You are on page 1of 24
Representing Steam Processes With Vacuum Models G.L, Stegemeier, SPE, Shell Development Co. D.D. Laumbach, SPE, shell Oi1Co. C.W. Volek, SPE, Shell Oi1Co, Abstract Scaled models of steam processes have contributed significantly to the design and implementation of ‘many field projects. These models provide a means of answering pertinent questions, including the effect of (1) injection rate, (2) production pressure, (3) completion interval, (4) pattern size and type, (5) aquifers, (6) heterogeneities, and (7) steam quality Parameters are presented for scaling up physical- model results to full scale and for relating one oil field to another. These relationships are generated by casting the governing equations in dimensionless form. A set of similarity parameters then are determined by inspectional analysis. In physical models, unfortunately, it is not possible to match all similarity parameters. Consequently, based on. engineering judgment, a set containing a reduced number of parameters, called scaling parameters, is generated that generally can be matched between scaled model and field prototype. Techniques 10 implement this scaling are discussed, including a description of the laboratory models, ‘typical materials, and procedures for conducting the experiments. Results of model studies for Mt. Poso and Midway Sunset prototypes are presented, Introduction Physical modeling technology has been developed to the extent that detailed descriptions of steam. processes can be provided for field projects in which the number of wells is large, patterns are irregular, or asymmetry occurs from dip or water influx. In many ‘of these cases, sufficient complexity can be in- troduced to provide both prediction of overall response and specific guidance for operating policies ‘on a well-by-well basis. The fine detail attainable in JUNE 1980 physical models arises from the large number of beads or sand grains, typically in excess of 10 million, that are used in the packed bed. By comparison, present-day numerical steam simulators are limited practically 10 about 1,000 grid blocks. Besides of- fering this capability of representing additional geometrical and geological complexity, the physical models have the advantage that physical phenomena are not constrained by specified relationships but are free to interact subject only to scaling factors. This, additional insight can be important in new processes for which relationships are not known or are difficult to formulate. Limitations of physical models arise because of the unavailability of materials and fluids having physical properties that will satisfy all scaling requirements. Effects of compromises in scaling often can be observed with simple geometric configurations in mathematical simulators. Conversely, improved mathematical simulation often is possible after determining important parameters experimentally Consequently, the two serve complementary roles in determining the important mechanisms for a par- ticular process. Our thermal models do not represent processes in which steam distillation, solution as, chemical reactions, or compressibility are important. The choice of whether to model physically or to calculate numerically depends on the actual process. being studied and the capabilities one has developed in each of these technologies. Scaling rules for steam-injection processes have evolved from those for isothermal and hot-water processes. Isothermal reservoir processes have been the subject of a number of scaling studies.'"10 Scaling for the hot-water drive has been reported in the work of Geertsma er al.,° Baker,? and Dietz; scaling for combustion processes has been given by Binder er al.!! Development of scaling groups by 1st dimensional inspectional analysis has been treated by Ruark!? Richardson, !? and Loomis and Crowell.!* Work dealing specifically with scaling of steam- injection processes and development of experimental techniques has been carried out by a number of Shell researchers during the past 20 years. Results of some of this early work have been reported by Niko and Troost!® and Harmsen. '* Other contributors to this, technology are lisied in the Acknowledgments. The earliest steam-model experiments were per- formed in rigid containers or in high-pressure vessels in which an overburden pressure could be applied external to the flow chamber. In these containers, temperatures equivalent 10, or approaching, field conditions could be achieved. P. VanMeurs showed that if the ratio of latent to sensible heat was properly scaled, comparable model results could be obtained at low temperatures with subatmospheric-pressure steam, F. van Daalen and H.R. van Domselaar then made a significant breakthrough in experimental technology by constructing packed-bed models confined by thin plastic sheets, which, upon im- position of a vacuum became rigid’ containers suitable for flow experiments with irregular geometric shapes. At VanMeurs’ suggestion, this vacuum technique was applied to scaled, low- temperature (about 200°F) steam experiments. In this manner, the technology of model fabrication with temperature-resistant materials and. the ex- perimental procedures that are used today were evolved. Recent developments have dealt with im- provements in experimental equipment, techniques, and scaling rules that extend the range of prototype conditions that can be modeled successfully. Model results of steam processes have been compared with field results under a variety of reservoir conditions and operational procedures. The comparisons provide insight into individual field behavior and permit generalizations that are useful in screening prospective thermal projects.'” The field projects that have been studied may be grouped as (1) relatively thin, dipping reservoirs, (2) horizontal reservoirs with a thin bottom-water layer, and (3) thick, gravity-drainage-type reservoirs, Steam drives in dipping reservoirs have been studied in model experiments for Tatums (OK), Coalinga (CA), and Mt. Poso (CA) fields. In each case, the effect of gravity along dip is significant. Some effects that were studied are (I) position of injection points along dip, (2) hot-water preflood, (3) steam-zone pressure, (4) permeability, (5) "oil saturations, and (6) aquifer influx. Model results have been consistent with field behavior in predicting much better performance from Mt, Poso than the ‘ovo earlier steam drives. Steam drives in horizontal reservoirs, containing viscous oil overlying thin water layer, have been studied for the Peace River (Alta.) and Slocum (TX) fields, In these reservoirs with low oil mobility, the water layer provides the injectivity essential to a steam drive. Model experiments were conducted with different injection procedures, well patterns, well spacing, well completions, and vertical distribution of permeability. 132 Steam soaks and steam drives also have been studied for thick reservoirs producing by gravity drainage. A thick, horizontal reservoir with a gas zone al the (op was modeled for the Yorba Linda (CA) field. Midway Sunset field has a thick, steeply dipping productive interval, with a gas zone updip. Soak volumes, completion intervals, and the effects of noncontinuous shales were investigated, In both fields, interwell communication enhanced oil recovery after four or five soak cycles. Conversion to steam drive thereafter accelerated oil production without greatly affecting the oil/steam ratio. Scaling Parameters The first step in modeling steam processes is the development of scaling parameters. The scaling parameters presented in this paper are obtained by the following procedure, 1, The governing equations of fluid flow and heat transfer are put in dimensionless form, 2. Similarity parameters are determined by in- spectional analysis 3. These similarity parameters are combined, or modified based on engineering judgment, to obtain a set of scaling parameters that generally can be matched between scaled model and field prototype The governing equations for steam processes are developed in Appendix A. Similarity parameters, derived by inspectional analysis of these equations, are given in Table 1. This table lists the equations in Appendix A from which the parameter originates the dimensionless group name commonly applied to it, and an assigned parameter number. The characteristic quantities, which are selected in an effort (0 make the dimensionless equations invariant between model and prototype, are given in Table 2. In most cases, the basis for selection is one that will give the best property