Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

GAS PRODUCTION TECHNOLOGY

OVERVIEW
For this feature, I have chosen three excellent papers that highlight the complex, but very common tendency of gas wells to load up. Given the explosive growth in new low-rate gas completions and the persistent move of all older high-rate completions toward the same condition, this topic seemed fit for a little more attention. Although not specifically addressed in these papers, the recent widespread use of very long (or thick), multihydraulically fractured completions intensifies these problems. If you think you do not have a well with a liquid-loading issue, you probably have not looked hard enough. Each of these papers addresses the issue from a different viewpoint. One uses video logging to see what is happening downhole, while another uses truly remarkable integrated transient wellbore and reservoir-simulation tools to predict mathematically the same behaviors observed in the first. The other paper stakes out a middle position of using comparatively simple analytical models to predict the exact location of a standing liquid level within a gas-producing section. The video-logging paper describes a series of production logs run in several horizontal wells. In the best spirit of data collection and research, the results were far from the expected outcome and revealed the predrill assumption of no potential loading problems to be incorrect. Also, the authors specifically put to rest that long-debated question of what happens to liquid and solids accumulations in horizontal wells with a toe-up trajectory. The paper on modeling the liquid-loading process builds on and complements N. Dousi et al. 2006 Numerical and Analytical Modeling of the Gas-Well Liquid-Loading Process [SPE Production & Operations 21(4): 475482 (SPE 95282-PA)]. In both papers, water loading up the wellbore is studied in detail. They both predict a metastable flow condition in wells that cannot lift all the liquid but have sufficient liquid injectivity (into the producing zones) to not die completely. This wellbore condition is now attributed to significant formation damage and well-performance deterioration after shut-in periods. The integrated-wellbore/reservoir-model paper takes the subject even further, with an exhaustive transient mathematical treatment of the load-up process. Coupling a reservoir simulator with a time-dependent wellbore model, the authors were able to prove that mathematical models can match an erratic production profile, although that is rarely the objective of the typical modeler. Whatever your mathematical inclination, one of these offerings should help you gain a greater appreciation of the complex process of liquid loadup in wellbores. JPT Scott J. Wilson, SPE, is a vice president at the Ryder Scott Company LP office in Denver. He specializes in well-performance prediction and optimization, reserves appraisals, simulation studies, software development, and teaching. Wilson has worked on oil and gas fields in all major producing provinces in his 25-year career as an engineer and consultant with Arco and Ryder Scott. He is Cochairperson of the SPE Reserves and Economics Technology Interest Group and serves on the JPT Editorial Committee. Wilson holds a BS degree in petroleum engineering from the Colorado School of Mines and an MBA degree from the University of Colorado. He holds two patents, and is a registered engineer in Alaska, Wyoming, and Colorado. Gas Production Technology additional reading available at the SPE eLibrary: www.spe.org SPE 107870 Analysis of Condensate Buildup and Flow Impairment of Retrograde Gases in Fissured Reservoirs by H. Ayala, Pennsylvania State University, et al. SPE 107169 Controls on Water Cresting in High-Productivity Horizontal Gas Wells by R.P. Sech, SPE, Imperial College London, et al. SPE 107274 Economic Viability of Gasto-Liquid Technology by Ali A. Garrouch, SPE, Kuwait University SPE 101497 Applications of Acoustic Liquid-Level Measurements in Gas Wells by J.N. McCoy, Echometer Company, et al.

88

JPT NOVEMBER 2007

GAS PRODUCTION TECHNOLOGY

Production and Video Logging in Low-Permeability Horizontal Gas Wells


Effective production-logging (PL) techniques are available for low-rate horizontal gas wells. Between June 2005 and September 2006, eight Sierra-field wells were evaluated with either traditional PL tools or downhole-videologging tools. Pressure, temperature, nuclear-fluid-density, single-capacitance, capacitance-array, fullbore-spinner, caliper, and gamma ray data were acquired with traditional PL tools. Video logging collected wellbore images as well as pressure and temperature information. The PL data indicated liquid accumulations; however, the different logging tools showed varying degrees of detail. Video-logging data provided definitive images of gas bubbles and slugs flowing through liquid.

