Documente Academic
Documente Profesional
Documente Cultură
Well Deliverability
ivind Fevang,* SPE, and C.H. Whitson, SPE, Norwegian Technical and Natural Science U.
o o
g rg
g g
and the BHFP reaches a minimum (dictated by surface constraints) p wf
and the well is forced to go on decline.
Muskat2 addresses the condensate-blockage problem in his dis- or in terms of black-oil PVT,
cussions of gas cycling, where he introduces a simple method for es- pR
timating the radius of condensate blockage as a function of time, gas
rate, and reservoir-rock and -fluid properties. Fetkovich3 uses Mus-
kats results to derive a rate- and time-dependent blockage skin for
qg + C Bk m R ) Bk m dp, . . . . . . . . . . . . . . . . . . (2)
o
ro
o
s
rg
gd g
p wf
use in the standard gas rate equation.
Kniazeff and Naville4 and Eilerts et al.5,6 were the first to numeri-
cally model radial gas-condensate well deliverability. These studies 2 p a1 k h
show radial saturation and pressure profiles as functions of time and where C + , . . . . . . . . . . . . . . . . . . . (3)
ln(r er w) * 0.75 ) s
*Now with Pera A/S. a1+1/(2 p 141.2) for field units, and a1+1 for pure SI units. The
Copyright 1996 Society of Petroleum Engineers
constant C includes basic reservoir properties, such as permeability,
k ; thickness, h; drainage radius, re ; wellbore radius, rw ; and other
Original SPE manuscript received for review 25 October 1995. Revised manuscript received
28 May 1996. Paper peer approved 6 August 1996. Paper (SPE 30714) first presented at the
constants. Relative permeabilities krg and kro are defined relative to
1995 SPE Annual Technical Conference and Exhibition held in Dallas, 2225 October. absolute permeability, not relative to permeability at irreducible wa-
ter saturation (this distinction is particularly important when corre- Calculating Pseudopressure. On the basis of our observations of
lating relative permeability data). Skin, s, is a composite factor that the three flow regions for many gas-condensate systems, we have
includes nonideal flow effects, such as damage, stimulation, drain- developed a simple method to calculate the pseudopressure integral
age geometry, and partial penetration. The traditional approach for in Eqs. 1 and 2 accurately. The approach is an extension of the
estimating12 or measuring13 composite skin for a well producing pseudopressure method proposed by Evinger and Muskat for solu-
single-phase fluid can be used to determine skin. The condensate- tion-gas-drive oil wells. First we break the pseudopressure integral
blockage effect is treated separately by the pseudopressure integral. into three parts, corresponding to the three flow regions discussed.
We show that the pseudopressure integral, if evaluated properly, is
B k m
pR
independent of well geometry for practical purposes. This greatly rg k ro
simplifies the treatment of gas-condensate well deliverability. Total D p p + ) R dp +
gd g Bo mo s
p wf
Flow Regions. An accurate yet simple model of a gas-condensate
well undergoing depletion consists of three regions (Fig. 1).
B k m
p*
Region 1: An inner near-wellbore region where both gas and oil
flow simultaneously (at different velocities). rg k ro
Region 1 ) R dp )
Region 2: A region of condensate buildup where only gas is flowing. gd g Bo mo s
p wf
Region 3: A region containing single-phase (original) reservoir gas.
For a given producing condition, one, two, or all three regions
pd
may exist. These three regions define pseudosteady-state flow
conditions, meaning that they represent steady-state conditions at a
given time but that the steady-state conditions change gradually
Region 2 Bk m rg
gd g
dp )
during depletion. p*
Region 1. The flowing composition (GOR) within Region 1 is
pR
constant throughout. This means that the single-phase gas entering
Region 1 has the same composition as the produced wellstream mix-
ture. Conversely, if we know the producing wellstream, then we
Region 3 k rg(S wi) B 1m gd g
dp. . . . . . . . . . . . . . . . . . . . . (4)
know the flowing composition within Region 1. Furthermore, the pd
dewpoint of the producing wellstream mixture equals the reservoir
Given the producing GOR, Rp , we know immediately p* because it
pressure at the outer edge of Region 1.
