TECHNOLOGY FOCUS
Artificial Lift There is nothing wrong with change, if it is in the right direction. —Winston Churchill In the artificial-lift community, do we understand the current technology well enough to know what improvements are needed? Are we advancing technology or just providing “band-aid” solutions because the root cause of failure is not understood? Where does the industry need to focus effort—technology development or reduction of failures caused by poor design, installation, and operating practices? Questions such as these were the driving force behind two industry consortiums to develop databases to collect thousands of electrical-submersible-pump (ESP) and progressing-cavity-pump (PCP) installation and failure records for performance analysis and benchmarking (www.esprifts.com and www.pcprifts.com, respectively). Data collection and its analysis from the databases have become very valuable in understanding the performance and reliability of these lift systems. Some of the lessons learned from the PCP consortium will be presented at the 2010 SPE Progressing Cavity Pumps Conference (PCPC) to be held in Edmonton, Alberta, Canada, in September. Both the PCP and ESP industries have been impaired by a lack of standardization, especially in the areas of equipment rating, testing, and nomenclature. In 2009, the International Organization for Standardization (ISO) published a revised ISO 15136-1 as an international standard for PCP equipment, and a session on its use will be presented at the 2010 SPE PCPC. Currently ISO 15551-1, which is expected to be the new international standard for ESP systems, is in development and will, undoubtedly, be a hot topic for discussion at the 2011 SPE ESP Workshop in Texas in April. The reliability-improvement consortiums and the development of new international standards are helping to define where we are today as an industry and provide clear direction for technology advancement. The three artificial-lift papers featured in this section contain unique solutions that may become starting points from which other developments may evolve: The geared centrifugal pump has a distinctive downhole-transmission assembly that offers an opportunity for other pump configurations to be driven by a rod string, a method to use gas lift above and below a packer in a single completion with long perforation intervals might expand to areas other than unloading tight gas wells, and the use of hybrid data-evaluation models to visualize and analyze production information better may become an essential tool for artificial-lift selection and optimization. The papers selected for additional reading also contain innovative artificial-lift solutions. JPT
Shauna Noonan, SPE, is a Staff Production Engineer for ConocoPhillips where she works as an artificial-lift specialist in the Completions and Production Technology group. Noonan’s responsibilities include development and validation of artificial-lift and completion systems for thermal applications and improving artificial-lift reliability. She has worked on artificial-lift projects worldwide at ConocoPhillips, and previously at Chevron, for more than 17 years. Noonan has chaired industry forums and committees and authored or coauthored numerous papers on artificial lift. She serves as a member of the SPE Production and Operations Advisory Committee, as an Associate Editor for SPE Prod & Oper, a member of the JPT Editorial Committee, and she will cochair the 2010 SPE PCPC. Noonan began her career with Chevron Canada Resources and holds a BS degree in petroleum engineering from the University of Alberta.
Artificial Lift additional reading available at OnePetro: www.onepetro.org SPE 124926 • “Real-Time Diagnostics of Gas Lift Systems Using Intelligent Agents: A Case Study” by G. Stephenson, SPE, Occidental Petroleum, et al. (See JPT, May 2010, page 55, and SPE Prod & Oper, February 2010, page 111.) SPE 124646 • “Application of Beam-Pumping Wells’ Energy-Saving Rebuilding Techniques in Daqing Oil Field” by Wang Fengshan, SPE, Daqing Oilfield Co., et al. SPE 124515 • “Plunger-Lift Modeling Toward Efficient Liquid Unloading in Gas Wells” by G. Chava, SPE, Texas A&M University (now at Chevron), et al. (See SPE Proj Fac & Const, March 2010, page 38.) 50
JPT • JULY 2010
ARTIFICIAL LIFT
The Geared Centrifugal Pump: A New High-Volume Lift System The geared centrifugal pump (GCP) is a high-volume artificial-lift system consisting of a progressing-cavity-pump (PCP) -style rotating rod string driving a bottom-intake electrical-submersiblepump (ESP) -style multistage centrifugal pump by use of a downhole speedincreasing transmission. The heart of the system is the unique transmission that uses a novel gearing configuration that allows high torque and power. The GCP provides the high-volume lift of an ESP but with better gas handling, simpler operation, and lower capital and operating costs.