Columbia. Approximately 700 producing wells have been drilled into the carbonate Devonian Jean Marie formation. The horizontal wells are drilled underbalanced to limit formation damage to the underpressured reservoir. Gas is produced from openhole completions. While drilling horizontally, well trajectory is steered on the basis of gas rate while drilling. The inclination is varied up and down until a good permeability streak is indicated by a spike in gas flow rate on the gas-ratewhile-drilling chart. After total depth is reached, tubing is installed and the well is brought on production. Typical gas-production rates decline from an initial rate of 1125 m3/d to 281 m3/d after 12 months and 221 m3/d after 36 months. Equipment and Processes Two types of equipment were used. The first was a typical full suite of PL tools run in memory mode, and the second was a downhole-video tool run in realtime-data mode. The following options for conveyance and operation were considered for the PL program. Wireline-with-tractor conveyance and real-time data collection. Coiled-tubing (CT) conveyance with electric line inside and real-time data collection. Continuous-sucker-rod conveyance with memory-mode data collection. Factors that influenced decisions on equipment selection and conveyance included the following. The cost to conduct the program the largest factor The need to pull the tubing to run a wireline-tractor assembly The helical-buckling and lockup limitations of smaller CT Ultimately, the cost of moving in a snubbing unit to pull the tubing and

Introduction In the past, operations in the Sierra field consisted of drilling, completing, and commissioning a large number of wells during the winter when the muskeg is frozen and land access is available. The initial focus was to bring on new production with very little engineering or research applied to mature wells. Incorrect assumptions resulted from this lack of engineering data. One such assumption was that the reservoir contained undersaturated gas and that, therefore, liquid-loading problems should be minimal. The greater Sierra area is approximately 90 km east of Ft. Nelson, British
This article, written by Technology Editor Dennis Denney, contains highlights of paper SPE 108084, Production and Video Logging in Horizontal LowPermeability Gas Wells, by D. Sask, SPE, Consultant; C.D. Hundt, SPE, and J.M. Slade, SPE, EnCana Corp.; and P. Daly, Consultant, prepared for the 2007 SPE Rocky Mountain Oil & Gas Technology Symposium, Denver, 1618 April.

run a wireline with a tractor-conveyed PL string outweighed the advantages of real-time data. The expected depth capability of a continuous-sucker-rod string was greater than that of CT. Therefore, with the reduced cost of memory compared to real-time logging, the decision was made to use the continous-sucker-rod string. The PL tools included pressure/temperature, full-bore flowmeter, radioactive density, single-capacitance liquid identification, circular-array liquid identification, and caliper. Three major types of video-logging tools were available. Full-video mode enables cameras run on optical wireline or coaxial cable to capture video at 30 frames/second (f/s). Standard mode enables cameras run on traditional mono conductor cable to capture images at approximately 1 f/s. Cameras run in memory mode capture up to 200 frames in total. There are three options for conveying the downhole camera. Standard wireline deployment is used in vertical wells. In horizontal wells with tubing sizes of 73.0 mm or larger, the camera can be deployed on wireline with a downhole tractor. Smaller tubing must be pulled from the well to run the camera and tractor. CT deployment in horizontal wells enables running the camera through existing 60.3-mm-diameter or larger tubing. However, there can be significant limitations on the horizontal distance that CT can traverse, depending on diameters and well trajectory. Because these wells are horizontal, standard wireline could not be used. The full-video-mode camera was not used because fiber-optic cable was not readily available in a CT configuration, nor was an option available for the well-tractor system.

For a limited time, the full-length paper is available free to SPE members at www.spe.org/jpt. The paper has not been peer reviewed. JPT NOVEMBER 2007 89

