SPE-196201-MS

Improved Data Mining for Production Diagnosis of Gas Wells with Plunger
Lift through Dynamic Simulations

Jianjun Zhu, University of Tulsa; Guangqiang Cao and Wei Tian, PetroChina Company Ltd.; Qingqi Zhao and
Haiwen Zhu, University of Tulsa; Jie Song, PetroChina Company Ltd.; Jianlin Peng, Zimo Lin, and Hong-Quan
Zhang, University of Tulsa

Copyright 2019, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Annual Technical Conference and Exhibition held in Calgary, Alberta, Canada, 30 Sep - 2 October 2019.

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
Plunger lift has been widely used in unconventional gas wells to remove liquid accumulation from the well..
Production surveillance provides large amount of data of production process and normal and abnormal
operations, which can be used in machine learning (ML) and Artificial Intelligence (AI) to develop
algorithms for anomaly diagnosis and operation optimization. However, in the surveillance data the majority
is related to daily operation and the data of failure cases are rare. Also the failure cases may not be repeatable
and many failure case signatures are not available until they happen. Large data size of anomaly cases
are needed to improve the ML model accuracy. Dynamic simulation of the plunger lift process offers an
alternative way to generate synthetic data on the specified anomalies to be used to train the ML model. It
also helps better understand the trends reflected in the surveillance data and their root causes.
From the available surveillance data of gas wells equipped with plunger lift, the simultaneous
measurements of different parameters at different points in a production system with normal and abnormal
occurrences can be analyzed and the correspondent trends/signatures can be identified. The typical
signatures that conform to pre-determined anomalous patterns can be obtained. Using a commercial transient
multiphase flow simulator, the actual field data of tubing/casing pressures can be matched through a tuning
process. Trial-and-error is needed to improve the dynamic plunger lift model so that a good agreement
with the production data can be achieved by adjusting the reservoir performance, plunger parameters or
surface pipeline boundary conditions. Following the validation under different flow conditions, synthetic
datasets for various operational and flow conditions can be generated by performing parametric studies.
Unlike the field data, the synthetic data from the dynamic simulations mainly comprise anomaly signatures
(e.g. tubing rupture, missed arrival of plunger, etc.), which can be added to the ML data pool to reduce the
data covariance and increase independency.

Introduction
As a gas reservoir depletes, the reservoir energy and bottom-hole pressures decrease gradually, resulting in a
lower gas production rate that is insufficient to carry the produced liquids to the surface (Zhang et al., 2018;
2 SPE-196201-MS

Zhu et al., 2019). If the gas flow rate is unable to remove these liquids from the well, they can accumulate and
buildup downhole, this adds backpressure to the decreased bottom-hole pressure and reduces the production
of gas further. Worse still, a vicious circle can be generated, causing intermittent or segregated flow patterns,
even lower gas production rates and eventually killing the well. Also, the accumulation of oil and/or water
at the bottom is responsible for many flow assurance issues in pipes, such as corrosion and terrain severe
slugging (Fan et al., 2018 and 2019), which further induces unstable production and potential failures.
The plunger lift system as shown in Figure 1 above has been applied to gas wells for liquid deliquification,
which is an intermittent artificial lift method that uses only the reservoir energy to produce the hydrocarbon
fluids (Lea and Nickens, 2011) from liquid-loaded gas wells. Plunger lift operates in a cyclic process with
the well alternately flowing and shut-in (Fu, 2016). It uses a free piston that travels up and down in the
tubing, which minimizes the liquid fallback and uses the reservoir energy more efficiently than slug or
bubble flow does. Compared to other artificial lift methods, plunger lift is able to remove liquids from the
wellbore so that the well can be produced at a lower bottom-hole pressure (Zhu et al., 2019).

