Oist~nguis ed Author Sertes

Proper Use a Equa fon 0

Sta for Com 051 I na~
Reservoir S~mulatlon
Peng Wang, SPE, and Gary A. Pope, SPE, U. of Texas at Austin

Su mm ary avaIlable EOS compositional simulators and in-house mod­

Equation-of-state (EOS) compositional reservoir simula­ els have been improved Significantly in terms of robustness,
tion is an accurate and powerful means to model compli­ effiCiency, and features . More frequent use of EOS composi­
cated phase and flow behavior involved in the displacement tional simulators should be expected in the future
of oil and gas in porous media. The use of EOS simulation However, uncertainties still exist that are associated with
becomes even more effective when the process involves sol­ the use of an EOS to model petroleum reservoir fluids
vent injection for miscible or near-miscible displacement of because of the complicated nature of such fluids and limi­
crude oil in reservoirs . The current trend of combining the tations of the most commonly used EOS's. Tuning the
simulation of reservoirs and surface facilities increases the adjustable EOS parameters to match experimental fluid
demand for the use of EOS compositional models. PVT data is a common practice. The proper use of OS
One ke use of 1 simulators
an phase behavior research topic since the 1970's. Although significant
of the fluids. This paper is a summary of experiences and progress in this area has been made , engineers still face sev­
developments in the use of EOS's for compositional reser­ eral challenging questions: How many fluid components
voir simulation studies. The reader is referred to the recent should be used 7 How much experimental data and what
monograph by Whitson and Brule 1 for a more complete kind of data are needed to tune an EOS7 How good is the
treatment of the volumetric and phase behavior of petrole­ predictive power of an EOS that matches all of the experi­
um fluids and the use of EOS's to model these fluids. mental data available? 'vVhat are limitations of the com­
monly used EOS's for hydrocarbons when the gas or oil
Introd ucti on contains Significant concentrations of components other
When the displacement process depends on pressure and than hydrocarbons? Which EOS should be used when
fluid composition, an EOS should be used to simulate the more than one option is available in a reservoir simulator?
equilibrium mass transfer between phases and the pres­
sure/volume/temperature (PVT) behavior of the fluids. PVT Representative- Sa mple Co llection
laboratory measurements are made on only a small portion The accuracy of any EOS model depends on the quality of
of the composition path encountered throughout the dis­ th,e laboratory PVT data and procedures used to obtain the
placement, whereas an EOS can be used to predIct behavior EOS parameters. The value of the PVT data depends on the
for the entire composition path and pressure range of the quality of the fluid samples. They must be representative of
process. Many field development projects have strong com­ the reservoir fluid Three common methods are used to
position dependence, such as production from gas/conden­ sample the reservoir fluid.
sate reservOirs, miscible and near-miscible gas injection, or
water-alternating-gas injection for enhanced oil recovery. Downhole Sampling. A sample collector with a capacity of
Until very recently, industry practice was to use a black-oil several hundred cubic centimeters is used to obtain a fluid
model and/or its modified versions 2 ,J or a limited-composi­ sample from within the wellbore. The downhole flOwing pres­
tion reservoir simulator 4 ,5 to approximate the complicated sure at the sampling point must be greater than the saturation
phase behavior. These Simplified approaches require less pressure of the fluid so that the sample is a Single phase At
computer time and memory and can provide a quick pre­ leasLthree fluid samples should be collected to detennine the
liminary evaluation of reservoir performance. However, appropriate fluid sample for the PYI measurements
efforts by industry and universities have led to rapid
advancemfnts in parallel-computer hardware and software WeUhead Samplmg. A fluid sample can be collected
during the last several years, which have increased simula­ directly from the wellhead The fluid needs to be In the sin­
tion efftcien cy considerably6-9 and now make composition­ gle-phase region at wellhead flowing conditions. Normally,
al simulation practical Also, most of the commercially the sample needs to be taken at a pressure of at least
100 psi greater than the bubblepoint to be valid gain,
Copyrigh t 2001 Society of Petroleum Engineers
taking multiple samples is recommended.
This is peper SPE 69071. Distinguished Author Series orod e s c f~ ge neral, dcscrtptive
r ~preS"e .,tohor'\S Ihot summarize Ihe stole of lhe art In 011 area 01 technology by describin
reaml deve!opme n.ls for r'aoders who ore noJ specmlists in If, e topics discussed Writte n by Separator ampling. A fluid sample is obtained by recom­
IrrdtVlduals recogni'Z.ed os experts In the: area, hese or1I1:les pro vide key re.erences- to moi'~
definitive wo, k o.nd p,esenl specific de.lails only 10 Ulusl rolt: the technology, Purpose: to
bining the gas and liquid taken at a surface separator at the
inform Ihe generol readership of recent odvo nces ,,"I various areas of pctrote..rn engineering. measured gas/oil ratio (GOR) Usually, GOR's are measured at

