Beruflich Dokumente
Kultur Dokumente
Most wells produce more than one phase. Oil and gas usually flow together, and often also with
water and solid particles. Figure 9.1 illustrates the typical flow patterns. Simplified sets are 3
main patterns of vertical flow:
- Continuous fluid, with gas bubbles / fast particles
- Continuous gas, with drops of liquid / solid particles and liquid film along the walls
- Discontinuous flow: liquid plugs, broken up by large gas accumulations
In horizontal flow, flow patterns will be similar, but with a tendency to stratification because of
density difference. At the relative low velocity, gas and oil may flow in separate, stratified layers
vs
3 fs
0.5
0.5
(9-1)
s : Density of particle
d : Diameter of the particle
fs : Friction factor between the particle and the surrounding fluid
Friction factor depends on the Reynolds number of the border layer between the particle and
fluid
Re s
vs d
(9-2)
Figure 9.3 shows the correlation between friction factor and Reynolds number, Lydersen /1976 /.
Correlation for spherical particles can also be expressed analytically
fs = 0.44
fs = 18.5 Res-0.6
Transition
fs = 24 Res-1
If sinking velocity of sand particles is greater than the flow velocity in a vertical well, then sand
will fall down and gradually fill up well. In an inclined well the falling velocity will drag
particles against the lower portion of the pipe cross-section. Turbulence will help to keep the
particles in dispersion. Transportation of particles further depends on whether sedimentation or
dispersion dominate.
Ff 1 2
f d vd2 Ad
It seems reasonable to relate the maximum drop size to the ratio of these forces. This provides
the critical Weber number. For a round drop, this can be expressed as
2
vd* d
We
F max
Ff
(9-3)
: Surface tension (N / m)
vd* : sinking velocity of the largest drops
Experiments indicate largest achievable (critical) Weber number in the range of 20-30. By setting
(9-3) in relation with sinking velocity (9-1), we can express maximum sinking velocity as
vd*
Kd
0.25
g l g
Kd
g2
(9-4)
4 We *
3 fd
0.25
Kd is dimensionless and can be considered as constant. For droplets that drop in static gas, this
has been experimentally determined as Kd = 2.75 - 3.1.
vb*
0.25
Kb
g l g
l2
(9-5)
Flow conditions around a gas bubble that rises in viscous liquid will be different from a liquid
drop that falls in small viscous gas, so the friction factor will be different. For gas bubbles the
rise in the stagnant liquid is found by Harmathy / 1960 /: Kb = 1.53.
stabilized by the pipe wall. Rise of large air bubbles in the water is primarily governed by liquid
flow around the bubble front. After passing the bubble front, the liquid will flow along the pipe
wall as a free-falling film, which has no influence on the rise velocity. Rise velocity has been
derived theoretically by Dumitrescu / 1943, (9-6). Dimension Analysis and measurements of
Davis & Taylor / 1950 / gave nearly identical results.
Figure 9.5 Large bubble in the pipe (Dumitescu bubble " / "Taylor bubble")
g
vD 0.347 g D 1
l
(9-6)
diameters. The effect of viscosity attributed to the Reynolds Number defined by fluid density,
liquid viscosity and the rising viscosity of the bubble: Re = vD D/. Viscosity had little
significance for Reynolds Number above 20. Zukoski found that the bubble velosity reach a
maximum at pipe slope around 35o compare to the horizon (inclination 55o)
Rise velocity for large oil accumulations in inclined pipe filled with water has been
experimentally found to be in reasonable compliance with Zukoskis results. In the vertical well,
oil was found to disperse.
(9-7)
u 2d
Break up forces are due turbulent velocity fluctuation: u 2 This can be relatioed to energy
dissipation
u2 d
(9-8)
Energy dissipation may have different causes: the flow through nozzle, mechanical stirring, wall
friction, pumping. For flow in the pipe energy dissipation per mass unit relates to friction loss
v dp 1 1 3
f
v
dx 2 D
(9-9)
By putting (9-8), (9-9) into the relationship between the turbulent forces and surface forces into
the Weber number (9-7), we can associate the maximum droplet size to the fluid in flow
conditions
1
d C 2
5
*
(9-10)
5.
Figure 9.6 below shows the maximum bubble size predictions, using the methods above, for the
parameters :g = 200 kg/m3, l = 900 kg/m3, f =0.02, = 40 .10-3 N/m, D = 0.1 m.
9.2 Fluid
vsl
vm
Qg
A
qo Rt Rs Bg
d 2 / 4
Ql qo Bo qw Bw
A
d 2 / 4
Ql Qg
A
vsg vsl
(9-11)
(9-12)
(9-13)
vg
vl
Qg
Ag
Qg A
Ag A
vsg
yg
Ql Ql A vsl
Al Al A yl
(9-14)
(9-15)
yg : gas fraction
yl : liquid fraction
Average density of fluid mixture in a pipe segment can be linked to fluid densities and fractions
TP
g Ag l Al
g y g l yl
A
(9-16)
Gas and liquid fractions can be measured indirectly, e.g.: by measuring the electrical impedance
for the mixture flowing. Since the bubble and the gas usually have different impedance, such
measurements indicate fractions of gas and liquid.
