Escolar Documentos
Profissional Documentos
Cultura Documentos
1
Department of Petroleum Engineering, Stanford University, Stanford, CA 94305-2220, USA
2
Schlumberger GeoQuest, 11 Foxcombe Court, Wyndyke Furlong, Abingdon, Oxfordshire, OX14 1DZ, UK
Abstract
Drift-flux modeling techniques are commonly used to represent multiphase flow in pipes and wellbores. These
models, like other multiphase flow models, require a number of empirical parameters. In recent publications we
have described experimental and modeling work on steady-state multiphase flow in pipes, aimed at the
determination of drift-flux parameters for large-diameter inclined wells. This work provided optimized drift-flux
parameters for two-phase water-gas and oil-water flows and a unified model for three-phase oil-water-gas flow for
vertical and inclined pipes. The purpose of this paper is to extend this modeling approach to transient countercurrent
flows, as occur in pressure build-up tests when the well is shut in at the surface. The experiments on which the
steady-state models are based also include transient flow data obtained after shutting in the flow by fast acting
valves at both ends of the test section. We first compare predictions from the existing steady-state drift-flux model
to transient data and show that the model predicts significantly faster separation than is observed in experiments. We
then develop a two-population approach to account for the different separation mechanisms that occur in transient
flows. This model introduces two additional parameters into the drift-flux formulation – the fraction of
bubbles/droplets in each population and a drift velocity multiplier for the small bubbles/droplets. It is shown that the
resulting model is able to predict phase separation quite accurately, for vertical and inclined pipes, for both water-
gas and oil-water flows. Finally, the model is applied to interpret a well test in which transient countercurrent
wellbore flow effects are important. It is demonstrated that … (to be added by Jon).
Keywords: Transient, Drift-flux, Countercurrent, Two-phase, Three-phase, Large diameter, Inclined, Steady state,
Water-gas, Oil-water, Oil-water-gas, Wellbore, Bubble, Shut-in, Phase redistribution, well testing, two-population
model
2 H. SHI, J.A. HOLMES, L.J. DURLOFSKY, K. AZIZ
The drift-flux technique is well-suited for modeling to investigate the ability of the drift-flux formulation
multiphase wellbore flow in reservoir simulators. to model the transient flow that occurs after the test
This is because the calculation of phase velocities is section is closed at both ends by fast-acting valves.
relatively simple and efficient and the equations are During this period, phases separate through
simulators. However, the drift-flux model includes a the flow that occurs when a well is shut in (as in a
number of empirical parameters, which need to be well test), so the ability to model it could improve
tuned to the particular conditions being modeled. numerical well test interpretation procedures. The
Prior to our recent work, the parameters reported drift-flux formulation is capable of modeling
in the literature and used in commercial simulators countercurrent flow as it describes the slip between
were (typically) determined from experimental data two fluids as a combination of a profile effect and a
in small-diameter pipes (5 cm or less) and might drift velocity. Our previous analysis was for steady-
therefore not be appropriate for large-diameter state cocurrent flow, but by modeling phase
wellbores. In previous publications1,2,3, we described separation we can test the applicability of the drift-
appropriate for large-diameter vertical and deviated example, flooding phenomena in countercurrent gas-
wells. This was based on steady-state in situ volume liquid annular flow) have been investigated
fraction data for a variety of water-gas, oil-water and previously4,5, compared to steady-state cocurrent
oil-water-gas flows in a 15 cm diameter, 11 m long flow, relatively few studies involving countercurrent
pipe at 8 deviations ranging from vertical to near- flow have been conducted. Transient cocurrent flows
horizontal1. We showed that the optimized have not received very much attention either.