JULY 2001
 dewpoint pressure. The
5,5oo,-- -- - - - -- - -- - -- - -- - -- - - - - - - - ------. sampling well needs to
produce fluid steadily
5,000
- 8-component EOS
for a long enough peri­
4,500 od of time and at a rate
- 5-component EOS sutTicie 11 tly high to carry
4,000
I'll
-12-component EOS any condensate in the
'[ 3,500 wellbofe to the surface.
Q)
~ 3,000
III
(I)
Labo ratory PVT
l!: 2,500 Expe riments
11.
2,000 The appropriate PYT
experiments are related
1,500 closely to the processes
1,000 implemented in the field
The primary purpose of
500
conductmg PYT experi­
o~-~~~~~-~-~-_-~-~-_-~-~-~--~~---~ ments is to generate reli­
-400 -300 -200 -100 0 100 200 300 400 500 600 iVO 800 900 .1,000 1,100 able fluid properties for
Temperature, OF use in reservoir simula­
tion and the evaluation
Fig. J-Effects of number of components on phase envelope of an actual gas condensQte. of the fluid properties.
The most commonly
availa ble PYT experi­
40 ments are the following.
- 8-component o Constant-mass ex­
CII
E 35 pansion (CME)-a series
~ - 5-component
g of volume measurements
..c
(5
30 - 12-component performed on the reser­
voir fluid at different
Q.
~
25 pressures Stalting at the
CII
0 saturation pressure
'0 20 o Differential libera­
~
0 tion (DL)-approxima­
ai 15
E tion of fluid -property
:::I
changes with pressure
g 10 fur oil during a
"C
'3 natural depletion.
0-
::i 5 o Constant-volume

depletion (CYD )- ap­

-_.- ------._-----.. _- 0
-f-----~====~i::Er-.~~-·-:.I:::-0::::0::0
, ==:::-:=====::::;~=~;:;:;;::=::;;::::,5::;;O-:;:;:0==4,0:;0;;0==4
0;500- ;:'000--2;500- 3:000
.;::; ' T,;50:~0:;:==;5~,0v0;C;;0-+-·prox-imat-lQn -G-f-gas/GEm---­
densate depletion when
Pressure, psla condensate is immobile
o Separator test- a
Fig. 2-Effects of number of components on liquid dropout. . series of flashes with the
liquid from one flash
stock-tank conditions and need to be corrected to separator becoming the feed for the next flash at different conditions.
conditions to be consistent with the gas and liquid collected o Swelling test-measurement of Oil-property variation

The separator gas and liquid sometimes are recombined to a with the amount of injection gas added. These properties
known reservoir bubblepoint pressure rather than to the might include saturation pressure , oil-volume expansion ,
measured GOR Occasionally, when separator gas is in short oil viscosity, and oil density
supply, synthetic separator gas made from pure components o Multicontact test (forward and backward)-simulation

can be used to combine with the separator oil or stock-tank of the dynamiC phase behavior during a gas-injection
oil to provide a synthetic crude oil. Because of difficulties asso­ process. The forward process mimics the mixing of rich gas
ciated with the accurate measurement of GOR, the dovvnhole at the gas front with the original oil The backward process
or wellhead sampling approach for oil reservoirs is recom­ mimics the continuous contact of the injected gas with the
mended when conditions are favorable. When large volumes oil left behind the gas front. The experimental data include
of oil are required for slim-tube or coreflood experiments, sep­ interfacial tension (lFT), viscosity, and gas and oil compo­
arator sampling may be the only practical alternative. sition at each contact. The minimum miscibility pressure
However, for gas condensates, the separator-sampling (MMP) might also be determined from such a test
approach is recommended because liquid condensation • Slim-tube test-a dynamiC phase-behavior test where
can occur in the well bore during the well cleanup, even the injection gas displaces oil from uniform sand or glass
though the reservoir pressure remains greater than the beads packed in a small-diameter tube to minimize non­

JULY 2001
 3,000

3,100
• /
Reservoir
Pressure • Experiment
- - Calculated
\
Dewpoint / ­
E 3.2W Pressure \ GaslOil E 3,200 Methane

..£ Contact 5.
E.
~ 3,300 Bubblepoint
• •
2)
3,300 • •
3,400
•
Pressure ~

Ex periment·
\ 3,400

- - Calculated
3.500 L -_ __ _ --'---'--L-'--'-~_'__'__~'___'___'____L........._ ' 3,500 . I
5,000 5.200 5.400 5,600 5, 8W 6.000 0 10 20 30 40 50 60 70 80
Pres~ ure . psia Mole Fraction, %

