JCPT01 - Advanced Two-Phase Well Control Analysis

Advanc ed Two-Phase Well Control Analysis 1/17
Jonggeun Choe, Seoul National University
A Revised Paper after the Peer Review
Abstract
Pressure responses during well control operations can be analyzed using the coupling solution of the conservation
of mass equations, the conservation of linear momentum equation, and the equation of state. For a realistic well
control simulation, a computer model should include the effects of well geometry, well control methods, formation
parameters, bubble rise velocity, and mud compressibility on well stabilization. Typical numerical problems
encountered in two-phase well control modeling are numerical dissipation, difficulty of selecting appropriate time step
and grid sizes, negative liquid velocity, and divergence due to distinct two-phase flow maps . Among these numerical
problems, the numerical dissipation is the most prominent and it should be handles properly for an accurate two-phase
well control simulation.
An unsteady state two-phase well control model is developed and compared successfully with the Nickens model,
the Santos model, and a single-phase model. However, an exact match is very difficult because of different two-phase
correlation models used and complexity of two -phase flow. Wellbore geometry, gas velocity, formation permeability,
and handling of numerical dissipation are found to be important factors in the prediction of the pressure behavior of
the kick fluids.
Introduction
As proven petroleum reserves decline through continued production, exploration for new oil and gas resources will
extend into environments which present significant economic risks and technical hurdles. A detailed study using
multi-company data disclosed about 8 to 10 billion bbls of oil equivalent in place for the deep water area of the Gulf of
Mexico outer continental shelf.(1) There are more than 15 known discoveries in water depths between 910 m and 2,300
m in the Gulf of Mexico(2) and there is great interest in accelerating the development of the known deep water
discoveries.
Since safety is one of the biggest concerns in drilling operations, the oil industry routinely trains its personnel in
areas which are critical for safe and economical drilling procedures. One of these major areas is well control. Well
control includes not only kick prevention and kick detection but also the process of removing kick fluids from the
kicking well and circulating heavy drilling mud under controlled conditions.
Well control simulation has received attention in recent years because of its applicability and flexibility, and many
computer models (3-10) have been developed to analyze the behavior of a kick. The main objective of a kick simulator is
to predict pressure and volume behavior of kick fluids as a function of time. One of the earliest mathematical models
was published by LeBlanc et al. assuming a known volume of gas kick as a single-phase.(3) They ignored effects of
frictional pressure loss (FPL), gas-mud mixture, gas slip velocity, and reservoir parameters.
Nickens presented a dynamic two-phase computer model of a kicking well.(4) He set up finite difference equations
(FDEs) for uniform wellbore geometry. Nickens also applied those equations for variable wellbore geometries by
changing the gas and liquid velocities at the locations of changes in area. He did not modify the conservation of
momentum equation for uneven wellbore geometries. The same equations were used by Starrett et al.(5) and Santos,(6)
but their research was limited to uniform wellbore geometry. Therefore, it is necessary to develop a well control model
to simulate realistic kick behaviors in variable wellbores with the consideration of mud compressibility and reservoir
parameters.
This paper presents a dynamic two -phase well control computer model to accurately analyze the behavior of kick
fluids based on a realistic assumption of unsteady state two-phase mixture flow. Two new sets of finite difference
equations are developed to account for the effect of changing flow geometry. The paper also investigates the
numerical problems associated with a two -phase well control model and their management.
Governing Equations
The appropriate assumptions and governing equations are critical to simulate realistic two -phase well control
operations. The two-phase model in this study is based on the following assumptions:
• unsteady state two-phase flow

• one-dimensional flow along the flow path
• water-based mud, where gas solubility is negligible
• incompressible mud
• known mud temperature gradient with depth
• kick occurs at the bottom of the well while drilling
Two-Phase Mixture Region

Eight variables describe the two-phase flow system completely. They are pressure, temperature, gas and liquid
fractions, gas and liquid densities, and gas and liquid velocities. Starrett et al.(5) started with incorrect governing
equations for their momentum balance equation for a two-phase mixture. However, they ended up with the correct FDE
for the momentum balance equation, which was taken directly from the Nickens paper. White et al. (7) considered the
flow area change in their continuity equations, but they disregarded the flow area term in the momentum balance
equation. This resulted in inconsistent units in their momentum balance equation.
There are still five unknowns such as gas and liquid velocities, gas fraction, pressure, and gas density based on
the above assumptions. Therefore, five equations are required to calculate the unknown variables with boundary
conditions.
The conservation of mass equation for mud:
∂
( Aρ m H m ) + ∂ ( A ρ m vm H m ) = 0 ……………………………………………………………………… (1)
∂t ∂x
The conservation of mass equation for gas:
∂ ∂
(
Aρ g H g + ) ( )
A ρ g vg H g = 0 ……………………………………………………………………….. (2)
∂t ∂x
The conservation of linear momentum equation for the mud-gas mixture:
∂p
∂
[ ]
A(ρ m vm H m + ρ g v g H g ) +
∂
[ ] ∂p
A(ρ m vm2 H m + ρ g v 2g H g ) + A + A f + A(ρ m H m + ρ g H g )g = 0 (3)
∂t ∂x ∂x ∂x
The two-phase correlation to calculate in-situ gas velocity:
v g = f ( vm , v g , θ , d o , d i , σ , p , T ) ……………………………………………………………………….. (4)
The equation of state to compute gas density:
γgp
ρ g = 3 .488 ………………………………………………………………….………………………… (5)
zT
Where, ρg is the gas density in g/cc, p is the pressure in mega Pascal (MPa), and T is the temperature in Kelvin.
Equations (1) through (3) have consistent units and the same form of equations are observed in references (8) and (9).
Single-Phase Region
Single-phase flow exists inside the drill-string and in part of the annulus. The annulus could have four regions: a
single-phase region above the two-phase mixture, the two-phase mixture region, a single-phase region for the old mud,
and a single-phase region for kill mud below the two-phase mixture.
For the single-phase flow, equations (1) through (5) above can be written in simple discrete form because of zero
gas fraction and incompressible mud.(9)
∆p ∆p v n +1 − v n
= 980 ρ + f + ρ …………………………………………………………………… (6)
∆L ∆L ∆t
Equa tion (6) is a general pressure gradient equation including hydrostatic pressure gradient, FPL gradient, and
acceleration loss gradient.
Frictional Pressure Loss