fit over the range of interest. In some cases, such as that of the reference saturation, there is but one choice for the characteristic quantity that will make the scaling invariant. The underlying principle of inspectional analysis is this: If the dimensionless equations and the associated constraints that describe the behavior in the model and prototype are identical, the dimen- sionless solutions also will be the same. In this manner, model results can be scaled directly to the field. The condition that all dimensionless equations be the same requires that the properties, as functions of pressure and temperature, be alike in dimen- sionless form, One of the most difficult relationships fo match, and one of the most important, is the pressure/temperature relationship for saturated steam. This match can be improved by operating the ‘model at subatmospherie pressure levels and at low temperatures, The major consequence of not scaling the pressure/temperature relation is that the flow rates will not be scaled at all points within the steam zone, Proper pressure/temperature scaling is particularly important after steam breakthrough. ‘When low pressure levels are used in the model, it is difficult to match the dimensionless steam-density relationship required by the dimensionless form of SOCIETY OF PETROLEUM ENGINEERS JOURNAL ‘TABLE 1 — SIMILARITY PARAMETERS. FOR STEAM PROCESSES. se Fomine mane 0) Eq. A-9. To correct for this problem, a new parameter can be incorporated that is based on a simple approximation to the pressure gradient in the steam zone and the adjacent liquid layers. This is given by Parameter IIf in Table 3. As in most complex problems, some similarity parameters in the present case are mutually exclusive. Much of this exclusiveness can be avoided by matching 44S between the model and the prototype. If this is not done, however, some compensating factors can be applied in particular problems. These are discussed in Appendix A. Because of the difficulties in matching the similarity parameters, a set of scaling parameters is, generated that generally can be matched between model and prototype for a large number of steam processes. These are given in Table 3. The development of this set of parameters is discussed at the end of Appendix A. Since there are a number of subtleties that arise in the application of the scaling rules to a particular problem, an example illustrating the details of their application is provided in Appendix B. Physical Models General Design Physical modeling of thermal processes in multiwell field projects necessitates complex equipment capable of controlling and measuring fluid flow and heat transfer at a number of points in the model, The design is interrelated with the scaling rules and dependent on the materials and fluids available for construction and operation, Because most field projects have more injection and production wells than canbe handled reasonably in model ex- periments, the smallest “element of symmetry” that Still represents overall project behavior is chosen. In field cases with formation dip or edge water, the smallest element of symmetry usually is a full dip section of the reservoir with lateral no-flow boun- JUNE 1980 ‘TABLE 2 — CHARACTERISTIC QUANTITIES FOR SCALING STEAM PROCESSES 5 son npstans perma, & dese pene He ‘TABLE 3— SCALING PARAMETERS. FOR STEAM PROCESSES. (aptamer or ree ow fot preaminrtess ay cae Me Gf a6 eee aries. Fig. | illustrates an element of symmetry that was chosen for a Midway Sunset field steam- injection project. To minimize scaling factors, the size of a model is selected as large as possible within the constraints of laboratory space and time available to perform the experiments. In most applications, a length scaling factor of 150 to 250 will result in a ‘model experiment of bench-scale proportions that can be run in a single day. The central component of the apparatus is a glass-bead pack contained by rigid fiberglass sides and thin, flexible Teflon® sheets at 13 Fig. 1—Prototype element of symmetry —L sand, Midway ‘Sunset fel Fig. 2 Cross section of atypical thermal vacuum model. the top and bottom (Fig. 2). Cap and base rock are mulated by layers of plate glass cemented together with a thin layer of silicon cement or by solid blocks of rock (e.g., limestone). Injection and production wells are built into the fiberglass frame at ap- propriate locations. In all cases, geometric similarity is maintained in three dimensions with a single scaling factor. Auxiliary flow equipment is required for injection and production of heated fluids (Fig. 3). Because of the complexity of operations and the number of injectors and producers, electronic equipment is used for control during the experiment and for measurement and retention of the large amount of data generated by a single experiment. Model Frame Construction of the model begins with the fabrication of an aluminum mold upon which a Jaminated U-shaped frame is built up from alternate layers of fiberglass and epoxy resin to a wall thickness of about % in. Prefabricated wells, machined from blocks of factory-made fiberglass, are cast integrally into this frame. If the prototype reservoir has dip, wells are set at a suitable angle since the model frame parallels the bedding planes. (Details of the well design are given in the next section.) After curing, circular windows are ‘machined along the sides of the frame. A sheet of Teflon is glued at the window inside the frame and another is placed on top of clear epoxy that is cast in the machined holes. This provides good optical surfaces through which fluid movements can be observed inside the model. Ports also are drilled at regular intervals along the frame to permit insertion of thermocouples inside the model. 7 ecto m Fig. 8~Schematic of injection and production equipment. Isa SOCIETY OF PETROLEUM ENGINEERS JOURNAL ‘Well Models Direct geometrical scaling of wells is not possible for the small size required for the model. For example, with a length scale of 200, a model well is only about 0.030 in. in diameter. In addition to the mechanical problem of constructing such a well, the well size loses significance when the diameter of beads composing the porous medium approaches or ex- ceeds the model-well diameter. Still, scaling the pressure drops in the near-wellbore vicinity is im- portant, especially for steam-soak processes. Two well-model techniques have been developed to circumvent this problem. In the first, a linear section fof the bead pack is extended beyond the scaled reservoir boundary so that the pressure drop properly ‘matches that of a radial system with the correct well diameter. An example quarter well is shown in Fig. 4 The approximate relationship (given by Van Meurs) of the linear well dimensions to a radial well is t bn Fagg wo The linear well technique has the disadvantage of having a disproportionate reservoir volume to be heated. Consequently, a second technique was developed that uses a’ slit well at the wall of the model. The effective radius of a slit well is a function of the slit width and the arrangement of the bead pack directly in front of the slit. With no effect from the beads, as in Fig. Sa, an approximate relationship between slit width and effective model-well radius is @) For a bead diameter that is twice that of the slit, as in Fig. Sb, the maximum value of the effective radius is r=. Although the exact value of the effective radius is not determined easily, the difference in relative inflow for these two cases is less that 20%. Dery =, Cap and Base Rock Cap and base rock heat losses usually are simulated by cementing Yin. glass plates together with clear silicon potting compound. Before the cement has hardened, the plates are loaded to extrude excess ‘compound, leaving only a thin cement layer. This thin layer ‘allows sufficient thermal expansion to avoid cracking of the blocks during the thermal experiment. Glass blocks about 9 in. thick introduce only small boundary effects for experiments not exceeding 15 hours. The associated error was estimated by D.G. Whitten, who compared cumulative heat loss to an infinite cap rock that has had its lower boundary temperature raised an amount, AT, 2k,ATVE Vira with the corresponding heat loss to a cap rock of finite thickness, z., for which the extermal boundary isheld at the original temperature: y= Mas (1 save FS inte TEE). Nao = iat Qe --@) a JUNE 1980 Fig, 5—Silit well model 5 seazst 8 2 2 Fig. 6 Heat loss error trom finite cap and base rock. wean seve orewnc, 3 Fig. 7~ Permeability as a function of bead size. 156, Eqs. 3 and 4 are the standard linear heat-flow ‘equations integrated over time. Errors in heat loss for typical experimental conditions are shown in Fig. 6. ‘The glass blocks also permit visual observations of the steam zone through the top and bottom of the model Bead and Sand Packing Several factors—including absolute permeability, relative permeabilities, and porosity—must be considered in scaling the porous medium. The ab- solute permeability is inversely proportional to the length scaling factor (see Eqs. B-l4 and B-16), Therefore, for typical reservoirs, model per- meabilities of the order of 1,000 darcies are required. ‘The relationship of sand grain or bead size to per- meability is derived from the data shown in Fig. 7. For a group of packs of 0.36 porosity, the relationship is, dances = 74,000 (dm)? 