Continuous gas lift performance analysis

Continuous gas lift performance analysis

Alcino Resende Almeida

An in-house gas lift simulator accurately computes contributions of each valve and reveals origin of several problems. Correct troubleshooting results and fine tuning enhances production and/or reduces injection rates

This article describes the basic principles and elements of a continuous gas lift simulator developed by the Petrobras Research Center in Rio de Janeiro, Brazil. The introduction overviews gas lift, why it is important to Petrobras’ offshore operations, and the limitations of most commercial software. The simulation approach and the concept and function of the computational package (steady-state simulator) which was developed are described, including use of available gas-lift dynamic models. Field examples from the PM-29 well in Pampo field and a well in Namorado field are discussed, and the data is illustrated.

The authors conclude that the inhouse developed software has shown reliable results in optimizing gas lift wells, beyond the capability of its “commercial counterparts.” However, it is recognized that an even better tool may be created by inserting the procedure into a “more general flow simulator” available in the market.


Gas lift is one of the most important artificial lift methods used worldwide and is considered a standard method for offshore applications.[1] In Brazil, about 80% of the oil production comes from offshore wells. Continuous gas lift is responsible for at least half of this figure, and its usage tends to increase due to the great number of offshore flowing wells that will produce by gas lift in the near future. Besides its flexibility and ability to handle several common operational problems, continuous gas lift reliability has warranted its wide offshore application, especially in subsea wells. However, gas lift efficiency is severely reduced when the well is located far away from the platform; some alternatives are being currently studied by Petrobras.[2]

After the gas lift installation kick-off, a performance analysis is frequently made, since, at design time, the engineer was faced with a lack of reliable data, especially for new wells. In these occasions, the flexibility of gas lift is indeed of great value. Sometimes, even old installations have to be routinely analyzed to check some operational occurrences, e.g., a problem in a gas lift valve, or to fine tune the well in an optimization procedure.

In the past, this analysis was limited since it relied on standard charts and hand calculations; graphical methods were a must. The progressive availability of powerful personal computers changed this scenario dramatically and new possibilities are opened to artificial lift designers. Nevertheless, some commercial computer algorithms are, in fact, simple translations of the old manual procedures and do not use all the potential offered by the computer. Moreover, recent developments in the areas of dynamic behavior of gas lift valves, transients and stability are being slowly introduced.

The great importance of continuous gas lift in Petrobras’ production profile, and the challenges related to the subsea wells, gave rise to adoption of an aggressive approach, including an innovative design method. Since 1986, research in this area has been developed at Petrobras Research Center (CENPES), and one of the most valuable issues of this extensive work is the computational package VGLC (an acronym of the Portuguese expression for Continuous Gas Lift Verifier). From the earliest version of this program run in mainframe computers to the latest version developed for Microsoft Windows 95, several improvements were made, incorporating experience of Campos basin personnel, and thereby creating a unique computational tool.


The VGLC is a steady-state simulator designed to check continuous gas lift installations that are already in operation. After collecting the required well data, one can run the program, obtain a picture of present well condition, and know if there is room for improvement. Sometimes, the objective is to reduce the amount of injected gas, instead of increasing production rate. With VGLC results, the user may decide which actions are most needed.

Multiphase flow simulators generally work with the concept of equilibrium point (or node), in which, inflow and outflow behaviors, given by relationships between flowing pressures and rates, are confronted against each other. The equilibrium solution is obtained as a flowing pressure/rate pair that simultaneously satisfies inflow (allowable behavior) and outflow (required behavior). At a glance, a convenient way of performing a nodal analysis is to select the bottom of the well as the equilibrium node, since pressure at the surface is normally known; then “marching” numerical techniques are usually applied to solve the momentum ordinary differential equation (ODE). However, the equilibrium point at the wellhead or at the separator is a better choice for a couple of reasons, as described below.

