SPE

Society of Petroleum Engineef'S

SPE 18422

Downloaded from http://onepetro.org/spersc/proceedings-pdf/89RS/All-89RS/SPE-18422-MS/3215450/spe-18422-ms.pdf/1 by Bandung Inst. of Tech. user on 31 December 2023

Compositional Modeling of High-Temperature Gas-Condensate
Reservoirs With Water Vaporization
by S. Bette, Mobil R&D Corp., and R.F. Heinemann, Mobil E&P U.S.
SPE Members

Copyright 1989, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Symposium on Reservoir Simulation in Houston, TX, February 6-8, 1989.

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, as presented, have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the
author(s). The material, as presented, does not necessarily reflect any position of the Society of Petroleum Engineers, its officers, or members. Papers
presented at SPE meetings are subject to publication review by Editorial Committees of the Society of Petroleum Engineers. Permission to copy is
restricted to an abstract of not more than 300 words. Illustrations may not be copied. The abstract should contain conspicuous acknowledgment of
where and by whom the paper is presented. Write Publications Manager, SPE, P.O. Box 833836, Richardson, TX 75083-3836. Telex, 730989 SPEDAL.

SUMMARY determination of well requirements, and in developing strategies to

improve product recoveries will be discussed.
An equation of state compositional model capable of simulating
water vaporization is presented. This approach treats water as a
component among other species in the hydrocarbon phases and INTRODUCTION
allows for mass transfer between the aqueous phase and the
hydrocarbon phases. The model is validated against single cell The Arun field is a large gas condensate reservoir located on the
PVT experiments. The impact of this formulation on the basic northern coast of Sumatra, Indonesia. The gas is produced under
compositional model equations and the solution of the nonlinear a gas cycling scheme to maximize condensate recovery and is taken
and linear system of equations will be discussed. The simulator from a limestone formation at a depth of approximately 10,000
incorporates a sophisticated well management scheme with a feet. The reservoir temperature at this depth is 352 • F. As the
model of a gas plant to evaluate performance under various facility pressure declines, the water content of the reservoir gas increases
constraints. significantly due to water vaporization at the relatively high
reservoir temperature.
The application of this simulator in the management of the Arun
gas-condensate reservoir in Indonesia is presented. The first phase The initial pressure of the Arun reservoir was 7100 psig and was
of this 5000 cell, 8 component, 50 well study involves establishing well above the dew point pressure. The initial stabilized
reservoir parameters by matching historical pressure performance condensate/gas. ratio was about 48 STB/MMscf of water-free gas.
of individual wells. Predicted total field wellstream composition Currently, wells produce 2700 MMscf/day of separator gas and
and cluster condensate gas ratios match historical data very well. 135,000 bbl/day of unstabilized condensate. The field contains ten
The accuracy with which this formulation simulates water gas injection wells which provide pressure support by re-injecting
vaporization is demonstrated by comparing predicted and 800 MM scf/day of separator gas. This lean gas changes the fluid
measured water vapor production. In the second phase, the composition within the reservoir, affects retrograde condensation
reservoir model is used to predict future performance. This and revaporization, and at breakthrough, lowers the condensate-
simulator has the unique capability of simulating gas processing gas ratio and product yields.
plant operation to calculate and predict product streams such as
LNG, LPG and stabilized condensate. This feature allows one to This paper presents modifications to the usual compositional
specify future demand for a product such as LNG. The equations to account for water vaporization in a simplified fashion.
specification is translated into exact individual well production The resulting mathematical model is then applied to simulate the
rates based on their produced well stream composition. Water historical and future performance of the Arun reservoir. Water
evaporation has a significant effect on reservoir performance and vaporization is shown to provide an additional energy source for
an even greater effect on tubing flow performance. The model the reservoir's production, and the effects of gas injection are also
predicts injection gas profiles, extent of retrograde condensation, discussed. An additional feature of interest in this study was the
subsequent revaporization by injection gas, and composition and production specifications in the prediction phase. These
saturation distributions across the reservoir. The utility of the specifications were given in terms of individual products from the
model predictions in the design of surface facilities, in the gas plant which services the field. The integration of the gas plant
material balances and the reservoir model are also presented.

References and illustrations at end of paper.

183
 COMPOSITIONAL MODELLING OF HIGH TEMPERATURE
2 GAS-CONDENSATE RESERVOIRS SPE 18422

THE MATHEMATICAL MODEL Using this expression in equation 4, we obtain

The thermodynamics and continuity equations which are solved in v (P-P )]

fa = fs e s .
compositional simulators are well-known [1,2]. These models
assume that mass transfer occurs between hydrocarbon phases
w w xp [
w
RT . . . . . . . . . . . . . . . . . . . ( 6)

presumed to be flowing in thermodynamic equilibrium. The

In this simplified case, the water vaporization is modelled by
equilibrium conditions may be described by equating the fugacity
solving two additional algebraic constraints. This approach was
of each component for the oil and gas phases. At high
tested by measuring the mole fraction of water in vapor phase at
temperatures, the equilibrium condition is more complicated since
declining pressures using the Arun reservoir fluid in a single-cell

