SPE-185485-MS

Deliquification in a Mature Gas Field: Comparison of Predicted and

Observed Rates after Well Workover to Install Velocity Strings

J. Burford, Imperial College London; G. Falcone, Cranfield University

Copyright 2017, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Latin America and Caribbean Petroleum Engineering Conference held in Buenos Aires, Argentina, 18-19 May 2017.

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
Liquid loading is a significant issue within gas and gas condensate wells, resulting in production decline
and rate instability. The Southern Gas Field serves as a case study, to investigate a field experiencing liquid
loadup. The field is located in the Southern North Sea, and is being produced via natural depletion with
no pressure support from the surrounding aquifer. In Yr. 9 production well A ceased to flow as a result
of liquid build up, and by Yr. 11 well B also showed signs of liquid loading. The load fluid consisted of
liquid condensate, and this was confirmed by wireline investigations. A decision was made by the operator
to deploy two 2 3/8" velocity strings, in an attempt to extend the field lifespan, by restarting well A and
returning rate stability to well B. Velocity strings are a relatively low cost deliquification method, that aim
to increase the gas flow rate above the critical flow rate, enabling the well to continuously unload liquids.
The design of velocity strings is crucial, as if too small a tubing is used, the velocity string can act as a
choke, suffocating the well. The velocity string run in well B proved a success, however the workover in
well A failed to restart the well. Batch foamer treatments were later tried in well A but proved ineffective.
The effectiveness of foam to deliquifying gas wells, depends upon the composition of the load liquid, as
both water and liquid hydrocarbons react differently to surfactants. Typically liquid hydrocarbons such as
condensate do not foam well, and must be agitated to maintain foaming.
This study aimed to make use of Petroleum Experts Integrated Production Modelling (IPM) software to
model the effect of the velocity string deployment, with regards to returning rate stability. Prosper models
were created and Systems Analysis was performed and confirmed that 2 3/8" velocity strings will lift both
well A and well B out of the Turner region. Prosper was also use to model the effect of using foam as
an artificial lift method, with the results identifying that the water gas ratio (WGR) of the field is too low
for effective deliquification using foams. A GAP model was also created for the Southern Gas Field to
determine the incremental reserve gains to be had by deploying velocity strings. The simulations identified
that well workovers can provide an additional 4.99 Bscf to the recoverable reserves.

Introduction
The Southern Gas Field is located in a water depth of 35m in the Southern North Sea. The offshore facilities
consist of a Normally Unmanned Installation (NUI) platform, 2 production wells (well A and well B), and
2 SPE-185485-MS

one 8" pipeline about 7km long running along the seabed and then upwards to the processing platform. Well
A commenced production in April Yr. 1, with well B coming online in August the same year. The wells
differ in completion style, with well A having open hole completions, and well B being completed over a
perforated interval. Both well A and well B were completed with dry trees.
The reservoir is producing under natural depletion, with no pressure support from the surrounding aquifer.
The average permeability of the reservoir is 13mD and the pay thickness is 23m. The petroleum fluid is made
up of gas and condensate, with the CGR reducing throughout the production life span. Seismic data mapped
a fault between the two producers, however pressure build-up from both wells demonstrated no significant
differences and no evidence of baffles. It was concluded both producers were in pressure communication
and that the fault separating them did not act as a barrier to flow, thus the 2 wells were adequate for the
drainage of the field. The field gas initially in place (GIIP) is about 100 Bscf based on seismic and confirmed
by P/Z analysis.
In Yr. 8 the gas rates from well A became erratic, declining more rapidly than forecast, and the well ceased
to flow by Yr. 9 and was subsequently shut in. The erratic behaviour was thought to be caused by liquid
loading, and this was confirmed by a wireline investigation that discovered condensate in the wellbore. A
decision was made to try and restart well A using a 2 3/8" velocity string. An identical velocity string was
also installed in well B at the same time, as the well was experiencing a decline in production rates and an
increase in liquid (condensate) production. The workover of both wells was conducted in Yr. 13.
Technical issues were encountered in well A, and the velocity string could only be placed 400ft above the
open hole completions, due to lower completion restrictions limiting access to the wellbore. Post workover,
well A failed to restart. Nitrogen injection and foaming agents were additionally used in later years to try
and restart well A, however both proved ineffective. The identical velocity string run in well B proved
effective however at returning rate stability.
This thesis aims to investigate the failure of well A velocity string and the unsuccessful foam treatment
to the well. One objective of this thesis is to review the wellbore completions diagrams and design, to
determine whether the 2 3/8" velocity string would have proved effective had it been run to the top of the
open hole completions in well A, or whether a 1" velocity string could have been run through the lower
completion restrictions. Alternative methods of gas well deliquification will also be reviewed, to provide an
insight into the most economic method to implement. Petroleum Experts IPM software will be utilised, to
produce Prosper models for both well A and well B, pre workover and post workover, to assess the benefits
of velocity string deployment. Prosper will also be used to trial foam lifting on well A, to determine whether
batch foamer treatments should have proved successful. A GAP model will be created and used to assess
the production of the well A and well B. Predicted cumulative production, reservoir pressure and gas flow
rates from GAP, will be compared with historic production and well test data. Typically velocity strings are
deployed prior to the death of a well, so workover timings will also be investigated.

Liquid Loading
Liquid loading occurs when the reservoir has insufficient energy to transport liquids to the surface, resulting
in the accumulation of liquid downhole. If a gas field is producing via natural depletion-drive, the transport
energy, allowing the fluids to be produced at surface will decline with time (Coleman et al., 1991). As the
static reservoir pressure declines the amount of condensed water increases exponentially, therefore natural
reservoir depletion compounds the problem of liquid loading (Coleman et al., 1991). A reduction in well
productivity can occur as liquids saturate the near-wellbore region, causing a reduction in gas relative
permeability.
Liquid loading occurs when the gas flow rate drops below the qgc required to remove liquids from the
wellbore (Lea et al., 2008). The liquids present can be the result of the reservoir pressure falling below the
dew point (condensate formation), water coning, water influx from aquifer support or as a result of free
SPE-185485-MS 3

formation water (Schinagl and Denny, 2007). The typical symptoms of liquid loading outlined by Hearn
(2010) include orifice pressure spikes, liquid slugging, fluctuating gas production, variance from decline
curve, a cease of liquid production, and pressure surveys revealing a heavier gradient in the tubing pressure.
Gas well liquid loading has been well documented throughout literature, with Turner et al. (1969) and
Coleman et al. (1991) exploring the criterion required for continuously unloading liquids. Turner et al.
(1969) compared two physical models for transporting fluids up vertical conduits, and deduced that the
onset of liquid load up could be best modelled and predicted using the liquid drop entrainment model. The
conclusions from Turner et al. in 1969 were that the minimum flow conditions required to unload a gas
well are those that provide a gas velocity sufficient to remove the largest liquid drops found in the well, and
that to ensure all liquid droplets are removed the critical gas velocity equation must be adjusted upwards by
20 percent. The terminal gas velocity with the 20 percent upward adjustment as suggested by Turner et al.
(1969) can be calculated using equation 1, and the critical gas rate can be calculated using equation 2.