Case Study: Well 1 Fig. 1 shows the wellbore trajectory and gas-flow rate while drilling for Well 1. Rate-while-drilling data were used to determine where gas inflow occurred in the wellbore while drilling. The wellbore trajectory is a profile of the wellbore; it shows changes in true vertical depth (TVD) vs. measured depth (MD) but does not illustrate changes in azimuth. The wellbore-trajectory chart is a powerful aid for interpreting production-logging and video-logging data. Interpretation of the static-pressure chart in Fig. 2 differentiates liquid-filled wellbore regions from gas-filled regions where changes in elevation occur along the wellbore. The pressure gradient for the interval from 1826.5 m MD (1506 m TVD) to 1923.5 m MD (1507 m TVD) was 9.5 kPa/m indicating a water-filled wellbore. The bold sections on the wellbore-trajectory line indicate where water was identified with the video camera. The flowing pressure shown in Fig. 3 provides insight on flowing-fluid type. In this well, the calculated pressure gradient through the liquid interval was 12.5 kPa/m TVD, which indicates water. It is worth noting that the pressure tool did not indicate the liquid interval at 2079 m MD because there was no meaningful elevation change across this interval. Video images from Well 1 also showed that the liquid was offset toward the toe of the section during shut-in periods. In contrast, video logging in Wells 2, 3, 5, and 6 shows liquid being pushed to the heel side of the sumps during shut-in periods. This phenomenon depends on gas flow into the respective regions of the wellbore and flow between the regions. The fullbore spinner displayed distinct liquid and gas phases under static conditions. When the well was opened to flow and logged again, the spinner readings became erratic in the liquid interval. The erratic behavior was interpreted as gas flowing through the liquid phase. There were no indications of gas or liquid inflow from the reservoir even though the well was flowing at 402 m3/d while logging. The temperature log indicated phase type by a change in temperature. Entering a liquid phase would cause a cooling effect on the tool. The cooling and warming effects observed in Well 1 did correlate with phase intervals observed by other tools, but the

Fig. 1Well 1: Wellbore trajectory and gas-flow rate while drilling.

Fig. 2Well 1: Pressure log at shut-in conditons.

Fig. 3Well 1: Pressure log at flowing conditons.

90

JPT NOVEMBER 2007

Fig. 4Natural fracture in Well 1.

changes in temperature were so gradual that the intervals could not have been picked from the temperature log alone. The capacitance-array tool (CAT) displayed various flow regimes through the use of software that provides videolike images of different colors moving through a wellbore. The 100%-gasfilled case was the most difficult to interpret because it closely resembled stratified flow. The only way to know that the wellbore was gas-filled was by correlating with the other PL tools. Video logging of Well 1 was completed approximately 31/2 months after the PL, and the camera was conveyed on CT. The tubing had a significant buildup of waxy solids, and the camera lens was coated with opaque material within 2 m of the top of the wellhead. A clear image was regained only after the camera was immersed in water downhole. The video camera was very effective in locating the sections of the wellbore that contained water. However, poor water clarity made it impossible to distinguish any gas within the water. Partially liquid-filled sections of the wellbore were evident and wellbore features including a natural fracture (as shown in Fig. 4), scaling, and a buildup of solid materials. Fig. 5 compares liquid levels observed by each tool under flowing and shut-in conditions. Under shut-in conditions, all interrogation methods identified a liquid interval in the 1820- to 1940-m-MD sump, but the exact location of the gas/liquid interfaces could not be identified consistently. From this compilation, the difficulty in interpreting production logs can be seen. Exact fluid locations were not consistent between the different production logs. The only method that could be relied upon exclu-

Fig. 5Summary of logging tools showing liquid regions in Well 1.

sively was video logging. The images showed definitive gas/liquid interfaces, so long as debris did not cover the lens. Because interpreting production logs was extremely time-consuming and the results were vague, it was decided to discontinue conventional PL. The videologging program was continued with the other seven wells. The full-length paper details testing of those seven wells. Conclusions 1. Video logging provided a significantly more conclusive evaluation of wellbore conditions than conventional PL in the wells logged in this program. 2. There was general agreement between the pressure, temperature, spinner, and density tools regarding the liquid boundaries of the largest water-containing section. However, the boundaries were not as well defined with the CAT. The PL evaluation was time-consuming, difficult, and open to conflicting interpretation. 3. Gas velocities in the horizontal openhole section were too low to energize the full-bore spinner and identify areas of gas inflow. The spinner reacted only to the frictional forces as the tools moved through water-filled sections.

4. Liquid inflow was not identified with either logging method. 5. In partially or completely liquidfilled wellbore regions, gas bubbles entering the wellbore and gas pockets moving over the top of the water-filled region were observed with video-logging images. These regions were not easy to distinguish or interpret with PL tools. 6. The video camera does not measure gas rates directly, but it is possible to recognize gas flowing past the camera as a result of a hazing or fog-like quality to the images. 7. Video logging is susceptible to wellbore debris covering the lens and hindering the quality of images. 8. The choice of conveyance for video-logging tools depends on several factors with no universal solution. Wireline-tractor deployment was best where extended reach was critical. CT deployment offered more flexibility to move quickly and to clean the camera lens during the logging program. 9. The use of wireline tractors can be enhanced with multiconductor wireline that allows tractoring and video capture concurrently. Single-conductor wireline does not allow video logging JPT while tractoring into the well.