Introduction The GCP is an artificial-lift system consisting of the rotating-rod-string drive of a PCP, driving the multistage centrifugal pump of an ESP by a downhole speedincreasing transmission. The transmission is required to increase the relatively slow rotational speed of the PCP rod string, typically less than 500 rev/min, up to the 3,500-rev/min operational speed of the centrifugal pump. The GCP was developed as an artificial-lift system to provide the high-volume lift of an ESP without the expensive and troublesome downhole electric motor and cable of the ESP system. The GCP uses efficient and relatively inexpensive industrial electric motors in the surface drive head, requiring none This article, written by Assistant Technology Editor Karen Bybee, contains highlights of paper SPE 123938, “The Geared Centrifugal Pump: A New HighVolume Lift System,” by J.C. Patterson, SPE, ConocoPhillips; W.B. Morrow, SPE, Harrier Technologies.; and M.R. Berry, SPE, Mike Berry Consulting, originally prepared for the 2009 SPE Annual Technical Conference and Exhibition, New Orleans, 4–7 October. The paper has not been peer reviewed.
of the special controls or transformers necessary for high-voltage ESP motors. The rotating rod string transmits power downhole with low power losses. This system combines the highly efficient GCP transmission assembly with a modern ESP pump. This combination results in a system with lift efficiency greater than 50%. Because all of the downhole components of the GCP are mechanical, the design is adaptable to very-high-temperature applications such as steam-assisted gravity-drainage (SAGD) projects. GCP Components Drive Head. The GCP uses the same drive system as a PCP system. A surface drive head turns a rotating rod string that drives the downhole assembly. The only difference between a PCP drive head and one built for the GCP is the load-capacity requirements of the thrust bearing carrying the rod load. In a PCP drive head, the thrust bearing must carry the weight of the rod string as well as the fluid load. Because the rod string of the GCP does not carry any fluid load, the thrust bearing in the GCP drive head can have a correspondingly lower capacity. Another characteristic of both PCP and GCP drive heads is prime-mover flexibility. Although the majority of well locations have electric power, the GCP can be powered by an internal-combustion engine, a hydraulic motor, or an electric motor. One of the features of the GCP drive head is the ability to change the pump rotational speed without having to install a permanent variable-speed drive (VSD). The typical drive head uses a belt-drive speed reducer between the prime mover and the rod string. If the sizing of the GCP installed in a well turns out to be less than optimum, the pump speed can be adjusted to opti-
mize the lift merely by changing the size of the motor sheave on the drive head. This capability allows the use of a portable VSD to determine the optimum pump speed on the basis of the productivity of the well, and then size the motor sheave to give that optimum speed. This capability gives the operational flexibility of having a permanent VSD without the associated cost. If, however, the sizing of an ESP is found to be less than perfect, optimizing lift by changing the pump speed requires a permanent-VSD installation, and it also may require changing the downhole motor voltage to optimize the power requirement. Optimizing a GCP to maximize the productivity of a well should be simpler and less expensive than for an ESP. GCP Transmission. The heart of the GCP is the speed-increasing transmission. Transmissions that have constraints on the size of the gears because of space limitations, such as the inside diameter of oilfield casing, but are nonetheless required to transmit significant power present a difficult problem. The most obvious potential solution is to use a planetary or related epicyclic configuration, which has a high power density and concentric input and output. However, the small diameters of oilfield casing severely limit even the power of a planetary transmission—a situation worsened by the fact that significant annular space must be provided between the transmission case and the outer pressure housing as a produced-fluid flow channel, further reducing the already small gearbox diameter. The problem is that an epicyclic transmission has only one gear set to handle all the input, and there just is not sufficient steel in the single gear set in these small diameters to handle significant torque and power.
For a limited time, the full-length paper is available free to SPE members at www.spe.org/jpt. JPT • JULY 2010
51
the D-tubes filled with the relative cool produced well fluid. This close proximity allows rapid and effective dissipation of the heat generated in the gear meshes and bearings. Keeping the gear train, and hence the lubricating oil, cool is critical to transmission life. Fig. 1—Schematic of PHG gear train after load balance is achieved.
