Analysis and Prediction of Gas Lift PDF

Analysis and Prediction of Gas Lift
Stability from Field Data
Eivind Litland Myhr
Petroleum Geoscience and Engineering

Submission date: June 2017
Supervisor: Harald Arne Asheim, IGP
Norwegian University of Science and Technology

Department of Geoscience and Petroleum
Preface
This Master’s thesis is the final work of the Master of Science degree in Petroleum Production
Engineering, at the Department of Geoscience and Petroleum at the Norwegian University of
Science and Technology (NTNU) in Trondheim, Norway.
First and foremost, I would like to thank my supervisor, Professor Harald Arne Asheim for his
inspirational guidance and for providing me with the model used in the thesis. This is much
appreciated. I would also like to thank Statoil ASA (operator of the Heidrun field) for pro-
viding the field data. The view expressed in this paper are the views of the author and do not
necessarily reflect the views of Statoil ASA.
Thank you to all my friends and fellow students.
Thank you to Helene, for all her support.
Lastly, I would like to thank my family for all their support and encouragement throughout my
years at NTNU.
Eivind Litland Myhr

Trondheim, June 2017
III
IV
Abstract
Producing oil wells on continuous gas lift may under certain operating conditions develop
flow instabilities. Prediction on beforehand may enable prevention by design and operational
changes. Stability criteria exist, but gas lift wells predicted stable by such criteria often turn
out unstable in practice.
A theoretical gas lift model has been tested against field data and the stability of a gas lift
well investigated. Based on existing stability criteria, the new model pretends to improve the
prediction power by considering the outflow variations explicitly. Incorporating the outflow
in the model is physically reasonable and seems to improve prediction capabilities in certain
cases. Dynamic production characteristics in the well have been explored in detail, and the
model response compared with measured dynamics. The dynamic response of the model is
predicted using numerical simulation as well as an analytical solution.
Analysis of the data showed the well might develop severe oscillations. These appear to be
initiated by the combination of varying density of the produced fluid mixture and disturbances
in the gas delivery system. The proposed model predicts the well stable for all stationary
production periods examined. In periods of flow instability in field data, the model prediction
is in disagreement with observed response.
A parametric sensitivity study has been performed and demonstrated the model is capable
of capturing the effect of main design and reservoir parameters on stability. When the outflow
response is excluded from the model, the prediction capabilities seems to be reduced.
V
VI
Abstrakt
Oljebrønner som produserer ved kontinuerlig gassløft kan ved enkelte driftsforhold utvikle
strømningsustabiliteter som forstyrrer produksjonen. Prediksjon av ustabilitet på forhånd er
ønskelig, og kan hindre ustabilitet ved hjelp av drifts- og designendringer. Ulike stabilitetskri-
terier eksisterer, men gassløftbrønner predikert stabile ved slike kriterier viser seg ofte å være
ustabile i praksis.
En teoretisk gassløftmodell har blitt testet mot feltdata og stabiliteten av en gassløftet
brønn har blitt undersøkt. Den teoretiske modellen forsøker å forbedre prediksjonen av sta-
bilitet ved å inkludere variasjonene ved utløpet av produksjonsrøret. Hovedforskjellen mellom
den nye modellen og eksisterende kriterier og tilnærminger er at variasjonen ved utløpet er
inkludert. Dette ser ut til å forbedre modellens prediksjonsevner i enkelte tilfeller.
Feltmålingene viser at brønnen klarer å opprettholde stabil produksjon bare i kortere peri-
oder før produksjonen utvikler alvorlig strømningsustabilitet. Analysen antyder at ustabiliteten
skyldes kombinasjonen av varierende tetthet av den produserte fluidblandingen og forstyrrelser
i gassleveransesystemet. Basert på feltmålingene predikerer den nye modellen brønnen stabil
i alle stasjonære produksjonsperioder som er undersøkt. I perioder med klar strømningsusta-
bilitet stemmer ikke modellens forutsigelser med oppførselen observert fra feltmålingene.
En parametrisk simuleringsstudie er utført og viser modellen er i stand til å vurdere ef-
fekten av viktige design- og reservoarparametere på stabilitet. Når variasjonene ved utløpet av
produksjonsrøret er ekskludert ser det ut til at modellens prediksjonsevner er redusert.
VII
VIII
Contents
1 Introduction 1
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2 The gas lift system 3
3 Gas lift instability 5

3.1 Hydrodynamic slugging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.2 System instability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.2.1 Casing heading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.2.2 Tubing heading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.3 Consequences of instability . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
4 Stability theory 9
4.1 Early observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
4.2 Asheim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
4.3 Blick et al. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.4 Gang and Golan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.5 Alhanati et al. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
4.6 Poblano et al. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.7 Fairuzov et al. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.8 Comparison of theories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
5 Active feedback control and stability 17

5.1 Dalsmo and Halvorsen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
5.2 Hu and Golan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
5.3 Eikrem et al. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
6 Dynamic gas lift model 19

6.1 General model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
IX
6.2 Current model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
6.2.1 Steady-state gas lift . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
6.2.2 Dynamic response . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
6.3 Comparison with existing criteria . . . . . . . . . . . . . . . . . . . . . . . . . 26
7 Gas lift stability study 27

7.1 The Heidrun case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
7.1.1 Data availability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
7.1.2 Well design and completion . . . . . . . . . . . . . . . . . . . . . . . 28
7.2 Production measured . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
7.3 Characterization and parameter estimation . . . . . . . . . . . . . . . . . . . . 31
7.3.1 Stationary production . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
7.3.2 Static pressure build-up . . . . . . . . . . . . . . . . . . . . . . . . . . 35
7.3.3 Oscillating production . . . . . . . . . . . . . . . . . . . . . . . . . . 36
7.3.4 Well stabilization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
7.3.5 Initiation of instability . . . . . . . . . . . . . . . . . . . . . . . . . . 41
7.3.6 Diminishing oscillations . . . . . . . . . . . . . . . . . . . . . . . . . 49
8 Dynamic response study 53

8.1 Model and prediction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
8.2 Numerical simulation for gas lift response . . . . . . . . . . . . . . . . . . . . 54
8.2.1 Stable production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
8.3 Parametric sensitivity study on stability . . . . . . . . . . . . . . . . . . . . . 60
8.3.1 Effect of gas injection rate . . . . . . . . . . . . . . . . . . . . . . . . 61
8.3.2 Effect of gas lift orifice size . . . . . . . . . . . . . . . . . . . . . . . 62
8.3.3 Effect of productivity index . . . . . . . . . . . . . . . . . . . . . . . 63
8.3.4 Effect of injection depth . . . . . . . . . . . . . . . . . . . . . . . . . 65
8.3.5 Effect of tubing size . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
8.4 Analytical solution for gas lift model response . . . . . . . . . . . . . . . . . . 68
8.4.1 Stable production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
9 Discussion 73
9.1 Observed characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
9.2 Initiation of instability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
9.3 Estimation of static and dynamic parameters from measured data . . . . . . . . 74
9.4 The dynamic model and dynamic characteristics prediction . . . . . . . . . . . 75
X
9.5 System versus model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
9.5.1 Further work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
10 Conclusion 77
Nomenclature 79
References 83
A Steady-state gas lift model 87

A.1 Model description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
A.1.1 Superficial velocity and mass flow . . . . . . . . . . . . . . . . . . . . 87
A.1.2 Fluid velocity, slip and volume fractions . . . . . . . . . . . . . . . . . 88
A.1.3 Flux fractions and mixed flow . . . . . . . . . . . . . . . . . . . . . . 89
B Dynamic model data 91

B.1 Response parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
B.2 Alternative production reference . . . . . . . . . . . . . . . . . . . . . . . . . 92
B.3 Production data for sensitivity study . . . . . . . . . . . . . . . . . . . . . . . 93
XI
XII
List of Figures
2.0.1 The gas lift system and main components (Eikrem et al., 2008). . . . . . . . . 3
4.4.1 Equilibrium points and static stability. . . . . . . . . . . . . . . . . . . . . . 12

4.6.1 Stability map, injection ports size and gas injection flow rate (Poblano et al.,
2002). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.7.1 Stability map, wellhead pressure and gas injection flow rate (Fairuzov et al.,
2004). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
6.2.1 Simplified gas lift model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
7.1.1 Well 6507/7-A-23 schematic. . . . . . . . . . . . . . . . . . . . . . . . . . . 28

7.2.1 Liquid production March - April 2016. . . . . . . . . . . . . . . . . . . . . . 29
7.2.2 Liquid production May - June 2016. . . . . . . . . . . . . . . . . . . . . . . 30
7.2.3 Liquid production January - February 2017. . . . . . . . . . . . . . . . . . . 30
7.3.1 Liquid production January - February 2017. . . . . . . . . . . . . . . . . . . 31
7.3.2 Liquid production rate 1-3. January 2017. . . . . . . . . . . . . . . . . . . . 32
7.3.3 Probability density function liquid production rate 1-16. January 2017. . . . . 32
7.3.4 Stationary pressure profile for production tubing and annulus. . . . . . . . . . 34
7.3.5 Measured well pressure 14-19. February 2017. . . . . . . . . . . . . . . . . 35
7.3.6 Measured and estimated well pressure 18. February 2017. . . . . . . . . . . . 36
7.3.7 Oscillating liquid production January-February 2017. . . . . . . . . . . . . . 37
7.3.8 Amplitude spectrum liquid rates January-February 2017. . . . . . . . . . . . 37
7.3.9 Downmhole and wellhead pressure oscillations 28. January 2017. . . . . . . 38
7.3.10 Amplitude spectrum pressure oscillations 28-29. January 2017. . . . . . . . . 38
7.3.11 Rate and pressure variations 28. January 2017. . . . . . . . . . . . . . . . . . 39
7.3.12 Rate and pressure variations 1. February 2017. . . . . . . . . . . . . . . . . . 39
7.3.13 Liquid production rate 30. January 2017. . . . . . . . . . . . . . . . . . . . . 41
7.3.14 Liquid production rate and gas production and injection rates 23. March 2016. 42
7.3.15 Downhole and wellhead pressures 23. March 2016. . . . . . . . . . . . . . . 42
7.3.16 Gas delivery and injection pressures 20-23. March 2016. . . . . . . . . . . . 43
7.3.17 Gas injection and liquid production rates 20-23. March 2016. . . . . . . . . . 43
XIII
7.3.18 Amplitude spectrum gas delivery and injection pressures 20-23. March 2016. 44
7.3.19 Amplitude spectrum gas injection and liquid production rates 20-23. March
2016. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
7.3.20 Oscillating liquid production rate 23. March 2016. . . . . . . . . . . . . . . . 45
7.3.21 Liquid production rate and gas production and injection rates 9. May 2016. . 46
7.3.22 Downhole and wellhead pressures 9. May 2016. . . . . . . . . . . . . . . . . 46
7.3.23 Gas delivery and injection pressures 6-9. May 2016. . . . . . . . . . . . . . . 47
7.3.24 Gas injection and liquid production rates 6-9. May 2016. . . . . . . . . . . . 47
7.3.25 Amplitude spectrum gas delivery and injection pressures 6-9. May 2016. . . . 48
7.3.26 Amplitude spectrum gas injection and liquid production rates 6-9. May 2016. 48
7.3.27 Oscillating liquid production 9. May 2016 . . . . . . . . . . . . . . . . . . . 48
7.3.28 Liquid production rate 2-17. June 2016. . . . . . . . . . . . . . . . . . . . . 50
7.3.29 Gas delivery and injection pressures 8-10. June 2016. . . . . . . . . . . . . . 50
7.3.30 Gas injection and liquid production rates 8-10. June 2016. . . . . . . . . . . 51
7.3.31 Amplitude spectrum gas delivery and injection pressures 8-10. June 2016. . . 51
7.3.32 Amplitude spectrum liquid production and gas injection rates 8-10. June 2016. 51
8.0.1 Mechanism of kinematic waves. . . . . . . . . . . . . . . . . . . . . . . . . 53

8.2.1 Simulated pressure response for stationary production 1-16. January 2017. . . 56
8.2.2 Simulated pressure response and measured pressure 1-16. January 2017. . . . 57
8.2.3 Measured well pressure 27-29. January 2017. . . . . . . . . . . . . . . . . . 58
8.2.4 Simulated pressure response 27-29. January 2017. . . . . . . . . . . . . . . . 59
8.3.1 Dynamic response - Injection rate: 8.85 104 Sm3 /d. . . . . . . . . . . . . . . 61
8.3.4 Dynamic response - Injection orifice size: 9.5 mm. . . . . . . . . . . . . . . 63
8.3.5 Dynamic response - Injection orifice size: 14.3 mm. . . . . . . . . . . . . . . 63
8.3.6 Dynamic response - Productivity index: 45 Sm3 /d/bar. . . . . . . . . . . . 64
8.3.7 Dynamic response - Productivity index: 5 Sm3 /d/bar. . . . . . . . . . . . . 64
8.3.8 Dynamic response - Injection depth: 2164 m MD. . . . . . . . . . . . . . . . 65
8.3.9 Dynamic response - Injection depth: 2850 m MD. . . . . . . . . . . . . . . . 66
8.3.10 Dynamic response - Tubing size: 5.5". . . . . . . . . . . . . . . . . . . . . . 66
8.3.11 Dynamic response - Tubing size: 7". . . . . . . . . . . . . . . . . . . . . . . 67
8.4.1 Analytical pressure response for stable production: Pˆw = 1 Pa, Pˆg = 1 Pa. . . . 70
8.4.2 Analytical pressure response for stable production: Pˆw = 2 Pa, Pˆg = 1 Pa. . . . 70
8.4.3 Analytical pressure response on oscillating production: Pˆw = 1 Pa, Pˆg = 1 Pa. 71
8.4.4 Analytical pressure response on oscillating production: Pˆw = 1 Pa, Pˆg = 2 Pa. 72
XIV
List of Tables
7.3.1 Input steady-state model 1-16. January 2017 . . . . . . . . . . . . . . . . . . . 33

7.3.2 Production statistics - oscillating production. . . . . . . . . . . . . . . . . . . . 37
8.2.1 Stationary production 1-16. January 2017 . . . . . . . . . . . . . . . . . . . . 55

8.2.2 Stationary production 27-28. January 2017 . . . . . . . . . . . . . . . . . . . . 58
8.3.1 Gas lift base case - sensitivity study. . . . . . . . . . . . . . . . . . . . . . . . 60
8.4.1 Dynamic parameters for analytical solution - stable production. . . . . . . . . . 69
8.4.2 Dynamic parameters for analytical solution - oscillating production. . . . . . . 71
B.1.1 Response parameters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

B.2.1 Oscillating production. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
B.3.1 Effect of gas injection rate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
B.3.2 Effect of productivity index. . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
XV
XVI
Chapter 1
Introduction
1.1 Motivation
In the specialization project in the autumn of 2016, the author investigated the stability of a gas
lift well in the Norwegian Sea. The analysis was based on existing criteria and showed strong
disagreement between predicted and measured well response. This thesis is a continuation of
the work and is intended to check a new gas lift model. The suggested model is a simplification,
and thus its applicability needs to be checked against measurements.
Field data have been provided, and this gives the opportunity to investigate the actual
dynamics in the well under specific operating conditions and enables the model response to be
compared with measurements. Utilizing the model aspects of instability will be studied and
compared with existing theories and standards. The model builds upon existing criteria but
includes some modifications that hopefully will improve prediction of stability in gas lift wells.
1.2 Outline
The report is divided into ten chapters. Chapter 2 gives a short introduction to the gas lift
method and the main components of the system. Chapter 3 provides the basic principles of in-
stability and describe the different types of flow instability in gas lift wells. The consequences
of unstable production will also be outlined in this section. Chapter 4 addresses selected sta-
bility theories from literature while Chapter 5 discuss the implementation of active control
systems in gas lift wells. The part on stability theory include both analytical criteria and graph-
ical solutions to predict the onset of instability. The theory discussed in Chapters 2-5 is partly
based on the author‘s previous project work (Myhr, 2016).
Chapter 6 form the theoretical basis for the new gas lift model to be tested against field
data. The model itself and the assumptions behind it are presented in detail and compared
against existing criteria. Chapter 7 accommodate the in-depth stability analysis of the well
1
in the Heidrun field and provide the operational basis for the proposed gas lift model. Chap-
ter 8 examines the gas lift model by checking it against the field measurements. Numerical
simulation explores the theoretical model against the data. Also, its prediction power and the
dynamical response is tested and checked with an analytical solution.
Chapter 9 summarizes the most significant findings from the stability study and the model
test results. Possible causes for instability in the Heidrun well are discussed, and the prediction
power of the gas lift model are explained. Model results that differ from field data or existing
theories are highlighted in this section. Lastly, Chapter 10 concludes the results of the field data
analysis and model test results, based on the discussion in Chapter 9.
2
Chapter 2
The gas lift system
In traditional gas lift systems, gas lift is accomplished by injecting gas at high pressure through
the wellhead and down into the annular volume between the tubing and casing. The compressed
gas mixture enters the production tubing through an orifice valve and mixes with the flowing
fluid in the tubing. The injection of gas to the tubing lowers the density of the flowing fluid
mixture in the tubing and reduces the hydrostatic pressure gradient in the well accordingly.
Figure 2.0.1 shows a simplified setup of a gas lift well and the main components.
Figure 2.0.1: The gas lift system and main components (Eikrem et al., 2008).
Injection valves provide communication between annulus and tubing. The size is selected
such that the pressure drop across the valve is an approximately 3-6 bar for a given gas injection
rate (Fairuzov et al., 2004).
3
The injection valve usually comprises an orifice port and a check valve. Proper sizing of
the injection orifice ensures a steady inflow of lift-gas while a check valve prevents the influx
of reservoir fluid from the tubing to the annulus (Bellarby, 2009).
Poettman et al. (1952) and Bertuzzi et al. (1953) proposed design principles that are still
valid. The design principles are used to verify the gas lift design and relate the gas injection
rate and associated oil rate and may be used to obtain the optimum combination of flow rates.
An efficient and functional gas lift system is important for various reasons. Too high injection
rate may both reduce oil production and increase the overall costs of the operation (Poettman
and Carpenter, 1952; Bertuzzi et al., 1953).
Injection of lift-gas increase the flow rate in the production tubing and will cause increased
frictional pressure drop. To obtain an efficient gas lift design, it is necessary to compare the
reduction in hydrostatic pressure and the increased frictional pressure drop.
For a gravity dominated system gas lift may be favorable and increasing the injection rate
may improve oil production. In the event of a well dominated by frictional pressure drop, i.e.
high gas-oil ratio, the benefit of gas lift is usually lower. In fact, under certain conditions,
the method may be counterproductive, and other artificial lift methods should be considered.
When too much gas is flowing in the tubing, the injection of lift-gas may cause phase slippage.
When slippage occurs, the frictional pressure drop overcomes the hydrostatic pressure drop in
the well, and less energy is available to transport the liquid along the production tubing.
Thus gas lift should preferably be considered as an artificial lift method in wells producing
heavy oils with low gas content or in wells with high water cut. (Saepudin et al., 2007).
4
Chapter 3
Gas lift instability
Gas lift systems comprise multiphase flow of oil and gas, and usually also water. Nearly all
multiphase flow systems experience rate and pressure variations because of redistribution of
gas and liquids. The variations cause relatively small flow changes with short durations. By
itself, this has little effect on production. In gas lifted wells, however, it may bring about flow
instabilities (Avest and Oudeman, 1995).
3.1 Hydrodynamic slugging
Hydrodynamic slugging denote instability that occurs on the gas-liquid interface in the flowing
fluid mixture. These variations may be described as a discontinuous redistribution of fluids and
may cause small density and pressure changes in the production tubing (Liu and Vandu, 2005).
Hydrodynamic slugging is natural occurring in gas lift systems and is usually not considered a
problem to production (Asheim, 1988).
3.2 System instability
System instability denotes variations that comprise the entire gas lift production system, often
as cyclic variations of large amplitudes. These may cause operational problems in the produc-
tion facilities and decreases the efficiency of the gas lift system (Alhanati et al., 1993).
Gang and Golan (1989) claimed system instability has a systematic background and are
caused by the inertia and feedback effects of the gas lift system. When such instability is
experienced, the system continues to oscillate about an average level periodically.
Two types of system instability are reported in the literature, casing heading and tubing
heading instability.
5
3.2.1 Casing heading
By casing heading is understood instability that involves pressure variations in the annulus,
detectable at the casing head. Gilbert (1954) described casing heading in flowing oil wells
and claimed that it could be suppressed by installing packers. Since then, casing heading has
been the subject of several studies (Asheim, 1988; Gang and Golan, 1989; Blick et al., 1988;
Alhanati et al., 1993).
Casing heading develops in the system as result of communication between the production
tubing and annulus at the injection valve. The communication between tubing depends on the
characteristics of the valve. Thus the valve has a significant impact on how changes in the
tubing influence the annulus. In continuous gas lift wells, heading may be observed if the flow
through the injection valve is sub-critical. Under this condition, the injection rate depends on
the downstream tubing pressure. At critical flow, however, any changes of the condition in the
production tubing will not affect the injection rate (Hu and Golan, 2003).
By traditional gas lift design, constant inflow of lift-gas to the tubing is assumed. The
pressure in the production tubing may exhibit temporary variations, causing temporary vari-
ations in the inflow of lift-gas. By casing heading, an increase in the inflow of gas leads to
increased pressure difference between the annulus and the production tubing. This causes the
inflow of lift-gas to increase further, and this positive feedback causes the well to exhibit un-
stable flow behavior. However, if an increased inflow of lift-gas causes decreased pressure
difference across the annulus and tubing the inflow of lift-gas will decrease. In this case, the
system is stabilized by negative feedback (Asheim, 1988).
The mechanism above may also be described analytically by considering the flow equa-
tions in the annulus and production tubing. By looking at incremental changes in flow rate
and pressure, it is possible to investigate the response of the system away from its steady-state
solution. In many cases, the well opposes the disturbances and have a dampening response to
small changes, which is called negative feedback. In some situations, however, the response of
the gas lift system may amplify the disturbances and flow instabilities may develop (Gang and
Golan, 1989).
3.2.2 Tubing heading