Fig. 3-Variation In saturation and reselVoir pressures Fig. 4-Compositional changes of methane and C1 + with
with depth for an actual field. depth.

ideal effects such as fingering. The MMP is detennined

from the recovery curve. The effluent composition vs.
p r> sure is provided also. • Experiment
Fluid density, molecular weight (MW), 1FT, and viscosi­ - - Calculated
ty are measured in some of the above experiments. Often,
the composition (component mole fractions) of the sample
fluid is obtained from the compositions of gas and uid at

tion of gas IS measured by use of a gas chromatograph. The

composition of liquid is obtained by use of either the sim­ BottomhoJe density
ulated distillation gas chromatograph if the sample amount
is small , or, if a large volume (a few liters) of the sample is
available, true-boiling-point distillation . The measurements 250 350 450 550 650 750 850 950

for oils often provide mass fraction rather than mole frac­ Density. kglrn 3

tion . To convert these measuremen ts into mole fractions,

the molecular weight of each fractio n must be used , which Fig. S-Variation in bottomhole and stock-tank densities
with depth.
can be measured or computed from a correlation.
For gas condensates, the mole fractions of components
up to at least C ll should be measured because the compo­ from the EOS match. Such a calculation can be perfonned
nent mole fractions often increase with carbon number for on PVT experiments that release gases, which include the
components up to C ll . For oil reservoirs with a potential DL, CVD, and separator tests A material-balance calcula­
for solvent injection, extended compositional analysis tion on CVD data can prOvide oil molar density and satu­
might be needed for components up to Co or even heav­ ration at each pressure stage. Dewpoint pressure, high­
ier. Mass transfer of these components berween the gas and pressure composition, molecular weight, and viscosiry are
oil dominates the displacement characteristics of miscible difficult to measure accurately. ACl.urate measurement of
or near-miscible floods. However, accurate measurement molecular weight is important to convert laboratory data
of such heavy components is not an easy task. Some com­ to molar units used in EOS calculations.
panies also request that PVT data, primarily CME data , be For a reservoir shOwing significant compositional varia­
provided at temperatures different from the reservoir tem­ tions, several fluid samples at different depths or locations
perature to validate the EOS model might be needed to validate the EOS model for the entire field.
Data from these expe riments can be used as input for a
black-oil model or to calibrate an EOS for a compositional EOS Development for Reservoir Fl uids
model. Depending on the simulation model and the field Usually, petroleum-reservoir fluids contain several hundred
development, only some of these experiments are needed. distinct components, which makes it impossible to model
or a wated10od , the eME, differential liberation, and sep­ the fluids rigorously as a mixture of these pure compo­
arator tests are adequate. For gas/condensate reservoirs , nents. Therefore , "pseudocomponents" are used in EOS
CME, CVD, and separator-test data are adequate. These models, espeCially for the heavier components. The most
expe riments describe the fluid-property change during widely used EOS models are cubic in volume, so only these
prE'ssure depletion For solvent injectton , data from the cubic EOS's are discussed in this paper This type of OS is
swelling tes t also are needed . For miscible or near-miscible computationally efficient and does a good job of predicting
processes , multicontact and slim-tube lests are needed. phase behavior of petroleum fluids over wide ranges of
Laboratory data from those experiments contain errors. pressure and temperature. During simulations with an EOS
Material-balance calculations can help an engineer evalu­ compositional model, two families of chemical compounds
ate the quality of each data point. If a suffiCiently large represent a reservoir flui.d : individual chemical compounds
error is identified in a data point, then it can be excluded with well-defined phYSical properties and several groups of

