A New Look at Predicting Gas-Well Load-Up Steve B. Coleman, SPE, consultant; Hartley B. Clay, SPE, and David G. McCurdy, SPE, Exxon Co. U.S.A.; and H. Lee Norris III, SPE, Exxon Production Research Co.
SPE 20280 Summary. This paper discusses results of field tests conducted to verify minimum flow rate (critical rate) required to keep low-pressure gas wells unloaded and compares results to previous work. This paper also covers liquid yield effects, liquid sources, verification that wellhead conditions control onset of load-up, and effects of temperature, gas/liquid gravities, well bore diameter, and packer/tubing setting depth.
Introduction As natural gas is produced from depletiondrive reservoirs, the energy available to transport the produced fluids to the surface declines. This transport energy eventually becomes low enough that flow rates are reduced and fluids produced with the gas are no longer carried to the surface but are held up in the wellbore. These liquids accumulate in the wellbore over time, and cause additional hydrostatic backpressure on the reservoir, which results in continued reduction of the available transport energy. In most cases, if this condition is allowed to continue, the wellbore will accumulate sufficient fluids to balance the available reservoir energy completely and cause the well to die. This phenomenon is known as gas-well load-up. As Fig. 1 shows, load-up can easily be recognized on a typical gas-well L-lO chart by the characteristic exponential rate decline caused by accumulating wellbore liquids. Numerous papers have offered methods for predicting and controlling the onset of load-up.1-6 Turner et al.'s 1 method for predicting when gas-well load-up will occur is most widely used. They compared two physical models for transporting fluids up vertical conduits: liquid film movement along the pipe walls and liquid droplets entrained in the high-velocity gas core. A comparison of these two models with field test data yielded the conclusion that the onset of load-up could be predicted adequately with an equation developed from liquid droplet theory (Stokes law), but that a 20% upward adjustment of the equation was necessary. Turner et al. also suggested that in most instances wellhead conditions controlled the onset of liquid load-up and that liquid/gas ratios in the range of 1 to 130 bbllMMscf did not influence the minimum lift velocity. Examination of Turner et al.'s published data indicates that most of the wells used in the comparison had wellhead flowing pressures (WHFP's) above 500 psi. Because gas-well load-up problems generally worsen with continued decline in reservoir energy, this paper focuses on wells with lower reservoir pressures that are experiencing liquid load-up and have WHFP's below 500 psi. Copyright 1991 Society of Petroleum Engineers
JPT • March 1991
Wellbore Liquid Sources Before examining the wellbore-liquid-Ioading mechanism, we must first consider the source of the liquids. There are two obvious sources: liquids condensed from the gas owing to wellbore heat loss and free liquids produced into the wellbore with the gas. Both liquid hydrocarbons and water may be present, depending on the specific reservoir. In examining these sources, one might tend to minimize the impact of condensed water, particularly at low reservoir pressures. Because the gas is saturated with water at reservoir conditions, a plot like Fig. 2 can be constructed to show the impact of condensed water for a typical8,OOO-ft, lowpressure gas well. As shown, the amount of water condensed increases exponentially as the static reservoir pressure declines. This is unfortunate because, as reservoir pressures decline, the amount of load fluid required to balance the reservoir hydrostatically and to kill a well also declines, compounding the problem. Other problems may also occur as a result of gas-well load-up. The near-wellbore region of the reservoir may begin to become saturated with liquids, causing the relative permeability to gas to decrease, further reducing the well's potential to remain productive. Also, condensed water can be damaging to formations containing swelling clays because it is low in total chlorides «500 ppm). Critical· Rate TheoryLlquld.Droplet Model
As Turner et al. showed, a free-falling particle in a fluid medium will reach a terminal velocity that is a function of the particle size, shape, and density and of the fluid-medium density and viscosity. Applying this concept to liquid droplets in a flowing column of gas, we can calculate the terminal velocity, vI' of the drop using v t = 1.912{[q'4 (PL -Pg) 'A ]/P g 'h}, . (1)
which assumes a fixed droplet shape, size, and drag coefficient and includes the +20 % adjustment suggested by Turner et al. Applying this terminal velocity equation to wellbores and correcting to standard conditions, we can determine the minimum gas 329
"Wellbores with concentric tubing strings that terminate some distance above the completion Interval will generally not follow a qc calculation based on flowing wellhead conditions." Fig. 1-Gas-well L-10 chart. flow rate, qc' for the continuous removal of liquids from a wellbore:
qc =3.06pvt AITz. . .............. (2) This rate, also known as the critical rate, can be used to predict the onset of gas-well load-up. Fig. 3 plots this critical rate for typical 2Ys-in. tubing with water as the droplet fluid.