Since the ability to absorb the gamma rays is also different for liquid and gas, gamma absorption
can also be used for measuring in situ fractions. A more direct method is to close a pipe segment
and measure the volumes of liquid and gas. Liquid fraction is equivalent to the fluid volume,
divided by the volume of the pipe segment.
Qg
Qg Ql
vsg
vm
Ql
v
sl
Qg Ql vm
(9-17)
(9-18)
g Qg l Ql
g g l l
Qg Ql
(9-19)
From the definitions (9-17), (9-18) follows that when fluid velocity approaches the velocity of
gas, the volume fractions, for example, liquid, approach flux fraction
yl
Al
Ql
v
Ag Al
Qg l Ql
vg
(9-20)
vl v g
9.2.4 Slip
Gas has less density and viscosity than the liquid and will usually flow faster. Zuber & Findlay /
1965 / proposed to link gas velocity to total superficial velocity
(9-21)
v g C o vm vd
Co: distribution parameter for bubbles in flow, usually: 1.0 <Co <1.2
vo : Buoyancy velocity of the gas bubbles, or sink velocity for droplets
By combining (9-13) and (9-14), we can express the liquid fraction directly
yl 1
vsg
(9-22)
Co vsg vsl vd
Z & F model has been widely used, but is misleading in many cases. Asheim /1986 / relates
gas velocity directly to the velocity surrounding liquids
(9-23)
v g Co vl vo
By combining the relationship between velocity and superficial velocity to the drop relationship
(9-23), the liquid fraction is expressed as
yl
1
2
vsg
Co sl 1
vo
vo
4Co
vsl
vo
v
1 v
sg Co sl 1
2 vo
vo
(9-24)
Relationship (9-24) is more complicated than (9-22), but avoids most shortcomings of the Z
& F model.
If the gas flows up and the liquid down, equation (9-24) gives one, two, or no solutions. One
solution predicts stable counter current flow. No solutions imply that counter-flow at the
given rates is impossible. Two solutions imply transition between continuous liquid and
continuous gas. One of these solutions will then usually be unstable, such that the flow
regimes will change
Ag dp Ag g g x dx Ag g v g dv g gw S gw dx i S i dx 0
(9-26)
9.3.2Mixed-flow
If gas and liquid flow together, we need a flow equation for mixture. We find it by putting:
Ag/A = yg , Al/A = yl into (9-25) and (9-26), and adding them. This eliminates the
interphasial shear. The mixed flow equation becomes
dp g y g l yl g x dx g v g y g dv g l vl yl dvl
g S gw lw Slw
dx 0
A
(9-27)
The last part in (9-27) contains shear stresses and wetted perimeters. If we presume that perimeter is
proportional with fractions (Sgw=ygS=ygd), and represent the velocity by (9-11), (9-12), we can
develop shear contribution as
1
g S gw l Slw
vsg
vsg
1
1
v
1
v
f g g vsg
d f l l vsl sl d
f g g vsg
f l l vsl sl
2
A
d / 4 8
yg
8
yl
2 d
yg
yl
When gas and liquid flow in the same direction, we can ignore the absolute values. It is useful to
relate shear to rate and flux density (9-17), (9-18). If we also assume equal friction factors for
liquid and gas : fg = fl =fo, we get
g S gw l Slw 1 o 1
2g
2
f
g
l l vm2
A
2
d
yg
yl
(9-28)
Often the shear contribution (9-28) expressed as for flow of homogeneous mixture
g S gw l S lw 1
1
fTP m vm2
A
2
d
(9-29)
Here fTP is thetwo phase friction factor, estimated from the correlation for single-phase flow,
with a correction factor for the drop:
(9-30)
fTP f o cTP
The comparable single phase friction factor fo is estimated by standard single-phase correlation
(for example : fo=0.16/Rem0.172) with Reynolds Number for the homogeneous mixture, usually
defined as
Re m
m vm d
g g l l
(9-31)
From (9-28), (9-29) follows the slip correction factor then becomes
g 2g l l2 g yl 1 l l y g l
cTP
m y g m yl
m yl 1 yl
2
(9-32)
With this theoreteical basis, we can express and calculate the pressure gradient
dv
2
dp
dv 1
(9-33)
The theory above involved many assumptions and approaches. Published models for steadystate two phase flow may deviations somewhat from the basis outlined above.
dp
dx
(9-34)
Lw
where:
Lw
dp
dx
:
p
Well pressure: p
pw pth
T Tth
, And temperature: T w
2
2
Since we do not know the average pressure in the well before we have estimate the pressure,
iterations are required: we may estimate the pressure gradient at the bottom, solve (9-34), and
then use the solutions to estimate the average pressure. In many cases it provides pretty good
estimate of pressure. Such step will be called a single-step Runge-Kutta solution.