parameters significantly improved in situ volume Therefore, not surprisingly, available data for
fraction predictions for two and three-phase flows2,3 transient countercurrent multiphase flow in large-
compared to predictions based on parameters derived scale systems are essentially nonexistent. Following
countercurrent and transient cocurrent flows, with liquid-gas countercurrent flows for stratified and slug
Steady-state countercurrent flows. Taitel and 5.7−12.1 cm and the maximum pipe inclination was
Barnea6 proposed models for three typical (bubble, xxx° from horizontal. Ghiaasiaan et al.15 conducted
slug and annular) vertical gas-liquid countercurrent vertical and deviated gas-liquid experiments in a 1.9
flow patterns. An additional flow pattern (semi- cm diameter pipe. The deviations were set to be 0°,
annular) was subsequently reported by Yamaguchi
28-30°, and 60-68° from vertical. In an attempt to
7,8
and Yamazaki from their experiments with vertical
apply the drift-flux model for hold up calculations for
water-air systems in 4 and 8 cm diameter pipes.
slug flow, they adjusted both the profile parameter C0
9
Hasan et al. developed a drift-flux model for
and the drift velocity Vd for different liquid
vertical countercurrent bubble and slug flow. The
viscosities to match their data.
value of the profile parameter C0 (discussed in detail
Zhu and Hill16 and Zavareh et al.17 performed oil-
below) was found to be 2.0 for bubble flow. They
water tests in an 18.4 cm diameter acrylic pipe at
10 11
concluded that the Harmathy and Nicklin
deviations of 0°, 5°, and 15° from upward vertical.
correlations for small bubbles and Talyor bubbles
Ouyang18,19 classified oil-water countercurrent flow
were valid for countercurrent flows. However, these
into five categories and developed models to compute
conclusions were based on experimental data with
the phase in situ volume fractions and pressure drop.
maximum mixture velocities of only 0.5 m/s. Kim et
His model predictions agreed well with the
12
al. also found that their experimental data from a 2
experimental data from Zhu and Hill16.
cm diameter vertical pipe were well fitted with the
Almehaideb et al.20 presented a coupled
11
drift-flux model with Nicklin’s correlation.
wellbore/reservoir model to simulate three-phase oil-
However, we are not aware of any published studies
water-gas countercurrent flow in multiphase injection
10 11
validating the Harmathy and Nicklin correlations
processes. Both a two-fluid model and a simple
for large-diameter, high flow rate liquid-gas systems.
mixture/homogeneous model were implemented for
Inclined countercurrent data are very limited.
wellbore flow. This comprehensive model considered
13,14
Johnston developed a semi-empirical model for
a black-oil system, in which the oil and water phases
4 H. SHI, J.A. HOLMES, L.J. DURLOFSKY, K. AZIZ
Transient cocurrent flows. Asheim and GrØdal21 applied to countercurrent steady-state or transient
used a modified steady-state drift-flux model to flow, specific flow regimes, such as bubble and slug
predict holdup in a transient vertical oil-water flow, were considered. Thus, a comprehensive drift-
system. The pipe used in the experiment was 4.3 cm flux model for such systems has yet to be presented.
in diameter. To investigate the performance of two- Furthermore, the Harmathy10 correlation, which is
phase transient flow models, Lopez et al.22,23 based on the single bubble rise velocity in a stagnant
considered numerical simulations using OLGA liquid, is commonly used to calculate drift velocity.
(based on a two-fluid model), TACITE (based on a In this type of correlation, all the gas bubbles/oil
drift-flux model) and TUFFP (based on a two-fluid droplets are considered to rise at the same velocity. In
model) against both laboratory and field data. They practical cases, however, all flow regimes can exist
concluded that all three models could match the simultaneously in the wellbore, with more than one
transient data from laboratory tests. However, only population of bubbles and droplets. We would expect
OLGA and TACITE were capable of simulating real different drift velocity mechanisms for
transient flows in long, large-diameter pipelines, with bubbles/droplets of different sizes. To apply the drift-
TACITE providing more accurate predictions than flux concept to transient countercurrent flows,
As indicated above, models for transient different sizes, as we will demonstrate below.