equals the dewpoint of the producing wellstream. Using black-oil
Region 1 is the main source of deliverability loss in a gas-condensate
PVT, with rs defined as the solution oil/gas ratio (OGR), we locate
well. Gas relative permeability is reduced owing to condensate buildup.
the pressure in the PVT table where rs +1/Rp and define this pres-
The size of Region 1 increases with time. For steady-state conditions,
sure as p*. In a compositional treatment, the dewpoint of the produc-
the condensate saturation in Region 1 is determined (as a function of
ing wellstream composition is defined as p*. If p*upR , then in-
radius) specifically to ensure that all liquid that condenses from the
tegration of the Region 1 integral should be only from pwf to pR ; in
single-phase gas entering Region 1 has sufficient mobility to flow
this case, Regions 2 and 3 dont exist.
through and out of Region 1 without any net accumulation. Region 1. The Region 1 pseudopressure integral is solved by use
Region 2. If it exists (as it usually does), Region 2 defines a region of the modified Evinger-Muskat approach. At pressures ptp*, the
of net accumulation of condensate. Effectively, only gas is flowing PVT properties Rs , Bo , rs , Bgd , mo , and mg are found directly. Next,
in this region because oil mobility is zero (or very small). Conden- the equation defining producing GOR,16
sate saturations in Region 2 are closely approximated by the liquid-
dropout curve from a constant-volume-depletion (CVD) experi-
ment,14 corrected for water saturation.
The size of Region 2 is largest at early times just after the reservoir
Rp + Rs ) kk mm BB (1 * r R ), . . . . . . . . . . . . . . . (5)
rg
ro
o
g
o
gd
s p
pressure drops below the dewpoint. It decreases in size with time be- is used to calculate krg /kro as a function of pressure.
cause Region 1 is expanding. The size and importance of Region 2
is greater for lean gas condensates.
The main consequence of Region 2 is that producing wellstream
k rg
k ro
( p) +
R p * R s m g B gd
1 * rs Rp mo Bo
, . . . . . . . . . . . . . . . . . . . . . . (6)
composition (GOR) is leaner than calculated by a simple volumetric
material balance (e.g., CVD measurements). Incorrect use of materi- where PVT properties are known as functions of pressure. It is readi-
al-balance GORs in calculating the pseudopressure significantly ly shown that Eq. 6 can be expressed in terms of the oil relative vol-
ume of the flowing gas during a constant composition expansion with 45 STB/MMscf and a maximum CVD liquid dropout of 2%.
(CCE), VroCCE +Vo /(Vg )Vo ) by Black-oil PVT properties for the two fluids were generated with the
methods proposed in Ref. 14. Key reservoir properties are given in
k rg
k ro
( p) + 1 * 1 mg .
V roCCE mo . . . . . . . . . . . . . . . . . . . . . . . (7) Tables 1 and 2. The gas/oil relative permeability data are calculated
with a Corey equation.17 Fig. 3 shows Set A curves; these curves are
From Eqs. 6 and 7, VroCCE can be expressed in terms of black-oil used in all calculations unless otherwise stated.
PVT properties, for any producing GOR Rp by
Compositional vs. Black-Oil PVT Formulation. Coats18 presents
1 radial-well simulations that show a modified black-oil PVT for-
R p * R s B gd mulation gives the same results as a fully compositional EOS PVT
V roCCE( p) + 1) . . . . . . . . . . . . . (8) formulation. Results are given for a rich gas condensate producing
1 * rs Rp Bo
on decline for 8 years. The EOS characterization uses seven compo-
As Evinger and Muskat show, relative permeabilities krg and kro nents with one C7+ fraction. Results from this example probably
can be expressed directly as functions of the ratio krg /kro (when both should be used with caution because only one C7+ fraction is used.
phases are mobile). This means that we can evaluate krg and kro di- With a more detailed C7+ split, oil viscosity differences between
rectly as functions of pressure in the Region 1 pseudopressure inte- black-oil and compositional formulations may yield noticeable dif-
gral, krg (p)+f [krg /kro ( p)] and kro ( p)+f [krg /kro ( p)], using Eq. 6. ferences in well deliverability.