(1)

(2)

The well data used as a comparison by Turner et al. (1969) mostly had wellhead flowing pressures
(WHFP's) greater than 500 psi, so Coleman et al. (1991) decided to focus on wells with WHFP's below
500 psi, because gas well loading issues worsen with depletion and a decline in reservoir pressure. Upon
comparison with field data the unadjusted liquid droplet model (no 20 percent adjustment), offers a better
match for low pressure wells with WHFP's below 500 psi (Coleman et al., 1991). The terminal gas velocity
without the 20 percent adjustment can be calculated using equation 3.

(3)

Prosper will be used to analyse the onset of liquid loading and model deliquification methods for the
wells of study. Within Prosper the Turner equation takes a slightly different form as shown by equation 4.

(4)

Equation 4 is taken from the original Turner et al. (1969) paper, however the units here are inconsistent
with the constant given, as the constant 20.4 relates to when surface tension is defined in units of ‘lbf/ft’
not ‘dynes/cm’ (IPM Prosper, version 14, March 2016). The Petroleum Experts 4 mechanistic model uses
a constant of 2.04 as this has been shown to give more reliable results in a wide range of examples (IPM
Prosper, version 14, March 2016).
The methodology for predicting the onset of liquid loading has been reviewed and extended by various
authors (Guo et al., 2006; Solomon et al., 2008; Zhou and Yuan., 2010; Sutton et al., 2010). Limitations of
current models predicting load up have been identified by Solomon et al. (2008), and the importance for
the need for a fully transient multiphase flow approach has been highlighted.
Well A from the Southern Gas Field showed signs of producing within the Turner region and suffering
from liquid loading issues, as the gas production rate became erratic in Yr. 8 as shown by figure 1. In order to
return rate stability a deliquification strategy must be implemented, to lift the well out of the Turner region.
4 SPE-185485-MS

Figure 1—. Fluctuating gas rate between Yr. 8 and Yr. 9 indicative of liquid loading

Gas Well Deliquification

Gas well deliquification techniques aim to delay or remediate the effects of liquid loading by reducing the
qgc, thus maximising production and the ultimate recoverable reserves (Vogelij et al., 2016).
Deliquification techniques have been well documented in literature and include velocity strings (changing
tubing size), surface compression, plunger lift, foam assisted lift, hydraulic pumping, beam pumps, gas lift,
electric submersible pumps and progressing cavity pumps (Lea et al., 2008). For the purpose of this project,
the focus will be on offshore deliquification techniques, as the field of study is located in the Southern North
Sea. Table 1 from Vogelij et al. (2016) outlines the most practical offshore deliquification methods.

Table 1—Summary of most practical offshore deliquification techniques, modified from Vogelij et al. (2016)

Advantages: Disadvantages:

Intermittent No installation costs Less recovery than other gas

Production/Cycling well deliquification methods

Surface Compression No well intervention, possible Constrained by cost

to benefit multiple wells,
generates reserves and capacity

Velocity String No operational costs, Loss of capacity especially

generates reserves in case of prolific
inflow performance

Batch Foam Low cost Chemical permitting and

Produced water Re-
injection needed, logistical
challenges offshore

Continuous Foam Can be operated remotely, Chemical permitting and

generates reserves Produced water Re-injection
needed, constrained by chemistry

Intermittent Production/Cycling
In order to cycle a well to effectively remediate loading, the well production rate, wellhead temperature and
wellhead pressure must be monitored (Schinagl and Denny, 2007). By shutting in a gas well suffering from
loading at the appropriate time, the aim is to allow pressure to build up and ‘charge’ the near-wellbore region,
so that upon opening the well, the increased pressure might lift liquids, and remediate loading (Schinagl
and Denny, 2007). Cycling can prove very effective if optimum shut-in and opening times are found, and
Schinagl and Denny (2007) presented an example where cycling doubled gas production in one North Sea
gas well. The advantage of intermittent production is that it can be used in any environment and for any
completion type (Schinagl and Denny, 2007). Cycling is not effective for wells that are heavily loaded and
SPE-185485-MS 5

cease to flow, and there will reach a point in the life time of a well where cycling will cease to remediate
loading (Schinagl and Denny, 2007).

Surface Compression
Compression acts by lowering the wellhead pressure, resulting in a lower bottomhole flowing pressure.
Gas velocity can be increased by surface compression, resulting in an increase in production and reserves
(Lea et al., 2008). Providing production rates are high at low pressures, surface compression can provide
an economically profitable deliquification method, despite the investment associated with the running and
maintenance of the compressor unit (Lea et al., 2008). On a systems graph surface compression causes
the inflow performance relationship (IPR) to remain fixed, but the vertical lift performance (VLP) curves
shift downwards with reducing wellhead pressure, leading to operating points with higher production rates
(Schinagl and Denny, 2007). Offshore compressors are often used on a field scale to reduce the gathering
point pressure, and with regards to the North Sea there has not yet been a purpose to install small compressors
on a well by well basis as typically done onshore (Schinagl and Denny, 2007). An issue with regards to
installing surface compressors in the offshore environment, is the lack of available platform space. Offshore
platforms may also have no spare power capacity to drive the local surface compressors.

