Sie sind auf Seite 1von 8

A Simple Model for Annular Two-Phase

Flow in Wellbores
A.R. Hasan, U. of Minnesota-Duluth; and C.S. Kabir, Chevron Energy Technology Co.

Summary
Annular flow is associated with production from both gascondensate and geothermal wells. Oil wells also experience it during high-gas-to-oil-ratio (high-GOR) production. The current
semimechanistic modeling approach requires estimation of film
thickness before computing frictional pressure drop as gas flows
past the wavy-liquid film surrounding the pipe wall. This study
intends to investigate this film thickness and its impact on pressure-drop computation in wellbores producing steam-water, gascondensate, and gas-oil mixtures.
Computational results show that this dimensionless liquid-film
thickness is most likely less than 0.06 in annular flow. For such
values of thin-film thickness, the computed friction factor is only
slightly higher than that estimated with a smooth-channel assumption. When the homogeneous model is used to compute pressure
gradient by ignoring the wavy-liquid film on frictional pressure
drop, good agreement is achieved with field data and with the
predictions of a semimechanistic model.
Introduction
Annular flow is dominant in gas-condensate and geothermal wells.
Oil wells also experience annular flow when high-GOR production
occurs after gas breakthrough or when gas lift is installed. In
general, the annular-flow pattern consists of a gas core in the
middle of the flow string with a thin liquid film flowing up the pipe
wall. Two issues appear to dominate the modeling needs. One
needs to estimate, first, the liquid entrainment in the gas core, and
second, the frictional resistance that the gas core experiences when
flowing past the wavy-liquid film. Note that the frictional gradient
is a very large contributor to the total pressure loss in annular flow
and therefore has obvious importance.
In the past, a few models treated this flow pattern assuming
zero slip between the two phases in the gas core. For instance, the
models of Duns and Ros (1963) and Aziz et al. (1972), who essentially adopted the Duns and Ros approach, fall into this category. Subsequently, the method of Hasan and Kabir (1988), based
on the approach of Wallis (1969), estimates both the entrainment
and the film-friction factors. However, the rigorous method of
Ansari et al. (1994) is rooted in sound modeling of film thickness
followed by accurate estimation of frictional and hydrostatic
heads. The same approach was adopted by Kaya et al. (2001). At
approximately the same time, Gomez et al. (2000) proposed a
method based on a two-fluid approach.
The intent of this study is to present an alternative approach to
modeling annular flow. We show that the liquid-film thickness is
generally too small to be of any consequence when computed with
the model of Ansari et al. (1994). The main objective is to demonstrate the application of a much simpler model with accuracy
comparable to a semimechanistic model. In fact, the authors recent study (Kabir and Hasan 2006) on gas-condensate wells has
shed some light on the possibility of simplified modeling of annular flow.

Copyright 2007 Society of Petroleum Engineers


This paper (SPE 95523) was first presented at the 2005 SPE Annual Technical Conference
and Exhibition, Dallas, 912 October, and revised for publication. Original manuscript received for review 31 May 2006. Revised manuscript received 27 August 2006. Paper peer
approved 29 August 2006.

168

Annular Flow Modeling


In annular flow, the gas phase flows through the central core of the
pipe while liquid flows along the wall as a film, as shown schematically in Fig. 1. A part of the liquid phase is entrained as
droplets in the gas core and flows at nearly the same velocity as
the gas phase. The system can be viewed as a homogeneous
flow of gas with liquid droplets through a pipe formed by the
liquid film with a diameter of d2d(1), where and (=/d)
are the liquid-film thickness and the dimensionless liquid-film
thickness, respectively.
Consider the core fluid (gas plus droplets) and assume entrained liquid droplet velocity to be same as that of the gas in the
core. Using the core fluid properties, the total pressure gradient
during annular flow can be written as

fcvc2c
dvg
dp
= gc sin +
, . . . . . . . . . . . . . . . . . (1)
+ cvg
dz
2d1 2
dz

where is the well inclination with the horizontal. An appropriate


limiting value for up to which Eq. 1 applies is 15. Using the gas
law, vg may be expressed as (ZRT/Ax p). Neglecting variation of Z
and T with axial position allows the following simplification:
ZRT dp
vg dp
dvg
=
. . . . . . . . . . . . . . . . . . . . . . . . . . . . (2)
=
dz
Ax p2dz
p dz
Combining Eqs. 1 and 2, the result is:

dp
fcvc2c
. . . . . . . . . . . . . . (3)
1 cv2g p = g c sin +
dz
2d1 2

Rearranging,

2
dp g c sin + fcvc c 21 d
=
. . . . . . . . . . . . . . . . . . . (4)
dz
1 cv2g p