Region 2. When Region 2 exists ( p*tpR ), the Region 2 integral Fig. 4, where oil viscosity is plotted vs. pressure, illustrates the prob-
is evaluated by use of krg (So ), where So is estimated as a function of lem. The solid line represents black-oil data, and the symbols represent
pressure from CVD relative oil volumes VroCVD ( p)+Vo ( p)/Vd , results taken from compositional simulation of Rich Gas A. The figure
yielding So (p)+[VroCVD (p)](1*Sw ). If VroCVD values are not shows that oil viscosities can change significantly during depletion.
known for the black-oil PVT data set, they can be calculated with Because the oil mobility required to flow condensed oil in Region
1 is basically fixed, a lower oil viscosity in the compositional simula-
N k1 * G k*1 (r s) k tions (particularly near the wellbore) results in a lower oil relative
(V ro CVD) k + (B o) k ,
1 * (r s R s) k permeability and lower oil saturation than in the black-oil simulations.
N k1 + V B
roCVD
o
)
1 * V roCVD
B gd
rs k1
,
TABLE 1RESERVOIR PROPERTIES USED IN SIMULATIONS
Water compressibility, psi*1 2.67 x 10*6
Rock compressibility, psi*1 5.00 x 10*6
V B
Reservoir height, h, ft 200
roCVD 1 * V roCVD Porosity, f, % 30
and G k1 + Rs ) , . . . . . . . . . . . . . (9) Absolute (horizontal) permeability, k, md 6
o B gd
k1 Relative permeability at Swi 0.8
where k represents the current pressure, k*1 represents the pre- Irreducible water saturation Swi , % 25
Reservoir area, acres 650
vious pressure, and (VroCVD )0+0. Gas plateau rate, MMscf/D 40
Region 3. Only gas PVT properties are found in the Region 3 inte- Minimum BHFP pwf min, psia 1,500
gral, where the traditional single-phase gas pseudopressure function
can be used.
TABLE 2KEY RESERVOIR FLUID PROPERTIES
Verification of Proposed Pseudopressure Approach Rich Gas Lean
To illustrate the proposed method for determining gas rate from Eq. A Gas B
2, with pseudopressure calculated by use of the proposed method Initial reservoir pressure, psia 6,500 5,500
outlined, we simulated several examples. We use two gas-conden- Initial reservoir temperature, F 266 315
sate fluids with radial, vertically fractured, and horizontal well ge- Dewpoint pressure, psia 5,900 5,400
ometries. Refs. 23 and 24 provide details of the two reservoir fluids Maximum CVD liquid dropout VroCVD , % 24 2
and all simulations reported here. Initial solution OGR rsi , STB/MMscf 175 45
STO API gravity, API 55 45
The two fluids are Rich Gas A, an undersaturated gas condensate
with 175 STB/MMscf and a maximum CVD liquid dropout of 23% Separator conditions. Stage 1 p+375 psia and T+108F, Stage 2 p+14.7 psia and
(Fig. 2), and Lean Gas B, a slightly undersaturated gas condensate T+60F.