Figure 1—Schematic of a typical plunger lift installation (figure courtesy petrowiki.org)

Conventional operation in the plunger lift system involves the cyclic processes that actuate the surface
control/motor valve to open or close intermittently. When the valve is closed, the plunger falls down from the
lubricator into the tubing to the bumper spring due to gravity, first through gas and then some accumulated
liquid (Lea and Nickens, 2011). When the motor valve is opened, the pressure difference across the plunger
causes it to rise to the surface, along with the liquids accumulated at the well bottom (Gupta et al., 2017;
Nandola et al., 2018). As the microprocessors and electronic controllers have advanced (Mower et al., 1985;
Morrow and Rogers, 1992), the plunger lift has gained a much wider use and broader applications in the
field. Also, the controllers have evolved from simple On and Off control to auto-adjust pressure control
and later the automated systems, aided by computers to sense the potential failures of plunger and make the
immediate adjustments (Clegg and Lake, 2006).
Success in plunger lift systems depends on proper candidate identification, proper well installation, and
the effectiveness of the operator, all of which may contribute to the plunger problems if inappropriately
treated. Lea and Nickens (2011) outlined the solutions to some of the more common problems encountered
with plunger lift systems, and grouped the issues with respect to the system components and particular
SPE-196201-MS 3

malfunctions. Table 1 below presents a brief summary of the potential failures in a conventional plunger lift
system. Besides the issues listed in the table, some other flow assurance issues such as hydrate formation
(Fu et al., 2019, Sun et al., 2019), sand or scale (Zhu et al., 2018) and paraffin (Chi et al., 2019) may
present further potential failure to the plunger lift systems, specifically the motor valve, tubing and plunger
components etc.

Table 1—Summary of the plunger lift failures

Several analytical models have been proposed in literature for prediction and optimization of plunger
lift productions, based on which the trouble-shooting and failure diagnosis can be conducted as well. Foss
and Gaul (1965) reported over 100 wells that used plunger lift in the Ventura Avenue Field operated by
Shell.9?). A simplified static model of plunger lift based on those field cases was proposed. Lea (1982)
proposed a dynamic model of plunger lift based on the force balance equation on the plunger, by solving
the plunger position, velocity, acceleration and pressure that could be obtained. Marcano and Chacín (1994)
proposed a new mechanistic model for predicting the conventional plunger lift process by incorporating
the liquid fallback during plunger rise as a linear function of average velocity. Charcin (1990) conducted
mathematical modeling and optimization of intermittent gas-plunger lift by combining the plunger lift
process and reservoir inflow performance. Gasbarri and Wiggins (2001) proposed a dynamic plunger lift
model that comprises four components in a complete production cycle, namely, upstroke, blowdown, build-
up and reservoir performance. A more recent dynamic model of plunger lift was proposed by Gupta et
al. (2017), based on which the online optimization for a plunger lift process in shale gas wells has been
conducted by Nandola et al. (2018).
In addition to the analytical models, the system design and trouble-shooting of plunger lift have been
assisted by the dynamic simulation and data-driven techniques. Especially, the field data such as flow rate,
4 SPE-196201-MS

pressure and temperature can be analyzed and interpreted by machine learning algorithms (Tian and Horne,
2015 and 2019) so that the reconstruction of historical production data becomes accessible. Nguyen and Del
Mundo (2016) used OLGA to evaluate the transient flow and the dynamic behavior of fluids with changes in
pressure, temperature, and composition, which helped to better understand the optimal wellbore conditions
for maximum plunger performance, optimized plunger lift cycles, and allowed engineers to develop a
consistent plunger lift strategy across gas fields. Singh (2017) presented a practical application of the
data-driven technique (CART, classification-and-regression-tree) as a method of root-cause identification
and production diagnostic tool. Case studies from the real field data for gas wells using the plunger-lift
system verify its superiority to the first-principles-based model by outlining the dominant variables that
potentially affect gas-production rates. Kamari et al. (2017) applied the least-squares supported-vector
machine (LSSVM) to predict the maximum possible liquid-production rate by use of plunger lift. Combined
with the coupled-simulated-annealing (CSA) optimization, the proposed model (LSSVM-CSA) showed a
novel feature selection mechanism on tuning the optimal parameters in plunger lift system.
Although many successful applications of dynamic models and data-driven techniques are available in
gas wells with plunger lift installed, the combined solution to better identify the examples of a variety
of failures and abnormal scenarios is rarely receives adequate attention. In this study, starting from the
available surveillance data, the simultaneous measurements of different parameters at different points in
a production system with known abnormal occurrences can be analyzed and the correspondent trends
and signatures can be identified. From this example signatures that conform to pre-determined anomalous
patterns can be provided. Meanwhile, the mechanistic models or dynamic simulators for multiphase flow
in wells and artificial lift equipment can be used to simulate the parameter trends in the system under
different flow conditions. Assuming an abnormal occurrence, the parameter trends at different locations can
be characterized. Specific parameters will be applied to the mechanistic models or dynamic simulators in
order to simulate anomalous patterns and failure events as described above. Simulated data based on these
input parameters can be generated. The resulting synthetic datasets are used to further train and refine the
machine learning models, with the goal of realizing improvements to better diagnose and predict potential
issues and failures of the plunger lift system.

