SPE 115567

Prediction Onset and Dynamic Behaviour of Liquid Loading Gas Wells

S.P.C. Belfroid, SPE, W. Schiferli, SPE, and G.J.N. Alberts, SPE, TNO Science and Industry,
C.A.M. Veeken, SPE, and E. Biezen, SPE, Shell-NAM

Copyright 2008, Society of Petroleum Engineers

This paper was prepared for presentation at the 2008 SPE Annual Technical Conference and Exhibition held in Denver, Colorado, USA, 21–24 September 2008.

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 have not been
reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers, its
officers, or members. Electronic reproduction, distribution, or storage of any part of this paper 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 SPE copyright.

Abstract
As reservoir pressures decrease in maturing gas wells, liquid drop-out forms an increasing restriction on gas production. Even
though virtually all of the world’s gas wells are either at risk of or suffering from liquid loading, the modeling of liquid
loading behavior is still quite immature and the prediction of the minimum stable gas rate not very reliable. Many wells start
liquid loading at gas rates well above the values predicted by classic steady state prediction models such as Turner. The
loading point is strongly dependent on inclination angle, flow regime transitions and the interaction between tubing outflow
behavior and the reservoir IPR. In the paper, the behavior of different natural gas wells and of an air-water test setup are
analyzed. Simulations were performed using both commercially available software and dedicated dynamic models. The onset
of liquid loading and the dynamic behavior of a flooded well during a restart were predicted. These were then compared to
actual production data. The influence of the reservoir parameters and of the tube inclination were of special interest.
The influence of dynamic disturbances on the stability are not taken into account by the classic prediction models.
Systems with high permeable reservoirs are less able to cope with disturbances. This leads to higher critical rates for those
systems. This corresponds to data from field observations. A maximum in the critical velocity is observed around an
inclination of 50° with a critical rate 40% higher than for a vertical well. To solve this, relations found from flooding
experiments are used to modify the current prediction models. Based on the current work an adaptation to the Turner
equation, which takes the inclination effects into account, is proposed. For the observed natural gas wells and for the air-
water experiments the modified Turner equation predicts the observed loading points within 20% accuracy.

Introduction
Liquid loading, that is the process when the gas is no longer able to lift liquid to the surface, is a major limiting production
factor for maturing gas wells. Solutions such as gas lift, soap injection, velocity string or plunger lift are required to solve this
problem. Accurate predictions of the onset of the liquid loading process allow for better planning and choosing the right
countermeasure. Currently, the most widely used model is still the classic Turner criterion, which is based on a force balance
on a falling droplet, although it is known to not always be correct.
In laboratories, liquid loading occurs due to the drainage of the liquid film which is present at the tubing walls in annular
flow (Belt 2008, Westenende 2008). In practice the production decline may also be due to other mechanisms, which may be
difficult to distinguish. The main mechanisms for the production decline are thought to be:
• Film drainage,
• System instability,
• Flow regime change (Toma 2007).

In film drainage the force balance on the liquid film results in a part of the liquid film with a negative (downwards) velocity.
System instability occurs when the inflow performance relation (IPR, reservoir curve) intersects the tubing performance
curve (TPC) to the left of the minimum in the tubing curve. In practice the liquid drainage point may be to the left or to the
right of the TPC minimum. The system stability is also governed by the pressure drop as is the force balance across the liquid
film. The flow regime change is a separate mechanism and is less determined by gravity but is more influenced by increased
hold up and wave formation. The flow regime change itself is more likely a result than an initiator. Slug formation can occur
when the liquid hold up increases. This increase is expected to be caused by the negative liquid film velocity. Therefore,
these three mechanisms may interact and coincide in field cases and the direct cause of a production decline may be difficult
to detect.
2 SPE 115567

All three mechanisms are affected by a large number of parameters. These include the total liquid content, the liquid
properties, the tube diameter, the inclination angle, the entrainment rate, the reservoir characteristics and many more. In this
paper two parameters are analyzed. First the influence of the reservoir characteristics is analyzed and secondly the effect of
tubing inclination angle.
A large number of models are used for the analysis of the test and field data. These include steady state models such as
the Turner criterion or models which describe the force balance across a liquid film (Westenende 2008). Besides the steady
state models also two full dynamic multiphase models are used. One is a model built in-house to simulate annular flow. The
second is a commercial multiphase package which is widely used in oil and gas industry.

