You are on page 1of 10

SPE 104605

On the Flow Performance of Velocity Strings To Unload Wet Gas Wells


P. Oudeman, SPE, Shell Intl. E&P

Copyright 2007, Society of Petroleum Engineers


This paper was prepared for presentation at the 15th SPE Middle East Oil & Gas Show and
Conference held in Bahrain International Exhibition Centre, Kingdom of Bahrain, 1114 March
2007.
This paper was selected for presentation by an SPE Program Committee following review of
information contained in an abstract submitted by the author(s). Contents of the paper, as
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to
correction by the author(s). The material, as presented, does not necessarily reflect any
position of the Society of Petroleum Engineers, its officers, or members. Papers presented at
SPE meetings are subject to publication review by Editorial Committees of the Society of
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper
for commercial purposes without the written consent of the Society of Petroleum Engineers is
prohibited. Permission to reproduce in print is restricted to an abstract of not more than
300 words; illustrations may not be copied. The abstract must contain conspicuous
acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O.
Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.

Abstract
In depleted gas wells, the produced gas rate and consequently
the velocity will drop to the extent that produced liquids are no
longer carried to surface. The liquids accumulate in the well
bore, increasing the sand face pressure. This further reduces
the inflow, so that more liquid collects and eventually the flow
dies down completely. This phenomenon is known as liquid
loading.
Velocity strings are a commonly applied remedy to liquid
loading in gas wells. By installing a small diameter string
inside the tubing, the flow area is reduced which increases the
velocity and restores liquid transport to surface. The
disadvantage of the velocity string is the increase in frictional
pressure drop, constraining production. Hence an optimal
velocity string has to be selected such that liquid loading is
delayed over a long period with a minimal impact on
production. This requires accurate methods to predict pressure
drop in the velocity string as well as tubing-velocity string
annulus.
The available methods to predict pressure drop in annuli for
gas-liquid flow are modifications of methods to predict wet
gas pressure drop in tubing. These modifications are usually
based on assumptions, which are strictly valid only for singlephase flow. Their validity for gas-liquid flow is questionable.
Hence to assess the validity of the methods a field test was
designed and executed. The results were compared with
various approaches to describe wet gas flow in an annulus.
This allowed selection of the best approach with an accuracy
comparable to the accuracy of methods to predict pressure
drop in tubing. Factors affecting the accuracy were identified.
Comparison with a field case provided further proof for the
validity of the approach.
This result is not only relevant for velocity string design, it is
important for all annular flow processes in wells such as flow
around a stinger, drill pipe, tool or coiled tubing string.

Introduction
When there is sufficient reservoir energy and gas wells can be
produced at medium to high rates, co-production of liquids is
seldom a problem, even at high liquid to gas ratios (LGR).
Although the liquid slips through the gas, effectively the gasliquid mixture tends to behave like a single phase liquid
flowing to surface, where the phases can be separated and
processed.
This changes when the reservoir depletes, the reservoir
pressure drops and the produced gas rates decline. The
velocity at which the gas moves upward approaches the
terminal velocity at which liquid droplets would fall
downward in a stagnant gas1,2. This means more liquid will be
retained in the casing or tubing. The consequence of liquid
accumulation in the well is an increase in the hydrostatical
pressure drop over the well. Since the well head pressure is
usually kept constant by the surface facilities, the increase in
pressure drop over the well leads to an increase of the pressure
at sand face. In turn an increased pressure at sand face gives
rise to a reduction of the inflow of gas and liquids, reducing
the gas velocity even further so that more liquid is
accumulated. The well is said to load up with liquid and flow
ceases altogether (or in the best case some gas continues to
bubble upward through a liquid column).
Several approaches have been suggested and tried to prevent
or delay the loading process, such as3,4:
5
The installation of siphon and velocity strings ,
6
Injection of surfactants to create foams ,
7
Well head compression ,
8
Plunger lift ,
Gas lift,
Work-over to a smaller tubing size, etc.
Of these approaches, installation of a velocity string, i.e. a
small diameter tubing or coiled tubing inside the actual tubing
to increase velocity and improve liquid transport, is one of the
most attractive options since it is low cost, can be carried out
under pressure (i.e. there is no need to kill the well) and
requires no further maintenance after installation.
Apart from mechanical considerations, such as interference
with the SSSV, the main drawback of the velocity string is
that the introduction of the string increases the frictional flow
resistance in the well. This inevitably leads to a reduction of
the productivity of the well. Hence the price for the
suppression of liquid loading is decreased production. This
makes selection of the optimum size of the velocity string
critical. It has to be selected such that liquid loading is avoided
or at least delayed over a considerable period of time, whilst
maintaining the highest possible production.

SPE 104605

The selection of the appropriate velocity string size is usually


based on nodal analysis, using two types of curves:
1. The inflow performance relation, IPR, which
describes the relation between the sand face pressure
and the amount of gas entering the wellbore and
2. The intake pressure curve, IPC, (also known as
vertical lift performance curve or J-curve) which
describes the relation between the amount of gas
produced to surface and the down hole pressure
required to produce this gas for a given fixed well
head pressure. The minimum of this curve gives the
lowest rate at which the well can be produced.
The intersection of these two curves gives the rate actually
produced. Schematically this looks as follows (Figure 1):
IPC tubing
+velocity
string

IPC tubing
Pressure
IPR,
now

IPR,
future
Gas Rate

Figure 1: Effect of a velocity string on production

The figure illustrates the dilemma for velocity string design.