Calculating each valve’s contribution. Most gas lift analysis programs assume gas injection at a single point with a user-supplied gas injection rate. Moreover, only recently, the dynamic behavior of gas lift valves has been included. This may prevent realistic simulating, since multi-point injection may happen in a practical application. Sometimes, the user is interested in evaluating the response of the well when submitted to a change in a parameter, e.g., injection gas pressure. This change may induce closed valves to open, and the simulator has to take this fact into account. In such a case, selecting the equilibrium node at a position downstream of the gas lift valves, usually at the separator or system’s first node, is advantageous.

An upstream calculation using a marching numerical technique will require a very complicated “shooting” approach since the gas rates crossing each gas lift valve are not a priori known – in fact, they are an important part of the response to be obtained from the simulator itself. Therefore, it is better to use such a “shooting” approach, departing from the bottom of the well and easily calculating the contributions of each valve while “marching” upstream. The possible existence of multiple solutions, as we will see later, is per sea strong argument for taking the equilibrium at the surface.

Temperature profile. This is another important issue. A gas lift simulator must take into account the energy balance along the well, since the mostused gas lift valves have a nitrogencharged bellows, which is very sensitive to temperature changes. This may have a remarkable effect on gas flow rate predictions, especially in the throttling flow regimes.[3] Solving the energy equation coupled with the momentum equation poses a two-point boundary value problem, since the pressure at the separator and the reservoir fluid temperature are usually the conditions which the differential equations have to satisfy.

Two basic classes of numerical methods may be used:[4] “shooting” and relaxation. It was considered preferable to adopt a shooting class method, with the starting point at the bottom of the well and the equilibrium point at the first node. Sometimes, especially in presence of long subsea flowlines, production stream temperature reaches equilibrium with the medium (sea) temperature at a certain distance from the first node, i.e., from this point, both sea and flow temperatures are the same, at least within machine precision. This reinforces the technique of calculating by marching from the bottom of the well, upward.

System unit concept. The VGLC program works with the concept of “system units.” A unit is a segment of a production system (well and flowlines) in which the characteristics, e.g., diameters, are the same throughout. Points between units are called nodes, the first node being at the surface (generally the separator), the last one at the bottom of the well (generally at the perforations). The user supplies system and gas-lift valve features in separated data sheets, and the VGLC creates additional units, subdividing the user system configuration and positioning each valve in the base of a new unit. Then, the user chooses some correlations and procedures, gives basic fluid properties, establishes the reservoir model (IPR), and the program is ready to run and finds out the equilibrium rate for a given gas injection pressure.

As stated above, the equilibrium node is taken at the first node of the system. Given a production rate and, as a consequence, a flowing pressure (from the IPR), the ODE’s momentum and energy are solved from the last node to the first using a marching numerical technique. In general, a discrepancy from the first node’s required flowing pressure (the separator pressure for most cases) arises, establishing a root-finding problem that is treated by means of the so-called Pegasus method which is a variant form of the secant method.[5]

Three models, correlations. As the integration proceeds from the bottom of the well to surface, each gas lift valve is checked with respect to eventual gas injection, and the corresponding gas rate is incorporated into the main stream. Three gas-lift valve dynamic models are available: TUALP,[6] Winkler & Eads,[7] and an in-house model developed from Nieberdings’s results.[8] Some Campos basin users have reported that the best fits have been obtained using the Winkler & Eads model. However, this is an indirect observation about the models and there is no consensus until now.

The two-phase calculations are performed by means of properties and gradient correlations previously chosen by the user. Some of these correlations were developed aimed at Campos-basin-specific conditions, but most of them are commonly found in the literature. Many of the related routines came from the TUFFP CORE library and were improved later.[9,10] The temperature profile is computed using a slightly modified procedure than that given by Alves et al.[11] Static conditions are assumed for the gas in the annulus when calculating injection gas pressure and temperature profiles. The bellows temperature is taken as the average between gas injection and production temperatures using weights supplied by the user, a contribution of 30% for the production temperature being the recommended for Camco BK-1 valves.[12]


Fig. 1 and the accompanying table summarize basic data for the PM-29 well in Pampo field, Campos basin. The best set of correlations to predict pressure drop was selected based on previous simulations by matching VGLC results against complete production test data, i.e., a reliable test with bottomhole pressure and temperature measurements. After a few months, the well began strong oscillations (heading) and the operational action taken, at that moment, was to choke the well, increasing the wellhead pressure. The VGLC shows the origin of this heading problem, Fig. 2. Lowering wellhead pressure, a simultaneous injection through Valves 1 and 2 was induced (Valve 1 probably lost pressure and Valve 2 was probably operating in an unstable condition).