The problem at hand then reduces to that of estimating the density


of the fluid in the core, c, and the friction factor, fc , for gas
flowing through a rough pipe, and the liquid film thickness, .
Note that appropriate conversion factors (e.g., gc) may be needed
to apply Eq. 4.
Homogeneous Model. For pressure-loss estimation purposes, the
homogeneous model makes the simplifying assumption that the
liquid film may be ignored and that all the liquid moves through
the core homogeneously with the gas phase with equal velocity.
Thus, in Eq. 4, is zero and
vc = vsc = vm = vsL + vsg, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (5)

c =

LvsL gvsg
+
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (6)
vm
vm

The two-phase friction factor, f=fc , is calculated with the explicit


correlation proposed by Chen (1979) as:
f=

4
d 5.0452

4 log
log
3.7065
Rem

. . . . . . . . . . . . . . . . . (7)

In Eq. 7, is defined in Appendix A and is the ordinary pipe


roughness factor, which ignores the wavy-liquid film. The Reynolds number, Rem, of the mixture is given by
Rem =

dmvm
dmvm
=
. . . . . . . . . . . . . . . . . . . . . . . . (8)
m
xg + 1 xL
May 2007 SPE Production & Operations

The values of f calculated from either of these two equations do not


differ significantly.
From geometrical considerations, 4 equals the liquid holdup,
fL. Therefore, Eq. 13 can also be rewritten as:
fc = f 1 + 75fL. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (16)
When entrainment, E, is less than 0.9, many have used the Whalley
and Hewitt (1978) expression for fc, which is given by:
fc = f 1 + 24L g1 3,

if E 0.9. . . . . . . . . . . . . . . . . . . (17)

Accurate estimates of the film thickness, , and the liquid


holdup, fL , are needed because the gas-core friction factor is very
sensitive to this entity. The approach adapted by Ansari et al.
(1994) and others to estimate is to note that the pressure gradient
for the gas core must equal that for the liquid film. In analogy to
Eq. 1, an expression for (dp/dz)L can then be written as:

Fig. 1Schematic representation of liquid film on pipe wall in


annular flow.

The mass-average viscosity for the Reynolds number of the mixture is used, where the gas mass fraction x is given by x[vsgg/
(vsgg+vsLL)].
Semimechanistic Models. Semimechanistic models, such as those
proposed by Ansari et al. (1994), Gomez et al. (2000), and Kaya
et al. (2001), attempt to incorporate a more rigorous physical description of annular two-phase flow than does the homogeneous
model. These models usually rely on the Steen and Wallis (1964)
correlation for estimating entrainment, E, the fraction of flowing
liquid entrained in the gas core. The following expression represents the graphical correlation presented by Steen and Wallis:
E = 1 e0.125vsgc1.5, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (9)
where the dimensionless gas velocity, vsgc, is given by
vsgc =

104vsgg
g L. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (10)

The core fluid superficial velocity, vsc, and in-situ core fluid
velocities are then given by
vsc = vsg + EvsL;

vc = vsc /1 22, . . . . . . . . . . . . . . . . . . (11)

and the core fluid density, c , is

c =

LEvsL gvsg
+
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (12)
vsc
vsc

For the gas-core friction factor, fc , the semimechanistic models


use the correlation proposed by Wallis (1969), which relates liquid-film roughness to the wall-friction factor, f, and the dimensionless liquid-film thickness, (/d), as follows:
fc = f 1 + 300 , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (13)
where the wall-friction factor, f, represents the friction that the core
fluid experiences when flowing through a smooth pipe (not
through a rough pipe as in the homogeneous model). The value of
f may be read off the curve for smooth pipes in a Moody friction
factor chart. Alternatively, f may be estimated using the Blasius
equation for very high Reynolds numbers (Eq. 14 for Re>50,000)
or moderately high Reynolds number (Eq. 15, 2100>Re>50,000)
as follows:
f = 0.184
f = 0.316

d vscsc
sc
d vscsc
sc

0.20

. . . . . . . . . . . . . . . . . . . . . . . . . . (14)

0.25

. . . . . . . . . . . . . . . . . . . . . . . . . . (15)