Introduction of the velocity string moves the intersection with
the current IPR curve to the left, i.e. the produced rate is
reduced. However when, due to depletion the IPR curve
changes, the tubing IPC curve would no longer intersect, i.e.
the well cannot produce, whereas with the velocity string the
well still produces. The choice is between a higher production
rate over a shorter period of time and a lower production rate
over a considerably longer period (and higher ultimate
recovery). This period can be even be lengthened by flowing
up the velocity string initially, until flow becomes unstable in
this string as well and switching to flowing up the annulus
between the velocity string and the original tubing in the final
stages of production.
This makes the accuracy with which the IPC curves can be
calculated of utmost importance. Assume e.g. that the multiphase pressure drop algorithms over-estimate the pressure
drop for the velocity string configuration; this would lead to
the selection of a string which would actually be too large,
resulting in a production rate more than anticipated after
installation of the string but a short lived effect of installation
of the string. Inversely selection of a string, which is too
small, would lead to an excessive decrease in production rate.
Reliable methods have been developed to calculate vertical
lift, IPC, performance curves for tubing8. These methods were
based on large sets of accurate field data collected during
dedicated field tests. Experience shows that in the vast
majority of cases an accuracy of +/- 10% in the pressure drop
over the tubing could be achieved. Given the accuracy of field

data, usually of the same order of magnitude, and unknown


factors such as the hydraulic roughness of the tubular, this was
considered satisfactory.
For the tubing-velocity string annular flow configuration such
data were not available, and no methods were developed
specifically for this configuration. To model the pressure drop
over the annular space between the tubing and the velocity
string, the methods developed for tubing strings were used
with an effective diameter to correct for the difference
between tubular and annular flow. Although such an approach
can be proven mathematically correct for simple cases of
single-phase incompressible flow, in multi-phase gas-liquid
flow serious doubts must be raised:
Flow regimes such as annular-mist and churn flow
and the transitions between them could be seriously
affected by the presence of a velocity string.
Likewise velocity distribution and interaction
between gas and liquid in an annular configuration
are totally different from those in an open tube
configuration.
This will be reflected in quantities such as liquid
hold-up and interfacial friction, which determine the
pressure drop in the tubing velocity string annulus.
Hence the validity of the effective diameter concepts cannot
be taken for granted. They can only be considered a sound
basis for design of velocity strings when they can be supported
with accurate field data. To collect such data, a dedicated field
test was designed and executed. This will be discussed in the
next section.
To interpret the results of this field test, the dependency on
flow geometry of one of the most reliable methods to model
wet gas well flow, the Gray correlations, was examined. Next
modifications of this method to take into account the annular
geometry are tested against the field test data. The best
approach was selected with this procedure. Comparison with a
field case provided further evidence for the validity of this
approach.
Dedicated field test set set-up and execution
Originally the test was set up to simulate and study the
dynamics of the kill process during hydraulic blowout control
9
. For this purpose a prolific gas well in the Slochteren field
was selected in which coiled tubing was run to total depth to
serve as a relief conduit. Kill fluid was pumped down the
coiled tubing at various rates and the response of the flowing
well was monitored. Failed kill attempts led to steady state
flow of kill fluid and gas up the tubing-coiled tubing annulus.
Hence every failed kill attempt was a data point for the
validation of methods to predict velocity string performance.
This broadened the scope of the test program.
Configuration
The essentials of the downhole configuration of the test are
shown in Figure 2. In the practically vertical well, a 1.75
coiled tubing carrying a PLT was run to a depth of 2800 m
(9190 ft) through the 5.5 (17 lbs/ft) tubing. Through an
injection valve just above the PLT a 3% KCl brine could be
injected at rates up to ca. 0.3 m3/min (2 bpm).

SPE 104605

gauge and a quartz gauge, so that their readings


could be cross-checked;
Tubing head pressure (FTHP) upstream of the
additional separators;
Flow line pressure (FLP) downstream of the
additional separators;
Gas flow rate, recorded at the local control room;
Liquid production rate, based on volumetric
readings of the separator level every 15 min.;
Flow line temperature (FLT) using a sensor
wrapped against the flow line directly at
wellhead;
Casing head pressure (CHP);
Injection pressure and injection rate measured at
the pump.
All measurements were stored in a data acquisition system.
Test programme and execution
A total of five trials were conducted at different initial
production rates and different injection rates, listed in Table 1.
Before each trial the well was produced clean of injected
brine, so that the point of steady state production, i.e. injected
fluid equals produced fluid, could be determined clearly. In
trial no.3 the production choke was reduced in a stepwise
manner while maintaining a constant injection rate, this
generated various test points.
Table 1: Test overview
Trial no.
1
2
3
4
5
Figure 2: Downhole configuration for the test