Velocity Strings
The installation of a velocity string, i.e. coiled tubing inside the original tubing, reducing flow area to
increase gas velocity, is an appealing option since it is low cost, can be installed without the need to kill the
well and no maintenance is required post installation and workover (Oudeman, 2007).
The hydrostatic backpressure on the formation will be higher after the installation of a velocity string in a
dead well (prior to reflowing the well), due to the fact that the fluid column will occupy a "greater height" in
a smaller diameter tubing (Schinagl and Denny, 2007). Under such conditions Schinagl and Denny (2007)
note production must be started with an alternative method such as nitrogen injection, coiled tubing gas
lifting or venting. The design of velocity strings is crucial, as too small a tubing diameter will cause a high
frictional pressure drop, and the velocity string may act as a choke (Lea et al., 2008). However the tubing
must still remain small enough to increase the gas velocity and prevent liquid loading. The disadvantages
of downsizing the tubing are identified by Lea et al. (2008) as the inability to run test tools in smaller
strings and the inability to swab or attempt nitrogen lift in smaller tubing (especially 1" tubing). For effective
deliquification and unloading with velocity strings it is required to downsize the string, to an even smaller
diameter as the reservoir pressure continues to drop (Schinagl and Denny, 2007).
North Sea regulations state that the subsurface safety valve (SSSV) must be fully operational at all times
(Fadipe, 2012). When installing velocity strings conventional SSSV's are a major drawback, meaning the
string cannot be hung from surface but must instead must be hung from below the SSSV (Oudeman, 2007).
Due to the limitations conventional SSSV's cause, SSSV's are being developed by companies, to allow a
uniform string diameter to be run through them, thus allowing the string to be run from surface to the desired
depth (Schinagl and Denny, 2007). Figure 2 from Fadipe (2012) shows how with a conventional SSSV the
velocity string must be hung below the SSSV. With regards to the Southern Gas Field the velocity string
could be set from surface for both well A and well B as Baker Hughes tubing-retrievable safety valves
(TRSV's) were installed.
6 SPE-185485-MS

Figure 2—Typical velocity string and tail pipe installation (Fadipe, 2012)

Coleman et al. (1991) stated that well completions have great importance, as liquid loading in wells
with concentric tubing strings terminating some distance above the completion interval are significantly
influenced by the larger diameter at the wellbore base. This point is crucial for the investigation of well A
velocity string, as the velocity string run terminated some 400ft above the openhole completions.

Chemicals
Surfactant injection can be applied via batch treatments, continuous injection or via the use of soapsticks
which are dropped down the wellbore. Micro strings (0.25" to 0.375" OD coiled tubing) can be used to
inject the foaming agents, with Germany and the Netherlands installing such completions for this purpose
(de Jonge and Tousis, 2007). The surfactant in each application reduces the surface tension of the liquid
present in the gas well, to allow more gas-liquid dispersion (Lea et al., 2008). The liquid and gas become
tied together, and the liquid is then held as a bubble-film, exposing it to a greater surface area, resulting in
less gas slippage and a lower density mixture, which can be unloaded via the wells' natural energy (Schinagl
and Denny, 2007). The liquid is transported as foam up the wellbore to surface, where it is then "released" by
the use of a defoamer (Schinagl and Denny, 2007). Vosika (1983) identified that foamer agent performance
depends upon foamer formulation, the concentration of the foamer, the condensate water ratio (CWR) and
the salt content of the solution.
Libson and Henry (1980) describe how foam is more effective for water based liquids, with water and
liquid hydrocarbon reacting differently to surfactant agents. Liquid hydrocarbons such as condensate, do
not foam well, this is especially true for light condensate hydrocarbons (Lea et al., 2008). Campbell et al.
(2001) confirmed foam is more effective for water based liquids, with hydrocarbons not foaming well due to
them being nonpolar and having weak forces of attraction between molecules. Water molecules however are
polar resulting in high film strengths with surfactants (Lea et al., 2008). Foaming agents do exist, designed
specifically for condensate loaded gas wells, and these agents are cationic (Wylde, 2010).
Vogelij et al. (2016) describe how ONEgas have had success in reducing the critical gas flow rate in
7 out of 8 wells producing with foamer injection, within the Southern North Sea. Foam injection reduced
the critical gas flow rate to 55% of the original rate, and resulted in an additional 104e6 Sm3 of gas being
recovered (Vogelij et al., 2016). This case study is evidence of the gains to be had with successful foamer
application. ONEgas have successfully applied batch foamer treatments within the Southern North Sea, but
there have been operational challenges and lessons learnt, as outlined in table 2.
SPE-185485-MS 7

Table 2—Operational challenges experienced by ONEgas (Vogelij et al., 2016)

Issue Impact Mitigation

include downhole siphon, adopt

Stops foamer injection, conservation and commissiong
Downhole creates liquid loading, reduces procedures, sample foamer quality,
blockages uptime, increases OPEX inspect surface foamer filter

Causes capillary string leaks and Use high grade subsurface

Sand Erosion possible fish, increases OPEX materials, avoid high flow velocities

Poor Foam
Separation Causes separator trips Inject anti-foamer

Methodology
Petroleum Experts (IPM) software was used for the modelling of the Southern Gas Field production system,
from reservoir, to wells and then to the surface network (IPM version 10.0, Petroleum Experts, May 2016).
A history-matched MBAL model was provided by the operator, as well as a base case GAP model, and
these were used as reference models, to compare the new models against.
The following models were created for the investigation of the Southern Gas Field:

• MBAL model to describe the reservoir

• Prosper model for well A original completion

• Prosper model for well B original completion

• Prosper model for well A 2 3/8" velocity string completion

• Prosper model for well A 1" velocity string completion

• Prosper model for well B 2 3/8" velocity string completion

• Prosper model for well A foam injection

• GAP model to define the surface pipeline network, and run a field forecast

Prosper
Models for each well completion pre-workover and post-workover were created using Prosper, which is a
Multiphase well and Pipeline Nodal Analysis tool (version 14.0, Petroleum Experts, May 2016). The PVT
within each Prosper model was kept the same for consistency.
Completions. Well A and B completions schematics provided by the operator were reviewed, and used to
produce the Prosper completions. It was found that wellbore restrictions had negligible effect on pressure
drop predictions, so were removed, (except for the SSSV's as these must remain fully operational) to simplify
the Prosper completions and reduce running times.
Generating Vertical Lift Performance (VLP) Curves. VLP curves were produced to represent the lifting
performance of the wells, both pre-workover and post-workover.
Multiphase Flow Correlations
To generate VLP curves, Prosper uses multiphase flow correlations (tubing correlations), which can then
be adjusted and corrected to match well test data, when available. Within a producing well the flowing
pressure gradient (loss) comprises of three terms: gravity, friction and acceleration. The gravity pressure
loss is due to the density of the production fluid mixture, the frictional pressure loss is a result of the shear
stress between the flowing fluids and the pipe wall and the acceleration pressure loss is a direct result of the
expansion of fluids as the pressure reduces (IPM Prosper, version 14, March 2016).
8 SPE-185485-MS