Field Information
The natural gas well studied in this paper is a tight gas reservoir, located in Changqing oilfield inChina.
Its original reservoir pressure is approximately 4,500 psig (30 MPa). The inflow performance curve can be
estimated by the inflow performance relationship (IPR) as below (Li et al., 2016):

(1)

where qAOF is the absolute open flow rate, pD is the dimensionless draw-down at the wellbore. The phase
envelop generated by PVTsim is shown in Figure 2. Since the methane occupies more than 93% of the
mole fraction, the phase envelope is shifted to the left-hand-side on the Pressure Temperature (PT) plot,
indicating that the gas phase dominates during the plunger lift production.
SPE-196201-MS 5

Figure 2—Phase envelop of the studied natural gas

The gas well is a horizontal well, as shown in Figure 3, with a total 8 stages of perforations. The total
measured depth (MD) is 16,218 ft, (Should you show metric units as well?) and the true vertical depth
(TVD) is 11,783 ft. The tubing has two diameters. The inner diameter of tubing vertical section is 3.5″, the
remaining tubing diameter is 2.875″. The inner diameter of casing is 7″.

Figure 3—Schematic of the gas well trajectory (Very good)

Although the studied gas well produces no oil condensate, it produces water which significantly reduces
the gas production due to its accumulation. Thus, the plunger lift system is necessary to be deployed for
deliquification. The plunger used in this well is a self-buffering piston that automatically lowers its falling
velocity at the tubing joint with different cross-section areas due to the impact effect. The basic geometry
of the plunger is 1.4 ft long, 2.32″ outer diameter, and 10 lb in weight.
The well has be stably produced for more than 3 years assisted by plunger lift. Figure 4 below
demonstrates the historical production data in terms of tubing (Pt) and casing pressures (Pc) within one
month. An obvious cyclic production feature is observed in both Pt and Pc plots as the plunger moves up
6 SPE-196201-MS

and down in the tubing. Since the liquid accumulates at bottom, the overall pressure difference of Pt and
Pc is more than 200 psi. Due to the non-disclosure agreement with the field company, the production data
natural gas flow rate is not presented in this paper.

Figure 4—Casing and tubing pressures in one month plot

Figure 5 shows the historical tubing pressure (Figure 5a) and casing pressure (Figure 5b) of one plunger
lift cycle. The data are stacked over multiple cycles to better represent the trends. As can be seen, the
production of plunger lift process can be decomposed into multiple cycles with each less than 1,000 minutes
(16.7 hours). Almost all cycles are congruent or similar to each other, although some cycles are inconsistent
with the majority due to the unstable production or abnormal operation of plunger lift. For every plunger lift
cycle, the shut-in time is at 200 (3.33 hours) min after the motor valve opens, and the re-open time interval
is approximately 1,000 min.