One potential way to obtain more gear sets and hence more steel in the gear train is to use a parallel-shaft configuration, with multiple gears on a common shaft. Although this option is appealing in concept, it has proved, in the past, to be impractical because of the difficulty in achieving load sharing among the multiple gears. That is, because of manufacturing tolerances, the multiple gears are not likely to be aligned perfectly and do not all engage simultaneously. The misalignment causes the gears to engage unevenly, resulting in some of them being loaded more heavily than they are designed for, causing premature failure as a result of excessive load or wear. Paired Helical Gearing (PHG). The solution to this problem is a new gearing configuration termed PHG. A PHG gear train consists of pairs of helically cut gears arranged along common shafts. Each half of the pair is cut in the opposite sense, so they look like herringbone gears, except that the two halves are not attached and are free to move axially relative to one another. The shaft is splined, as are the gears, so that the gears are free to move laterally along the shaft, but can drive the shaft or be driven by it. The basic principle that results in load sharing in a PHG gear train is that helical gears exert an axial force that is proportional to their radial, or torque, load. This axial force is in a direction that would move the gear away from engagement were the gear free to move axially on its shaft. Conversely, if the gear were pushed axially in the opposite direction, the gear would move into greater engagement. If a load were imposed on such a gear train, the first thing to happen, before any significant rotation occurred in either shaft, would be the gears reacting to the tooth loads and immediately moving axially along their respective shafts until they bumped into one another (Fig. 1). At this point, the gears would begin to bear increas-
52
ing but varying tooth loads, depending upon the relative and uneven timing of each gear. The gears that bear the higher loads push axially more strongly than their neighbor gear, moving away from engagement, while at the same time pushing those neighbor gears into more engagement. This process continues among all the gear pairs until the axial forces are opposite and equal, and no further lateral movement occurs. Because the axial force of a helical gear is proportional to the torque load on the gear, when axial forces among the gears are balanced, gear torsional loading also is balanced. Because multiple gears can be mounted on a common shaft, and load sharing ensured, the capacity of a PHG transmission is not limited by gear strength or wear issues but by the ultimate strength of the input shaft—a much less restrictive criterion. If a more powerful transmission is desired, all that is required is the addition of more pairs of gears until the input-shaft strength is reached. Little additional design work is needed to increase the capacity of the transmissions significantly, and the incremental manufacturing costs are minimal. The ability to add pairs of gears as needed also has profound effects on bearing issues. Bearing life and capacity frequently are the limitation of conventional transmissions, particularly epicyclic gear trains where the heavily loaded planet gear bearings are the Achilles heel of the designs. The PHG allows the addition of pairs of gears solely to increase bearing capacity and life. The PHG layout allows easy cooling of the gear train. Planetary transmissions, with the heavily loaded planet gears “buried” deep inside the gearbox and shrouded by the planet carrier and annulus gear, have problems with cooling. With the parallel-shaft configuration of the PHG transmissions, all the gears and bearings are in close proximity to the external cooling medium—in the case of the GCP, the lubricating oil the gear train is immersed in and
Field Trial The first field trial of a GCP was performed in a west Texas well formerly equipped with an ESP. The well had been producing approximately 1,300 BFPD with a 98% water cut and a very low gas rate from 4,600 ft when it was converted to GCP lift. The principal purposes of the conversion were to understand the installation and operational issues of a GCP in an actual oil well, determine the “real-world” efficiency of this new artificial-lift system, and test some of the features unique to the high-temperature GCP. The well, a producer in a carbonate waterflood, was chosen because it has 7-in. casing through the productive interval and it has sufficient total-fluid productivity to test the operational lift capacity of the GCP. The GCP installed in the well is a normal-temperature version of the 100-hp high-temperature model being planned for use in a SAGD project in Alberta and, hence, is designed for the 7-in. or larger casing used in that project. The unit ran for 4 weeks before failing as a result of rod-coupling fatigue at the top of the stab-in connector. The rod-connector coupling was redesigned, and the well was put back on production but went down after 2 more weeks because of a faulty weld in one of the D-tubes. The bad weld allowed water into the transmission, resulting in multiple bearing failures. Tear down of the entire assembly showed that the seals and compensator operated as designed, and that the transmission was showing normal wear when the water contamination caused bearing failure. During the 6 weeks of operation, the performance of this first installation of a GCP went as might be expected. The construction flaw in the D-tube was disappointing, but no fundamental design problems were discovered. Modifications to the prototype D-tube construction have been made, which should minimize this problem with future prototypes. Production GCP transmissions will most likely use JPT shaped seamless tubes.
