Академический Документы
Профессиональный Документы
Культура Документы
Liquid Loading
Desheng Zhou, SPE, Xian Petroleum University, and Hong Yuan, SPE, IHS
upward forces are smaller than the downward force. The balance As shown in Fig. 2b, there are two droplets (A and B) in a gas stream
of the forces (FD FB FG) yields the droplet model (Eq. 1), at (assuming all the gas streams in Fig. 2 and Fig. 3 satisfy its critical
which the droplet will keep its velocity, and gives the critical gas gas velocity). For a laminar gas stream, the two droplets may move
velocity to sustain the droplet. According to the model, a liquid in the same direction in the well and will flow out of the well. For
droplet will be carried out by a gas stream if the stream flows faster gas wells, gas velocity is usually very high and the flow is turbu-
than the critical gas velocity. lent (Nosseir et al. 2000). In a turbulent gas stream, liquid droplets
Turner’s model is based on the force balance on a single liquid move not only upward with the gas stream, but also in all directions
droplet, but what if there is more than one droplet in the gas stream? irregularly. The nearby liquid droplets may encounter each other
FG
4000 Liquid
film
2000
0
0 2000 4000 6000 8000
Turner’s Model Rate, Mscf/D B
Turbulent A
Fig. 1—Incorrectly predicted wells by Turner’s model. gas
(a) (b)
and coalesce into a bigger droplet. As illustrated in Fig. 2b, liquid
droplet A and droplet B move in different ways and may encounter Fig. 2—Encountering two liquid droplets in turbulent gas
and coalesce into a bigger droplet AB at the top of the well. stream.
The newly formed droplet (droplet AB in Fig. 3a) may fall
down in the gas stream because its bigger size needs higher gas where Hl is liquid holdup and vsl and vsg are superficial liquid and
velocity to suspend it. gas velocities, respectively.
During falling, the newly formed bigger droplet may shatter Liquid-droplet concentration is the control factor in droplet
into small droplets (droplet AB shatters into droplets 1, 2, and 3 encounters. The higher the concentration of liquid droplets in a
in Fig. 3a) by velocity pressure (Turner et al. 1969), and the small turbulent gas stream, the greater the chance that the droplets will
droplets may be picked up again by the drag forces from the gas combine and fall.
stream. If there are other droplets in the upstream (droplets C, D, The concentration of liquid droplets in a gas stream may be the
E, and F in Fig. 3a), the bigger droplet or its shattered droplets third mechanism contributing to liquid loading, in addition to the
may encounter other droplets during their falling and the encoun- liquid-film mechanism and Turner’s liquid-droplet mechanism.
tered droplets may coalesce and keep falling. As shown in Fig. Turner’s entrained-liquid-droplet model is based on the force
3a, droplet AB may encounter droplet C and coalesce into a new balance on a single droplet and does not include the encounter
droplet ABC before droplet AB shatters, or its shattered droplet 3 effect. For low liquid-droplet concentration, the chance of encoun-
may encounter upstream droplet D. Because only a few droplets ters is low and Turner’s model works well. However, when the
are in the well, the droplets will be carried out finally. liquid concentration reaches a certain value, the encounter coalesc-
As shown in Fig. 3b, if there are more liquid droplets in the ing falling process of liquid droplets in a gas stream will dominate
gas stream, the chance of the process of liquid-droplet encounter- the entrained-liquid-droplet movement, and hence Turner’s single
ing, coalescing, falling, and shattering increases. As the number liquid-droplet model losses its function, even with gas-stream
of liquid droplets in a gas stream, called liquid-droplet concentra- flows faster than critical velocity.
tion here, increases to a threshold value þ, the process of droplets In addition to the liquid-droplet concentration, flowing conduit
encountering, coalescing, falling, and shattering will continue and length is also a major factor. The longer the mixture of gas and
bring those liquid droplets down to the well bottom. liquid droplets travels, the higher the chance of liquid-droplet
Liquid holdup can be used to represent the liquid droplet con- encounters. In petroleum engineering, well depth is considered
centration in a gas well. Liquid holdup is defined as long enough for entrained-liquid-droplet encounters.