By tubing heading is understood instability that does not involve annulus pressure variations,
thus without variation of the gas injection rate. This is a rather new phenomenon discovered in
gas lift wells in the North Sea, producing from depleted reservoirs. Unlike casing heading, the
tubing heading instability may develop in gas lift wells in which the inflow of lift gas is critical
(Hu and Golan, 2003).
6
When the inflow of lift-gas is constant, any variation to the liquid inflow from the reservoir
results in a density change in the fluid mixture at the bottom of the well. The density of the
fluid mixture will vary due to variation to the phase fractions. The density change will travel
along the tubing as density wave and reach the exit of the tubing after a period of time. This
generates a small change to the pressure drop in the tubing and will affect the inflow to the
wellbore.
Hu and Golan (2003) claimed that the occurrence of tubing heading does not necessary
shift the gas lift system to become unstable since the well has a self-controlling effect. This
means that any decrease in the pressure drop due to decreased mixture density results in an
increase of the liquid inflow from the reservoir, and opposite. However, this mechanism is
out of phase due to the delay between the inflow and outflow in the production tubing. Under
certain conditions, the self-controlling mechanism might break down and the well exhibit flow
instabilities (Hu and Golan, 2003).
3.3 Consequences of instability

Instability in continuous gas lift wells has several disadvantages. Operating under cyclic con-
ditions are associated with operational issues like unplanned shutdowns and increased main-
tenance. Periods of reduced liquid production followed by large plugs of gas and liquid may
result in poor separation and limit the separator capacity (Slupphaug and Bjune, 2006). Also,
periods of very high gas rates may lead to excessive flaring. Pressure surges and rate varia-
tion in the production facilities are associated with production control, gas allocation, and can
prevent reliable production measurements and well tests. The variations, in turn, can make it
difficult to predict future production and the storage and transportation needed (Alhanati et al.,
1993).
Beside operational problems, Hu and Golan (2003) pointed out production loss as another
important consequence of unstable gas lift wells. The authors simulated the production loss due
to casing heading and claimed that up to 35 % production loss could be expected. Guerrero-
Sarabia and Fairuzov (2013) did a similar study and discovered a reduction in oil production
of 27 %. Thus, to reach the desired recovery from the field/well the tail production must be
prolonged to compensate for the reduced production rates. This may increase the costs of the
project due to the extended ongoing operational costs. The overall consequence of unstable
production is safety and profit-reducing issues and reduced efficiency of the gas lift system (Hu
and Golan, 2003).
7
8
Chapter 4
Stability theory
4.1 Early observations

Since early 1950, numerous studies on instability in gas lift wells have been reported. Among
them, the first comprehensive study of instability was given by Gilbert (1954).
In his study, Gilbert made observations and characterized the instability processes in natu-
rally flowing oil wells completed without production packers. Gilbert observed that associated
gas could accumulate in the annulus. The accumulation initiated flow instabilities when the
gas in the annulus reached sufficiently high pressure and started to flow into the tubing. The
accumulation of gas in annulus changed the rate and pressure conditions in the well. Gilbert
claimed instability could be eliminated by the use of packers between the production casing
and tubing.
In wells already producing, Gilbert suggested installing pressure actuated valves. At a
preset lower annulus pressure the tubing outlet choke will close such that reservoir fluids flow
into the annulus and accumulate. The influx of reservoir fluids will cause a rapid pressure
increase in the annulus, after which the valve is opened at a given maximum pressure (Gilbert,
1954). According to Gilbert, this mechanism could eliminate instability in oil wells completed
without packer. Today, however, all wells are equipped with packers and this eliminates the
need for valves of this type in the production tubing.
Bertuzzi et al. (1953) described the mechanism of heading in gas lift wells and observed
that heading would occur if the injection rate were reduced below a certain minimum. In the
event of severe heading, the liquid production would eventually stop. They claimed that a drop
in the tubing pressure brought about a sudden flow of gas into the tubing. The amount of gas
flowing into the tubing depends on the volume and the pressure in the annulus. If the annular
pressure dropped too much, gas stopped to flow into the production tubing (Bertuzzi et al.,
1953).
9
Several attempts have been made to understand and quantify flow instability in gas lift
wells with numerical approaches and techniques, similar to what Grupping et al. (1984) did.
They succeeded in demonstrating instability by numerical means and suggested stabilizing
measures for gas lift wells (Grupping et al., 1984). Fitremann and Vedrines (1985) performed
linear stability analysis on a mathematical model of a gas lift system. The results were shown to
resemble small-scale laboratory experiments. No comparisons with field data were attempted
(Fitremann and Vedrines, 1985).
4.2 Asheim
As a simplification of Fitremann’s analysis, Asheim (1988) developed two explicit criteria for
stability in gas lift wells. The criteria are derived by considering how the gas lift system,
initially in equilibrium, respond small changes in the well conditions.
The derivation is based on how the injection system and reservoir respond to a change
in the tubing pressure. Transient inflow is neglected, and the relationship between the well
pressure and flow from the reservoir is assumed to be described by the steady-state inflow rela-
tionship. Grupping et al. (1984) and Fitremann and Vedrines (1985) did the same assumption in
their study. Furthermore, slippage is neglected, and the flow through the gas lift valve assumed
isothermal and incompressible. The system is considered gravity dominated, and acceleration
and frictional forces are disregarded.
Inflow response
The first criterion quantifies stability as a consequence of the inflow responses of injection gas
and reservoir fluid inflow. The inflow from the reservoir acts stabilizing because it introduces
heavier fluids into the production tubing. If the inflow of reservoir fluid is more sensitive
to pressure than the gas injection rate, the density of the fluids in the production tubing will
increase. The inflow causes an increase in the pressure in the tubing and has a stabilizing effect
on the system. Stabilization by the inflow response is given as
2
ρgsc Bg qgsc J
F1 = > 1. (4.2.1)
qLsc (EAi )2
10
By equation 4.2.1, stability is favored by a high inflow rate of lift-gas, small injection
port size, and high productivity index. Also, criterion F1 show a high density of the lift-gas is
stabilizing. However, the gas density is in general low compared to the liquid density.
Depletion response
If equation 4.2.1 is not satisfied, a decrease in the pressure in the tubing will cause the inflow
of lift-gas to increase more than the liquid rate from the reservoir. The increased inflow of gas
will deplete the pressure in the annulus and decrease the tubing pressure. If the pressure in the
annulus depletes faster than the pressure in the tubing, then the pressure difference between the
tubing and annulus will decrease. This reduces the inflow of lift-gas, and stabilizes the system.
Stabilization by the pressure-depletion response is given as
Vt 1 pti qf i + qgi
F2 = > 1. (4.2.2)
Vc gD (ρf i − ρgi ) qf i (1 − F1 )
By equation 4.2.2, stability is stimulated by a high inflow rate of lift-gas and a small
volume of the annulus. While confirming that a high injection rate is beneficial the inequality
also shows that a high tubing pressure is stabilizing.
4.3 Blick et al.

Blick et al. (1988) approached the stability problem differently than Asheim. They developed a
mathematical model consisting of differential equations describing the transient responses in a
gas lift system. The purpose of the model was to analyze the influence of small variations from
the steady-state of the injection rate, pressure at the injection point, and the reservoir inflow.
The differential equations are Laplace transformed and then combined by Cramer‘s rule to
obtain the solutions as a linear system of equations. The procedure produces a characteristic
equation with three coefficients, and the model predicts the gas lift system stable when the
coefficients are all positive or negative. If one of the coefficients have a sign that is different,
the well is predicted unstable and small flow perturbations increase with time. Blick et al. used
their model on hypothetical gas lift wells but never compared the result with any field data
(Blick et al., 1988).
4.4 Gang and Golan

Gang and Golan (1989) worked out a criterion for assuring smooth and stable operation of gas
lift wells. The purpose of the study was to aid the design of continuous gas lift wells by giving
recommendations on the selection and settings of the production choke, downhole gas lift ori-
11
fice, and injection choke (Gang and Golan, 1989). Two types of instabilities are addressed in
the study, static and dynamic. Gang and Golan used a graphical approach to describe the sta-
bility of gas lift wells, the procedure originally presented by Gilbert (1954). By analyzing the
inflow characteristics (IPR), lifting relationship (TPR), gas injection relationship (GPR), and
discharge of gas (DPR) through the gas lift valve, they could determine the equilibrium points
in the system. These points describe the conditions at which steady-state flow occur. Gang and
Golan plotted (figure 4.4.1) the GPR and DPR as a function of gas injection rate, and picked
the two intersection points as steady-state solutions of the system.
DPR
Pressure
GPR
Gas Injection Rate
Figure 4.4.1: Equilibrium points and static stability.
Under the condition that the annulus pressure is constant, they claimed that the left point
represents a stable solution while the right equilibrium point is subjected to static instability.
At equilibrium point to the right, a temporary increase in the flow rate above the equilibrium
rate requires less pressure to preserve the flow than the pressure available (from the reservoir).
Thus, the flow rate is demanded to increase further and will not return to the original value.
The opposite applies to the point to the left (Gang and Golan, 1989). Based on the analysis,
Gang and Golan discovered that a small injection port size and choking at the wellhead are
stabilizing.
The authors suggests that the static stability should be investigated by considering the
required and available pressure-rate relationship in the production system, and proposed the
following criterion for static stability.
dpavailable dprequired
< . (4.4.1)
dqf low dqf low
12
Gang and Golan focused their study on steady-state flow and static instability but empha-
sized that even if an equilibrium is statically stable, it may still be unstable due to systematic
instabilities. They mentioned the interaction between the surface and downhole injection ori-
fice and claimed that the sizes must be compatible if not severe oscillations could occur (Gang
and Golan, 1989).
4.5 Alhanati et al.

Alhanati et al. (1993) investigated formerly proposed stability criteria in the literature and ex-
tended the criteria developed by Asheim (1988). They wanted to arrive at a more general
stability criteria and extended Asheim‘s criteria by taking into account different flow regimes
at the surface injection valve and the downhole injection valve. Like Gang and Golan (1989),
Alhanati et al. analyzed the effect of inflow and outflow on stability in gas lift wells. They
claimed steady-state procedures are inadequate and postulated; "Examining only inflow and
outflow performance curves, as was done in the way suggested by Gang and Golan, is not suf-
ficient for a proper continuous gas lift stability analysis; rather, a different approach than that
used for flowing wells is necessary" (Alhanati et al., 1993).
Alhanati et al. compared the criteria developed by Asheim (1988) and Blick et al. (1988)
and claimed that Asheim‘s criteria are easier to use as they only involve parameters employed in
the design and operation of the well. The derivation of the new criteria, however, was inspired
by the mathematical model of Blick et al., but the results are still closer to Asheim‘s. Like
Asheim, they neglect frictional forces in the production tubing which they agree will make the
criteria more conservative since friction has a dampening effect on the system (Alhanati et al.,
1993).
Using the criteria, Alhanati et al. show that casing heading will not occur if the flow
through the downhole injection valve is critical. This is inherent in the criteria developed by
Asheim. If the injection rate does not change by a reduction in the well pressure, the fluid
density in the well will increase as a result of increased inflow of reservoir fluid. This satisfies
stability criterion F1 (equation 4.2.1). The authors recommend ways of stabilizing the well,
and these include decreasing the volume of the annulus, reducing the size of the injection port,
and increasing the flow rate of lift-gas. Alhanati et al. compared the criteria against field data
and concluded the criteria gave qualitative indications on ways to avoid instabilities in gas lift
wells (Alhanati et al., 1993).
13
4.6 Poblano et al.
Poblano et al. (2002) developed diagrams to facilitate the application of stability criteria. The
diagrams, called stability maps, demonstrate regions of stable and unstable flow in gas lift wells,
in addition to the operational limits of the system. By a numerical model of a gas lift well, field
data, and existing stability criteria, stability maps were constructed to determine the stability
of the well under certain conditions. A stability map, based on the criteria by Alhanati et al.,
is shown in figure 4.6.1. The map shows stability given as a function of the gas injection rate
and the size of the downhole injection port. Any change to other well parameters will generate
a new map, with a different stability limit associated with the operation (Poblano et al., 2002).
Fig. 6—Stability map based on the Alhanati et al.8 criteria (subcritical flow at the superficial valve) for the tested well.
Figure 4.6.1: Stability map, injection ports size and gas injection flow rate (Poblano et al., 2002).
v ! velocity, L/t, m/sec [ft/sec] $g ! gas density, m/L3, kg/m3 [lbm /ft3]
Vc ! volume of casing annulus between the wellhead and $gi ! lift gas density at injection point, m/L3, kg/m3[lbm /ft3]
injection point, L3, m
According to Poblano et al., the benefit of stability maps is a rapid and efficient assessment
Vt ! tubing volume between the wellhead and injection
point, L3, m Subscripts
y ! ratio of downstream and upstream pressure
ofinjection
of the operation, and it can be used in the designcii !
yc ! critical pressure ratio, Eq. A-10 the point
gas lift system. The stability maps
! casing at injection point depth
Y ! gas-expansion factor ! casing at surface (downstream the surface choke)
confirm that a small injection port size and high co
z ! gas compressibility factorinjection
ch ! surface gas rate
injectionof lift gas stimulates stability.
choke
!g ! specific gravity of gas m ! gas allocation network
"g ! gas viscosity, M/Lt, Pa sec [cp]
Also, the maps show a reduction in diameter of the
ti !production
#p ! pressure drop, m/Lt2, Pa [psi]
tubing
tubing at injection point depth has a stabilizing effect.
v ! gas-lift valve
$fi ! reservoir fluid density at injection point, m/L3, kg/m3
The study showed the stability criteria by Asheim and Alhanati et al. correlate well with the
Acknowledgments
actual conditions in the test well. Poblano et al. E.port

Poblano is grateful to Pemex E&P for providing financial sup-
declared stability
during his stay maps asU. aof useful
at the Natl. Autonomous Mexico. tool in well
planning and design. References
1. Bertuzzi, A.F., Welchon, J.K., and Poettmann, F.H.: “Description and
Analysis of an Efficient Continuous-Flow Gas Lift Installation,”
Trans., AIME (1953) 198, 271–278.
2. Gruping, A.W., Luca, C.W.F., and Vermulen, F.D.: “Continuous-Flow
Gas Lift: Heading Action Analyzed for Stabilization,” Oil & Gas J. (23
4.7 Fairuzov et al. July 1984) 47–51.
3. Gruping, A.W., Luca, C.W.F., and Vermulen, F.D.: “Continuous-Flow
Gas Lift: These Methods Can Eliminate or Control Annulus Heading,”
Oil & Gas J. (23 July 1984) 186–192.
Fairuzov et al. (2004) modified the stability maps developed by Poblano et al. and used the
4. Filtremann, J.M. and Verdines, P.: “Nonsteady Gas-Liquid Flow in
Pipes and Gas lifted Wells,” Proc., Second Intl. Conference on Mul-
tiphase Flow, London (1985) 245–262.
wellhead pressure as the ordinate axis in the new maps. They believed this modification was
5. Asheim, H.: “Criteria for Gas Lift Stability,” JPT (November 1988)
1452–1456.
more appropriate since the wellhead pressure and injection rate are parameters that easily can
6. DeMoss, E.E. and Tiemann, W.D.: “Gas Lift Increases High-
Volume Production from the Claymore Field,” JPT (April 1982) 696–
Fig. 7—Predicted fractional-valve opening for stability map
shown in Fig. 6. 702.
74 14 February 2005 SPE Production & Facilities

be measured and controlled during production. New stability criteria were developed, and data
from a deep offshore gas lift well in the Akal field were used for comparison with existing
criteria. Figure 4.7.1 show the stability boundaries predicted by the different criteria. It is
worth noting that the new stability map also shows the conditions in which the well cannot
operate due to low gas rate in combination with high wellhead pressure.
Figure 4.7.1: Stability map, wellhead pressure and gas injection flow rate (Fairuzov et al., 2004).
Fairuzov et al. discovered that existing stability criteria underestimated the unstable region
and claimed the new criteria provides a more accurate prediction of the stability threshold.
Among their findings, they observed that increased productivity index was strongly stabilizing,
which confirms the observations made by Asheim. Fairuzov et al. believed that this information
could be used to identify well damage and the effect of stimulation. They also pointed out that
instability can be caused by a drop of the pressure in the gas supply network, e.g., in the event
of a temporary shutdown of compressors or other disturbances (Fairuzov et al., 2004).
4.8 Comparison of theories