Lower oil saturation results in higher gas relative permeability and bet- estimated. The dot-dashed line uses the same CVD MB method but
ter well deliverability for the compositional simulations (Fig. 5). with pwf +pwfmin+1,500 psia for all times. The difference in rates
This problem can be improved with a modified mo ( p) relationship calculated with the proposed method and the CVD MB method is
in the black-oil simulator. The dashed line in Fig. 4 passes through largest at early times. The reason is that Region 2 is largest at early
the important compositional results (data at pressures lower than times and decreases in size with time.
the point where mo reaches a minimum). This same trend can be de- The same radial-well simulation is run with Rich Gas A. Results
termined by use of a PVT simulator. When the modified mo ( p) rela- are given in Fig. 7, where the proposed method for evaluating
tion is used in black-oil reservoir simulation, well performance is pseudopressure reproduces the simulated results almost exactly.
closer to compositional results (dashed line in Fig. 5). The simplified CVD MB method gives good results for only a short
time while the reservoir is still sufficiently undersaturated that pro-
Pseudopressure Method vs. Fine-Grid Simulations. Gas rate was ducing GOR equals the initial solution GOR (1/rsi ). As soon as Rp
calculated with the proposed method for determining pseudopres- deviates from 1/rs ( pR ), the CVD MB method starts to overestimate
sure in Eq. 2. The same black-oil PVT data were used as in the simu- deliverability loss.
lations. Producing GOR, BHFP, and average reservoir pressure as
Fine-grid 2D simulations of a vertically fractured well and fine-
functions of time are taken from the simulator. For comparison, gas
grid 3D simulations of a horizontal well gave the same results as for
rate was also calculated with the same pseudopressure function, but
with producing GOR set equal to 1/rs (evaluated at pR ). This ap- the radial-well simulations.23,24 The proposed method for calculat-
proach is equivalent to using a material balance based on a simple ing pseudopressure (where producing GOR was known accurately)
CVD depletion process [the CVD material-balance method(CVD resulted in accurate prediction of production performance. This ver-
MB)]; this implies Region 2 doesnt exist. ifies that the well geometry term C in Eq. 2 can be separated from
The Lean Gas B production forecast for a radial well is shown in the pseudopressure integral.
Fig. 6, with simulated black-oil results shown as symbols. The solid Our simulations also show that the well with a 50-ft vertical frac-
line represents gas rate calculated with the proposed method to eval- ture half-length has the same deliverability as the 1,000-ft horizon-
uate pseudopressure (with Rp , pR , and pwf from the simulator). The tal well. This result is for a typical kv /kh ratio of 0.1. With kv /kh +1,
productivity index (PI) constant, C in Eq. 2, is calculated from Eq. the plateau period increases from 3 to almost 10 years. This extreme
3 with s+0. Results are excellent. The dashed line represents gas sensitivity to kv /kh does not exist for vertically fractured wells, and
rate calculated with the same pseudopressure function (Eq. 2), also if horizontal wells are considered in the development of a gas-con-
with pR and pwf from the simulator, but with Rp +1/rs (with rs eva- densate reservoir, the kv /kh ratio should be determined with certainty
luated at pR ). Results are poor; well deliverability is highly under- to avoid overly optimistic production forecasts.
Fig. 6Well performance for a radial well with Lean Gas B. Com- Fig. 7Well performance for a radial well with Rich Gas A. Com-
parison of fine-grid simulation with proposed pseudopressure parison of fine-grid simulation with proposed pseudopressure
method (and with approximate CVD MB method). method (and with approximate CVD MB method).
Application to CoarseGrid Field Models Coarse Radial Grid. The size of the first grid cell in a radial simula-
The main conclusion from the previous comparisons is that the pro- tion can be important in modeling well deliverability of gas conden-
posed method for calculating the gas rate pseudopressure function sates. This is shown in Fig. 8, where Lean Gas B is used with a first-
for a gas-condensate well is accurate as long as the producing GOR grid-cell radius of 100 ft (vs. 0.7 ft in the fine-grid simulation); the
is known accurately, independent of well geometry and production remaining grids are spaced logarithmically. The plateau period is
mode. Given this observation, we decided to evaluate the accuracy more than doubled from 2.5 for the fine-grid simulation to 6.25
years for the coarse-grid simulation. Even for an 18-ft inner radius,
of producing GORs predicted by coarse-grid simulations.
the plateau period is overpredicted by more than 1 year.