countercurrent phase separation are useful for the This paper proceeds with a brief description of the
interpretation of well tests (the models of experimental setup and some sample transient data
Almehaideb et al.20 and Hasan and Kabir24 can be for two-phase water-gas and oil-water systems and
applied under limited conditions). However, the three-phase oil-water-gas flows. The drift-flux model
amount of published transient countercurrent data for used in this work is then reviewed. It is shown that
small-diameter, vertical pipes is quite limited. To our predictions of water-gas and oil-water separation
knowledge, there has been no published data for during transient flow are not adequately modeled
large-diameter, inclined pipe, transient countercurrent using the steady-state drift-flux parameters. A two-
5 DRIFT-FLUX MODELING OF TRANSIENT MULTIPHASE FLOW IN WELLBORES 5
population drift-flux model is then proposed and ends with fast-acting valves. These two-valves,
evaluated for two-phase flows. Finally, the which were normally open, were simultaneously
application of the transient model to phase separation closed to trap the fluid instantaneously (the incoming
in a well during a build-up test is discussed. fluids were led to a bypass system to minimize water
The detailed experimental work was described in water fraction. The probes were placed perpendicular
Oddie et al.1 Sample data for steady-state two-phase to the pipe axis and positioned at 1, 2, 3, 4, 5, 6, 7,
water-gas and oil-water flows, and three-phase oil- 7.75, 9 and 10 m along the test section. These probes
water-gas systems were shown in our previous were one source for determining the steady-state in
modeling work2,3. In this paper we briefly explain the situ volume fraction. This quantity was also
transient data, which will be used for the transient measurements and measurement of the final position
flow model. of the interface after the fluids settled to their final
Experimental setup. The test apparatus used in this flow data during phase separation after shut-in.
diameter of 15 cm. Experiments were performed with Transient data. In this study, vertical flows are
kerosene, tap water and nitrogen. The viscosity of emphasized in this study because separation of the
the oil is 1.5 cP at 18°C and the density is 810 kg/m3. phases is generally the slowest in vertical pipes,
Tests were conducted with pipe inclinations of 0° though deviations of 5°, 45°, 70°, 80°, 88° are also
(vertical), 5°, 45°, 70°, 80°, 88°, 90° (horizontal), and considered. The flow rate ranges for the water-gas
92° (downward 2°). Data at 90° and 92° flows were tests are: 2.0 m3/h ≤ Qw ≤ 100.0 m3/h and 2.6 m3/h ≤
strongly impacted by end effects1 and were therefore Qg ≤ 72.2 m3/h. The tests for oil-water flow were
not used for the determination of model parameters. conducted in the range of 2.0 m3/h ≤ Qo ≤ 40.0 m3/h
The test section, shown schematically in Fig. 1, and 2.0 m3/h ≤ Qw ≤ 130.0 m3/h. For oil-water-gas
was of clear acrylic pipe that could be closed at both flow, the data are in the range of 2.0 m3/h ≤ Qo ≤ 40.0
6 H. SHI, J.A. HOLMES, L.J. DURLOFSKY, K. AZIZ
m3/h, 2.0 m3/h ≤ Qw ≤ 40.0 m3/h, and 1.8 m3/h ≤ Qg ≤ register nonzero h/D at the end of the transient. This
Three sets of transient data are shown in Figs. 2-4 and provides an estimate of the error associated with
to illustrate the probe response with time for vertical the probe data.
flows of water-gas, oil-water and oil-water-gas, Fig. 3 shows the transient profile of a vertical oil-
respectively. The figures show dimensionless water water test. The water and oil flow rates are almost the
depth (h/D) with h/D = 0 corresponding to the bottom same for this test (Qo =40.2, Qw =40.4), and the flow
of the pipe and h/D = 1 to the top of the pipe. Each rates are relatively high. For this case, oil and water
figure represents the probe responses for a particular were observed to be totally mixed to form a
set of Qo, Qw, Qg. homogeneous phase. The shut-in water volume
Both steady-state pre-shut-in and transient data fraction value is 51%, which confirms a
for a water-gas test are plotted in Fig. 2. Fig. 2 (a) homogeneous flow pattern with the flowing volume
shows steady state data over a ten second interval. fraction equal to the in situ volume fraction. An
The response from each probe varies in time as the interesting phenomenon is apparent in Fig. 3.