Figure 5—Tubing and casing pressures in one plunger lift cycle, (a) tubing pressure, (b) casing pressure

Dynamic Model Simulation

An OLGA model was built to match the important parameters of the typical plunger cycles of a conventional
plunger lift system. Figure 6(a) shows the well trajectory based on the survey data and Figure 6(b)
corresponds to the well details, including the inner and outer diameters of tubing and casing, as well as the
SPE-196201-MS 7

reservoir position etc. It should be noted that a standing valve is positioned at the end of the vertical section
of the tubing to prohibit the plunger falling back into the wellbore or annulus. In reality, the plunger will sit
on top of the bumper spring after it falls back from the well head.

Figure 6—Well trajectory and geometries, (a) survey data, (b) well geometries

To better match the field reality, the surface line and control valves along with the control schematics
should be provided in the dynamic simulation. As shown in Figure 7, both the motor valve and plunger are
operated by the manual control in OLGA. For motor valve control, the time-series data in terms of open and
close versus varying time are given in valve control configuration. As for the plunger, the multi-launches are
required to simulate multiple plunger lift cycles. Other time-series data regarding the launch time interval
are initialized. Furthermore, the valve control and plunger control schematics should conform to perform
successful simulations.
8 SPE-196201-MS

Figure 7—Schematic representation of the dynamic plunger lift model in OLGA

The downstream surface line, valve and separator are also required, although the detailed field
information of surface infrastructure is unavailable. Assuming the necessary components based on the
boundary condition, the length of surface line, and the opening of surface valve can be determined. A trial
and error process should be conducted to reach agreement of simulation results against the field production
data.
The total simulation time is more than 8 plunger lift cycles until the last several cycles achieve stable
production. A comparison can be performed by plotting the simulated casing and tubing pressure by the
OLGA dynamic simulator in single plunger lift cycle, as shown in Figure 8. The blue scattered points
correspond to field data, and the red ones are from OLGA simulation results. A good match can be reached
as long as the plunger lift model in OLGA simulator has been properly tuned and configured. Thus the
authors are confident to conduct the dataset generation using the dynamic simulator, which can produce the
adequate normal and abnormal cases in plunger lift operations easily.

Figure 8—Comparison of simulated casing and tubing pressure against the field data, (a) tubing pressure, (b) casing pressure

For multiple plunger lift cycles, the comparison of OLGA simulated casing and tubing pressure with
the corresponding field data can be found in Figure 9 below. A first investigation implies that OLGA can
SPE-196201-MS 9

simulate the plunger lift cycles with excellent agreement with field data in both trends and amplitudes.
Although the mismatch can be detected in the simulated frequency of plunger lift cycle, it can be ascribed to
the arbitrary manual control scheme in OLGA simulation. In OLGA, a uniform open and close time interval
was assigned in the manual control of motor valve. However, the motor valve is usually automated by the
SCADA (Supervisory Control and Data Acquisition) system, which outputs an inconsistent time-interval
series for the motor valve opening and closing. Thus, it is expected to see the mismatch of particular periods
of plunger lift cycles.

Figure 9—Comparison of multiple plunger lift cycles with OLGA simulated casing
and tubing pressure again field data, (a) tubing pressure, (b) casing pressure

Synthetic Data for Abnormal Cases

In this section, we present several examples of plunger lift anomalies, such as liquid loading, tubing rupture,
stuck motor valve and line pressure increase etc. All these abnormal cases have been encountered, although
from different gas wells.

Liquid Loading
For gas wells, the continuous depletion of reservoir reduces the reservoir pressure and gas flow rate, which
in turn makes the gas unable to carry the liquids to the surface. The accumulated liquid in the well adds the
hydrostatic pressure and decreases the drawdown pressure. When the liquid level in a gas well increases,
there is a significant departure of tubing pressure and casing pressure which can be obvious, as shown in
Figure (10). Initially, the gas well suffers from moderate liquid loading as the casing pressure is above the
tubing pressure during the motor valve closing periods. After a sudden increase of WGR (water gas ratio) in
the bottom-hole, the pressure difference at tubing and casing surface becomes increasingly large, indicating
that severe liquid loading occurs.
10 SPE-196201-MS