Results show that coarse-grid GORs are generally very accurate.
With the multiphase pseudopressure method (based on a pseudo-
Consequently, the proposed method for calculating pseudopressure
tubing curve), the correct plateau period of 2.5 years is predicted and
function of a gas-condensate well can be used to convert coarse-
the rate/time performance overlays the fine-grid simulation results.
grid-cell pressures to BHFPs (individually for each well grid cell)
With the same coarse radial grid (r1+100 ft) and Rich Gas A, the
accurately. Conversion from grid-cell pressure to BHFP for a gas proposed pseudopressure method also predicts the entire rate/time
well is usually made with the radial-flow equation and a well index, performance accurately.
J, where pwf +pgrid*qg /J. J is given by
k rg Coarse Cartesian Grid. A coarse Cartesian grid with a 200 200
J+C , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (10) ft well grid cell was used to compare the proposed method with the
m g B gd
fine-grid radial simulation for Lean Gas B. Results are shown in Fig.
where krg , mg , and Bgd are evaluated at conditions in the well grid 9, where the rate/time performance is calculated accurately with the
cell. Our proposal is simply to replace this formulation with the coarse-grid pseudopressure method. Note that the standard well
pseudopressure integral defined in Eqs. 2 and 4, evaluating the inte- treatment (Eq. 10) results in a plateau of 6 years compared with the
gral from pwf to pgrid. correct plateau of 2.5 years. By use of a coarse Cartesian grid with
Although Peacemans equation(s)19 can be used to calculate J (or a 500 500 ft well grid cell, the proposed method was compared
C), we consistently found that it was better to determine the well in- with 2D fine-grid simulation results of a vertically fractured well
dex with results from single-phase simulations with a fine grid. In with Rich Gas A. Fig. 10 shows that the rate/time performance is
a simulator, all PVT and relative permeability properties are avail- calculated accurately with the coarse-grid pseudopressure method.
The standard well treatment results in a plateau of 6 years compared
able in each grid cell. For the sake of efficiency, the pseudopressure
with the correct plateau of 3 years. A coarse Cartesian grid with
function D pp can be calculated during initialization for several pro-
333 333 25 ft well grid cells was used to compare the proposed
ducing GORs, and then stored as a 3D table Dpp ( pwf , pgrid, Rp ) or
method with fine-grid 3D horizontal well simulation results for Rich
Dpp ( pwf , pgrid, Sg ). Conceivably, a different D pp function needs to
Gas A. Results are shown in Fig. 11, where the rate/time perfor-
be generated for all PVT/relative permeability regions. In the most
mance is calculated accurately with the coarse-grid pseudopressure
general case, Dpp can be stored as a 4D table to handle changing wa- method. The standard well treatment results in a plateau of almost
ter saturations, D pp (pwf , pgrid, Rp , Sw ). 6 years compared with the correct plateau of 3 years.
We wanted to test the proposed application using a commercial res-
ervoir simulator. Interas ECL100 was used, but it was necessary to Discussion of Coarse-Grid Pseudopressure Method. We have
incorporate the pseudopressure table as a pseudo-tubing table be- shown that local grid refinement in gas-condensate wells is not nec-
cause we did not have access to source code. First, we introduced a essary. The only limitation of the pseudopressure approach is that
large (infinite) well index, J, so that the model-calculated BHFP producing GOR from the coarse-grid model is reasonably accurate
equals the well-grid-cell pressure. The pseudotubing table then con- (compared with fine-grid simulation).