May 2007 SPE Production & Operations

dp
dz

= gL sin +

fcv2c1 2c

, . . . . (18)
8d1
128d31 23
fLFv2sL1 E2L

where fLF represents the friction between the liquid film and the
pipe wall. Note that Ansari et al. (1994) neglects kinetic head loss
to arrive at Eq. 18.
Neglecting the last term or kinetic head in Eq. 1 for pressure
gradient in the gas core and equating it with (dp/dz), Eq. 18 may be
written as:
gc sin +

fcv2c c
fLFv2sL1 E2L
= gL sin +
2d1 2
128d31 23
2
fcvc1 2c

. . . . . . . . . . . . . . . . . . . . (19)
8d1

Rearranging,
fLFv2sL1 E2L
fcv2c c
= g sinL c +
. . . (20)
2d1 1 2
128d31 23
Eq. 20 is then solved iteratively for the single unknown, . However, the solution of the nonlinear form of Eq. 20 poses some
challenges because various terms are very sensitive to . Ordinarily, the equation-solver capability of most spreadsheets can be
used to obtain a solution, provided a reasonable initial guess is
given as the starting value.
For very high gas velocities, liquid entrainment may approach
a value of unity, causing the third term in Eq. 20 to become
negligible. In addition, under such circumstances, both 1 and
12 will approach 1.0, resulting in the following simplification of
Eq. 20:
2
c
fcvsc
= 5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (21)
2dgL c

Eq. 21 will often produce nonphysical results, such as >0.5. An


explanation of such nonsensical values of is in order. When
entrainment approaches 1.0, no liquid film exists on the pipe wall,
resulting in simple homogeneous flow of a gas/droplet mixture.
Therefore, Eq. 21 truly has an upper bound when the wall has
become dry; in this case, the equation no longer applies.
Computational Results
The Tulsa U. Fluid Flow Project (TUFFP) database was used in
this research to generate solutions and test hypotheses for simplified modeling of oil wells. To simplify computations and aid comparisons, efforts were concentrated on those tests that are likely to
have annular flow in the entire well.
Three simple criteria were used to select these tests, such as the
superficial gas velocity at the transition to annular flow, the producing gas/liquid ratio, and the liquid-film thickness. The first
criterion was the presence of a superficial gas velocity, vsg, of 20
ft/sec or higher at the wellhead. With the Taitel et al. (1980)
transition criterion, all the selected 425 tests exhibited annular
flow, at least at the wellhead. Fig. 2 shows that the Ansari et al.
169

Fig. 2Acceptable fit using the Ansari et al. model for the wellhead-vsg range: 136.5 to 20 ft/sec.

Fig. 3Good fit with homogeneous model for the wellhead-vsg


range: 136.5 to 20 ft/sec.

(1994) model fitted the data with moderate success. In contrast,


Fig. 3 suggests that the homogeneous model performed somewhat
better for the same dataset.
Note that in these and subsequent comparisons, we presupposed that annular flow applies to the entire well. In other words,
discrimination of flow patterns is not made because of highvelocity flow in these wells and for simplicity. Further justification
of this approach is discussed in the context of the liquid-film
thickness computation. However, we acknowledge that such a
simple approach results in some inaccuracies in the computed
results and the resultant inferences. These inaccuracies do not
compromise the goal of this research, which is to compare the
performance of the homogeneous model with the Ansari et al.
(1994) model.
When similar comparisons are made based on a producing
gas/liquid ratio (GLR) in excess of 2,000 as the criterion for annular flow, the homogeneous model appears to perform better.
Figs. 4 and 5 show the statistical fitting of a subset of the same
database. We do not, however, recommend the use of GLR arbitrarily as a discriminating tool, because it does not tell us anything
about pressure-dependent fluid velocity.
As discussed earlier, liquid-film thickness is perhaps the most
important discriminator in annular flow. Cumulative distribution
function (CDF) of liquid-film thickness showed that a maximum
of about 6% of pipe inside diameter (ID) occurs at the well bottom,
as shown in Fig. 6. This finding is in remarkable agreement with
Wallis assessment of 5%. The most likely or p-50 -value is
approximately 3.2% in this population. Obviously, if is this small
at the well bottom, the remainder of the wellbore will experience
mist flow or no slip between the two phases.

When the distribution of at shallower depths is investigated,


the quality of the solution begins to degrade. Figs. 7 and 8 show
results for at each wells midpoint and wellhead respectively.
Although the p-50 value is somewhat lower at the midpoint than
at the well bottom, data scatter is quite evident. At the wellhead,
the scatter of the data is a reflection of the difficulty of computing
. This point is discussed further in the next section. Fig. 9 depicts
wellbore profiles of for a few wells drawn from this population.
This 379-test dataset, based on a maximum of 6%, is modeled
with both homogeneous and Ansari et al. methods. Results shown
in Figs. 10 and 11 suggest that the notion of mist flow or flow
homogenization is highly consistent with these models. We think
that the difficulty of computing manifests in terms of scatter, as
shown in Fig. 11.

