history-dependent saturation functions." SPEJ, Trans.,
AIME 261, 37- 48 (February 1976).
6. Skauge, A., Dale, E. I.: "Progress in Immiscible WAG Modeling." SPE 111435 (October 2007).
7. Spiteri, E. J., & Juanes, R.: "Impact of Relative Permeability Hysteresis on the Numerical
Simulation of WAG Injection." SPE 89921 (January 2004).
8. Kralik, J.G., Manak, L.J., Spence, A.P., and Jerauld, G.R.: "Effect of Trapped Gas on Relative
Permeability and Residual Oil Saturation in an Oil-Wet Sandstone." SPE 62997 (October 2000).