JULY 2001
 heavy components (pseudocomponents) that are Iden tified A recent practice in the industry17 is to start with more
by use of a sillgle carbon nu mber. An EOS requires the crit­ components (eg. , 15 to 20), then successively reduce the
ical le mpera ture (Ie)' critical press ure (Pe)' and acentric number of components until the EOS loses Its acceptabll
fac tor (eli) for each component. However, the physical representatio n of the measured PVT d La Oftm, this
propertiD avatlable for the heavy groups usually are only process requires extended analysis of fluid composition
the mole fra ction, M'vV, and specific gravi.ty eSG), The crit­ during the laboratory tests and targets complicated sol­
ical properties of the heavy pseudocompon nts are diffi­ vent-inJ ection displacements Figs. 1 and 2 show typical
cult, if not impossible , to obtain experimentally Therefore , effects of the number of components on phase envelope
a bridge is needed between the measurable properties and and liquid dropout of gas condensates .
the parameters m an tOS. A practical number of compo­
nen ts to use in large compositional reservoir simulations Tuning EOS Pa rameters
typically is between 6 and 15. The EOS still needs to repre- As discussed in the previous section, a link is needed
sent the behavior of the fluids vvith an acceptable accuracy between the measurable properties, such as MW and SG ,
even when limited in the number of components The crit- and the EOS parameters, such as Ie' Pc' and w for the
ical volume and parac hor of these components also are heavy groups. Often, such a link is provided in the form of
needed for the calculation of viscosity and interfacial ten- empirical correlations, \,vhich inevitably involve
sion . Therefore , the following steps usually are involved. some uncertai.nties .
1. Develop pseudocomponents for the heavy fraction . Additionally, the well-defined compounds used to gener­
a. Split the heavy fraction into many pseudocomponents alize the parameters of the original cubic EOS are usually
by use of a distribu ti on function that honors the measured lighter than C 1s 18,19 Extension of the OS to heavier com­
MW and SG . 1,10-13 The MW and SG of each split pseudo- ponents existing in the petroleum fluids relies on the extrap­
component are ob tained. olation of the EOS parameters . The vapor pressure and the
h. Estimate the Te , Pc' and w for these split pseudo com- conditions at the critical point have been used to develop
ponents by use of an empirical correlation 10 ,1l,l4 two-parameter cubic EOSs , while the volumetric propen ies
-I~-----&-b~tmi3--+I1I~-5<e- s plit j3set1-d-eetrmpm'te-nts--int
- a Ee
t;,""',,---.t"ynpIW·ca-Ity--are-
w not ased~~TItt, a fIXed cntlcal compress­
groups. ibility factor (0.303 for the Peng-Robinson EOS and 0.333
d. Compute the Ie' Pe, and w for these lumped groups for the Soave EOS) is used. Therefore , these EOS's are not
by use of mixing rules_ expected to provide accurate estimates of denSity, especially
2. Validate this EOS model against experimental PVT for the liquid phase or when a large fraction of heavy com­
data and determine if tuning of the EOS parameters ponents is in the mixture. To improve the accuracy of the
is needed. liquid-density predictions, a volume-shift parameter was
3. Tun.- the EOS parameters added to two-parameter cubic EOSs20,21 Three-parameter
a. Select suitable experimental data to tune the EOS. cubic EOS's, such as that developed by Patel and Tqa 22 also
b Choose the adjustable EOS parameters that will be can be used to improve volumetric accuracy, but these EOS's
used to match these data_ commonly are not used in reservoir simulators.
c. Perform tuning by use of an optimization technique. Because of these uncertainties and others not discussed
4. Evaluate the predictive capability of this EOS model here, even with the good fluid-characterization procedures
It appears that Step 1 is independent of the EOS model available today, EOS predictions often are found to differ,
used . However, a lumping scheme that uses K-values 15 sometimes significantly, from the experimental data,
does require a specific EOS to provide the K- values. Therefore, tuning the EOS by adjusting some of the input
It should be added that, because o f the cost associated parameters to the experimental data is required.
with PVT-data collection and the time required to develop Tuning the heavy-grou p propert:es that relate to EOS
the EOS m odel , it is wise to Justify both the purpose and parameters requires the following.
importance of the effort. • Selection of tuning parameters and the ra nges in
which the selected parameters m ay vary
N umber of Pseudocomponents • Selection of experimental data for matching the tuned
The following gui.delines can be used to select the number EOS parameters .
of pseudocomponents. • Judgment regarding the predictive capability o[ the
1. At least two, often three or more , pseudocomponents tuned EOS model
are needed for the C 7+ fraction. Also , the intermediate
components between 2 and C6 can be grouped into two Tuning-Parame.Ler Selection. It is no t practical to m easure
pseudocomponents unless surface separation requires the parameters I e' Pe' and w for pseudocomponents. How­
detailed knowledge about them ever, three indirect parameters (MW, SG, and true boiling
2. lI1Jection-gas properties need to be modeled accurately point) for each heavy group are measurable. The Tc ' Pc' and
3. To adequately describe mass transfer between phases w values for each pseudocomponent can be detern1ined
[or vapori2.ed- and condensed-gas drives and their com­ from these indirect parameters by use of empirical correla­
bined prucesses , more intermediate components (C 5 tions . Alternatively, some correlations use carbon numbers
through ( 1) may be required for this purpose. The volume-shift parameter, c, also is
4. For p rocesses mvolving surface natural-gas liquid and available for tuning. Thus, seven parameters are available
ltquefied pelroleum gas extrac tion, explicit representation for each heavy group, three direct and four indirect. Binary
of these components in the EOS m odel may be reqUired interacti.on coefficients also can be considered as tunmg
unless a scheme is used to delump the produ ctio n stream parameters, but should be used with extreme caution as
at the surface for these components .16 discussed below.