So far, the discussion has focused on the applicability of the homogeneous model to gas/oil systems. This section examines
whether this modeling approach can be extended to other systems,
such as stream/water flow in geothermal wells and gas flow in
condensate wells.
Data from A-10 well, which was recently presented by Garg
et al. (2004), show a very good correspondence between the homogeneous and Ansari et al. models for field data. Fig. 12 shows
the quality of the match, and Appendix B presents a sample calculation at the wellhead. As a further illustration of the quality of
agreement, Fig. 13 shows error (measured pressure minus model
pressure) as a function of depth. The maximum error is approximately 26 psi, with both models tracking each other well.

Fig. 4Good fit with homogeneous model for the GLR range:
1.12E62,000 scf/STB.

Fig. 5Ansari et al. model shows large scatter for the GLR
range: 1.12E62,000 scf/STB.

170

Homogeneous Model in Other Systems

May 2007 SPE Production & Operations

Fig. 6Range of liquid-film thickness at the well bottom for 379


tests.

Fig. 7Range of liquid-film thickness at the midwell depth for


379 tests.

As discussed in Appendix B, the homogeneous model estimates


higher static head than the Ansari et al. model, but underestimates
frictional head, thereby yielding very similar results.
Earlier, we (Kabir and Hasan 2006) showed that gas/
condensate wells lend themselves to homogeneous modeling over
a wide range of operating conditions. Fig. 14 shows that three
independent datasets, encompassing a very wide range of flowing
bottomhole pressures, liquid contents (1.3 to 256 STB/MMscf),
and flow rates (0.5 to 30 MMscf/D), are well represented by the
homogeneous model.

pressure gradients requires knowledge of: (1) the extent of liquid


entrainment, E, (2) the liquid-film friction factor, ffL , (3) the gas/
liquid interface friction gradient, and (4) the core fluid velocity,
which depends on the core volume fraction, that is, on film thickness.
Estimating each of these parameters introduces significant uncertainty. For instance, entrainment, which is needed for calculating core fluid density and liquid film thickness, is a particularly
difficult parameter to estimate. Collier (1973) mentioned that the
available correlations for E do not account for the influence of a
number of variables, especially tubing diameter. Considering the
difficulty of gathering accurate entrainment data (using invasive
probes) to provide a basis for the correlations, the lack of confidence in the accuracy of E estimates becomes easy to understand.
Wallis (1969) suggested that the usual range for liquid holdup
in annular flow is from 0 to 0.2, with a corresponding dimensionless liquid-film thickness (/d) ranging from 0 to 0.05. Estimating
friction between the pipe wall and such thin liquid films, which can
be in laminar, transition, or turbulent flow, introduces another
source of inaccuracy.
Estimating the gas/liquid interface friction gradient poses two
problems: first, estimating the interface friction factor, fc, and second, estimating the difference in velocities between gas and liquid
at the interface. While the empirical correlation of Wallis (1969)
for fc has achieved general acceptance, its validity can be established only indirectly because direct data cannot be gathered. Similarly, while the bulk velocities of the liquid film and the core fluid
can be estimated with good accuracy, the local values of these entities
at the liquid film/core fluid interface are often difficult to estimate.
The various exponents of the terms containing in the denominator of Eq. 20 pose yet another problem. Mathematically speaking, several possible values of may satisfy Eq. 20, of which more
than one may be real and positive. For the problems investigated,

Discussion
As stated earlier, a number of sophisticated models attempt to
describe the physics of annular flow much more accurately than
does the simple homogeneous model. More than three decades
ago, Steen and Wallis (1964) noted that for horizontal annular flow
at very high flow rates of both phases, almost all the liquid is
entrained and the pressure drop and gas-volume fraction are well
represented by the homogeneous model. For vertical annular twophase flow, the homogeneous model represents the frictional gradient better than the separated (Lockhart-Martinelli) flow model,
although gas-volume fraction is somewhat overestimated (Steen
and Wallis 1964). Therefore, the suitability of homogeneous
model for annular flow has been recognized for a long time.
Our rationale for adopting the homogeneous model for annular
flow stems primarily from the simplicity of the approach and the
uncertainty inherent in the sophisticated approaches, such as the
Ansari et al. (1994) model. The Ansari et al. model equates the
pressure gradient for the gas core with that for the liquid film to
compute liquid-film thickness, and then goes on to calculate gasvolume fraction and pressure drop. Unfortunately, estimating these

Fig. 8Range of liquid-film thickness at the wellhead for 379


tests.
May 2007 SPE Production & Operations

Fig. 9Distribution of liquid-film thickness along well depth.