Downloaded from http://onepetro.org/spersc/proceedings-pdf/89RS/All-89RS/SPE-18422-MS/3215450/spe-18422-ms.pdf/1 by Bandung Inst. of Tech. user on 31 December 2023

the liquid water can transfer to the hydrocarbon phases. This
PVT experiment. The experimental and modelling results are
strictly requires a three phase constraint as follows
shown in Figure 1 and are in excellent agreement.
v
£. i = 1, ... , N . . . . . . . . . . . . . . (1) The compositional model solves the above thermodynamic
l..
equations in conjunction with a set of species continuity equations.
The hydrocarbon balances are
where N is the number of hydrocarbon components used in the
fluid description. The three phase approach obviously doubles the
number of thermodynamic equations and unknowns in the model.
The solution of this expanded set of equations with a Newton-based
algorithm is not attractive. i 1, N - 1 . . . . . . . . . . (7)

In this work, we have assumed that the vaporization of water is not

The water balance becomes
influenced by any hydrocarbon components that are present in the
aqueous phase and that water is in equilibrium in all three phases.
The thermodynamic equations then simplify to
a
at [rp(pOSO~+pgSgyw+pwSw)]

1, ... , N + 1
v· (ToXwV~o+TgYwV~g+Tw~w) + qw. . . . . . . . . . . . . (8)

The total hydrocarbon balance completes the model

(2)

a
The expressions for f~ and fY are available elsewhere. at [rp(poSo( 1 -Xw)+pgSg( 1 -Yw)l
The fugacity of pure water is given by
V"(T 0 (1-~)V~ 0 +T
9 (1-yw)~ 9 ) + qBC
where
R'l' l.n
qi PoXiQo + PgYiOg
or equivalently qw P 0 ~Q 0 + PgYwOg + PwOw
qBC Po( 1 -Xw>Oo + Pg(l-yw)Og· · ·· ·· ····· ..
Jp [ pRT] Jp [ c pRT] (9)
= 0 s v - dP + p v - dP (3)
s The model thus solves N + 2 thermodynamic constraints and N + 1
continuity equations. This set of equations requires very little
The first integral provides the fugacity of saturated vapor and the increase to Jacobian generation in a Newton algorithm since only
second gives the compression correction for the condensed phase. two algebraic equations have been added to the usual
If one assumes the partial molar volume is independent of compositional model [ 1,2] .
pressure, the expression can be written as
The model employs a strongly coupled well formulation where
. . . . . . . . . . . . . . . . . . . (4)

where rps is the fugacity coefficient of the saturated vapor that

describes the deviation from ideal behavior. The exponential in and
the above accounts for the compression effects and is referred to as
the Poynting correction. rps can be obtained by applying the Peng- . . . . . . . . . . . . . . . (10)
Robinson equation of state to describe the fugacity of a pure
component. The result is where Kj represents an effective transmissibility for the well, for
phase j, and includes a skin factor to account for formation
fs damage or improvement, limited entry to flow, drainage area
effects, non-Darcy flow effects, etc. m(P) is the gas pseudo-
l.n 1/Js l.n w (Z-1) - l.n(Z-B)
p potential described by Al-Hussainy, Ramey and Crawford [3].
s The upwind finite difference scheme, Newton algorithm and the
l. [Z+2.414Bl iterative solver used to solve the above equations are described in
. . . . . . . . (5) references [1,2].
n z-.414B ·

184
SPE 18422 S. BETTE' AND R. F. HEINEMANN 3

RESERVOIR MODEL In the vertical direction, the reservoir is represented by six layers.
Earlier cross sectional simulation studies comparing both twelve
Geologic Description: and six layer models concluded that six layers were sufficient to
adequately account for vertical flow effects. A north-south cross
The Arun field is located in Aceh province, North Sumatra, section is shown in Figure 3. The thickness of each cell for each of
Indonesia. The gas-bearing reservoir, locally more than 1000 feet the six layers is provided by the geologic model.
thick, is a reefal carbonate complex of Early to Middle Miocene
age. It rests on Early Miocene Shale of the Bampo Formation and This three-dimensional model (48 x 17 x 6) results in 4896 grid

Downloaded from http://onepetro.org/spersc/proceedings-pdf/89RS/All-89RS/SPE-18422-MS/3215450/spe-18422-ms.pdf/1 by Bandung Inst. of Tech. user on 31 December 2023

is overlain by Middle Miocene Baong shales which are considered blocks. Each grid block is assigned an average value of porosity,
to be the source rock for the hydrocarbons. Current structural permeability, water saturation, thickness and depth of the top face,
and facies relationships suggest the complex was tilted south- based on the geologic description. Cells with non-zero porosity
southwest during a post-depositional tectonic event. yielding a positive pore volume are active; there are 2726 aCtive
cells.
The extensive core analysis indicates three main depositional
facies: Reef, Near Reef and Lagoonal. The Reef and Near Reef For the historical period the daily total well stream production
facies were formed by the growth, accretion and attrition of corals rate of each producer and daily injection rate of each injector is
and algae. These forms are still identifiable in the Reefal and Near specified as a function of time. The field separator pressure and
Reef facies rocks, which comprise mainly boundstone, grainstone temperature data are also specified as functions of time.
and packstone. The lagoonal facies consists predominantly of fine
- to very fine - grained lime mudstone. Reservoir parameters seem For the prediction period, LNG BTU demand is specified along
to be in part facies - dependent and hence it was desirable to with gas injection rates. The total well stream production rates are
determine facies distribution throughout the field. To this end, a adjusted in the model to meet the LNG demand. The surface
method was developed for identifying facies in uncored wells, on facilities model performing these calculations will be discussed
the basis of E-log character. later.