JULY 200-1
 The direct parameters are functions of the indirect para­ a. For gas injection processes, fit the PVT data of the
meters. ror example, w can be determined once the M\V, SG, injection gas , if available. The Peng-Robinson EOS can sat­
and true boiling point are available . Hence, tuning of all these
isfactorily predict the PVT properties of pure carbon diox­
parameters by use of a regrf'..5sion program may break the ide , methane, and nitrogen, so do not adjust their physical
intemal consistency among them and result in a substantial properties during the EOS tuning process
loss of predicuve power of the O. A reasonable guideline b. Fit saturation points, K-values of the key compo
15 to tunc as few parameters as possible and as little as possi­
nents, phase molar fractions, phase molar volmll(,s , gas
ble to achIeve an acceptable match with the data while still moles removed, and Z-factors by sequentially tuning the
maintaining intemal consistency among the parameters critical properties of the heavy pseudocomponents Some
swelling test and multicontact data also can be added here
Match ·d-Dala Seleclion . A reliable EOS requires a good if the OS predictions do not differ Significantly from the
description of both phase volumes and phase mole measured data
amounts . Accurate phase volumes are needed to predict c. Match the liquid dropout data for gas condensates and
phase sa turation , which directly relates to phase relative all other volumetric properties by adjusting the volume­
permeability Therefore, the matched data need to include shift parameters of the components. In practice, sometim
both phase volumes and mole amounts , one EOS model must be used for the reservoir fluid and a
Normally, the CYD test reports the follOwing data at differem EOS model used for the surface fluid
each depletion pressure d, Refine the EOS model by tuning the MWs of the
• Devvpoint pressure and ltquid dropout heavy pseudocomponents to improve the match of prop­
• Composition, Z- factor, and moles of the gas removed erties, such as mass densities , that are related to MW
from the PYI cell. e, Repeat Steps b through d until consistent results are
• M\V and SG of the heavy end for the gas phase obtained
From the CVD data , material- and volume-balance cal­ f. Match other data such as viscosity and 1FT data.
culations can be performed [or valuable properties. These Adjustment of the component critical volumes is used to
properties are the K-values of all components, molar vol­ match the viscosity data for both gas and oil The parame­
umes of both gas and oil, and mole and volume fractions ters of the viscosity correlation can be adjusted if such an
of the oil phase option is available in the software. To fit the measured 1FT
The swelling test and multicontac t experiments also data , tune the component para chors.
should be matched for solvent-injection processes. 1FT 6. Test the EOS against the swelling test, multicontact,
data should be considered because of the strong depen­ and slim-tube data . The match to these data might be
dency of the relative pnweabilities on 1FT at low 1FT's, improved by adjusting the binary interaction coefficients
such as close to a critical poinr 23 The slim-tube test is a between nonhydrocarbons and hydrocarbons and between
measure of the dynamiC phase behavior included in the light hydrocarbons (methane and ethane ) and the
swelling and multicontact tests. MMP or minimum misci­ heavy pseudo components
biEty enrichment (MME) data can be included because an 7. Evaluate the predictive capability of the tuned model
analytical approach for computing the MMP or MME with During each step of the tuning, it is crucial to maintain
an EOS has been proposed 24 ,25 the adjusted parameters within reasonable physical limits
and vvith a monotonically increasing or decreaSing trend
Tuning Heavy-Group Properties. Many procedures for with carbon number, As the carbon number increases , the
tuning the EOS paramete.rs to match the experimental critical temperature and acentric factor increase , while the
PVT dala -have bce n-rep-oTtect-l-&,2e~G-As-desuibed-tn (h-e~Ttrlcal-pressure clecreases.- ­
previous section, the number of possible adjustable pa­ When several samples are collected at vanous depths or
rameters is large , which gives the tuning process a high from wells with different compositions, the reference fluid
rate of success in terms of matching the data . However, the should be the one whose PVT data are more complete and
accuracy and predictive power of the tuned EOS model of better quality An EOS model for the reference fluid is
still remains uncertain . Experience suggests that the fol­ then developed first 1 be tuned model is next applied to
lowing procedure often leads to a good EOS model. predict the PVT properties at other depths or locations.
1. valuate the reliability of the available PYT data. The MW of the heavy groups can be adjusted to match the
2. Group the measured PYT data into phase-equilibrium measured fluid compositions. Figs. 3 to 5 show an exa m­
data (saturauon pressures and K-values) molar and volu­ ple of the use of this approach for a reservoir that exhibits
metric data (molar and volume fractions , and molar vol­ large changes in fluid composition with depth
umes), MW-related data (mass denSity), and other data The regression process for tuning OS parameters to
such as viscosity and 1FT The surface separator and stock- measured PYI data re({uires speCific software, orne of thal
tank data, if any, might be grouped toge-Lher software provides a "sensitivity matrix ," which helps the
3. Predict PYT data by use of the initial EOS with all user assess the effect of adjusted variables on the PVI
zero binary interaction coefficients between hydro car­ properties of the fluid .
bon components,
4. Identify the major source of the EOS errors in terms Remarks
of the data type described in Step 2, Ihe saturation pressures and K-values are sensitive to the
5. Match the measured PYT data in a stepwise manner. binary interaction coeffiCients , so it has become common
Aft\ I each step, perform a PYT calculation by use of the practice to tune them to fit these data 3 l As noted by Ped­
modified model for all the available data to delermine the ersen et aI., 10,11 this practice might not always be good ,
need for continuation or termination of the tuning process. even though it is sometimes effective In fact. expenence