The main features of the above stability theories are in general the same, and they all say gas lift
instability is a phenomenon that entails production problems. The theories indicate that proper
well design may help ensure stable production throughout the well’s life. The stability theories
also agree on the stabilization measures that may be implemented in the system. Among oth-
ers, the most suggested measures are increased gas injection rate, decreased injection orifice
size, and reduced annular volume. Decreasing the annular volume, however, may usually be
accomplished only in the design phase.
Thus, thorough analyses should be performed in the design stage to identify potential
stability problems before the well is completed. In this way, unnecessary interventions and
production shutdowns are avoided.
15
The stability criteria derived by Asheim (1988) are often the easiest to use in practice and
form the basis for several of the more recent and more complex theories. Common to all criteria
is that they are based on several simplifying simplifications and assumptions, and the extension
by Alhanti et al. (1993) include many of the same assumptions as the criteria by Asheim.
Poblano et al. (2002) used the criteria of Asheim and Alhanati et al. as limit values in
their stability criteria, and discovered the criteria correlated well with the conditions in the
test well. Fairuzov et al. (2004), on the other hand, claimed the criteria underestimated the
unstable region in the map. With attention to Asheim’s criteria, they are both based solely
on the downhole response and does not take into account the conditions at the outlet of the
production tubing. This may affect their prediction power.
In more recent times in his doctoral thesis, Hu (2004) claimed that existing stability cri-
teria are difficult to use today because wells have become far more complicated than before.
Especially regarding completion and flow. The conditions will vary from well to well, and often
other factors need to be taken into account when choosing the stabilization method. Also, the
prediction of future well conditions is often very challenging, and this too makes it harder to
select the best way of stabilization. Therefore, the consequence of unstable gas lift wells may
give considerable uncertainties in the production forecast and operational costs (Hu, 2004).
16
Chapter 5
Active feedback control and stability
During the 1990s, reports suggested cybernetic approaches to eliminate flow instability in gas
lift wells. Studies showed that equipping wells with an automatic feedback control system
could both reduce instability and increase production. Automatic feedback control implies
that settings of one or more adjustable elements in the well, i.e. choke or orifice opening, are
automatically decided based on one or more measurements in the same system (Dalsmo and
Halvorsen, 2002).
5.1 Dalsmo and Halvorsen

Dalsmo and Halvorsen (2002) employed an active feedback control solution on a gas lift well
in the Brage field. To control the production, downhole pressure measurements in the well
were used to estimate the opening of the production choke at the wellhead. In an attempt to
control the test well manually, pressure and flow rates showed severe oscillations, and after a
short production period, the well had to be shut-in.
When the control system was activated the pressure and flow variations were eliminated,
and the flow stabilized (Dalsmo and Halvorsen, 2002). Dalsmo et al. concluded that the benefit
of employing automatic feedback control in unstable gas lift wells was evident, and decided to
launch another test program utilizing similar control system.
5.2 Hu and Golan

Hu and Golan (2003) investigated the benefit of active feedback control on production loss
in unstable gas lift wells. Using a dynamical simulator, the bottomhole flowing pressure was
controlled by the production choke to simulate the effects of feedback control on production.
17
By simulation Hu and Golan demonstrates that both casing heading and tubing heading
instability may be stabilized by the active feedback controller. For casing heading, the simu-
lation shows that 17 % of production can be saved using feedback control on the well. In the
case of tubing heading instability, 20 % of production is saved. Hu and Golan concludes that
active feedback control gives a new option to both stabilize the well and optimize production
(Hu and Golan, 2003).
5.3 Eikrem et al.

Eikrem et al. (2008) proposed three simple control structure to stabilize gas lift instability in
simulations and laboratory experiments. The authors investigated the use of downhole, casing-
head, and differential pressure as the set point for the controller to stabilize flow instability.
Both the simulations and the laboratory experiments gave promising results and all control
structures successfully stabilized production. Eikrem et al. claimed the findings shows that it is
possible to use different control structures to eliminate instability. In the event sensor failure or
other issues, it is possible to switch from one control structure to another (Eikrem et al., 2008).
18
Chapter 6
Dynamic gas lift model
Regarding existing stability criteria, experience seems to show that when they predict instabil-
ity, the gas lift well will most likely be unstable. When the criteria predict stability, however,
the well may turn out to be unstable in practice (Alhanati et al., 1993).
The existing stability criteria do not take into account the flow at the outlet of the produc-
tion pipe, and this has been a motivation for the work undertaken in the thesis. The model to
be tested considers the outflow variations explicitly, and it is reasonable to believe that this will
improve the prediction power.
6.1 General model

Hjalmars (1973) investigated the flow instability in airlift pumps by perturbation analysis. Fitre-
mann and Vedrines (1985) developed a numerical model of a gas lift well in similar manner, and
Asheim (1988) developed simple stability criteria for gas lift by considering how the system
respond to small perturbations. Since then, extensions and improvements have been suggested
to predict stability in gas lift wells (Fairuzov et al., 2004; Poblano et al., 2002; Alhanati et al.,
1993).
The existing criteria considers change due to current state and may be expressed as
d
X (t) = AX (t) . (6.1.1)
dt
Where X is a vector of state variables such as pressures or rates and t is the current time.
A is a matrix of coefficients relating the well design and reservoir parameters.
19
With active feedback control, control variables such as valve opening can continuously be
adjusted based on measurements. This constitute to an additional term in the model.
d
X (t) = AX (t) + BU (t) . (6.1.2)
dt
Where U is a vector containing the control variables and B is a matrix specifying how the
control variables affect the system.
6.2 Current model

Figure 6.2.1 represents the gas lift system considered. Gas is pumped down into the annulus
at a constant rate, and the produced fluids flow into a pressure-controlled separator. Thus,
the tubing head pressure may be considered constant. The lift-gas enters the tubing through a

downhole injection valve and mixes with the fluids from the reservoir. This reduces the density
of the fluid mixture and thereby lowers the pressure at the bottom of the well such that the
inflow from the reservoir increases.
Pth To seperator
Lift-gas
Annulus
Tubing
Casing
Pw Injection depth
Pwf
Figure 6.2.1: Simplified gas lift model.
20
6.2.1 Steady-state gas lift
The relations used apply to stationary/steady flow at given input pressure, temperature, and
flow rate. A steady-state flow model has been constructed, and along the production pipe, it is
assumed the mass flow is constant, while pressure and temperature change. The model enables
calculation of pressure changes along the production pipe, and thus also phase relationship,
volume rate, and velocity. The model accounts for slip between gas and liquid while liquid
holdup has been neglected. To match the model output with the provided well data, the slip
parameters are tuned. Although the model output matches the measured data, this does not
necessarily mean the flow modeling is correct. However, it provides a basis for estimating the
conditions down in the well.
A slightly different steady-state gas lift model has also been constructed. The model esti-
mates pressure and flow rates in the same manner as the model described above. However, for
certain downhole pressure, wellhead pressure, gas-oil ratio, and water cut the model estimates
the required oil rate for a given injection rate.
The mixture of injected gas and reservoir fluid flows up along the production tubing. At
steady-state flow, the pressure gradient in the well is estimated as in equation 6.2.1.
dp 1 ρT P 2
+ ρT P gx + fT P v = 0. (6.2.1)
dx 2 d m
The bottomhole pressure may be expressed by integrating equation 6.2.1 along the pro-
duction tubing.
1 ρT¯P 2
pw = pth + ρT¯P gx + fT¯P v¯m L. (6.2.2)
2 d
The inflow of reservoir fluid is expressed by the steady-state pressure drop between the
reservoir and the bottom of the well.
Ql
pw = pr − . (6.2.3)
Jw
Where Jw is the productivity index, indicating downhole rate. Lift-gas is injected at the
wellhead and flows down the annulus. Assuming the flow velocity is low, the pressure at the
injection depth x can be approximated by the barometric equation.
Mg x
pg (x) = pwh exp . (6.2.4)
zRT
21
At the depth of the injection point the lift-gas flows into the production tubing through an
orifice valve. For a constant valve opening, the pressure drop across the valve may be expressed
by the orifice equation.
1 ρg 2
∆pc = pg − pw = Q . (6.2.5)
2 A2c g
6.2.2 Dynamic response

The dynamic response is estimated by considering how the gas lift system at steady-state is af-
fected by small disturbances. The current model is developed and provided by Asheim (2017).
The model includes outflow variations and results in a time delay. The model structure may be
expressed as
d
X (t) = AX (t) + DX (t − ∆t) . (6.2.6)
dt
Where D is the delay-matrix depending on well design and reservoir parameters. A, t, and
x are defined as in equation 6.1.1.
Production tubing
The steady-state pressure drop along the tubing may be expressed by equation 6.2.2. Differ-
entiating equation 6.2.2 and perturbating gives the tubing pressure response due to change in
density of the fluid mixture.
2

ϑ ϑ 1 vm ϑ ρT¯P ϑ 1 ρT¯P ϑ 2
δpw = δptp + gx + fm L δ ρT¯P + fm vm L δvm + fm L δv .
ϑt ϑt 2 d ϑt d ϑt 2 d ϑt m
(6.2.7)
Density change will occur if the relationship between the flowing gas and liquid changes.
This can be quantified using the continuity equation, ϑ/ϑtρT P +ϑ/ϑx(ρm vm ). Integrating from
inlet to outlet, along the tubing, this gives
ϑ 1
δ ρT¯P = (ρg vsg + ρl vsl )x=0 − (ρg vsg + ρl vsl )x=L . (6.2.8)
ϑt L
The outflow, at x = L, is a mixture that can be quantified by the product of the density,
ρm = (ρg vsg + ρl vsl )/vm , and the mixture velocity, vm = vsg + vsl . Neglecting slip, the mixture
in tubing has the velocity vm and will therefore reach the outlet after time, ∆t = L/vm . The
relation in equation 6.2.8 may be expressed as
ϑ 1
δ ρT¯P = (ρg vsg + ρl vsl )t − (ρm vm )t−∆t . (6.2.9)
ϑt L
22
If the inflow changes the density and velocity of the fluid mixture will change. The ve-
locity change will propagate as pressure waves and will quickly affect the flow at the outlet.
Changed inflow gas-oil ratio and mixture density will follow the stream, thus propagate much
slower. Assuming changes in velocity are noticed immediately at the outlet, perturbation of
equation 6.2.9 gives
ϑ 1
δ ρT¯P = (ρg δvsg + ρl δvsl )t − (ρm vm )t − (ρm vm )t−∆t . (6.2.10)
ϑt L
According to the definitions of mixture density and velocity, perturbation gives
(vm δ (ρg vsg + ρl vsl ) − (ρg vsg + ρl vsl ) δvm )

∆ρm = 2
, (6.2.11)
vm
and
δvm = δvsg + δvsl . (6.2.12)
Substituting the above results into equation 6.2.7 relates the change of well pressure to the
inflow variation. To express in terms of volume rates, the superficial velocities are multiplied
by the tubing cross sectional area. Using the definition of flux fraction, λi = Qi /(Qg + Ql ),
this gives
ϑ ρT¯P ϑ ∆ρgx L
δpw = fm vm L δvm + (λg δQl − λl δQg )t − (λg δQl − λl δQg )t−∆t .
ϑt d ϑt Vt
(6.2.13)

ϑ ρT¯P ϑ ϑ ∆ρgx L
δpw = fm vm L δQg + δQl + (λg δQl − λl δQg )t − (λg δQl − λl δQg )t−∆t .
ϑt At d ϑt ϑt Vt
(6.2.14)
Annulus
Pressure change in the annulus will propagate with sonic speed. This gives such a rapid re-
sponse that the pressure can be considered in stationary equilibrium and expressed by the equa-
tion of state pg Va = znRT . At the wellhead, constant lift-gas rate is assumed to enter the
annulus. At steady-state conditions, equal rate of lift-gas flows out through the injection valve,
thus mass conservation. If the rate through the injection rate deviates, the amount of gas in the
annulus will change.
23
According to the equation of state, the pressure response becomes
ϑ pg
δpg = − δQg . (6.2.15)
ϑt Va
Where δQg express the deviation from steady-state injection rate at downhole temperature
and gas pressure, pg .
Inflow
Fluid flow from the reservoir may be expressed as in equation 6.2.3. Within shorter periods of
time , the reservoir pressure and productivity index may be considered constant. Perturbating
then gives the dynamic response
δQl = −Bo Jδpw = Jw δpw . (6.2.16)
The inflow of gas usually takes place through a nozzle at the injection valve. Perturbation
of equation 6.2.5 show that the pressure deviations, δpg and δpw , affect gas inflow as follows
δQg = Jg δpg − Jg δpw . (6.2.17)
Where,
A2c
Jg = . (6.2.18)
ρg Qg
Here, Ac is the orifice opening and Jg the "gas inflow index".
System response
The dynamic response of the gas lift system results from interaction between tubing and annu-
lus. By substituting equation 6.2.17 into equation 6.2.15 provides the annular pressure response
in terms of pressure deviation.
ϑ pg
δpg = − (Jg δpg − Jg δpw )t . (6.2.19)
ϑt Va
By substituting equation 6.2.16 and 6.2.17 into 6.2.14 provides the tubing response.

ϑ ρT¯P L ϑ ρT¯P L ϑ
δpw = aw δpw −ag δpg −fm vm (Jg +Jw ) δpw +fm vm Jg δpg
ϑt d At ϑt d At ϑt t
(6.2.20)
− [aw δpw − ag δpg ]t−∆t .
24
By combining equation 6.2.15 and 6.2.17 the tubing response may be further expressed as

ϑ ρT¯P L ϑ ρT¯P L
δpw = aw δpw − ag δpg − fm vm (Jg + Jw ) δpw + fm vm Jg (cδpw − cδpg )
ϑt d At ϑt d At t
− [aw δpw − ag δpg ]t−∆t .

(6.2.21)
By expressing the parameter groups in terms of response coefficients, equation 6.2.19 and
6.2.21 becomes
ϑ pg
δpg = − (Jg δpg − Jg δpw )t = (cδpw − cδpg )t , (6.2.22)
ϑt Va
and
ϑ
(1 + ft Jg + ft Jw ) δpw = (aw + cft Jg )t δpw − (ag + cft Jg )t δpg − (aw δpw − ag δpg )t−∆t .
ϑt
(6.2.23)
The response of the gas lift system may described by the two equations above. With
linearization around a steady-state solution of the system, the coefficients become constants.
On matrix form, the two equations form a delay-differential equation given as
       
ϑ δp w a + cf J − (a + cf J ) p a −a p
−1  = w t g g t g
  w −  w g
  w (6.2.24)
ϑt δp c −c pg 0 0 pg
g
t t−∆t
Where = (1 + ft Jg + ft Jw ). Setting a1 = (aw + cft Jg ), b1 = (aw + cft Jg ), a2 =

aw , and b2 = ag the differential equation may be expressed as
       
ϑ δpw  a1 −b1  pw  a2 −b2 p
  w
= − (6.2.25)
g
t t−∆t
The first part of equation 6.2.25 refers to the current time t and considers the variations
downhole at the injection depth. The second part, t − ∆t, accounts for the kinematic delay
between the inflow and outflow. Mixture variations generated downhole will propagate along
the tubing and exit at a later time. The changes are accounted for by the delay term.
The system parameters , aw , ag , c, ft , Jg , and Jw relate functionally to reservoir, fluid
properties and well design. Given fluid and reservoir properties and gas lift design, the response
parameters and matrix coefficients may be estimated. The dynamic response of the gas lift
system may then be estimated from equation 6.2.25.
25
6.3 Comparison with existing criteria
The derivation of the current model is based on the perturbation method by Asheim (1988), but
contain some differences and improvements. Most important, the model includes the delayed
response that represents outflow density variations. Also, the suggested model accounts for
flow friction along the production tubing. The annular response, however, is the same to what
Asheim (1988) considered in his criteria. If the delayed response is neglected, equation 6.2.24
simplifies as
    
ϑ δpw   (aw + cft Jg ) − (ag + cft Jg ) pw 
= (6.3.1)
ϑt δp c −c pg
g
t
Neglecting flow friction gives ft = 0 and leads to, = 1, a1 = aw , and b1 = ag . Equation

6.3.1 may than be written as
    
ϑ δpw  aw −ag  pw 
= (6.3.2)
ϑt δp c −c pg
g
t
The expression above represents a homogeneous linear equation, and the stability of such
systems is completely determined by its eigenvalues (Hirch and Smale, 1974). The eigenvalues
may be expressed as

1
q
λ= (aw − c) ± (aw − c)2 − 4c(ag − aw ) . (6.3.3)
2
The eigenvalues may be real or complex of the form λ = α + iω. In many cases the
eigenvalues are complex, and stability of the system requires all eigenvalues to have negative
real part, α = aw − c < 0. Negative real part implies damped oscillations and thus stable
system. Inserting the expressions for aw and c into the inequality gives (Asheim, 2016)
A2c Ql 1 pg A2c

1 ∆ρgx L
− Jw Qg > . (6.3.4)
Vt Qg + Ql ρg Qg Va ρg Qg
The above equation imitates the depletion response, equation 4.2.2, proposed by Asheim
(1988) and shows the model reduces to existing criteria under certain considerations. The
inequality illustrates how important well and reservoir parameters affect stability in gas lift
wells. The purpose of the comparison is simply to show that the current model is an extension of
existing stability criteria and that some modifications have been made to improve its prediction
power hopefully. In the following chapters, however, the delay-differential equation 6.2.25
form the basis for the gas lift stability analysis.
26
Chapter 7
Gas lift stability study
The analysis in the subsequent chapters investigate the stability of a gas lift well and is intended
to check the suggested model. The theoretical model is a simplification, and the dynamic
response predicted by the model needs to be checked against measured data.
The characteristics of different production modes will be discussed, and estimates of static
and dynamic parameters are presented. The initiation of flow instability is examined in detail,
and the wells ability to oppose variations of particular characteristics are explored. The findings
provide the basis for further analysis, and in Chapter 8 the theoretical model is applied to the
production data.
7.1 The Heidrun case

The provided data consists of measurements from well 6507/7-A-23 in the Åre formation,
located in the Heidrun field in the Norwegian Sea. The Åre formation consists of lower and
middle Jurassic sandstone, and the reservoir is heavily faulted. The primary recovery strategy
in the formation is water injection The well is operated by Statoil and has been producing oil
since late 2000. Due to declining reservoir pressure and reduced deliverability, the well was
put on gas lift in 2010 to improve production (Norwegian Petroleum Directorate, 2017). Since
early 2016, the well has occasionally been subjected to flow instabilities.
7.1.1 Data availability