This brine mixed with the gas produced by a perforated


interval from 2810-2930 m (9220 9613 ft) and the multiphase mixture entered the tubing shoe at 2790 m (9154 ft).
Well effluents
The gas produced by the Slochteren reservoir is extremely
lean and contains only 1 m3/MMsm3 (0.18 bbl/MMscf)
condensate of 800 kg/m3 (45 API). The gas consists mainly of
methane, 81.3%, contains a small amount of CO2, 0.9%, and a
considerable amount of N2, 14.3%. The molecular weigth is
18.6 kg/kmol (s.g. 0.644, air =1).
Although there is a huge aquifer, no free water is
produced. At the reservoir temperature, 102.3 0C (216 0F) and
pressure, 163.7 bara (2374 psia), the saturated gas contains 6
m3/MMsm3 (1.25 bbl/MMscf) water. Hence the LGR of the
gas water mixture flowing up the annulus is determined almost
completely by the injection rate through the coiled tubing.
Measurements
The following signals were recorded during the test:
Flowing bottom hole pressure (BHP) and
temperature (BHT), using a PLT connected to the
coiled tubing nose, below the brine outlet (see
figure 2). Two pressure gauges were run, a strain

Initial gas rate


(MMm3/d)
1.25
1.20
1.26
1.30
0.53

Average liquid injection rate


(m3/min)
0.197
0.243
0.272
0.275
0.292

When injection started, the FTHP would decrease until the


well stabilised and the produced fluid at surface equalled the
down hole injection rate. This concluded each trial. The FTHP
dropped at least 10 bar in each of the tests, i.e. a substantial
increase of the liquid hold-up in the tubing-CT annulus was
achieved during each test.
Results and analysis
Once the liquid rate recorded at surface equalled the injection
rate, no further changes were made until the well had
completely stabilised in terms of gas and liquid rates,
pressures and temperatures. The stabilised rates, pressures and
temperatures are given in Table 2.
Table 2: Test results
Test no
0
1
2
3
4
5

Rate
LGR
6
(MMsm3/d) (m3/10 sm3)
1.25
7
1.08
270
0.97
370
0.89
480
0.8
550
0.585
740

FBHP
(bara)
162.3
162.4
162.5
162.6
162.7
162.9

FWHP
(bara)
96.8
87.5
85
82.6
88.2
94.9

FWHT
0
( C)
60
45
43
42
42
41

SPE 104605

Here test 0 denotes the conditions prior to injection.


These test results will now be used to validate various
approaches to describe multi-phase flow in the annular space
between the tubing and a coiled tubing velocity string.
Essentially these approaches consist of modifications to
methods to describe wet gas well flow in tubing such as the
Gray correlation.
The Gray correlation
One of the most commonly used methods for gascondensate well pressure profile prediction is the Gray
correlation10. According to Gray, the effect of holdup on the
friction gradient can be adequately expressed as a change in
pipe wall roughness rather than a change in the friction factor.
The Gray correlation utilizes the following dimensionless
parameters:

g ( l g ) D 2
m2V 4 sm
V + Vsw
Nv =
; ND =
; R = so
g ( l g )

Vsg
.(1)
Where (in any consistent unit system):

g
Vsm

l,g

=
=
=
=

Mixture density
Gravitational acceleration
Superficial mixture velocity
Mixture surface tension

=
Liquid and gas densities
D
=
Pipe diameter
Vso,w,g =
Oil, water, gas superficial velocity
The velocity number Nv is the square of what is known as
the Kutateladze number that is used in annular-mist flow
regime modelling or to describe the hanging film
phenomenon. Nd is the diameter or Bond number which
indicates the ratio (gravitational force) / (surface tension
force) and is used in momentum transfer in general and
motion of bubbles and droplets calculations in particular. The
parameter R is a dimensionless liquid to gas ratio
The liquid volume fraction (hold-up) is estimated using:

Hl = 1
Where:

(1 Y )
(R + 1)

.(2)

Y = Exp 2.314 A B ....(3)


205

A = N v 1 +
N D

..(4)

730 R
B = 0.0841 1 0.0554 ln1 +

R + 1 .(5)

The mixture surface tension is defined as:

m =

qo o + 0.617 qw w
qo + 0.617qw

(6)
where qo and qw are the in-situ volumetric rates of
condensate and water.

Based on data from Katz11, Gray developed the following


correlations for the water and condensate surface tensions:

w = (2.115 0.119 p ) (0.174 2.09 10 4 (T 460 ))


............................(7)

c = 0.044 1.3 10 4 (T 460)

pD p
( p D 2120)2.5

.....................(8)
Where:
=
Dew point pressure
pD
T
=
Temperature (R)
The hydrostatic gradient is now calculated with:

dp
= g ((1 H l ) g + H l l )
dz H
...(10)
The Nikuradze equation for fully rough wall flow friction
is used to calculate the friction factor with pseudo wall
roughness, r, evaluated as follows:

r = rl = 28.5

m Vm2

r = rg + (rl rg )

for R > 0.007 .(11)

R
0.007 for R < 0.007 (12)

Subject to the limit r > 2.77 * 10-5 ft where rg is the dry


hydraulic tubing roughness. This formulation for the pseudo
wall roughness reflects the observation that the frictional
pressure drops observed in smooth laboratory piping exceeded
the friction in corroded field piping. The assumption was that
the liquid film at the wall reduced the friction and that this
effect increased at higher gas velocities. The frictional
pressure gradient is calculated with the set of equations (1315):

C3
dp
=
dz f m C 4
C3 = 1.325

..(13)

g Q g + c Q g + w Qw
64.34 DH A 2

D
C 4 = 1.309 + ln H
r

...(14)