JULY 2001
 shows that an acceptable match can result without adjust­ ferent phases flow at different velocities in the direction
ing binary U1teraction coefficients of the hydrocarbons if that leads to a decrease in the flow potential of each phase.,
the EOS m odel includes some pseudocomponents that are the equilibrium between phases changes with time. To
heavy enough. The Gaussian quadrature method 32 for achieve this new equilibrium, mass transfer of these com­
lump ing me ts this requirement compared with other ponents between phases needs to take place co nstantly An
popular lumping schemes. EOS is used to solve this equilibrium part of the equations
The volume-shift technique moves the bubblepoint governing the multiphase and multicomponent flow at a
curve of a pure component, usually to the left, to match the given pressure and overall composition After the solution
sa turation liquid denSity. Then, a linear mixing rule is used of the eqUilibrium equations is reached, the EOS prOVides
to compute the volume-shift parameter for a mixture. Use the number of phases, phase molar and mass densities,
of this property does not affect the calculation of saturation phase saturations, and the composition of each phase.
pressures and phase molar amounts , but it does affect the Subsequently, phase viscosities and interfacial tension
properties computed wirh the molar or mass densities, such between phases can be determined by use of correlations.
as viSCOSity and 1FT This parameter also affects multiple
flashes if the input to each flash is a function of the phase Futu re Trends
volumes from the previous flashes. 1t should be noted that This paper has treated the proper use of traditional cubic
this parameter bears an empirical nature in both its defini­ EOS's for compositional reservoir simulation. It is worth
tion and its extension to a mixture . To maintain a phYSical­ noting that many improved EOS's have been developed lD
ly acceptable density, it is suggested that the volume-shift recent years Orbey and Sandler34 and Sandler35 discuss
parameter be limited. to within the range and illustrate some of these developments , In particular,
they show how liquid-activity-coefficient models can be
cup pcr =
I
ZCi RTc"
p c .i
EOS
C, I
)
- V . .. (1 modelswith used new mixing rules in the EOS formalism Such
greatly improve the accuracy of calculations for
mL"Xtures containing polar components, such as water, as
---f- -- _..-.Linking..1h~EOS_C.Olllposilioua.Lre"gp,tGir simulat-i.0ftwi.t.fr.-- weH--as-greatlJriTIuersnlTepm1ictlve power ot the EOS.
the surface-facility simulation to optimize the field devel­ These new mixing rules can be combined easily with fa mil­
opment plan is now feasible 16 Usually, the surface simula­ iar cubic EOS's, such as the Peng-Robinson EOS. Software
tion demands the use of a large number of components for volymetric and phase-behavior calculations that use the
(e.g.) l5 to 20) which is far more than that typically used in Wong-Sandler approach are readily available . It seems like­
full-field reservoir simulations, To compensate for this need ly that such improv€ments will gradually be Incorporated
and also to make the reservoir simula tion affordable, it is into at least some compositional reservoir simulators. J~'f'
natural to del1.1mp or expand the production stream from
the reservoir model that used pseudocomponents into the Nomenclature
surface model, which requires many components. To make C(ower= lower bound of the volume-shift parameter for
the del1.1mping possible and unique,33 it is necessary to use component i
the concept of the three-equation flash that requires zero ciupper= upper bound of the volume-shift parameter for
binary interaction coefficients. To approximate this process, component i
small binary interaction coefficients can be used. For this Mi= molecular weight for component i
purpose, it is better to limit the value or not to use the bina·· P( i = critical pressure for component i
ry interaction coefficients during the EOS development. R= universal gas constant
Coats and Smart26 proposed tuning the parameters Q a Te i= critical temperature for component i
and Q b of the Peng-Robinson EOS, This method has been ve'i= critical molar volume for component i
used widely in the petroleum industry It is easy to show that Zc'i= critical compressibility for component i
this procedure is somewhat equivalent to tuning Te and Pe Yi= speCific gravity for component i
as follows. w = acentric factor
Q a= an EOS constant
Q b = an EOS constant
, .. (2)
Superscripts
E05 = property computed from an EOS
Changes in the parameters Q a and Q b should be bound­ t= property after tuning
ed to guarantee a monotonic increase of Te and a mono­ u= properties before tuning
tonic decrease of Pe as the carbon number increases, but
this condition is often violated when Q a and Q b are tuned. References
Although tuning Q a and Q b for carbon dioxide and 1. Whitson, C H. and Brule, M.R: Phase BehavIOr, Monograph
methane can be very effective in improving the match, it is Series, SPE, Richardson, Texas (2000) 20.
much better to tune the critical properties directly rather 2. Todd, M.R and Longstaff, W]. "The Development,Testing, and
than Q a and Q b for the heavy components Apphcation of a Numerical Simulator for Predicting Miscible
Flood Performance," 1PT Ouly 1972) 874: Trans, AIME 253.
link Between EOS -PVT and 3. Fayers, FJ et aL: "Use of the Four-Component Todd and
Compositional Simulation Longstaff Method as an Upscaling Technique in Simulating
Reservoir simulators are built on the basis of mass or Gas IllJection Projects," paper SP E 59340 presemed at the
m olar balances of each component Such a balance is 2000 SPEIDOE Improved Oil Recovery Symposium, 1l11sa,
assumed at any moment during the fluid flow. Because dif- Oklahoma , 3- 5 April.