JPT NOVEMBER 2007

91

GAS PRODUCTION TECHNOLOGY

Improved Model for the Liquid-Loading Process in Gas Wells

An improved model for describing the liquid-loading process in gas wells has been introduced. The new model improves how the Dousi model handles the downhole inflow and outflow between the well and the reservoir and makes more-realistic and -detailed assumptions with respect to the production and injection interval. Also, the new model can be used to predict future behavior of the well in a more realistic fashion. The difference between the new model and the Dousi model is that flow-rate changes occur slower when liquid loading begins, reflecting realistic inflow-performance assumptions.

Top of reservoi r

Production

Top of reservoi r

Production

Water

Water

Injection

Injection

Introduction Liquid loading is a serious problem in maturing gas fields. The liquid-loading process occurs when the gas velocity within the well drops below a certain critical gas velocity. The gas then is unable to lift the water that is coproduced with the gas (either condensed or formation water) to surface. The water will fall back and accumulate downhole. A column is formed that imposes a backpressure on the reservoir and reduces gas production. The process eventually results in intermittent gas production, and the well dies. Methods of reducing liquid loading include production-string resizing in which a smaller tubing size is chosen to increase the gas velocity above the critical Turner rate, compressor installation
This article, written by Technology Editor Dennis Denney, contains highlights of paper SPE 106699, An Improved Model for the Liquid-Loading Process in Gas Wells, by Frank van Gool, SPE, and Peter K. Currie, SPE, Delft University of Technology, prepared for the 2007 SPE Production and Operations Symposium, Oklahoma City, Oklahoma, 31 March3 April.

Dousi Model (2005)

Improved Model (2007)

Fig. 1Bottomhole production-/injection-interval models.

to lower the tubinghead pressure thereby increasing the gas velocity above the critical Turner rate, plunger lift to lift all the liquids by use of the gas pressure during shutdown of the well, pump installation to pump the liquids during production, foaming the liquids so that it is easier for the gas to lift all the fluids thus reducing the critical Turner rate, and gas lifting with gas from other wells to decrease the pressure loss in the tubing and increase the velocity. The best solution for a given well depends on the properties of that particular well. Metastable Flow During the liquid loading of a mature gas well, the gas velocity in the tubing drops when the reservoir pressure drops, and, in time, the gas velocity may fall below the critical Turner rate. At that time, the produced liquids fall back down the wellbore and accumulate at the bottom of the well. This liquid column forms a backpressure on the reservoir, thus slowing the production rate and accelerating the problem.

However, in parts of the well, the pressure can be lower than in the reservoir such that production still occurs even though at the bottom of the well the pressure from the liquid column is high enough that liquid injection can take place. If the liquid-injection rate equals the rate of liquid coproduction with the gas, metastable production occurs. This metastable flow rate is observed in the field. Data analyzed from gas wells in The Netherlands, including production profiles, temperature, and pressure, showed several examples indicating the existence of a metastable flow rate. Dousi Model This model describes the liquid loading of a gas well and the pressures and flow rates that occur during this process. This model uses the following assumptions. The bottom part of the wellbore is assumed to be vertical. The tubinghead pressure is equal to the surface export pressure when the well is flowing.

For a limited time, the full-length paper is available free to SPE members at www.spe.org/jpt. The paper has not been peer reviewed. 92 JPT NOVEMBER 2007