verts this (well-grid-cell) pressure to actual BHFP (with Eq. 2), on the Effectively, Region 2 is eliminated in the well grid cell when this
basis of gas rate and producing GOR. The pseudotubing-table ap- approach is used. Surrounding grid cells, however, automatically
proach can be used as a general solution to problems where the well treat the Region 2 pressure losses. The more important Region 1 be-
produces from layers that are in vertical communication (i.e., experi- havior is treated accurately in the well grid cell. However, if the size
encing reservoir crossflow). However, the approach is not recom- of the well grid cell becomes too large, Region 1 pressure losses will
mended for wells producing from layered no-crossflow systems. The be overestimated and calculated well deliverability underestimated.
best general solution is to have the well pseudopressure tables gener- In the limit of one grid cell describing the entire drainage area, this
ated at initialization, so that any grid cell that becomes a well grid will method becomes equivalent to the CVD MB method, which always
automatically have the multiphase pseudopressure method available. underestimates well deliverability.
Deciding an appropriate well grid size will depend on (1) the lean- Simulations of rate/time performance with relative permeability
ness of the gas condensate, (2) the minimum well-plateau length, Sets B and B are shown in Fig. 17. The well performs identically,
and (3) the degree of undersaturation. Smaller well grids are needed for practical purposes, with both sets of relative permeabilities.
for lean gas condensates, short plateau periods, and initially satu- These results are not a special case but are generally true for all gas-
rated fluids. In other words, a saturated lean gas condensate that condensate reservoirs. krg vs. krg /kro is the fundamental relative
goes on decline immediately will require the smallest well grid size. permeability relationship dictating Region 1 behavior and well de-
A few sensitivity cases can be run to determine the required well liverability, not krg and kro as a function of saturation.
grid size for a given gas-condensate reservoir. These cases should
evaluate wells producing at the maximum expected rate (dictated by Critical Oil Saturation. It has been suggested that Soc is an impor-
equipment constraints, such as tubing diameter and erosional veloc- tant parameter in defining relative permeabilities of gas conden-
ity). To evaluate a late well being drilled after depletion to the sates. The only reason this claim can be made is that Soc has been
dewpoint, a case with initial pressure equal to the dewpoint should used in parametric equations where effective oil saturation S o* is
be used, even for highly undersaturated reservoirs. normalized with Soc ; e.g.,
1 *S S
(2 ) l) l
Relative Permeability Effects k ro + (S o*) 2 o
parameter in sensitivity studies, the effect it has on well deliverabil- Gas/Oil IFT. An often misunderstood concept in gas-condensate well
ity should be recognized as the effect Soc has on the krg +f (krg /kro ) deliverability is the importance of gas/oil IFT on relative permeabili-
relationship, and not on kro or krg values near Soc . ties. Numerous authors have discussed the potential effect, functional
Because Region 1 flow behavior dictates well-deliverability loss dependence, and methods for measuring the effect of IFT on relative
caused by condensate blockage and because oil saturation, oil rela- permeabilities, but noone has yet shown that well performance is sig-
tive permeability, and oil mobility are all at a maximum in Region nificantly improved by straightened-line relative permeabilities ow-
1, the low-oil-saturation relative permeabilities (near Soc ) are irrele- ing to low IFTs. The two reservoir mechanisms that are affected by rel-
vant to condensate blockage. For richer condensates, Region 2 may ative permeabilities in gas-condensate reservoirs are (1) well
have oil saturations somewhat greater than Soc . Even so, the oil rela- deliverability and (2) gravity segregation of condensate that theoreti-
tive permeability in this saturation region is unimportant because oil cally can occur in high-permeability or fractured reservoirs.