6) Glass beads approximately 0.1 em in diameter typically will “satisfy the _absolute-permeability requirement for the model; however, relative per- meabifities and especially end-point saturations are not ordinarily matched to those from low- permeability, partly consolidated, natural sands or sandstones. Ideally, the shape’ of the relative- Permeability curve ‘based on movable saturation would be matched between mode! and prototype; however, the effect of minor deviations in shape is not expected to be significant ‘The mobile saturation range appears in the scaling with porosity as the factor #45. Complications arising from differences in AS between model and prototype are discussed in Appendix A. This scaling analysis indicates that AS should be matched as nearly as possible. Since little can be done to reduce AS, which typically is too large with glass beads compared with reservoir rocks, one method of achieving a lower AS is to reduce porosity with a binary bead pack. It is well known'® that the imum porosity for a uniformly sized pack of spheres ranges from 0.2595 for hexagonal close- packed t0 0.4764 for cubie-packed spheres. Of course, these extremes cannot be achieved without hhand-packing the beads individually. A random packing of spheres characteristically yields a porosity of up to 0.37. Using the packing screen technique of Wygal and Naar,'2 porosity as low as 0.33 can be obtained under nearly ideal conditions. A theoretical minimum porosity for a binary pack is derived by assuming two bead sizes having such a large difference in size that the void space of the larger bead structure is unaffected by, but completely filled with, the smaller bead pack, as illustrated in Fig. 8. Since the large beads effectively diminish the volume to be filled by a pack of the small grains, the overall porosity is, B= ON bg eee a O) For an ordinary packing efficiency of SOCIETY OF PETROLEUM ENGINEERS JOURNAL y= $420.35, the resulting theoretical minimum porosity is only 0.12. This minimum porosity should occur when the proportion of large beads present is such that they just “fill” the container if packed alone at their attainable porosity, $,. Thus, the fraction of the total volume occupied by the large beads is (1 The small beads are proportioned so that ther a sufficient number to fill the voids between the large beads. Therefore, the fraction of the total volume occupied by the small beads is ¢,(1—¢,). The proportion by weight of the larger beads (assuming equal density for both sizes) is (1-6) Ds (1 6,6,) Again, if ,=0,=0.35, the theoretical weight fraction of large beads from Eq. 7 is 74¥% compared with 73¥o determined experimentally by Naar and Wygal.2° As dy/d, becomes small, the boundary effect of the large beads on the small-bead packing becomes important and higher porosity results. Fig. 8 shows some experimental values for porosity as a function of the dy/d, ratio and weight fractions of sand grains, ‘These data demonstrate that reduced porosities can be obtained but that boundary effects still prevent attainment of minimum porosity. Since the average permeability of the pack is determined primarily by the small-bead diameter, these beads are chosen first. The other beads then are chosen to be as large as possible, within the constraint of avoiding large nonequilibrium effects that result from the finite time needed to heat the center of these large beads For layered prototypes, various bead sizes may model changes in permeability over the vertical interval. Shale breaks can be simulated with plastic sheets or fiberglass spacers. Often, the extent of shale breaks is not defined well enough in the field to model this effect precisely; however, upper and lower limits of process behavior can be investigated. o Model Assembly The base blocks are positioned on a rigid, horizontal steel frame, which may be pivoted to permit erection fof the model to the proper dip after the model has been completed. Careful leveling of these base blocks is essential to avoid sharp edges that can result in leaks in the lower side of the model. One 0.020-in. sheet or two 0,005-in. sheets of bondable FEP Teflon film are cemented with epoxy to the lower flange of the fiberglass frame of the model. After curing, the frame is placed on the base blocks, which have been coated with a silicon potting compound. The lower Teflon sheet is made to conform to the glass blocks by applying mild loading to extrude excess silicon. Upon setting, thermocouples and valves are con- nected and the model is ready for packing. This is accomplished with a modified Wygal screen device that can be moved back and forth along the model and that is equipped with two independently motor-driven conveyor belts. These proportionally deliver the predetermined ratio of large and small JUNE 1980 beads into the model. After the beads are packed up to the upper flange, minor skimming is needed to smooth them even with the upper flange. The upper sheet then is bonded to the flange with epoxy cement as before. After the epoxy has set, the model is evacuated. With the application of a vacuum, the model becomes rigid and the glass beads are im- mobilized as long as subatmospheric pressure is maintained. Leaks often are a troublesome factor in large models, Before the cap blocks are placed, the model is checked thoroughly for leaks with a vacuum gauge. In the event of a leak, the source is found either by pulling water into the model and observing bubbles or, as a last resort, by slightly overpressuring the model with freon and sensing with an electronic freon probe. Placement of the cap blocks is accomplished in a manner similar to that used on the lower side. The upper film, however, is irregular because of the identation of the grains resulting from the vacuum. Sufficient silicon cement is placed on the sheet to fill the indentations, and the cap blocks are lowered carefully onto the model. Again, silicon cement is extruded, this time by the weight of the cap block. After curing, the model is connected to the injection and production equipment. The sides of the model are insulated with about 4 in. of Styrofoam™ to minimize heat loss across these ‘‘no-flow” boun- daries. At this point, the model is ready for fluid saturation preparatory to the thermal experiment. Steam Injection Equipment Injection equipment for steam consists of two calibrated storage vessels (one for water and one for water plus polyethylene glycol), injection pumps, a steam generator, and electronic control equipment for timing and metering flow rates. Steam is generated by pumping pure water (0 to 3 mL/s) through an 8-ft length of 4-in.