Detailed petrography study has shown that most of the reservoir Fluid PVT Characterization:
porosity is secondary produced by a number of diagenetic events.
Three principal types of porosity are recognized in the Arun The three parameter Peng-Robinson (P-R) EOS, as utilized in the
reservoir: vuggy, moldic and intercrystalline. Vuggy porosity is compositional simulator, was used to model the original in place
developed mainly in the reefal facies, whereas most of the moldic reservoir fluid. Modelling consists of determination of a
porosity occurs in the Lagoonal facies. Intercrystalline porosity, component breakdown, characterization of the heavy hydrocarbon
which accounts for about two thirds of the total, is present in both. plus fraction, and matching of experimental PVT properties.

Based on correlations between log-derived porosity and core The component breakdown for fluid modelling involved a number
permeability, data transforms were developed for each of the three of considerations. First, the significance of water vaporization
reservoir rock facies. requires that water be a component. Next, propane and butane
are products and thus require separate tracking, although it was
There are no identifiable barriers to gas flow in any direction not necessary to model the isomers for matching purposes. The
within the Arun formation. No laterally - extensive permeability quantity of nitrogen was small (less than 0.3 mole %) and could
stratification has been identified in the sense of zones separated by reasonably be combined with C02 . Lastly, it was necessary to split
impermeable shales or other tight layers, nor are there faults or the heavier hydrocarbon fractions into two groups (C 5 -C 10) and
other permeability barriers to horizontal flow. Subdivision of the C 11 +) to adequately characterize the heavy ends for the
reservoir into correlative zones is made difficult by the complexity condensate. Further splitting of the (C 5 -c 10) fraction into c 5 -C6
of varying facies and a wide range of permeabilities. By the and c 7 -c 10 fractions did not provide an improved
gamma ray log, however, reinforced with core data where characterization. Note that with the existing process facilities and
available, a zonal correlation was developed and extended over the conditions, almost all of the C 5 + component ends up in the
entire reservoir. Using this correlation the Arun reservoir was condensate. This component definition.. of the fluids results in an
subdivided into twelve layers and the facies distribution within eight component model. Table 1 details the component breakdown
each zone mapped. ·The correlation zones vary little in thickness of the fluid.
across the field and the reservoir is tilted generally to the south.
The first step in establishing the EOS parameters for this fluid was
The appropriate transform for each facies was then applied to map to match the dew point pressure. The dry (water free) gas dew
the permeability distribution. In addition, structure, isopach, point pressure of 4462 psia at 352 ° F was matched by adjusting the
porosity, porosity-feet, permeability and hydrocarbon porosity- binary interaction coefficients. The next step was to match the
feet maps were also developed for each layer. initial condensate gas ratio (CGR) ("'58"-66 bbliMMSCF at 1255
psia, 70 ·F) and the measured Z-factors. By adjusting the shift
Simulation Model Set Up: parameters, a reasonable match of both CGR and Z-factor was
obtained. The measured and the calculated Z-factors are plotted
The Arun field is areally described by a 48 x 17 grid as shown in in Figure 4 as functions of pressure.
Figure 2. Finer grids were chosen around the closely spaced
production wells to obtain better resolution. The orientation and Water vaporization in the Arun reservoir is so significant that it is
grid size were chosen based on reservoir shape and well locations. essential that this phenomenon is modeled accurately. The EOS
The smallest grid is 300m x 300m, i.e., approximately 984 ft x 984 calculated values are compared with experimental values in Figure
ft. 1. The match is excellent, considering the complexity of the phase

185
 COMPOSITIONAL MODELLING OF HIGH TEMPERATURE
4 GAS-CONDENSATE RESERVOIRS SPE 18422

behavior. Note the exponential increase in mole fraction of water The model-predicted unstabilized CGR is compared to actual data
vapor at pressures below 2000 psia. This will have a significant for each cluster and on a field average basis in Figures 10 to 14.
effect on field operations during the later stages of depletion. The This good match is obtained by using the parameters from the
initial water saturated (wet) composition of the fluid is listed in pressure match without further adjustment. Condensate and gas
Table 1. streams for each cluster are computed by flashing the cluster total
well stream production at the specified historical separator
Single Cell Material Balance Calculations: conditions. The condensate to gas volume ratio depends on the
produced fluid composition, pressure and temperature.
Overall material balance calculations using a single cell are used to