mobility is practically zero. Furthermore, gas relative permeability The physics of IFT effect on relative permeabilities is not well un-
at low oil saturations is not affected directly by Soc . derstood. Measurements quantifying the effect in a systematic way
To illustrate the insignificance of Soc , Rich Gas A was simulated are lacking, and the data available are not reliable enough to build a
with radial and vertically fractured well geometries by use of Set A theoretical (or empirical) model for predicting the effect. The existing
conceptual model states that IFTs must be lower than a threshold
relative permeabilities (Soc +10%). Rate/time performance for the
IFT, s*, before relative permeabilities are affected. Furthermore, as
two wells are shown as solid lines in Fig. 18. For this fluid, the impor-
IFT approaches zero the relative permeabilities approach straight
tant region of krg /kro is less than 3 for most of depletion. On the basis
lines with zero residual saturations. The model is given by
of this observation, we modified the kro curve at saturations where
krg /kro is greater than 3 (So t [ 40%). Extrapolations of kro with k r + F k r , Immiscible ) (1 * F) k r , Miscible ,
Soc +0% and Soc +20% are shown in Fig. 19. As seen in Fig. 18, the
simulated rate/time performance is practically identical for Soc +0% S r + F S r , Immiscible ,
and 20% compared with the base case with Soc +10%. We also ran
compositional simulations at lower BHFP (250 psia) to see if low oil k r , Miscible + (S * S r)(1 * S wi * S r) ,
saturations in Region 1 from vaporization would result in significant
ss* , s t s *; F + 1, s w s *.
n
differences in well deliverability for the three oil relative permeability
and F + . . . . . . . . . . . . (12)
curves with different Soc values. The effect was very small.
Fig. 15Two sets of gas/oil relative permeability curves that Fig. 16Identical krg +f(krg /kro ) relationship for two completely
have identical krg +f(krg /kro ) relationships. different sets of krg (Sg ) and kro (Sg ) curves.
Some measurements indicate that s* ranges from 0.1 to 0.3 dyne/ ect economics when the IFT/relative permeability phenomenon is
cm. Exponent n+0.1 is recommended, though this is based more on so poorly understood.
physical intuition than experimental evidence.
Concerning the effect of IFT on well deliverability, we have made Velocity/IFT Effect. Schulte20 questioned the validity of using only
the following observations (which are supported by simulations IFT effect on relative permeabilities for gas-condensate systems. He
presented below). (1) Gas-condensate reservoirs will generally nev- presents arguments that indicate an additional improvement in krg (in
er experience IFTs lower than threshold IFTs of 0.1 to 0.3 dyne/cm addition to low IFT) that might be expected because of high velocities
in near-wellbore Region 1, once BHFP reaches a minimum and the (i.e., pressure drops) experienced near the wellbore. Effectively, he
well goes on decline. (2) Low IFT in Region 2 and the resulting im- claims that the capillary number (given by the ratio of viscous to cap-
provement in gas relative permeability has only a minor effect (if illary forces) should be the correlating parameter instead of IFT (cap-
any) on deliverability. illarity) alone. Ref. 21 presents more detailed results.
Simulation of a worst case with Rich Gas A was made for radial Subsequently, Henderson et al.22 provided experimental results of
velocity and velocity/IFT effects on relative permeability using Berea
and vertically fractured wells. The IFT model given by Eq. 12 uses
sandstone for a five-component, synthetic gas-condensate mixture.
Set A immiscible relative permeability curves, s*+0.3 dyne/cm,
The two systems had IFTs of 0.05 and 0.4 dyne/cm, with flowing ve-
and n+0.1. Simulated rate/time performance for the vertically frac-
locities ranging from 3 to 120 ft/D (3.28 10*5 to 1.31 10*3 ft/
tured well is shown in Fig. 20, where little effect of IFT-corrected sec). The experiments combined a stabilized steady-state-flow test
relative permeabilities on well deliverability is seen. Absolutely no with a later interpretation of an unsteady-state-flow test.
effect of IFT on well deliverability is found for Lean Gas B, because Fig. 22 shows a plot of the 0.4-dyne/cm data plotted as krg vs.