.(15)

It has been noted that for very low velocities the pseudo
roughness could exceed the diameter in equation (15).
Practically this is of little relevance since at such low
velocities the wells will be loaded with liquids, the flow
regime will be intermittent or even bubble flow and the Gray
correlation can not be considered valid any more. Dedicated
field testing indicated this to happen at velocity numbers less
than 9.6.
The Gray correlation has been known to give good results
in both water and condensate wells for condensate ratios up to

SPE 104605

around 50 bbl/MMscf and high produced water ratios (200


bbl/MMscf). The initial development of this correlation used
an empirical fit to match 173 data points. The accuracy of the
data fit was manifested as an average tubing pressure drop
error of -0.3% and a standard deviation of 5.3%.
Various approaches are feasible to adapt the correlation to
annular flow.
Equivalent area approach
The simplest method of modelling annular velocity string
flow with the Gray correlation is assuming flow through the
annulus is similar to flow through a tubular which has an
internal diameter such that the flow area of the tubular equals
the flow area of the tubing velocity string area. For well
performance simulators this is a convenient assumption; no
code changes are required and parameters such as velocities
are automatically calculated correctly.
To judge the quality of the assumption the error in
calculated pressure drop, p equals FBHP-FWHP, over the
conduit will be used:

Error = 100

pcalculated pmeasured
pmeasured

(16)

Note that this means that if the assumption under-predicts


the pressure drop, i.e. is too optimistic, the error is negative.
The result using the equivalent area assumption to predict the
pressure drop over the tubing-velocity string annulus for the
tests is shown in Table 3 (a hydraulic roughness of 0.000938
has been used as originally recommended by Gray).
Table 3: Test on equivalent area diameter approach
Test no.

FBHP (bar)

0
1
2
3
4
5

162.3
162.4
162.5
162.6
162.7
162.9

FWHP meas.
(bar)
96.8
87.5
85
82.6
88.2
94.9

FWHP calc.
(bar)
111.6
96.2
95.8
94.4
95.4
96.6

Error %
-22.6
-11.6
-13.9
-14.8
-9.7
-2.5

The table shows that the pressure drop is systematically


under-predicted. Over the data set the average error equals
12.5% with a standard deviation of 6.6%, i.e. a far cry from
the accuracy of the original Gray correlation.
It is noted that the under-prediction is largest for the tests
with the highest gas rates (low test numbers) and tends to
improve for the lowest rates. This indicates that friction, which
depends on the flow rate squared, is not taken into account
correctly. This is introduced by the equivalent area
assumption: although the flow area is correct, the wetted
perimeter, i.e. the amount of surface area over which the
tubular and the flowing fluids are in contact and friction
develops, is severely underestimated.
To maintain the equivalent area approach, an attempt could
be made at increasing hydraulic roughness to compensate for
the under-estimation of the wetted perimeter. By selecting a
hydraulic roughness, which minimises the error for the highest
rate in test no. 0, the quality of the prediction for the lower rate
tests should also improve. By raising the hydraulic roughness
tenfold from the recommended value, i.e. to 0.01 the error in

the predicted pressure drop for test no. 0 is reduced to 1.8%.


The effect on the prediction for the other tests is shown in
Table 4.
Table 4: Equivalent area diameter with increased hydraulic roughness
Test
no.
0
1
2
3
4
5

FBHP
(bar)
162.3
162.4
162.5
162.6
162.7
162.9

FWHP
meas. (bar)
96.8
87.5
85
82.6
88.2
94.9

FWHP
calc. (bar)
95.6
95.8
95.8
94.2
94.8
96.6

Error
%
1.8
-11.1
-13.9
-14.5
-8.9
-2.5

Although the prediction for the highest rate test has


improved considerably, the overall average error still equals
8.2% with a standard deviation of 6.5%. For the intermediate
rates there is hardly any improvement at all. The reason is that
in the Gray correlation the hydraulic roughness for high
LGRs is determined mainly by the liquid film at the tubing
wall, according to equations (11) and (12) of the previous
section. Hence raising the hydraulic roughness has practically
no influence for high LGRs. This means that the
shortcomings of the equivalent area diameter approach cannot
be corrected with an increased hydraulic roughness. An
alternative approach is required.
Hydraulic diameter approach
To model the pressure drop in water or air flowing through
channels of all shapes and sizes, the frictional pressure drop is
calculated using a hydraulic diameter which is the ratio of the
area occupied by the flowing fluids, A, and the wetted
perimeter, P:

DH =

A
P

.(17)

For a concentric annular configuration, between the outer


diameter of the inner tube, D1, and the inner diameter of the
outer tube, D2, it can easily be shown that:

DH = D2 D1 .(18)
The first place to adapt the Gray correlation to annular
flow with this hydraulic diameter is in the friction calculation,
equation (13). Frictional pressure loss is inversely proportional
to the hydraulic diameter and since the hydraulic diameter will
be smaller than the equivalent area diameter, discussed in the
previous section, the friction gradient will increase. Of course
the area, A, in equation (13) should be the real flow area.
It is less obvious how the hydraulic diameter should
influence the hold-up calculation. The Bond number, ND, also
depends on diameter. In typical situations where the Bond
number plays a role, such as considerations on the stability of
droplets or bubbles, the diameter in the bond number is the
diameter of the droplet or bubble. High Bond numbers
indicate a low sensitivity to surface tension effects, low bond
numbers indicate a high sensitivity to surface tension. In the
annular geometry the maximum size of a bubble would be
difference between the diameters, i.e. the hydraulic diameter.