JULY 200]
 4. Tang. D.E. and Zick, AA: "A New Limited CompoSitional 24. Jessen, K et aL: "Calculation of MMP," paper presented at IntI.
Reservoir Simulator," paper SPE 25255 presented at the 1993 Conference on Production of Reservoir Fluids in rontier
SPE Reservoir Simulation Symposium, New Orleans, 28 Feb­ Conditions, Rueil-Malmaison, France (4-5 December 1997)
ruary-3 March. 25. Wang, Y and Orr, FM.: "Calculation of Minimum Miscibili­
5. Jerauld, G .R.: "A Case Study in Scaleup for Multicontact Mis­ ty Pressure." paper SPE 39683 presented at the 1998
cible Hydrocarbon Gas Injection," paper SPE 39626 present­ SPEIDOE Improved Oil Recovery Symposium, 1ulsa, Okla­
ed at the 1998 SPEIDOE Improved Oil Recovery Symposium, homa, 19-22 April.
Tulsa, Oklahoma, 19-22 April. 26. Coats, KH. and Sman, GT : "Application of a Regre sion-
6. Wang, P et a/. : "A Fully Implicit Parallel EOS Compositional Based EOS PVT Program to Laboratory Data," SPE (May
Simulator for Large Scale Reservoir Simulation," paper SPE 1986) 277.
51885 prest:llted at the 1999 ST'i ' Reservoir Simulation Sym­ 27. Agarwai, RX, Li , Y-K, and Nghiem , L: "A Regression Tech­
posium, Houston , 14-17 February. nique With Dynamic Parameter Selection for Phasr- Behavior
7. Dogru , A.H : "Megacell Reservoir Simulation, " 1PT (May Matching," SPERE (February 1990) ll5.
2000) 54. 28. Merrill , RC , Hartman, Kj. , and Creek , JL: "A Comparison
8. Abate , J, Wang, P, and Sepehrnoori, K: "Parallel Composi­ of Equation of State Tuning Methods," paper SPE 28589 pre­
tion Reservoir Simulation on Clusters of PC's, " The IntI. I sented at the 1994 SPE Annual Technical Conference and
High Pe10rmance Computing Applications (Spring 2001) Vol. Exhibition, New Orleans, 25-28 September
15, No. 1 29. Joergensen, M_ and Stenby, E "Optimization of Pseudocom­
9. Shiralkar, GS et al. "Falcon A Production Quality Distribution ponent Selection for Compositional Studies of Reservoir Flu­
Memory Reservoir Simulator ' SPEREE (October 1998) 400. ids," paper SPE 30789 presented at the 1995 SPE Annual
10. Pedersen, KS., Thomassen, P, and Fredenslund, A. : "Char­ Technical Conference and Exhibition, Dallas, 22-25 October.
acterization of Gas Condensate ML:tures," paper presented at 30. Liu , K: "Fully Automatic Procedure for Efficient Reservoir
the 1988 AIChE Spring National .vieeting, New Orleans, Fluid Characterization, " paper SPE 56744 presented at the
16-19 March . 1999 SPE Annual Technical Conference and Exhibition,
11 . Pedersen, KS., Fredenslund , A. , and Thomassen, P: Proper­ Houston, 3-6 October
ties of Oils and Natural Gases , Gulf Publishing Co, Houston, 31. Slot-Petersen , C: "A SystematiC and Consistent Approach To
Texas (1989) . Determine Binary Interaction Coefficients for the Peng­
12. Whitson, ca, Anderson, TF, and Soreide, L: "C7+ Charac­ Robinson Equation of State," SPERE (November 1989) 488.
terization of Related Equilibrium Fluids Using the Gamma 32. Behrens, RA. and Sandler, S.L: "The Use of Semicontinuous
Distribution," paper presented at the 1988 AIChE Spring Description to Model the C 7+ Fraction in Equation of State
National Meeting, New Orleans, 16-19 March . Calculations," SPERE (August 1988) 1041.
13 Khan, S.A. , Pope, GA , and Sepehrnoori, K "Fluid Charac­ 33. Leibovici, CE, Barker, JW , and Wache , D.: "A Method for
terization of Three-Phase CO/Oil Mixtures," paper SPEIDOE Delumping the Results of a Compositional Reservoir Simula­