JPT • JULY 2010
ARTIFICIAL LIFT
Annular-Velocity Enhancement With Gas Lift as a Deliquefication Method It is common practice in the industry to complete multiple reservoirs in a single wellbore to establish commercial production rates from tight gas completions. In some cases, these zones are separated vertically by several hundred to more than 1,000 ft. The deliquefication of these completions is one of the most challenging problems facing our industry. Many of the deliquefication methods currently being used are effective over only a portion of the wellbore.
Introduction It is common practice to complete tight gas wells with the tubing tail above the bottom perforation. This completion practice allows static liquid to stand across the lower set of perforations. Static liquid standing across the perforations promotes increased near-well liquid saturations through a process of spontaneous imbibition. The imbibition process is a damage mechanism that can be present upon initial completion and can continue throughout the productive life of the well if allowed to form through ineffective liquid-removal/completion practices. There are a number of different methods currently in use for the deliquefication of gas wells. Some of the more common are plunger lift, chemical-foam lift, and conventional pumps (i.e., rod This article, written by Assistant Technology Editor Karen Bybee, contains highlights of paper SPE 130256, “Annular Velocity Enhancement With Gas Lift as a Deliquification Method for Tight Gas Wells With Long Completion Intervals,” by S.A. Pohler, SPE, W.D. Holmes, SPE, and S.A. Cox, SPE, Marathon Oil Corporation, originally prepared for the 2010 SPE Unconventional Gas Conference, Pittsburgh, Pennsylvania, 23–25 February. The paper has not been peer reviewed.
pumps, electrical submersible pumps, and progressing-cavity pumps). Each method has drawbacks when it comes to deliquefying wells with long completion intervals. Effective liquid removal can improve overall well performance significantly. However, an effective system cannot ignore the potential presence of a static liquid column. Current completion technology has been found to have limited effectiveness when long productive intervals are involved. Therefore, a modified continuous-flow gas lift system is proposed that includes a packer and crossover system and that allows for gas lifting the well below the bottom perforation. The details of this completion process and its field applications are discussed in the body of the full-length paper. New Artificial-Lift System An alternative type of lift system has been deployed to effectively deliquefy wells completed over long intervals. The system is a modified continuousflow gas lift system that allows for lifting intervals below a packer using side-pocket mandrels and wirelineretrievable gas lift valves both above and below the packer. Lifting below the production packer allows the well to be gas lifted below the bottom perforation, which effectively addresses the staticliquid-column issue. This is accomplished by installing a bypass access mandrel immediately above the packer in the well. A wireline-retrievable crossover assembly is inserted into the bypass access mandrel. This assembly extends through the packer into a lower seal receptacle that is located below a slotted nipple. Side-pocket mandrels are installed above and below the packer. The gas lift valves installed above the packer are standard tubing-flow valves, while the gas lift valves below the packer are casing-flow valves. The flow from
the formation below the packer is up the tubing/casing annulus. To lift the liquid from the wellbore effectively, the tubing must be sized such that critical velocity in the annulus is approximately the same as the critical velocity in the tubing above the packer. This is accomplished either by running larger tubing below the packer or by wrapping the tubing below the packer externally with fiberglass to the desired outside diameter (OD). To obtain maximum lift, the tubing below the packer is run to a point below the lowermost perforation in the well. A gas lift orifice with a back check is installed at the bottom of the string as the planned operating gasinjection point. The critical velocity for the casing/ tubing annulus cross-sectional area was assumed to be 30 to 50% less than the critical velocity in tubing with the same cross-sectional area. On the basis of this assumption the recommended tubing installation inside of 51/2-in., 17-lbm/ ft casing is 31/2-in. flush-joint tubing or 27/8-in. tubing externally wrapped with fiberglass to an OD of 4 in. Below-Packer Annular Gas Lift With Velocity Enhancement. Fig. 1 shows a a normal annulus gas lift system above the packer, where gas is injected down the annulus and then the gas crosses over and is injected below the packer through the tubing. This allows the well to be lifted from the bottom of the perforated interval. The increased OD of the tubing increases fluid velocity in the annular area below the packer by reducing the cross-sectional flow area. Crossover System. Fig. 2 shows how the gas is injected through the crossover system and then is injected down the inside of the larger tubing below the packer. Below the packer, production flows up the annulus, crosses
For a limited time, the full-length paper is available free to SPE members at www.spe.org/jpt. 54
JPT • JULY 2010
27/8-in. tubing
Lift Gas
Production
Lift Gas
Side pocket gas lift mandrel
Larger OD tubing
Fig. 1—Normal annular gas lift system.
