SPE/lADC 29342 Gas Migration: Fast, Slow or Stopped

Ashley Johnson and Ian Rezmer-Cooper, Schlumberger Cambridge Researeh; Tim Bailey, Sedco Forex; and Dominic McCann, Anadrill
SPE/lAOC
Co@aht 1S6 Membsrs

SPEMIDC

Drilling Conference. held in Amatwdam, 28 FelNwy-2 March 1SS5.

TIIis paper was pm$nred for presentation at fhe 1SS5 SPEflAOC Orllllng Caferense

man 8bafracI aubmlffad by lb aufhotls). Curfenm of the paper, TM papar was 9abco3d preearmbn byan SPE/lAOC Prugmn COmmHtOetollowlng revbw G+Infomrafbm cdatned by Ihe lnf6mafic+rd AaacciMbn of Drflllng Curtractaa w the Society of Pefmleum Englnmm and are wb+oct to oxradkm by the MirOr@). The ~~ti,hmtitin~ PSWM m~ ~ s~~ MOOWS IVO s@Paf fO w~~bn mtwid, - PWWntd, doM not mxesswiiy tafbct anY podfkn of the SPEorWC, [Link] w mm-. MtmbmmaymfbcoW.~~swW ravbw byEdMwlal Commfttaas of the SPE and lADC. PermksJmI to copy b msffwdto anabsfmof ofnofmomtkwti. U.S.A. Tef% 1SS24S SPEUT. . . . ....--.A , A.... qu IUIWWi. nr~ad. Write [Link], SPE, P.0, S4X ~, Rk_fh n 79XMS26, contain conspicw uaacimdsdgmror WIIOIVEI. *J . .. ... . . _ .. . . . . . . . . .

for

Abstract
We review the conflicting literature on gas migration vehxities during kicks. We consider the laboratory and large scale test data that shows that for any locat gas void fraction of more than 10%, the influx migrates at approximately 100 fthnin. We atso review the evidence from field experience that shows that gas can migrate much more sIowly (the typicat rule of thumb suggests that gas bubbles move at approximately 15 fthnin) and in some cases remain stationary. We show that the yield stress of the drilling mud which holds cuttings in suspension whilst mrddngconnections,can also hold gas bubbles in suspension, and report an expefimentst study of these gas suspension effects. Significant volumes of gas can be held in suspensionduring a gas kick uiitii tiie find k ~lithis trapped gas remaining Statkriiar-y culated out of the well. We consider the implications of this for well control operations, and present field data where gas was injected into a marine riser, it dispersed and remained stationary until circulated out. We show that a single bubble migration model, which neglects gas suspension, predicts that as the gas rises and eXptUNIS it unloads the risa. By simulating the gas suspension characteristics we model the field data. We conclude that gas in moderate concrxmations(more than 10%)migrates quicldy, typicatly at 100 ft/min. This migrating influx leaves a trait of suspended gas in the mud that remains stationary. For small kicks in deep wells the entire influxcan be distributed, at a low concentration, and remain in suspension until the gas-cut mud is circulated out of the well.
Referencesand illustrationsat the cod of paper