The Cullender-Smith well-flow model describes the flow from the bottom of the wellbore to the surface. A single gas-production point is used at the top of the reservoir, a simple Darcy model for single-phase gas flow is used to model the production rate, and the coproduced water is assumed to be at a constant water/gas ratio. A single water-injection point is used at the bottom of the well, and a simple Darcy model for single-phase liquids is used to model the injection rate. If the gas flow rate is higher than the critical Turner rate, then all the coproduced liquids are produced to the surface. If the flow rate is lower than the critical flow rate, all the liquids fall back to the bottom of the well to form a water column. The numerical scheme used to calculate the predictions of this model takes discrete timesteps and solves the model equations in an iterative manner. Input values describe the details of the particular well (e.g., water cut and productivity). The tubinghead pressure is specified as a function of time. Improved Model The Dousi model makes simple assumptions about the behavior in the production/injection interval. The improved model spreads this interval as shown in Fig. 1. The original model had one production point at the top of the reservoir and one injection point at the bottom of the well. In the improved model, it is possible either to produce gas/water or to inject water at any point of the interval, depending on the local pressure. The pressure for a sample loaded well is shown in Fig. 2. The numerical scheme breaks the interval down into separate elements, each with its own pressure calculation. The flowing pressure is calculated for each element. If the pressure in the reservoir is higher than the well pressure in the element, gas is produced together with water. The amount of water is a fixed fraction of the produced gas (i.e., constant water/gas ratio). If the pressure in the reservoir is lower than the pressure in the calculated element, which is possible only if there is water accumulated in the wellbore, water is injected into the reservoir at that particular point. The equations used in the improved model are the same as those used in the Dousi model except for production and injection. Those equations were rewrit-

Pressure/Depth Graph
Pressure, bara 50 3200 Well Depth, m 3400 3600 3800 4000 60 70 80

Unloaded

Perforated interval
4200 Flowing gas gradient unloaded Flowing gas gradient loaded Pore pressure

Loaded

Fig. 2Pressure distribution in the reservoir interval.

ten for a unit area of wellbore and are detailed in the full-length paper. Model Predictions To evaluate the two models, the original Dousi data set was used. It includes parameters of the well and the reservoir. The gas in this data set is dry gas,

and the liquid is coproduced water. Initially, the pressure is low enough to ensure that the flow rate is higher than the critical Turner flow rate. After 1 day, the flowing tubinghead pressure is increased and the gas flow rate falls below the critical Turner rate. Then, the coproduced water falls back down the

4850 East Fulton Street, Ada, Michigan 49301-9108

JPT NOVEMBER 2007

93

wellbore, building up a liquid column. This column forces a backpressure on the reservoir, and when the pressure at the well bottom is higher than the reservoir pressure, water is injected into the reservoir. In time, a new stabilized metastable gas flow rate is reached when the water injection becomes equal to the water production. Then, between Days 3 and 3.5, the well is closed in, stopping the gas inflow and raising the tubinghead pressure. The higher pressure in the wellbore compared with the reservoir pressure causes water to be injected into the reservoir, and the wellbore is cleared of water. The differences between the two models occur when the gas flow rate falls below the critical rate. In the Dousi model, the flow rate initially is only slightly decreased while the water column is rising. Only after the water column rises above 170 m, the position of the production point, is the effect of backpressure noticed and gas production reduces to a new metastable flowrate at approximately Day 3. Meanwhile, because of the rising water column, the pressure at the bottom of the wellbore

is high enough to start injection of the liquids. At approximately Day 3, the injection of the liquids is equal to the production of liquids, which stabilizes the height of the water column. Because all of the processes are more or less stable at this time, this situation is called metastable. With the improved model, there is an immediate effect of the liquid column. Because production occurs from the entire well interval, production is decreased immediately when the first liquid drop falls back and accumulates in the wellbore. This reduced gas flow rate causes a lower liquid production, thus a slower build-up of the water column. Ultimately, the gas flow rate is much lower in the improved model than in the Dousi model. In the improved model, the pressure regime over the entire production interval is used in the calculation, compared with the single point in the Dousi model. Between Days 2 and 3, the process reaches its metastable state. After Day 3, the well is shut in and the pressure in the reservoir is equal to the pressure in the wellbore just above the water column.

Because the reservoir is assumed to be gas-bearing only, the water now flows into the reservoir because the pressure at the bottom of the water column is higher than the pressure in the reservoir. During this 3-day period, the Dousi model nearly completely removes the liquid column, while the improved model requires much longer. This longer period is because the total water injectivity is averaged over the whole interval. In the original model, there is a single injection point, at the bottom of the interval, at which the pressure difference is highest. In the improved model, this pressure difference is calculated for every point in the interval, which gives a much lower average difference resulting in a lower injection rate. In applying either model, the waterreinjection resistance factor should be chosen carefully because it will have a large effect on the liquid-injection rate. Conclusions This extension of the Dousi model makes more-realistic and -detailed assumptions with respect to the production and injection interval by improving modeling of the downhole inflow and outflow between the well and the reservoir. From the simulation performed with the two models, it was seen that with comparable well parameters, the new model predicts faster decline in production rate during liquid loading; longer time to reach metastable flow rate; and slower water-reinjection rate during cleanup, and thus a longer time to clean up. The implication is that the model and the parameters chosen for the reservoir section of the wellbore are very important and have a significant effect on the predictions of the liquid-loading behavior. With inappropriate assumptions, the well may take a longer time to clean up than expected, resulting in delays in planning, thereby increasing costs. If the operator plans to flow the well at the metastable flow rate, the flow rate could be lower than expected. The interaction between the reservoir and the well is, therefore, crucial in modeling liquid-loading behavior. The proposed model for the reservoir interval still is only an approximation. The predicted liquid-loading behavior may be affected significantly if further improvements are introduced such as incorporating a full model of the nearJPT wellbore region.