Even though frictional pressure loss is small in the annulus for a large well diameter at low kill rate, FPL is critical
for slim-hole wells or inside the choke line for offshore wells. FPL is considered to achieve more realistic simulation of
kick behavior for all flow geometries and flow rates. The Power-law fluid model is assumed. Detailed equations are
available in reference (10). The estimation of two-phase FPL is required to calculate the two-phase mixture momentum
balance equation. The two-phase well control model utilizes the Beggs and Brill correlation.(11)
Gas Properties
Even though gas density is small compared to drilling mud density, the correct evaluation of gas density is
essential to calculate the hydrostatic pressure of the two-phase mixture and to predict bubble rise velocities. The gas
compressibility factor, which is function of pseudo-reduced properties, is calculated from the equation proposed by
Dranchuk et al .(12) Gas viscosity is obtained using the Lee et al. correlation.(13) Surface tension, which is necessary to
estimate bubble rise velocity, is determined by the Katz et al. Method.(14)
Gas slip velocity is one of the parameters needed to describe a two-phase system. It also affects initial gas
distribution and kick migration velocity during well shut-in. Since the two -phase flow phenomenon is very complex,
much experimental work has been done. After an intensive literature survey, the Hasan and Kabir model(15) was chosen
in this study for the following reasons:
§ Since they did a lot of experimental work for two-phase flow, their model was good for pipe flow and annulus
flow for vertical and deviated wells. They also ran tests for different annulus geometries.
§ They provided all the equations so that those equations are ready to use.
§ They clarified some contradictions for two-phase flow in deviated wells.
Solution Procedures
Drilling to Kick Detection
The two-phase model starts the simulation by taking a kick while drilling after 2 minutes of drilling. Gas inflow rate
is calculated by assuming an infinite-acting homogeneous reservoir.(16) The gas distribution is calculated by the Hasan
and Kabir correlation using the mud circulation rate and gas influx rate from the formation. All parameters in the two-
phase mixture region are evaluated at the middle point of the two-phase mixture weighted by the effective gas fraction.
Since initial gas kick volume is relatively small, the above approximation gives excellent results to compute pressure of
the kick and the flowing bottomhole pressure. The effective flow rate for the single-phase region above the two-phase
region is the summation of mud circulation rate and gas inflow rate.
One of the primary kick warning signs is increased mud return rate. The next step is to confirm that the well is
flowing after the surface pump is shut down. This is the same as the “Drilling” stage except for reduced flow rate
without circulating mud. The same calculations are repeated here.
Shut -in
The next important step after detecting a kick is to shut the well in to prevent further influx from the formation.
However, there is some flow from the formation as long as the BHP is less than the formation pressure. Since the total
system volume is the same after well shut-in, further inflow results in BHP increase. If BHP rises up to the formation
pressure, the system has reached pressure equilibrium. At this point shut-in drill-pipe pressure (SIDPP) and shut-in
casing pressure (SICP) are recorded.
The amount of pressure build -up for the given duration can be calculated from equation (7), if a gas kick is assumed
compressible.(9)
z n +1 n n+1
p n +1 = p n ………………………………………………………………………………………… (7)
z n nn
Where, n is the number of moles of the gas kick.
In this study, mud compressibility is also considered to compute pressure buildup during well shut in.
∆Vkick
p n +1 = p n +
(CmudVmud + CkickVkick )
………………………………………………………………… (8)
1 1 dz
C kick = − ……………………………………………………………………………………… (9)
p z dp
The two-phase region is discretized after well stabilization with the length specified by input data. An additional
grid is given wherever flow area changes to make the calculation of grid block volume easy and to reduce complexity
of FDEs. All grid information is assigned for each grid point as initial conditions.
Circulation
A constant BHP and kill rate are pressure and flow rate boundary conditions at the bottom of the well, respectively.
If there is a single-phase region below the two-phase mixture, the pressure at the bottom of the two -phase mixture is
calculated from equation (6). For the two-phase region, a fully implicit finite difference method is applied for each
discretized grid segment. Here are the calculation procedures for two-phase flow when the annular area increases or
stays the same.
1. Calculate new gas fraction at i-th cell from equation (A2).