SPE 104605

Hence it seems appropriate to calculate ND with the hydraulic


diameter as well.
The result of using the hydraulic diameter in the friction
and hold-up calculations is shown in table 5.

the correction factor, and consequently the frictional pressure


drop, decreases with increasing eccentricity. Eccentricity gives
larger open spaces for flow so that there is less shear between
the flowing liquids and the pipes.
1

Table 5: Hydraulic diameter approach

0
1
2
3
4
5

FBHP
(bar)
162.3
162.4
162.5
162.6
162.7
162.9

FWHP
meas. (bar)
96.8
87.5
85
82.6
88.2
94.9

FWHP
calc. (bar)
101.4
82.2
83.2
82.4
84.2
90

0.9

Error
%
-7.0
7.1
2.3
0.2
5.4
7.2

Correction

Test
no.

0.8
0.7
0.6
0.5
0

Using the hydraulic radius in the Gray correlation leads to


an average error of 2.5% with a standard deviation of 5.4%.
This means that 99% of the predicted pressure drops will
deviate between -14% and +19% from the real value.
Wellhead pressures will be predicted with a much higher
accuracy since these are mainly determined by the FBHP in
gas wells.
In contrast with the equivalent area diameter approach, the
hydraulic diameter approach has a small positive error, i.e. in
general pressure drops are over-predicted. The only exception
is test point no. 0. Increasing the hydraulic roughness to
0.0018, which is still a very acceptable value for tubular
steel, can actually eliminate the error in this prediction. As
explained in the previous section, this will hardly affect the
prediction for the high LGR test points. The small but
systematic over-prediction of the pressure drop can probably
be attributed to the failure of the assumption that the tubing
and the coiled tubing are concentric.
Single-phase flow in eccentric annuli has been studied
extensively for drilling purposes12. In most studies the
frictional pressure drop in the eccentric annulus is related to
the flow in the concentric annulus by means of a correction
factor, Cecc:

dp
dp
= Cecc
dz ecc dz conc
(19)
Correlations have been sought for the correction factor
which depends on fluid rheology, eccentricity and pipe
diameters. A popular correlation for Newtonian fluids is:
0.1852

D
e 1.5 1
D2

0.4

0.6

0.8

Figure 3: Correction on the frictional pressure drop for


eccentricity

It appears that for the maximum eccentricity the frictional


pressure drop is some 70% of the pressure drop in the
concentric annulus. When the pipe axes are misaligned by 1,
which corresponds to an eccentricity of 0.3, the frictional
pressure drop is some 10% less than in the concentric case, i.e.
a slight misalignment can give rise to a substantial reduction
in pressure drop. In the field misalignment must be considered
standard, the coiled tubing may even be buckled inside the
tubing and over-prediction is to be expected.
Summary
Three approaches to adapt the Gray correlations for wet
gas well flow in tubings to flow in the annulus between a
velocity string and the tubing have been tested against a data
set obtained during a dedicated field test:
1.
Assuming an effective diameter such that the flow
area equals the flow area of the velocity stringtubing annulus area.
2.
As 1 but with an increased hydraulic roughness to
compensate for the wall area not taken into
account with that approach.
3.
Assuming a hydraulic radius equal to the
difference between the inner radius of the tubing
and the outer radius of the coiled tubing.
For each approach actual pressure drop over the annulus
and calculated pressure drop were compared for all test points
and the errors made with each approach were calculated.
The overall result of this comparison is depicted in Figure
4.

0.2527

D
e + 0.96 1
e3
D2
(20)
2

The eccentricity, e, is defined as the distance between the


centres of the pipes divided by the hydraulic radius, i.e. e
equals zero for concentric pipe and e equals unity when the
pipes touch. The correlation is valid for eccentricities from 0
to 0.95, pipe diameter ratios of 0.3 to 0.9.
For the test set-up the diameter ratio is 0.36. In Figure 3
the value of the correction factor is plotted as a function of
eccentricity for this diameter ratio. The first observation is that

Eccentricity

Error %

0.8454

D
Cecc =1 0.0072 1
D2

0.2

25.0
20.0
15.0
10.0
5.0
0.0
-5.0
-10.0
-15.0
-20.0
-25.0

Hydraulic radius
Eq. Area radius
Eq. Area + roughnes

0.5

0.7

0.9

1.1

Gas rate MMsm3/d

Figure 4: Comparison of approaches

1.3

SPE 104605

Comparison with a field case


The case discussed here is an offshore gas well, drilled in
the Dutch sector of the North Sea. The well was drilled to
develop a gas reserve with properties very similar to the
Slochteren field discussed above (gas s.g. 0.59, air=1,
condensate gravity 59 API, CGR 1 m3/MMsm3, WGR 12
m3/MMsm3). After drilling to a depth of 3340 m, 9 5/8 casing
was set. To maintain under-balance with gas lift a 7 string
was run inside the 9 5/8 casing. An 8.5 horizontal hole was
drilled under-balanced to a total depth of 4774 m. The well
was strongly deviated with a kick-off point at 1500 m,
building to 90 degrees at 3270 m, maintaining the angle to
total depth.
To test the well, the drill string was pulled and a pressure
gauge was run to a depth of 3400 m on 3.5 drill pipe with a
number of 4 drill collars and 5 connectors. The flow
configuration is shown in Figure 5.