l g 1/ 4
o(p p )1.593 1/ 2
vcrit N vcrit T pg for Hl þ
(a) (b)
, . . . . . . . (5)
H
Fig. 3—Liquid loading when liquid-droplet number reaches a vcrit N vcrit T ln l a for Hl þ
threshold value. þ
2
A (ft ) vsqwh vcrit-T vslwh vcrit-N qcrit-N
Well qtest qcrit-T yc+yw Bqwh [Eq.A-3, (ft/sec) (ft/sec) (ft/sec) HI (ft/sec) (Mscf/D)
No. p (psi) (Mscf/D) (Mscf/D) (bbl/MMscf) [Eq.A-2] Eq.A-4] [Eq.A-5] [Eq.A-6] [Eq.A-9] [Eq.A-10] [Eq.A-11] [Eq.A-12]
1 725 775 779 6 0.0200 0.0325 5.5122 5.5407 0.0093 0.0017 5.5407 779
2 400 417 583 18 0.0356 0.0217 7.9201 11.0730 0.0225 0.0028 11.0730 583
3 108 568 306 22 0.1204 0.0325 24.3548 13.1207 0.0250 0.0010 13.1207 306
4 540 712 661 21 0.0266 0.0217 10.1100 9.3859 0.0448 0.0044 9.3859 661
5 450 442 419 11.3 0.0318 0.0217 7.4917 7.1019 0.0150 0.0020 7.1019 419
86 500 400 2184 14 0.0287 0.1042 1.2755 6.9643 0.0035 0.0027 6.9643 2184
87 500 800 1726 5 0.0287 0.0588 4.5208 9.7535 0.0044 0.0010 9.7535 1726
88 660 4300 6367 3.5 0.0219 0.1841 5.9201 8.7659 0.0053 0.0009 8.7659 6367
89 280 500 2083 28 0.0501 0.1042 2.7847 11.6009 0.0087 0.0031 11.6009 2083
90 210 470 3248 24 0.0657 0.1841 1.9430 13.4272 0.0040 0.0020 13.4272 3248
where vcrit-N is the critical velocity from the new model in ft/sec, Hl As shown in the far-right column in Table 1, there are 12 incor-
is liquid holdup that reflects the liquid-droplet concentration, þ is rectly predicted wells from the new model. The results are better
the threshold value of liquid-droplet concentration for petroleum than Turner’s model (24 unpredicted wells) and Turner’s adjusted
production wells, and a is a fitting constant. a 0.6 and þ 0.01 model (13 unpredicted wells). Among the 12 unpredicted wells,
were estimated from the Turner et al. (1969) data. The maximum three wells were tested as unloaded. They are Well 14, Well 49,
liquid holdup is 0.24. When liquid holdup becomes higher than and Well 75. Fig. 4 shows the 12 incorrectly predicted wells from
0.24, the two-phase flow changes to slug- or churn-flow pattern the new model.
(Barnea 1987). Fig. 5 is a combination of Fig. 1 and Fig. 4. As shown in Fig. 5,
The new model is composed of two parts. When liquid holdup from the new model, five wells are almost on the diagonal line of
is less than or equal to the threshold value, þ, the critical-velocity the graph, which means that although the five wells are incorrectly
model is the same as Turner’s model. When liquid holdup is greater predicted mathematically, the differences are very small between
than the threshold value, þ, the critical velocity varies with the the calculated critical rates from the new model and the test results.
liquid holdup and can be calculated from the new model. The five wells are almost predicted by the new model. For Turner’s
The critical-rate correlation for the new model is the same as adjusted model, only two wells are almost predicted. If we take
that by Turner et al. (1969), as shown in Eq. 2. away the almost-predicted wells from the incorrectly predicted
wells, the new model has seven incorrectly predicted wells, but
3060 pv A Turner’s adjusted model has 11 incorrectly predicted wells.
qcrit N crit N (6)
Tz Application of Coleman et al. (1991) Data. Coleman et al.
evaluated Turner’s model and Turner’s adjusted model with their
Application of Turner et al. Data. Table 2 shows the application 56-well data. They found that Turner’s model matched their data
of the new model to the well data from Turner et al. (1969). Only pretty well, but Turner’s adjusted model did not work. They con-
the first and last five wells are listed here. Appendix A gives the cluded that, instead of using Turner’s adjusted model as suggested
equations for the calculations in the table. by Turner et al., Turner’s model should be used directly for low
The first four columns in Table 2 are the data from Turner et al. wellhead pressure wells.