Comparison to Field Data To verify this theory and Turner et at. 's conclusions for wells experiencing liquid loadup with WHFP's below 500 psi, two sources of field data were collected and analyzed. First, 17 tests were conducted to determine the rate and WHFP at which load-up would begin. These tests examined wells that would consistently flow at a stable rate above their calculated qc' For the test, gradual, stepwise increases in WHFP were introduced until the wells exhibited the typical exponential rate decline denoting the onset of load-up. WHFP and gas-rate chart recorders were used to document the test and to provide a time history of the variables. Fig. 1 shows a typical test chart. The second source of data consisted of numerous gas-well 8-day L-lO production charts examined for similar rate-decline tendencies. In each case, an interpretation of the rate and WHFP was made for the point that indicated the onset ofload-up. Because the production charts for these wells were generally located some distance from the 330
wellhead, historical well test records were used to obtain WHFP's. Table 1 presents a summary of the data collected from both the critical-rate tests and production-chart data base. Fig. 4 plots these same data, comparing the observed and calculated critical rates. A first examination of the plot shows that the unadjusted liquiddroplet model tends to offer a better match of the field data for these low-pressure wells. Thus, the terminal-velocity equation (Eq. 1) can be rewritten without the 20% adjustment:
v t = 1.593{[u \4 (PL -Pg) \4]/ Pg y, }.
.
(3)
A closer examination of the plot and raw data yields several other valuable observations and explains some of the anomalies. First, the data examined support the assumption that in most cases wellhead conditions control the onset of load-up. This is consistent with Turner et at. 's observations. Second, in several of the critical-rate tests, a relatively large step was inadvertently imposed when the WHFP was raised to force the well into load-up. In doing so, the exact point of load-up was overshot. This helps explain why some of these data points (solid circles) plot below the theoretical curve. Third, it was observed that a number of the data points (open points), which lie significantly below the theoretical line, exhibited liquid-slugging behavior denoted by a ragged trace on the L-IO chart. These wells should be not expected to fit the liquid-droplet model theory and are considered anomalies.
Next, as Table 1 shows, the liquid/gas ratio of the wells examined ranged from 1 to 22.5 bbl/MMscf. This had no influence in determining the onset of load-up. In other words, regardless of the amount of liquid present, the onset of load-up will occur at the same gas flow rate, all other factors being equal. This is consistent with Turner et at. 's findings. Next, a comparison of the actual water production rates with calculations for condensed-water volumes indicates that the primary source of load fluid was condensed water in most cases. Finally, for several of the critical-rate tests, the wells were produced to a threephase separator to measure gas and liquid production. In these tests, liquid production stopped completely once the well began to exhibit load-up behavior. This verified that droplet terminal velocity had been reached and produced liquids were being held up in the wellbore. Additional information regarding wellbore liquid-holdup behavior is provided in Part 2 of this series. 8
Influence of Important Variables This examination of critical-rate theory would not be complete without a discussion of the influence of different variables on the accuracy of the calculation. Up to this point, the discussion of qc has been directed toward using water as the load fluid. This is a valid assumption in most cases. However, some situations do exist for rich-gas reservoirs where the wellbore conMarch 1991 • JPT
ditions cause condensation of the heavy hydrocarbons, but not water. This was observed in several wells examined as a part of this work. Turner et at. noted that in these situations the density and surface tension of the liquids should be adjusted accordingly in calculations of qc' In most cases, ifboth water and liquid hydrocarbon are present, the denser water controls the onset ofloadup. Examination of other variables, such as temperature, gas gravity, and interfacial tension (1FT), indicates that in most cases they have only a minor influence on the accuracy of the critical-rate calculation. The wellbore cross-sectional area obviously is one of the most important variables in the critical-rate calculation. Fig. 5 shows a comparison of various wellbore diameters. Smaller-diameter wellbores have better droplet lift efficiencies because of increased transport-gas velocities. This direct relationship leads to a very important observation. Wellbores with concentric tubing strings that terminate some distance above the completion interval will generally not follow a qc calculation based on flowing wellhead conditions. Because of the significant impact of diameter, the qc for these situations should be calculated with the flowing conditions of the largest-diameter segment of the wellbore, rather than the wellhead conditions. Setting the tubing/packer of the concentric tubing string a significant distance above the completion interval can lead to premature liquid-loading problems. This may seem somewhat inconsistent with the theory that condensing liquids are a primary source of load fluids, because these fluids are generally present only in the upper portion of the wellbore. We have found that system upsets, well shut-ins, human intervention, etc., will cause liquids to be deposited periodically in the lower portion of the wellbore. If there is a significant difference in diameter between the upper and lower portions of the wellbore, the largest of the two will generally control the onset of load-up.