The provided well data were measured throughout the year 2016 and January-February 2017,
and consist of measurements taken once every minute. The data include measurements of
temperatures, pressures, surface flow rates, and choke settings. Temperature and pressure data
are available at the wellhead and downhole conditions. The pressure in the annulus and the
gas lift supply pressure, are also provided. The flow rate measurements were acquired using a
multiphase flow meter and include lift-gas, gas, oil, and water rates at surface conditions.
27
The accuracy of the multiphase meter is unknown. However, the author is aware of the
issue and that there might be inconsistencies between measured and actual rates.
Data on the particular reservoir are scarce in the literature. However, the necessary fluid
properties have been provided by Statoil. The lift-gas has been assumed to have the same
properties as the produced reservoir gas.
7.1.2 Well design and completion

The well is deviated, and the total measured depth (MD) is 3022 meters. The true vertical depth
(TVD) is 2485 meters. The well is perforated at approximately 2900 meters MD. The downhole
temperature and pressure gauge are set at 2670 meters MD. The well is equipped with a 1.5"
orifice injection valve with a 23/64" nozzle installed at 1826 meters MD. The well is completed
with a 5.5" tubing and a 9 5/8" production casing. Figure 7.1.1 shows a simplified model of the
well. The figure is neither in the correct scale nor deviated and is only intended to illustrate the
main components of the gas lift system. The schematic also shows the position at the wellhead

and downhole at which the pressure and temperature measurements were acquired, indicated
with PT/TT.
PT
TT
Pwh
Production
Twh
Lift-gas inflow
5.5’’ production
tubing
7 5/8’’ production
casing
Injection depth
23/64’’ injection
1826 m MD
orifice
PT
TT
Pw
Measurement depth
Tw
2670 m MD
Pwf Perforation depth

2900 m MD
Figure 7.1.1: Well 6507/7-A-23 schematic.
28
7.2 Production measured
The provided data have been split into sub-periods, on a two-month basis, to give a better
overview of the production characteristics. The stability analysis focuses on three out of the in
total seven sub-periods of measurements. The selected periods consist of March-April 2016,
May-June 2016, and January-February 2017. These data sets have been selected for further
analysis because they contain different production characteristics and thus provide a basis for
examining the well under various conditions.
The figures that follow show measured liquid (oil and water) rates. The purpose is to give
the reader a quick overview of what is available and may be expected in the well. The red and
blue dotted lines illustrate the standard deviation and the mean of production respectively.
Figure 7.2.1-7.2.3 below and on the next page show periods of varying production rates.
Alternating periods of stationary production appear to develop sudden oscillations. Often, the
oscillations are only temporary, and production returns to normal after a couple of hours. At
other times, however, the system develop oscillations that are more persistent and last for sev-
eral days to weeks.
Total liquid production: Mar-Apr 16

Data
Mean
4000 Std
3500
3000
Flow rate, Sm3/d
2500
2000
1500
1000
500
0
0 200 400 600 800 1000 1200 1400
Time,hrs
Figure 7.2.1: Liquid production March - April 2016.
29
Total liquid production: May-June 16
4000
Data
Mean
3500 Std
3000
Flow rate, Sm3/d

2500
2000
1500
1000
500
0
0 200 400 600 800 1000 1200 1400
Time,hrs
Figure 7.2.2: Liquid production May - June 2016.
Total liquid production: Jan-Feb 17

3000
Data
Mean
Std
2500
2000
Flow rate, Sm3/d
1500
1000
500
0
0 200 400 600 800 1000 1200 1400
Time,hrs
Figure 7.2.3: Liquid production January - February 2017.
An interesting observation is that significant oscillations occur almost suddenly and with-
out any gradual buildup. The measurements also consist of intervals of what seems to be
stationary production, of relatively long duration. The different intervals facilitate for compari-
son between various production modes and may enable us to understand under what conditions
the system becomes unstable. Thus, it could be of interest to examine intervals of production
shortly before oscillations develop in the system.
At times the data show intervals in which the well is shut-in after periods of severe oscilla-
tions. The production stop may be due to operational problems in the processing facilities and
other surface equipment, or an attempt to gain control over the well.
30
7.3 Characterization and parameter estimation
To characterize the production in the well, data in the interval January-February 2017 have been
selected for a more in-depth study. Figure 7.3.1 illustrates the liquid production for the period.
The period seems to include ranges of different production modes, and this makes it possible
to study the behavior of the system under different circumstances. The various conditions
also allow for estimation of system parameters and may provide a significantly better insight
than analysis based on only a single production mode. The period January-February 2017
seems to be comprised of ranges of stationary production, static pressure build-up/shut-in, and
oscillating production.
Total liquid production: Jan-Feb 17

3000
Data
Mean
Std
2500
2000
Flow rate, Sm3/d
1500
1000
500
0
0 200 400 600 800 1000 1200 1400
Time,hrs
Figure 7.3.1: Liquid production January - February 2017.
The production measurements January-February 2017 seems to include the following pro-
duction modes:
• Stationary production: 0 - 400 hours.

• Oscillating production: 680 - 710 and 785 - 805 hours.
• Static pressure build-up: 1177 - 1190 hours.
31
7.3.1 Stationary production
Figure 7.3.2 shows the liquid rate variation during the first three days of January 2017. An
interval of 72 hours has been selected to visualize the variations around the mean of the pro-
duction better. The mean and the standard deviation of the liquid rate in the interval t = 0-400
hours is estimated to be 819.4 Sm3 /d and 26.9 Sm3 /d respectively.
Total liquid rate: 1-3. January 17

1000
Liquid rate
Mean
950 Std
900
Flow rate, Sm3/d
850
800
750
700
650
600
0 10 20 30 40 50 60 70
Time,hrs
Figure 7.3.2: Liquid production rate 1-3. January 2017.
To check for stationarity, figure 7.3.3 shows the probability density function (pdf) of the
flow rate data in the assumed stationary production interval. The measurements exhibit the
characteristic bell shape of normally distributed data. The red line is a theoretical pdf fitted to
the data with same mean and standard deviation. The good fit indicates the measurements are
close to perfect normal distributed and that the data vary stochastic around the mean.
PDF Total liquid rate 1-16. January 17

0.018
Data
N dist. fit
0.016
0.014
Probability density function
0.012
0.01
0.008
0.006
0.004
0.002
0
720 740 760 780 800 820 840 860 880 900 920
Flow rate, Sm3/d
Figure 7.3.3: Probability density function liquid production rate 1-16. January 2017.
32
Analysis of the frequency spectrum does not reveal any stable oscillation present in the
data. Based on the discovery above, the production may be considered stationary during the
first 17 days of January 2017. It is worth noting the average fluids production in the interval
t = 0-400 hours was 819.4 Sm3 /d and the average in the interval t = 0-72 hours was 820.0
Sm3 /d. The closeness supports the assumption of stationary production in the first 400 hours
of January 2017.
Stationary reference
The stationary production period provides a basis for estimating parameters in the well using
the steady-state gas lift model. The stationary reference is built on averages of the production
data in the interval t = 0-400 hours. Table 7.3.1 summarizes the input parameters for the steady-
state model.
Table 7.3.1: Input steady-state model 1-16. January 2017
Parameter
Wellhead pressure, [bar] 22.3
Downhole pressure, [bar] 122.5
Oil rate, [Sm3 /d] 263.3
Water rate, [Sm3 /d] 556.1
Gas rate, [Sm3 /d] 1.38 105
Injection rate, [Sm3 /d] 8.85 104
Gas injection pressure wellhead, [bar] 118.9
Injection orifice diameter, [mm] 9.1
Figure 7.3.4 on the next page shows estimated pressure profiles in tubing and annulus
for the rate and pressure conditions in the table above. Adjusting the slip parameters gives a
pressure of 122.5 bars at the downhole measuring point. This is the same as the average of the
measured well pressure. Injection of lift-gas at a depth of the injection valve (1826 meters MD)
will reduce the pressure gradient up the production tubing, and this may be observed from the
change in the blue pressure curve.
The tubing pressure at the injection depth is estimated to 68.6 bars, and the pressure at the
bottom of the well is estimated to 138.2 bars. The figure also shows pressure in the annulus
for injection pressure at the wellhead as given above. It is assumed gas-filled annulus down
to the injection valve and liquid-filled below. At the injection depth, this provides a pressure
difference annulus-tubing of 64.6 bars.
33
According to the orifice formula, equation 6.2.5, injection of lift-gas through the given
orifice size should cause a pressure drop of 7.6 bars. The inconsistency of the two pressure
differences is obvious and may be due to errors in estimating the pressure gradient in the an-
nulus or an error in the provided data. In the model calculations later in the report, however,
the pressure drop given by the orifice equation has been assumed correct and will be used as a
reference for the pressure in the annulus.
Stationary pressure profile - Well 6507/7-A-23

0
Tubing pressure
Annulus pressure
Injection depth
-500 Measurement depth
Reservoir depth
Estimated annulus pressure
- orifice equation
Depth below RKB, m
-1000
-1500
-2000
-2500
-3000
0 50 100 150 200 250
Pressure, bar
Figure 7.3.4: Stationary pressure profile for production tubing and annulus.
Stationary productivity index
Based on stationary production, interval t = 0-400 hours, the stationary productivity index in the
well may be estimated. During stationary production, the gas lift model estimates the pressure
at the bottom of the well to be 138.2 bars. Average liquid production was 819.4 Sm3 /d.
From the pressure build-up period (section 7.3.2), the reservoir pressure has been approx-
imated to around 171 bars. When the measurements were acquired the stationary production
and static pressure buildup was close in time. Thus the reservoir pressure was likely to be
similar. Utilizing the steady-state inflow relationship gives the resulting stationary productivity
index
3 3
ql 819.4 Sm
d
Sm
J= = = 25.06 d . (7.3.1)
pr − pw (171.0 − 138.2)bar bar
34
Analysis of other periods of steady production gives similar values of the productivity
index. Considering the uncertainty in the flowing bottomhole and reservoir pressure estimates
it may be assumed that 25 Sm3 /d/bar is a good representation of the well deliverability.
7.3.2 Static pressure build-up

Figure 7.3.5 shows measured well pressure 14-19. February 2017, interval t = 1100-1200 hours.
The measurements show the well was shut in for 13 hours 18. February. The static pressure
build-up provides a basis for estimating the reservoir pressure in the surrounding formation.
Well Pressure: 14-19. February 17

155
Pressure
150
145
140
Pressure, bar
135
130
125
120
115
1100 1110 1120 1130 1140 1150 1160 1170 1180 1190 1200
Time,hrs
Figure 7.3.5: Measured well pressure 14-19. February 2017.
Figure 7.3.6 on the next page shows measured downhole pressure and estimated bottom-
hole pressure. The estimate assumes the gas in the production tubing rises rapidly when the well
is not producing, such that the production tubing between the bottom of the well and measuring
point only contains static liquid. By utilizing the steady-state model, the average density of the
liquid phase between the two points is estimated to be 983.2 kg/m3 . The estimate is based on
the stationary reference case in table 7.3.1. A liquid density of 983.2 kg/m3 corresponds to a
static pressure difference of 22.2 bars between the measuring point and the perforations.
35
Shut-in Well Pressure: 18. February 17
180
Measured well pressure
Estimated bottomhole pressure
170
160
Pressure, bar
150
140
130
120
1176 1178 1180 1182 1184 1186 1188 1190
Time,hrs
Figure 7.3.6: Measured and estimated well pressure 18. February 2017.
Figure 7.3.6 indicates an initial transient, possibly due to redistribution of gas and liquid
in the wellbore. After that, a steady pressure buildup to 171 bars. The estimate is based only on
measured pressure data and other simplifying assumptions using the steady-state model. Thus,
there is lots of uncertainty in the estimate, and the static reservoir pressure is probably a little
higher. Considering the information available it has been determined the estimate is sufficient
for further analysis.
7.3.3 Oscillating production

Figure 7.3.1 show several intervals in which the production oscillate and illustrate considerable
variation from the mean value. Figure 7.3.7 on the next page shows the production rate as
measured in the intervals 28-29. January and 1-2. February, the intervals t = 680-710 and t =
785-805 hours respectively. To better visualize the oscillations only five hours of production
28. January and 1. February has been plotted. The trend of the oscillations, however, remains
constant throughout the two intervals mentioned.
The oscillating nature of the liquid rate is obvious, and the amplitudes appear to be rather
constant during the periods. The production in the interval t = 680-710 was eventually shut-in
for a short time before the well was put on production again. The shutdown was probably due to
operational problems and an attempt to control production. Then, a period of stable production
follows before the well start to oscillate once again, t = 785-805 hours.
36
Total liquid rate: 28. January 17 Total liquid rate: 1. February 17
2500 2500
Liquid rate Liquid rate
Mean Mean
Std Std
2000 2000
Flow rate, Sm3/d

Flow rate, Sm3/d
1500 1500
1000 1000
500 500
0 0
680 680.5 681 681.5 682 682.5 683 683.5 684 684.5 685 785 785.5 786 786.5 787 787.5 788 788.5 789 789.5 790
Time,hrs Time,hrs
Figure 7.3.7: Oscillating liquid production January-February 2017.
Visual inspection of production 28. January and 1. February 2017 show clear similarities.
Although the well was shut-in after the first period the production quickly started to oscillate
similarly. Table 7.3.2 below summarize the production characteristics for the two periods. The
mean of production increases slightly while the standard deviation decreases.
Table 7.3.2: Production statistics - oscillating production.
Period Mean production [Sm3 /d] Std production [Sm3 /d]

28-29. January 2017 827.0 634.4
1-2. February 2017 858.6 617.9
The standard deviations tell there is a significant variation in the liquid production for both
periods. To better examine the oscillations the data have been transformed to the frequency do-
main using the Fourier Transform. Figure 7.3.8 show the amplitude spectrum of the measured
flow rates. The figure show dominant amplitudes of about 483 Sm3 /d and 623 Sm3 /d. Both
periods oscillate at equal frequencies, 5.66 hr−1 , with a duration of 10.6 minutes.
Amplitude spectrum, Liquid rate; 28-29. January 17 Amplitude spectrum, Liquid rate; 1-2. February 17
600 700
600
500
500
400
Amplitude, Sm3/d
Amplitude, bar
400
300
300
200
200
100
100
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Period, minutes Period, minutes
Figure 7.3.8: Amplitude spectrum liquid rates January-February 2017.
37
For gas lift systems subjected to flow instability, pressures and flow rates usually oscillate
at equal frequency, but with some phase-shift. Figure 7.3.9 shows the measured pressure at the
wellhead and downhole for five hours 28. January 2017. As for the liquid rate, the pressures
seems to oscillate with a rather constant amplitude around the mean.
Figure 7.3.10 shows the pressure amplitude spectrum’s 28-20. January and reveal dom-
inant amplitudes of around 7 and 2 bar for the wellhead and downhole pressure respectively.
The oscillatory period for both the wellhead and downhole pressure is 10.6 minutes. The same
as for the liquid rate, as expected. Investigation of the pressures 1-2. February 2017 show same
attributes.
Wellhead pressure: 28. January 17 Downhole pressure: 28. January 17
55 130
Pressure Pressure
Mean Mean
50 Std Std
128
45
126
40
Pressure, bar
Pressure, bar
35 124
30 122
25
120
20
118
15
10 116
680 680.5 681 681.5 682 682.5 683 683.5 684 684.5 685 680 680.5 681 681.5 682 682.5 683 683.5 684 684.5 685
Time,hrs Time,hrs
Figure 7.3.9: Downmhole and wellhead pressure oscillations 28. January 2017.
Amplitude spectrum, Wellhead pressure; 28-29. January 17 Amplitude spectrum, Downhole pressure; 28-29. January 17
8 3
7
2.5
2
5
Amplitude, bar
Amplitude, bar
4 1.5
3
1
0.5
1
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Figure 7.3.10: Amplitude spectrum pressure oscillations 28-29. January 2017.
Rate and pressure variations
The discovery above reveals constant oscillations in both pressure and flow rate when severe
flow instability develops. In the two periods investigated the entire gas lift system seems to
oscillate at a dominant frequency, affecting both the production rate and pressure in the well.
The figures on the next page illustrate rate and pressure variations for the two periods examined.
38
Figure 7.3.11a and 7.3.12a shows the variations in liquid production and injection rate.
Figure 7.3.11b and 7.3.12b shows how the downhole pressure and injection rate varies. Fre-
quency analysis reveals that all rates and pressures oscillate at the same frequency. The oscil-
lation period may also be visualized from the figures, but with phase lags. An interval of two
hours has been used to visualize the oscillations better.
Liquid production and injection rate 28. January 17 Downhole pressure and injection rate 28. January 17
2000 2000 2000 5
Liquid rate Injection rate Well pressure Injection rate
1500 1500 1500 3.75
Deviation injection rate, Sm3/d

1000 1000 1000 2.5
Deviation well pressure, bar

Deviation liquid rate, Sm3/d
500 500 500 1.25
0 0 0 0
-500 -500 -500 -1.25
-1000 -1000 -1000 -2.5
-1500 -1500 -1500 -3.75
-2000 -2000 -2000 -5

685 685.2 685.4 685.6 685.8 686 686.2 686.4 686.6 686.8 687 685 685.2 685.4 685.6 685.8 686 686.2 686.4 686.6 686.8 687
Time, hrs Time, hrs
(a) Liquid production and injection rate. (b) Downhole pressure and injection rate.
Figure 7.3.11: Rate and pressure variations 28. January 2017.
Liquid production and injection rate 1. February 17 Downhole pressure and injection rate 1. February 17
2000 2000 2000 5
Liquid rate Injection rate Well pressure Injection rate
1500 1500 1500 3.75

1000 1000 1000 2.5
Deviation well pressure, bar

Deviation liquid rate, Sm3/d
500 500 500 1.25
0 0 0 0
-500 -500 -500 -1.25
-1000 -1000 -1000 -2.5
-1500 -1500 -1500 -3.75
-2000 -2000 -2000 -5