The aim of VLP curves is to calculate the pressure drop along the wellbore, such to best mimic the
wells performance. Empirical and mechanistic multiphase flow correlations exist. Mechanistic models use
fundamental physics to calculate the holdup, whereas empirical methods are based on experimental data.
Petroleum Experts 4 and 5 flow correlations are mechanistic in nature, resulting in long computational time,
compared to empirical correlations.
Tubing Correlation Comparison
Due to the field having no downhole pressure gauges, the bottomhole pressure (BHP) predictions from the
various tubing correlations could not be compared with downhole measured data, to identify the correlation
that would most accurately predict the pressure drop in the wells under investigation for this study.
The following tubing correlations were neglected as they were deemed unsuitable for gas wells with
loading issues:

Table 3—Discarded tubing correlations, (IPM Prosper, version 14, March 2016)

Correlation Reason for discarding

Fancher Brown Correlation used for quality control purposes

Hagedorn Brown Should not be used for condensates, under predicts VLP at low rates
and should not be used for predicting minimum stable rates

Duns and Ros Modified Should not be used for lift curve generation, should only be used for
quality check of input well test data

Petroleum Experts 4 Advanced mechanistic model, long computational time

Petroleum Experts 5 Longer computational time when generating VLP curves within GAP

The Gray, Petroleum Experts, Petroleum Experts 2 and Petroleum Experts 3 correlations were compared,
for wells A and B, for both the original completions and velocity string completions, using the tubing
correlation comparison option within Prosper.
Flowing well head pressures and gas flow rates from production data were entered into the tubing
correlation comparison within Prosper and the correlation was selected on the basis of it falling in the centre
of the correlation spread.
For the original completions for well A and well B, the Petroleum Experts 2 correlation was selected,
as it fell within the centre of the correlations spread, thus providing the average pressure drop for the well
(Figure 3 and Figure 4). However for the velocity string completions for both well A and well B, the Gray
correlation fell within the centre of the correlation spread, and was therefore selected (Figure 3 and Figure 4).
SPE-185485-MS 9

Figure 3—Tubing correlation comparison for well A, original completions (left), 2 3/8" velocity string (right)

Figure 4—Tubing correlation comparison for well B, original completions (left), 2 3/8" velocity string (right)

Generating Inflow Performance Relationship (IPR) Curves

Several models exist for the generation of IPR curves. The Jones IPR model was selected in this case, which
required the input of various reservoir parameters. Alternative IPR models were available such as the multi-
rate C and n model, but this was discarded due to the limited field data available. Other IPR models such
as the Hydraulically Fractured Well model, Horizontal well model, Multi-layer Reservoir model and Dual
Porosity model, were discarded as they were not applicable for the field under investigation. The Jones IPR
varied between well A and well B, due to the wells having a different permeability and perforation interval.
The permeability input was based on well test data.
The Technical Open Flow Potential (TOFP) determined for each well in Yr. 3 was compared with the open
flow potential calculated using the Jones IPR model (with Yr. 3 input parameters). There was approximately
10% difference between the well test open flow potential and IPR calculated value, so it was deemed that
the Jones IPR model be sufficient at modelling well inflow (Table 4).
10 SPE-185485-MS

Table 4—Comparison of AOFP for well A and well B

Parameter Value Units Percentage Difference

TOFP well A Yr. 3 well test: 540000 Nm3/d -

TOFP well B Yr. 3 well test: 770000 Nm3/d -

TOFP well A Jones IPR Model 480000 Nm3/d -11.1

TOFP well B Jones IPR Model 850000 Nm3/d 9.4

MBAL
A new MBAL model was created using the same PVT data set within Prosper, to ensure consistency. The
GIIP and transmissibility between Tank A and Tank B were inherited from the operator MBAL model, as
the operator had history matched to the latest production data available.

GAP
GAP is a multiphase oil and gas optimiser tool that is used to model the surface gathering network of
field production systems (Petroleum Experts, 2016). GAP can be linked with well models from Prosper
and reservoir models from MBAL to achieve a full field production optimisation and forecast (Petroleum
Experts, 2016).
Schedule. The GAP model provided by the operator included a schedule for the production life of both
well A and well B. It was decided that the schedule be maintained and used within the new GAP model
created, as the schedule included historical shut-ins from well tests conducted during the life cycle of both
well A and well B. DP controls were also found within the schedule, and were used within the new model,
as they included some level of history-matching to the wells' actual production data.

History Matching
The new GAP model, containing the new Prosper models generated, was history matched to field production
data spanning from the 29th of April Yr. 1 to the 11th of June Yr. 17. The model generated was found to be
more productive than reality, with well A ceasing to flow in Yr. 11, as opposed to Yr. 9. For the purpose of
history matching the model, well A was artificially turned off in Yr. 9, and the velocity string for well A was
also turned off, using the GAP scheduling. The purpose of this was to history match the actual field output.

Production Rate Comparison

The GAP predicted production rate for well A is lower than the actual production data rate at the beginning
of production in Yr. 1 (Figure 5). This could be due to the Jones IPR model describing the inflow into the
well. Within GAP the Jones IPR model is transformed to a C and n model, with both the C and n variables
remaining constant with time. The permeability of well A is changing with time, and this is not captured
within the constant C and n IPR model. The permeability of the reservoir in Yr. 1 was 13 mD, however
the Jones model used a value of 5 mD. This value of 5 mD was selected on the basis of Yr. 3 well test
data which determined the permeability of well A to be 4-5 mD. A sensitivity investigation within Prosper
proved that the perforated interval also has a significant impact on the gas production rate. There was a
level of uncertainty in defining the perforated interval from the completion schematics provided by the
operator, and for well A the perforated interval in the Jones IPR model was inherited from the operator
model. Discrepancies between the predicted and actual rates are therefore best explained by, the uncertainty
surrounding the perforated interval defined in the Jones IPR model, the permeability value defined in the
Jones IPR model, and the transformation of the Jones IPR model to a C and n model within GAP (which
assumes constant C and n parameters).
SPE-185485-MS 11

Figure 5—Well A gas rate history match results

For well B the initial production rate predicted by the GAP model is higher than the actual production rate,
and this is most likely due to the interpretation of the perforated interval from the completions schematics.
The perforated interval for well B was a major parameter changed from that found in the operator GAP
model. The over prediction of the gas rate during initial production is offset by the under prediction of the
gas rate between Yr. 2 and Yr. 8 (Figure 6). The model provides a good match between Yr. 9 and Yr. 17,
with the model successfully predicting well B velocity string to be a success.