24130 presented at the 1992 SPEIDOE Eighth Symposium on tion," paper SPE 49068 presented at the 1998 SPE Annual
Enhanced Oil Recovery, Tulsa, Oklahoma, 22-24 April. Technical Conference and Exhibition, New Orleans ,
14 Twu, CH. "An Internally Consistent Correlation for Predict­ 27-30 September.
ing the Critical Properties and Molecular Weights of Petrole­ 34. Orbey, H. and Sandler, S.L: Modeling Vapor-Liquid Equilibria:
um and Coal-Tar Liquids," Fluid Phase Equilibria (1984) 16, Cubic Equations of State and Their Mixing Rules, Cambridge U.
137. Press, New York City (1998)
15. Newly; TMJ and Merrill, R.C : "Pseudo component Selection 35 . Sandler, S.L: Chemical and Engineering Thermodynamics, third
for Compositional Simulation," SPERE (November 1991) edition,john Wiley &: Sons Inc., New York City (1999)
490; Trans, AIME, 291.
6_--'BaIkerT+W_and-l~ibO-v:iGi-;.-c..~k>%lumI*ng-GemJ30Si+ienal.--Peng-Wang-rs-cneservotrs-rmrJiutio-nengineeTarDeGblyer-
Reservoir Simulation Results: Theory and Applications ," and MacNaughton. His areas of interest include reservoir sim­
paper SPE 51896 presented at the 1999 SPE Reservoir Simu­ ulation, equation-of-state development, enhanced oil-recov­
lation SympOSium, Houston, 14-17 February ery processes, and well deliverability for gas condensates.
17. Jerauld, G.R.: "Timing of Miscible Hydrocarbon Gas InJec­ Previously, Wang was with the Elf Geoscience Research Cen­
tion After Wated100ding," paper SPE 59341 presented at the ter in London and the U of Texas at Austin_ He holds a BS
2000 SPEIDOE Improved Oil Recovery Symposium, Tulsa, degree in chemical engineering from Tsinghua U, Beijing, an
Oklahoma, 3-5 April. MS degree in petroleum engineering from Petroleum U of
18 Peng, D.-Y and Robinson, D.B. : "A New Two-Constant Equa­
China, Beijing, and a PhD degree in chemical engineering
tion 01 State," Ind. Eng. Chem. Fund. (1976) 15, 59-64.
from Technical U of Denmark. Gary A. Pope is director of
19 . Soave, G. "Equilibrium Constants From a Modified Redlich­
the Center for Petroleum and Geosystems Engineering at U of
Kwong Lquation of State," Chel11. Eng Sci., 27 (1972) 1197.
20. Peneloux, A , Rauzy, E , and Freze, R. : "A Consistent Correc­ Texas at Austin. His areas of interest include enhanced oil
tIon for Redlich-Kwong-Soave Volumes," Fluid Phase EqUilib­ recovery, reservoir engineering, fluid flow in porous media,
ria (1982) 8, 7-23. numerical simulation, phase behavior and fluid properties,
21 Jhaverl, BS and Youngren , G X: "Three-Parameter Modifi­ water-soluble polymers, surfactants, surfactant-enhanced
cation of the Peng-Robinson Equation of State to Improve aquifer remediation, characterization of nonaqueous-phase
Volumetric Predictions," SPERE (August] 988) 1033. liquids in groundwater, and the use of tracers for characteri­
22 . Patel, N.C and Teja , A.S .. "A ew Cubic Equation of State for zation of aquifers. Previously, Pope was with the U of Texas
Fluids and Fluid Mixtures," Chem . Eng. Sci, Vol. 37, No.3 and Shell Development Co. He holds a BS degree in chemical
(1982) 463. engineering from Oklahoma State U and a PhD degree from
23 . Pope, GA et al.: "Modeling Relative Permeability Effects in Rice U Pope has received the SPE Distinguished Achievement
Gas-Condensate Reservoirs With a New Trapping Model," Award (7996), SPE Distinguished Member Award (7995), and
SPERl: E (April 2000) 1 71 the SPE Reservoir Engineering Award (7997).

JULY 2001