If the pressure gradient change much, it would be desirable to assume the pressure and flow
conditions at intervals along the pipe. The algorithm can then formally be expressed as
pi 1 pi
dp
dx
Ti 0.5
Li 1 Li
(9-35)
pi 0.5
Figure 9.9 shows the estimated pressure profiles. The parameters for calculation are: pR = 310
bar, J = 20 Sm3/d/bar, Rt = 200, d = 0.1 m, o = 0.9, g = 0.58, Co = 1.1, vo
=0.1 m for parameters: pR = 310 bar, J = 20 Sm3/d/bar, qo = 200 Sm3/d, Rt =
200, d = 0.1 m, o = 0.9, g = 0.58, Co = 1.1, vo =0.1 m . At low pressure,
pressure gradient will decline significant. This is because the gas is released and is expanding, so
that the average density of the fluid mixture is declining. We observe that the pressure at
production rate 200 Sm3/d is larger than at smaller rates .
Figure 9.9 Pressure, estimated with the flow model outlined above
Figure 9.10 below shows how the liquid fraction changes along the pipe, from 0.9 at the bottom,
to 0.1 near the top ,with the production 200 Sm3/ d.
Figure 9.10 The estimated liquid fraction along the production pipe
Figure 9.11 shows the superficial velocity along the pipe.
Superficial velocity for liquid decreases with increasing pipe diameter, because disolved gas
evaporates. Actual velocity of gas and liquid can be estimated by (9-14) and (9-15). The actual
flow velocity of gas will be slightly over the sum of superficial velocity; for liquid, a little bit
smaller
Figure 9.13 Wellhead pressure characteristics of the well which produces oil
and gas
The figure above shows that for the 10 bar pressure, rate can be both 50 Sm3/d and 400 Sm3/d.
Probably the well head rate will be stable only at the separator pressure of 10 bar. It will be
difficult to start well, with separator pressure 10 bar
pipe, to reduce the average density of the mixture, doing that we usually increase GOR. Figure
9.15 illustrates the tubing head pressure characteristics as a function of gas / oil ratio.
bubbles will rise. In a flow with 95% gas and 5% liquid, droplets will sink relative to the gas
velocity. To estimate rising or sinking velocity, we need to know wheter gas or liquid is
continuous.
Figures 9.16 and 9.17 below illustrates flow regimes. A main classification principle is wheter
liquid or gas flow is the larger. The relationship: vsg = vsl has been drawn into both figure 9.17 and
9-17,, and we see that both regime maps relate to this.
Figure 9.16 Regime map for vertical gas/liquid flow, Duns & Ros / 1963/
Regime map shown in Figure 9.16 is based on measurements. Expressing the map in terms of
dimensionless variables, surface tension and liquid density variations are in principle included
Figure 9.17 Regime Map for vertical gas/liquid flow at atmospheric pressure
Regime map shown in Figure 9.17 is based on simple criteria related to the liquid fraction,
densities and velocity. The criteria are linked together by Kabir & Hasan / 1990 /.
Regimes based models should provide better connection to flow mechanisms. However it is not
always so that regime based models offer better prediction. Theses at this institute have shown that,
for example, Kabir & Hasan model applied to the Ekofisk data, prediction of pressure drop less
accurately than a homogenous flow model neglecting regime transitions.
9.5 References
1943 Dumitrescu, DT: "Strmung an einer Luftblase in senkrecthen Rohr"
Z. angew. Math. Mech., 1943, vol. 23, no. 3, pp 139-149
1950 Davis, R.M., Taylor, G.I.:
"The Mechanics of Large Bubbles Rising Trough
Extended Liquids and Through Liquids in tubes "
Proc. Royal Soc., London, vol. 200 series A, 1950, pp 375-390.
1955 Hinze, J.O.:
"Fundamentals of the Hydro Dynamic Mechanisms of
Splitting in Dispersion Processes ",
AICHE J. (Vol 1, No. 3), 1995, pp 289-295
1960 Harmathy, T.Z.:
"Velocity of Large Drops and Bubbles in Media
Of Infinite or Restricted Extend "
AICHE, no. 6, p. 281, 1960.
1962 Nicklin, DJ, Wilkes, JO, Davidson, JF: "Two Phase Flow in Vertical tubes"
Trans. Instn. Chem. Engrs. Vol 40, 1962, pp 61-68.
1963 Duns, H., Jr. and Ros, N. C. J.:
Vertical Flow of Gas and Liquid Mixtures in Wells,
Proc. Sixth World Pet. Congress, Frankfurt (Jun. 19-26, 1963) Section II, Paper 22-PD6.
1965 Zuber, N. & Finlay, J.:
"Average volumetric concentration in two-phase flow systems"
Trans ASME, J. Heat Transfer. C87, 453-468