Figure 6—Well B gas rate history match results

Cumulative Production Comparison

Figure 7 shows how the cumulative production predicted for well A and well B, falls within 10% of the
production data cumulative.
12 SPE-185485-MS

Figure 7—Cumulative production history match, with 10% error bars

Reservoir Pressure Comparison

GAP generated reservoir pressures were compared with P/Z reservoir pressures. The P/Z reservoir pressures
consist of reservoir pressures derived from shut in wellhead pressures. The predicted reservoir pressures
from GAP fall within the 15 percent error bars for both well A and well B (Figure 8 and Figure 9). With
production rates, cumulative production and reservoir pressure history matched for both well A and well B,
various scenarios could be run within GAP, to determine the impact of running velocity strings.

Figure 8—Well A Reservoir Pressure History Match, with 15% error bars

Figure 9—Well B Reservoir Pressure History Match, with 15% error bars
SPE-185485-MS 13

Results
Systems Analysis was performed within Prosper to determine the impact of foam treatment and velocity
string deployment, on the operating point. The intersection of the IPR and VLP curves corresponded to the
expected production rate of the system. Systems Analysis could be used to determine whether the well was
operating within the Turner region or not. Specific GAP simulations were then run, to determine the impact
of workover timing on the overall recoverable reserves for the field.

Prosper well A and well B Velocity String Comparison

In order to perform Systems Analysis the following steps were taken:
1. Identify the onset of liquid loading (erratic flow) using production data, (Yr. 8 for well A and Yr.
11 for well B)
2. Use available production data to determine the wellhead pressure at the time of loading for both wells
3. Determine the reservoir pressure at the time of liquid loading for both wells, and adjust the IPR models
Figure 10 shows the system plots for well A for the various completion types. In Yr. 8 the original
completions for well A, were flowing within the Turner region, allowing the well to load up (Figure 10 a).
The deployment of the 2 3/8" velocity string would lift the well out of the Turner region sufficiently, and
result in a reduced flow rate (Figure 10 b). It can be noted that the deployment of the 1" velocity string
(Figure 10 c) would lift the well even further away from the Turner region, so the well is less likely to load
up again as the reservoir pressure depletes further.

Figure 10—Systems Plots for well A original completions (a), 2 3/8'' velocity string (b), 1'' velocity string (c)

For well A both velocity strings are therefore effective at lifting the well out of the Turner region, and
returning stable flow. The gas flow rate is reduced by 34% when running the 2 3/8" velocity string and by
14 SPE-185485-MS

51% when running the 1" velocity string (Table 5). The Systems Plots for well B are shown in figure 11,
and once again confirm that in Yr. 11, well B was experiencing liquid loading, and was producing within
the Turner region. The deployment of the 2 3/8" velocity string brings well B out of the Turner region and
reduces the gas flow rate by 42% (Table 6).

Table 5—Systems Analysis input parameters determined from production data for well A

Input Parameters Value Units

Top Node Pressure 44 Barg

WGR 1 STB/MMscf

CGR 12.3 STB/MMscf

Reservoir Pressure 100 Barg

Operating Point

Original Completions 3.16 MMscf/d

2 3/8" Velocity String 2.07 MMscf/d

1" Velocity String 1.54 MMscf/d

Table 6—Systems Analysis input parameters determined from production data for well B

Input Parameters Value Units

Top Node Pressure 32 Barg

WGR 1 STB/MMscf

CGR 12.3 STB/MMscf

Reservoir Pressure 70 Barg

Operating Point - -

Original Completions 2.06 MMscf/d

2 3/8" Velocity String 1.19 MMscf/d

Figure 11—System plots for well B original completions (a), 2 3/8" velocity string (b)

Prosper Foam Treatment Modelling

Foaming agents failed to restart well A, and this section serves to investigate whether in theory foam lift
should have proved successful. Within Prosper ‘Foam Lift’ can be selected as an artificial lift method.
Tubular flow was selected as the flow type, as well A produces purely through the tubing. There are
limitations within Prosper for modelling foam lift however, as it is not possible to select the foam type,
SPE-185485-MS 15

and Prosper simply uses a default model for modelling the effect of the surfactant on the surface tension
and foam density. This default model is outlined in Petroleum Society Paper 2007-118 by Yang and Zhang.
Yang and Zhang (2007) used a pre-determined calibration curve of the surface tension and foam density
versus the surfactant concentration, to then determine the minimum critical velocity. A significant limitation
is that the model defined by Yang and Zhang (2007) which is used within Prosper, is for the use of foams to
dewater gas wells, and in the case of the Southern Gas Field, the reservoir liquid of concern is condensate
not water. The aim within Prosper was to determine the optimum surfactant mass percentage that would
decrease the critical gas flow rate.
Outside the Prosper limitations for modelling foam treatment, it should also be noted that traditional
foamers tend to be ineffective as the condensate-to-water ratio increases (Orta et al., 2007). Liquid
condensate can significantly impact the ability of conventional foamers to unload loaded wells, and Orta
et al. (2007) developed a novel foamer specifically designed to foam condensates. The foamer developed
by Orta et al. (2007) helped unload a gas well that had a condensate-to-water ratio of 2:1, however the
Southern Gas Field has a much higher condensate-to-water ratio (12.3:1). The CGR of the Southern Gas
Field therefore appears to be sufficiently high to render foam treatments ineffective, unless an alternative
novel foam is identified, which can be used for high condensate-to-water wells.
From Figure 12 it is evident that foam lift is having negligible effect on the gas flow rate, and is therefore
not acting as a deliquification solution. This result is expected, as well A has a condensate-to-water ratio of
12.3:1. Had the Southern Gas Field had a high WGR, Prosper would have shown a significant improvement
in gas velocity upon the use of foams.