13.375 in Csg
72.00 lbm/ft
1150 m

9.625 in Csg
53.50 lbm/ft
3340 m

8.500 in
4774 m

3.500 in DP
10.20 lbm/ft
3400 m

7.000 in Csg
32.00 lbm/ft
3401 m

Figure 5: Well schematic

After clean-up the well was produced at a number of choke


settings. Bottom hole and well head pressures and
temperatures, gas and liquid rates were measured
continuously.

At a number of more or less stabilised gas production


rates, pressures and temperatures were noted. Flowing bottom
hole pressures as a function of gas rate are shown in Figure 7.
A Forcheimer type inflow performance:
2
p R2 p wf
= A q g + F q g2

Was fitted to the data points with a least squares method.


This had the following result:
pR = 291 bar (4220 psia)
A = 26.3 bar2/1000 m3/d (156000 psi2/MMscf/d)
F = 0 bar2/(1000 m3/d)2 (0 psi2/(MMscf/d)2)
The quality of the fit is shown in Figure 6.
Inflow performance
350
300
Pressure (bar)

Here the calculated errors are plotted as a function of the


gas flow rate. It appears from the figure that the equivalent
area approaches systematically under-estimate the pressure
drop over the annulus. This is an undesirable feature since it
would lead to the selection of velocity strings, which are too
large, which would restrict production unnecessary.
The hydraulic radius approach performs considerably
better, the average error is only 2.5%. This small but
systematic over-prediction can be attributed to the eccentric
position of the coiled tubing in the tubing. Misalignment of the
pipes gives rise to smaller pressure drops. To what extent this
affects application of the hydraulic diameter concept will be
demonstrated with a field case.

250
200

Measured

150

A & F fit

100
50
0
0

0.5

1.5

Gas rate MMsm3/d

Figure 6: Flowing bottomhole pressure versus gas rate

Note that there is considerable scatter in the data points.


Apart from random measurement errors this can be attributed
to:
Poor stabilisation, the test was intended to be an
indication of the well potential rather than a fully
stabilised multi rate test.
Variations in surface back pressure.
This makes the data less suitable for an exact comparison
with model calculations, i.e. along the lines the dedicated field
test data were analysed in the foregoing, since the scatter will
tend to indicate a large standard deviation of the error. An
overall comparison with calculations to detect trends is
feasible.
In Figure 7 calculated flowing tubing head pressures
(using a modified Gray correlation and the hydraulic diameter
approach) are compared with measured flowing tubing head
pressures. With the value for hydraulic roughness of the
tubulars recommended by Gray, ca. 0.001, a fair match with
the data is obtained for gas flow rates less than ca. 1.0
MMsm3/d. This confirms the observations made on the
dedicated field test discussed in the foregoing sections. Underprediction due to misalignment of the 7 casing and the 3
test string is not observed here.

SPE 104605

FTHP vs. gas rate


300

FTHP (bar)

250
200

r = 0.001"

150

r = 0.01"

100

Measured

50
0
0.0

0.5

1.0

1.5

2.0

Gas rate (MMsm3/d)

flowing well head pressure, according to the nodal analysis


presented in Figure 8.
The gas rate drops to 0.55 MMsm3/d when the reservoir
pressure has dropped to 200 bars. At lower rates the intake
pressure curve for the 7-inch tubing tends to increase slightly
with decreasing rate. This signals the onset of instability, the
well may become sensitive to slight changes in e.g. surface
pressure, i.e. a small decrease in rate gives rise to an increase
in pressure which in turn decreases rate etc. The well can
probably kept in production under these circumstances.
When the reservoir pressure declines to 176 bar (2465
psia), the production rate drops to 0.3 MMsm3/d, the well can
no longer be produced.

Figure 7: Calculated and measured FTHPs

Hence a hydraulic roughness exceeding the roughness of


continuous clean steel pipe is to be expected in this case.
Unfortunately hydraulic roughness of piping can rarely be
predicted with a high accuracy. Even clean piping may exhibit
high roughness caused by the manufacturing process. As
discussed, for high LGR gas flow this will become less
important since it is the liquid film at the wall effectively
determining the hydraulic roughness.
To demonstrate application of the improved prediction of
annular pressure drop in velocity string design, an example
will be discussed.
Example velocity string design
The vertical example gas well runs to a total depth of 3060
m . The 9 5/8, 32 lbs/ft production casing is perforated from
3000 to 3050 m. The produced gas (s.g. 0.75 air = 1) contains
56 m3/106sm3 (10 bbl/MMscf) condensate (density 59 API,
742 kg/m3) and 560 m3/106sm3 (100 bbl/MMscf) solids-free
water. The initial reservoir pressure is 350 bar (5000 psi) and
temperature is 100 deg. C (212 deg. F). To obtain a high initial
production rate, the well is completed with 7 tubing (ID
5.92) set down to 2900 m. Given a Forcheimer A factor for
inflow performance of 50 bar2/1000 sm3/d, this allows a
production rate of 1.7 MMsm3/d against 70 bar (1000 psi)