Influence reservoir parameter

All production wells encounter some degree of instability. This could be among other things due to small pressure changes in
the flowline pressure, changes in the choke settings or small variations in the liquid input. The reservoir characteristics
determine how the reservoir reacts to changes in the tubing operations. A system with a high permeability reservoir will react
much faster and much stronger to small changes than a system with a low permeability reservoir. From field data it has been
observed that liquid loading in wells with high permeability reservoirs happens faster than for wells with low permeability
reservoirs. In Figure 1 the liquid loading point of a large number of natural gas wells is plotted as function of a reservoir
parameter A. The loading point is made dimensionless with the Turner criterion. The reservoir parameter A is defined with:
Pres2 − Pbot
2
= AQ gas , sc + FQ gas
2
, sc ,
with Pres the reservoir pressure [bar], Pbot the flowing bottomhole pressure, Qgas,sc the gas flow rate [e3Sm3/day] and A and F
are the Darcy and non-Darcy inflow resistance parameters.
Although the spread in the data is large, the trend is unmistakably that for lower A factors the critical rate is higher. That
is, liquid loading occurs already at higher velocities.
3.0

2.5
Qmin,field / Qmin,turner (-)

2.0

1.5 y = 1.4425x-0.0667

1.0

0.5

0.0
0 20 40 60 80 10
A (bar2/e3m3/d)
Figure 1: Liquid loading point, made dimensionless with the Turner Criterion, as function of the reservoir parameter A (Courtesy of
C.A.M. Veeken, NAM).

This can be explained by assuming that the liquid influx is not a steady flow but that there are short periods of incremental
liquid content. In that case, the short period of liquid content can result in an unsteady condition which can kill the well. This
can graphically be explained by examining the behaviour of the TPC and IPR intersection. If we assume a production steady
state of 4 kg/s gas at a liquid content of LGR=500 Sm3(liquid)/e6Sm3(gas) and a reservoir characterized with
A=4 bar2/e3Sm3/day, the production remains stable even with periodic liquid increases up to LGR=750. However, at the same
operating point with a reservoir of A=0.1 bar2/e3Sm3/day the same liquid increase results in a non-producible situation as the
IPR and TPC no longer intersect. Often a limited increase in pressure drop due to a liquid accumulation is already too high to
restart stable production, and the well does not recover. Of course the liquid variations expected in practice are lower than
described above, but the same happens for smaller disturbances.

In Figure 3 the dimensionless critical flow rate is plotted as function of the reservoir A parameter for a 3km long vertical
well with an ID=0.1571m with an initial liquid load of LGR=100 Sm3/e6Sm3/day and a liquid upset to LGR=150
Sm3/e6Sm3/day. The same trend observed in the field data is seen in the numerical results. The critical rate increases for low
SPE 115567 3

A factors to twice the Turner criterion. The analysis presented above is based on a quasi steady analysis although the
mechanism also is present in the steady state case. For a fully dynamic system the behavior is even worse as in that case the
IPR and the TPC are no longer fixed straight lines but can become ellipsoids (Sturm 2004). The steady state cases are also
simulated with the two fully dynamic two fluid models. Both cases with a reservoir parameter A=4 bar2/e3Sm3/day and A=0.1
bar2/e3Sm3/day are subjected to a disturbance of a liquid increase from LGR=100 to 150 Sm3/e6Sm3/day (Figure 4). In
Figure 5 the liquid hold up is plotted for both cases as function of time. The A=4 case is able to produce the liquid upwards
out of the tubing whereas for the A=0.1 case the liquid keeps dropping to the bottom of the tube where it accumulates and
finally kills the well even after the liquid inflow content is reduced again.

60
LGR = 100
LGR = 150
LGR = 500
50 LGR = 750
A = 0.1
A=4

40
FBHP [bar]

30

20

10

0
0 1 2 3 4 5 6 7 8 9
Qgas [kg/s]

Figure 2: TPC and IPC curves for different liquid content and different reservoir parameters.

1.8

1.6

1.4
Qmin/Turner [-]

1.2

0.8

0.6

0.4

0.2

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Ares [bar2/(1e3 Sm3/day)]

Figure 3: Liquid loading point, made dimensionless with the Turner Criterum, as function of the reservoir parameter A.
4 SPE 115567

160

150

140
LGR [m3/E6 Sm3]

130

120

110

100
0 2000 4000 6000 8000 10000 12000
time [s]

Figure 4: Liquid inflow content as function of time.