785 785.2 785.4 785.6 785.8 786 786.2 786.4 786.6 786.8 787 785 785.2 785.4 785.6 785.8 786 786.2 786.4 786.6 786.8 787
Time, hrs Time, hrs
(a) Liquid production and injection rate. (b) Downhole pressure and injection rate.
Figure 7.3.12: Rate and pressure variations 1. February 2017.
Other periods in January 2017 subjected to oscillations show similar behavior and oscilla-
tions of more or less the same frequency present in the measurements.
The length of the production pipe from the injection point is 1826 meters, and the steady-
state well model estimates the flow speed to be around 3.4 m/s. The flow time from the inlet
to the outlet then becomes, ∆t ≈ 9 minutes. Thus, oscillation period is of order-of-magnitude
similar to flow time along the tubing. This indicates that the oscillations might be due to the
varying inflow of gas-oil ratio, suggesting kinematic waves as an underlying mechanism of the
instability (Asheim, 2016).
39
Dynamic productivity index
When the pressure in the well varies, the inflow of reservoir fluid will also vary. The produc-
tivity index, equation 7.3.2, quantifies the dependence of well pressure on the liquid inflow
rate.
ql
J= (7.3.2)
pr − pwf
However, when pressure and flow rate oscillate around an average, it is more practical to
evaluate the pressure-rate dependence by the relationship between the variations, a dynamic
productivity index (Asheim, 2000). The dynamic productivity index is defined as the ratio
of the change in inflow rate to bottom well pressure. The expression can be derived using
Darcy’s equation, and it can be shown that the dynamic productivity (at established oscillations)
index depend on the frequency, not the amplitude. There will also be a phase shift relative to
stationary production. At stationary production, the maximum inflow rate is achieved when the
well pressure is at a minimum. At oscillating production, however, maximum production is
achieved at maximum well pressure (Asheim, 2000).
If we neglect the phase difference, the dynamic productivity may be approximated as
ql (t) − q̄l δql Aql

j= = ≈ (7.3.3)
pw (t) − p̄w δpw Apw
Where Aqw and Apw are the amplitude of the oscillations in flow rate and well pressure,
respectively. Substituting in the amplitudes obtained 28-29. January and 1-2. February we
get values of around 254 Sm3 /d/bar and 260 Sm3 /d/bar. The dynamic productivity index
is approximately ten times higher than the stationary productivity index. The magnitude of
the productivity index tells us that liquid production responds much stronger to oscillating
pressure variations than stationary pressure variations. This may be explained due to fluid
expansion/compression around the wellbore.
40
7.3.4 Well stabilization
Both periods examined above show oscillatory production, separated by an interval that ap-
pears to contain stationary production. Figure 7.3.13 shows the production was shut-in for
approximately 1 hour following the unstable production 28-29. January 2017.
Total liquid rate: 30. January 17
2000
Liquid rate
1800
1600
1400
Flow rate, Sm3/d
1200
1000
800
600
400
200
0
730 731 732 733 734 735 736 737 738 739 740
Time,hrs
Figure 7.3.13: Liquid production rate 30. January 2017.
When production is resumed, the analysis shows that the well produces stably, comparable
to production the first 400 hours at the beginning of the month. Thus, temporary shutdown of
production may stabilize the well. After only 35 hours, however, oscillations start again.
7.3.5 Initiation of instability

There might be minor variations and disturbances in the system which later leads to larger
oscillations in the tubing. Two different periods, 20-23. March 2016 and 6-9. May 2016, have
been inspected to give an explanation to how and why oscillations initiate in the well. In this
way, it is ensured that the conditions and dynamics that leads to oscillations are not random,
but something that recurs before the instability.
Well production 20-23. March 2016
Figure 7.3.14 and 7.3.14 on the next page shows pressure and production rates 23. March 2016.
The plots illustrate how the conditions in the well change when production goes from stable to
highly variable. The red line shows an exponential moving average fitted to the data.
41
Total liquid rate: 23. March 16: 560-565 hr Gas production and injection rate: 23. March 16: 560-565 hr
3000 9
Flow rate
Gas production
Moving average Injection rate
8 Moving average
2500
7
Flow rate, 105 Sm3/d

2000 6
Flow rate, Sm3/d
5
1500
4
1000 3
2
500
1
0 0
560 560.5 561 561.5 562 562.5 563 563.5 564 564.5 565 560 560.5 561 561.5 562 562.5 563 563.5 564 564.5 565
Time,hrs Time,hrs
Figure 7.3.14: Liquid production rate and gas production and injection rates 23. March 2016.
Downhole pressure: 23. March 16: 560-565 hr Wellhead pressure: 23. March 16: 560-565 hr
129 45
Pressure Pressure
Moving average Down-stream choke pressure
128 Moving average
40
127
35
126
Pressure, bar
Pressure, bar
125
30
124
25
123
122
20
121
15
120
119
560 560.5 561 561.5 562 562.5 563 563.5 564 564.5 565 10
560 560.5 561 561.5 562 562.5 563 563.5 564 564.5 565
Time,hrs
Time,hrs
Figure 7.3.15: Downhole and wellhead pressures 23. March 2016.
The moving average captures the change in the mean by giving more weights to the most
recent measurements and provide an indication of the trend as the rate and pressure changes. An
exponential moving average has been used because it reacts more quickly to changes compared
to other moving averages.
Figure 7.3.14 and 7.3.15 show the oscillations are initiated at approximately 562 hours
and as the oscillations grow it may be observed that the average well pressure goes up by
about 4 bars. The average pressure at the wellhead declines slightly from 29 bars to 26 bars.
The moving average of the liquid and gas production rates remain rather constant, but the
amplitude of the variations grow large. The data show the conditions change dramatically over
a relatively short period. Oscillations seem to be well established after 30 minutes, from 562 to
562.5 hours.
42
Production history March 2016
To investigate the cause of the emerging oscillations the production history from 20-23 March
2016 has been further examined. Figure 7.3.16 and 7.3.17 show lift-gas delivery pressure at the
wellhead, gas injection pressure, liquid rate, and injection rate of lift-gas for the period. Com-
pared to the oscillations observed above, these measurements show less variation and smaller
amplitude. At times, however, the amplitudes appear to be somehow constant, and this may
suggest that the variations oscillate repetitively.
Delivery pressure: 20-23. March 16: 500-560 hr Injection pressure: 20-23. March 16: 500-560 hr
152 131.5
Delivery pressure Injection pressure
Moving average Moving average
150 131
130.5
148
130
146
Pressure, bar
Pressure, bar
129.5
144 129
128.5
142
128
140
127.5
138
127
136 126.5
500 510 520 530 540 550 560 500 510 520 530 540 550 560
Time,hrs Time,hrs
Figure 7.3.16: Gas delivery and injection pressures 20-23. March 2016.
Injection rate: 20-23. March 16: 500-560 hr Total liquid rate: 20-23. March 16: 500-560 hr
1.2 1040
Injection rate Liquid rate
1.15
1020
1.1
1.05
1000
Flow rate, Sm3/d
0.95 980
0.9
960
0.85
0.8
940
0.75
0.7 920
500 510 520 530 540 550 560 500 510 520 530 540 550 560
Time,hrs Time,hrs
Figure 7.3.17: Gas injection and liquid production rates 20-23. March 2016.
The figures on the next page illustrate the amplitude spectrum’s and clarify the variations
observed in figure 7.3.16 and 7.3.17. The delivery pressure, injection pressure, and gas injection
rate appears to oscillate at the same frequency, with an oscillation period of 12.5 minutes.
Although the amplitudes are small, they seem to be consistent. The amplitude of the injection
rate is considerably larger than the other amplitudes. This is because the injection rate is in the
order of 105 Sm3 /d. The amplitude spectrum of the liquid rate, however, appear more chaotic
and shows no stable oscillation frequency.
43
Even though the injection rate oscillates with a relatively large amplitude, the liquid pro-
duction still appears more or less stationary at this time.
Amplitude spectrum, Delivery pressure 20-23. March 16: 500-560 hr Amplitude spectrum, Injection pressure 20-23. March 16: 500-560 hr
1 0.25
0.9
0.8 0.2
0.7
Amplitude, bar
Amplitude, bar
0.6 0.15
0.5
0.4 0.1
0.3
0.2 0.05
0.1
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Figure 7.3.18: Amplitude spectrum gas delivery and injection pressures 20-23. March 2016.
Amplitude spectrum, Injection rate 20-23. March 16: 500-560 hr Amplitude spectrum, Liquid rate 20-23. March 16: 500-560 hr
2500 4
3.5
2000
3
Amplitude, Sm3/d
Amplitude, Sm3/d
2.5
1500
1000
1.5
1
500
0.5
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Figure 7.3.19: Amplitude spectrum gas injection and liquid production rates 20-23. March 2016.
Figure 7.3.20 on the next page shows liquid production 23. March ten hours later, t =
570-590 hours. The oscillations seem to be well established and have a period of 9.5 minutes
and a dominant amplitude of around 400 Sm3 /d. Examination of the well pressures indicates
oscillations of the same frequency.
44
Total liquid rate: 23. March 16: 570-590 hr Amplitude spectrum, Liquid rate 23. March 16: 570-590 hr
3000 500
Liquid rate
Mean 450
Std
400
350
2000
Amplitude, Sm3/d
Flow rate, Sm3/d
300
250
200
1000
150
100
50
0 0
570 572 574 576 578 580 582 584 586 588 590 0 5 10 15 20 25 30
Time,hrs Period, minutes
Figure 7.3.20: Oscillating liquid production rate 23. March 2016.
In the period from 20-23. March the gas delivery system show small variations with an
oscillation period of 12.5 minutes while the liquid rate shows no stable oscillations. A few
hours later, however, the liquid rate develops severe oscillations with a period of 9.5 minutes.
The difference in frequency might be due to the characteristics of the system and the erratic
nature of the multiphase flow. Although the gas delivery system experiences variations of a
particular period it does not necessarily imply the flow in the tubing will develop oscillations
of the same frequency. This has to do with the system’s natural frequency.
However, the measurements in figure 7.3.16 and 7.3.17 show the variations in the gas
delivery system level off just before (around five to ten hours) the oscillations in the tubing
develop. Closer analysis of the interval 557-562 hours confirms this. Also, small oscillations in
the liquid rate seem to start to develop in this interval. This may indicate that the disturbances
observed earlier in the gas delivery system have already initiated smaller flow instabilities be-
fore the large oscillations develop.
Well production 6-9. May 2016
The figures on the next page show pressures and production rates 9. May 2016. The conditions
in the well show similarities with the measurements observed in March.
The average well pressure goes up by about 8 bars, and the average pressure at the well-
head decline from 23 to 20 bar, analogous to the period in March. The oscillations seem to be
established after a relatively short time.
45
Total liquid rate: 9. May 16: 218-223 hr Gas production and injection rate: 9. May 16: 218-223 hr
4000 12
Flow rate Gas production
Moving average Injection rate
3500 Moving average
10
3000

8
Flow rate, Sm3/d
2500
2000 6
1500
4
1000
2
500
0 0
218 218.5 219 219.5 220 220.5 221 221.5 222 222.5 223 218 218.5 219 219.5 220 220.5 221 221.5 222 222.5 223
Time,hrs Time,hrs
Figure 7.3.21: Liquid production rate and gas production and injection rates 9. May 2016.
Wellhead pressure: 9. May 16: 218-223 hr

Downhole pressure: 9. May 16: 218-223 hr 40
130
Pressure
Pressure Down-stream choke pressure
Moving average
128 Moving average
35
126
124 30
Pressure, bar
Pressure, bar
122
25
120
118 20
116
15
114
112
218 218.5 219 219.5 220 220.5 221 221.5 222 222.5 223 10
Time,hrs 218 218.5 219 219.5 220 220.5 221 221.5 222 222.5 223
Time,hrs
Figure 7.3.22: Downhole and wellhead pressures 9. May 2016.
Production history May 2016
Figure 7.3.23 and 7.3.24 on the next page shows lift-gas delivery pressure at the wellhead,
gas injection pressure, liquid rate, and injection rate of lift-gas in the period 6-9. May 2016.
Compared to the measurements 20-23. March, the variations appear somewhat smaller and
maybe less consistent.
46
Delivery pressure: 6-9. May 16: 160-218 hr Injection pressure: 6-9. May 16: 160-218 hr
147 131.5
Moving average 131.4 Moving average
146.8
146.6 131.3
146.4 131.2
Pressure, bar
Pressure, bar
146.2 131.1
146 131
145.8 130.9
145.6 130.8
145.4 130.7
145.2 130.6
145 130.5
160 170 180 190 200 210 220 160 170 180 190 200 210 220
Time,hrs Time,hrs
Figure 7.3.23: Gas delivery and injection pressures 6-9. May 2016.
Injection rate: 6-9. May 16: 160-218 hr Total liquid rate: 6-9. May 16: 160-218 hr
1 1060
Injection rate Liquid rate
Moving average 1050 Moving average
0.995
0.99 1040
0.985 1030
Flow rate, Sm3/d
0.98 1020
0.975 1010
0.97 1000
0.965 990
0.96 980
0.955 970
0.95
160 170 180 190 200 210 220 960
Time,hrs 160 170 180 190 200 210 220
Time,hrs
Figure 7.3.24: Gas injection and liquid production rates 6-9. May 2016.
Figure 7.3.25 and 7.3.26 on the next page show the corresponding amplitude spectrum’s.
The spectrum’s show a dominant amplitude at equal frequencies. The oscillation period is
estimated to be 13.3 minutes and is slightly higher than for the data examined in March.
An interesting observation can be made from the amplitude spectrum of the liquid pro-
duction rate. Figure 7.3.24 shows that the liquid rate oscillates at the same dominant frequency
as the gas delivery system. This characteristic was not the case for the liquid rate in March
and may indicate that the flow in the tubing is more prone to flow instability if the gas delivery
system oscillates at lower frequencies.
47
Amplitude spectrum, Delivery pressure 6-9. May 16: 160-218 Amplitude spectrum, Injection pressure 6-9. May 16: 160-218
0.1 0.02
0.09 0.018
0.08 0.016
0.07 0.014
Amplitude, bar
Amplitude, bar
0.06 0.012
0.05 0.01
0.04 0.008
0.03 0.006
0.02 0.004
0.01 0.002
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Figure 7.3.25: Amplitude spectrum gas delivery and injection pressures 6-9. May 2016.
Amplitude spectrum, Injection rate 6-9. May 16: 160-218 Amplitude spectrum, Liquid rate 6-9. May 16: 160-218
150 4
3.5
100
Amplitude, Sm3/d
Amplitude, Sm3/d
2.5
1.5
50
0.5
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Figure 7.3.26: Amplitude spectrum gas injection and liquid production rates 6-9. May 2016.
Figure 7.3.27 shows liquid production at some hours later 9. May. The oscillations that
develop have a dominant amplitude of around 650 Sm3 /d and the oscillation period is esti-
mated to 7.8 minutes. Just before the oscillations develop closer analysis show the variations in
the gas delivery system gets smaller and less systematic, similar to what discovered in March.
Total liquid rate: 9. May 16: 240-260 hr Amplitude spectrum, Liquid rate 9. May 16: 240-260 hr
800
Liquid rate
4000 Mean
Std 700
600
3000
Amplitude, Sm3/d
Flow rate, Sm3/d
500
2000 400
300
1000
200
0 100
0
240 242 244 246 248 250 252 254 256 258 260 0 5 10 15 20 25 30
Time,hrs Period, minutes
Figure 7.3.27: Oscillating liquid production 9. May 2016
48
The production in March and May 2016 show similar characteristics ahead of oscillating
production. Minor variations in the lift-gas delivery pressure, gas pressure, and injection rate
may be a trigger for the oscillations observed in the well. Small variations in the gas delivery
system may cause disturbances in the annulus, which affect the injection of lift-gas to the
tubing. If these variations are amplified at specific frequencies, this could be an explanation to
the instability.
There are, however, some interesting differences in the two periods. Ahead of the insta-
bility in March, the gas delivery system oscillates with a period of 12.5 minutes. The liquid
oscillations that later develop in the tubing have a dominant amplitude of around 400 Sm3 /d.
In May, before the well becomes unstable the gas delivery system and the liquid rate show
oscillations with a period of 13.3 minutes. When the well later becomes unstable the liquid
oscillations have a dominant amplitude of around 650 Sm3 /d.
The discovery may suggest that the characteristics of the instability observed in the well
depend, to some degree, on the history of the gas delivery system. As the gas delivery system
oscillates at lower frequencies, the amplitudes that later develop in the tubing are larger.
7.3.6 Diminishing oscillations

The findings in the previous section show periodic variations in the gas delivery system and an-
nulus ahead of larger oscillations in the well. This may indicate the well is prone to instabilities
if the well is subjected to small variations/disturbances with period of 12-13 minutes.
In general, the measured flow data show at times short lasting and temporary oscillations.
Figure 7.3.28 on the next page shows liquid production 2-17. June 2016. An interesting obser-
vation can be made in the interval t = 900-1137 hours. The period show at times small and brief
oscillations that cease to exist after only a short period. The behavior seems to be repeating,
most evident at 990 and 1030 hours until the variations no longer vanish, but grow into larger
oscillations at t = 1137 hours. To gain control of the well the measurements indicate the well
was shut-in after 14 hours of oscillating production.
49
Total liquid rate: 2-17. June 16: 800-1160 hr
4500
Flow rate
Moving average
4000
3500
3000
Flow rate, Sm3/d

2500
2000
1500
1000
500
0
800 850 900 950 1000 1050 1100 1150 1200
Time,hrs
Figure 7.3.28: Liquid production rate 2-17. June 2016.
The figures that follows show the lift-gas delivery pressure at the wellhead, gas injection
pressure, liquid rate, and injection rate of lift-gas for 8-10. June, the interval t = 945-985 hours.
Compared to figure 7.3.17 and 7.3.16 the variations appear less constant and more scattered
around the moving average. Also, visual inspection of the measurements shows the variations
have smaller amplitudes than in March and May. Although the measurements do not indicate
large fluctuations, there seem to be disturbances in the gas system affecting the delivery and
injection of lift gas. Sporadically, the delivery pressure drops in line with the injection rate.
Delivery pressure: 8-10. June 16: 945-985 hr Injection pressure: 8-10. June 16: 945-985 hr
153 132
152
131.5
151
150
Pressure, bar
Pressure, bar
131
149
130.5
148
147
130
146
145 129.5
945 950 955 960 965 970 975 980 985 945 950 955 960 965 970 975 980 985
Time,hrs Time,hrs
Figure 7.3.29: Gas delivery and injection pressures 8-10. June 2016.
50
Injection rate: 8-10. June 16: 945-985 hr Total liquid rate: 8-10. June 16: 945-985 hr
1.04 1350
Injection rate Flow rate
Moving average 1300 Moving average
1.02
1250
1
1200
0.98
Flow rate, Sm3/d

1150
0.96
1100
0.94
1050
0.92
1000
0.9
950
0.88
900
0.86 850
945 950 955 960 965 970 975 980 985 945 950 955 960 965 970 975 980 985
Time,hrs Time,hrs
Figure 7.3.30: Gas injection and liquid production rates 8-10. June 2016.
Figure 7.3.31 and 7.3.32 show the associated amplitude spectrum’s, prior to the short
oscillation cycle at 990 hours. The figures show no particularly clear amplitudes, but it seems
to be a dominant amplitude with a period of 23.7 minutes. The amplitude spectrum for the
liquid production, however, appear chaotic and with no stable oscillation period. Thus the
liquid production still appears random and more or less stationary.
Amplitude spectrum, Delivery pressure 8-10. June 16: 945-985 hr Amplitude spectrum, Injection pressure 8-10. June 16: 945-985 hr
0.35 0.12
0.3 0.1
0.25
0.08
Amplitude, bar
Amplitude, bar
0.2
0.06
0.15
0.04
0.1
0.02
0.05
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Figure 7.3.31: Amplitude spectrum gas delivery and injection pressures 8-10. June 2016.
Amplitude spectrum, Injection rate 8-10. June 16: 945-985 hr Amplitude spectrum, Liquid rate 8-10. June 16: 945-985 hr
1000 4
900 3.5
800
3
700
Amplitude, Sm3/d
Amplitude, Sm3/d
2.5
600
500 2
400
1.5
300
1
200
100 0.5
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Figure 7.3.32: Amplitude spectrum liquid production and gas injection rates 8-10. June 2016.
51
Compared to the initiation of larger oscillations (March and May 2016), it turns out the
gas delivery system in the period 8-10. June oscillates with a period of 23.7 minutes. Approx-
imately one-half of the oscillation frequency compared to in March and May. The measure-
ments in figure 7.3.28 show the fluid production has a short-lasting oscillation period in the
subsequent period t = 990 to 995 hours. A similar response can be observed in the pressure.
However, since the response is neither persistent nor leading to greater oscillations, this might
indicate the well can oppose disturbances of this characteristics.
Similar behavior may be observed in the subsequent period up to June 16, t = 1137 hours.
Alternating intervals of similar response may be detected in the data until large oscillations
develop. During the production leading up to the oscillations, the variation in delivery pressure,
gas injection pressure, and gas injection rate increases and becomes more established. In a
similar fashion as for the intervals examined in March and May the same year.
52
Chapter 8
Dynamic response study
Periods of unstable production in well 6507/7-A-23 show oscillations with a period of about
8-11 minutes, observed in figure 7.3.8 and 7.3.10. The gas injection rate and reservoir pressure
are more or less constant in the periods considered. Still, production rates and pressures show
severe oscillations. The analysis in the previous chapter suggests disturbances in the gas de-

livery
system and kinematic waves as possible mechanisms of the instability in the well. The
illustrations below show the mechanism for the kinematic wave propagation in the well.
High-density outflow Low-density outflow

(Liquid-rich) (Gas-rich)
Lift-gas inflow Lift-gas inflow
Reservoir inflow Reservoir inflow

Pw Pw
(a) Liquid-rich outflow at time t. (b) Gas-rich outflow at time t + ∆t.
Figure 8.0.1: Mechanism of kinematic waves.
53
Increased lift-gas inflow to the tubing reduces the density of the fluid mixture. Thus the
bottom-hole pressure decreases. At a time later, the fluid mixture of reduced density reach the
outlet and will cause the well pressure to increase. A decrease in well pressure, however, will
cause a higher inflow of lift-gas. If the delivery pressure of lift-gas at the wellhead is constant,
this will cause the pressure in the annulus to decrease. After some time, this will cause the
inflow of lift-gas to the tubing to drop. In combination with varying inflow from the reservoir
and flow at the outlet, this may amplify the destabilization of the gas lift system (Asheim,
2016).
8.1 Model and prediction

The dynamic response on pressure disturbances may be expressed by the delay-differential
equation below.
       