IPC, 7 in

400

IPR, 350 bar

350

IPR, 200 bar

300
FBHP, bar

At gas rates exceeding 1.0 MMsm3/d, flowing well head


pressures are over-predicted using the standard hydraulic
roughness, i.e. pressure drops are under-predicted. By
increasing the hydraulic roughness tenfold, to 0.01, a value
commonly used for rusted steel or galvanised iron, a
considerably better match with the high rate data is obtained,
while the results at lower rates are hardly affected. For several
reasons a higher value for hydraulic roughness is justified in
this case:
Prior to the test, the annulus was filled with a
drilling fluid containing solids such as weighting
material and cuttings. In part these solids will
stick to the piping after the clean up, increasing
the hydraulic roughness of the piping walls.
The test string consists of 3.5 drill pipe with a
number of 4 drill collars and 5 connectors.
All crossovers between these pipe sizes will
create additional turbulence, adding effectively to
the hydraulic roughness.

IPR, 167 bar

250
200
150
100
50
0
0

Gas rate, MMsm3/d

Figure 8: Nodal analysis for the well as originally completed

At rates less than 0.3 MMsm3/d, the flow regime in the


well changes from continuous liquid removal in annular(-mist)
flow to intermittent (plug/slug) flow, with a dramatic increase
of the pressure drop over the tubing, i.e. the well loads up with
liquids and only some residual production from gas bubbling
through a liquid column can be maintained.
At or before this point it is considered to install a velocity
string to stabilise production, prolong the productive life of the
well and delay more expensive options such as wellhead
compression. To determine the optimum size of the velocity
string the following criteria are applied:
1. The effect of the velocity string, apart from
stabilisation, should be a substantial prolongation of
the well life.
2. The adverse effect on production should be minimal.
The nodal analysis, as shown in Figure 8 is used to
determine the effect of velocity string size. The following
options are considered:
5.5 tubing
4.5 tubing
3.5 tubing
2 7/8 tubing
2 3/8 tubing
For all sizes the reservoir pressure and rate at which
production may become unstable as well as the reservoir
pressure and rate at which the well will cease to flow are
determined. The result is shown in Table 6.

SPE 104605

OD

ID

PR-load

Q-load

in

in

PR-unstable Q-unstable
bar

MMsm3/d

bar

MMsm3/d

5.92

200

0.55

167

0.29

5.5

4.9

163

0.25

147

0.15

4.5

3.92

149

0.14

141

0.1

3.5

2.9

137

0.065

137

0.06

2.875

2.441

137

0.047

137

0.047

2.375

1.99

137

0.031

137

0.031

First it is noted that with decreasing size of the velocity


string the difference between the rate at which the well may
become unstable and the rate at which it will load up with
liquids, vanishes. This is a consequence of the increasing
friction with smaller tubing size, which causes the pressure
drop to increase with increasing rate for all rates above the
point of loading.
This also gives rise to the fact that no further gain in
lowering the reservoir pressure at which the well loads up is
made by installing a velocity string smaller than 3.5, the
additional friction offsets the beneficial effect of a higher
velocity. Hence the 3.5 string seems the obvious choice for
the velocity string. The downside of this string however is that
at a reservoir pressure of 180 bar, the production rate is
limited to ca. 0.2 MMsm3/d, i.e. directly after installation of
the velocity string, the production of the well drops by more
than 30%. This can be avoided by installing a more flexible
configuration. This is shown in Figure 9.

I: Annular flow

II: Tubular flow

Figure 9: Velocity string flow configurations

The idea is basically that initially the large reduction of


flow area by installing the 3.5 string is not required. Hence
by setting a plug in this well and opening a SSD above this
plug, the well can be produced through the larger area of the
7x3.5 annulus. Once the velocity in this annulus drops and
unstable behaviour and/or loading occur, the plug is pulled
and the SSD closed, forcing flow through the velocity string
itself. The benefits of this strategy are illustrated in Figure 10,

Gas rate, MMsm3/d

in which production rate as a function of reservoir pressure is


plotted for the two configurations.

Table 6 : Effect of velocity string size

0.35
0.3
0.25
Loading
0.2
0.15
0.1
0.05
0
130 140 150

3.5" string
7"x3.5"
annulus

160

170

180

Reservoir pressure, bar


Figure 10: Rate as a function of reservoir pressure

It appears from the graph that flowing through the 7x3.5


initially, i.e. directly after installation of the string at a
reservoir pressure between 170 and 180 bar, gives a rate of ca.
0.3 MMsm3/d, hardly less than the rate through the 7 string,
and considerably more than flowing through the 3.5 string.
With decreasing reservoir pressure this rate decreases until the
well loads up when the rate has decreased to 0.16 MMsm3/d
and reservoir pressure is ca. 150 bar. By switching to flowing
through the 3.5 string at this stage, a stable rate of 0.13
MMsm3/d is obtained. Stable production can be maintained
down to a rate of 0.06 MMsm3/d and a reservoir pressure of
137 bar. Installation of a simple velocity string has resulted in
lowering the reservoir pressure with some 35 bar, i.e. some
10% of the GIIP, before more expensive options such as
plunger lift or wellhead compression have to be considered.
For wellhead compression the assembly can be left in the hole,
flowing through both the string and annulus initially and
repeating the sequence annular flow-string flow at a later
stage, making optimal use of the installed compression
capacity.
CONCLUDING REMARKS
Velocity strings can be a cost effective means to
delay liquid loading in wet gas wells, extend the
well life and increase ultimate recovery.
Selection of the proper string size is critical.
When the flow conduit size is too small
production is restricted unnecessarily and when
the string is too large the beneficial effect of
installing the string will be short lived.
Nodal analysis to select the proper string size
requires accurate methods to predict the pressure
drop over the velocity string as well as the
velocity string-tubing annulus.
The methods to predict the pressure drop over the
velocity string-tubing annulus are usually
modifications of methods to predict pressure drop
over tubing. Since there are considerable
differences between annular and tubing flow, the
accuracy of the modified methods needs to be
assessed.
For this assessment a field test to obtain data on
annular flow of wet gas was designed and
executed.