-OD, 1/16-in. ~ID, stainless steel tubing that is coiled inside an insulated, box. This tubing, in addition to serving as a conduit for the fluid, serves as an electrical resistance heater. A: to 100-A, 24-V power supply, connected at the ends of the tubing, provides the thermal energy to convert the water to superheated steam. Water pump rates and elecirical current are controlled to yield the desired quantity and heat content of steam. The superheated steam then is mixed with a polyethylene glycol solution to yield the required steam quality and aqueous-phase viscosity for injection into the model Steam is diverted to individual wells by a manifold system with automatically controlled valves. All injection lines and valves are insulated to minimize heat losses before the steam enters the model. Excess steam that is generated, especially when a well is shut in, is bypassed through a backpressure regulator, condensed, and collected in a calibrated reservoir. Production Receivers Each model well is equipped to inject steam or to produce oil, water, and steam, Production is ist collected in alternate graduates by a system of isolated, double-vacuum receivers that permit changes of graduates during the run, Production- well pressures are sensed with pressure transducers located in the wells, and vacuum is regulated in the receivers with a vacuum regulator driven by the electronic control system, Production lines to the receivers are heated to prevent cold oil blocking and a check valve prevents accidental backflow of air into the model during graduate changes or in the event of excessive pressure in the receiver. Experimental Procedure Model experiments of multiwell thermal projects involve considerable numbers of operations, as shown in the flow sheet in Fig. 9. Preparation for an experiment and processing of results normally takes 2 to. 3 weeks, although the actual scaled run is per= formed in a single day. Operations mainly consist of (1) saturating the model with water and oil, (2) cooling to a uniform initial temperature, and @) injecting steam and producing steam, oil, and water while measuring volumes, temperatures, pressures, and steam-zone movements. Saturation Following preparation of a new model as described in the previous section, the frame is tilted to saturate the model with water from the downdip end with a gravity-stabilized front. Air is excluded from the model by introducing CO, and maintaining a vacuum in the downstream collection vessel. A mixture of polyethylene glycol and water, scaled to yield the proper aqueous-phase viscosity (ap- proximately 2 cp at room temperature), is mixed in a barrel and pulled by vacuum into the upstream receiving vessels. After a suitable deaeration period, the aqueous phase is introduced into the model slightly below atmospheric pressure at the upstream end and with a deep vacuum (~ I psia) at the outflow end The measured volume of injected water from the receivers is checked with calculated porosity to verify the completeness of saturation. Injection of oil is somewhat more complicated because of its high viscosity. Viscosity of a typical blend of two DUTREX® oils might be as high as 1,000 cp at room temperature. To hasten the saturation, oil may be injected and withdrawn at intervals along a model. Again, all parts of the model must be maintained at vacuum to retain rigidity of the bead pack; however, the shorter distances for flow result in higher pressure gradients and higher velocities. In some cases, the entire model is enclosed ina portable, insulated air bath and heated to about 150°F. By reducing the oil viscosity, the time required for saturation can be reduced greatly. The oil injection ordinarily saturates the model up to 0.90 or 0.95. Depending on the prototype saturations required, the model then may be (1) waterflooded in a controlled manner to reduce overall oil saturation, (2) waterflooded to position edge or bottom water, ot (3) gas driven to create an initial gas cap. 138 Cooling Since model temperatures are limited to about 200°F by the subatmospheric pressure requirement, the ‘model temperature range is extended by cooling to just above the freezing point. This extended te erature range improves the scaling match of viscosity, steam quality, and the pressure/tem- perature relationship for saturated steam. After all auxiliary systems (transducers, ther- mocouples, and production- and injection-well controls and insulation) are connected, the model is closed in an insulated air bath. Portable air con- ditioning units then refrigerate the entire model to about 40°F over a period of several days. When the thermocouples at the center of the model indicate uniform cooling, the model is ready for the thermal experiment. Conduct of Experiments Initial operation in an experiment may involve a brief period of cold-water injection and production to establish base lines for well productivity. In steam- soak processes, cycles of injection and production are controlled automatically, but frequent graduate changes must be made manually in each well, On large models, especially during start-up, three or four operators may be needed to monitor injection and production, to record data, to change graduates, and, to observe and record steam-zone growih. Since a typical prototype year is about 30 minutes, duration of a single steam-soak injection is commonly less than 5 minutes. As discussed previously, the total length of an experiment is limited by the finite thickness of the cap and base blocks, which results in sizeable heat-loss error after 12 to 14 hours; however, few prototype processes are expected to extend past this time (~25 years), The produced mixture of viscous oil and water is collected in the graduates as an emulsion, At the end. of an experiment, these produced fluids are treated with a commercial emulsion breaker and sodium. chloride (10 increase the density difference), heated to 150°F for 12 hours, and cooled. Oil and water production are then read and recorded. Other data collected during the experiment consist of the following. 1. Thermocouple readings of temperature as a function of time are stored on magnetic tape. 2. Pressure transducer data by well are recorded on strip charts, 3. Injected and bypass fluid-level readings are recorded manually. 4, Injection timer records are stored on tape. Alll these data are processed, cards are punched, and most calculated results ‘are computed and tabulated by standard data reduction methods. Applications We have performed scaled experiments for design or assistance in operation of steam drives and steam soaks fora number of fields including Tatums,2! Coalinga,” Slocum,?? Mr. Poso,4 Peace River,?5 Yorba Linda,*® and Midway Sunset. SOCIETY OF PETROLEUM ENGINEERS JOURNAL ‘e:npedoid yeyuawi8dxs 10} wesSeIp mos 6 “BL ‘oo 4 139 JUNE 1980 Le n CCS ane seats Aw) S| nee at) Cos Fig. 10 Mt. Poso field model These projects represent the whole spectrum of response from exceptionally good to poor. The experience derived from the field tests, together with the insight gained from the model studies, has helped us in recent years to choose the more favorable reservoirs for projects and to operate them in a ‘manner that will maximize oil/steam ratios and/or recovery.'7 In the case of Mt. Poso field, the model studies directly influenced the decision to proceed with a steam drive there. In spite of previous un- successful or marginally successful steam drives in the other fields, the model studies for Mt. Paso field provided sufficient contrast to warrant a recom: ‘mendation for the steam drive. Examples of the type of information that can be obtained from model studies are presented for results from two of our large model studies on Mt. Poso and Midway Sunset fields. Mt. Poso Field The Mt. Poso laboratory model is illustrated in Fi 10 for the 250-ft-wide, half-well, dip element of symmetry given in Ref. 24. Wells are spaced at 500-ft intervals from updip (Well 1) to downdip (Well 9) Model and prototype conditions and laboratory materials used are given in Tables 4 and 5. ‘A sseries of about 50 experiments demonstrated that steam layover would be a significant factor, with gravity causing the oil and water to flow downdi 160 ‘TABLE 4 ~ STEAM-DRIVE PROTOTYPE PROPERTIES FOR MOUNT POSO, UPPER VEDDER RESERVOIR SOCIETY OF PETROLEUM ENGINEERS JOURNAL ‘TABLES MOUNT POSO MODEL GLASS BEAD ‘AND FLUID DATA, The experiments also showed that injection into two wells along dip would be needed to heat the reservoir with a reasonable project life. Fig. 11 shows typical results of several experiments in which different wells were used as injection points. Driving cold-oil banks from two updip injectors resulted in high steam-zone pressures and consequently lower oil/steam ratios. Heating only downdip improved the oil/steam ratio but depleted only part of the reservoir. The optimum flooding method for dealing with the water influx was studied in the experiments shown in Fig. 12, Steamflooding the upper half of the reservoir (Experiment 37) permitted a low steam-injection pressure (125 psi) toward the end of the project. This increased the oil/steam ratio, but cumulative oil recovery was low because of lack of downdip sweep. Other experiments with steam injection at Wells 3 and 6 gave higher oil/steam ratios when reservoir pressures were reduced by downdip water produc- tion. In these runs (Runs 26, 35, and 39), steam-zone pressures decreased to about 100 psi after 8 years. In Run 36, where water influx maintained higher pressures, the oil/steam ratios decreased at later times and were about 20% lower. Effect of oil-saturation distribution along dip was another prototype uncertainty that was investigated (Fig. 13). High