Figure 10—Simulated severe liquid loading effect in a gas well assisted by plunger lift

Tubing Rupture
The tubing in gas wells suffers from rupture due to scaling, erosion or corrosion (Zhang et al., 2019; Wang
et al., 2019). When it occurs, the plunger lift system performs differently in terms of tubing and casing
pressure. As shown in Figure 11, the first 5 plunger lift cycles with/without tubing rupture have been plotted,
where the blue curves demonstrate the plunger cycles with and withoutt tubing rupture, and the yellow ones
correspond to the anomaly induced by the bottom tubing rupture. The rupture was assumed to be located at
the end of tubing vertical section with the rupture size of 0.2 inch. Clearly, the difference of tubing and casing
pressure during the shut-in periods of gas wells is minimized once the tubing has been cracked or penetrated.
Moreover, it is observed that the tubing rupture tends to decrease the casing pressure and increase the tubing
pressure. It is expected to see this behavior as the rupture allows more fluids to flow from casing to tubing.

Figure 11—Simulated tubing rupture effect in a gas well assisted by plunger lift

Figures 12 and 13 below show the parametric studies of rupture size and position effects on plunger lift
cycles. Three different rupture sizes (0.2, 0.4 and 0.6 inch) and positions (tubing top, middle and bottom)
are included in the calculations. As can be seen in Figure 12, the rupture size effect seems to be stable as
the size of rupture increases since the casing and tubing pressures for 0.4 and 0.6 inch of rupture sizes are
almost identical. It can be inferred that the tubing rupture affects the plunger lift system in terms of tubing
and casing pressures at the very beginning of rupture growth. Other parameters, such as production rates of
gas/liquid are necessary for comparison, based on which the rupture sizes can be distinguishable.
SPE-196201-MS 11

Figure 12—Effect of tubing rupture size in a gas well assisted by plunger lift, (a) tubing pressure, (b) casing pressure

Figure 13—Effect of tubing rupture position in a gas well assisted by plunger lift, (a) tubing pressure, (b) casing pressure

As to the tubing rupture position, it is possible that the rupture can grow anywhere along the tubing.
We look at three typical positions, namely the top, middle and bottom of the tubing vertical section. As
seen in Figure 13, the difference of casing and tubing pressure due to varying rapture position is trivial.
Thus, additional data are required to distinguish the anomalies in plunger lift due to tubing rupture positions.
Specifically, the production rates may provide the necessary information for machine learning algorithm to
better diagnose the anomaly details.

Stuck of Motor Valve

As discussed above, the motor valve being stuck may occur during its opening or closing, either of which
affects the plunger lift cycles significantly. Figure 14 shows the signature of tubing and casing pressures
in the plunger lift system when the motor valve suffers from being stuck (i.e. won't close or won't open).
Figure 14(a) demonstrates the plunger lift cycles with the motor valve always open. As seen, the production
continues as long as the motor valve remains open, leading to the tubing pressure being close to the line
pressure. As a boundary condition, the line pressure is set to be constant at 190 psia. Thus, the tubing
pressure drops to a valve and remains constant during the motor valve open. However, the casing pressure
slightly increases due to the reservoir fluid inflow. If the motor valve resume its functionality, the gas well
12 SPE-196201-MS

ceases production as it closes. The casing and tubing pressures build up quickly. The liquid loading may
occur since the difference of tubing and casing pressure increases further.

Figure 14—Effect of motor valve stuck in a gas well assisted by plunger lift, (a) won't close, (b) won't open

If the motor valve won't open, the gas well behaves differently. Since the well shut-in becomes longer,
the build-up of gas well continues. The tubing and casing pressures increases gradually before reaching a
stable valve. Once the valve opens again, the gas well gains sufficient energy to blow the gas and liquid
stored in tubing or annulus out. Then, the accumulated liquid can be unloaded and removed. Actually, the
continuous shut-in of gas well is practical in gas field to deliquify the liquid from the well bottom.