Figure 12—Systems plot showing well A 2 3/8" velocity string completions flowing within the turner region with a surfactant
mass percentage of 0.5 (a) and the sensitivity plot highlighting the negligible effect foams are having on the flow rate (b)

GAP Simulations
The history matched GAP model was used to run various scenarios to assess the impact of workover timing
on the recovery factor for the Southern Gas Field. Within the GAP model well A velocity string was now
turned on in the schedule, to determine the contribution of both well A velocity string and well B velocity
string to the total recovery. The permeability was set to 2 mD in the Jones IPR model for well A velocity
string, to attempt to capture the effect of condensate banking and/or skin as a result of the workover, which
resulted in a reduced inflow to well A.
The various scenarios run were as follows:
16 SPE-185485-MS

• Scenario 1: no velocity strings run, so original completions run until production ceases

• Scenario 2: both well A and well B velocity string run on the original schedule used by the operator

• Scenario 3: both well A and well B velocity string run when original completion wells cease to flow

• Scenario 4: well A velocity string never run, well B velocity string run when original completions
well B ceases to flow
• Scenario 5: both velocity strings run earlier in time, at initial signs of loading issues (Yr. 8 for well
A and Yr.11 for well B).
Within all scenarios except scenario 5 well A was turned off in the GAP schedule in Yr. 9, as denoted by*
The results of each simulation are tabulated below:
Table 7 shows that without the deployment of velocity strings the field recovery factor is expected to be
around 80%. Velocity strings are shown to add 4.6% (Table 10) to 5.4% (Table 9) to the total recovery factor
of the field, which amounts to a maximum additional reserve gain of 4.99 Bscf. Table 8 and Table 11 show
that the workover timing has no significant impact on the recovery factor. The significant finding however,
is that not running well A velocity string (Table 10), only drops the recovered reserves by 377 MMscf,
compared to the reserves recovered deploying both velocity strings in Yr. 13 as done by the operator.

Table 7—Scenario 1 GAP Simulation Results

Well Start Date End Date Cumulative Production (MMscf)

well A Prediction 29/04/Yr. 1 *10/06/Yr. 9 26428

well B Prediction 1 6/08/Yr. 1 24/12/Yr. 1 247250

Total Predicted Cumulative 29/04/Yr. 1 31/01/Yr. 13 73678

Predicted Recovery Factor - - 0.797

Table 8—Scenario 2 GAP Simulation Results

Well Start Date End Date Cumulative Production (MMscf)

well A Prediction 29/04/Yr. 1 *10/06/Yr. 9 26428

well B Prediction 16/08/Yr. 1 24/12/Yr. 1 247250

well A VS Prediction 26/01/Yr. 13 01/01/Yr. 22 584

well B VS Prediction 31/01/Yr. 13 01/01/Yr. 22 4056

Total Predicted Cumulative 29/04/Yr. 1 01/01/Yr. 22 78319

Predicted Recovery Factor - - 0.847

Table 9—Scenario 3 GAP Simulation Results

Well Start Date End Date Cumulative Production (MMscf)

well A Prediction 29/04/Yr. 1 *10/06/Yr. 9 26428

well B Prediction 16/08/Yr. 1 24/12/Yr. 12 47062

well A VS Prediction 11/06/Yr. 9 01/01/Yr. 22 1246

well B VS Prediction 25/12/Yr. 12 01/01/Yr. 22 3923

Total Predicted Cumulative 29/04/Yr. 1 01/01/Yr. 22 78659

Predicted Recovery Factor - - 0.851

SPE-185485-MS 17

Table 10—Scenario 4 GAP Simulation Results

Well Start Date End Date Cumulative Production (MMscf)

well A Prediction 29/04/Yr. 1 *10/06/Yr. 9 26428

well B Prediction 16/08/Yr. 1 24/12/Yr. 12 47254

well A VS Prediction - - -

well B VS Prediction 24/07/Yr. 12 01/01/Yr. 22 4260

Total Predicted Cumulative 29/04/Yr. 1 01/01/Yr. 22 77942

Predicted Recovery Factor - - 0.843

Table 11—Scenario 5 GAP Simulation Results

Well Start Date End Date Cumulative Production (MMscf)

well A Prediction 29/04/Yr. 1 01/10/Yr. 8 26159

well B Prediction 16/08/Yr. 1 15/10/Yr. 11 45745

well A VS Prediction 01/10/Yr. 8 01/01/Yr. 22 1692

well B VS Prediction 15/10/Yr. 11 01/01/Yr. 22 4848

Total Predicted Cumulative 29/04/Yr. 1 01/01/Yr. 22 78444

Predicted Recovery Factor - - 0.848

Figure 13 shows a comparison of Scenario 1 compared to Scenario 3 and 4, and highlights that velocity
string deployment will bring about rate stability, and prolong production. In Figure 13 there is an increase
in the average production rate in Yr. 16 and Yr. 18 for both Scenario 3 and 4, which is the result of the
separator pressure being reduced in the GAP schedule. This scheduling of the separator pressure reduction
was inherited from the GAP model provided by the operator. It is usual that the separator pressure be reduced
as the field continues to deplete as it reduces backpressure on the formation.

Figure 13—Combined well A and well B cumulative production, and average rates scenario comparison
18 SPE-185485-MS