probe is subjected to different flow conditions. The Though the pipe is eventually half filled with water
observed flow pattern for this test is elongated (water at the bottom and oil at the top), probes 1–5,
bubble. The flow is statistically steady and most of which are eventually immersed in water, reach their
the oscillations are around an h/D value of 0.4–0.5. final state more quickly than probes 6–10, which are
The shut-in water volume fraction ( α w ) is 49% for finally immersed in oil. This phenomenon occurs due
Fig.2 (b) shows the electrical probe signals from compared to oil-in-water emulsions, as discussed in
the time of shut-in to a time after the phases are Oddie et al.1
completely settled. The settling time for this case is An oil-water-gas test is displayed in Fig. 4. The
water and oil flow rates are the same for this test as
around 50 seconds. Since αw = 49%, the profiles of
for the oil-water test shown in Fig. 3. The flow
probes 1–5 reach h/D = 1.0 as they are fully
pattern here was elongated bubble/slug. The
immersed in water, while probes 6–10 are totally in
relatively high gas flow rate (26.2 m3/h) has very
the gas phase. Note that signals from probes 6–10
7 DRIFT-FLUX MODELING OF TRANSIENT MULTIPHASE FLOW IN WELLBORES 7
little effect on the overall flow. Compared with the review both the original26 and optimized liquid-gas
oil-water vertical flow case (Fig. 3), the settling time and oil-water models before illustrating the
is almost the same for this three-phase flow case. The performance of the steady-state models for transient
expectation was that the settling time for this three- flows. The emphasis here is on vertical flows, though
phase transient process would be longer due to the deviated flows are also considered.
smaller droplets. Similar setting times may be Liquid-gas flow. Zuber and Findly25 correlated
observed because of complex emulsion behaviors actual gas velocity Vg and mixture velocity Vm using
that occur for the oil-water system around the phase two parameters, C0 and Vd:
inversion point are more difficult to separate, leading divided by total pipe area) and α g is the gas in situ
From the sample data discussed above, we can depends on the use of appropriate values for C0 and
extremely complicated, especially for oil-water and In the original (Eclipse26) model, C0 generally
oil-water-gas systems. Our goal is to develop a varies from 1.0 to 1.2, so we have
relatively simple model for these systems that is 1.0 ≤ C 0 ≤ 1.2 (2)
consistent with our previous models for steady-state and Vd is computed via:
flow.
(1 − α g C 0 ) C 0 K (α g ) Vc (3)
Vd = ⋅ m(θ )
ρg
α g Co + 1 − α g C0
ρl
Steady-state drift-flux models
The original26 and optimized steady-state drift-flux where K (α g ) = 1.53 C0 when α g ≤ a1 and
models for two-phase water-gas, oil-water and three- K (α g ) = K u ( Dˆ ) when α g ≥ a2 . Parameters a1 and a2
the dimensionless pipe diameter D̂ . The dependency where Vo is the in situ oil velocity and Vl is the liquid
K u ( Dˆ ) on D̂ is given in Shi et al.2 Vc is called the mixture velocity. The original value for C0′ is in the
characteristic bubble rise velocity, which was same range as C0 for liquid-gas flows:
the profile parameter, for which we obtain C 0 = 1.0 . flow, the optimized value of m′(0) is not much
This lower value of C0 directly leads to a much different from its original value of 1.0. However, the
higher Vd value. For example, the optimized value of the exponent n is reduced from 2.0 to 1.0.