over at the packer, and flows up the tubing above the packer. The internal part of the crossover assembly is wireline retrievable, which allows wellbore access below the packer. New-System Benefits The below-packer annular gas lift system with velocity enhancer makes it possible to produce zones several thousand feet long. The system uses standard wireline-retrievable gas lift valves below the packer and allows for running pressure/temperature memory tools below the packer. It also allows access to the wellbore below the perforations and reduces the lift gas required to reach critical velocity. In addition to deliquefying the completion effectively, this system provides an excellent opportunity to chemically treat the entire wellbore for corrosion. This can be accomplished by atomizing corrosion inhibitor into the gas-injection stream. The injected gas then will carry the corrosion inhibitor through the entire wellbore. The below-packer annular gas lift system with velocity enhancer has minimal restrictions through the system, and it removes more liquid and results in increased production compared with conventional completions. The system can be run both in vertical and horizontal wells. In horizontal
JPT • JULY 2010
Production
Fig. 2—Gas injected through the crossover system.
wells, the packer and the crossover are set in the vertical section of the hole, the tubing is run to the lowest point in the horizontal completion so that gas can be injected to sweep the liquid from the horizontal section, and an orifice is run either at the end of the tubing or in the crossover assembly. The system can be run in 41/2-in. and larger casing sizes. To date, 16 systems have been run in east Texas; one of the systems is in a horizontal well. The majority of the installations have been in wells equipped with 51/2-in. casing and 27/8-in. tubing above the packer. Incremental rate after the installation of the system has averaged approximately 200 Mcf/D. Field Application The first example well is producing from the Travis Peak reservoir from a depth of approximately 9,600 ft. The well is completed in six intervals, with a vertical separation of 2,058 ft. A rod pump was installed on the well at 1,600 days of production. After 2,400 days, the well was recompleted to add additional zones to the well. The rod pump was removed as part of this recompletion. Before the installation of the new gas lift system, the well was producing with foam injection down the casing.
The foam injection was unsuccessful in maintaining stable production from the well. The base decline for this well was 35%/yr, established while the well was rod pumped. The gas lift installation resulted in an initial rate of 500 Mscf/D and a stable decline of 11%/yr. The incremental recovery from the gas lift system is estimated at 1.3 Bscf on the basis of decline-curve projections. The second example well is a horizontal completion producing from the Pettet reservoir at approximately 13,386 ft measured depth, with a horizontal displacement of approximately 4,400 ft. The well was completed with an uncemented liner and was fracture stimulated. As a result of the limited formation-flow capacity, the well was unable to flow naturally at rates sufficient to unload the wellbore and remained shut-in. During shut-in, the well would build pressure but it was not able to be returned to production because of liquid loading. The decision was made to install tubing and gas lift in an attempt to return the well to production. As a result of this work, the well currently is producing at a stable rate of 180 Mscf/D. The full-length paper contains wellbore diagrams showing how the completion system works in vertical and JPT horizontal wells.