Downloaded from http://onepetro.org/spersc/proceedings-pdf/89RS/All-89RS/SPE-18422-MS/3215450/spe-18422-ms.pdf/1 by Bandung Inst. of Tech. user on 31 December 2023

estimate the original gas in place, to check consistency between The initial CGR match confirms the PVT characterization of the
measured pressure and production data, and to verify fluid PVT original reservoir gas. (Note that the early deviation observed in
description. Furthermore, the importance of water vaporization Cluster III CGR is due to lack of separator pressure and
on overall reservoir performance can be evaluated. temperature data.) The fluctuations in CGR are mainly due to
changes in separator conditions. However, the sustained decline in
A single cell depletion model with the approximate Arun field pore Cluster I and Cluster IV CGR is due to the changing composition
volume and fluid PVT description was set up and the depletion of the produced well stream. A comparison of the field average
process was simulated. P/z plots, both corrected and uncorrected, July 1986 produced well stream composition presented in Table 2
are shown in Figure 5. The P/z plot which had been corrected for indicates a very good match. The accuracy in the model-calculated
water vaporization yields a straight line that could be extrapolated CGR and and well stream composition to actual reservoir
to zero pressure to determine gas in place. Water vaporization performance validates the calculated compositional distribution.
provides an additional energy source, acting like a gas injector in
each cell of the reservoir. This phenomenon is equivalent to Figure 15 illustrates the match of historical water production
adding approximately 8 to 10% of additional gas to the obtained with the compositional model. The figure indicates that
productivity of the ,reservoir. the thermodynamic model employed by the compositional model
accurately accounts for the water vaporization effects.
The gas in place determined from the P/z plot corrected for water Calculations indicate that in 1987 the water saturation around the
vaporization effects was exactly equal to the initial volume of the injectors has decreased considerably and in some cases that area is
single cell, and the material balance at various stages of depletion completely dry. This was confirmed by cores from a recent
was also very accurate, thus proving a material balance closure replacement injector drilled close to an existing injector. Injection
within the compositional model. gas, which is dry (water free), continuously evaporates interstitial
water and reduces water saturation.

HISTORY MATCHING PAST PERFORMANCE Individual well inflow performance were established by adjusting
skin and non-Darcy coefficients to match measured flowing bottom
The history matching objectives were to validate the reservoir hole pressures.
geologic and fluid description, to determine the gas in place, and to
verify the pressure and compositional distribution within the
reservoir. This was achieved by matching historical performance PREDICTION OF FUTURE PERFORMANCE
data such as static bottom hole pressure, separator gas, condensate
and water production, flowing bottom hole pressure, and Reservoir simulators, in general, do not perform calculations
composition of produced fluids. The simulation procedure beyond field separator trains. They can only compute separator
involves imposing historical well production and injection rates for gas and condensate production rates. In future performance
each well and adjusting model parameters, such as permeability, predictions, gas production rates can be specified. In Arun, the
porosity (volume) and initial fluid characterization, to match primary product is liquified natural gas (LNG). Secondary
performance data. There were nine and one half years of products are liquified petroleum gas (LPG) and condensate. Here,
historical performance data. prediction of future performance is based on meeting projected
LNG and/or LPG demand. Specifically LNG demand is based on
A match of individual well shut-in bottom hole pressures (SBHP) total BTU/year with additional constraint on the quality or the
and the overall pressure distribution across the reservoir was heating value (BTU/SCF). Therefore, a method is necessary to
obtained with relative ease due to the excellent geological compute gas volume production requirements for specified LNG
description. To obtain a history match, the permeability demand projections. This conversion depends strongly on the
distribution had to be modified (especially by lowering the plant and field facilities operation. LNG plant simulation is often
permeability of layer 1 and 2), to match the north-south pressure performed using detailed process simulators. An iterative
gradient between the injectors and the producers. After these procedure between the presented reservoir simulator and such a
adjustments, minor porosity adjustments, particularly process simulator would have been extremely inefficient and
redistribution of volumes from the north and south to the west and impractical. Therefore, a field and plant facilities simulation
middle part of the reservoir, provided a good match. Lastly, model was constructed and coupled with the reservoir simulation
porosity and permeability distribution were fine tuned for model. This combined feature can then dynamically specify and
continuity and to refine the SBHP matches. A typical match for a allocate well production rates to meet the specified LNG BTU
production well, an injection well and an observation well are production.
shown in Figures 6, 7 and 8. Similar matches were obtained for all
the wells in Arun. Consequently, the calculated well rate averaged
reservoir pressure matches the field calculated average reservoir
pressure, as shown in Figure 9.

186
SPE 18422 S. BETTE' AND R. F. HEINEMANN 5

Field and Plant Facilities Simulation: values are shown in Figure 17. The gas production predicted by
the model meets the LNG requirement almost exactly. The sharp
The overall scheme was to determine and adjust production well variations in LNG production were chosen to exhibit the capability
rates to meet the specified LNG BTU production and heating value of the algorithm to dynamically adjust the well rates to meet LNG
while providing gas for plant and field fuel, national projects, and requirement. The calculated (predicted) total well stream rate
injection. In addition, LPG and condensate production was to be corresponding to the LNG specification is shown in Figure 18. The
calculated. The total gas production rate is mainly influenced by gas production profile follows the LNG profile closely. In general,
LNG, the primary product. LPG and condensate are by-products when LNG demand remains constant over several years, we see an
and are a consequence of LNG production. increase in total well stream production. This is because the