deviation effect is Compared with the original model, this makes Vd′
m(θ ) = m(0)(cosθ ) 0.21 (1 + sinθ ) 0.95 (5) decrease linearly and much more rapidly with
the optimized Vd value is 1.85 times higher than the During transient flow after shut-in, there is no net
original Vd for vertical liquid-gas flow. flow, so Vm = 0. Hence there is no effect of the profile
Oil-water flow. The general form of the drift-flux depends only on the drift velocity. Therefore, the key
model applied to oil-water flows is: to modeling the transient process is to model the drift
9 DRIFT-FLUX MODELING OF TRANSIENT MULTIPHASE FLOW IN WELLBORES 9
velocity accurately. reasonably well, but the predicted speed of the water
Comparison with experimental observations. model predicts even higher velocities for both the gas
Eclipse26 applies the same drift-flux models (the and water interfaces. This is perhaps surprising, since
original steady-state models) for both steady-state the optimized model is more accurate for steady-state
a sequence of steady-state flows. One of the Oil-water vertical flow. A sample comparison for
objectives of our work is to test the validity of this model predictions with experimental data for vertical
We proceed by identifying two interfaces for two- volume of the two fluids in the system is about the
phase flows. The gas interface is the interface same. As in the previous case, the speed of the water
between the pure gas and the mixture of gas and interface with the original model is much higher than
liquid. Similarly, the liquid interface is defined as the that observed in the experiment. Furthermore, the
interface between the pure liquid and the mixture of optimized model yields even higher velocities for
gas and liquid. Therefore, during the transient both oil and water interfaces.
process, the gas interface moves down and the liquid An explanation for the disagreement between
interface moves up. The two interfaces meet when transient experiments and steady-state model
the phases are completely separated. predictions can be offered by considering the drift-
Liquid-gas vertical flow. Fig. 5 shows a sample C 0 = 1.0 for the optimized model, i.e., there is no
comparison of experimental data with predictions for profile slip. Hence m(θ ) , the Vd multiplier, must
vertical water-gas flow. Both the original and
increase accordingly, and for vertical flow, it is
optimized steady-state models are considered. In this
almost twice the value as in the original model.
case the volume of gas and water in the system is
Therefore the optimized model predicts much faster
25
almost the same. We see that the original model
settling. For oil-water flow, the major reason for the
predicts the speed of the gas interface height
prediction of faster separation by the optimized
10 H. SHI, J.A. HOLMES, L.J. DURLOFSKY, K. AZIZ
model compared to the original model is the velocity used in both steady-state models (original
reduction in the exponent n from 2.0 to 1.0. and optimized) does distinguish between large and
From these comparisons of experimental data small bubbles/droplets. Fig. 7 (a) shows that a linear
with model predictions, we see that our steady-state interpolation is used to connect the bubble flow
models do not fully capture the mechanics of regime and liquid flooding curve2 over the range
countercurrent transient flows. These findings are a1 < α g < a2 . Since bubble size increases with α g at
system. In fact, in reality there will exist a bubble separation ( C 0 S = 1.0 ) due to the large bubble
distribution of bubble sizes, with the smaller bubbles separation with the mixture, a general drift-flux
having even lower drift velocity4. The dashed line in model is obtained (see Appendix A for details):
transient flows using one drift-flux model. The two- Here C 0 L is the profile parameter for the separation
population model discussed below is a unified model of large bubbles from the mixture of small bubbles
for steady-state and transient flows. This unification and liquid. V dL and V dS define the drift velocity of
is especially important for reservoir simulation, in large bubbles and small bubbles respectively. This
which a smooth transition between steady-state and equation reduces to the original form when there is
transient flows is required. only one kind of bubble and there is no profile slip
steady-state and transient flows we can conclude that Two-population model for oil-water systems. The
in the separation of water and gas, two processes two-population oil-water model is similar to the
occur. First, large gas bubbles separate from the gas- liquid-gas model, but the mechanisms involved in
water mixture, and next the entrained small gas oil-water separation are different. Specifically, large
bubbles separate from the water. This can be modeled water droplets move down while the small water
by dividing the total gas fraction into two parts, droplets entrained in the oil move up with the oil