Benet from the knowledge


gained through our highly successful North American joint industry project focused on tight gas sands. Core Laboratories is now offering the industrys most comprehensive evaluation of tight gas sands for projects worldwide.

Reduce Risk and Increase Success


Access to the Largest Tight

Gas Sands Database in the Industry Detailed Core-Based Reservoir Characterization Rock-log Calibration and Petrophysical Modeling 3-D Fracture Design and Completions Fracture Production Forecasting Real Time Monitoring Post-Frac Evaluation INTEGRATED RESERVOIR SOLUTIONS DIVISION
UNITED STATES Tel: +1-713-328-2673 UNITED KINGDOM Tel: +44-173-785-2390

www.corelab.com/IRS/studies/tgs email: tgs@corelab.com

94

JPT NOVEMBER 2007

GAS PRODUCTION TECHNOLOGY

Integrated Wellbore/Reservoir Model Predicts Flow Transients in Liquid-Loaded Gas Wells


An integrated wellbore/reservoir model was used to investigate liquid loading in a gas well. The well produces from a storage reservoir, and it experiences water coning from an aquifer. The integrated model showed how the water cone caused the gas-flow rate from each gas layer to decrease and the liquid holdup in the wellbore to increase. Depending on reservoir conditions, the well may enter into a mode of unsteady production, during which the gas-flow rate cycles over a period of several days. The simulation revealed the reason for this unsteady flow.

PRODUCTION WELL CONTROL WELL

CONTROL WELL

INJECTION

PRODUCTION
IM P

ERM EAB L

GAS+WATER

EC O VE R

Introduction Liquid loading in gas wells is a challenge in mature fields. Several techniques have been developed to address the issue. There is a growing demand for better simulation tools to optimize the operations. A dynamic wellbore/reservoir integrated simulation is required when studying the transient liquid-loading processes in gas wells. A reservoir model is needed to simulate the changes in well productivity and phase mobility as fluid saturations gradually change in the near-wellbore region. A transient wellbore multiphase-flow model is required to predict the onset of loading and the flow transients resulting from liquid accumulation.
This article, written by Technology Editor Dennis Denney, contains highlights of paper SPE 110461, Integrated Wellbore/Reservoir Model Predicts Flow Transients in Liquid-Loaded Gas Wells, by Gal Chupin, SPE, Bin Hu, SPE, and Tor Haugset, SPT Group; Jan Sagen, IFE; and Magali Claudel, SPE, Gaz de France, prepared for the 2007 SPE Annual Technical Conference and Exhibition, Anaheim, California, 1114 November.

WATER

Fig. 1Aquifer storage for natural gas.

Further, the mechanisms of the many liquid-loading-mitigation techniques (e.g., plunger lift, compression, pumping, and gas lift) are dynamic processes in which the variables change and interact constantly. The complete system, from the reservoir to the receiving facility, must be considered rather than studying individual components separately. A coupled model of the wellbore and near-wellbore reservoir is required to simulate the dynamics accurately, analyze the phenomena, and numerically test the remedial actions or control schemes. A dynamic well/reservoir model has been developed to simulate oil production from thin oil rims subject to gas and water coning. The model was used to develop gas- and water-coningcontrol schemes. However, the transientwellbore-flow model was too simple to be used widely on complicated well flows. A reservoir simulator was linked with a commercial transient-multiphaseflow simulator to simulate the formation heading and liquid loading in a gas

well. Both the reservoir and the wellbore model are comprehensive and have strong functionalities. Unfortunately, the link itself is explicit, which limits the simulation speed and numerical stability. An integrated dynamic wellbore/reservoir model is presented. Wellbore/Reservoir Model Near-Wellbore Reservoir Model. The model can simulate transient threephase (gas/oil/water) flow in a porous medium. The flow equations can be solved in one, two, or three dimensions, yielding saturations and pressures varying in space and time. No energy equation is solved; instead, the reservoir is assumed to be isothermal. The equations are solved by use of the Newton-Raphson iterative method at each timestep. Input data to the model are permeabilities and porosities of the porous medium, fluid properties of the flowing phases, and, if needed, thermal properties of the rock and fluids. Boundary