(1969). Column 5 lists liquid yields that are the sum of condensate Table 3 shows the application of the new model to the Coleman
yields and water yields from the Turner et al. data. All the other et al. data. All the wells have a tubing inside diameter (ID) of 2.441
columns were calculated from the first five data columns, and their in. in the Coleman’s data. From Eq. A-3, the cross-sectional areas
equations are given in Appendix A. of all the wells are A 0.032498 ft2. Again, only the first and last
The far-right column is the calculated critical rates from the five wells are listed in Table 3.
new model. For comparison, it was copied into Table 1 as the The first five columns in Table 3 are data from Coleman et al.
ninth column. All the other columns are calculated values from the equations in
4000 4000
2000 2000
0 0
0 2000 4000 6000 8000 0 2000 4000 6000 8000
New Model Rate, Mscf/D Model Rate, Mscf/D
Fig. 4—Incorrectly predicted wells by the new model. Fig. 5—Comparison between new model and Turner’s models.
Appendix A. The last column presents the calculated liquid holdup with pressure and temperature. However, for a constant produc-
(Hl ) at the wellhead. ing gas rate (qg), producing gas velocity (vg) depends on flowing
Fig. 6 gives the calculated liquid holdup for each well. As conduit area, vg qgTz /(3,060pA). The larger the flowing conduit
shown in Fig. 6, all the liquid-holdup values are less than the area, the smaller the gas flowing velocity is.
threshold þ value (0.01) in Eq. 5. From Eq. 5, the new model is The areal effect is useful in gas-well design. For a desired
the same as Turner’s model when Hl 0.01. Therefore, the new producing gas rate, one may increase producing gas velocity to
model matches the Coleman et al. data and gives the same conclu- satisfy unloading-condition (Eq. 7) by choosing smaller tubing.
sion using Turner’s model, not Turner’s adjusted model, as that of For tubing production, the producing gas velocity varies inversely
Coleman et al. Therefore, the new model is an improvement over with the square of tubing diameter. For a well with varying flowing
Turner’s adjusted model for Coleman’s 56 wells. area, the unloading condition should be evaluated at the place of
the largest flowing area.
Discussion For a constant flowing area, such as tubing sets at the perfora-
As mentioned previously, for the liquid-droplet model (Eq. 1 or the tion, from wellhead to bottomhole both vg and vcrit decrease with the
first part of Eq. 5), Turner et al. (1969) suggested using the model at increase of pressure and temperature. The question is, which velocity
the wellhead. Coleman et al. (1991) supported that point of view, but reduces faster to determine the evaluating place (wellhead or bot-
suggested applying the model at the largest-diameter segment if there tomhole)? The comparison associates calculations of surface tension,
is more than one flowing area, such as when the tubing bottom is set densities, pressures, and temperatures along the tubing. Numeric cal-
significantly above perforation. Lea et al. (2003) pointed out that the culations for a gas well in Table 4 are given here for illustration.
wellbore bottom is the controlling place for liquid loading. What fol- In Well A (Table 4), gas flows up inside the casing (ID of 4.78
lows is a discussion of the application range of the new model. in.), first from the perforation depth of 8,467 ft to the tubing bottom
of 8,410 ft, and then flows up the tubing (ID of 2.441 in.) from
Turner’s Model (the First Part of the New Model). The first part the tubing bottom to the wellhead. Two flowing areas are used to
of the new model (Eq. 5) is the same as Turner’s liquid-droplet study the effect of conduit sizes.
model (Eq. 1). The discussion is a summary of where Turner’s Table 5 shows the calculated results of a popular commercial
droplet model should be evaluated—wellhead or bottomhole? software package. The critical velocity (vcrit-T) and critical flow rate
Liquid-unloading condition is where producing gas velocity, vg, (qcrit-T) are based on water properties because water is heavier than
is higher than critical velocity, vcrit, as the condensates in the well. Fig. 7 plots the calculated producing
gas velocities and critical velocities.