sus* at all ptpd (Fig. 21). krg /kro (the 0.05-dyne/cm data are totally irrelevant to well deliver-
The potential effect of IFT on well deliverability will be greatest ability). The curve is shifted to higher krg values at higher velocities,
for rich, near-critical fluids producing initially on decline (no pla- without changing the shape of the curve significantly. Unfortunate-
teau). However, for this type of low-permeability well the IFT effect ly, 0.4 dyne/cm is probably the smallest IFT that can ever be ex-
on relative permeabilities will be only one of several major uncer- pected in Region 1 (when a well is approaching the minimum
tainties (absolute permeability, fracture dimensions and conductiv- BHFP). The net effect of high velocities at more relevant (higher)
ity, rock relative permeabilities, and oil viscosity). Practically, it will IFTs has yet to be studied experimentally; we suspect that the effect
be impossible to separate IFT effects from these other effects, and may disappear at higher IFTs found in most of Region 1, as men-
it will be difficult to sell the optimistic effect of low IFTs to proj- tioned by Henderson et al.
k
k
Conclusions
1. Gas-condensate wells producing with BHFP lower than the
dewpoint have up to three flow regions. Region 1 has a constant
flowing composition (GOR) where both gas and oil flow simulta- Fig. 23Combined velocity/IFT effect on krg as a function of gas/
neously. Most of the deliverability loss is caused by reduced gas oil capillary number for Ref. 22 data.
permeability in Region 1. Region 2 is where condensate accumu-
lates but has no mobility. Region 3 is the outer region where reser- 7. Gas/oil IFT usually has little effect on gas-condensate well de-
voir pressure is greater than the dewpoint and only gas flows. liverability.
2. Gas-condensate well deliverability can be determined accu- 8. Deliverability of horizontal gas-condensate wells is strongly
rately with a simple rate equation, Eq. 1 or 2. The multiphase affected by the kv /kh ratio. Severe deliverability loss results for a typ-
pseudopressure function is easily calculated from producing GOR ical kv /kh ratio of 0.1 (compared with the performance of the same
(composition) and PVT properties. The effect of reduced gas
horizontal well with kv /kh +1).
permeability (condensate blockage) is incorporated in the pseudo-
pressure function. All other well terms (well geometry, damage
skin, etc.) are accounted for in the productivity constant C. The Nomenclature
method is shown to work for radial, vertically fractured, and hori- Bgd + dry gas FVF, L3/L3, RB/scf
zontal wells. Bo + oil FVF, L3/L3, RB/STB
3. The multiphase pseudopressure function is calculated in three C+ gas rate constant
parts, based on the three flow regions. Region 1 pseudopressure is F+ IFT correlating parameter
calculated with the Evinger-Muskat approach, modified for gas- G+ Current surface gas in place in CVD cell, L3, scf
condensate systems. Region 2 uses the krg (So ) relationship and So (p) h+ reservoir thickness, L, ft
estimated from the liquid-dropout curve from a CVD experiment. J+ productivity index, L2/m-t, scf/D-psi
Region 3 pseudopressure is the same as that for single-phase gas. k+ absolute permeability, L2, md
4. Local grid refinement is not needed for gas-condensate wells
kr + relative permeability (generic)
in full-field models. The proposed pseudopressure method calcu-
lates well deliverability accurately in coarse-grid models, without krg + gas relative permeability
any near-well grid refinement. Examples are given for radial, verti- kro + oil relative permeability
cally fractured, and horizontal wells. kv /kh + vertical-to-horizontal permeability ratio
5. The primary relative permeability relationship affecting con- Mg + gas molecular weight, m/n, lbm/lbm mol
densate blockage (in Region 1), and thus the primary cause of re- Mo + oil molecular weight, m/n, lbm/lbm mol
duced well deliverability, is krg as a function of krg /kro . Saturation n+ exponent in IFT correction
does not enter the calculation. N+ Current stock-tank oil in place in CVD cell, L3, STB
6. Critical oil saturation, Soc , itself (i.e., kro at low oil saturations) Nc + Dimensionless viscous-to-capillary number
has no effect on gas-condensate well deliverability. p+ pressure, m/Lt2, psia