oil saturation mid-dip, or especially updip, improved the process compared with a uniform oil saturation. Experiments on the effects of keeping production wells pumped off revealed that failure to capture oil updip by lowering fluid levels could result in problems of cold-oil bank buildup before the oil is produced downdip. In these experiments, the prototype permeability was 5 darcies. The effect of increasing permeability from $ to 10 darcies in one experiment was an crease in oil/steam ratio, which mainly reflected differences in steam-zone pressure. Even higher permeabilities (20 darcies) were used in later ex: periments as additional data became available from the field The final process design is based on these later experiments. Field performance to date is compared «with the corresponding model experiment in Fig. 14. In summary, the experiments with Mt. Poso field models delineated effects of uncertainties in JUNE 1980 . (es) Fig, 11Etfect of steam injection point ~ Mt. Poso field. i Fig. 12~Etfect of methods of handling water Influx ~ Mt. oso field 161 Fig. 13~Effect of high oil saturation middip-Mt. Peso fie. Mount poso PHASE 1 STEAM ORIVE Fig. 14~ Actual and model performance. se a Fig. 15 Model, fiold, and prototype relationship between oll viscosity and temperature, 162 prototype properties and in modes of operation and led to the conclusion that conditions for a steamflood in the Upper Vedder reservoir are very good, Midway Sunset Field The large model study of one of the steeply dipping sands in the Midway Sunset field is an example of scaled experiments on steam soak and steam drive in a viscous oil reservoir. The element of symmetry chosen is shown in Fig. 1, Model and prototype conditions and laboratory materials used are given in Tables 6 and 7 and Fig. 15. A series of 26 experiments, investigated the options of infill drilling or con- verting_an existing steam-soak operation to steam drive. Existing well locations are on the 2.5-acre spacing shown in Fig. 1 by Wells 2B, 3A, 4B, and SA. The first series of experiments, given in Fig. 16, compared steam soaking on 2.5-aere spacing with soaking on 1.25-acre spacing. Using an operating policy of one 12,500-bbl steam soak per year, the heating rate is doubled by the infill project. With the exception of Run 12, in which a single downdip well malfunctioned, oil/steam ratios (as indicated by the cumulative-oil/cumulative-steam graph) are slightly higher because of interwell heating. Compared with cold gravity-drainage production, steam soaking greatly increases oil production’ rate, yet project life is still extremely Jong. Note that'the steam-soak oil-production rate did not diminish even after 25 years. Typical steam zone growth and prototype temperatures for infill Experiment 10 (1.25-acre, one 12,500-bbI/well/yr soaks) are shown in Fig. 17 after 4 years of steam soaking, A second group of experiments studied the effects of a number of different methods of heating (Fig. 18). Again; the quantity of steam injected rather than the method of application is the most important factor determining the quantity of oil produced. Some differences, however, are observed. EX Periment 17, a 2.$-acre, 12,S00-bbl soak with a 2- year production period, was the most thermally efficient (highest oil/steam ratio), although the amount of recovery was least. Next, Experiments 4 and 9 are base-case 12,S00-bbl soaks on 2.5-acre spacing. Doubling steam-soak volume in Experiment 13 and quadrupling the soak volume in Experiment 11 successively reduced oil/steam ratios. Although injection rate per day was the same in each case, higher final injection pressures at the end of the longer injection periods diminished the steam-zone volume obtained per barrel of steam injected. Another factor tending to reduce oil/steam ratios in the large soaks was the additional displacement of oil from the vicinity of the wells. The highest yearly rate (50,000 bbi/yr), when achieved by double soak volumes into 1.25-acre wells (Experiment 19), was more efficient than the 2.5-acre, quadruple soak volume (Experiment 11). Oil response by well for the Experiment 19 steam soak (Fig. 19) illustrates the importance of downdip wells after the reservoir has been heated sufficiently for long-range gravity drainage to predominate. ‘Two experiments in which steam soak was con- SOCIETY OF PETROLEUM ENGINEERS JOURNAL TABLE 6 ~ STEAM SOAK MODEL AND PROTOTYPE ‘TABLE? - MIDWAY SUNSET MODEL GLASS BEAD PROPERTIES FOR MIDWAY SUNSET RESERVOIR ‘AND FLUID DATA. verted to steam drive after 4 years (Experiment 5) and after 11 years (Experiment 3) are contrasted to the base-case soaks in Fig. 20. The greatly accelerated rate of heat injection causes large increases in oil- production rate at some cost in oil/steam ratio. Project life was reduced substantially by the drives. Comparable oil recoveries would have required more than $0 years for 2.5-acre steam soaking at base rates. The steam-zone growth and temperatures for the steam drive in Experiment $ are given in Fig. 21, 4 years after starting the steam drive. ‘Conclusions as to merits of steam soaking or conversion to steam drive are dependent on economic considerations that are not treated in this paper. The detailed data, which have been demonstrated by the above examples from model experiments, can be combined with economic data to provide a basis for project design and for operating decisions. As ‘mathematical simulators become capable of handling a larger number of grid blocks, they undoubtedly will perform many of the functions of the present physical models; however, for the present, the laboratory models have unique capabilities that are useful in studying the most complex processes. Conclusions Physical models of steam-injection processes are useful tools for depicting behavior of complicated field projects. Information from model experiments can be used to (1) screen prospective field projects, (2) optimize design, and (3) provide guidance for operating policies. Nomenclature b = width of linear well model, L specific heat, L?/2°T well diameter or bead diameter, L. rate of energy injection, mL?/t° steam quality, dimensionless net-pay-interval/total-interval, dimensionless JUNE 1980 Fig. 16-Comparison of reference and infill soaks— Midway Sunset fel. 163 fraction of larger beads in binary pack, dimensionless Partial well fraction (well angle/360°), dimensionless acceleration due to gravity, L/1? enthalpy per unit mass, L/1? permeability, L? thermal conductivity, mL/2 7 relative permeability, dimensionless length of linear well model, L = length or distance latent enthalpy per unit mass, L?/12 arbitrary variable, Eq. A-28, or mass distance, direction normal to boundary, L pressure, m/Lt? conductive heat flux, m/13 cumulative heat loss, mL? /1? heat-loss rate, mL?/2 > " L Fig, 17~Temperature distribution after 4 years of steam ‘soaking ~ Midway Sunset field 1 = effective well radius (Fig. 5), L true geometric scale model well radius, Z saturation, dimensionless jemperature volumetric flux (Darcy velocity), L/¢ ‘mass rate of flow, m/t = coordinate distances, L z= coordinate distance normal to bedding planes, L reservoir thickness normal 10 bedding planes, L a = thermal diffusivity, L2/1 = function equal to ratio of values of its argument (m) in the prototype to that in the model, dimensionless dip angie, degrees viscosity, m/Lt Fig. 18 Comparison of various soak sizes and rates — ‘Migway Sunset tela SOCIETY OF PETROLEUM ENGINEERS JOURNAL p= density, m/L3 6 = porosity, dimensionless Subscripts aqueous phase for quality control bottomhole ‘cap or base rock dimensionless finite cap and base rock initial condition irreducible water phase (oil, water, or steam) lateral boundary movable saturation model normal to boundary oil residual oil saturation to steam produced prototype reservoir rock, except with k reference quantity used to obtain dimensionless number (see Eq. A-28) steam saturation temperature or pressure calculated steam temperature by Eq. B-9 in Table 8 total interval upper or lower boundary water total water injected net-pay-thickness/gross-thickness large-bead diameter small-bead diameter rst time interval or injection rate Acknowledgments Other than those referenced in the text, a large number of persons have contributed’ to the development of _ physical-model experimental techniques and scaling rules for steam processes. At the risk of omitting some, we acknowledge the work of C.A. Chase, R.H. Hammerle, H. Hooykaas, J.W. Hulett, G.F.R. Kingelin, F. Lehner, L. Schenk, D. Teeuw, F. van Daalen, H.R. van Domselaar, C. van Egmond, P. VanMeurs, KH. Wang, and D.G. Whitten, whose unpublished reports have aided in the preparation of this paper. We also express our appreciation to Shell Development Co. for per- mission to publish this paper References 1. Leverett, M.C., Lewis, W.B., and True, M.E.: “Dimensional Mode! Siudies of Oi! Field Behavior," Trans., AIME (1942) 146, 175-193, 2. Rapoport, LA, and Leas, WJ: “Properties of Linear Waterflocals,” Trans, AIME (1983) 198, 139-148 3, Rapoport, LA "Scaling. Laws for Use in Design and Operations of Water-Oi Flow Models," Trans., AIME (1985) 204, 143-150 4. Croes, GA. and Schwarz, N. “Dimensionally Scaled EX- JUNE 1980 Fig, 19~ Effect of well location along dip ~ Midway Sunset field Fig. 20-Effect of time on conversion to steam drive ~ Midway Sunset fel. 16s Fig, 21 ~ Temperature distribution 4 years alter conversion tossteam crive~ Midway Sunset field periments and the Theories on the Water Drive Praces Trans., IME (1988) 204, 35-42. 