ϑ  δpw a1 −b1 pw a2 −b2 p
=   −    w (8.1.1)
g
t t−∆t
Where pw and pg are pressure deviations in the tubing and annulus respectively. The kine-
matic delay between inflow and outflow in the well is expressed by ∆t. The model coefficients,
a1 , a2 , b1 , b2 , and c relate the dynamic response to fluid and reservoir properties and well design.
The parameters may be tuned to investigate the effect of alternative well design and operating
conditions. The model may be solved numerically, thus simulating the dynamic response of the
well to initial pressure disturbances.
8.2 Numerical simulation for gas lift response

A way of predicting the dynamic response in the well is by numerical simulation of equation
8.1.1 in time. The dynamic response of a well may also be predicted by discretization in time
and space such as Hu and Golan (2003) did in OLGA. This approach, however, includes a
much more detailed model and will make it harder to interpret the results.
8.2.1 Stable production

The model simulates pressure variations in the production tubing and annular volume at the
injection depth. The average of fluid densities and flow rates, from the injection point to outlet,
are used in the calculation of the coefficients.
54
Considering the stationary production interval 1-16. January 2017 (table 8.2.1), lineariza-
tion around the stationary solution gives the matrix coefficients; a1 = −0.000667, b1 =
0.000750, a2 = −0.000846, b2 = 0.000571, and c = 0.00130. The time delay is calculated
from the injection depth to outlet and is estimated to be ∆t = 6.3 minutes.
Table 8.2.1: Stationary production 1-16. January 2017
Parameter
Bottomhole pressure, [bar] 138.2
Productivity index, [Sm3 /d/bar] 25.0
Gas rate, [Sm3 /d] 1.38 105
Gas injection pressure wellhead, [bar] 118.9
Figure 8.2.1 on the next page illustrates the pressure response subjected to an initial dis-
turbance, δpw = 1 Pa and δpg = −1 Pa. As shown in Chapter 6, the flow rate depends on
pressure and will, therefore, show a similar response.
The y-axis shows the simulated pressure deviation from stationary conditions and the x-
axis illustrate the time course. The red and blue line represents the annular and tubing pressure
response at a depth of the injection valve, respectively. Figure 8.2.1a shows the dynamic re-
sponse with the time delay neglected (computed with a Runge-Kutta 4th order scheme). Figure
8.2.1b illustrates pressure variations predicted with the delayed response included (computed
with Matlab dde23 script). The simulation time is set to two hours to give a better illustration
of the initial pressure response. After stabilization, the response remains constant for longer
simulation times.
55
Pressure response without delay: 1-16. January 17 Pressure response with delay: 1-16. January 17
1.5 1.5
Well pressure Well pressure
Annulus pressure Annulus pressure
1 1
Pressure deviation, Pa
0.5 0.5
0 0
-0.5 -0.5
-1 -1
0 20 40 60 80 100 120 0 20 40 60 80 100 120
Time, min Time, min
(a) Response without delay. (b) Pressure response with delay.
Figure 8.2.1: Simulated pressure response for stationary production 1-16. January 2017.
With the delayed response neglected the imposed disturbances decreases with time, and
the well appears dynamically stable. The applied disturbances disappear after a short period,
and the deviations drop to zero. When incorporating the time delay in the model, the simulation
show similar response. However, the deviations stabilize at a value close to the tubing pres-
sure disturbance. The simulation depends on initial conditions, and therefore stabilization will
depend on the magnitude of the imposed disturbances. Nevertheless, damped responses like
in figure 8.2.1 indicates stability, and this corresponds with the dynamics observed in January
2017.
Assuming a gas-rich fluid mixture reaches the outlet and flows out the tubing. This causes
the average density in the production tubing to increase, thus increasing the pressure in the
well. If the injection of lift-gas through the orifice is sensitive to tubing pressure, an increase in
tubing pressure will cause the injection rate to decrease. This will, in turn, cause the pressure
in the annulus to increase. Figure 8.2.2a on the next page shows the response when the annulus
and tubing pressure are subjected to the disturbances, δpw = 10 Pa and δpg = 5 Pa. Only
the well pressure is showed to compare the simulated response to the measured well pressure.
Figure 8.2.2b shows measured well pressure 1-16 January 2017. The simulation time is set to
400 hours to compare the pressure response with measured data. The pressure response without
time delay predicts zero pressure deviations and has not been included.
56
Simulated pressure response: 1-16. January 17 Measured well pressure: 1-16. January 17
9.88 124.5
9.86
124
9.84
9.82 123.5
Pressure, bar
9.8
123
9.78
9.76 122.5
9.74
122
9.72
9.7 121.5
0 50 100 150 200 250 300 350 400 0 50 100 150 200 250 300 350 400
Time, min Time,hrs
(a) Simulated downhole pressure response. (b) Measured downhole well pressure.
Figure 8.2.2: Simulated pressure response and measured pressure 1-16. January 2017.
By including the variations at the outlet as in figure 8.2.2a, it seems the well model can
capture the dynamics observed in the well at certain initial conditions. Although the simulation
indicates persistent pressure oscillations, the variations are so small that production may be
regarded as stationary. Applying a positive pressure disturbance to both the tubing and annulus
seems to initiate small variations in the pressure. The variations appear more chaotic than
systematic, and frequency analysis showed no stable oscillation.
Multiple simulation runs suggest that increasing the magnitude of the disturbances or in-
creasing the simulation time does not cause the variations to increase.

In an attempt to model oscillating production as observed in the well, production data ahead of
the oscillations 28-29. January 2017 have been used as a reference. Figure 8.2.3 on the next
page shows measured well pressure 27-29. January the same year. At roughly 680 hours the
average well pressure increases from 120 to 123.5 bar. The pressure increase may be due to
a sudden density change in the production tubing. Varying inflow and outflow can cause the
conditions in the well to change and may help to explain the abrupt increase in tubing pressure.
Investigation of the measured data reveals a stationary production interval, t = 660-678
hours, ahead of oscillating production. Using the stationary interval as an operational reference,
it is possible to investigate if the model can replicate the unstable flow behavior.
57
Measured well pressure: 27-29. January 17
130
Well pressure
Moving average
128
126
Pressure, bar
124
122
120
118
116
660 665 670 675 680 685 690 695 700
Time,hrs
Figure 8.2.3: Measured well pressure 27-29. January 2017.
Considering the stationary gas lift scenario in table 8.2.2 provides the matrix coefficients;
a1 = -0.000610, b1 = 0.000708, a2 = -0.000736, b2 = 0.000583, and c = 0.00110. The time
delay is estimated to, ∆t = 5.4 minutes.
Table 8.2.2: Stationary production 27-28. January 2017
Parameter
Gas rate, [Sm3 /d] 1.59 105
Gas injection pressure, [bar] 129.0
The response predicted by solving the model equation yields similar results as for the
stationary production scenario 1-16. January, and appear dynamically stable. Multiple simula-
tions run with different disturbances applied on the system reveals that the model consistently
predicts the well stable for the given gas lift scenario. In fact, analysis of various periods lead-
ing to oscillatory production results in close to identical model coefficients as for the scenario
above, and thus the model anticipate damped pressure responses.
58
To model pressure oscillations similar to the measured well pressure, an alternative pro-
duction scenario has been constructed with some adjustments to the production scenario in table
8.2.1. The size of the injection orifice has been increased to 11.5 millimeters and the injection
rate decreased to 4.74 104 Sm3 /d. Also, the time delay was estimated from the perforations
to the outlet, not from the injection valve as earlier. Keeping all other parameters constant,
i.e. water cut, gas-oil ratio, and wellhead pressure, the steady-state gas lift model was used to
estimate the required production rates to flow the well. The steady-state solution provided the
coefficients: a1 = 0.0048, b1 = 0.0057, a2 = 0.0020, b2 = 0.0030, and c = 0.0068. The time delay
is estimated to, ∆t = 16.7 minutes. Data on the alternative production scenario are located in
Appendix B.
Figure 8.2.4 illustrates oscillations of the same order of magnitude (in bar) as measured
in the well. The simulated pressure response shows increasing oscillations, and the deviation
resembles the measured oscillations in figure 8.2.4. In the simulation, the initial pressure dis-
turbances were set to δpw = 9 pascals and δpg = −9 pascals. It should be mentioned that
the model is not able to limit oscillations as long as the second order terms are neglected in
the perturbation analysis. Thus it is not possible to simulate stable oscillation cycles using the
model.
Simulated pressure response: 27-29. January 17
10
Well pressure
8 Moving average
4
Pressure deviation, bar
-2
-4
-6
-8
-10
0 5 10 15 20 25 30 35 40
Time, hrs
Figure 8.2.4: Simulated pressure response 27-29. January 2017.
Analysis of the oscillating production 28-29. January 2017, section 7.3.3, showed an
oscillation period of around 10.6 minutes for pressure and flow rates. The oscillation period
predicted by the model equation, however, is around 30 minutes.
The simulations of stable and oscillating production show the dynamic model can recreate
the flow dynamics in the well to a certain extent. Initiated at the stationary production reference
1-16. January 2017, the model is able to capture the erratic behavior of the measured well
pressure when small pressure disturbances were imposed on the system. To model oscillations
that increase with time, however, an alternative production reference is required.
59
8.3 Parametric sensitivity study on stability
Over recent decades, gas lift stability has been the subject of many studies. Along with simula-
tion studies, stability criteria and theories have provided valuable insight regarding stability of
gas lift wells. Parameters recognized to affect stability are gas injection rate, injection orifice
size, productivity index, injection depth, tubing size. Among others, Asheim (1988) and Alha-
nati et al. (1993) showed how the listed parameters influence stability and verified their analytic
criteria against reported field data. However, some wells predicted to be stable by existing cri-
teria often turn out unstable in practice. The purpose of the current model is to correct this
by including the outflow response. By performing a simulation study, the subsequent sections
examine the effect of the above parameters on stability.
To observe a significant change in the model response, simulation results in Chapter 8.2
demonstrated that production parameters had to be changed rather much. The reason for the
lack of similarity between simulation and measurements is unknown and difficult to tell. It may
be due to errors in measurement data and well design parameters, or due to the model’s ability
to capture the dynamic pressure response.
However, to investigate the effect of the mentioned parameters adjustments have been
made to the stationary production reference 1-16. January 2017. Table 8.3.1 summarizes the
stationary gas lift scenario chosen as the base case. The orifice diameter has been increased to
from 9.1 to 11.9 mm and the time delay is estimated from the depth of the perforations to the
outlet. All other parameters are the same.
Table 8.3.1: Gas lift base case - sensitivity study.
Parameter
Bottom hole pressure, [bar] 138.2
Gas rate, [Sm3 /d] 1.38 105
Gas injection pressure at wellhead, [bar] 118.9
60
8.3.1 Effect of gas injection rate
Increasing the injection rate may improve stability if the gas lift well is producing near its
most efficient point, and attributes mainly two factors. Firstly, increased injection rate causes
increased friction loss in the tubing, and this has a dampening effect on instability. The other is
that increased flow rate in annulus reduce the delay effect, between inflow and outflow, in the
annular volume (Hu and Golan, 2003).
In the current model, only the effect of frictional pressure loss is relevant. The figures
that follows gives the simulation results for three different gas injection rates, in which the first
corresponds to the base case in table 8.3.1. For the different injection rates the steady-state gas
lift model has been used to find the well pressure and the corresponding production rates of
oil, water, and gas, assuming the water cut and gas-oil ratio stays the same. The steady-state
production data under different gas injection rates are located in Appendix B.
Pressure response without delay - Injection rate 8.85*10 4 Sm3/d Pressure response with delay - Injection rate: 8.85*10 4 Sm3/d
4 4
3 3
2 2
1 1
0 0
-1 -1
-2 -2
-3 -3
-4 -4
0 20 40 60 80 100 120 0 20 40 60 80 100 120
Time, min Time, min
Figure 8.3.1: Dynamic response - Injection rate: 8.85 104 Sm3 /d.
Pressure response without delay - Injection rate 6.85*10 4 Sm3/d Pressure response with delay - Injection rate: 6.85*10 4 Sm3/d
4 4
3 3
2 2
1 1
0 0
-1 -1
-2 -2
-3 -3
-4 -4
0 20 40 60 80 100 120 0 20 40 60 80 100 120
Time, min Time, min
61
4 4
Pressure response without delay - Injection rate 4.85*10 Sm3/d Pressure response with delay - Injection rate: 4.85*10 Sm3/d
4 10
3 8
6
2
4
1
2
0 0
-2
-1
-4
-2
-6
-3
-8
-4 -10
0 20 40 60 80 100 120 0 20 40 60 80 100 120
Time, min Time, min
The first case, figure 8.3.1, has the highest injection rate. It shows the disturbances de-
creases with time both with and without the time delay included. The well is stabilized at this
injection rate. Figure 8.3.2 shows the case when the injection rate is 6.85 104 Sm3 /d, reduced
by around 25 %. Without delay, the initial amplitude increases slightly before the response is
stabilized. With the time delay included, damped oscillations emerge, and the effect of decreas-
ing the injection rate is clear. The well, however, still appear stabilized.
Figure 8.3.3 shows the case with the lowest injection rate, 4.85 104 Sm3 /d. With the
delay term included, the applied disturbances lead to oscillations that increase with time, and
the system appears dynamically unstable. The oscillation period is estimated to 37 minutes.
Without the time delay included the response is approximately as for the cases with higher gas
injection rates.
8.3.2 Effect of gas lift orifice size

The orifice size of the downhole injection valve has a strong impact on stability in gas lift wells.
Injection valves of smaller orifice diameter promote well stability since it can suppress large
flow variations. The effect of orifice size on stability is well documented, and proper valve
design is crucial for the operation.
Figure 8.3.1 on the previous page shows the simulation of the gas lift base case with an
orifice size of 11.9 millimeters. Figure 8.3.4 and 8.3.5 on the next page shows the pressure
response for two cases with different orifice sizes, 9.5 and 14.3 millimeters respectively.
62
Pressure response without delay - Orifice size: 9.5 mm Pressure response with delay - Orifice size: 9.5 mm
4 4
3 3
2 2
1 1
0 0
-1 -1
-2 -2
-3 -3
-4 -4
0 20 40 60 80 100 120 0 20 40 60 80 100 120
Time, min Time, min
Figure 8.3.4: Dynamic response - Injection orifice size: 9.5 mm.
Pressure response without delay - Orifice size: 14.3 mm Pressure response with delay - Orifice size: 14.3 mm
6 10
8
4
6
4
2
2
0 0
-2
-2
-4
-6
-4
-8
-6 -10
0 20 40 60 80 100 120 0 20 40 60 80 100 120
Time, min Time, min
Figure 8.3.5: Dynamic response - Injection orifice size: 14.3 mm.
It is clear that the oscillations are decreased when a smaller orifice size is used in the well.
Using an orifice size of 9.5 millimeters, the response with and without time delay is more or
less identical. For the case with an orifice size of 14.3 millimeters, there is a big difference
in the response. By including the variations at the outflow, an increase in orifice size causes
increasing oscillations to develop, and the model predicts the well dynamically unstable.
8.3.3 Effect of productivity index

The productivity index also affects the stability of gas lift wells, and this is quantified by cri-
terion F1 , derived by Asheim (1988). Injection of lift gas to tubing decrease the hydrostatic
pressure drop and promote flow instability. A high productivity index introduces more reser-
voir fluid to the well and increase the average density of the flowing fluid mixture. This will
increase the pressure in the tubing and compensate for the increased gas fraction and should
stabilize the flow (Asheim, 1988).
63
The figures below shows the pressure response for a productivity index of 45 Sm3 /d/bar
and 5 Sm3 /d/bar. The steady-state gas lift model has been used to estimate downhole pres-
sures and flow rates for the different cases. The injection rate and wellhead pressure are the
same as for the base case scenario. Production data on the different cases are located in Ap-
pendix B.
Figure 8.3.6 shows the case for a productivity index of 45 Sm3 /d/bar. The response, with
and without delay, seems very similar to the simulation of the base case in figure 8.3.1. Figure
8.3.7 shows the simulation results for a productivity index of 5 Sm3 /d/bar.
Pressure response without delay - PI: 45 Sm3/d/bar Pressure response with delay - PI: 45 Sm3/d/bar
4 4
3 3
2 2
1 1
0 0
-1 -1
-2 -2
-3 -3
-4 -4
0 20 40 60 80 100 120 0 20 40 60 80 100 120
Time, min Time, min
Figure 8.3.6: Dynamic response - Productivity index: 45 Sm3 /d/bar.
Pressure response without delay - PI: 5 Sm3/d/bar Pressure response with delay - PI: 5 Sm3/d/bar
4 4
3 3
2 2
1 1
0 0
-1 -1
-2 -2
-3 -3
-4 -4
0 20 40 60 80 100 120 0 20 40 60 80 100 120
Time, min Time, min
Figure 8.3.7: Dynamic response - Productivity index: 5 Sm3 /d/bar.
64
A smaller productivity index should, according to literature, destabilize the system, but
this does not appear to be the case. By including the outflow variations, the response looks
more or less identical to the case for a productivity index of 25 and 45 Sm3 /d/bar. Without
the delay term incorporated in the model, the response also looks similar to the two other cases.
The initial amplitudes decrease slightly less with time, but the system still appears dynamically
stable.
Multiple simulations with various values of productivity index and injection rates confirm
the response observed above. Thus it seems the suggested model is not particularly sensitive to
the reservoir inflow.
8.3.4 Effect of injection depth