10

SPE 104605

A comparison between the data and various


approaches to model annular flow demonstrated
the appropriate method to calculate pressure drop
over the tubing-velocity string annulus.
Validation with a field case provided further
evidence that this was the correct approach,
although at the highest rates hydraulic roughness
plays a major role, a parameter, which is rarely
known to a high accuracy.

Nomenclature
A
Area, m2
B
Coefficient in Gray correlation, dimensionless
BHP
Bottom hole pressure, Pa
C
Coefficient
to
correct
pressure
gradient,
dimensionless
D
Diameter, m
e
Eccentricity, dimensionless
f
Friction factor, dimensionless
g
Gravitational acceleration, ms-2
G
Pressure gradient, Pa/m
H
Hold-up, dimensionless
LGR
Liquid-gas ratio m3/sm3
N
Dimensionless number in the Gray correlation
P
Perimeter, m
PD
Dew point pressure, Pa
p
Pressure, Pa
q
Flow rate, m3/s
R
Liquid-gas velocity ratio, dimensionless
r
Hydraulic roughness, m
T
Temperature, K
V
Velocity, m/s
WHP Well head pressure, Pa
Y
Coefficient in Gray correlation, dimensionless
Z
Distance, m
Greek symbols

Density, kg/m3

Surface tension, N/m


Subscripts
conc
Concentric
D
Diameter
e
Effective
ecc
Eccentric
g
Gas
H
Hydraulic
l
Liquid
m
Mixture
o
Oil
v
Velocity
w
Water
SI Metric conversion factors
ft 3.048* E-01=m
ft2 9.290 304* E-02=m2
ft3 2.831 685 E-02=m3
lbf 4.448 222 E+00=N
lbm 4.535 924 E-01=kg
psi 6.894 757 E+00=kPa=0.01 bar*
*Conversion factor is exact.

References
1.
Turner, R.G., Hubbard, M.G. and Dukler, A.E.:
"Analysis and Prediction of Minimum Flow Rate for the
Continuous Removal of Liquids from Gas Wells", SPE
paper 2198, Journal of Petroleum Technology,
November 1969.
2.

Coleman, S.B., et al.: "A New Look at Predicting GasWell Load Up," SPE paper 20280, Journal of Petroleum
Technology, March 1991.

3.

Lestz, R.E.: "Using capillary strings to unload gas wells


and increase production ", World Oil Magazine, February
2003

4.

Lea, J.F. and Nickens H.V.: Solving Gas-Well LiquidLoading Problems Journal of Petroleum Technology,
Volume 56, Number 4, April 1991, pp 30-36

5.

Garg, D., Lea, J.F., Cox, J. and Oetama T.: New


Considerations for Modeling Plunger Performance
paper SPE 93997, presented at the 2005 SPE Production
Operations Symposium, 16-19 April, Oklahoma City,
Oklahoma

6.

Rignol, J; Krepa, J M; Hogan E and Den Besten H.:


Using Coiled Tubing Equipment To Run Complex
Jointed Tubing Velocity Strings, Paper 93586 presented
at the SPE International Coiled Tubing Association
(ICOTA) Coiled Tubing Conference 12-13 April 2005,
Houston, TX

7.

Mejia, L. L.M. , Perez, E. V., Solese, M; Martinez, O. M.


and Galvan, J. L. M.: Eliminating Liquid Loading Hikes
Burgos Basin Production, Oil & Gas Journal, Volume
104, No. 4, pp 42-50

8.

Oudeman, P.: Improved prediction of wet-gas-well


performance, SPE Petroleum Engineering, August 1990,
pp 212-219.

9.

Oudeman, P. and Mason, D.: "Results of a Field Test to


Improve Hydraulic Blowout Control Calculations," SPE
50577, presented at the 1998 SPE European Petroleum
Conference, The Hague, The Netherlands, 20-22
October 1998.

10.

Gray, H. E.: Vertical Flow Correlation in Gas Wells,


Users Manual for API 14B Surface Controlled
Subsurface Safety Valve Sizing Computer Program, 2nd
Edition, (Appendix B), American Petroleum Institute,
Dallas, TX, June 1978.

11

Katz, D. A.: Natural Gas Engineering, McGraw-Hill,


(1951), New York, NY.

12

Haciislamoglu M. and Langlinais J.: "Non-Newtonian


flow in eccentric annuli" Journal of Energy Ressources,
Volume 112, pp 163-169, 1990

You might also like