Additional simulation suggested that improvement was possible by replacing Valve 1 with a dummy, and injecting through Valve 2 alone, Fig. 3. A strong increase in production was expected, i.e., to 1,190 [m.sup.3]/d from 700 [m.sup.3]/d (to 7,500 bpd from 4,400 bpd). After wireline operation, to introduce the above change, a stable production of 1,150 [m.sup.3]/d (7,200 bpd), at 56% watercut, was obtained, confirming simulator predictions.

Occurrence of multiple solutions (multiple possible operating points) is not unusual, especially for high productivity wells. A good design may avoid this problem, although uncertainties frequently impair achievement of this goal. Moreover, the gas lift string may have to be designed to attain good performances over a long period of time without changes (typical for subsea wells). Some adjustments in design rules to cope with related changes in well conditions may introduce additional operating points. Transient unloading simulators are valuable[13,14] in determining the correct unloading procedure to achieve the desired operating point.

Fig. 4 illustrates a multiplesolution case for a well in Namorado field in Campos basin. Considering the required pressure at surface, four solutions are possible. Those solutions are close together. The system may operate in one (stable) of those four points and a change in well conditions (casing pressure, for example) may move this operating point. The new point may be unstable, and even small system perturbations may, again, easily move the operating point into another stable or unstable point.

This introduces oscillations in the well response; and misleading interpretations by the operating personnel may result. Although transient, the problem origin can easily be seen in the outcome of the steady-state simulator VGLC. One may suggest as a possible solution, an increase in the casing pressure. Fig. 5 shows that this action may introduce more solutions, e.g., injection through an upper valve is now possible. In such cases, a review of the string design is the best option.


A continuous gas lift simulator was developed to fit the needs of Petrobras; and it has become the most-used software in Campos basin, for this type of application. Typically, when speaking about simulators, many improvements may be devised, especially the incorporation of transient analysis. Despite this, the Continuous Gas Lift Verifier (VGLC) has shown reliable results and has helped operating personnel optimize gas lift wells in a fashion not accomplished by its commercial counterparts.

Although the effort of developing an in-house software was thoroughly justified, a better tool may be obtained by inserting the VGLC procedure into a more general flow simulator available in the market. At this point, enhancement of VGLC will continue within Petrobras due to its strategic importance for daily operations.

Basic data for Well PM-29, Pampo field, Campos basin

Valve configurations/dimensions

No. Designation Size, in. Press., kgf/[cm.sup.2]

1 R-20 5/16 97.2

2 R-20 5/16 94.7

3 R-20 5/16 93.0

4 RDO 20/64 –

Fluid properties (29 [degrees] API)

Formation water gravity = 1.04 Injected gas gravity = 0.68 Formation gas gravity = 0.74 Watercut, % = 64 Formation gas-liquid ratio, [m.sup.3]/[m.sup.3] = 50

IPR (Vogel)

Static pressure, kgf/[cm.sup.2] = 175 Bubble pressure, kgf/[cm.sup.2] = 189 Potential rate, [m.sup.3]/d = 8,090