Discussion
The Prosper foam lift model identified that batch foamer treatments require a low condensate-to-water ratio
to be successful, as condensates can actually act as a de-foamer. Foam lift was found to have a negligible
effect on the gas flow rate, with increasing surfactant concentration (Figure 12). Regardless of the limitations
for modelling foam treatment within Prosper, a review of literature identified that high condensate-to-water
ratios are undesirable. It is therefore of great importance that the load fluid be known, so that an appropriate
surfactant can be selected.
The new GAP model (including new Prosper well models) created for this investigation, was able to
predict the cumulative production for well A with original completions to within 6% of that actually
produced (Figure 7). The prediction for well B, with original completions was within 3% of that produced
(Figure 7). The new GAP model was therefore sufficiently history-matched to investigate workover
scenarios, with regards to total cumulative recovery. Due to the steady-state nature of the GAP model, the
model was unable to correctly predict the year in which well A ceased to flow, however still confirmed that
the cessation of flow would be attributed to loading and flow within the Turner region.
The Prosper models created, confirmed that velocity string intervention would prove successful at lifting
both well A and well B out of the Turner region. For well A, the 1" velocity string was able to lift the
well further from the Turner region, compared to the 2 3/8" velocity string. This was a consequence of the
2 3/8" velocity string terminating 400ft above the openhole completions, and the unloading ability being
significantly influenced by the larger diameter at the base of the well. The 1" velocity string completions
included a 1" section of coiled tubing run through the lower completion restrictions, thus bypassing the
larger diameter openhole section, which was limiting the wells ability to unload. It is therefore important to
note that when modelling a velocity string that terminates some distance above the perforated interval, the
larger diameter section at the base of the wellbore must be modelled as well, as the larger diameter section
is highly influential to the unloading ability, as identified by Coleman et al. (1991).
Upon running Scenario 1 with the new GAP model, it was found that without velocity string intervention,
production from the field would cease by late Yr. 12, as a result of liquid loading. The GAP predicted
recovery factor without well workovers amounted to 79.7%. Velocity string deployment was seen to extend
field life (Scenario 2 to 5), and this was a direct consequence of the velocity strings being able to lift the
wells out of the Turner region, returning stable flow. Velocity string intervention also resulted in an increase
in the recovery factor, with Scenario 3 giving a recovery of 85.1 %. Well workovers could therefore result
in an additional recovery of 5.4%. An important point to note is the relative contribution of well A velocity
string compared to well B velocity string, with the recovery from well B velocity string far outweighing that
from well A. The contribution from well A velocity string is marginal in some cases, and upon comparing
Scenario 2 (operator workover timings), with Scenario 4 (no well A workover), the additional well A
workover only amounts to 0.4% (377 MMscf) increased recovery, thus rendering it uneconomical.
A sensitivity analysis was conducted and reservoir permeability was found to have a significant impact
on the gas flow rate and well inflow performance. The operator concluded that the failure of well A velocity
string was a possible result of condensate banking and reduced relative gas permeability, which would have
resulted in a reduced well inflow. The Jones IPR model for well A velocity string used a permeability of
2mD to attempt to reflect a reduced inflow into the well. The process of condensate banking and reduced gas
relative permeability is a transient process, which is not captured by purely switching well completions from
original completions to velocity string completions within the GAP model. Solomon et al. (2008) concluded
that within the industry there is a poor understanding of the complex phenomena associated with liquid
loading, especially with respect to the dynamic interaction between reservoir and wellbore. Both reservoir
kh and CGR were both decreasing with time. This transient effect is not captured within the GAP model,
and is most likely the reason for well A ceasing to flow in Yr. 11 as opposed to Yr. 9.
SPE-185485-MS 19

Velocity string deployment could add a maximum of 4.99 Bscf to the recovered reserves (Table 9). Well
A velocity string performs poorly compared to well B velocity string, and Table 10 identifies how a recovery
factor of 84.3% can be achieved without the need to workover well A. Due to the marginal nature of well
A velocity string, a simple economic analysis is conducted below, to determine the minimum gas return
required by a velocity string in order to pay off the workover costs.

Economics
Using a gas price of 6 £/MMBTU, and assuming workover costs of £2.15m per well (Wood Mackenzie,
2012), and the units conversion 1030 BTU/ft3 suggests that 344 MMscf would need to be produced from
a workover for a single well, to pay back the workover costs. Table 8 shows that well A velocity string
only produces 584 MMscf, meaning the economic gains of running well A velocity string on the operator's
schedule would be marginal.

Conclusions
• Without the deployment of velocity strings, field production would have ceased by Yr. 12,
providing a final recovery of 79.7% as identified by Scenario 1.
• Installing velocity strings to workover well A and well B, would extend field life (wells still flowing
at end of prediction in Yr. 22), thus resulting in a maximum additional recovery of 5.4%.
• The 2 3/8" velocity string run in well A would successfully unload the well, however the unloading
performance is hindered by the velocity string terminating 400ft short of the perforated joint. The
larger diameter section at the base of the wellbore could become more problematic as the reservoir
pressure declines, and we could possibly see the well re-enter the Turner region.
• A 1" velocity string run in well A would prove more successful at lifting the well out of the Turner
region, however the operator decided not to install the 1" velocity string, as the technical challenges
and operational risks outweighed the benefits.
• The identical 2 3/8" velocity string installed in well B, provided the majority of the additional
reserves post-workover in all Scenarios 2 to 5. In practice the operator found well B velocity string
installation to be a success, and this is captured in the new GAP model. The GAP predictions are
therefore representative of reality.
• This investigation highlighted that well A velocity string gives marginal additional recovery, and it
is believed that the failure of the workover as found by the operator can be attributed to condensate
banking and skin associated with the change in completions.
• Sensitivity investigations identified permeability and perforation interval to have the most
significant impact on the gas flow rate, and these parameters were deemed to have the highest level
of uncertainty within the GAP model.
• The most economic strategy would have been to only workover well B, and to shut-in well A in
Yr. 9. This is due to the marginal nature of well A velocity string in all scenarios run, as a result
of the reduced inflow.
• In order to select an appropriate surfactant for batch foam deliquification, the full composition of
the load fluid must be known, as water and liquid hydrocarbon react differently to certain foams.

Recommendations and future work

Solomon et al. (2008) identified how loading is a transient process, and there is a need for a fully transient
multiphase flow approach when modelling load up, and this cannot be achieved using GAP, MBAL and
Prosper as they are all forms of steady-state software. Future work could look at trying to incorporate a
‘transient’ element to the GAP model, by adding further wells to represent the changing properties over
20 SPE-185485-MS

time. Multiple Prosper models could be created for the original completion wells, each with unique Jones
IPR properties to attempt to capture the evolution of reservoir properties such as CGR and permeability
over time. Additionally multiple MBAL tanks could be created, to represent the change in relative gas
permeability through time. The major limitation with regards to future work is the availability of field data,
to fully capture the changing properties overtime. Transient modelling software such as the OLGA dynamic
multiphase flow simulator could be investigated to determine their applicability to the Southern Gas Field.
Numerical reservoir simulations (radial flow model) could be used to attempt to model the effects of
condensate banking over time on the Southern Gas Field reservoir. Effects of liquid dropout rate and gas
relative permeability on gas production (well productivity) could then be investigated, and provide an insight
into whether condensate banking could have provided a sufficient reduction in well productivity, such as
to explain the cessation of flow in well A.