55
ARTIFICIAL LIFT
Integrating Data Mining and Expert Knowledge for an Artificial-Lift Advisory System The full-length paper describes a new workflow to accelerate and improve decisions regarding where and when to apply an artificial-lift system in fields with a considerable number of active wells. The workflow deploys a hybrid combination of a user-driven expert system and a data-driven knowledgecapturing system calibrated with historical data. The two systems interact to determine the right point in time to support a particular well with an artificial-lift system.
Introduction Typically, asset expertise is stored in spreadsheets that define the limits for condition parameters within which a specific artificial-lift system needs to be installed. As a first step toward decision automation, these limits are implemented in a rule-based reasoning system deploying if-then-else logic to automate the knowledge. After successful validation of the rule-based approach, the information about the limits and the combination and interaction of the parameters is fed into a Bayesian belief network (BBN) capable of performing a more robust reasoning process under uncertainty or missing parameters. The setup process of the BBN requires frequent interacThis article, written by Assistant Technology Editor Karen Bybee, contains highlights of paper SPE 128636, “Integrating Data Mining and Expert Knowledge for an Artificial Lift Advisory System,” by E. De la Vega, G. Sandoval, and M. Garcia, SPE, Pemex, and G. Nunez, SPE, A. Al-Kinani, SPE, R.W. Holy, SPE, H. Escalona, SPE, and M. Mota, Schlumberger, originally prepared for the 2010 SPE Intelligent Energy Conference and Exhibition, Utrecht, The Netherlands, 23–25 March. The paper has not been peer reviewed.
tion with the asset team to ensure the validity and consistency of the system recommendations. While the iterative validation is necessary for the BBN setup, it also enables the asset team to review and visualize their complex decision-making process thoroughly and helps to define more-efficient decision logic. In parallel, a data-mining analysis on historical data is performed to find the decision logic as it was applied in the field. The data-mining study incorporates operational data (e.g., tubinghead pressures, line pressures, and calculated gas rates), well-test data, and information from well-intervention reports. The objectives of the datamining analysis are: • To identify patterns in operational and well-test data that lead to the decision of a new artificial-lift system. • To identify potential correlations of well-intervention activity and incremental production after intervention to quantify the success of a specific artificial-lift installation under specific conditions. Case Study Activo Integral Burgos (AIB) is a typical example of a large gas field where production declined as a result of gasloading backpressure and reduced permeability in the target formation. The fast decline of the gas wells during their first year of production drove a change from a reactive to a proactive management approach to monitor the field and select candidates for workovers and artificial-lift systems. However, the large number of wells in AIB (approximately 3,500 active wells) and the fact that 95% of the data is obtained by manual surveillance suggest the need for some level of automation to aid decision making.
Current Situation. Under the current working environment, two main situations challenge the asset team: 1. Analysis of the huge amount of data involves a significant time delay, which leads to lost production because of delayed well intervention (e.g., installing the right artificial-lift system in time). 2. There is insufficient processed data to make an informed decision, which means that the asset team initiates well interventions too early and hence blocks the intervention resources for other wells that might require the intervention investment sooner. Currently, AIB has a contractual scheme in which several service companies participate by measuring well parameters in the field and creating engineering analysis of candidates to install the most appropriate artificiallift system. Because of the large number of wells in the asset, the associated capture of required measurements, and the distance to the location of interest, the existing analysis process was performed at an average rate of only two wells per week. The current strategy focuses on accelerating the quantity and improving the quality of the recommendations at the well level that help increase or at least maintain the production of the current asset. The asset team wants to take advantage of its investment in the automation of current processes related to operational data storage and engineering workflows, such as gasrate estimation and event detection and notifications. An investigation of current conditions also shows that while all decisions are made on the basis of field data, analysis relies heavily on an engineer’s experience and judgment. Engineers typically move from one job position to the next, and the arrival
For a limited time, the full-length paper is available free to SPE members at www.spe.org/jpt. JPT • JULY 2010
57
Gas Rate After (MMscf/D)
Gas Rate Before (MMscf/D)
Line Pressure Before (psi)
0.1 0.3 0.6 0.8 1.1 1.3 1.6 1.8 2.1 2.3
0.1 0.4 0.7 1.0 1.3 1.6 1.9 2.2 2.5 2.8
122 294 465 636 807 978 1,149
0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Wellhead Pressure After (psi)
Wellhead Pressure Before (psi)
Liquid Level Before (ft)
Last Well-Test Gas Rate (MMscf/D)
190 487 783 1,080 1,377 1,874 1,971
178 478 778 1,078 1,378 1,678 1,978
89
Type Numeric ( )
503 917 1,330 1,744 2,157 2,571 0.1 0.5 0.8 1.1 1.5 1.8 2.1 2.5 2.8 3.1
Last Well-Test Water Rate (STB/D)
Type Numeric Red: AL System A Green: AL System B Blue: AL System C 1
19 38
56 74 92 111 129 147
Fig. 1—SOM to investigate historical decision making regarding artificial-lift installations.