2. Compute new liquid fraction, Hm = 1.0 - Hg .
3. Estimate new liquid velocity at the i-th cell from equation (A1).
4. Obtain new gas velocity from the Hasan and Kabir correlation.
5. Determine the gas fraction with updated values from equation (A2).
6. Check the convergence of gas fractions. If they are close enough, go to step 7, or repeat steps 1 through 5.
7. Evaluate new pressure from equation (A3).
8. Check the convergence of pressure.
9. If pressure does not converge, repeat the whole procedure until a converged pressure solution is obtained.
The procedures are repeated for the adjacent downstream cell. The iteration carries on to the end of the two-phase
region. Material balance error in terms of total weight of the kick in the two-phase region is calculated to check degree
of numerical dissipation. If absolute material balance error is greater than 0.01 %, the whole calculation is repeated until
the criterion is satisfied. If the flow area decreases, equations (A4) through (A6) should be employed instead of
equations (A1) through (A3), respectively.
If there is a single -phase region above the two-phase mixture, equation (6) is applied from the top of the two-phase
mixture to the surface. The effective flow rate in the single-phase above the two-phase region is the summation of the
kill circulation rate and increased flow rate due to the kick expansion over the time step.
Numerical Problems and Their Management

Divergence
Many numerical problems have been observed while programming this dynamic two-phase well control computer
simulation. One of them is a divergence problem. The Hasan and Kabir correlation (15) utilizes a flow pattern map which
has a distinct boundary with different gas slip velocity. Therefore, the iterations between gas fraction and gas velocity
do not converge near the flow regime boundary.
A modification of the flow map based on the gas fraction is necessary to have a converged solution. The slip
velocity between any two flow regimes is assumed to vary linearly with gas fraction.
If Hg < 0.25 bubble flow

If 0.55 < Hg < 0.75 slug flow
If Hg > 0.9 annular flow
Negative Liquid Velocity

Sometimes the liquid velocity from equation (A1) or equation (A4) becomes negative and causes numerical
problems for functions which require a positive argument. Negative velocity is possible physically because gas moves
faster than liquid while circulating mud. In other words, the liq uid flows back to fill the void portion created by gas
migration. If the calculated liquid velocity is negative, two modifications are made. One is that gas velocity is assumed
the same as the slip velocity at that grid point. The other is that FPL for that grid is negligible. As iteration carries on
to the next downstream grids, the liquid velocity becomes gradually positive due to gas expansion.
Time Step and Grid Size

The selection of an appropriate time step and grid size has been an important question in numerical simulation.
Starrett et al.(5) and Santos(6) computed liquid and gas velocities first in their calculation procedures, and then they
evaluated the gas fraction from the two-phase flow correlation selected. As can be seen from equations (1) and (2), the
fluid velocity does not have any time derivative. Therefore, as can be seen from equations (A1) and (A4), a large time
step gives a numerically stable solution compared to a short time step. Starrett et al. continuously increased the time
step size to make all velocities positive if there was any negative velocity of liquid or gas.
In this study the gas fraction which has a time derivative is computed first. The gas holdup calculation yields a
numerically stable solution for a small time step, which is very logical. On the other hand, a very small time step could
cause a negative liquid velocity and produce a large output file. To solve the time step selection problem, the time step
is chosen so that the top interface of the two-phase mixture moves exactly one grid spacing in one time step. About 15
to 25 m grid spacing generally gives good results.
Numerical Dissipation
Numerical dissipation is the phenomenon whereby the magnitude of the numerical solution tends to decrease and
to spread out over a wide range with a small magnitude. This means that grid points with non-zero gas holdup exist
above the top interface assumed. Gas holdup at the bottom interface of the two-phase mixture has a small va lue that
eventually becomes zero.
To solve the gas holdup dissipation problem, the movement of the top interface is limited to move one grid spacing
according to the time step size determined before. The movement of the bottom interface is determined by a minimum
gas holdup of 0.005 or one-tenth of the initial gas holdup, whichever is smaller. One side effect observed is that
material balance error increases by limiting gas movement at the top and bottom interfaces. Therefore, additional
iterations are repeated until the summation of absolute relative material balance error is less than 0.01 %.
Sensitivity Runs
After managing all the numerical problems discussed above, several runs are made to check the sensitivity of
surface choke pressures and surface pit volume gains. Table 1 shows all default data used in this study unless
otherwise specified. Table 1 represents typical well configurations in the Gulf of Mexico for deep water wells. Table 2
shows relative % errors on maximum surface choke pressures and maximum pit volume gains with the corresponding
time. Constant gas influx of 1,136 liters/min (300 pgm) and 1.59 m3 of kick volume gain are used to minimize effects of
other parameters. The relative % errors are calculated based on values at the grid size of 10 m.
As can be seen in Table 2, the two variables vary a little as grid size changes. However, the relative % errors for
maximum choke pressure and pit volume gain are less than 0.8 % as long as the grid size is less than 25 m. Large grid
size also implies large time step size as described above. Since kick behaviors during well control are affected by many
factors, an exact match is very difficult especially for gas velocities for the whole depth range of 4,572 m. Relative %
errors on the corresponding time can be as high as 4.7 %, if grid size is 30 m.
Comparison with Other Models