Gas Migration Velocity - Literature Controversy

There is a major controversy in the published literature ova gas migrationrates during kicks while drMng. Experimental tests in small flow loops and in teat wells ahow the gas migration velocity is around 100 ft/min, while field estimates suggest that gas rises at around 15 fthin or more slowly. We consider this dqancy. Johnson and White [1] showed rha~ in typicat drilling g~ ometries, in reatistic tilting fluids, for gas concentrations larger than 10%, gas migration velocities were around 100 Mnin, significantly larger than the equivalent migration rates in water. The viscosity of dritting mud hinders the bubble break-up proms allowing gas to migrate as bigg= bubbles (which travel faster). They also observed that the yield stress of the drilting mud woutd hold low concentrations of gas in suspensionwith no migration. Radef et al. [2] reported similar resutts for gas migration in a 3.7 m, (12 ft) flow loop and a 1800 m (6,000 ft) well. The gas veloeity in the well was measured using the time of flight principle. Hovkmd and Rommetveit [3] reported large state tests in a 1500 m (5000 ft) deep test well, which had a maximum deviation of 63. They used the time of flight between pressure transducersmounted at different depths in the well to measure a gas slip velocity of 0.55 xn/s(110 Mnin). A widely acceptedrute of thumb used in the field says that u---- , DUDDKS IN@te ti 0.0% WJS,(15 fttftiifi). Iii@Xtt [4] gSSL claimed that he had evidence of gas migration rates of around 0.014 rnk (3 fthnin), sKhough he did not specify how these were derived. In field situations an accurate estimation of gas migration during a well control incident is very difficult. Velocities 93

GAS MIGIUTION: FAST, SLOW, OR STOPPED

are sometimes derived from casing pressure rise rates during shut-in, using a simple correlation which assumes that the wellboreis a rigid leaktight vessel filled with incompressible mud. However, Johnson and Tarvin [5] showed that Uds procedure could underestimatethe actual migration velocity by more than 10 times. A time of flightmeasure of gas velocity is more reliable than surface pressure interpretation. However, in a well control incident, it may be difficult to identify when the gas first entered the well. In terms of the gas arrival time, there are many well control incidents where EARLY GAShas arrived at surface indicating a larger velocity than expected, but there are also cases where the gas has taken many days to appear, often only arriving at the top of the well when the mud is being circulated out. Although the precise time of gas entry into the well may not be known, tlds evidence shows that in some cases the gas migrates very quickly while elsewhere it moves extremely slowly or, more likely, not at all.

The magnitudeof the yield stress is dependent upon not only the mud type and consistency,but for a bentonite based mud it is also dependent on the shear history. If a Bentonite mud remains unsheared then the yield stress will increase. Figure 1 shows the evolution of yield stress for a 6% bentonite mud. We see that over a period of 1 hour the yield stress of the mud increases considerably. In terms of a well control operation when tbe well is shutin and the influx has ceased, the yield stress of the mud column increases. This has implications for the volumes of gas which can be held in suspension. In the field it is conventional pracdce to measure the 10 second and 10 minute gel strengths. These give some information on the short time characteristics of the mud, but they give no information on the long time (i.e. shut-in time scale) characteristics. For typical field fluids the 1 hour gel strength can be in the range 10 to 100 lbs/[Link], although typical values would be below 50 lbs/100sqft.

Experimental Gas Migration - Bimodal Bubble Distri~ti$;~~,

Many authors, notably Johnson and White [1] and Phillips et al [6] have reported that visual observations of gas drilling mud flows show a bimodal distribution of bubble sizes. For gas void fractions larger than 10% the flow is dominated by large bubbles which almost fill the pipe and migrate up the well at a high velocity. In addition to this the liquid phase holds a suspension of very small bubbles which remain stationary relative to the fluid. These bubbles me typically smaller than 2 mm in diameter and are held in suspension by the effective yield stress of the mud. Initially, the bulk of the gas is carried in the large bubbles. However, in shedding a trail of small gas bubbles (which can become significant in total volume) the gas cloud becomes smaller. If a well is deep enough, all of the migrating influx WKIbecomesuspended and gas migration Wli cease completely (for a large influx the well would have to be very deep). The suspended gas void fraction will be dependent on the theological characteristics of the drilling mud so we must consider these characteristics and how they can vary during a well control oneration. =_-. .-
Wm.,+7. w ,,,aw

Tests
..UWJ

am .aw-i-ant.l au *p ,1.,.1-

.h,t+.rtfl [Link]; oatr=the . . . ..U5CAW b..

hllhhle ...