5. Geerisna, J.. Crocs, G.A., and Schwarz, N. ‘Theory of Dimensionally Sealed Models of Petroleum Reservoirs,” Trans, AIME (1956) 207, 118-123, 6. Perkins, F.M. Jt and Collins, R.E-: “Scaling Laws for Laboratory Flow Models of Oil Reservoirs,” Trans., AIME (1960) 219, 383-388 7. Carpenter, C.W, Jr, Bail, PLT. and Bobek, J.B “A Verification “of Waterflood Scaling in Heterogeneous Communicating Flow Models," Soc. Pet. Eng. J. (March 1962) 9-12; Trans, AIME, 225 8. Nieben, RL. and Tek, M.R.: “Evaluation of Scale-Up Laws {ot Two-Phase Flow Through Porous Media,” Soc. Pet. Eng J. (Dune 1963) 164-196; Trans, AIME, 22, 9. Baker, PLE. "Heat Wave Propagation and Losses in Thermal Oil Recovery Processes.” Proc., Seventh World Pet, Cone, Mexico City (1967) 3, 459-469 10, Diets, D.N.: "Hot Water Drive." Proc., Seventh World Pet ‘Cong., Mexico City (1967) 3, 481-457, G-G. Jr, Elzinga, E.R., Tarmy, B.L., and Willman, “sealed Model Tests of In Situ Combustion in Massive Unconsolidsted Sands," Proc., Seventh World Pet. Cong., Mexico City (1967) 3, 47-485, 12, Ruark, A-E.:“lnspectional Analysis," J Elisha Mitchell St Soc. (Aug. 1951) 1, No. 1, 127-133, 13, Richardson, J.G.: “Flow Through Porows Media,” Streeters Handbook ‘of Fluid Dynamics, V.L, Swrecter (ed). fist ition, McGraw-Hill Book Co. tne-, New York City (1961) Sec. 16, 161-112. 14. Loomis, A.G, and Crowell, D.C.: “Theory and Application ‘of Dimensional Inspectional Analysis to Model Stady of Fluid Displacements in Petroleum Reservoirs,” RI 6546, USBM (1368), 15, Niko, H. and Troost, PJ.P.M.: “Experimenta Investigation ‘of Steam Soaking ina Depletion-Type Reservoir,” J. Pet Tech: (Aug. 1971) 1006-1018; Trans., AIME, 281. 166 16. Harmsen, 6.1: “Oil Recovery by Hot-Water and Steam Injection." Proc., Eighth World Pet, Cong., Moscow (1971) 3.243251 17, Myhill, NLA. and Stegemeier, G.L.: ‘Steam Drive Correlition and Prediction,” J. Pet. Tech, (Feb. 1978) 173 1m. 18, Slichier, C.S.: “Theoretical Investigation of the Motion of Ground Waters," 19th Annual Report, USGS (1897-1898) Part 2, 295.84, 19, Wygal, RJ: "On the Construction of Models That Simulate Oil Reservoirs," paper presented. at 23rd Technical Con ference on Petroleum Production, Pennsylvania State U., University Park, Sept 1962 20. Naar, J.'and Wygal, RJ: “Structure and Properties of Unconsolidated Agaregates,” Can. J. of Physics (1962) 40, sI8831 21. French, M.S. and Howard, R.L.: “The Steamflood Job, Hefner Sho-Vel-Tusm,” Oil and Gas J (Duly 17, 1967) 65, No. 29, 646 22. /Afoeju, B.L.: “Conversion of Steam Injection 10 Waterflood, ast Coalinga Feld," J Pet, Tech. (Nov. 1974) 1227-1232, 23. Hall, AL and Bowman, R.W.: “Operation and Performance ‘of the Slocum Thermal Recovery’ Project,” J. Pet. Tech (April 1973) 402-408, Stokes, D.D., Brew, J.R., Whitten, D.G., and Wooden, L.W.: "Steam Drive as a Supplemenial Recovery Process in an lotermediate Viscosity Reservoir, Mount Poso. Fed California,” J. Pet. Tech. Gan. 1978) 125-131, 25. Prats, M.! “Peace River Steam Drive Scaled Model Ex periments,” paper presented at 281h Annual Technical Meeting, Petoleum Society of CIM, Canada: Venezuela Oil Sands Symposium 1977, Edmonton, Alta, May 1977 26, Stokes, D:D. and Doscher, T.M.: "Shell Makes a Suceess of Steam Flood at Yorba Linda,” Oil and Gas J (Sept. 2, 1974) 7178. 27. Land, N.S: “A Compilation of Non-Dimensional Num bers," SP-274, NASA (1972), Determination of Sealing Parameters The similarity parameters are determined by in- spectional analysis, consisting of (I) formulation of the governing differential equations in dimensionless form, (2) identification of the independent dimen- sionless parameters, and (3) selection of the characteristic quantities based on the most important property values and constraints. These similarity parameters then are modified by engineering judgment to obtain a set of scaling parameters that generally can be matched between model and prototype. A Cartesian frame of reference, with the x and y axes oriented parallel to the bedding plane, is selected for this analysis and is applied to an element of symmetry typically illustrated in Fig. 1. The model element of symmetry is scaled geometrically to that of the prototype so that all characteristic lengths (including layer thicknesses and completion intervals) are geometrically similar. The following assumptions are made. I. Three phases may exist: an oleic phase, an aqueous phase, and a steam-vapor phase (no volatile hydrocarbons). 2. There is no partitioning into or out of the oil phase (dead-oil assumption), 3. Rock compressibility and thermal expansion are negligible, 4. Darcy's and Fourier’s equations are valid, 5. Capillary-pressure effects are negligible. 6. The system is in local thermodynamic equilibrium. SOCIETY OF PETROLEUM ENGINEERS JOURNAL. 7. Kinetic energy, potential energy, and viscous dissipation energy are negligible compared with the thermal energy. 8. The enthalpy and internal energy are essentially equal for the oleic phase and for the aqueous phase and are linear functions of the temperature. 9. The difference between the steam enthalpy and internal energy can be neglected. 10. The time rate of change of the specific steam enthalpy in the steam zone is negligible. 11, The internal energy of the rock is a linear function of the temperature. 12. The saturated steam temperature is the ‘maximum temperature at any location. 13. Relative permeabilities depend exclusively on the saturations. 14. Soy and Sjy are constant and uniform throughout the model. 15. Critical saturation for steam flow is assumed tobe zero. 16. The changes in the density of the immovable water and residual oil are negligible. Governing Equations With the above assumptions, conservation of mass applied to the oil phase yields A(PoS0) oat where the subscript o denotes oi The continuity equation for the water (in both the liquid and vapor phases) is given by AowSw) ar $V (Doig) =0, 2 (AD (0,85) ar 49+ (Pyiy) +O +9 (pgiis) =0, (A2) where the subscript w signifies water in the liquid phase and s signifies water in the vapor phase (team). From the assumption that Darcy's equation ap- plies we may write for any species j, where j=0, w, ors, kky ~~ (vp~pj2) (A3) * ij For the assumptions stated, conservation of enerey for the reservoir can be expressed as ar (1 8)2,C, +8 (0 CoS0 +n CwS 0) 5, (058s) zo wa eee] + (BoCollg + PyCwily) VT + bile Th + V-G=0. -(A4) By assumption, the conductive heat flux is ex- pressed by the Fourier equation, ky VT. (As) JUNE 1980 When three phases are present, the saturation identity becomes So4Sy AS, A6 AL saturated conditions, the pressure and tem- perature are interdependent and are related by the Clausius-Clapeyron relationship, which in functional formis Psa = Psat Tsai): (A-7) ‘The remaining relationships required to describe the system are the constitutive equations that express the dependence of the material properties on the thermodynamic-state variables. These can be ex- pressed in functional form, = duy2), + (A-8) 2) = o(R.T), (AD) k= k(x), (10) ky = hy (Ss (AU) 4 =D» (012) he = he(D), =(A-13) Ly = Ly(1), -(A-ld) ky = ky (92), (As) Pr = PARE» + (A16) where § denotes dependence on the phase saturations. The above expressions represent. 