,. 25
... ... CJ
VI
Z
2.
CJ
Z
a
u
...'" ~ ...a ...u ,.,.
..
~
,.
VI
;;;. -'
JPT • March 1991
..
.+---------~--------~--------~--------~--------_r--------~ , I •• 2 •• I ••
•
•••
•••
STATIC RESERVOIR PRESSURE, PSIA
Fig. 2-Condensed·water volume: typical 8,000·ft well. 7
....
,T-------~--------
__
--~--
__
~------------
__------__
--~
__,
.. -.-- .. -..... -- .... ...........•. ~
:--NON LOAD-UP REGION
,..,.
+-------~----~~
__
..
,
~~~~~--------~--_+--~~--~~~
1000
FLOWING WELLHEAD PRESSURE, PSIA
Fig. 3-Wellbore critical rate: 27fa·in. tubing.
.
----r------__--------__...__--__
Conclusions 1. The minimum flow rate, known as the critical rate, qc' required to keep lowpressure gas wells unloaded can be predicted adequately with the Turner et at. liquiddroplet model without the 20% upward adjustment. 2. A primary source ofload fluid for lowpressure gas wells can be condensed water. 3. Liquid/gas ratios below 22.5 bbl/ MMscf have no influence in determining the onset of load-up. 4. Wells that exhibit slugging behavior may not follow the liquid-droplet model because of a different transport mechanism. 5. Such variables as temperature, gas and liquid gravity, and 1FT have little effect on the critical rate, whereas wellbore diameter and pressure have a direct and significant impact. 6. In most cases, wellhead conditions can be used to predict the onset of load-up. How-
Source: Mck_tta-W.h. Baal I; 0.' Sp. Gr. Ga. 200·r R• .." rlmp. SO Pwlg W.llhead P,.. .. 90-r W.llhead Tlmp.
1 2 0 0 Q - r - - - - - - - - - - - - - - -__- -____
....•.. ..........
~'7I
10000+-----;----'-----+---;--+--,-t-----'---+---c--....__-+-~"""''_I •
Critical Rat.T •• fa
o
Critical Rat. T.... - Slugglnll
... Production Chart Data
...
. "•. ··ii:~
::·.·.·.~·.~~·.,)il.
~
{j. PrOduction Chart Data - 5111"ln,
~
•
TII,no, Data - Unload.d
o
Turn.r Data - loaded
() U)
:2 cr 0 UJ
1000
> a:
UJ
U)
aJ
a
CALCULATED q,. MSCFD
Fig. 4-Critical·rate data base. 331
TABLE 1-CRITICAL-RATE DATA BASE
Gas Specific Gravity Depth Tubing ID Condensate (in.) (bbVMMscf) Test (air ... 1) ~ - 1 0.582 7,812 2.441 0.1 2.441 2.9 2 0.595 8,021 2.441 5.7 3 0.628 8,437 5.7 4 2.441 0.628 8,437 2.441 3.5 5 0.620 8,042 2.441 1.5 6 0.602 5,538 2.441 1.5 7 0.602 5,538 2.441 2.0 8 0.654 6,446 2.9 2.441 9 0.668 6,026 2.441 2.9 10 0.668 6,026 2.441 1.3 11 0.602 6,499 2.441 0.0 12 0.628 6,764 5.9 0.672 5,678 2.441 13 2.441 0.0 14 0.610 8,507 2.441 0.0 0.610 9,445 15 5.0 16 2.441 0.628 6,984 2.441 2.9 17 0.620 6,034 2.441 0.0 0.643 5,338 18 2.441 0.6 19 0.654 7,632 0.0 2.441 20 0.652 5,342 0.0 21 0.646 5.147 2.441 2.441 0.0 0.595 7,763 22 0.0 23 0.595 7,763 2.441 0.0 2.441 24 0.595 6,261 2.441 0.0 0.622 6,900 25 3.4 2.441 0.661 7,428 26 2.2 2.441 0.682 4,680 27 1.7 2.441 0.650 5,011 28 2.441 2.9 29 0.634 5,745 2.441 2.9 30 0.658 ·6,491 4.0 2.441 31 0.628 6,443 4.0 0.628 6,443 32 2.441 2.441 8.7 33 