A=4 A = 0.1
0.04 0.35

0.035 0.3

0.03 0.25
Liquid
accumulates at
liquid hold-up [-]

liquid hold-up [-]

0.025
0.2
bottom and
0.02
0.15 ‘restart’ is not
0.015
0.1 achieved
0.01
0.05

0.005
0 2000 4000 6000 8000 10000 12000 0
time [s] 0 2000 4000 6000 8000 10000 12000
time [s]
2 3
Figure 5: Liquid hold up as function of time for a system with a reservoir parameter of A = 4 bar /e3Sm /day (left) and A = 0.1
2 3
bar /e3Sm /day (right). The system parameters are Length L=3000m, tube diameter ID=0.1571m, Mole weight MW= 21.63 g/mole,
Tubing pressure Ptop=10 bar, Reservoir pressure Pres=35.6 bar (A=4 case), Pres=18.6 bar (A=0.1 case).The different lines are for
results at different heights in the tubing.

Influence of inclination angle

In deviated wells different mechanisms become important. In going from a vertical to a horizontal wellbore, gravity becomes
less important. For a horizontal well almost by definition no liquid loading can occur as there is no driving force for the
liquid to counterflow the gas. The predominant flow regime changes significantly with varying inclination. For a vertical
well, the liquid is distributed along the full circumference of the tubing whereas for a horizontal flow the dominant flow is a
stratified flow. For inclinations between these extremes the liquid film at the bottom of the tube becomes progressively
thicker compared to the top of the tube. Furthermore, secondary gas circulation will appear in the gas phase which will
influence the deposition and entrainment rates (Westenende 2008).
Both mechanisms of decreased gravity and of increased film thickness will influence the critical rate. In Figure 6 the
critical rate is plotted as function of the inclination angle for an air-water system (Westenende 2008). The critical rate
increases for the medium inclination, which is likely due to the increased film thickness, whereas the reducing effect of
gravity is seen for the larger inclinations.
SPE 115567 5

20
Experimental data TU Delft

18
Critical gas velocity [m/s]

16

14

12

10

8
-10 0 10 20 30 40 50 60 70 80 90
Inclination angle (0 = vertical) [degrees]

Figure 6: Critical gas velocity as function of inclination angle. The system parameters are Length L=12m, tube diameter ID=0.05m,
Mole weight MW= 28.96 g/mole, Tubing pressure Ptop=1.01 bar (data Technical University Delft [Westenende]).

The critical gas velocity is generally predicted using the Turner criterion or similar criteria. In Figure 7 predictions using
different models are compared to the experimental results for an air-water experiment (Figure 6). The different models
include besides steady state predictions such as the Turner criterion or a liquid film force balance model (Westenende 2008)
and two full dynamic multi-fluid models. Finally, a prediction based of flooding experiments in small diameter tubes as
presented by Fiedler is also included (Fiedler 2004).

The Turner criterion prediction compares very well with the experimental value for the vertical case, however there is no
angle dependency in the Turner criterion. The steady state model which describes a force balance across a liquid film gives
the gravity dependence but fails to capture the increase in the critical velocity for the mid inclination angles. This also holds
for the dedicated dynamic model. In the results of the commercial model the same trend is simulated. Finally, the prediction
based on small tube diameter flooding experiments is off with respect to the absolute value but captures the angle
dependency reasonably. Since the Turner criterion yields critical velocities in the measured range, and the Fiedler model
appears to capture the relationship between critical velocity and inclination well, it was decided to combine the Fiedler shape
function to the conventional Turner prediction. The critical velocities predicted by this model fit the experimental data very
well for all inclination angles.

The angle-corrected Turner criterion is given by:

M g ,min = 3.1A ( ρ g (gσ (ρ l − ρ g )) )

(
1sin(1.7 β ) )
4
0.38
, (1)
0.74
With A the tube area [m2], ρg the gas density [kg/m3], ρl the liquid density [kg/m3], σ the surface tension [N/m], g the
gravitation acceleration [m/s2] and β the inclination angle [º] with β=0º is the horizontal. The second term is the correction
term.
6 SPE 115567

20

18

16

14
Critical gas velocity [m/s]

12

10 Experimental data TU Delft

Film model (steady state)
TNO Dyn. simulator
8 Commercial Dyn. simulator
Turner
Turner corrected
6
Fiedler

0
-10 0 10 20 30 40 50 60 70 80 90
Inclination angle (90 = horizontal) [degrees]

Figure 7: Comparison between experimental and model predictions

The above results were obtained by comparison with air-water data. In order to verify the applicability to gas wells, results
were compared to field data. In Table 1 an overview is given of the characteristics for the air-water case and of the two
natural gas wells. The two gas wells differ in internal diameter but especially in the trajectory and in the reservoir
characteristics. The well trajectories of both wells are plotted in Figure 8. The majority of the tubing of both wells has an
inclination of 40º with Well-1 ending in a short horizontal section.