Downloaded from http://onepetro.org/spersc/proceedings-pdf/89RS/All-89RS/SPE-18422-MS/3215450/spe-18422-ms.pdf/1 by Bandung Inst. of Tech. user on 31 December 2023

produced well stream gas is becoming progressively leaner due to
Figure 16 is the flow diagram of the surface facility simulation. dilution by injection gas and increasing C0 2 and water vapor
The produced well stream gas (1) is flashed in the field separator, concentration. The fact that we need more of the leaner gas to
at specified pressure and temperature, to yield separator gas (2), make the same amount of LNG is clearly evident.
unstabilized condensate (3), and water (22). A portion (4) of the
separator gas is processed in the Field NGL (natural gas liquids) The predicted pressure distribution, in general, indicates a
plant to extract LPG and condensate (10). The lean gas (7) from pressure sink around the cluster I and II production wells located
the NGL plant is used for field fuel (8) and injection (9). A major in the northern side of the field. As the reservoir is depleted,
portion of the separator gas (2a) and unstabilized condensate (3) is reservoir pressure declines with the same general pressure
pipelined to the plant. A part of the separator gas supplies gradient. Higher pressures exist in the southern end of the field
national projects (5). Then gas (6) is combined with condensate (3) due to higher injection in that area. The average P/z relationship
and processed to produce LNG, LPG and stabilized condensate. with cumulative recovery and (Piz>corr. corrected for the effects of
water vaporization are shown in Figure 19. Notice that the actual
In this simulation, units T 1 , T 2 , and T 3 serve the purpose of uncorrected P/z curve shows the additional energy due to water
separating individual components and then constituting them to vaporization as it curves upward during later stages of depletion.
form various products. T 1 splits c 1 and c 2 (12) from C 3 + (13). Upon correction for this energy source, we obtain a P/z
Some c 1 is taken from (12) to provide for plant fuel (16). The representation which is essentially linear.
remaining C 1 and c 2 (15) are used for producing LNG. The field
NGL (10) is added to the C 3 + stream (13). T 3 separates C 5+ and The migration of injection gas in the reservoir, and consequent
C 11 +from (14) to make stabilized condensate (21), which also sweep and dilution of reservoir gas, has a significant impact on
contains small amounts of c 4 . Stream (19) contains c 3 and c 4 reservoir performance. Therefore, it is important to clearly
which make up LPG. A portion of this c 3, c 4 stream is spiked (18) establish injection gas profiles in the reservoir. We need a method
into LNG as required to make up the specified heating value of to determine these profiles without actually tracking a tracer
LNG. The remaining c 3 , C 4 forms LPG (20). component. In lean gas condensate systems such as this, saturation
pressure or dew point pressure is a good measure of compositional
In this scheme, the primary specification is the yearly LNG BTU changes due to dilution by injection gas. The decrease in dew
production (17) and its heating value. Other specified streams are point pressure is a function of the amount of injection gas in the
plant fuel (16) (BTU/yr), national projects (5) (MMSCF/D), field mixture. Therefore, by plotting saturation pressure distribution
fuel (8) (MMSCF/D), and injection gas (9) (MMSCF/D). These across the reservoir, one can obtain a reasonable map of injection
specifications are translated to well stream (1) production in gas distribution. It should be mentioned, however, that utilization
MMSCF/D. The LNG (17) heating value requirement determines of dew point pressure to indicate injection gas after the system has
the amount of C3, C4 spike (stream 18). Plant fuel (16) is mainly formed two hydrocarbon phases is not as accurate as for a single
made up of C 1 and inerts such as nitrogen. Plant fuel requirement phase. The profiles indicated that injection gas migrated
is a function of LNG and LPG production and is also specified. preferentially in the high permeability layers. In addition, gravity
override was also evident. These profiles clearly established the
In this surface process simulation scheme LNG BTU production, unswept areas and also indicated causes of injection gas
LNG heating value, plant fuel, field fuel, injection gas, and breakthrough. This information is vital for future development
national projects' requirement are specified yearly. At the strategies.
beginning of the prediction period initial well rates are allocated
among the producers. The algorithm adjusts these rates monthly Injection gas minimizes condensate drop out and further aids in
to meet the yearly LNG BTU and heating value demand. The well revaporization of the condensate. When the reservoir pressure
rates are increased if the BTU production rate is lower than falls below the dew point pressure, condensate drops out in the
required to meet the target, and vice versa. The shortfall or excess reservoir, especially in the areas which have not been swept by
BTU production during a given month is prorated to the following injection gas. Condensate drop out significantly affects condensate
month and rates adjusted accordingly. In almost all cases, the production. The stabilized condensate production rate is shown in
LNG demand was met exactly, even for different production Figure 20. The good match during the historical period reiterates
allocation schemes. This scheme also allows for initiation of new the accuracy of the compositional model along with the surface
wells, abandonment of wells and new allocation among all facilities model. The observed decline in condensate production is
production wells. Coupling this algorithm with the numerical due to two reasons: one, condensate drop out in the reservoir, and
scheme of the compositional simulator is very complex. two, injection gas breakthrough. This compositional information
is crucial for optimizing future production profiles.
Performance Predictions:
As pressure declines, interstitial water evaporates and water
For purposes of discussion in this paper, future performance saturation declines. This process is accelerated by the injection of
predictions were performed based on hypothetical LNG dry (water free) gas. Around injectors and along injection gas flow
requirement. LNG specification along with model-calculated streams we observed a significant reduction in Sw and some parts