vg vcrit .............................................................................................................................................. (7) At the wellhead, producing gas velocity vg 8.304 ft/sec and
critical velocity vcrit-T 7.876 ft/sec. Because vg vcrit-T, there is no
Liquid-loading condition is vg vcrit. liquid-loading problem if evaluated at the wellhead. However, as
For Turner’s droplet model, as shown in Eq. 1 or the first part shown in Table 5 or Fig. 7, both velocities reduce to 7.152 ft/sec
of Eq. 5, the liquid-droplet model is independent of flow area and
is a function of interfacial tension liquid and gas densities that vary
TABLE 4—WELL A CONDITIONS
0.012 Production rate 1659.5 Mscf/D
Perforation depth 8467 ft
0.01
Tubing depth 8410 ft
Threshold value β=0.01
Liquid Holdup
at the depth of 5,651 ft. From the depth of 5,651 ft to the tubing tions and temperature gradients may change the relative trend of the
bottom at 8,410 ft, vg vcrit-T and liquid loading appears. Therefore, two curves. If evaluating unloading at the wellhead is acceptable
a well may be loaded up downhole even though the producing-gas for simplification, a larger downhole flowing section should not
velocity satisfies the unloading condition at the wellhead and in be ignored during unloading evaluation.
the same tubing set down to perforation.
As shown in Fig. 7, both producing gas velocity and critical
velocity decrease with depth. And producing gas velocity decreases Gas Velocity, ft/sec
faster than the critical velocity does. Therefore, if vg vcrit-T at tub- 0 2 4 6 8 10
ing bottom, all the producing gas velocities above should be greater 0
than corresponding critical velocities. Bottomhole is the controling
location for unloading evaluation. 1000
The example also shows the effect of flowing area on produc- 2000 Producing velocity
ing gas velocity and critical velocity. As shown in Table 5 and Turner’s critical velocity
Fig. 7, the critical velocity decreases to 6.838 ft/sec at the depth 3000
of 8,460 ft and to 6.837 ft/sec at the bottomhole of 8,467 ft from
Depth, ft
6.845 ft/sec at the tubing bottom at 8,410 ft. The flowing diameter 4000
changes from 2.441 in. at the tubing bottom to 4.78 in. in the sec-
5000
tion-tubing bottom to wellbore bottomhole. The flowing area does
not have effect on the critical velocity. 6000
However, the flowing area has a significant effect on the
producing gas velocity as shown by the lowest points in Fig. 7. 7000
The producing gas velocity decreases from 6.701 ft/sec at tubing
8000
bottom of 8,410 ft to 1.745 ft/sec and 1.744 ft/sec at the depths of
8,460 ft and 8,467 ft, respectively. 9000
The producing gas velocity curve and the critical-velocity curve
in Fig. 7 need complex numeric calculations, and different correla- Fig. 7—Effect of depth and flowing area on liquid loading.
0.009 at wellhead, which is smaller than the threshold value þ, but dtbgID tubing inside diameter, in .
the liquid holdup increases to 0.01 at 2,950 ft and 0.011 at 6,291 dtbgOD tubing outside diameter, in2.
ft. Therefore, as for the liquid-droplet model, the bottomhole is T temperature, ˚F
also the controlling location for the new model. Hl liquid holdup at wellhead
When taking into account the effect of pressure and tempera- p wellhead pressure, psia
ture on liquid rate, the new model has parameters of a 0 and þ qc condensate rate, B/D
0.01. For the example well, from 2,950 ft to 6,291 ft, the new qcrit-N gas critical rate from the new model in this paper,
model gives the same critical velocity as vcrit-N vcrit-T ln(0.01/ þ), Mscf/D
and ln(0.01/ þ) 0. From the depth of 6,291 ft to the bottomhole qcrit-T gas critical rate from Turner’s model, Mscf/D
8,467 ft, the critical velocity from the new model vcrit-N vcrit-T q gas critical rate from Turner’s model, Mscf/D
ln(0.011/ þ) vcrit-T 0.0953.