24 independent equations in 24 unknowns; thus, they constitute a complete set for describing the reservoir behavior, Since there are no movable fluids in the cap and base rock, a reduced number of equations is required in these regions and it is convenient to write a separate set of governing equations. The equations of significance are the energy equation, given by cay. Coy and the heat-flux equation for q, given by Eq. A-5, (A17) Constraints Analysis of the partial differential equations — representing conservation of mass, momentum, and energy ~shows that the system can be represented by three of the five primitive variables~p, 7, and the three S,. Since the actual selection depends on whether there is steam in the system, conditions be given on all five. All the time derivatives of these variables are first order; thus, only one initial condition needs to be specified for each, These can be represented by SHUHIZ)=SH(HIZs « “(A-18) PURI) =P), oe vec cee ee AAI) and T(ixy.2) =T,=constant, (020) where the subscript / denotes the initial value, 167 The necessary boundary conditions are treated ‘most easily in a general manner by considering the mass and energy fluxes at the boundaries. If these fluxes are specified at all boundaries, then the problem is well posed, With the exception of the Wells, the mass Mux is zero at all boundaries; hence, al” =0, bjttpl? 2 (A21) where /b signifies lateral boundary and ub signifies upper and lower boundaries. The subscript n refers to flow in a direction normal to the boundary Similarly, the convective heat flux is zero at all lateral boundaries exclusive of the wells; thus, ary | (A-22) Atthe upper and lower boundaries, ar yu ar yuo kur Sal, =. he Sn |e (A-23) AL the injection wells, the mass injection rate of wet steam is given by % ae werd) * (oatlan * Pstisn) oe (a2) a At the production wells, either the mass flux or the pressure can be specified, For simplicity here, the bottomhole pressure is specified such that P=Py(O (28) The rate of energy injection at the injection wells is aiven by Bard| (oyttynCyAT+05tanhs) WUgLy+CyAT), (0.26) where f, is the average quality of the steam at the injection-well face and AT is the temperature dif- ference T,—T;, with T, the injection-steam tem- perature and; the initial reservoir temperature. With the bottomhole production pressure specified, the mass rates of production and the energy production rate can be determined without further specification, In steam processes, there will be a time when the steam injection rate is reduced to zero. If, is the first time this occurs, this temporal constraint cean be expressed as (A27) Dimensionless Form Each variable, parameter, and operator can be made dimensionless’ by dividing it by a characteristic reference quantity, mg. This dimensionless ratio for the variable mis designated mp (A28) mp The conservation-of-mass equation for oil in dimensionless form then becomes, 168 (2e5abe ) A(2oDSomp) Unle ap + VD UKopitep) =O. oe eevee (829) Note that movable oil saturation has been used. Since residual oil and immovable water generally are not the same between model and prototype, the use of movable saturations improves the match of the relative permeabilities. For the water, the dimen- sionless form is » dkeuDS wo + expSsp) OpSel, Cee Uple +p Wupiinn* Pspiigp) =O. ---..(A30) Darey’s equation in dimensionless form becomes wrtale)s _ koky watialer iz, — Koka Ty yp, (289 ta am Prealp z, . onto). A3l Ce veo] (a3) where the relative permeabilities are dimensionless and have not been further reduced. ‘The energy equation, combined with Fourier’s equation, becomes in dimensionless form PeRCR. 3 =) [A 0) PC, +O (Coors (Sane, I e.C, +9(0 ay ; +oxCwSve)] 5? +0 (BopCoDSomo p ap (Lo town, z PavCwpS umd) 50 *( b HespSsp) Cami wate (Lon SeSnbn Cele ) anit V(bep+CupTo) + (ee) onSala *(PopCopiton + PupCuviiwn) -V oTo Kner 2 (Sea ee re enw V0 Tp =0, ..(A-32) where pceCep has been used to nondimensionalize the terms in the first set of brackets The saturation identity is already dimensionles however, the movable saturations must be scaled Thus, the saturation identity is written as So—Sors 5k I +Swmp+Ssp= —— (A-33 Se (A333) SOCIETY OF PETROLEUM ENGINEERS JOURNAL where the subscript m signifies movable. Here, the immovable steam saturation (critical gas saturation) hhas been taken equal to zero. The remaining sup- porting relationships governing reservoir behavior remain identical in form to those shown in Eqs. A-7 through A-16, with the exception that all the quantities have the subscript D. ‘The energy equation in the cap and base rock, Eq. AIT, becomes (A334) The initial conditions given by Eqs. A-18 through A-20 retain their form but, being dimensionless, carry subscripts D. Boundary conditions Eqs. A-21 and A-22 also retain their form with the D subscripts added. Eq. A-23 becomes aT py po aT p yx ‘WD Gn |, = Khed (A-35) Sap |r ~*™P aap le ‘The equation forthe injection rate becomes dp (ou end (A336) Eq. A-25 for the bottomhole pressure remains un- changed in form, with the quantities acquiring D subscripts. The’ energy injection-rate equation becomes (3) [( oteo +Cuoato] 20 hg erty) [rnoCwoonodT n+ (Ge ae Psp" saphsn] HAD) Eq. A-27 becomes tai Wp =0; Ep (438) ion of the Independent less Parameters With the governing equations in dimensionless form, the dimensionless parameters are simply the coef- ficients of the terms. It remains to determine which of these are independent. In general, this ean be done by observation; however, if necessary, it can be done rigorously using the Buckingham Pi theorem. The independent parameters that result arc tabulated in Table 1, with the name most commonly used in- dicated (see Ref. 27 for a comprehensive collection of dimensionless numbers). In Table 1, up has been eliminated through the use of the’ dimensionless parameter on the left side in Eq. A-31, which is, equivalent 0 substituting Darcy’s equation for the velocity in all relationships. JUNE 1980 If all the parameters in Table 1 could be matched, and, further, if Eqs. A-6 through A-27 were the same in dimensionless form, then the scaling requirements would be satistied completely subject 10 the assumptions stated. Selection of the Characteristic Quantities The object in the selection of characteristic quamtities is to choose them so that the system is invariant as the scale is changed from the model to the prototype. To. illustrate this, consider Eq. A-33. If Sg is chosen as Sa=1-Sors~ Sins (A239) then the saturation identity becomes Somb + Symb *S:0 (A-40) Thus, Eq. A-40 is invariant between model and prototype, provided S,, and S;,, are constant. This hhas been assumed to be the case. Thus, the initial condition on the oil saturation, given by Eq. A-18, becomes. Soli Ir2) ~Sors _ Soi Is , .(Acdl) 1-Sors—Siw 1— Sors — Siw and to have similarity between model and prototype, the following condition must be satisfied: (42) where 7(m) is defined as the ratio of the value of m in the prototype to that in the model In a similar fashion co that chosen for the saturation scaling, the value of pg is taken to be the difference between the maximum and minimum expected pressure: PR=Pmax ~ Prin = SP mas (A-43) These values do not have to be known exactly, but it is beneficial to have reasonably good values so that the best decisions can be made in matching properties between the model and prototype. All pressures are referred to the minimum pressure level so that the initial condition given by Eq. A-19 becomes 2) Prin Piir Panax ~Pavin Pi(6I2) —Prin (Add) Prax ~Pavin For similarity, then, the following initial condition ‘must be met Pi%IZ) ~Pavin l- “(&-45) Pmax ~ Prin If compressibility effects are not scaled, this condition can be relaxed. In a similar fashion, the characteristic temperature, T, is chosen to be the difference between the maximum and minimum. values. In the general case, the minimum temperature will be the initial reservoir temperature and the maximum temperature will correspond to the saturated steam temperature at the maximum pressure, With this choice for the characteristic temperature, Eq. A-20 becomes, in dimensionless form, 169 THI) Tin T ‘max ~ Tin we The initial condition on the temperature then will be satisfied automatically in both model and prototype. Proceeding in a similar fashion, Lg is selected to be z), fq is chosen as t), wp is taken to be w,, and the Fest of the characteristic quantities are selected based on the best property fit or value. The characteristic quantities determined in this manner are tabulated in Table 2. Selection of Scaling Parameters It is not always possible to satisfy all the similarity requirements, for the following reasons. First, the pressure drop in the model must be much less than in the field to scale gravitational effects. To obtain a good match of the Clausius-Clapeyron equation for ‘small pressure changes, the pressure must be reduced to vacuum pressure levels. At these low pressure levels, the steam density in dimensionless