187
 COMPOSITIONAL MODELLING OF HIGH TEMPERATURE
6 GAS-CONDENSATE RESERVOIRS SPE 18422

were completely dry. Further, due to low initial water saturations, account for water vaporization in high temperature gas reservoirs.
coupled with decreasing water saturation with depletion, there is The formulation results in two additional algebraic constraints
little or no free water production. The produced water is a over the usual compositional equations. The accuracy and the
consequence of water vapor production along with gas production. utility of the model is demonstrated by applying it to the Arun gas
Water production rate is shown in Figure 21. The rate increases as condensate reservoir in Indonesia. The model provides an
gas production increases and increases further even when gas excellent history match of past performance. To predict future
production is fairly constant because of increasing water performance, the model's well management scheme is coupled with
vaporization with declining reservoir pressure. The water vapor a surface facilities model which enables one to predict various
content in the produced gas as a function of average reservoir product streams. This coupling facilitates the evaluation of

Downloaded from http://onepetro.org/spersc/proceedings-pdf/89RS/All-89RS/SPE-18422-MS/3215450/spe-18422-ms.pdf/1 by Bandung Inst. of Tech. user on 31 December 2023

pressure approximately follows the profile shown in Figure 1. We reservoir performance under various product demand scenarios.
observe an exponential rise in water mole fraction below 2000 psia. In addition, alternate reservoir management strategies to optimize
The predicted water production rate is crucial to the design of one or more of the products can be evaluated on a regular basis.
water handling facilities in the field. Water vaporization in this high temperature gas reservoir has a
significant effect on future performance. It is equivalent to an
The predicted yearly LPG production profile is shown in Figure additional 8 to 10% of original gas in place. It is shown that the
22. The general decline in LPG production is a consequence of accurate prediction of water vapor production is crucial in the
declining C3 and C4 components in the total well stream. This design of future facilities requirement. At the later stages of
profile is also impacted by the constraint on LNG heating value. depletion, the water vapor content increases rapidly, and has a
Since the produced well stream composition declines with time, significant effect on tubing flow performance. Rigorous tubing
more and more c 3 and c 4 are needed to spike (stream 18 in flow calculations have enabled us to calculate ultimate recovery
Figure 16) LNG for maintaining its heating value. factors. The model predictions for 1987 based on specified LNG
production are within 4% of actual measured values; thus are
The recovery of individual components as a function of reservoir within the limits of measurement accuracy. The versatility of the
pressure can be calculated. The overall dry gas recovery as a compositional model in concert with the facilities model as a
function of average reservoir pressure is shown in Figure 23. The reservoir management tool is clearly demonstrated.
simulation also included the determination of well performance
and its affect on the overall recovery of hydrocarbons. The
procedure involves establishing the inflow performance of all the NOMENCLATURE
production wells followed by tubing flow performance evaluation
using a 3-phase, fully compositional flow calculation. Tubing flow A,B Peng-Robinson equation-of-state
performance calculations showed that the produced water vapor parameters
can condense when pressure and temperature fall as the gases rise fj fugacity of component i in phase j
I
in the tubing. This has a significant effect on pressure drop in the k absolute permeability (mD)
tubing. The effect of tubing flow and associated pressure drops on krj relative permeability of phase j
recovery factor is also shown in Figure 23. For any specified K· effective well transmissibility of phase j
tubing head pressure, the ultimate recovery can be determined, mj(P) gas pseudo potential
and the corresponding average reservoir pressure can be obtained N number of hydrocarbon components
from the same plot. p pressure (N/m2 )
ph flowing bottom hole pressure (N/m2)
In this study, 1987 was part of the prediction period. At present, qi ith component molar injection/production rates
we have the opportunity to compare model predictions to actual (kgmoles/s)
production. For specified LNG production, the calculated total volumetric injection/production rate of phase j
well stream gas production and stabilized condensate production (m3/s)
are within 4% of actual measured values. Thus they are within the gas constant
limits of measurement accuracy. saturation of phase j
time (s)
The coupling of the facilities simulation to the reservoir simulation temperature
model enables one to determine future performance for a specified transmissibility of phase j
product demand and involves a single simulation run for the entire partial molar volume
prediction period. Thus, for different product demand scenarios, molefraction of component i in the liquid phase
reservoir performance can be determined relatively easily. In molefraction of component i in the gas phase
addition, production profiles of secondary products can be compressibility factor
calculated. On the other hand, for a given LNG demand scenario, density of phase j (kgmoles/m3)
alternate reservoir management strategies to improve production porosity
of LPG and condensate can be studied in a time and cost effective fugacity coefficient of the saturated vapor
manner. The model is presently being used on a regular basis to flow potential for phase j
evaluate different reservoir management strategies in terms of well gravity potential
future facilities requirement, well requirement, optimal location of
wells, gas injection strategy, economic feasibility, etc. Finally, an Subscripts
optimum depletion strategy will be developed.
0 oil phase
g gas phase
CONCLUSIONS w water phase
s saturated vapor
We have presented a compositional model which will accurately