Table 1: Characteristics test cases.

Air- Water Well-1 Well-2
Inner diameter [m] 0.05 0.112 0.074
Length [m] 12 3743 3545
Inclination angle [°] 0, 30, 60, 90 Variable Variable
Gas [g/mole] Air, 28.9 18.6 17.5
Liquid content [Sm3/E6 Sm3] Water, Water, 21 Water, 17
176 – 35218 Cond., 1
Wellhead pressure [bara] 1 14 21
Wellhead temperature [°C] ≈ 20 45 30
Reservoir temperature [°C] ≈ 20 120 110

3
Water density [kg/m ] 1000 1020 1050
3
Condensate density [kg/m ] - 820 800
Liquid viscosity [Pa s] 0.001 0.00033 0.00041
Gas viscosity [Pa s] 1.8e-5 1.4e-5 1.3e-5
Surface tension [N/m] 0.073 0.062 0.064

2 3
Reservoir parameter A [bar /(e3m /d)] - 175 35
2 3 2
Reservoir parameter F [bar /(e3m /d) ] - 0 0.0693
SPE 115567 7

90
Well-1
Well-2
80 Air-water(90)
Air-water(60)
70 Air-water(30)
Air-water(0)
inclination angle [deg] (0=vertical)

60

50

40

30

20

10

0
0 500 1000 1500 2000 2500 3000 3500 4000
AHL [m] (Air-water [cm])

Figure 8: Well trajectories of the different test cases.

With the full dynamic models the loading point was determined by two means. In the first method a semi-steady approach
was used in which gas mass flow rates were used to determine both the TPC as well as the minimum gas flow rate. This
method gives an uncertainty due to sensitivity to initial conditions for the gravity-dominated part of the TPC and because the
reservoir characteristics are not taken into account. Therefore it was opted to use a decline simulation. In this simulation the
reservoir characteristics are used and the reservoir pressure is reduced over a larger period of time. In this way the
initialization assumptions are eliminated. In Figure 9 the results of the decline simulation of Well-2 are given using the
commercial dynamic simulation tool. In this graph the bottomhole pressure is plotted as function of the gas flow rate. In the
simulation, the gas flow rate starts at high flow rates and decreases over time. Furthermore, the liquid film and liquid droplet
mass flow rates are plotted. In the decline curve several effects are observed. With a decreasing gas flow rate both the film
and droplet flow rate decrease with a constant liquid gas ratio. At a certain velocity the film rate decreases rapidly and
becomes negative. The produced liquid (liquid to gas ratio) remains at a constant level down to a lower gas rate at which it
starts to drop rapidly. The negative film velocity is compensated for by sharply increasing upward droplet transport. At a
lower gas velocity, the software predicts that this transport is no longer possible, and liquid transport ceases. The point at
which droplet transport is no longer possible corresponds to a predicted change in flow regime to slug flow. The point of the
negative liquid film flow rate corresponds in this case with the minimum in the TPC. It must be noted that this point is at
higher rates than anticipated from the steady state TPC. The point at which the liquid film rate becomes negative is denoted
as the loading point.
8 SPE 115567

60 20
Total liquid
18

Liquid film mass [*10 Sm3/day]; Liquid

50
16
Droplets
Pressure (Bottomhole) [bara]

ratio (LGR) [e6Sm3/Sm3]

14
40
TPC decline 12
curve
30 10

TPC (ss) 8
20
6
Liquid film
4
10
2

0 0
40000 50000 60000 70000 80000 90000 100000 110000 120000
Gas flow rate [Sm3/d]

Figure 9: Simulation results decline simulation for natural gas Well-2. Green indicates the liquid film mass flow rate, blue indicates
3 3
the droplet mass flow, pink indicates the liquid gas ratio (Sm /mmSm ), dark blue is the steady state TPC and red indicates the
dynamic TPC.