The effect of the injection depth on stability is illustrated in the figures below. Increasing the
depth of the injection point from 1826 (base case) to 2164 and 2850 meters MD destabilizes
the system, and may cause the well to be unstable under certain conditions. This should be
accounted for when designing a gas lift well. In general, increasing the depth of the injection
point boosts the efficiency of the gas lift operation. However, the engineer should always verify
if the well is capable of producing without experiencing oscillations with the selected depth of
the injection valve.
Figure 8.3.8 and 8.3.9 shows the current model is capable of capturing the effect of the
injection depth. For an injection depth of 2850 meters (figure 8.3.9) damped oscillations de-
velop and the well still appear stable. However, the analysis shows that a slight decrease in the
injection rate causes the well to experience oscillations that increase with time at this depth.
Without time delay, the effect of injection depth is hardly noticeable.
Pressure response without delay - Injection depth: 2164 m MD Pressure response with delay - Injection depth: 2164 m MD
4 4
3 3
2 2
1 1
0 0
-1 -1
-2 -2
-3 -3
-4 -4
0 20 40 60 80 100 120 0 20 40 60 80 100 120
Time, min Time, min
Figure 8.3.8: Dynamic response - Injection depth: 2164 m MD.
65
Pressure response without delay - Injection depth: 2850 m MD Pressure response with delay - Injection depth: 2850 m MD
4 4
3 3
2 2
1 1
0 0
-1 -1
-2 -2
-3 -3
-4 -4
0 20 40 60 80 100 120 0 20 40 60 80 100 120
Time, min Time, min
Figure 8.3.9: Dynamic response - Injection depth: 2850 m MD.
8.3.5 Effect of tubing size

Already in the 1950s, Bertuzzi et al. (1953) demonstrated that a decrease in the annular volume
might help stabilize gas lift wells. Since then, several studies have investigated and verified the
effect of tubing/casing size on stability. To quantify the influence of tubing size on stability the
oscillating production scenario in figure 8.3.3 has been used as the base case. Figure 8.3.10 and
8.3.11 demonstrate the effect of changing the tubing size from 6 5/8" (base case) to 5.5" and
7", respectively.
For a tubing size of 5.5" outer diameter, the annular volume is increased, and simulation
shows the well is further destabilized. By including the outflow variations, the oscillations in-
crease more rapidly with time and the amplitude of the oscillations increase. When the outflow
is not incorporated in the model simulation shows pressure oscillations that decrease with time.
Figure 8.3.11 shows that using a 7" outer diameter tubing has a strong stabilizing effect on the
well. For a 7" tubing the annular volume is reduced and the model predicts minor oscillations
that decrease with time when the outflow is included.
Pressure response without delay - Tubing size: 5.5" Pressure response with delay - Tubing size: 5.5"
4 10
3 8
6
2
4
1
2
0 0
-2
-1
-4
-2
-6
-3 -8
-4 -10
0 20 40 60 80 100 120 0 20 40 60 80 100 120
Time, min Time, min
Figure 8.3.10: Dynamic response - Tubing size: 5.5".
66
Pressure response without delay - Tubing size: 7" Pressure response with delay - Tubing size: 7"
4 10
3 8
6
2
4
1
2
0 0
-2
-1
-4
-2
-6
-3 -8
-4 -10
0 20 40 60 80 100 120 0 20 40 60 80 100 120
Time, min Time, min
Figure 8.3.11: Dynamic response - Tubing size: 7".
67
8.4 Analytical solution for gas lift model response
The structure of equation 8.1.1 as well as observed and simulated gas lift instability indicate
cyclic and/or exponential responses. In addition to simulation, solutions for the tubing and
annular pressures may be expressed as complex exponential functions, but with different am-
plitude and some phase delay. The solutions may be expressed as δpw = Pˆw e(α+iω)t and
δpg = Pˆg eαt ei(ωt+ϕ) . Where Pˆw and Pˆg are the initial amplitudes, α is the attenuation factor,
ω = 2πf is the oscillation frequency, and ϕ is the phase delay. Using trigonometric identities
the expressions may be rewritten as
δpw = Pˆw e(α+iω)t = Pˆw eαt (cos(ωt) + isin(ωt)) . (8.4.1)
δpg = Pˆg eαt ei(ωt+ϕ) = Pˆg eαt (cos(ωt + ϕ) + isin(ωt + ϕ)) . (8.4.2)
Substituting equation 8.4.1 and 8.4.2 into equation 8.1.1 we obtain after separating real
and imaginary components,
Pˆw αcos(ωt) − ωsin(ωt) − a1 cos(ωt) + a2 e−∆tα cos(ω(t − ∆t)

(8.4.3)
+ Pˆg b1 cos((ωt) + ϕ)b2 e−∆tα cos(ω(t − ∆t) + ϕ) = 0.

Pˆw αsin(ωt) + ωcos(ωt) − a1 sin(ωt) + a2 e−∆tα sin(ω(t − ∆t)

(8.4.4)
+ Pˆg b1 sin((ωt) + ϕ)b2 e−∆tα sin(ω(t − ∆t) + ϕ) = 0.

Pˆg [αcos(ωt + ϕ) − ωsin(ωt + ϕ) + cos(ωt + ϕ)c] − cPˆw cos(ωt) = 0. (8.4.5)
Pˆg [αsin(ωt + ϕ) − ωcos(ωt + ϕ) + sin(ωt + ϕ)c] − cPˆw sin(ωt) = 0. (8.4.6)
From the above expressions follows that the pressure response may exhibit oscillations at har-
monic frequencies. If the attenuation factor is positive, the variations will increase with time,
and the system may be considered unstable. If the attenuation factor is negative, however, the
oscillations decrease with time, such that the system may be regarded as stable. With lineariza-
tion around a steady-state solution of the gas lift system equation 8.4.3 through 8.4.6 results in
four equations and four unknowns, α, ω, ϕ, and t.
68
The system of equations may be solved iteratively by considering various initial guesses
of the unknowns and different ratios of the initial amplitudes. Investigating the deviation from
steady-state production the well and annular pressure may be expressed as
pw (t) = p¯w + Pˆw e(α+iω)t . (8.4.7)
pg (t) = p¯g + Pˆg eαt ei(ωt+ϕ) . (8.4.8)
Where p¯w and p¯g are the steady-state tubing and annular pressures respectively.
8.4.1 Stable production

Initiated at stationary production January 2017, the simulations predicted the pressure response
dynamically stable, and the applied disturbances quickly vanished. Figure 8.4.1 on the next
page shows the response predicted by the analytical solution of the model equation for initial
amplitudes of Pˆw = 1 Pa and Pˆg = 1 Pa. The pressure response is estimated by plotting the real
part of equation 8.4.1 and 8.4.1. To solve the system of equations the built in Matlab function
fsolve was utilized.
Table 8.4.1 summarizes the dynamic parameters from the analytical prediction. The table
also shows the parameters change when the initial amplitude of the tubing pressure is increased
to Pˆw = 2 Pa. Increasing the amplitude ratio, Pˆw /Pˆg , from 1 to 2 seems to drive both the
attenuation factor, angular frequency, and phase lag towards zero.
Table 8.4.1: Dynamic parameters for analytical solution - stable production.
Pˆw /Pˆg [-] Atten.fac. [1/s] Angular freq. [1/s] Phase lag [rad/s]
1/1 -0.0004 0.0009 -0.8942
2/1 -0.0001 0.0005 -0.7109
69
Analytical pressure response: Pw/Pg = 1/1
4
Well pressure
Annulus pressure
3
1
-1
-2
-3
-4
0 20 40 60 80 100 120
Time, min
Figure 8.4.1: Analytical pressure response for stable production: Pˆw = 1 Pa, Pˆg = 1 Pa.
Figure 8.4.1 shows similar response to what was predicted by the simulations in Chapter
8.2.1. Solving the analytical solution gives negative attenuation factors, and thus the system
may be considered dynamically stable. The initial amplitudes level off quickly and become
apparently constant after 60 minutes. So, the analytical solution confirms that the system is
damped as predicted by simulation of stationary production 1-16. January 2017.
Figure 8.4.2 shows the solution when the amplitude ratio is increased to 2/1. The solution
confirms that the system is stable. However, the time for the amplitudes to level off is consid-
erably longer, and the variations become constant after approximately 500 minutes. Increasing
the amplitude ratio further causes the pressures to stabilize immediately at their initial ampli-
tude value as if the system stabilizes at another stable steady-state. Decreasing the amplitude
ratio to Pˆw /Pˆg = 1/2, however, causes the amplitudes to decline more rapidly, and the system
appears strongly stabilized.

4
Well pressure
Annulus pressure
3
2
-1
-2
-3
-4
0 100 200 300 400 500 600
Time, min
Figure 8.4.2: Analytical pressure response for stable production: Pˆw = 2 Pa, Pˆg = 1 Pa.
70
To model oscillating production and to obtain a visually interesting response in the well, the
parametric simulation study was initiated using a modified gas lift scenario. Based on this
scenario, and by setting the injection rate to 4.85 104 Sm3 /d, the model equation predicted the
system to be dynamically unstable.
Figure 8.4.3 shows the response predicted by the analytical solution initiated under the
conditions just described. Table 8.4.2 shows the dynamic parameters associated with the solu-
tion. Increasing the amplitude ratio seems to increase the attenuation factor while the angular
frequency and phase lag decrease.
Table 8.4.2: Dynamic parameters for analytical solution - oscillating production.
Pˆw /Pˆg [-] Atten.fac. [1/s] Angular freq. [1/s] Phase lag [rad/s]
1/1 0.0001 0.0072 -0.8551
2/1 0.0023 0.0015 -0.2588

4
Well pressure
Annulus pressure
3
2
Pressure deviation, pa
-1
-2
-3
-4
0 20 40 60 80 100 120
Time, min
Figure 8.4.3: Analytical pressure response on oscillating production: Pˆw = 1 Pa, Pˆg = 1 Pa.
The result illustrated in figure 8.4.3 is similar to the simulated response in figure 8.3.3.
The solution results in positive attenuation factors such that the system may be considered dy-
namically unstable. We observe that the oscillations increase slowly and thus the analytical
solution confirms that the system i undamped. The oscillation period is estimated to 14.4 min-
utes. Thus, the oscillation period predicted by the analytical solutions is closer to the measured
period (8-11 minutes) compared to the oscillation period predicted by simulation.
71
Figure 8.4.4 shows the response for an amplitude ratio of 2/1. The solution confirms
that the system is unstable, but the magnitude of the long-term amplitudes seems to be of no
physical meaning. However, it may indicate that the system is further destabilized for higher
amplitude ratios.

500
Well pressure
400 Annulus pressure
300
200
100
-100
-200
-300
-400
-500
0 20 40 60 80 100 120
Time, min
Figure 8.4.4: Analytical pressure response on oscillating production: Pˆw = 1 Pa, Pˆg = 2 Pa.
72
Chapter 9
Discussion
9.1 Observed characteristics

Analysis of production in the well show pressures and flow rates oscillate severely in certain
periods. The months the data were acquired, the measurements reveal the well could maintain
stable production only for short periods of time before the production becomes unstable. The
oscillations that develop seems to be stable and establishes quickly in the system. They also
appear to fluctuate with approximately constant amplitudes and frequencies.
To control production, the field data show Statoil shut-in the well several times. The
effect of shutting down production, however, only seems to have a temporary effect on the flow
instability.
In the event of stationary production, the analysis shows flow rates are close to perfect
normal distributed and that the measurements vary stochastic around the mean. Also, no sys-
tematic oscillation frequency is present in the data in the cases of steady production.
9.2 Initiation of instability

The analysis revealed systematic disturbances in the gas delivery system ahead of flow insta-
bility. The gas delivery pressure, gas injection pressure, and the gas injection rate all shows
variations of the same oscillation periods. The flow instability that later develops in the tubing
seems to depend on the characteristics of the disturbances in the gas delivery system. In the
event lower frequency variations in the gas delivery system the liquid oscillations that develop
have larger amplitudes compared to the case of variations of higher frequency in the gas deliv-
ery system. The oscillation periods of the variations in the gas delivery system were estimated
to be 12.5 and 13.3 minutes. For an oscillation period of 12.5 minutes, the liquid oscillations
that later develop have an amplitude of about 400 Sm3 /d. In the event of an oscillation period
of 13.3 minutes, the liquid oscillations have an amplitude of 650 Sm3 /d.
73
The investigation also identified short lasting and diminishing oscillations in the measure-
ments when the gas delivery system is subjected to small variations with a period of around
24 minutes. This may indicate that the flow in the production tubing is more sensitive to dis-
turbances of higher frequencies, which is more similar to the oscillations that develop in the
tubing.
The characteristics of the disturbances imply the gas delivery system oscillate at a lower
frequency compared to the oscillations that later develop in the tubing. Although the variations
in the gas delivery system are small and have different characteristics, it may have a significant
impact on the communication between annulus and tubing. Considering tubing and annulus
are connected via the injection valve downhole the conditions in the annulus, and thus the
gas delivery system will affect the flow in the tubing. Also, the oscillation period was found
similar to the flow time along the production tubing and kinematic waves have been suggested
to maybe be one of the reasons for the instability in the well.
9.3 Estimation of static and dynamic parameters from mea-

sured data
Both stationary and dynamic parameters have been estimated from the measurements. These
estimates depend on the steady-state model’s ability to simulate pressure in the well. Estima-
tion of the stationary productivity index involves estimates of the reservoir pressure and flowing
bottomhole pressure. The steady-state model use measured production rates and wellhead pres-
sure to estimate the pressure at the bottom of the well. The reservoir pressure was estimated by
considering a period of static pressure build-up.
The dynamic productivity index was estimated in certain periods of oscillating produc-
tion. Using the liquid and the downhole pressure amplitudes gave a dynamic productivity
index of around 260 Sm3 /d/bar, versus the stationary productivity index of 25 Sm3 /d/bar.
The magnitude of the dynamic productivity index illustrates the severeness of the oscillations
and indicate that the liquid production responds very strongly to the oscillating pressure varia-
tions in the well. The liquid amplitudes were estimated to be around 400-650 Sm3 /d while the
wellhead and downhole pressure amplitudes were measured to around 7 bars and 2 bars respec-
tively. In periods of unstable production, frequency analysis shows pressures and production
rates oscillates at equal frequencies. This indicates the flow instability affect the entire well.
The oscillation periods were estimated to be 7.8-10.6 minutes.
The provided field data were acquired using temperature/pressure transducers and mul-
tiphase meter, and the accuracy of the measurements will depend on how well calibrated the
equipment was. This may affect the estimation of the parameters. The simplifications and
74
assumptions regarding the steady-state model will also influence the pressure estimates. A
more complex model, i.e. including holdup and other flow relations, may give more accurate
estimates.
9.4 The dynamic model and dynamic characteristics predic-

tion
Initiated at stationary production, the dynamic model predicts the well dynamically stable for
all cases examined. Damped pressure responses like in figure 8.2.1 indicates stability. The
simulated response in figure 8.2.2 show that the model can capture the erratic flow behavior
observed in the field data. The simulated pressure shows small variations around an average
value as seen in the measurements. The response, however, indicates dynamic stability.
Simulation of the model equation depends on initial conditions such that stability predic-
tion by simulation may be misleading, and may give a wrong idea of the dynamics in the well.
Therefore, different stationary references have been subjected to various initial disturbances to
check the simulation results. Common to all simulations, initiated at field measurements, is that
the model predicts the well stable. In certain periods, this is not in agreement with the observed
oscillations in the data.
The uncertainties discussed in Chapter 9.3 may also affect the estimation of the response
parameters in the dynamic model and thus also the prediction. However, for the model to simu-
late increasing oscillations the parameters had to be changed rather much. Thus, the accuracy of
the measurements is probably not crucial to the disagreement between measured and simulated
response.
To model oscillations that increase with time modifications had to be made. The modifica-
tions may suggest that the model is particularly sensitive to injection orifice size, gas injection
rate, and time delay. When the time delay is estimated from the injection depth to the outlet,
the model predicts oscillations that decrease with time. Increased gas injection rate will both
reduce the average mixture density and the average flow velocity of the flowing fluid. Reduced
flow velocity will increase the time delay in the model. The increase in density will have a
big impact on the pressure conditions in the well and thus also the response parameters in the
dynamic model. Increasing the diameter of the injection orifice has a significant influence on
the inflow of lift-gas to the tubing. This was demonstrated in the sensitivity study.
The sensitivity study illustrates that the model can capture the effect of main design and
reservoir parameters on stability. Changing the injection orifice size, gas injection rate, injec-
tion depth, and tubing diameter give model responses which correspond well with the literature
discussed. The effect of productivity index on the model response, however, does not seem to
75
coincide with earlier studies. The reason for the disagreement is hard to say and can only be
speculated about.
9.5 System versus model

The dynamic gas lift model seems to capture the dynamics observed in the well in certain
cases. Initiated at stationary production the model predicts the well stable for all cases. To
model unstable production similar to the measurements adjustments had to be made regarding
some parameters.
The suggested model seems to encounter a similar problem as existing criteria. In certain
production periods, when the model predicts stability the field data show the well was unstable
in practice. Although the model considers the variations at the outlet of the tubing, it fails to
predict the response in case of oscillating production. However, by including the outflow vari-
ations, the model seems to be more sensitive to changes in the reservoir and design parameters.
This may be considered as an advance, and a step towards better prediction of stability in gas
lifted wells.
It ’s hard to give a proper assessment to the difference between the model results and
field data. The analysis of the field data revealed an inconsistency regarding provided gas
injection pressure and estimated annulus pressure downhole. Thus, there might other errors in
the field data that affect the results. It may also be that the instability in the well is due to other
factors than what the model takes into account. Another reason to the disagreement may be the
simplifications and assumptions in the models.
9.5.1 Further work