break-up and suspension processes and to evaluate the effects of mud properties on the gas void fractions which can be held in suspension by the mud. The experimental programme was executed in the muitiphase flow loop test facility at Schlumberger Cambridge Research, Figure 2, which forms a universal multiphase flow test centre. The present test configuration, described in detail by Johnson and White [1], is used for gas-liquid flows, although solid-liquid and liquid-liquid flows can also be evaluated. The facility offers a straight flow length of almost 15 m [49 ft], 13.5 m [45 ft] of which is perspex to permit visual evaluation of the flows. The piping is mounted on a table which can be pivoted, enabling tests to be carried out in all orientations from horizontal to verticti, although the teats reported here were made with deviations horn verticat to 60. It has been designed to permit tests in Newtonian and non-Newtonian fluids. Shear degradable liquids can also be used without damage to either the fluid or the facility. fhe facility is designed to operate at pressures up to 10 bar [145 psi]. lle working section used for the experiments reported here comprises 5 perspex pipe sections of 200 mm [7.81 internal dhmeter with a removable 89 mm [3.5~centrebody. The tests were made by suspending a column of gas-fkee mud in the loop, and then injecting a single bubble at the bouom. After the bubble had risen up through the column l.a --.:-:. ....-a..AA. . . ..,.:A &,.,.*: ,... .... . . ..a .....d LUG 1GUlalllulg NqJGuusal go v Ulu u abuuu w aa Jluiasu w
G

Mud Rheology
Drilling mud has a yield stress to prevent sedimentation of cuttings when pumping is suspended. The small bubbles are held in suspension in a similar manner.

using differential pressure transducers. We used two Xantham Gum solutions as the liquid phase for the tests. These fluids have a yield stress, so they simulate the low shear behaviour of conventional drilling muds. 94

A. B. JOHNSON & I. REZMER-COOPER&T. BAILEY &D. MCCANN

However, they offer some significant advantages over conventional Bentonite fluids. Primarily, this yield stress is independent of shear history, making the experimentalanalysis much more simple. Also the fluids are transparent allowing visual observations of the fluids before during and after the large bubbles have risen through the column. A rheogram of the fluids is shown in F@ure 3. We made a comprehensive experimental study of the gas -----: --- -La --.4 10U** I141W mnm!ar and tool joint .,:O= ffi. Suspcllslull Ularavb=! AU. y.~-, ------geometries, for single bubbles, trains of bubbles, and for bubbles of different sizes. Figure 4 shows the suspended gas fraction from a large number of tests with the two muds in an annular drilling geomety. The results are presented as the mean gas fraction (averaged over many tests) and the limits of scatter in this data, plotted against the yield stress of the mud. The scatter in the measurements was random, with large variations in the measurementsat different points in the column, even for the same test. This scatter did not correlate with any of the experimentalvariables, or location along the column. We apply a quadratic fit to the data. If we consider that in a well control operation the yield stress can vary in the range 10 to 50 lb/100sq f~ then we can expect the suspended gas fraction to vary in tbe range 0.5 to 5%.

bbls of influx could remain in suspension before the influx migrated up to the top of the well. This is in addition to any influx which dlssolvea into the mud. For a 5% suspended gas fraction the suspendedinflux volume could be 1I.8 bbls. Once this gas is held in suspension,it will remain there until the mud rheology breaks down, not arriving at the surface until the gas cut mud is circulated out of the well. (2) llte effect of the suspended gas on the wellbore pressure rise rate is more difficult to calculate. Johnson et al [51 showed that the system compressibWy (mud compressfDiiity + wellbore elasticity) and fluid loss were all-important in determining the wellbore pressure rise rate. They did not separate the effects of a suspended and migrating gas fraction in their model. Neglecting the effects of wellbore elasticity and fluid 10SS we extend their model to show that the shut-in pressure rise rate, P,, can be written as:
P. = Xg Vg~/2gV,lip Xmudvm.d +

Xgvgt . ...