171

Fig. 10Performance of homogeneous model based on film


thickness, 379 tests.

only one physically acceptable (0.0<<0.5) solution for was


observed, perhaps because the likely values of lie between 0 and
approximately 0.05. The possibility of multiple solutions for is
still an unsettling question.
In reality the liquid film ascending up a tubing wall has a wavy
interface, with a thickness fluctuating over time, a characteristic
that gives rise to a high friction factor. The contention here is that
the film thickness is a statistical average property, not a single
value attributable to a particular location in the well at all times.
For values as small as 0.003, the average value may not retain
as much physical sense as might be desired.
The preceding discussion regarding the complexity of semimechanistic models and issues surrounding the estimation of film
thickness, friction factor, and other parameters pertains to any
annular two-phase-flow problem. In other words, annular flow in
geothermal and gas/condensate wells is no different than in oil
wells. This observation leads to the recommendation to adopt the
homogeneous model for all systems that experience annular flow.
All the examples presented in this study pertain to vertical flow.
To account for flow in deviated wellbores, we think that the angular correction as indicated in Eq. 1 is sufficient with the limit
indicated. This is because the difference in film thickness owing to
gravity may not be sufficient to warrant separate treatment. After
all, the most likely value of is only 3.2% of the pipe ID. This
approach is supported by the work of Steen and Wallis (1964),
who reported success of the homogeneous model even in horizontal pipes.
Conclusions
1. The proposed homogeneous modeling approach for annular
flow shows accuracy comparable with existing semimechanistic
models, yet retains simplicity and transparency. Verification of
the homogeneous modeling approach has been performed using
data from wells involving gas/oil, steam/water, and gas/
condensate flows.

Fig. 12Performance of two models for A-10 geothermal well


(Garg et al. 2004).
172

Fig. 11Performance of Ansari et al. model based on film thickness, 379 tests.

2. The liquid-film thickness criterion appears the best approach to


characterizing annular flow. The maximum film thickness is about
6% of the pipe diameter, with a most likely value of about 3.2%.
These values are in good agreement with Wallis 5% criterion.
Nomenclature
Ax cross-sectional area for flow, ft2
d tubing ID, ft
E entrainment factor, defined as the fraction of total
liquid flowing as entrained droplets in the gas
core during annular flow, dimensionless
f Moody friction factor for smooth pipes,
dimensionless
fc Moody friction factor for the core fluid in annular
flow, dimensionless
fL liquid holdup, dimensionless
fLf Moody friction factor for the liquid film in
annular flow, dimensionless
g gravitational acceleration, ft/sec2
gc conversion factor, 32.17 (lbm-ft)/lbf/sec2
h enthalpy of steam/water mixture, Btu/lbm
hg enthalpy of steam, Btu/lbm
hL enthalpy of water, Btu/lbm
p pressure, psig
dp/dz total pressure gradient, psi/ft
(dp/dz)A accelerational pressure gradient, psi/ft
(dp/dz)F frictional pressure gradient, psi/ft
(dp/dz)H hydrostatic pressure gradient, psi/ft
T absolute fluid temperature, oR.
vc in-situ velocity of the core fluid, ft/sec

Fig. 13Error in A-10 well modeling.


May 2007 SPE Production & Operations

Fig. 14Performance of the homogeneous model for 167 gas/


condensate well tests (Kabir and Hasan 2006).

vsc
vsL
vsg
vsgc
vm
dv/dz
x

z
Z

g
L
sc

g
L
m

superficial velocity of the core fluid, ft/sec


superficial velocity of liquid, ft/sec
superficial velocity of gas, ft/sec
critical velocity of gas, ft/sec
velocity of gas/liquid mixture, ft/sec
velocity gradient, ft/sec
mass fraction [vsgg/(vsgg+vsLL)] or steam
quality{(hhL)/(hghL)}
vertical well depth, ft
gas-law deviation factor, dimensionless
liquid-film thickness, ft
dimensionless thickness of the liquid film (%
/d)
well inclination with horizontal, degrees
gas viscosity, cp
liquid viscosity, cp
core fluid viscosity, cp
pipe-roughness factor, ft
mixture density of the fluid in the gas core,
lbm/ft3
gas density, lbm/ft3
liquid density, lbm/ft3
mixture density, lbm/ft3
surface tension, lbm/sec2

Acknowledgments
We thank Chevrons Bolaji Oloyede for providing computational
assistance. We are also indebted to Chevron management for permission to publish this work.