corresponding to large bubbles and small bubbles: phase. This is also consistent with the observation by
α g = α gL + α gS (9) Zhu and Hill16 and Zavareh et al.17. In addition, the
where subscript L represents the large bubbles and S entrained small water droplets further separate from
We can apply the drift-flux model, Eq. (1), for We divide the water droplets into two
and apply the oil-water drift-flux model, Eq. (8), to These are the fraction f of large bubbles/droplets to
both settling processes with the assumption that the the total bubbles/droplets in the system and the drift
profile slip of small droplets is 1.0 due to the velocity multiplier mS for small bubbles/droplets
disruption of large water droplets separating with the (where VdS = mSVdL( α g = 0 )). These parameters
mixture. The resulting two-population model for oil- depend, in general, on the shut-in holdup, though in
water separation is (see Appendix A): many cases constant values suffice. Using the two-
α wV w = V m − (1 − α wL − α wS )( C 0′ LV m + V dL′ + V dS′ ) (12)
population model with these two parameters, we can
where C 0′ L is the profile parameter for the separation achieve close matches to the transient experimental
of large water droplets from the mixture of oil and data. In the following figures, the model results are
water. V dL′ and V dS′ represent the drift velocity of the shown in terms of interface height. Predictions by the
droplets respectively.
Vertical water-gas flow. For all water-gas cases, a
We note that the two-population model described
single set of optimized parameter values
here represents a considerable simplification of the
true transient process, in which a continuous (independent of αg and αw) was determined:
distribution of bubble or drop sizes presumably f = α gL α g = 0.3 and m S = 0.3 . These values indicate
exists. Nonetheless, as shown below, this model does that most (70%) of the gas bubbles in the water-gas
appear to capture the key transient effects observed in systems are small bubbles.
the experiments. This is likely because the “two The water-gas results are illustrated in Figs. 7-9.
populations” of bubble/drop sizes (and corresponding Each figure corresponds to a particular value of
adjustable parameters) represent, in some sense, an α g (as indicated in the figure). The first example is
appropriate sampling of the true continuous for a relatively low α g ( α g = 0.18 ). We see from Fig.
distribution.
8 that the optimized steady-state model predicts very
The amount of water and gas in the system is ( m S′ = 0.03 ). We attribute this to our expectation
about the same for the last example displayed in Fig. that the phase inversion point is around 50% for this
10. The results from the steady-state models for this oil-water system (the fine oil and water droplets
case were presented in Fig. 4. Here the movement of separate very slowly around the phase inversion
the gas interface is predicted by the two-population
point). Table 1 gives m S′ values for seven oil-water
model to be too slow at the beginning of the
tests. It clearly demonstrate that m S′ reaches a
separation but overall the results for both the gas and
minimum at around α w = 50%. Accurate results are
water interfaces are in reasonable agreement with the
also obtained in the case of high oil fraction, as
experiments.
shown in Fig. 13.
systems are represented very well by the two- • A new unified two-population drift-flux model
population models. We again emphasize that the was developed for transient two-phase flows.
models in this case are consistent with the steady- The model reduces to the steady-state model in
state models, as m(θ ) is the same in both cases. appropriate limits. The model predictions match
Application to well testing both vertical and deviated water-gas and oil-
* Why phase redistribution can be important • Application to well testing (Jon’s contribution)
* Simulation results A concern with this model (or many wellbore flow
* What tweaks to d-f are necessary to match the models) is that the model parameters are based on
Conclusions and recommendations acting valves may not represent actual conditions in
From this study, we can draw the following the field. It is therefore possible that the model
• The drift-flux model is well suited for steady- applications. This can only be gauged by testing the
state concurrent flows as well as transient model against other experimental data sets, which are
countercurrent flows in wellbores and pipes. not currently available. Even though the model
• Experimental data from large-diameter pipes parameters may require tuning for a particular
suggest that wellbore transient flow cannot be application, it is still reasonable to expect that the
represented by a series of steady-state flows. two-population model presented here (or a very
• Experimental observations show that gas exists similar model) can be used to represent transient
as large and small bubbles during the settling countercurrent wellbore flows.