For a limited time, the full-length paper is available free to SPE members at www.spe.org/jpt. The paper has not been peer reviewed. JPT NOVEMBER 2007 95

conditions at the well and at the outer reservoir must be given. When coupled to the wellbore-flow model, the wellbore-flow model will determine the boundary condition at the well. Typical time-dependent boundary parameters are injection and production flow rates, pressure, saturation, and temperature. This model supports radial (cylindrical) and rectangular grid types. Wellbore-Flow Model. The wellbore-flow model can simulate transient three-phase (gas/oil/water) flow in pipes. At each timestep, a set of five coupled mass-conservation equations is solved for the gas continuous phase, the water droplets, the oil droplets, the oil film, and the water film. One momentum equation is solved for the gas/droplet field and another for the liquid bulk. The model is closed by an appropriate set of closure laws describing the friction laws at the wall and the interface, droplet and bubble entrainment, and droplet deposition. One energy-balance equation is solved for the fluid mixture. Wellbore-/Reservoir-Model Coupling. The near-wellbore reservoir model is considered a plug-in to the wellboreflow model. The basic principle is that the wellbore model provides the pressure boundary for the reservoir model while the reservoir model provides the flow into the wellbore together with the fluid temperature. The coupling can cope with flow in both directions for injection wells or certain flow transients in a production well. The coupling between the reservoir model and the wellbore-flow model is implicit. With implicit coupling, the reservoir model calculates a sensitivity coefficient for the production rate with respect to wellbore pressure, which then is entered into the wellbore-flow model. At the next timestep, the wellbore model uses this sensitivity coefficient to solve the new wellbore pressure. The sensitivity coefficient is extracted from the Jacobian matrix of the nearwellbore reservoir model at the last iteration. A coupling level determines the size of the near-wellbore domain contributing to the rate/pressure-sensitivity calculation. Currently, the coupling level is a user input. Workflow. The implicit coupling between the near-wellbore reservoir

model and the wellbore-flow simulator enables the model to carry out integrated transient-flow simulations. The coupling method provides the required speed and numerical stability for such integrated simulations. Because the integrated model is used mainly to account for the fast-flow pressure transient in the wellbore and near-wellbore reservoir area, it is not necessary to include the full-scale reservoir in the near-wellbore reservoir model. The far deep reservoir has little effect on fast-flow pressure transients in the wellbore and near-wellbore region. In addition, modeling a large area of the reservoir increases the simulation burden and reduces the grid resolution in the near-wellbore region that is of interest. In most situations, it is sufficient to include only a region (approximately 10 to 100 m) around the wellbore in the model to perform simulations. However, most of the input information to the near-wellbore reservoir model is maintained in a full-scale reservoir simulator. It then is of interest to determine a pragmatic way to transfer data between the full-scale reservoir model and the near-wellbore reservoir model. The workflow for building the near-wellbore reservoir model must contain a stage in which timestep data from commercial simulators are transferred to the near-wellbore simulator before performing the integrated simulation. The workflow can be described as follows. 1. Identify a region of the full-scale reservoir around the wellbore that is relevant for the coupled simulation. Typically, the region to consider is that for which information from the wellbore will not travel to the outer boundary of the region in less than the simulation time. Trial and error may be required to delineate the proper region. 2. Select the timestep of interest from the reservoir simulation as the initial condition to the transient simulation. 3. Transfer data from the full-scale reservoir simulator to the near-wellbore simulator. Data include grid information, pressures in each phase, phase saturations, relative permeability curves, porosity, and permeability. 4. Regrid the near-wellbore region if the grid in the full-scale reservoir simulation is too coarse. 5. Connect the near-wellbore reservoir model to the wellbore-flow model.