The Nickens Model
The two-phase well control model has been compared with the Nickens well control model and a single-phase
model. A single-phase model assumes that kick fluid enters into a well as a single phase and remains as a single slug
throughout the well control operations. The pressure and volume of the kick are two major unknowns. Those are
determined from dynamic equilibrium with a specified bottomhole pressure.(9-10)
p b z xT x
Vx = Vb …………………………………………………………………………………………. (10)
p x zb Tb
pb = p x + ∆p f + ∆phy ……………………………………………………………………………….. (11)
Table 3 lists all the data used by the Nickens model, the single-phase model, and the two-phase model for
comparison. Fig. 1 shows the comparison of choke pressures for the three models by the driller’s method. The two -
phase model matches the overall trend well with the Nickens model, but there is an almost constant difference in choke
pressure. The initial SICPs are very clo se for both the two-phase model and the Nickens model. There is about 3
minutes time difference for maximum choke pressure. This can be explained by both 4-minute time delay by the
Nickens model after well stabilization and the different gas slip models us ed. Nickens utilized his own gas slip model
partially combined with regression from published data.
Nickens did not explain the frictional pressure loss in the small choke line. Even if the 304.8 m, 7.62 cm (3-inch ID)
choke line is one-tenth of the total well depth, frictional pressure drop is large because of small inner diameter. From
the data in Table 3, FPL in the choke line is 476 kPa (69 psi), which is 71 % of the total FPL in the annulus for the kill
circulation rate. When all the gas kick is removed from the wellbore, the choke pressures from the two models should
coincide because there is only single-phase 1.2 g/cc (10 ppg) mud in the wellbore. As can be seen in Fig. 1, there is still
the same amount of difference. Another possible reason is that Nickens might use a different BHP as a pressure
boundary condition at the bottom of the well. The author assumes the BHP during the kick circulation out is the same
as the formation pressure.
For the single-phase model in Fig. 1, no gas slip velocity is considered. As expected, two distinct differences are
the magnitude of the choke pressure and time delay due to two -phase mixture effects and gas slip velocity, both of
which are ignored in the single-phase model.(9) However, the single-phase model closely ma tches the two-phase model
at the beginning of mud circulation after well stabilization, because the gas kick density is high and the kick does not
expand much under high pressure s in the wellbore. One very important point is that the single-phase model exa ctly
matches the required choke pressure except for the time delay when all the kick is removed out of the wellbore.
The Santos Model

The Santos model(6) was only valid for driller’s method with constant well geometry. Table 4 lists all input data
used by the three models. As can be seen in Table 4, Santos specified SIDPP, SICP, and initial pit volume gain. Two
out of the three will be enough to determine the rest value. He also specified initial gas fraction without considering
formation parameters.
Fig. 2 shows the comparison of choke pressures for the three models. The two -phase model matches the choke
pressure well until the top interface of the two-phase mixture arrives at the surface. After that, the Santos model
predicts high choke pressure for a long time. The two-phase model uses 568 liters/min (150 gpm) circulation rate and
1,704 liters/min kick influx rate to provide the given initial gas fraction 0.75.
Compared to the single-phase model, the gas kick by the Santos model arrives at the surface sooner than the
single-phase model, but the end point of the Santos model to remove all kick from the well is the same with the single-
phase model because Santos did not consider the gas slip velocity at the bottom of the two -phase mixture. All the
three mo dels almost overlap after the kick removal from the well because there is a single-phase liquid flow.
Results and Discussion

Several numerical problems and their management have been discussed before. Some of them are expected and can
be handled without much difficulty. Among them the numerical dissipation is the most pronounced and should be
mentioned here again. Fig. 3a shows surface choke pressure versus time with and without the management of
numerical dissipation. The choke pressure without numerical dissipation (i.e. with the management of numerical
dissipation) shows a typical choke pressure trend for offshore wells with a small ID choke line. The choke pressure
rapidly increases as the gas kick fills the small ID choke line to compensate for the hydrostatic pressure loss. The
choke pressure also rapidly decreases as the old mud replaces the kick in the choke line.
With the numerical dissipation (i.e. without the management of numerical dissipation), the choke pressure even
decreases without changing wellbore geometry. Note that it takes a longer time to remove all the kick out of the
wellbore because the kick spreads out with low gas holdup, especially at the bottom part of the two-phase mixture.
Low gas holdup also results in low gas slip velocity.
If the choke pressures in Fig. 3a are compared, it could give a false impression that the numerical dissipation is not
severe and it just results in less choke pressure. When pit volume gains are compared, Fig. 3b clearly shows the
numerical dissipation phenomenon. As the simulation time goes on, the pit volume gain with the numerical dissipation
decreases gradually. When the kick expansion is dominant near the surface, the final volume gain at the surface is less
than the initial pit volume gain which is clearly wrong. Therefore, the numerical dissipation should be managed
properly to predict the kick behavior correctly. Starret et al.(5) presented a shallow-gas kick simulator. Although they
ignored the numerical dissipation, their choke pressures predicted are not so strange because they assumed
continuous gas influx from the formation for an uncontrolled shallow gas. In other words, the numerical dissipation
phenomenon is shadowed by the continuous formation influx.
Three different formation permeabilities are employed to see the effects of gas kick influx rates. In order to simulate
actual field well control operations in detail, the following scenario is assumed:
1. Drilling to the target depth (2 minutes duration)