industrial affiliates of the Stanford Project on the m′ = drift velocity multiplier for oil-water flows
L = large bubbles/droplets
Co = profile parameter
m = mixture
D = pipe internal diameter
o = oil
f = fraction of large bubbles/droplets
S = small bubbles/droplets
Ku = Kutateladze number
Greek
L = test section length
α = in situ fraction or holdup
m = drift velocity multiplier for water-gas flows
σ = interfacial tension/surface tension
16 H. SHI, J.A. HOLMES, L.J. DURLOFSKY, K. AZIZ
Exhibition, Houston, TX, 2-5 October. presented at the 1998 International Petroleum
17. Zavareh, F., Hill, A.D. and Podio, A.: “Flow Regimes Conference and Exhibition of Mexico, Villahermosa,
SPE paper 18215 presented at the 1988 SPE Annual 24. Hasan, A.R. and Kabir, C.S.: “Modeling Changing
Technical Conference and Exhibition, Houston, TX, Storage During a Shut-in Test”, SPEFE (1994) 9, 279-
18. Ouyang, L.B.: “Mechanistic and Simplied Models for 25. Zuber, N. and Findlay, J.A.: “Average Volumetric
Countercurrent Flow in Deviated and Multilateral Concentration in Two-Phase Flow Systems”, J. Heat
Wells”, SPE paper 77501 presented at the 2002 SPE Transfer, Trans. ASME, (1965) 87, 453-468.
Annual Technical Conference and Exhibition, San 26. Schlumberger GeoQuest, ECLIPSE Technical
19. Ouyang, L.B.: “Mechanistic and Simplied Models for 27. Hasan, A.R. and Kabir, C.S.: “A Simplified Model for
countercurrent flow in deviated and multilateral Oil/Water Flow in Vertical and Deviated Wellbores”,
wells”, Petroleu Sci.& Tech, (2003) 21, 2001-2020. SPE Prod. & Fac. (February 1999) 56-62.
20. Almehaideb, R.A., Aziz, K. and Pedrosa, O.A.: “A 28. King, M.J.S., Hale, C.P., Lawrence, C.J., and Hewitt,
Reservoir/Wellbore Model for Multiphase Injection G.F.: “Characteristics of Flow Rate Transients in Slug
and Pressure Transient Analysis”, SPE paper 17941 Flow”, Int. J. Multiphase Flow, (1997) 24, 825-854.
presented at the 1989 SPE Middle East Oil Technical 29. Pal, R.: “Pipeline Flow of Unstable and Surfactant-
Conference and Exhibition, Manama, Bahrain, 11-14 Stabilized Emulsions”, AIChE J.(1993) 39, 1754-
March. 1764.
valve valve
differential gamma
pressure densitometer pressure
(a)
(b)
Fig. 2. Water-gas data for θ=0°, Qw= 40.4 m /h, Qg= 58.0 m /h
3 3
(αw=52%).
20 H. SHI, J.A. HOLMES, L.J. DURLOFSKY, K. AZIZ
m /h (αw =51%).