Line Pressure Change

The line pressure is related to the surface separator, which usually stays constant for the sake of stable
production and transportation of hydrocarbon fluids. However, it also fluctuates due to unexpected flow
assurance or safety issues. When line pressure changes, it significantly affects the plunger lift production
cycles. The effect of line pressure change should be well studied in order to better diagnose or predict the
plunger lift system anomaly.
Figure 15 displays the tubing and casing pressure trends after the surface line pressure increases or
decrease. The green curve represents the line pressure change, from which a peak trend is observed. The
change of line pressure has significant effect on plunger lift cycles. As seen, the departure of casing and
tubing pressure before line pressure change diminishes. The line pressure change also alters the plunger lift
production signature by increasing the casing pressure. After the change, the casing and tubing pressures
are almost identical during the shut-in periods, indicating that the accumulated liquid in gas well bottom is
unloaded. The production resumes without severe liquid accumulation in the bottom hole.
SPE-196201-MS 13

Figure 15—Simulated line pressure change effect in a gas well assisted by plunger lift

Conclusions
In this paper, we presented an improved data preparation technique using dynamic simulation to generate
the synthetic data, which can be coupled with the real field data to feed the machine learning algorithm
for anomaly detection and failure prediction in plunger lift systems. The methodology proposed in this
study starts from the surveillance data of gas wells with plunger lift installed. The normal and abnormal
occurrences of plunger lift production cycles are analyzed, based on which the correspondent trends and
signatures are identified. The typical signatures that conform to pre-determined anomalous patterns can be
obtained. Utilizing the commercial transient multiphase flow simulator (OLGA), the actual field data in
terms of tubing/casing pressures can be matched through a trial and error process. With the dynamic plunger
lift model tuned, different operational scenarios of plunger lift can be simulated, where the synthetic datasets
comprising plunger lift anomalies can be generated. Compared to the overwhelming normal operation
data of plunger lift, the synthetic data can be fed to the data pool for machine learning to reduce the
data covariance and increase independency, which improves the precision and accuracy of the data-mining
algorithms.

Acknowledgement
The authors wish to thank the Tulsa University Artificial Lift Projects (TUALP) members for their financial
and technical support.

References
Chi, Y., Yang, J., Sarica, C. and Daraboina, N., 2019. A Critical Review of Controlling Paraffin Deposition in Production
Lines using Chemicals. Energy & Fuels.
Clegg, J.D. and Lake, L.W., 2006. Petroleum Engineering Handbook: Production Operations. Society of Petroleum
Engineers.
Fan, Y., Pereyra, E. and Sarica, C., 2018. Onset of liquid-film reversal in upward-inclined pipes. SPE Journal.
Fan, Y., Pereyra, E., Sarica, C., Schleicher, E. and Hampel, U., 2019. Analysis of Flow Pattern Transition from Segregated
to Slug flow in Upward Inclined Pipes. International Journal of Multiphase Flow.
Foss, D.L., and Gaul, R.B. 1965. Plunger-life performance criteria with operating experience – Ventura Avenue Field.
Drilling and Production Practice. American Petroleum Institute. API-65-124.
Fu, W., 2016. Gas Well liquid loading onset and plunger lift modeling, Master thesis, The University of Tulsa.
Fu, W., Wang, Z.Y., Yue, X., Zhang, J. and Sun, B., 2019. Experimental Study of Methane Hydrate Formation in Water-
continuous Flow Loop. Energy & Fuels.
Gasbarri, S., and Wiggins, M.L. 2001. A dynamic plunger lift model for gas wells. SPE production & Facilities, 16(02)
89–96. SPE-72057-PA.
Gupta, A., Kaisare, N.S. and Nandola, N.N., 2017. Dynamic plunger lift model for deliquification of shale gas wells.
Computers & Chemical Engineering, 103, 81–90.
Kamari, A., Bahadori, A. and Mohammadi, A.H., 2017. Prediction of maximum possible liquid rates produced from
plunger lift by use of a rigorous modeling approach. SPE Production & Operations, 32(01), pp.7–11.
Lea, J.F. 1982. Dynamic analysis of plunger lift operations. Journal of Petroleum Technology, 34(11), 2617–2629.
14 SPE-196201-MS