2. Taking a kick while drilling
3. Kick detection by a preset pit volume warning level (10 bbls)
4. Stop drilling
5. Shutting pump down to confirm the kick (10 seconds time duration)
6. Shutting the well in (20 seconds time duration)
7. Well stabilization (to the formation pressure)
Fig. 4a displays pit volume gain based on the scenario above. More problems are expected for high formation
permeability. For 1,000 md formation permeability, it takes only 1.6 minutes for a 1.61 m3 (10.14 bbls) gas kick gain. It
takes a short time so that the rig crew should watch all kick indicators very closely. There is an additional 1.24 m3 (7.8
bbls) kick due to 30 seconds reaction delay from kick detection to well shut-in. Pit volume gain at the surface remains
constant after well shut-in while the number of moles of the gas kick increases continuously until well stabilization.
For 10 md formation permeability, it takes about 13.2 minutes for 1.59 m3 (10.02 bbls) of gas kick. There is only 0.16
3
m (1.0 bbl) of additional pit gain after kick detection. Note that the kick volume detected is very close to the preset pit
warning level. It also takes a longer time for well stabilization. Therefore, a well control team in the field has more time
to prepare any necessary actions to bring the kicking well under control. Several minutes of time delay will be
affordable.
Fig. 4b shows surface casing pressure based on the scenario above. Before closing a blowout preventer, surface
casing pressure is zero. Shut-in casing pressure at the surface increases after well shut-in until the BHP balances with
the formation pressure. The lower the formation permeability, the longer the time to stabilize the well pressures. The
high formation permeability gives higher choke pressure mainly because of larger initial pit gain, and higher gas
holdup. SIDPP build-up shows the same trends as SICP build-ups in Fig. 4b. However, stabilized SIDPPs are the same
for all three cases regardless of initial pit volume gain because SIDPP is a function of formation pressure, mud density
in use, and well depth.
Fig. 5 shows a comparison of surface casing pressure buildup after well shut-in with and without mud
compressibility. Formation permeability is 10 md and initial pit volume gain is 1.75 m3. If mud compressibility is ignored
(Eq. 7), the surface choke pressure gives fast stabilization with less number of gas kick moles than that with mud
compressibility. For realistic pressure buildup, mud compressibility should be considered (Eq. 8). However, the final
stabilized SICP and SIDPP will be the same except for time delay, if the remaining conditions are the same.
There are small ID choke and kill lines on offshore wells. These lines are characterized by small volume capacity
and high flow friction during well control operations. Typical choke pressure trends can be seen in Figs. 1 and 3a. If
well control crew is not educated and trained to handle these situations, there is high probability of having an
additional kick influx or underground blowout by breaking down the formation.
Two ways are suggested in the field to minimize operational problems for offshore well control due to the choke
line. One is to practice very low pump rates to have more time to displace the choke line. The other is to utilize both
lines as kick fluid paths which will enlarge volume capacity and displacement time. Fig. 6 shows comparison of choke
pressures when choke line and kill line are used together. If the both lines are used, the corresponding flow rates have
to be determined based on the following equations. They can be used for multiple choke and kill lines.
q total = q choke + q kill …………………………………………………………………………………….. (12)