. . . ...(2)

Suspended Gas - Well Control Implkations

There are two effects that the suspended gas fraction will have during a well control operation. (1) As the gas migrates up the well it will leave a trail of gas behind, reducing the volume of gas migrating. (2) Also the gas left in suspension will act to increase the compressibility of the mud in the well and therefore reduce the rate of the shut-in surface pressure rise. We consider these two effects separately. (1) To calculate the volume of an influx which can be suspended in a well we must consider how the intlux volume changes as it migrates. We consider a case where fluid is being bled off through the choke to maintain bottom hole pressure. As the intlux rises up the wellbore its pressure falls, and volume increases. We need a calculation of the influx volume which can be suspended, in terms of the initial volume at bottom hole conditions. This is derived in Appendix A. The suspended gas volume under bottom hole conditions, v, is:
V* = +8Ad

X and V are the volumetric compressibilityand volume respectively of the mud and gas identified by subscripts mud and g. The denominator of thk model is the total system compressibility, so Xgt Vgt is the compressibility of M of the gas in the wellbore. The numerator is the growth rate of the migrating gas, so Vgm is the volume of migrating gas. As the migrating gas cloud becomes smaller, as it leaves a trail of suspended gas, the system compressibility will the wellh(ye not change, but V$~ WUI IX iediiced &c will -... . . . . . . . . -. pressure rise rate.
. ...11 Ua -

We have calculated (see Appendix B) the effect of Udsgas on the wellbore pressure evolution as an influx migrates up the well. For simplicity we neglect the effects of wellbore elasticity and fluid loss. For a 10 Bbl irdhtx, migrating at 100 ft/min, in a 10,000 ft well we calculate the pressure increase as the gas migrates. In figure 5 we show 3 traces, assuming no suspension, 2% and 5% gas suspension. The effect of gas suspension on the wellbore pressure rise rate is very significant. If there is nq gas going into suspension the surface pressure would rise by nearly 5,000 psi. With 2% gas suspension the surface pressure will rise by only 3:000 psi, before the influx reaches the top of the well. But for the 5% suspended gas fraction the maximum pressure increase is less than 1000 psi, with all of the gas going into suspension. For the cases of a trail of suspended gas remaining in the well the rate of surfacepressure rise would give an inaccurate(low) estimate of the actual gas migration rate.

. . . . . . . . . . . . . . ...(1)

cr$is the suspended gas fraction, A the annular cross section aod d the well depth. For a 10,000 ft weil with 8.5 hole and 5 drill pipe, if the mud could sustain a suspended gas fraction of 2%, 4.7 95

Suspended Gas - Field Evidence

An analysis was made of data horn Amoco teats [7] conducted on a deep water riser to examine the effect of gas release from a BOP. Various gas volumes were injected into

GAS MIGRATION FAST, SLOW, OR STOPPED

a 3118 ft, 183 x 17* marine riser. The influx was injected into 13.2 ppg water based mud. When 10 Bbl of gas were injected no gas appeared at the surface during the test period of 3.6 hrs. 200 Bbls of mud were then circulated through the riser, and gas finally appeared at the surface as small bubbles boiling out of the mud. A simulation of this test with a single bubble model shows that the intlux would migrate up the well until the top of the bubble was at 335 ft from the top. It would then expand very rapidly, unloading the top 670 ft of the riser. There was no measurement of the long time scale yield stress, so an accurate prediction of the suspended gas fraction is not possible. Figure 6 showsthe influxvolume which can be suspendedin the riser as a function of the suspended gas fraction. This is calculated from equation 1. If the mud can sustain a suspended gas fkaction of 2.5% then all of the injected gas would be suspended in the well with none reaching the surface. When 20 Bbls of gas were injected into the riser, the mud at surface started to bubble after 9 minutes, but it did not unload the riser and the riser woutd not flow during the next 3.6 hours. When the influx was circulated out the mud continued to bubble. Later, after pumping more than 500 Bbls of mud, an estimated 6 Bbls of gas came out of the well in a slug. Clearly all 20 Bbls of gas were not behg held in suspension, so (from figure 6) the suspended gas tiaction was less than 5%. The gas cloud rose up the riser at a very high speed. From the injection time and the surface arrival time we find a gas velocity of 1.7 rnh (330 fthnin). This is not surprising for such a large annular geometry. As the gas cloud migrates it leaves a tail of gas in suspension in the well and tfds comes out more S1OW1 y as the mud is circulated out. he rapid slugging, later in the circulation, occurs as gas from close to the bottom of the riser, which had been held in suspension,expands, to form a large gas cloud which rises rapidly through the riser.