0.683 6,582 2.441 1.8 34 0.599 6,898 1.8 2.441 35 0.599 6,898 2.441 14.8 36 0.663 6.351 2.441 5.9 37 0.709 6,722 2.5 2.441 0.600 7,600 38 2.441 4.3 39 0.660 6,120 0.0 2.441 40 0.617 6,880 2.441 41 0.674 6,556 0.0 1.8 0.662 6,301 2.441 42 2.441 2.4 43 0.642 4,751 44 2.441 3.5 0.651 5,065 0.0 45 2.441 0.610 6,285 2.441 0.0 0.600 6,335 46 7.5 2.441 47 0.651 8,439 4.6 0.620 8,158 2.441 48 2.441 0.0 0.750 8,508 49 2.4 2.441 0.625 8,466 50 2.4 51 0.625 8,466 2.441 2.441 4.5 0.621 8,504 52 4.5 2.441 53 0.621 8,504 5.3 0.717 8,440 2.441 54 2.441 1.1 55 0.688 6,796 2.441 6.5 0.674 6,381 56
Observed Calculated BHSP* WHFP Water qc qo (psig) Comment (bbVMMscf) (psi a} (MscflD) (Mscf/D) 874 1.9 275 726 548 qc test well 744 350 qc test well 0.0 205 660 737 650 qc test well 3.9 212 585 3.9 150 468 618 650 qc test well 691 355 qc test well 7.1 185 573 593 619 315 qc test well 5.6 145 617 5.6 145 619 315 qc test well 412 175 qc test well 250 9.2 70 140 607 580 223 qc test well 7.0 223 qc test well 7.0 138 575 600 306 qo test well 0.0 130 635 586 200 qc test well 583 563 7.0 125 329 qc test well 2.3 649 628 165 1,031 647 400 q c test well-ragged trace 395 0.0 612 821 450 q c test well-ragged trace 0.0 255 613 qc test well 17.6 355 952 962 182 q c test well-ragged trace 520 3.6 105 430 L-l0 chart-ragged trace 396 494 400 99 0.0 410 650 L-l0 chart-ragged trace 2.7 70 164 350 L·l0 chart 43 329 323 0.0 284 L-l0 chart 0.0 52 267 356 663 L-l0 chart 352 640 983 0.0 225 615 780 663 L·l0 chart 0.0 1,072 1,174 466 L-l0 chart 495 0.0 488 N/A L-l0 chart 94 748 0.0 276 395 568 L·l0 chart 12.3 65 N/A L-l0 chart 500 371 3.4 59 1.9 366 348 168 L·l0 chart-ragged trace 50 311 284 L-l0 chart 5.4 39 324 1,000 L-l0 chart-ragged trace 484 9.8 97 90 421 L-l0 chart-ragged trace 389 17.5 60 220 478 421 L·l0 chart-ragged trace 17.5 90 355 341 122 L-l0 chart 50 338 9.7 N/A L·l0 chart 11.0 401 398 60 N/A L·l0 chart 450 460 11.0 80 508 257 L·l0 chart 5.1 107 471 372 0.0 306 L·l0 chart-ragged trace 553 135 781 131 518 590 L·l0 chart 1.6 51;j2 L·l0 chart-ragged trace 462 130 330 5.7 181 511 460 L·l0 chart 0.0 82 461 273 L·l0 chart 558 0.0 90 362 L·l0 chart 100 493 491 1.5 627 676 673 L·l0 chart 6.5 183 120 518 542 202 L·l0 chart 1.8 212 L-l0 chart 47 358 349 1.0 440 L·l0 chart 0.0 315 885 924 712 638 447 L·l0 chart 0.0 165 L-l0 chart 450 75 408 438 4.9 1,100 L·l0 chart 924 2.3 380 666 630 725 L·l0 chart-ragged trace 5.9 155 648 725 L·l0 chart-ragged trace 145 564 608 5.9 781 782 728 L·l0 chart 0.0 235 755 764 728 L·l0 chart 0.0 225 610 725 L-l0 chart 165 620 1.0 244 L·l0 chart 3.8 49 430 335 397 372 154 L·l0 chart 8.6 59
• Bottomhole static pressure.
ever, for concentric tubing strings where the tubing/packer is a significant distance from the completion interval, flowing conditions of the largest-diameter segment should be used to predict the wellbore critical rate.