National Engineering and Environmental Laboratory, Report EXT-0401760.


Gomez, L.E., Shoham, O., Schmidt, Z., Chokshi, R.N., and Northug, T.
2000. Unified Mechanistic Model for Steady-State Two-Phase Flow:
Horizontal to Vertical Upward Flow. SPEJ 5 (3): 339350. SPE65705-PA. DOI: 10.2118/65705-PA.
Hasan, A.R. and Kabir, C.S. 1988. A Study of Multiphase Flow Behavior
in Vertical Wells. SPEPE 3 (2): 263272; Trans., AIME, 285. SPE15138-PA. DOI: 10.2118/15138-PA.
Kabir, C.S. and Hasan, A.R. 2006. Simplified Wellbore-Flow Modeling in
Gas/Condensate Systems. SPEPO 21 (1): 8997. SPE-89754-PA. DOI:
10.2118/89754-PA.
Kaya, A.S., Sarica, C., and Brill J.P. 2001. Mechanistic Modeling of TwoPhase Flow in Deviated Wells. SPEPF 16 (3): 156165. SPE-72998PA. DOI: 10.2118/72998-PA.
Steen, D.A. and Wallis, G.B. 1964. Pressure Drop and Liquid Entrainment
in Annular-Two-Phase Flow. AEC Report NYO-3114-2.
Taitel, Y., Barnea, D., and Dukler, A.E. 1980. Modeling Flow Pattern
Transition for Steady Upward Gas-Liquid Flow in Vertical Tubes.
AIChE J. (46): 345354.
Wallis, G.B. 1969. One-Dimensional Two-Phase Flow. New York:
McGraw-Hill.
Whalley, P. and Hewitt, G.F. 1978. The Correlation of Liquid Entrainment
Fraction and Entrainment Rate in Annular Two-Phase Flow. UKAEA
Report, AERE-R9187, Harwell, Oxfordshire, England.

Appendix AOil Well Example


TUFFPs #918 is an 11,130-ft vertical well producing 35.4o API
oil at 2,181 stock tank barrels per day (STB/D) and 17 STB/D of
water through a 2.992-in. ID tubing. The gas/oil ratio (GOR) is
1,601 scf/STB, and the gas specific gravity is 0.815. The following
fluid properties and other parameters are computed at the wellhead
where pressure is 45.3 psig:
vsL = 1.79 ft sec

vsg = 122.7 ft sec

vm = vsL + vsg = 124.49 ft sec

L = 52.95 lbm ft3

g = 0.25 lbm ft3

d = 2.992 12 = 0.2493 ft

L = 25.6 cp

g = 0.01019 cp

= 28.4 dynes cm = 0.0626 lbm sec2


The Homogeneous Model.
Quality, x = vsgg vsgg + vsLL = 0.247.
Homogeneous gas volume fraction = vsg vm = 0.986
Homogeneous mixture density,

m = 0.986g + 0.01452.95 = 1.011 lbm ft3


Mixture viscosity,

References
Ansari, A.M., Sylvester, N.D., Sarica, C., Shoham, O., and Brill, J.P. 1994.
A Comprehensive Mechanistic Model for Upward Two-Phase Flow in
Wellbores. SPEPF 9 (2): 143151; Trans., AIME, 297. SPE-20630PA. DOI: 10.2118/20630-PA.
Aziz, K., Govier, G.W., and Fogarasi, M. 1972. Pressure Drop in Wells
Producing Oil and Gas. J. Cdn. Pet. Tech. (79): 3848.
Chen, N.H. 1979. An Explicit Equation for Friction Factor in Pipe. Ind.
Eng. Chem. Fundamentals 18 (3): 296297.
Collier, J.G. 1973. Convective Boiling and Condensation. 1st ed. New
York: McGraw-Hill, 8182.
Duns, H., Jr. and Ros, N.C.J. 1963. Vertical Flow of Gas and Liquid
Mixtures in Wells. Paper presented at the Sixth World Petroleum Congress, Tokyo, Japan.
Garg, S.K., Pritchett, J.W., and Alexander, J.H. 2004. Development of
New Geothermal Holdup Correlations Using Flowing Well Data. Idaho
May 2007 SPE Production & Operations

m = xg + 1 xL = 0.2470.01019
+ 0.75325.6 = 19.286 cp
Rem = vmmd m
= 124.51.0110.249319.286 0.000672 = 2,421
Using Eq. 7, the result is
f=

4
d
5.0452
4 log

log
3.7065
Re

= 0.0468.