Case Study 1: Gas-Storage Reservoir For storage purposes, natural gas often is injected into underground reservoirs that have pressure support from an aquifer. Fig. 1 shows a typical aquifer storage system. In periods of high gas demand, the aquifer storage is produced with vertical wells drilled into the gas cap. Separators may have limited capacity and efficiency such that the amount of water produced at surface must be maintained below a specified quantity. During the production part of one cycle, the produced-water/-gas ratio (WGR) increases sharply after approximately 60 days of steady production. Aquifer storage is characterized by high absolute permeabilities with fast pressure-depletion rates. Therefore, a possible cause of the WGR increase was water coning into the perforated interval. Another possible explanation was the overall rise of the water-table level as gas was produced from the storage. To investigate the coning phenomenon at the reservoir level, a reservoir model of the complete aquifer storage was built with a reservoir simulator. The near-wellbore region close to the producer was gridded finely to capture possible formation of a water cone. Two production wells were included: a typical production well (Well 1) and a generic well (Well 2). Well 2 represents production from all the other wells. When running the reservoir model, constant gas-flow rates from the production well and the generic well were assumed and were achieved by regulating the bottomhole pressure such that the drawdown across the perforation was large enough to achieve the target gas flow. If the required bottomhole pressure fell below a floor value, the simulation was stopped. After 100 days, water had invaded the reservoir blocks immediately below the bottom perforation. After simulating 500 days of production, strong water coning toward the production wells was indicated by the simulation. In order of magnitude, the WGR scaled correctly with production data. However, the WGR increase in the field data was much sharper than in the simulations. This disagreement suggested that something was overlooked in the simulations and that more resolution was required. To improve the results, the reservoir simulation was linked with a wellbore

96

JPT NOVEMBER 2007

hydrodynamic simulation. This linking enabled investigating the following issues. Ability of the well to lift the water produced from the reservoir Tendency toward liquid loading Flow stability Effect of well hydraulics on reservoir delivery Validity of the constant-gas-flowrate assumption in the standalone reservoir simulation Results. For the first 400 days of production, a good match was achieved between the amount of water produced into the wellbore (full-scale reservoir simulation) and the amount of water produced at the wellhead (integrated simulation). However, after 450 days, the flow rate from the integrated simulation fell below that of the full-scale reservoir simulation. The reason was that after 400 days, the well could no longer achieve the gas-production flow rate assumed in the reservoir simulation. From 450 to 500 days, because of reservoir depletion and increased pres-

sure drop, the well operated at fullopen choke and incurred backpressure from the wellhead pressure boundary. The operating point of the well was displaced toward lower gas-flow rates. As the gas-flow rate decreased, the amount of water produced by the reservoir also decreased. The total liquid volume exhibited a steady increase with time. However, the well flow did not become unstable, and the increasing liquid holdup was caused primarily by the increasing water production from the reservoir. Case 2: Two-Layer Reservoir In certain cases, the gas flows in a quasisteady state at a velocity less than the Turner velocity. In these cases, the flow is referred to as metastable. As detailed in the full-length paper, the well could reach a stable operating point. However, compared with the reservoir-model approach, the production and injection flow were constant. In reality, the inflow close to the near-wellbore region was not constant because saturations and pressures were changing with time in the near-wellbore reservoir medium.

This phenomenon can be simulated properly only with an integrated tubing/reservoir model. Conclusions 1. Dynamic integrated wellbore/ reservoir simulations were achieved through an implicit coupling between the wellbore and reservoir models. 2. Integrated simulations add detail to standalone reservoir simulations. In liquid-loaded gas wells, the onset of loading was predicted, as was how the well loading evolved with time and whether the well eventually died. The approach provided a more realistic boundary condition on the reservoir such that production was determined more accurately. 3. The integrated model revealed transient phenomena associated with liquid loading. Metastable flow could be reproduced in accordance with field experience. However, although the operational point was stable, the well entered a mode of unsteady production during which the produced-gas and -liquid flow rates cycled over a period JPT of several days.

What would industry be without fluids ?


Bringing quality products to customers depends on the ability to keep missioncritical fluids flowing through industrial systems. No matter which products you work with, whether they are abrasive, fragile, viscous, corrosive, hot or heavy, they require special handling. And because fluids are your companys lifeblood, PCM has been at your side for over 70 years, developing pumps, pump systems and advanced fluid handling technology. www.pcm.eu

JPT NOVEMBER 2007

97

September 2007

You might also like