Nomenclature
A = tubing cross-sectional area, ft2
CGR = condensate gas ratio, stb/MMscf
d = tubing internal diameter, ft
DP = pressure gradient
krg = relative gas permeability, mD
krw = relative water permeability, mD
MD = measured depth, ft
MMscf = 106standard cubic feet
OD = outer diameter, inches
OPEX = operating expenditure
p = surface pressure, psia
PVT = pressure volume temperatur
qgc = critical gas flow rate, MMscf/day
T = fluid temperature, °R
TVD = true vertical depth, feet
νt = critical velocity, ft/s
WGR = water to gas ratio, stb/MMscf
z = gas compressibility factor, dimensionless
ρg = gas density, lbm/ft3
ρl = liquid density, lbm/ft3
σ = liquid/gas surface tension, dynes/cm

References
1. Campbell, S., Ramachandran, S., Bartrip, K.: "Corrosion Inhibition/Foamer Combination
Treatment to Enhance Gas Production", SPE-67325-MS presented at the 2001 Production and
Operations Symposium. Oklahoma City, Oklahoma, Mar. 24-27.
2. Coleman, S.B., Clay, H.B., McCurdy, D.G. and Norris III, H. L.: "A New Look at Predicting
Gas-Well Load-Up", Journal of Petroleum Technology (March 1991), 329.
3. De Jonge, R. M., & Tousis, U. A. M: "Liquid Unloading of Depleted Gas Wells in the North
Sea and Continental Europe, Using Coiled Tubing, Jointed Pipe Velocity/Insert Strings, and
Microstrings". SPE 107048-MS (January 2007). Europec/EAGE Conference and Exhibition, held
in London, United Kingdom, 11-14 June 2007
SPE-185485-MS 21

4. Fadipe, S. 2012. Candidate Selection and Assessment of the benefits of velocity strings on the
Durban field. MSc thesis, Imperial College, London, (September 2012).
5. Guo, B., Ghalambor, A. and Xu, C.: "A Systematic Approach to Predicting Liquid Loading
in Gas Wells", SPE-94081-MS presented at 2005 Production and Operations Symposium.
Oklahoma City, Oklahoma, U.S.A., Apr. 17-19.
6. Hearn, W.: " Gas Well Deliquification," paper SPE 138672 presented at the 2010 Abu Dhabi
International Petroleum Exhibition and Conference held in Abu Dhabi, UAE, 1-4 November
2010.
7. Lea, J.F., Nickens, H.V., Wells, M.R.: Gas Well Deliquification, Second Edition, Gulf
Professional Publishing, Jordan Hill, Oxford, UK (2008).
8. Li, H., Yang, D., & Zhang, Q. "A Theoretical Model for Optimizing Surfactant Usage in a Gas
Well Dewatering Process". Petroleum Society of Canada. PAPER 2007-118 presented at the
Petroleum Society's 8th Canadian International Petroleum Conference. Calgary, Alberta, Canada,
June. 12–14
9. Libson, T.N., Henry, J.R.: "Case Histories: Identification of and Remedial Action for Liquid
Loading in Gas Wells – Intermediate Shelf Gas Play", 1980 JPT 32 (4): 685-693. SPE-7467-PA
10. Orta, D. R., Fosdick, M. L., Yang, J., Ramachandran, S., Salma, T., Long, J. J., Salinas, O.: "A
Novel Foamer for Deliquification of Condensate-Loaded Wells", SPE 107980 presented at the
2007 SPE Rocky Mountain Oil and Gas Technology Symposium held in Denver, Colorado,
U.S.A., 16-18 April 2007
11. Oudeman, P.: "On the Flow Performance of Velocity Strings To Unload Wet Gas Wells," SPE
104605 presented at the 2007 SPE Middle East Oil & Gas Show and Conference held in Bahrain
International Exhibition Centre, Kingdom of Bahrain, 11-14 March 2007.
12. Petroleum Experts. 2016. GAP, http://www.petex.com/products/?ssi=5 (accessed 20 July 2016).
13. Petroleum Experts. 2016. IPM Prosper User Manual, Version 14, 663. Edinburgh: Petroleum
Experts Limited.
14. Schinagl, W. and Denny, M.: "Securing the Future in Mature Gas Fields", SPE 107467 presented
at the SPE Annual Conference and Exhibition held in London, 11-14 June 2007.
15. Solomon, F.A., Falcone, G., and Teodoriu, C.: "Critical Review of Existing Solutions to Predict
and Model Liquid Loading in Gas Wells". Presented at the SPE Annual Technical and Exhibition,
Denver, Colorado, 21-24 September. SPE-115933.
16. Sutton, R.P., Cox, S.A., Lea, J.F., Rowlan, O.L.: "Guidelines for the Proper Application of
Critical Velocity Calculations", 2010 SPE Prod & Oper 25 (2): 182-194. SPE-120625-PA.
17. 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", Journal of Petroleum Technology
(November 1969), Trans., AIME, 246, 1475.
18. Vogelij, N. A., Veeken, C. A., Islamov, R. I., Lugtmeier, B., Ros, O., De Vries, G., & Zhidkova,
N. A.: "Four Years of Continuous Foamer Application in Southern North Sea Offshore Gas
Wells". SPE-180031-MS presented at the SPE Bergen One Day Seminar. Bergen, Norway, April.
20
19. Vosika, J.L.: "Use of Foaming Agents to Alleviate Liquid Loading in Greater Green River TFG
Wells", SPE/DOE-11644-MS presented at the 1983 Symposium on Low Permeability. Denver,
Colorado, Mar. 14-16.
20. Wood Mackenzie (2012), Asset Analysis, Europe (UK) Upstream Service, August 2012
21. Wylde, J.J.: "Increasing Lazy Gas Well Production: A Field Wide Case History in Northern
Alberta", SPE-134143-MS presented at 2010 Canadian Unconventional Resources &
International Petroleum Conference. Alberta, Canada, Oct. 19-21.
22 SPE-185485-MS

22. Zhou, D., Yuan, H.: "A New Model for Predicting Gas-Well Liquid Loading", 2010 SPE Prod &
Oper 25 (2): 172-181. SPE-120580-MS.