∆ p f ( q choke ) = ∆p f ( q kill ) ……………………………………………………………………………… (13)
As expected, the choke pressure with the choke line only requires lowest surface casing pressure before the kick
fills the choke line because the choke line gives more friction loss. High friction loss also justifies the use of the
dynamic kill method and the kill rate should be chosen carefully so as not to exceed the formation fracture pressure.
When the kick starts to fill the both lines, the surface choke pressure is lower than that of choke line alone because of
increased capacity by the both lines. It also takes a longer time to displace both lines. Since 7.62 cm ID kill line has
small capacity, the effect on choke pressure by using the additional kill line is not prominent. However, its effects on
surface choke pressure for the additional 10.7 cm ID kill line can be clearly seen.
Conclusions
The following conclusions have been drawn from the two-phase well control study:
• A dynamic two-phase well control model has been developed. It can handle variable well geometry with a realistic
assumption of unsteady state two-phase flow. Two new sets of finite difference equations are derived and utilized
to handle variable flow geometry.
• The typical numerical problems for two -phase well control simulation have been identified such as numerical
dissipation, difficulty of selecting appropriate time step and grid sizes, and divergence due to a distinct two-phase
flow map. Their management has been discussed. Among several numerical problems identified, the numerical
dissipation is the most prominent and it should be handled properly for a realistic two-phase well control
simulation.
• The two-phase model has been compared successfully with the Nickens model, the Santos model, and the single-
phase model. However, an exact match is not possible because of different two-phase flow correlations used and
different assumptions.
• Gas slip velocity is important to estimate the time of maximum choke pressure and it should be considered both at
the top and bottom of the two -phase mixture.
• Mud compressibility should be considered for a realistic simulation of pressure build up after well shut in.
• Choke pressures highly depend on kick height in the annulus so that well and drill-string geometry should be
considered in detail for the well control study.
• A kick with high formation permeability could result in a very large pit gain if the kick is not detected and reacted
to quickly. It also gives fast well stabilization after well shut in.
• Both choke line and kill line can be used to reduce maximum surface choke pressure but the amount of choke
pressure reduction is highly dependent on kill line sizes.
NOMENCLATURE
A = area of the conduit, cm2
C = mud compressibility, 1/(dynes/cm2)
di = inner diameter of the annular, cm
do = outer diameter of the annular, cm
g = gravity acceleration, 980 cm/s 2
H = holdup or volume fraction
n = number of moles
p = pressure, dyne/cm2
q = flow rate, liters/min
T = temperature, Kelvin
t = time, second
V = volume, cc
v = velocity of the fluid, cm/s
x = vertical coordinate
z = gas deviation factor
∆L = measured length of an interval, cm
∆phy = hydrostatic pressure, dymes/cm2
∆x = vertical height of a fluid, cm
θ = inclination angle, degrees
σ = surface tension, dynes/cm
γ = specific gravity
ρ = density of a fluid, g/cc
Subscripts
b = bottomhole
f = friction
g = gas-phase or gas kick
i = index of each interval or grid
m = liquid-phase or mud
x = depth of analysis
Superscripts
n = old time level
n+1 = new time level
REFERENCES
1. Lawrence, D.T. and Anderson, R.N., Details Confirm Gulf of Mexico Deepwater as Significant Province; Oil & Gas
J., pp. 93-97, Vol. 91, No. 21, 1993.
2. Salama, M.M., Some Challenges and Innovations for Deepwater Developments; Paper OTC 8455, 1997 Offshore
Technology Conference held in Houston, Texas, 5-8 May.
3. LeBlanc, J.L. and Lewis, R.L., A Mathema tical Model of a Gas Kick; AIME Trans., pp. 888-898, Vol. 243, 1968.
4. Nickens, H.V., A Dynamic Computer Model of a Kicking Well; SPEDE, pp. 159-173, Vol. 2, No. 2, 1987.
5. Starrett, M.P., Hill, A.D., and Sepehrnoori, K., A Shallow-Gas-Kick Simulator In cluding Diverter Performance;
SPEDE, pp. 79-85, Vol. 5, No. 1, 1990.
6. Santos, O.L.A., Well-Control Operations in Horizontal Wells: SPEDE, pp. 111-116, Vol. 6, No. 2, 1991.
7. White, D.B. and Walton, I.C., A Computer Model for Kicks in Water- and Oil-Based Muds; Paper SPE/IADC
19975, 1990 SPE/IADC Drilling Conference held in Houston, TX, Feb. 27-Mar. 2.
8. Rommetveit, R., Kick Simulator Improves Well Control Engineering and Planning: Oil & Gas J., pp. 64-71, Vol. 92,
No. 34, 1994.
9. Choe, J. and Juvkam-Wold, H.C., A Modified Two-Phase Well-Control Model and Its Computer Applications as a
Training and Educational tool; SPE Computer Applications, pp. 14-20, Vol. 9, No. 1, 1997.
10. Choe, J., A Comparison of Theoretical Kill Sheet and Conventional Kill Sheet of Vertical, Directional, and
Horizontal Wells; The Korean Institute of Mineral and Energy Resources Engineers, pp. 492-501, Vol. 32, No. 6,
1995.
11. Beggs, H.D. and Brill, J.P., A Study of Two-Phase Flow in Inclined Pipes; JPT, pp. 607-617, Vol. 25, No. 5, 1973.
12. Dranchuk, P.M. and Abou-Kassem, J.H., Calculation of Z Factors For Natural Gases Using Equations of State; J. of
Canadian Petroleum Technology, pp. 34-36, Vol. 14, No. 3, 1975.
13. Lee, A.L. and Gonzalez, M.H., The Viscosity of Natural Gas: AIME Trans., pp. 997-1000, Vol. 243, 1966.
14. Brill, J.P. and Beggs, H.D., Two-Phase flow in Pipes; 3rd printing, pp. 2-76, U. of Tulsa Press, OK, 1984.
15. Hasan, A.R. and Kabir, C.S., A Study of Multiphase Flow Behavior in Vertical Oil Wells: Part I - The oretical
Treatment; Paper SPE 15138, 56th California Regional Meeting of the SPE held in Oakland, CA, April 2-4, 1986.
16. Lee, J.W., Well Testing; 1st printing, p. 144, SPE Textbook Series, Dallas, Texas, 1982.
Appendix A: Finite Difference Equations of the Governing Equations for Variable Wellbore Geometry
The governing equations (1) to (3) are one-dimensional flow equations. The solution of the governing equations
(1) to (5) involves non-linear partial differential equations. Therefore, an analytic solution is impossible. A fully implicit
finite difference method is utilized to solve the governing equations.
The FDEs in this research are obtained by a central average-in-time and backward-in-space discretization scheme.
The Eulerian approach with grid-centered mesh is also used. Figure A1 shows time and spatial mesh for the FDE. A
variable flow area for each grid can be either flow area enlargement (Fig. A2a) or flow area reduction (Fig. A2b). If flow
area changes a lot, it requires extra work to calculate a grid block volume.
A simple finite difference scheme on the time derivative can be used to solve the governing equations because its
FDEs are simple. However, the scheme turns out causing oscillations of gas holdup and choke pressure calculations,
especially for very high gas fractions. The use of central average-in-time smoothes out the oscillations of gas holdup
and choke pressure.
n n+1
i
space x - unknown
i-1 x - known
time
FIGURE A1: Time and space mesh for finite difference equations.
Flow Area Enlargement

When the flow area enlarges (Fig. A2a), Ai should be used to calculate the volume of the grid block by nodes i and
(i-1). The final FDEs are Eqs. (A1) through (A3).
i i+1
i-1 i
i-2 i-1
(a) (b)
FIGURE A2: Annulus flow geometries: (a) flow area enlargement and (b) flow area reduction.
( A ρ m H m )ni−+11 −
n +1
vm i
n +1
= v m i −1
∆ xi
[(ρ m H m )ni+1 + (ρ m H m )ni−+11 − ( ρ m H m )in − ( ρ m H m )in−1 ] (A1)
( A ρ m H m )i 2∆t (ρ m H m )i
n +1 n +1
n +1
(A ρ v H )
g g
n+1
g i −1 −
Ai ∆xi
2∆t
[( ]
ρg H g )ni−+11 − (ρg H g )ni − (ρg H g )ni−1
Hgi = …………………………….. (A2)
n +1  n +1 ∆xi 
Ai ρg i  v g i + 
 2∆t 
∆xi
pin +1 = pin−+11 −
2∆t
(RVi n +1 + RVi n−1+1 − RVi n − RVi −n1 ) −
1
2 Ai
(RV 2 ni+1 + RV 2in−+11 − RV 2ni − RV 2 ni−1 )
∆ xi  dp dp  980 ∆ xi
n +1 n +1
( )
n n
dp dp
−  + + + − RV 3 ni +1 + RV 3 ni−+11 + RV 3 ni + RV 3 ni−1 (A3)
4  dL i dL i −1 dL i dL i −1  4
Where, RV = ρ m vm H m + ρ g v g H g
RV 2 = Aρm vm2 Hm + Aρg vg2 H g