Suggestions for further work that may improve the results could be to include slippage, hold-
up, and other flow relations in the dynamical model as well as the steady-state model. This may
give a more realistic simulation of the conditions in the well.
To verify the model results, it would also be interesting and useful to test the model against
another gas lift well that exhibit flow instabilities. In this way, the model’s prediction capabili-
ties are checked against independent measurements under other circumstances. Comparing the
model results from different cases may be useful considering improvements and updating of
the model.
76
Chapter 10
Conclusion
From the discussion above, the following may be concluded:
• Unstable production in the well exhibits large liquid oscillations with amplitudes of
around 400 and 650 Sm3 /d. Pressures and production rates oscillate at equal frequen-
cies and have an oscillation period of about 10 minutes, similar to the flow time along the
tubing. Kinematic waves seem likely to be a mechanism of the instability in the well.
• The variations in the gas delivery system also seem likely to be a cause of the instability.
Ahead of unstable production, the gas delivery system shows systematic variations with
an oscillation period of around 12-13 minutes. The oscillation period seems to influence
the size of liquid amplitudes that later develop in the tubing.
• The suggested model seems to improve prediction of flow conditions in certain cases.
Initiated at stationary production the dynamic gas lift model predicts the well dynami-
cally stable for all cases examined. The model seems to be able to imitate minor and
random variations as observed in the measurements.
• To model oscillations that increase with time modification of the parameters was re-
quired. The oscillations simulated by the model equation seems to resemble the charac-
teristics of the measurements to some extent.
• The analytical solution gave results similar to simulation, but the oscillation periods pre-
dicted by the analytical solution are closer to the measurements.
• The sensitivity study seems to demonstrate that the model is capable of capturing the ef-
fect of main design and reservoir parameters on stability. Excluding the outflow response
77
appears to reduce the prediction capability of the model.
• The simplifications and assumptions in the models will affect the estimation of the re-
sponse parameters in the dynamic model and may be a cause of the disagreement between
measured and simulated response. It is likely that the accuracy of the field measurements
is not a crucial factor to the model results.
78
Nomenclature
α attenuation factor, 1/s
∆t time delay, s
λl , λg flux fractions, −
µg gas viscosity, P a s
µl liquid viscosity, P a s
ω angular frequency, rad/s
ρf i reservoir fluid density at injection point, m3 /kg
ρgi lift-gas density at injection point, m3 /kg
ρgsc lift-gas density at standard conditions, m3 /kg
ρg lift-gas density, m3 /kg
ρl liquid density, kg/m3
ρm fluid mixture density, kg/m3
ρT P two-phase fluid mixture density, m3 /kg
a1 , a2 , b1 , b2 , grouping of response parameters
Ac injection port size, m2
Ag area occupied by gas phase, m2
ag , aw , c, f t response parameters
Al area occupied by liquid phase, m2
Bg gas formation volume factor, m3 /Sm3
Bo oil formation volume factor, m3 /Sm3
Co distribution parameter for bubbles in flow, −
cT P slip correction factor, −
D vertical depth to injection valve, m
79
d tubing diameter, m
E injection orifice efficiency factor, -
F1 , F2 stability criteria
fm fluid mixture friction factor, −
fT P two-phase friction factor, −
g acceleration due to gravity, m/s2
J productivity index, Sm3 /d/bar
Jg gas inflow index at downhole conditions, m3 /d/bar
Jw productivity index at downhole conditions, m3 /d/bar
L measured depth, m
Mg gas molecular weight, kg/kmol
pg annulus pressure downhole, bar
pr reservoir pressure, bar
pth tubing head pressure, bar
pti tubing pressure at injection point, bar
pwh wellhead pressure, bar
qf i flow rate of reservoir fluids at injection point, m3 /d
qgi flow rate of lift-gas at injection point, m3 /d
Qg gas flow rate at downhole conditions, m3 /d
qLsc liquid flow rate at standard conditions, Sm3 /d
QL liquid flow rate at downhole conditions, m3 /d
R universal gas constant, J/K/mol
Rs solution gas-oil ratio, Sm3 /Sm3
Rt produced gas-oil ratio, Sm3 /Sm3
T temperature, K
t current time, s
Vc annulus volume, m3
vm fluid mixture velocity, m/d
vo buoyancy velocity of the gas bubbles, or sink velocity for droplets, m/s
vsg gas superficial velocity, m/d
80
vsl liquid superficial velocity, m/d
Vt tubing volume, m3
x measured depth to injection point, m
yl , yg volume fractions, −
81
82
References
Alhanati, F., Schmidt, Z., Doty, D., and Lagerlef, D. (1993). Continuous Gas-Lift Instability:
Diagnosis, Criteria, and Solutions. Society of Petroleum Engineers. doi: /10.2118/26554-MS.
Asheim, H. (1986). MONA, An Accurate Two-Phase Well Flow Model Based on Phase Slip-
page. Society of Petroleum Engineers.doi: 10.2118/12989-PA.
Asheim, H. (1988). Criteria for Gas-Lift Stability. Society of Petroleum Engineers. doi:
10.2118/16468-PA, vol 40(11).
Asheim, H. (2000). Analytical Solution of Dynamic Inflow Performance. Society of Petroleum

Engineers. doi: 10.2118/63307-MS.
Asheim, H. (2016). "11. Gas Lift", Lecture notes, Unpublished, TPG4245 NTNU, URL:
http://www.ipt.ntnu.no/ asheim/prodbr.html.
Asheim, H. (2017). Personal communication Spring 2017.
Avest, D. and Oudeman, P. (1995). A Dynamic Simulator to Analyse and Remedy Gas Lift
Problems. Society of Petroleum Engineers. doi: /10.2118/30639-MS.
Bellarby, J. (2009). Well Completion Design 1st ed. Elsevier Science, ISBN-9780444532107.
Bertuzzi, A., Welchon, J., and Poettman, F. (1953). Description and Analysis of an
Efficient Continuous-Flow Gas-Lift Installation. Society of Petroleum Engineers. doi:
10.2118/953271-G, vol 5(11).
Blick, E., Enga, P., and Lin, P. (1988). Theoretical Stability Analysis of Flowing Oil Wells and
Gas-Lift Wells. Society of Petroleum Engineers. doi: 10.2118/15022-PA, vol 3(4).
Dalsmo, M. and Halvorsen, E. (2002). Active Feedback Control of Unstable Wells at the Brage
Field. Society of Petroleum Engineers. doi: 10.2118/77650-MS.
Eikrem, G., Aamo, O., and Foss, B. (2008). On Instability in Gas Lift Wells and Schemes for
Stabilization By Automatic Control. Society of Petroleum Engineers.doi: 10.2118/101502-
PA, vol 23(2).
83
Fairuzov, Y., Guerrero-Sarabia, C., Carmona-Diaz, R., Cervantes-Baza, T., Miguel-Hernandez,
X., and Rojas-Figueroa, A. (2004). Stability Maps for Continuous Gas-Lift Wells: A New
Approach to Solving an Old Problem. Society of Petroleum Engineers. doi: 10.2118/90644-
MS.
Fitremann, J. and Vedrines, P. (1985). Non Steady Gas-Liquid Flow in Pipes and Gas-Lifted
Wells. Conference on Multi-Phase Flow.
Gang, Z. and Golan, M. (1989). Criteria for Operation Stability of Gas Lift Wells. Society of
Petroleum Engineers.
Gilbert, W. (1954). Flowing and Gas-Lift Well Performance". Drill. and Prod. Prac. American
Petroleum Institute.
Grupping, A., Luca, C., and Vermuelen, F. (1984). Heading Action Analyzed for Stabilization.
American Petroleum Institute, Oil and Gas Journal.
Guerrero-Sarabia, I. and Fairuzov, Y. (2013). Linear and non-linear analysis of flow

instability in gas-lift wells. Journal of Petroleum Science and Engineering. doi:
10.1016/j.petrol.2013.01.012, vol 108.
Hirch, M. and Smale, S. (1974). Differential Equations, Dynamical Systems, and Linear Alge-
bra. Elsevier, ISBN-9780080873763.
Hjalmars, S. (1973). The Origin of Instability in Airlift Pumps. ASME. J. Appl. Mech.
doi:10.1115/1.3422994, vol 40(2):399–404.
Hu, B. (2004). Characterizing gas-lift instabilities. Doctoral Thesis NTNU.
Hu, B. and Golan, M. (2003). Gas-lift Instability Resulted Production Loss and Its Remedy
by Feedback Control: Dynamic Simulation Results. Society of Petroleum Engineers. doi:
10.2118/84917-MS.
Liu, H. and Vandu, C. (2005). Hydrodynamics of Taylor Flow in Vertical Capillaries: Flow
Regimes, Bubble Rise Velocity, Liquid Slug Length, and Pressure Drop. Ind. Eng. Chem.
Res. doi: 10.1021/ie049307n, vol 44(14).
Myhr, E. (2016). "Gas Lift and Instability", Project report, Unpublished, TPG4560 NTNU.
Norwegian Petroleum Directorate (2017). Heidrun: Gen-

eral information, Retrieved March 4, 2017, from
http://factpages.npd.no/ReportServer?/FactPages/PageView/fieldrs:Command=Renderrc:
Toolbar=falserc:Parameters=fNpdId=43771IpAddress=129.241.228.11CultureCode=en.
84
Poblano, E., Camacho, R., and Fairuzov, Y. (2002). Stability Analysis of Continuous-Flow Gas
Lift Wells. Society of Petroleum Engineers. doi: /10.2118/77732-MS, vol 40(11).
Poettman, F. and Carpenter, P. (1952). The Multiphase Flow of Gas, Oil, and Water Through
Vertical Flow Strings with Application to the Design of Gas-lift Installations. American
Petroleum Institute.
Saepudin, D., Soewono, E., Sidarto, K., Gunawan, A., Siregar, S., and Sukarno, P. (2007). An
Investigation on Gas Lift Performance Curve in an Oil-Producing Well. International Jour-
nal of Mathematics and Mathematical Sciences. doi: 10.1155/2007/81519, vol 07(Article
ID 81519).
Slupphaug, O. and Bjune, B. (2006). Active Slug Management. Society of Petroleum Engineers.
doi: 10.2118/96644-MS.
85
86
Appendix A
Steady-state gas lift model
A.1 Model description

The usual way to estimate two-phase steady-state pressure losses is to start with the momentum
equation for the mixture in the pipe. Along a pipe segment of length x the following momentum
balance has been used.
dp 1 ρT P 2
+ ρT P gx + fT P v =0 (A.1.1)
dx 2 d m
In equation A.1.1, the pressure drop due to acceleration has been neglected. Due to lim-
ited information regarding the elevation of the well, the acceleration due to gravity has been
normalized considering the relationship between the true vertical depth and measured vertical
depth, expressed as gx .
The brief theory presented in the subsequent sections are included to give the reader an
explanation of the functionality of the model and how it simulates pressure and volume rate.
A.1.1 Superficial velocity and mass flow
Considering a pipe segment in the well in which oil, water, and gas flow. At steady-state
conditions, the mass flow through any cross section of the pipe is assumed constant. With
knowledge of the production rates at surface conditions, the volume streams down in the well
can be expressed using the black oil model. In the well, it is convenient to describe the volumes
by superficial velocities.
Ql qo Bo + qw Bw
vsl = = (A.1.2)
At At
Qg qo Bg (Rt − Rs )
vsg = = (A.1.3)
At At
87
The velocity of the mixture are then expressed as the sum of the two individual velocities.
Ql + Qg
vm = = vsl + vsg (A.1.4)
At
A.1.2 Fluid velocity, slip and volume fractions
The flow velocity of gas and liquid are linked to the cross-sectional area in the pipe occupied
by the two phases. The velocities are defined as the flow rates divided by the cross-sectional
area. Expressed in terms of superficial velocities and volume fractions we get.
Ql
Ql At vsl
vl = = Al
= (A.1.5)
Al At
yl
Qg
Qg At vsg
vg = = Ag
= (A.1.6)
Ag At
yg
Gas will usually flow faster than liquid due to the difference in density and viscosity.
Asheim (1986) proposed the following relationship combining the gas velocity to the velocity
of the surrounding liquid.
vg = Co vl + vo (A.1.7)
Where Co describes the distribution of bubbles in the flow, and vo is the buoyancy velocity
of the gas bubbles or the sink velocity for the droplets. Asheim (1986) combined the rela-
tionship between the superficial velocity and the phase velocity to the drop relationship above,
expressing the liquid fraction in the flow as
s 2
1 vsg vsl vsl 1 vsg vsl
yl = ± + Co − 1 + 4Co − + Co −1 (A.1.8)
2 vo vo vo 2 vo vo
The fraction of the gas simply gets

yg = 1 − yl (A.1.9)
Knowing the fluid fractions, the average two-phase density of the fluid mixture in the pipe
segment can be expressed in terms of fluid densities and fractions.
ρl A l + ρg A g
ρT P = (A.1.10)
At
88
A.1.3 Flux fractions and mixed flow
Flux fractions are also linked to superficial velocity and relates flow rate for each phase to the
total flowing volume.
Ql vsl
λl = = (A.1.11)
Ql + Qg vm
Qg vsg
λg = = (A.1.12)
Ql + Qg vm
The average mixture density is linked to the flux fractions and is given as
ρl Ql + ρg Qg
ρm = = ρl λl + ρg λg (A.1.13)
Ql + Qg
The two-phase friction factor, fT P , in equation A.1.1 has been estimated using a correla-
tion for single-phase flow, with a slip correction factor (Asheim, 2016).
0.16
f T P = f 0 cT P = c
0.172 T P
(A.1.14)
Rem
Where cT P is the slip correction factor and is defined as
rhog yl (1 − λl )2 + ρl yg λl2
cT P = (A.1.15)
ρm yl (1 − yl )
The two-phase friction factor has been estimated using the Reynolds Number correlation
for a homogeneous flow.
ρm vm dt
Rem = (A.1.16)
µl λl + µg λg
The theory above form the basis of the steady-state model used in the thesis. The model in-
volves many assumptions and simplifications and is not intended for modeling the flow regimes
in the well. The purpose of the model is pressure drop and volume prediction only. The in-
jection of lift-gas down in the well has also been taken into account. At steady-state flow
conditions, the model simulates the pressure at the injection point and the bottom of the well.
Among several input variables, it uses the pressure at the wellhead and the surface flow rates.
For a given wellhead pressure the model calculates two-phase properties and the pressure
at the next step down the well. At the new pressure, two-phase properties are estimated, and the
pressure is then recalculated. The procedure is repeated along the tubing down the well until
the iteration reaches the perforated section of the well. At the gas injection point, the lift-gas
is simply subtracted from the fluid stream to account for the increased gas fraction above the
injection valve. This may be observed graphically (7.3.4) by a sudden change of the pressure
89
gradient in the well. To match the output pressure with the pressure measured in the well, the
slip parameters are adjusted.
90
Appendix B
Dynamic model data
B.1 Response parameters

Table B.1.1 shows the definitions of the response parameters in the model equation and their
relations to reservoir and design parameters.
Table B.1.1: Response parameters.
Response parameter
QJ
ag (ρl − ρg ) gVxtL Qg l+Q
g
l

1 (ρl −ρg )gx L A2c Ql
aw Vt Qg +Ql ρg Qg − J w Qg
pg
c Va Jg
ft f¯m ρTd¯P v¯m ALt
91
B.2 Alternative production reference
Table B.2.1 below summarizes the production conditions for the simulated pressure response
in figure 8.2.4.
Table B.2.1: Oscillating production.
Parameter
Gas rate, [Sm3 /d] 1.01 105
Gas injection pressure, [bar] 129.0
92
B.3 Production data for sensitivity study
Table B.3.1 summarizes the production data used in the simulations in Chapter 8.3.1.
Table B.3.1: Effect of gas injection rate.
Injection rate [104 Sm3 /d]

Parameter 4.85 6.85 8.85
Wellhead pressure, [bar] 22.3 22.3 22.3
Downhole pressure, [bar] 129.3 125.0 122.5
Bottom hole pressure, [bar] 145.7 141.2 138.2
Oil rate, [Sm3 /d] 203.8 239.6 263.3
Water rate, [Sm3 /d] 430.1 505.5 556.6
Gas rate, [Sm3 /d] 8.68 104 1.13 105 1.38 105
Table B.3.2 summarizes the production data used in the simulations in Chapter 8.3.3.
Table B.3.2: Effect of productivity index.
Productivity index [Sm3 /d/bar]

Parameter 5.0 25.0 45.0
Wellhead pressure, [bar] 22.3 22.3 22.3
Downhole pressure, [bar] 129.3 122.5 131.9
Bottom hole pressure, [bar] 108.9 138.2 148.2
Oil rate, [Sm3 /d] 99.9 263.3 328.0
Water rate, [Sm3 /d] 210.7 556.1 692.2
Gas rate, [Sm3 /d] 1.07 105 1.38 105 1.50 105
93

Analysis and Prediction of Gas Lift PDF

Uploaded by

Copyright:

Available Formats

Analysis and Prediction of Gas Lift PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Analysis and Prediction of Gas Lift PDF

Uploaded by

Copyright:

Available Formats

Analysis and Prediction of Gas Lift

Stability from Field Data

Eivind Litland Myhr

Petroleum Geoscience and Engineering

Norwegian University of Science and Technology

Thank you to all my friends and fellow students.

Thank you to Helene, for all her support.

Eivind Litland Myhr

2 The gas lift system 3

3 Gas lift instability 5

5 Active feedback control and stability 17

6 Dynamic gas lift model 19

7 Gas lift stability study 27

8 Dynamic response study 53

A Steady-state gas lift model 87

B Dynamic model data 91

4.4.1 Equilibrium points and static stability. . . . . . . . . . . . . . . . . . . . . . 12

6.2.1 Simplified gas lift model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

7.1.1 Well 6507/7-A-23 schematic. . . . . . . . . . . . . . . . . . . . . . . . . . . 28

8.0.1 Mechanism of kinematic waves. . . . . . . . . . . . . . . . . . . . . . . . . 53

7.3.1 Input steady-state model 1-16. January 2017 . . . . . . . . . . . . . . . . . . . 33

8.2.1 Stationary production 1-16. January 2017 . . . . . . . . . . . . . . . . . . . . 55

B.1.1 Response parameters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

The gas lift system

Gas lift instability

3.1 Hydrodynamic slugging

3.2 System instability

3.2.2 Tubing heading

3.3 Consequences of instability

4.1 Early observations

4.3 Blick et al.

4.4 Gang and Golan

Gas Injection Rate

Figure 4.4.1: Equilibrium points and static stability.

4.5 Alhanati et al.

actual conditions in the test well. Poblano et al. E.port

74 14 February 2005 SPE Production & Facilities

4.8 Comparison of theories

Active feedback control and stability

5.1 Dalsmo and Halvorsen

5.2 Hu and Golan

5.3 Eikrem et al.

Dynamic gas lift model

6.1 General model

6.2 Current model

Figure 6.2.1: Simplified gas lift model.

6.2.2 Dynamic response

(vm δ (ρg vsg + ρl vsl ) − (ρg vsg + ρl vsl ) δvm )

δQl = −Bo Jδpw = Jw δpw . (6.2.16)

δQg = Jg δpg − Jg δpw . (6.2.17)

− [aw δpw − ag δpg ]t−∆t .

Where  = (1 + ft Jg + ft Jw ). Setting a1 =  (aw + cft Jg ), b1 =  (aw + cft Jg ), a2 =

Neglecting flow friction gives ft = 0 and leads to,  = 1, a1 = aw , and b1 = ag . Equation

Gas lift stability study

7.1 The Heidrun case

7.1.1 Data availability

7.1.2 Well design and completion

Pwf Perforation depth

Figure 7.1.1: Well 6507/7-A-23 schematic.

Where = (1 + ft Jg + ft Jw ). Setting a1 = (aw + cft Jg ), b1 = (aw + cft Jg ), a2 =

Neglecting flow friction gives ft = 0 and leads to, = 1, a1 = aw , and b1 = ag . Equation