Directorate, Bailey et al [8] reinforced some of the above ideas. Using a kick simulator it was shown that when a (5 Sxtidi~OUnt Ofg= trapped iUidei ~i =imi~- pieiK5iib3i bbl) is released into a risw containing water-base mud, in very deep water (> 1000 m) the gas will become d@ersed in the riser. In certain cases, where some gas migrates into the riser on PIWGUW1, ~L wUeluvllou & ,11 opening he UIIIIUIiU deep water the gas becomes dispersed to such a degree that it represents less of a threat than possibly supposed.
---..1 ... . . . .. . . . . :. . . ... A . . ... . ..t..l+aA * .* ;

Conclusion
Gas migration through drilling muds is complex and cannot be described by a single slip velocity or rule of thumb. . At large gas concentrations (> 10%) the gas wilt rise FAST, at around 0.5 m/s (100 ft/min) in a typicat drilling geometry. . This rapidly moving gas cloud will leave a trail of bubbles suspended in the well by the yield stress of the mud. llese small gas bubbles will be STOPPED.
G

MIs interpretation of surface pressures during shutin will indicate that gas migration is SLOW.

The suspended gas has a huge effect on the shut-in surface pressure rise rate, reducing the terminal wellbore pressure and the rate of pressure rise significantly. In deep wells, or wells with large annular geometries the volume of gas suspended can become very significant in relation to the totat influx volume. In some cases the entire influx can become suspended and remain stationary indefinitely, not aniving at surface until the gas cut mud is circulated out of the well. ~is is particularly important in deep water well control operations when gas is released from the BOP. The size of the suspended gas fraction is dependent on the yield stress of the mud. To make an accurate prediction of thk level we need a measure of the gel strength of mud left at rest for a similar period to the shut-in time.

Implications
If an influx is taken and a well shutin, the maximum surface pressure occurs if the gas cloud is allowed to rise to the surface. If the effects of gas suspension and system corrlpreasibility are neglected the influx will carry the bottom hole pressure to the BOP. Gas suspension has a dramatic effect (a huge reduction) on these maximum wellbore preasurea. With a knowledge of the mud gel strength, and the gas suspensioncharacteristics(discussedhere) we show that the maximum wellbore pressures can be much smaller. In a modelling study conducted as part of a co-ordinated project on deep water drilling for the Norwegian Petroleum
96

References
[1] A.B. Johnson and [Link]. Gas rise velocity during gas kicks. SPEDriUing Engineering, December(20431), 1991.
[2] D.W. Rader, A.T. Bourgoyne Jr, and R.H. Ward. Fac-

tors affecting bubble-rise velocity of gas kicks. J. of

Petroleum Technology, May 1975.

A. B. JOHNSON & I. REZMER-COOPER&T. BAJLEY &D. MCCANN

[3] F. Hovland and R. Rommetveit. Analysis of gas - rise velocities from full scale kick experiments. SPE Technical Con., Washington, (SPE24580E331 340, 1992. [4] E. Blount. Editorial comment. SPE Drifling Engineering, page 236, 1991. [5] A.B. Johnson and J.A. Tarvin. Field calculations underestimate gas migration velocities. IADC Etqooeun Well Control Con&, 1993. [6] J. Philip, J. M. proctor, K. Niranjan, and J. F. Davidson. Gas holdup and liquid circulation in inter@ loop reactors containing highly viscous newtonian and - ----.,.*-; O 1in..iri. 45(3):651-664, 1990.
.tu.,-.i~w [Link] ..yusuo. ~ham+nl wr=rr.m--e J7n ur.~ oinn~nno -w-, .,.s Ginnrv u-.--.. -,