Correlations used

Gas solubility, bubble pressure, oil formation volume factor: Standing

Dead oil viscosity: Beggs and Robinson

Vertical flow: Hagedorn and Brown

Dynamic behavior of gas-lift valves: Winkler and Eads


The authors acknowledge Petrobras for its permission to publish this paper. They also thank M. Brennand who provided the field example data, G. A. Peixoto for his valuable comments, and K. Minami for a careful revision of the manuscript. Many people made contributions during development of the VGLC, especially in the earliest versions; these contributors include: F. J. S. Alhanati (now with C-FER), E. C. Capucci, J. R. Fagundes Netto, D. Lopes, B. R. Motta Filho, T. Nakaya, G. P. H. A. Oliveira, G. A. Peixoto and J. A. P. Prince. This list is certainly incomplete and the authors apologize for any omissions. The participation of the Campus Basin Artificial Lift team (E&PBC/GESCOM) in all development phases of VGLC was of paramount importance, and the authors are grateful to all of them.


1 Clegg, J. D., S. M. Bucaram and N. W. Hein, Jr., “Recommendalions and comparisons for selecting artificial-lift methods,” JPT, December 1993, pp. 1, 128-1,657.

2 Camargo, R. M. T., I. N. Alves and M. G. Prado, “Advances in artificial lift and boosting systems for subsea completion,” paper OTC 8475, presented at the Offshore Technology Conference, May 5-8, 1997, Houston, Texas.

3 Sugar, R. K., “Improved dynamic model of gas-lift valve performance,” M.Sc. Thesis, The University of Tulsa, Tulsa, Oklahoma, 1991.

4 Press, W. H., et at., “Numerical recipes in Pascal,” Cambridge University Press, Cambridge, 1994.

5 Gjumbir, M. and Z. Olujic, “Effective ways to solve single non-linear equations,” Chemical Engineering, July 23, 1984, pp. 51-56.

6 Bertovic, D., “Gas-lift valve performance – Version 1.5 (Software description),” TUALP – The University of Tulsa Artificial Lift Projects, 1995.

7 Winkler, H. W. and P. T. Eads, “Applying the basic performance concepts of single-element, unbalanced gas-lift valves for installation design,” SPE Production and Facilities, August 1993, 211-6.

8 Nieberding, M., “Normalization of nitrogen loaded gas-lift valve performance data,” M.Sc. Thesis. The University of Tulsa, Tulsa, Oklahoma, 1988.

9 Jefferson, L., F. Kung and A. L. Corcoran, III, “TUFFP CORE – Software users Manual (Version 2.0),” Tulsa University Fluid Flow Projects, The University of Tulsa, Oklahoma, March 1989.

10 Almeida, A. R., “Simple hints improve computational performance of two-phase flow routines,” In Situ (submitted), 1998.

11 Alves, I. N., F. J. S. Alhanati and O. Shoham, “A unified model for predicting flowing temperature distribution in wellbores and pipelines,” paper SPE 20632, presented at the 65th SPE Annual Technical Conference and Exhibition, New Orleans, Louisiana, 1990.

12 Bertovic, D., “A unified model for gas-lift performance incorporating temperature effects,” M.Sc. Thesis, The University of Tulsa, Oklahoma, 1995.

13 Capucci, E. C. and K. V. Serra, “Transient aspects of unloading oil wells through gas-lift valves,” paper SPE 22791, 1991.

14 Hall, J. W. and K. L. Decker, “Gas-lift unloading and operating simulation as applied to rnandrel spacing and valve design,” paper SPE 29450, 1995.

The authors

Alcino Resende Almeida, an artificial lift teamleader at Petrobras R&D Center (CENPES) in Rio de Janeiro, Brazil, holds BS and MS degrees in mechanical engineering from the Federal University of Rio de Janeiro. He joined Petrobras in 1984, and worked beth as a field and design engineer, mainly in the artificial lift area, in Bahia state, Brazil, until 1988, when he was assigned to his present position.

Ivan Slobodcicov, a specialist in artificial lift software development at the Petrobras R&D Center, holds a BS degree in electronic engineering from the Federal University of Rio de Janeiro. He joined Petrobras in 1987, and worked as a field petroleum engineer in Rio Grande do Notre state, before moving into his present position in 1992.

COPYRIGHT 1998 Gulf Publishing Co.

COPYRIGHT 2000 Gale Group