Lea, J.F. and Nickens, H.V., 2011. Gas well deliquification. Gulf Professional Publishing.
Li, Y., Xiao, F., Xu, W. and Wang, J., 2016. Performance evaluation on water-producing gas wells based on gas & water
relative permeability curves: A case study of tight sandstone gas reservoirs in the Sulige gas field, Ordos Basin. Natural
Gas Industry B, 3(1), pp.52–58.
Marcano, L. and Chacín, J., 1994. Mechanistic design of conventional plunger lift installations. SPE Advanced Technology
Series, 2(01), 15–24.
Morrow Jr, S.J. and Rogers Jr, J.R., 1992. Increasing Production Using Microprocessors and Tracking Plunger-Lift
Velocity. In SPE Mid-Continent Gas Symposium. Society of Petroleum Engineers.
Mower, L.N., Lea, J.F., Beauregard, E. and Ferguson, P.L., 1985. Defining the characteristics and performance of gas-lift
plungers. In SPE Annual Technical Conference and Exhibition. Society of Petroleum Engineers.
Nandola, N.N., Kaisare, N.S. and Gupta, A., 2018. Online optimization for a plunger lift process in shale gas wells.
Computers & Chemical Engineering, 108, pp.89–97.
Nguyen, H.T. and Del Mundo, F.C., 2016, October. Improving artificial lift design through dynamic simulation. In SPE
North America Artificial Lift Conference and Exhibition. Society of Petroleum Engineers.
Singh, A., 2017. Application of data mining for quick root-cause identification and automated production diagnostic of
gas wells with plunger lift. SPE Production & Operations, 32(03), pp.279–293.
Sun, B., Fu, W., Wang, Z., Xu, J., Chen, L., Wang, J. and Zhang, J., 2019. Characterizing the rheology of methane hydrate
slurry in a horizontal water-continuous system. SPE Journal.
Tian, C. and Horne, R.N., 2015, September. Machine learning applied to multiwell test analysis and flow rate
reconstruction. In SPE Annual Technical Conference and Exhibition. Society of Petroleum Engineers.
Tian, C. and Horne, R., 2019. Applying Machine-Learning Techniques To Interpret Flow-Rate, Pressure, and Temperature
Data From Permanent Downhole Gauges. SPE Reservoir Evaluation & Engineering.
Wang, Z., Zhang, J., Shirazi, S.A. and Dou, Y., 2019. Experimental and numerical study of erosion in a non-Newtonian
hydraulic fracturing fluid. Wear, 422, pp.1–8.
Zhang, J., Darihaki, F. and Shirazi, S.A., 2019. A comprehensive CFD-based erosion prediction for sharp bend geometry
with examination of grid effect. Wear.
Zhang, Z., Sun, B., Wang, Z., Gao, Y., Liu, S., and Yin, Z., 2018. Whole wellbore liquid loading recognition model for
gas wells. Journal of Natural Gas Science and Engineering, 60, 153–163.
Zhu, H., Zhu, J., Rutter, R., Zhang, J. and Zhang, H.Q., 2018, July. Sand Erosion Model Prediction, Selection and
Comparison for Electrical Submersible Pump (ESP) using CFD Method. In ASME 2018 5th Joint US-European
Fluids Engineering Division Summer Meeting (pp. V003T17A003–V003T17A003). American Society of Mechanical
Engineers.
Zhu, J., Zhu, H., Zhao, Q., Fu, W., Shi, Y. and Zhang, H.-Q., 2019, March. A Transient Plunger Lift Model for Liquid
Unloading from Gas Wells. In International Petroleum Technology Conference. International Petroleum Technology
Conference.