of new engineers makes knowledge transfer to those in charge of the field additionally difficult. Therefore, the following main questions need to be asked: • How to capture the knowledge of experienced personnel and make it available to the whole organization? • How to apply this knowledge in a general and unbiased manner throughout the assets of the organization? • How to keep this knowledge updated by use of new experience? The solution presented in the fulllength paper is based on an expert system and automated reasoning tools that are able to capture and apply knowledge and make it available to the whole organization. In general, the combination of these two technology-based approaches enables the asset team to process the field data and apply the knowledge systematically at the field level, obtaining results faster and on time for all wells in the field and independently of asset, well, or engineer. An automated workflow consisting of data capture, data validation and processing,
58
expert-knowledge application and reasoning analysis, and finally reporting is implemented to run once a day. Results are presented to the production engineer in the form of a list in which candidate wells are ranked by priority of intervention according to their critical condition. The engineer then is empowered to initiate the necessary engineering workflows (e.g., nodal analysis) more effectively and make the final decision on how to distribute resources. User-Driven Expert System The user-driven expert system is an automated workflow that fully relies on the knowledge and experience of the engineer. The asset team has the ability to incorporate its complex business reasoning into surveillance rules to detect potential artificial-lift candidates. The advantage of this system is the trustworthiness and, most significantly, the traceability of decisions leading to the artificial-lift candidates detected by the system, because the inference rules have been created in close cooperation with the asset engineers and fully reflect their provided
criteria. Additionally, the user-driven expert system can be revised easily and changes can be undertaken should the need arise (e.g., technical or economical limitations). During the implementation of the user-driven expert system, two different methods have been applied: (1) a rule-based approach that can be considered deterministic because the outcome is either “artificial lift=true” in case the input conditions are favorable or “artificial lift=false” in case they are not and (2) a stochastic approach implemented through a BBN in which the outcome is not just true or false but a probability distribution denoting the “degree of applicability of an artificial-lift system” that takes into account the uncertainty of a certain decision or choice. Rule-Based Approach. The rule-based approach consists of implementing the engineer’s reasoning into an automated routine comprising a sequence of ifthen-else loops. In this way, complex knowledge can be captured and used in real-time surveillance and detection workflows. These workflows can be
JPT • JULY 2010
scheduled to run at any time and as often as desired, enabling immediate detection when the conditions of a well indicate need of an artificial-lift system and enabling early notification to the asset team and engineers. Data-Driven Expert System. Datadriven approaches are used widely to reveal hidden information in large data sets. Typically, in a brownfield with a large number of wells and a production history spanning many years or decades, the amount of data available overwhelms the conventional approach to data screening (e.g., spreadsheets or production plots). A data-driven approach allows the engineer to visualize data in different contexts to facilitate the analysis of correlations and the process of knowledge discovery. The presented workflow applied an unsupervised clustering algorithm and a special visualization technique using Kohonen’s self-organizing maps (SOMs) SOMs. SOMs display multidimensional data sets in a 2D map. The SOM uses the Euclidian distance between two points in the multidimensional space as a measure of their similarity. The more similar the measurement, the closer the points are placed together in the 2D projection—the “map.� In this work, SOMs have been applied to identify a pattern in the historical decision making of the asset team. The objective is to discover the pattern under which a decision for a specific artificial-lift installation was typically made. Ideally, this pattern corresponds to the rule-based expert system as discussed earlier; however, sometimes it does not as a result of exceptional issues (e.g., no infrastructure available to support a certain installation). Fig. 1 presents the depiction of several realizations of the same SOM. Each realization corresponds to a parameter that has been used to calibrate the SOM (gas rate before and after artificial lift installation, line pressure before artificial-lift installation, wellhead pressures before and after artificial-lift installation, liquid level in the tubing before installation). The output can be seen in the top-right realization of the parameter (i.e., “type numeric�, representing the artificial-lift system that was chosen for the particular well).