In Table 2 the prediction results using the different models are compared to the field data. In the table also the results of the
air-water case are included. Based on the field observations the two wells had onset velocities of 90000 Sm3/day and 45000
Sm3/day respectively for Well-1 and Well-2. The TNO dynamic model was not able to predict the loading point due to a very
high hold-up for Well-1. This high hold-up resulted in a non-annular flow and the model was therefore out of its application
area. The commercial tool, in general, over predicts the onset of liquid loading for the natural gas cases. It is thought to be
due to a tube diameter effect and possibly due to the fact that in the model it is assumed that all liquid is at the bottom of the
tube. This has the effect that the critical rate will be over predicted (SPTgroup 2007). As already previously described, the
annular flow model was also not able to capture the inclination effects. The adapted Turner criterion (Eq. 1) based on
bottomhole conditions is able to predict the onset for both the natural gas wells as well as the air-water cases including the
angle dependency correctly. The total uncertainty in the prediction for the current cases is less than 20%. This is much better
than the current predictions. In a next phase more gas wells will be used to validate the angle dependency.

Table 2: Comparison prediction onset rate liquid loading, made dimensionless with the test or field data,
for the two natural gas wells and the air-water test case.
Inclination Experiment Commercial TNO dynamic Turner
Angle /field data dynamic model (bottomhole,
model [-] angle adapted)
[-]

Well-1 40 90000
3
Sm /day 2.1 - 1.2
Well-2 40 45000
3
Sm /day 1.6 - 0.9
Air –Water 0 13.3 m/s 1.3 0.9 1.1
30 17.5 m/s 0.9 0.9 1.1
60 15.5 m/s 0.9 0.7 1.2
80 11 m/s 1.0 0.4 1.1

Error ±110% -60% ±20%

SPE 115567 9

Conclusions
The onset of liquid loading is determined by the transport of the liquid film. Two main parameters which influence the onset
of liquid loading have been investigated. The first parameter is the reservoir characteristics. Systems with high permeability
reservoirs will show loading behavior much faster than low permeability reservoirs. Systems with low permeability reservoirs
are much more resilient to small changes in operating conditions or liquid content. The systems with a high permeability
reservoir will not be able to cope with small changes and will therefore become unstable much faster. The second parameter
is the tube inclination. The onset of the loading could be correctly predicted for vertical tubes for both air-water systems as
well as natural gas wells. For larger inclinations from the vertical, the effect of gravity is reduced and therefore the critical
gas rate is lower. However, at larger inclinations the liquid film thickens at the bottom of the tube, compared to the top,
which increases the critical gas rate. The total results in the highest critical rate occurring at mid inclinations. None of the
tested models predicted this effect correctly for all test cases. The Turner criterion could be adapted with a shape function.
With the angle dependent Turner criterion based on bottomhole conditions the liquid loading velocity could be predicted for
the different cases within a 20% error. The applicability of this criterion is currently tested on a larger set of field data.

Acknowledgement
We would like to acknowledge the support of the Technical University of Delft and especially Jos van ‘t Westenende for
allowing the use of the air-water experimental data.

References
R.J. Belt, “On the liquid film in inclined annular flow” PhD thesis TU Delft (2008)

N. Dousi, C. Veeken, P. Currie, “Numerical and Analytical Modeling of the Gas-Well Liquid-Loading Process”, SPE Production &
operations November 2006

S. Fiedler, H. Auracher, “Experimental and theoretical investigation of reflex condensation in an inclined small diameter tube”, Int. J. Heat
and mass transfer 47(2004) 4031-4043

Peters, M., e.a. ,“Gas lifted gas wells”, First European Deliquification Conference 2006 (Zeegse)

SPT Ggroup , Internal communications, September 2007

Sturm, W.L. , e.a., ”Dynamic Reservoir Well Interaction”, SPE 90108 presented at the 2004 Annual Technical Conference and Exhibition
2004, Houston USA.

Toma, P., E. Vegas, E. Kuru, “Prediction of Slug-to_annular Flow Pattern Transition (STA) for Reducing the Risk of Gas-Lift Instabilities
and Effective Gas/Liquid Transport for Low-Pressure Reservoirs”, SPE Production & operations November 2007.

Westenende, J. Van ‘t , “Droplets in annular-disperserd gas-liquid pipe-flows “ PhD thesis TU Delft (2008)