Summing over each section of the wellbore(where nu and d are the upper and lower limits of the element n) with a different cross section. Assuming a constant annular cross section, tlds can be simplified to
-an Ad . . . . . . . . . . . . . . . . .(7) 2

t), =

Appendix B: Wellbore Pressure Rise

We consider a case where the influx has risen from the bottom of the well d to depth d~. The wellbore pressurehas increasedby AP, so the bottom hole pressure is ~~h+ fi~ and the gas pressure P~ is:
Pg spbh + () +Ap.., . . . . . . . ..(q

[7] J.M. Shaughnessy. Test of the effect of gas in a deepwater risa-. Technical repofi Amoco Internal Memorandum, 1986. [8] T. Bailey et al. NPD Voring Plateau Deep Water Drilling Project, Well Control, Final Report. Technical Report NPD/REP/Final 1, Sedco Forex, 1994.

We consider conservation of gas in terms of the gas under initird bottom hole conditions, i.e. volume at pressure Pbh, identified by subscript M. So we wriw vbh= v,bh+v~bh . . . . . . . . . . . ...(9)
Vbh is the initial influx volume, and subscripts s and m indicate suspendedand migrating gas respectively. We also consider conservation of wellbore volume

Appendix A: Suspended Gas Volume

If we consider a worst case, where the wellbore pressure remains constant as the gas migrates then for a well of depth d with bottom hole pressure p, then the pressure at depth dd (iOcdj is:

.,. . ., -1.,
Uoh .S T .tn

AY T

.1/ [Link]--.

D . . . . . . .(!!!)

Pd=p+.................(3) X~.~
For the gas we assume that pv = const, then the volume of an influx under bottom hole conditions, W,which has a volume Vdat depth dfj is:

and are the mud compressibility and volume respectively. (For simplicity we neglect wellbore elasticity and fluid 10sS here).
Vm.d

We can also write

v, =~tA(d-dg)............(ll)

V=V~;=Vd+ ..............(4)
In a section of the well of cross sectional area A, of length dl at depth dd, the suspended gas fraction v,l is
v,l=a,Adl . . . . . . . . . . . . . . ...(5)
V,bh =

CY.A ~(d-d;)+~(d-ds) [

..(12)

v~ = Vmbh

()
Pbh

. . . . . . . . . . . ..(13)

P9

a, is the suspended gas fraction. Substituting U,I for Udin 4 and integrating over the length of the wellbore we find the suspended gas volume under bouom hole conditions, u, is:

Solving the above for AP we can calculate the wellbore pressure rise rate as the gas bubble migrates.

11 3a2zA(d~-d~~)

(6)

97

GAS MIGWITION FAST, SLOW, OR STOPPED

50

100 Time (rnins)

150

o~
o 500

1000

Figure 1: Evolution with time of yield stress for a 6% bentonite mud.

shearRate(1/s)
Figure 3: Rheogram of Xanthum solutions used for experimental tests.

5.0

4.0

3.0 2.0 1 /

jAK--J o

10 m 30 Yield Stress (lb/l(M) ftA2)

40

50

F@ure 4: Suspended gas fraction from a large numbtx of tests with the two muds in an annular drilling geometry. fa-

98

A. B. JOJ-LNSON & I. REZMER-COOPER&T. BAILEY & D. MCCANN

5000

n O

20

40 60 80 Tme (reins)

100

120

Figure 5: Shut-in wellbore pressure rise rate for casea of no suspension, 2% and 5% gas suspension.

\ 20 -

10 -

2 4 /.-,, .-J-J. - .. . iiuspenoeuwas rmcuon IYOJ

Figure 6: Influx volume which can be suspended in the marine riser used for field test. Volume calculated from equation 1.

99