Note that in Eq. 7,


=

7.149
d1.1098
+
2.8257
Re

0.8981

173

A pipe roughness factor of 0.0018 in. was used, resulting in a value


of 0.000737 for /d.
dp dzF = fmvm2m 2gc d 144 = 0.317 psi ft.
dp dzH = m g sin gc 144 = 0.0070 psi ft.
dp dzA = mvmdvm dz gc 144 = 0.0070 psi ft.
dp dz = 0.3241 psi ft total pressure gradient at wellhead.
The Ansari et al. Model. For the wellhead conditions, the superficial gas velocity requirement for annular flow to exist is given by
the Taitel et al. (1980) model as:
vsg 3.1 g L g 2g 1 4
= 3.1 32.2 0.062652.95 0.253 0.06251 4
= 19.78 ft sec.

Appendix B Geothermal Example


Data reported by Garg et al. (2004) are used below to illustrate a
sample calculation for a geothermal well. Well A-10 is a 4,265-ft
inclined well which is vertical for the first 1,427 ft from the wellhead. It has an ID of 15.112 in at the wellhead. The wells diameter
changes to 12.4 in at a depth of 1,427 ft, and it exhibits three other
deviations from the vertical at various depths.
The well delivers 1.151106 lbm/hr of a steam-water mixture at
the wellhead with an enthalpy of 461.8 Btu/lbm. The following
values of fluid properties, obtained from interpolation of steam
tables, and of other parameters are computed at the wellhead,
where the pressure is 166.9 psia.
Saturated steam/water property values:
hg = 1196.3 Btu lbm

Therefore, actual flow at the wellhead is obviously in annular regime.


To calculate the liquid-film thickness, , Ansari et al. used Eq.
20. With vcvsc /(12 )2, Eq. 20 may be rewritten in the following
form, which is suitable for iterative calculations:
d=

ent to E, which is estimated using an empirical approach, renders


the models accuracy open to debate.

hL = 339.3 Btu lbm

L = 55lb ft

g = 0.3636 lb ft3

L = 0.144 cp

g = 0.015 cp

= 41 dynes cm = 0.09032 lbm sec2

2
fcvsc
c

d = 15.112 in

21 1 25gL c
2
fLFvsL
1 E2L

12831 23g sin L c

. . . . . . . . . . . . . . . . . . . . (A-1)

Film-friction factor, fc, in Eq. A-1 is calculated using the Wallis


correlation, Eq. 13 (or the Whalley and Hewitt correlation, Eq. 17,
when E<0.9), in combination with Eq. 13, fc=f (1+300). Thus, the
value of fc strongly depends upon .
In addition to fc, estimating using Eq. A-1 requires an estimate of entrainment, E, using the Steen and Wallis correlation.
For this example, vsgc = 9.279; E = 0.6218.
Using Eq. 11, the core fluid superficial velocity, vsc, and in-situ
core fluid velocities are calculated and then used to obtain core
fluid density, c , using Eq. 12, with the result:

c = 0.7271 lbm ft .

Ax = 1.247 ft2

qms = 1.151 106 lbm hr


x = h hL hg hL
= 461.8 339.3 1196.3 339.3
= 0.143
qg = qms x g
= 1.151 106 0.0249 0.3636 3600
= 125.7 ft3 sec
vsg = qg Ax = 100.9 ft sec
vsL = qL Ax = 4.0 ft sec

The Homogeneous Model.

With these property and parameter values, Eq. A-1 was solved
iteratively for and fc, using a value for diameter d of 0.2493 ft.
Values of 0.00640010 and fc0.100 were obtained.

Gas-volume fraction = vsg vm = 0.962


Mixture density, m = 0.962 g + 0.038L

With these values,


dp dzF = fcvc2c 2gcd1 2 144 = 0.5118 psi ft.
dp dzH = c g sin gc 144 = 0.0050 psi ft.

= 2.442 lbm ft3


Mixture viscosity, m = x g + 1 xL
= 0.36360.144 + 0.63640.015

dp dzA = m vmdvm dz gc /144 = 0.0070 psi ft.

= 0.1255 cp

dp dz = 0.5239 psi ft total pressure gradient at wellhead.