RV 3 = ρ m H m + ρ g H g
Eqs. (A1) through (A3) are liquid velocity, gas fraction, and pressure equations, respectively.
Flow Area Reduction

When the flow area reduces (Fig. A2b), Ai-1 is employed to determine the volume of the grid block by nodes i and
(i-1). The final FDEs are Eqs. (A4) through (A6).
 n +1 ( ρ H )n +1
n +1
vm i = vm i −1 m m in−+11 −
∆xi
( ρ m H m )i 2∆t ( ρ m H m )i
n+1
[ 
]
( ρ m H m )ni+1 + ( ρ m H m )in−+11 − (ρ m H m )in − ( ρ m H m )ni−1  Ai−1 (A4)
  Ai
(ρ v
g g H g )ni−+11 −
∆xi
2∆t
[ ]
(ρg H g )in−+11 − (ρg H g )ni − (ρg H g )ni−1
H g ni +1 = …………………………………….. (A5)
n +1  n+1 Ai ∆xi 
ρg i  v g i + 
 Ai −1 2∆t 
∆xi
pin +1 = pin−+11 −
2∆t
(RVi n +1 + RVi n−1+1 − RVi n − RVi −n1 ) −
1
2 Ai −1
(RV 2in+1 + RV 2ni−+11 − RV 2 ni − RV 2in−1 )
∆xi  dp n +1 dp n +1 dp n dp n  980 ∆xi
− 
4  dL i
+ + + −
dL i −1 dL i dL i −1  4
(
RV 3 in+1 + RV 3 ni−+11 + RV 3 ni + RV 3 ni−1 ) (A6)
The above FDEs (A1) through (A6) are identical to the FDEs proposed by Nickens (4) if flow areas are the same (i.e.
Ai-1 = Ai ). If the FDE for uniform flow geometry is utilized to calculate velocity at grid face i (vi ) for flow area
enlargement, the velocity at grid face (i-1) should be modified first by the area ratio (Ai-1/Ai ), then substitute it in the
FDE for constant geometry. Eq. (A1) shows the mathematical expression.
If the flow area reduces, vi is evaluated from FDE for constant flow geometry, then the actual velocity at grid face i
can be obtained by multiplying by the area ratio of (Ai-1/Ai). The FDE for pressure can not be obtained from the FDE
for constant geometry and velocity modification by the area ratio because of the velocity square term in the equation.
Therefore, the detailed equations above should be used for accurate simulation.
TABLE 1: Default input data in this study

Mud density, gm/cc 1.68
Plastic viscosity, cp 23.0
Yield point, dynes/cm2 76.6
Bit nozzle opening number 3
Bit nozzle diameter, 1/32nd in. 14.0
Well vertical depth, m 4572
Length of drill-collars, m 137.2
Depth of last casing seat, m 3658
ID of last casing, cm 22.659
Open hole diameter, cm 21.59
OD & ID of drill-pipe, cm 12.7 x 11.201
OD & ID of drill-collar, cm 16.51 x 5.08
Pump rate while drilling, liters/min 1514
Kill mud pump rate, liters/min 757
Water depth, m 914.4
Marine riser diameter, cm 48.26
Choke line ID, cm 10.16
Kill line ID, cm 7.62
For kick analysis:
Formation over pressure, kPa 5378
Pit volume warning level, m3 3.18
Gas specific gravity 0.65
Mud compressibility, 1/kPa 0.87E-6
Surface temperature, oC 21.1
Sea floor temperature, oC 6.1
Bottomhole temperature, oC 72.8
Formation permeability, md 100
Formation skin factor 2
Formation porosity, fraction 0.25
Rate of penetration, m/hr 18.29
TABLE 2: Relative % errors on maximum choke pressures and pi t volume gains

grid max. % error for time for max. % error max. pit % error time for % error
size, choke max. choke choke for the gain, m3 for max. max. pit for the
m pressure, pressure , pressure, min time, % pit gain, min time, %
kPa % gain, %
10 11357 0.0 98.73 0.0 4.01 0.0 99.68 0.0
15 11319 0.33 100.04 - 1.33 4.02 - 0.40 101.17 - 1.49
20 11290 0.59 101.31 - 2.61 3.98 0.79 102.32 - 2.65
25 11346 0.10 102.64 - 3.96 4.01 0.0 103.93 - 4.26
30 11327 0.29 103.89 - 5.23 3.94 1.59 104.39 - 4.73
TABLE 3: Input data used by the Nickens model