188
SPE 18422 S. BETTE' AND R. F. HEINEMANN 7

SuQerscriQts

v vapor
L liquid
TABLE 1
a aqueous
s saturated vapor Reservoir Fluid Composition

ACKNOWLEDGMENTS

Downloaded from http://onepetro.org/spersc/proceedings-pdf/89RS/All-89RS/SPE-18422-MS/3215450/spe-18422-ms.pdf/1 by Bandung Inst. of Tech. user on 31 December 2023

Composition
The authors wish to thank Jack L. Harper for his contributions Component Drx Wet
and his dedication to the development of the model. Thanks to
Thor Sutan-Assin for his contribution and participation in the H2 0 0 4.06
Arun Simulation study. We would like to thank Rafi Al-Hussainy C0 2 14.89 14.29
and Aziz S. Odeh for their advice and encouragement. Also, cl 71.50 68.6
thanks to the Reservoir Management team at Mobil Oil Indonesia c2 5.61 5.38
and the Simulator Development group at the Dallas Research c3 2.58 2.48
Laboratory for their assistance during the study. Finally, we wish c4 1. 47 1. 41
c5-C1o 3.14 3.01
to thank Mobil Research and Development Corporation for
C11+ 0.81 0.78
permitting us to publish this paper.

REFERENCES

1. Harper, J.L., Heinemann, R.F., Ray, M.B., and Stephenson,

P.E., "A Compositional Simulator for Performing Large Field
Studies in a Vector Computing Environment", SPE 13714,
Middle East Oil Technical Conference, Manuma, Bahrain
(1985).
TABLE 2
2. Bette', S., and Heinemann, R.F., "The Utility of Compositional
Simulators in the Analysis of Rich Gas Condensate Reservoir", Wellstream Composition Comparison
Proceedings; First Wyoming Sim. on Enhanced Oil Recovery,
Laramie, Wyoming, March 1986. .
3. Al-Hussainy, R., Ramey, H.J., Jr. and Crawford, P.B., "The Lab Measured Predicted
Flow of Real Gases Through Porous Media", J. Pet. Tech., Component KOL % MOL %
May 1966.
N2 0.31 0.34
C02 14.06 14.60
Cl 72.65 71.82
C2 5.76 5.59
C3 2.50 2.54
IC4 0.59 0.62
NC4 0.76 0.81
IC5 0.41 0.40
NC5 0.27 0.30
C6+ 2.71 2.98

Total 100.00 100.00

189
 NW-SE STRIKE SECTION • CEU(X)•12
40 ~ _l ___ /r-- -- ,----

Downloaded from http://onepetro.org/spersc/proceedings-pdf/89RS/All-89RS/SPE-18422-MS/3215450/spe-18422-ms.pdf/1 by Bandung Inst. of Tech. user on 31 December 2023

0 32
e LAB MEASURED
~
,v<--r--
---r--r--
N
I
0
- PREDICTION
I/ I"- t--~" """-,
I--"
!
~

1- 24 ~r-.
z :::::f:::: r--
~::::
~
w ---r-- __ f::::~
0 r--~"
a::
w .._ r--....
~
r-......._ " N
Q_
16 r--......_
---~--
~--

::::~
~
~
0
2 8 '---r-- ---r--
r-."
0
0 1600 3200

PRESSURE, psia
4800

Fig. 1-Comparison of measured and calculated mole fraction of water in the vapor phase.
6400 8000
~
'\
I\_1\
~
L 1
: V1 m~I T 11 ·- - - - ·- ·- ·- ·- ·- - ·-
1- 1\ -f--
- Fig. 3-North-south cross section of Arun field.

:g -
\ -
).,!

~ ~
1.15
::::

. ;T
.......
=I/ 1.10

1.05
-- a::
0
1-
0 1.00
/
~
. -
--
<(
lL
I
N
~ - - EXPERIIVENTAL
0.95
'\ .. - 0 CALa.LATED

~ 0.90
~ =l
\ -_I/ 0.85
3000 4000 5000 6000 7000 8000

~ 7 PRESSURE. psia

I ( · v
1\J~ I illll~
Fig. 4-Comparlson of measured and calculated Z·factors.
l I I
Fig. 2-Areal simulation grid.
 Downloaded from http://onepetro.org/spersc/proceedings-pdf/89RS/All-89RS/SPE-18422-MS/3215450/spe-18422-ms.pdf/1 by Bandung Inst. of Tech. user on 31 December 2023
6500.------------------------------------------. 7500.--------------------------------------------,

-<::::, - - UNCORRECTED lU
6100 -<::::,
-<::::, "iii 6800
~ - - · CORRECTED c.
~
~
~
~
E
::1
lU
"iii
5700 "" ---...::
---...:: .......-
iU
"0
6100
c.
.......
.......
N .......
....... uj·
0: 5300 .......
....... (j) 5400
.......
.......
w 0 ACTUAL
([
....... MEASLRED
....... 0.
....... uj
4900
.......
.......
.......
w
([ 4700 CALCULATED

4500~-------L--------~------~--------L-------~ 4000~------~--------~------~--------~-------J
0 100.0 200.0 30QO 400.0 500.0 1977 1979 1981 1983 1985 1987
NET TWS PRODUCTION YEAR

Ag. 5-P/z performance of the slnglfH:ell depletion model. Ag. 7-Pressure match of an InJection well.