crit-T20%
For wells with tubing end at perforation, the new model may qg producing gas rate, Mscf/D
need to be evaluated at wellhead because the effects of both pro- ql liquid rate, B/D
ducing gas velocity and liquid holdup downhole are small. When qtest gas-test rate, Mscf/D
evaluating at the wellhead, an upward adjustment may be needed qw water rate, B/D
to account for the downhole effect. T temperature, °R
vcrit critical speed, ft/sec
Application Procedure. To apply the new model, one may follow
this procedure: vcrit-N critical speed from the new model in this paper, ft/s
1. Determine where to evaluate liquid loading: If fluid proper- vcrit-T critical speed from Turner’s model, ft/sec
ties and gas-flowing calculations are not available or are unreliable, vcrit-T20% critical speed from Turner’s adjusted model, ft/s
evaluate at wellhead. If there is a larger flowing area downhole, vg producing-gas velocity, ft/s
liquids may accumulate in that section. Otherwise, when fluids and vsg gas superficial velocity, ft/s
flowing calculations are available, evaluate at bottomhole. vsl liquid superficial velocity, ft/s
2. Use the first part of Eq. 5 (Turner’s model) to evaluate a vsgwh gas superficial velocity at wellhead, ft/s
well’s unloading condition (vg vcrit-T or qg qcrit-T). vslwh liquid superficial velocity at wellhead, ft/s
3. If the unloading condition is not satisfied, the well will be yc condensate yield, bbl/MMscf
loaded up. Stop here. Note, choosing smaller tubing will increase yw water yield, bbl/MMscf
the producing gas velocity. Otherwise,
z gas z factor
4. Calculate liquid holdup and check if concentration model
a parameter in new model, 0.6 or 0
should be used (Hl þ). If Hl þ, stop here. Otherwise, þ the threshold value of liquid droplet concentration, 0.01
5. Use the concentration model (the second part of Eq. 5) to
calculate a new critical velocity vcrit-N. The parameters are a 0.6 for petroleum gas wells
and þ 0.01 when using the stock-tank-liquid rate as liquid rate in pg gas density, lbm/ft3
holdup calculation, or use a 0 and þ 0.01 when using in-situ- pl liquid density, lbm/ft3
liquid rate (i.e., account for the effect of pressure and temperature o interfacial tension, dynes/cm
on liquid rate).
6. Evaluate unloading condition by the new critical velocity (vg Acknowledgments
vcrit-N or qg qcrit-N). If it is satisfied, liquids will be brought out The authors would like to thank IHS for permission to publish
of the well, otherwise, the well will be loaded up. this paper.
Engineers of AIME.
where
Taitel, Y., Barnea, D., and Duckler, A.E. 1980. Modelling flow pattern
p
Rμ 1.4 100.0125 API 0.00091T
transitions 1.204
Journal 26 for
(3):steady upward
345–354. gas-liquid flow in vertical tubes. AIChE
doi:10.1002/aic.690260304.
g .
18.2
s
Turner, R.G., Hubbard, M.G., and Dukler, A.E. 1969. Analysis and Pre-
diction of Minimum Flowrate for the Continuous Removal of Liquids
from Gas Wells. J Pet Technol 21 (11): 1475–1482; Trans., AIME, 246. Water formation volume factor can be calculated as (McCain
SPE-2198-PA. doi: 10.2118/2198-PA. 1990)
Wallis, G.B. 1969. One Dimensional Two-Phase Flow. New York: McGraw-
Hill. Bw (1 AVwt )(1 AVwp ),
and Tsc520°R. For Turner et al. wells, wellhead temperature was Superficial liquid velocity at the wellhead, v in ft/sec, can be
lwh
taken as 120°F. Using 0.9 for the z-factor, the gas formation volume
calculated from the liquid-flow rate, q l in B/D, at wellhead,
factor at wellhead, Bgwh , is
5.615ql 5.615qtest (yc yw ) . . . . . . . . .(A-9)
Bgwh
(0.0283)(0.9)(580) , ................................................(A-2) vslwh
p 24 60 60 A 24 60 60 1000 A
Rearranging Eq. A-6, we can obtain the critical rate, qcrit-N in Mscf/
where dtbgOD is tubing outside diameter in in. and dcsgID is casing D, if the critical velocity, vcrit-N in ft/sec, at wellhead is known,
inside diameter in in.. Both the casing ID and tubing OD were
given in the Turner et al. data. qcrit N (24 60 60)vcrit N A / (1000Bgwh ) ....................... (A-12)