Bit nozzle opening number 3
Bit nozzle diameter, 1/32nd in. 12.0
Length of drill-collars, m 137.2
Depth of last casing seat, m 1524
ID of last casing, cm 27.94
OD & ID of drill-pipe, cm 12.7 x 11.201
OD & ID of drill-collar, cm 20.32 x 7.137
Pump rate while drilling, liters/min 1514
Formation over-pressure, kPa 2427
Formation permeability, md 300
Formation porosity, fraction 0.25
Penetration rate, m/hr 9.14
Pit volume gain, m3 4.563
Gas specific gravity 0.728*
Surface temperature, oC 21.1
Sea floor temperature, oC 16.1
Bottomhole temperature, oC 71.1
Water depth, m 304.8
Marine riser diameter, cm 48.26
Choke line ID, cm 7.62
* Calculated from kick weight given.
TABLE 4: Input data used by the Santos model

OD of drill-pipe, cm 12.7
SIDPP, kPa 1379
SICP, kPa 2206
Pit volume gain, m3 2.4645
Initial gas fraction, fraction 0.75
Gas specific gravity 0.7
Surface temperature, oC 21.1**
Bottomhole temperature, oC 51.1**
** Not available and assumed.
9000 Nickens model

The two-phase model
Surface choke pressure, kPa
7500 Single-phase model
kick at surface
6000
SICP
4500
3000
1500 Driller's method

304.8 m water depth all kick has been
4.563 cu. m gas kick removed
0
0 40 80 120 160
Time, min.
FIGURE 1: Comparison of the two-phase model with the Nickens model and the single-phase model.
6000
1,646 m well
1.08 g/cc mud
5000
2.46 cu. m initial pit gain
1,379 kPa SIDPP
4000
3000
2000
Santos results
1000
The two-phase model all kick is removed
Single-phase model for single-phase model
0
0 20 40 60 80 100
Time, min.
FIGURE 2: Comparison of the two-phase model with the Santos model and the single-phase model.
12000
driller's method
4572 m well depth
5378 kPa formation

9000 over-pressure
1.68 g/cc mud
6000
3000
Without numerical dissipation

With numerical dissipation
0
0 30 60 90 120 150
Time, min.
FIGURE 3a: The effect of numerical dissipation on the surface choke pressure.

Pit volume gain, cubic meters
914 m water depth

10.16 cm choke line
4.5
1.5
Without numerical dissipation
With numerical dissipation
0
0 30 60 90 120 150
Time. min.
FIGURE 3b: The effect of numerical dissipation on pit volume gain.

3
k = 1000 md
Pit volume gain, cubic meters
k = 100 md
k = 10 md
2
1.59 cu. m pit warning level

1 18.29 m/hr penetration rate
5,378 kPa formation
over pressure
2.0 skin factor
0
0 10 20 30 40
Time from the beginning of simulation, min
FIGURE 4a: Pit volume gain from drilling to pressure stabilization after well shut-in.
8000
6000
4000
k = 1000 md
2000
k = 100 md
k = 10 md
0
0 10 20 30 40
FIGURE 4b: Surface choke pressure from drilling to pressure stabilization after well shut-in.
with mud compressibility

8000
without mud compressibility
6000
4000
10 md permeability
2000 well shut in 1.75 cu. m pit gain
0.87E-6 1/kPa mud
compressibility
0
0 10 20 30 40
FIGURE 5: The effects of mud compressibility on the surface choke pressure build up.
12000
driller's method
4572 m well depth

914.4 m water depth
9000
10.16 cm ID choke line
6000
In addition to the choke line:

3000
No kill line
7.62 cm ID kill line
12.7 cm ID kill line
0
0 30 60 90 120 150
Time, min.
FIGURE 6: Surface choke pressures for multiple choke lines.

JCPT01 - Advanced Two-Phase Well Control Analysis

Transféré par

Informations du document

Description originale:

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

JCPT01 - Advanced Two-Phase Well Control Analysis

Transféré par

Droits d'auteur :

Formats disponibles

Advanc ed Two-Phase Well Control Analysis 1/17

Jonggeun Choe, Seoul National University

A Revised Paper after the Peer Review

• unsteady state two-phase flow

Two-Phase Mixture Region

The conservation of mass equation for mud:

The conservation of mass equation for gas:

The conservation of linear momentum equation for the mud-gas mixture:

The two-phase correlation to calculate in-situ gas velocity:

The equation of state to compute gas density:

Frictional Pressure Loss

1. Calculate new gas fraction at i-th cell from equation (A2).

Numerical Problems and Their Management

If Hg < 0.25 bubble flow

Negative Liquid Velocity

Time Step and Grid Size

to 25 m grid spacing generally gives good results.

Comparison with Other Models

The Santos Model

Results and Discussion

1. Drilling to the target depth (2 minutes duration)

q total = q choke + q kill …………………………………………………………………………………….. (12)

Flow Area Enlargement

RV 2 = Aρm vm2 Hm + Aρg vg2 H g

Flow Area Reduction

TABLE 1: Default input data in this study

TABLE 2: Relative % errors on maximum choke pressures and pi t volume gains

TABLE 3: Input data used by the Nickens model

TABLE 4: Input data used by the Santos model

9000 Nickens model

7500 Single-phase model

1500 Driller's method

5378 kPa formation

Without numerical dissipation

3.18 cu. m initial pit gain

914 m water depth

FIGURE 3b: The effect of numerical dissipation on pit volume gain.

1.59 cu. m pit warning level

with mud compressibility

4572 m well depth

In addition to the choke line:

FIGURE 6: Surface choke pressures for multiple choke lines.

Vous aimerez peut-être aussi