7500.------------------------------------------, 7500.------------------------------------------,

lU lU
"iii 6800 "iii 6800
c. c.

~ ~
E
iU
"0
6100 lU
"0
6100 0

uj uj
(j) 5400 t" o ACTUAL (j) 5400 0 ACTUAL
w
~
0.
MEASURED ([
0.
MEASURED
uj - - - CALCULATED uj
w w - - CALCULATED
([ 4700 ([ 4700

4000~------~------~--------~------~-------J 4000~------~--------~------~--------~------_J
1977 1979 1981 1983 1985 1987 1977 1979 1981 1983 1985 1987

YEAR YEAR

Ag. &-Pressure match of a production well. Ag. 8-Preseure match of an observation well.
 Downloaded from http://onepetro.org/spersc/proceedings-pdf/89RS/All-89RS/SPE-18422-MS/3215450/spe-18422-ms.pdf/1 by Bandung Inst. of Tech. user on 31 December 2023
7500 100

7000 80
u::
IU 0
·u; (f)
a. 2
6500 2 60
ui ::J
([
m
::J
(f)
(f)
m
w 6000 0 40
([ --ACTUAL
a. • ACTUAL i=
<(
([
MEASl.BED
- CALCULATED
5500
20f • CALCULATED

5000 0
1977 1979 1981 1983 1985 1987 1977 1979 1981 1983 1985 1987

YEAR YEAR

Fig. 9-Average reservoir pressure match. Fig. 11-Condensate gas ratio match of Cluster II.

:s

100 100

80 80
u:: u::

~
0 0
(f) (f)
2 2
2
::J
60 • 2
::J
60
m m
m m
0 40 0 40
--ACTUAL i= --ACTUAL

20~
i=
<( <(
([ MEASURED ([ MEASURED
20
• CALCULATED • CALCULATED

0 0
1977 1979 1981 1983 1985 1987 1977 1979 1981 1983 1985 1987

YEAR YEAR

Rg. 10-Condensate gas ratio match of Cluster I. Rg. 12-Condensate gas ratio match of Cluster Ill.
 Downloaded from http://onepetro.org/spersc/proceedings-pdf/89RS/All-89RS/SPE-18422-MS/3215450/spe-18422-ms.pdf/1 by Bandung Inst. of Tech. user on 31 December 2023
100 50

--ACTUAL
80 40 tvEASLRED
iL
0 e CALCULATED

~
(/)
~
60
~
_j
0
::J
30
m m
m m
I
0 40 2 20
;:::: --ACTUAL
<( MEASLRED
a:
20 e CALCLLATED 10

0 0
1977 1979 1981 1983 1985 1987 1977 1979 1981 1983 1985 1987
YEAR YEAR

Rg. 13-Condensate gas ratio match of Cluster IV. Fig. 15-Water vapor production match.

g::

FIELD AND PLANT FACILITY SIMULATION

100

NATIONAL
FIELD FUEL PROJECTS PLANT FU,EL
80
Ll
0
(/)
~ 53 ® I®
60
~
_j
m
m 1 ....., , sz LNG lc1,c 2 ,c3 ,c4,
0 40
;:::: --ACTUAL
<(
a: tv1EASlRED ) ~ LPG lC3,C4)
20
e CALCULATED

WELL STREAM
0 STABILIZED
WATER ' ~ CONDENSATE
1977 1979 1981 1983 1985 1987 (C4,Cs +,Cn+)

j
0
YEAR

FIELD PLANT
Rg. 14-Reld average condeneate gaa ratio match.

Fig. 16-Field and plant facilities model.

Downloaded from http://onepetro.org/spersc/proceedings-pdf/89RS/All-89RS/SPE-18422-MS/3215450/spe-18422-ms.pdf/1 by Bandung Inst. of Tech. user on 31 December 2023
 Downloaded from http://onepetro.org/spersc/proceedings-pdf/89RS/All-89RS/SPE-18422-MS/3215450/spe-18422-ms.pdf/1 by Bandung Inst. of Tech. user on 31 December 2023
100
- - Calculated

80 e ACTUAL

0 60
::J
m
m
I
:2 40

20
8000
0- -B- FESERVOIR
7000 PAESSLRE
1977 1983 1989 1995 2001 2007
--+-- n.BING
YEAR 6000 PAESSLRE

Ill
Fig. 21-Predlcted water vapor production. "iii
0. 5000
u.i
([
::J 4000
(f)
(f)
CD w
([ 3000
"" 0..
2000

1000

0
0.00 0.20 0.40 0.60 0.80 1.00
([
<( CUMULATIVE RECOVERY OF JGIP, %
w
>- Rg. 23-Calculated racovery factor curvaa.
(fj
z
0
!::::
(.9
0..
_j

1989 1991 1993 1995 1997 1999

YEAR

Fig. 22-Predlcted LPG production.