Therefore, compared to the Ansari et al. model, the homogeneous
model underestimates the frictional gradient while overestimating
the static head. However, it should be noted that the two terms on
the right side of Eq. A-1 (for 00640010 and fc0.100) are:
fcvc2c
= 55.3389,
21 1 2gL c
2
fLFvsL
1 E2L

12831 23g sin L c

= 55.0897.

Thus, the small value of 0.2493 ft for the tubing diameter was
matched by subtracting two large values, which could be a source
of significant inaccuracy. Note that a 10% higher estimate for E for
this case leads to an decrease of approximately 8% in the film
thickness and interface-friction factor estimates (0.00595 and
fc0.086), leading to a 3% decrease in frictional pressure gradient
{(dp/dz)F0.4937}. This high sensitivity of the pressure gradi174

Rem = vmmd m = 104.92.4421.26 0.1255 0.000672


= 3,830,066.
Using Eq. 7, the result is
f=


4 log

4
5.0452
d

log
3.7065
Re

= 0.01281.

A pipe-roughness factor of 0.0018-in was used, resulting in a


value of 0.000737 for /d.
dp dzF = fmvm2m 2d = 4.256 psf ft = 0.0295 psi ft.
dp dzH = m g sin = 2.445 psf ft = 0.0168 psi ft.
dp dzA = m vm dvm dz = 0.004 psi ft.
dp dz = 0.0496 psi ft total pressure gradient at wellhead.
May 2007 SPE Production & Operations

The Ansari et al. Model. For this case, the superficial gas velocity
above which annular flow exists is given by the Taitel et al.
(1980) model:
vsg 3.1 g L g 2g1 4
= 3.1 32.2 0.0903255 0.3636 0.363621 4
= 18.26 ft sec.
Thus, flow at the wellhead is in annular regime.
As in the case of Well 918, liquid-film thickness, , is calculated using the Ansari et al. model with the following expression
for well diameter that satisfies the pressure gradient for the gas
core and the liquid film:
2
c
fcvsc

2 1 1 2g sin L g

With these values, pressure gradients can be computed as follows:


dp dzF = fcvc2c 2d1 2 = 0.038 psi ft.
dp dzH c g sin = 0.0117 psi ft.
dp dzA = m vm dvm dz = 0.002 psi ft.
dp dz = 0.0527 psi ft total pressure gradient at the wellhead.
In this example, as in the last one, the homogeneous model
estimates a higher value for static head, which is compensated for
by the lower estimate of the frictional head. Overall, the homogeneous model underestimates the total pressure gradient by 2.15%
compared with the Ansari et al. model. This trend is true for the
entire wellbore and is evident in Fig. 12. In general, however, the
computations show that the differences in pressure-drop estimation
between the homogeneous and Ansari et al. models are well within
the limits of engineering accuracy.

fLFv2sL1 E2L
12831 23g sin L g

= d. . . . . . . . . . . . . . . . . . . (B-1)

As before, the expression fcf (1+300), and Eqs. 9 and 10 can be


used to calculate entrainment and critical-gas velocity as follows:
E = 0.6227; vsgc = 30.5 ft sec.
The core fluid superficial velocity, vsc , and in-situ core fluid
velocities are calculated using Eq. 11. The core fluid density, c ,
is then calculated from Eq. 12.

c =

LEvsL gvsg
+
vsc
vsc

c = 1.676 lbm ft3.


With these property/parameter values, 0.00826 and
fc0.023.

May 2007 SPE Production & Operations

Rashid Hasan is professor and head of chemical engineering


at the U. of Minnesota-Duluth. He has 30 years of teaching and
research experience in fluid and heat flows in wellbores and
pressure-transient testing. He has consulted with, and offered
short courses for, oil operating and service companies. He has
also worked with NASA on various aspects of multiphase flow
and thermohydraulic transients. He holds MS and PhD degrees
in chemical engineering from the U. of Waterloo, Canada.
Hasan has published extensively and has served various SPE
committees, including editorial review for SPEPF and SPEJ.
Shah Kabir is a petroleum-engineering consultant with Chevron
Energy Technology Co. in Houston. He received commendation as an outstanding technical editor five times for two different journals, and also received the SPE Western Regions
Service Award in 2002. He has 30 years experience in the oil
industry. He holds a masters degree in chemical engineering
from the U. of Calgary, Canada. Kabir has published extensively, including the 2002 SPE book entitled Fluid Flow and Heat
Transfer in Wellbores. He has served on various SPE committees,
including editorial review committees for SPEPF, SPEREE, and
SPEJ. He is an SPE distinguished lecturer for 200607.

175

Das könnte Ihnen auch gefallen