3
21 DRIFT-FLUX MODELING OF TRANSIENT MULTIPHASE FLOW IN WELLBORES 21
11
10 Experiment_gas
9
Interface Height (m)
8 Original_gas
7
Optimized_gas
6
5
Experiment_water
4
3 Original_water
2
1 Optimized_water
0
0 10 20 30 40 50 60 70 80
Time (s)
11
10 Experiment_oil
9
Interface height (m)
8 Original_oil
7
Optimized_oil
6
5 Experiment_water
4
3 Original_water
2
Optimized_water
1
0
0 100 200 300 400 500 600 700 800
Time (s)
0.7
large bubbles
0.6
0.5
small
0.4
bubbles
Vd
0.3
0.2
0.1
a a
0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
αg
0.7
0.6
0.5
0.4
Vd
small bubbles
0.3
0.2
smaller bubbles
0.1
a1 a2
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
αg
11
10
Interface Height (m)
9
Experiment_gas
8
7 Optimized_gas_ss
6
Optimized_gas_t
5
4 Experiment_water
3
Optimized_water_ss
2
1 Optimized_water_t
0
0 10 20 30 40 50 60 70 80 90 100
Time (s)
11
10 Experiment_gas
9
Interface Height (m)
8 Optimized_gas_ss
7
Optimized_gas_t
6
5
Experiment_water
4
3 Optimized_water_ss
2
1 Optimized_water_t
0
0 10 20 30 40 50 60 70 80 90 100
Time (s)
11
10 Experiment_gas
9
Interface Height (m)
8 Optimized_gas_ss
7
Optimized_gas_t
6
5
Experiment_water
4
3 Optimized_water_ss
2
1 Optimized_water_t
0
0 10 20 30 40 50 60 70 80
Time (s)
11
10 Experiment_oil
9
Interface Height (m)
8 Optimized_oil_ss
7
Optimized_oil_t
6
5 Experiment_water
4
3 Optimized_water_ss
2
Optimized_water_t
1
0
0 100 200 300 400 500 600 700
Time (s)
11
10 Experiment_oil
9
Interface Height (m)
8 Optimized_oil-ss
7
Optimized_oil-t
6
5
Experiment_water
4
3 Optimized_water_ss
2
1 Optimized_water_t
0
0 100 200 300 400 500 600 700 800
Time (s)
11
10 Experiment_oil
9
Interface Height (m)
8 Optimized_oil-ss
7
Optimized_oil_t
6
5
Experiment_water
4
3 Optimized_water_ss
2
1 Optimized_water_t
0
0 100 200 300 400 500 600 700 800
Time (s)
11
10 Experiment_gas
Interface Height (m)
9
8 Optimized_gas_ss
7
Optimized_gas_t
6
5 Experiment_water
4
3 Optimized_water_ss
2
1 Optimized_water_t
0
0 10 20 30 40 50 60 70 80
Time (s)
Appendix A α gS V gS = α gS C 0 S V mS + α gS V dS (A-5)
Derivation of two-population drift-flux Assuming that there is no profile slip for small
models
bubbles since the profiles are disrupted by large
Liquid-gas flow. Because of small gas bubbles that bubbles, C 0 S = 1.0 , Eqn (A-5) becomes:
bubbles travel with a velocity VgS, which can also be Oil-water flow. Our experiments show that water
computed by drift-flux model: entrained in the oil phase, and the water-in-oil
35 DRIFT-FLUX MODELING OF TRANSIENT MULTIPHASE FLOW IN WELLBORES 35
dispersions/emulsions separated much slower than where Vw is the average water velocity of both large
pure phases. Therefore, we can assume that in the water droplets and small water droplets .
overall system mixture of oil and small water Combining Eqn (A-12), (A-13), (A-14) and (A-
droplets rises while large water droplets sink. 15), and assuming that the profile slip for the oil and
Similarly to the treatment of the liquid-gas small water droplets system is disrupted by large
system, let there be two populations of water water droplets ( C 0′ S = 1.0 ) we obtain the following
droplets: large water droplets with volume fraction of two-population model for oil-water flow:
α w = α wL + α wS (A-12)
of the mixture.
be determined from:
Vo = C 0′ S Vom + V do (A-14)
relationship:
V m = α wV w + (1 − α w )Vo (A-15)