Nomenclature A = flow area of conduit, ft2 p = pressure, psia qc = critical gas flow rate, MMscflD T = temperature, OR 332
v(
= terminal
z= = = a =
Pg PL
velocity of particle, ft/sec gas compressibility factor gas phase density, Ibm/ft3 liquid-phase density, Ibm/ft3 interfacial tension, dynes/cm
Acknowledgments We express our appreciation to the management of Exxon Co. U.S.A. and Exxon Production Research Co. for the support and encouragement given while we conducted
this study. Special thanks go to the Operations personnel and the Production/Reservoir Technology support staff of Exxon's South Texas Div., who provided invaluable cooperation and assistance in gathering and manipulating data and in preparing the necessary technical review packages.
References r. Turner, R. G. et al.: "Analysis and Prediction of Minimum Flow rate for the Continuous Removal of Liquids From Gas Wells," JPT (Nov. 1969) 1475-82; Trans., AIME, 246. March 1991 • JPT
.. .. . ,····fSS2~~~';""';mss:': ~~ : ".-.
... ~.
...... ....
. . .;. . ·;··:nT. -·~--·:··t-1-(
~.
t········~···· ~
~
~
.
~
~
~ ••
.;•. !.~ .
.. .:... .. ~-.:. ~ . ~ .. ~ .. -~.
.. :- .. -:-- ..
:
.... ,. ;.L-,...,....,.,---,--..,.---,-...,...J,.
1·±,.--....;..-;.....;......;....+....;...~1.~.1'-1I-"-1A....;..-..;........;.....;....;....;...;...;.1.l-••--...;....-i-.;.....i-+~..;,........ FLOWING WELLHEAD PRESSURE. PSIA
Fig. 5-Wellbore critical rate
VS.
WHFP. . McCurdy
2. Hutlas, E.J. and Granberry, W.R.: "A Practical Approach to Removing Gas Well Liquids," JPT (Aug. 1972) 916-22. 3. Iiobi, M.l. and Ikoku, C.U.: "Minimum Gas Flow Rate for Continuous Liquid Removal in Gas Wells," paper SPE 10170 presented at the 1981 SPE Annual Technical Conference and Exhibition, San Antonio, Oct. 4-7. 4. Libson, T.N. and Henry, J.R.: "Case Histories: Identification of and Remedial Action for Liquid Loading in Gas Wells-Intermediate Shelf Gas Play," JPT (April 1980) 685-93. 5. Weeks, S.G.: "Small-Diameter Concentric Tubing Extends Economic Life of High Water/Sour Gas Edwards Producers," JPT (Sept. 1982) 1947-50. 6. Orris, P.W. and DeMoss, E.E.: "Liquid Removal From Gas Wells-Gas Lifting With Reservoir Gas, " paper presented at the Southwestern Petroleum Short Course, Texas Tech U., Lubbock, Apri118-19, 1%8. 7. Engineering Data Book, 10th edition, Gas Processors Assn., Tulsa (1987). 8. Coleman, S.B. et al.: "Understanding GasWell Load-Up Behavior," JPT(March 1991) 334-38.
JPT • March 1991
51 Metric Conversion Factors bbl x 1.589 873 ft x 3.048* ft3 x 2.831 685 OF (OF-32)/1.8 in. x 2.54* psi x 6.894 757
E-Ol = m3 E-Ol = m
E-02 = m 3 = °C E+OO = em E+OO = kPa
Norris
Steve .. ca ...... lsacOnsultlngengio neer In Corpus Christi, TX.He previous. "worked at Exxon's SoutIt T'-Dlv. In Corpus Chl'lstl. hoIds~a as
degree in mechnlcal
from
Tax_Alelll. engineer in the Raa_oII',.r•.-omrtOlCIGY
'Conversion factor is exact.
Provenance Original SPE manuscript, Optimizing Recovery of Natural Gas From DepletionDrive Reservoirs: Part I-A Study to Verify the Minimum Flow Rate Required for Continuous Removal of Liquids From Low-Pressure Gas-WeDs, received for review Nov. 16, 1989. Paper (SPE 20280) accepted for pUblication Feb. 9, 1990. Revised manuscript received Dec. 21, 1990.
JPT
", H.. .... 1IG11'1'1tl III ta 8118f1l1Ol"4 f'8 Be arcb SpecIaIIat In the Fa.ciIItIes De-
SIgn SectIon at Exxon PJoduction Research Co. In Houston. Be holds a degree from the U. of· In AuStm
as
and MSand
Stanford ~~. alt~.~_~.·_fl:teerl...
333