It can be seen in Fig. 1 that there is a distinction between when to use Method A (red) and when not to (green or blue). Artificial Lift Method A typically is applied for low- to verylow-gas-rate wells with low pressures but with a moderately high staticliquid level in the well. However, it was considerably more difficult to clearly identify the conditions under which Method C was applied in the well to decrease the likelihood of liquid loading in the well. The operating conditions that lead to the choice of Method C are very much like those of Method B (higher gas rates, higher pressures). The data that were used to calibrate the SOM then were clustered to find groups of similarly behaving wells. The wells per cluster are selected by the clustering algorithm to minimize the variation of the individual items (wells) in each cluster. Using the clustering information, the problem space could be reduced from several hundred individual wells to five clusters of similarly behaving wells. This generalization helps the asset team (and the surveillance engineer) to classify a well immediately according to its production features (which are driving the cluster number) rather than having to investigate all available production information of a well before knowing how to classify it. Conclusions After analyzing the implemented workflow and the results obtained, the following conclusions were reached: 1. The main benefit from applying this methodology is the reduction of lost production by the ability to install an artificial-lift system sooner. 2. It is possible to capture the knowledge and field experience and implement them in an automated workflow, which can be used in a systematic way, acting as a fast field screening tool. The asset team then can focus its attention on the candidates identified, saving valuable time. 3. As knowledge is captured— including multiparameter criteria— and applied systematically, the methodology can be applied in an unbiased way to the whole field, allowing identification of new opportunities for artificial lift systems that could not have been identified simply by selecting one parameter as the rankJPT ing criterion.
SPE Bookstore
www.spe.org/store
NEW BOOK
Transient Well Testing Monograph Series, Vol. 23 Medhat M. Kamal Transient well testing is one of the most important diagnostic tools used by petroleum engineers to characterize hydrocarbon assets and predict future performance. This new monograph has the latest reference information on designing, running, and analyzing different types of transient tests in oil and gas reservoirs. Contents: #BTJD DPODFQUT PG USBOTJFOU UFTUJOH t 8FMM UFTUJOH NFBTVSFNFOUT t 7BMVF PG JOGPSNBUJPO t #BTJD JOUFSQSFUBUJPO PG IPNPHFOFPVT SFTFSWPJST t 8FMMCPSF FGGFDUT t 0VUFS CPVOEBSJFT t $PNQVUFS BTTJTUFE JOUFSQSFUBUJPO t 1SFTTVSF USBOTJFOU BOBMZTJT TPGUXBSF t /BUVSBMMZ GSBDUVSFE SFTFSWPJST t 8FMM UFTU BOBMZTJT PG IZESBVMJDBMMZ GSBDUVSFE XFMMT t %FMJWFSBCJMJUZ UFTUJOH BOE VOEFSHSPVOE HBT TUPSBHF t 4MBOUFE BOE QBSUJBMMZ QFOFUSBUJOH XFMMT t )PSJ[POUBM XFMMT t 5FTUJOH VOEFS NVMUJQIBTF n PX DPOEJUJPOT t *OUFSQSFUBUJPO PG TJNVMUBOFPVTMZ NFBTVSFE USBOTJFOU n PX SBUF BOE QSFTTVSF EBUB t 1FSNBOFOU HBVHFT BOE QSPEVDUJPO BOBMZTJT t .VMUJMBZFS SFTFSWPJST t *OKFDUJPO XFMM UFTUJOH t 'PSNBUJPO UFTUJOH t %SJMMTUFN UFTUJOH
Visit our online bookstore at
JPT • JULY 2010
www.spe.org/store