form is simply not the same as the prototype steam density. As a consequence, a scaling group must be in troduced to compensate for this. Second, a related problem arises in trying to simultaneously match Parameters 3 and 7 in Table 1. Parameter 7 can be matched by adjusting fye: however, Parameter 3 eannot be well-matched, Some compensation for this error is provided by the parameter introduced to compensate for the steam- density difference between model and prototype. ‘As Tong as the pressure drop in the steam zone is small compared with the overall pressure drop, the steam-density error is small, since the proportion of injected steam that remains in the vapor phase is very small. However, once steam breaks through into a production well in a steam-drive process, the steam density plays an important role in establishing. the pressure gradient. If flow is segregated, the gradient established will correspond closely between model and prototype, provided the following parameter (due to P, VanMeurs) is matched: Sonitsron yer, This parameter is obtained by equating an ap- proximation for the pressure gradient in the steam zone to that in an underlying liquid layer. Examination of Eq. A-32 shows that if the tem- perature gradients in the steam zone are small and the rates of change of the steam density and saturation are slow, then the major term involving Parameter 3 is the flux term in the second set of brackets. Use of the group given in Eq. A-47 then yields Parameter 7 as the dimensionless coefficient to be satisfied. In addition, the conservation-of-mass equation for the liquid water and steam vapor (Eg. A-30) is ap- proximately scaled through the use of Eq. A-47. Third, when GAS is not the same between the model and the prototype, Parameters 4 and $ in Table 1 cannot always be satisfied simultaneously because of limitations of available material. Binary packs can be used to minimize this problem; however, it usually is not possible to make +(¢AS) (a4) 170 quite reach unity. To treat this case requires consideration of two factors. When A$ is not matched, it must be determined whether the heating of the reservoir or the cap and base rock predominates. In the former case, Parameter 4 should be honored, In the latter, Parameter 5 should be used. If early-time behavior (or reservoir heating) is of primary concern, Parameter 7 should be honored as is. However, if heating of the cap and base rock is of concern, Parameter 7 should be modified as indicated in Parameter Il in Table 3. It is clear from the above that a concerted effort should be made to match @AS between model and prototype. Based on the preceding discussion, the similarity parameters given in Table 1 have been modified to yield a set of scaling parameters as shown in Table 3. A discussion of their application is provided in ‘Appendix B. APPENDIX B Example Scaling Procedure ‘Once a set of scaling parameters has been generated, a number of engineering decisions must still be made to implement the scaling. An example has been generated to demonstrate how physical models of steam-recovery processes can be scaled. ur scaling technology has evolved over the past 15 years; therefore, the procedures for scaling the various models described in the body of this paper differ slightly. The example given below is our current method of accommodating the compromises necessary for scaling. Field Characterization The first step in making calculations is selection of the prototype values to represent the field. Quite often, a single value of a quantity is not characteristic of the whole field or the element of symmetry to be modeled. In such a case (e.g., if the quantity was the permeability), a decision must be made as to whether the model should be constructed with a similar distribution or a constant “average” value. The field quantities to be specified are well spacing, net and gross pay thickness, porosity, permeability, thermal conductivity, heat capacity, movable saturation, oil viscosity and density as a function of the temperature, initial reservoir tem- perature, and range of the injection pressure. With the exception of injection pressure, these quantities generally are known and will not be discussed fur- ther. The process pressure range usually is bounded on the upper side by the fracture-gradient restraint on the injection pressure and on the lower side by the bottomhole production pressure. Because _of mechanical or operating limitations, the production well often cither is not pumped off or has a shut-in casing annulus. For the example calculation given here, a 50-psia backpressure on the formation will be used. (2p) p= 50 psia -B) SOCIETY OF PETROLEUM ENGINEERS JOURNAL ‘TABLE 8 - EXAMPLE MODEL AND PROTOTYPE SCALING VALUES Selection of Length Seale (Bp) = 1 sia. The selection of a length scale for the model, although somewhat arbitrary, is determined by temperature/pressure relations and by the physical model size considerations. If the prototype production pressure is 100 psia or less, then to match the pressure/temperature relation ‘for saturated steam, (L) should be made as small as possible (about 100 or less). This means the model should be as large as possible, within the practical constraint of cost and time necessary to saturate, cool, clean, etc. Further, the increased possibility of leaks and the increased structural requirements of a larger model must be considered. Suppose, for this example, that the reservoir thickness is 229 ft. Further, that because of laboratory ceiling clearance, cap and base block size, and drainage into receiver considerations, the model reservoir thickness should be no greater than 12in, The minimum length scale, therefore, is 229 “Tit Lp L)= = 229) oo eeeeee ee (BD) Lv where the subscripts P and M signify prototype and model, respectively. Selection of Model Production Pressure The selection of the model production pressure is straightforward and appears to be arbitrary in that the scaling does not directly dictate what value should be chosen. However, to scale the temperature within the model when there is an appreciable pressure drop in the steam zone, the pressure/temperature relation for saturated steam must be scaled. For prototype pressures as low as 50 psia and typical length scales of 100 to 200, the pressure/temperature relation at saturation can be matched best if one selects as low a value for the model production pressure as possible. From a practical standpoint, the lowest value that can be maintained because of vacuum pump limitations, vapor pressure considerations, etc., is about 1 psia. For an initial value in scaling calculations, this value is recommended. If the length scale is quite small or the production pressure is quite high, a higher value ‘may be chosen. For our first iteration, then, JUNE 1980 Calculation of Model Differential-Pressure Once a value for the model production pressure has been chosen, model pressure scaling is calculated from Scaling Parameter 1 in Table 3. ap) = 2=Pede . epsple (-Pp)m— ePMsuieM Since the length scale is known from Eq. B-2 and the ratio of the gravitational accelerations is essentially unity, the density ratio is the only unknown in Eg. B-4, It is dependent on some of the quantities yet t0 be calculated. Consequently, at this stage of the calculations, a value is assumed that may be corrected later. Theoretically, if all similarity parameters were being matched, the density ratio of any of the fluids present at any temperature could be used. However, in vacuum modeling, since steam density is not scaled, it should not be used. In fact, in selecting characteristic quantities, any quantity that has been compromised appreciably should not be used, except in expressions that correct for such compromises. In general, it will not be necessary to compromise very much, if at all, on the oil and water density and either one may be used. For our example, we have chosen the average oil-density ratio. Typically y(pg) is close to unity, Assuming Vlog) =0.9 for our frst iteration, we have irom Eq, Bed (Bt) = _=Pp)p__ (Po) m= Tyo sy aay ~ OMS (PP) P : sees -(B5) ‘Combining Eqs. B-1 and B-3 with Eq. B-S, Py =0.757 +0.00485p p. 8) Here, pyy is the model pressure corresponding to a prototype pressure pp, where pyy and pp are in psia. In carrying out scaling calculations, it is beneficial to make up a table as shown in Table 8. Here, several prototype pressures covering the range of scaling are tabulated along with their corresponding saturated- steam temperatures, T,, the enthalpy of water at steam temperature (relative to any consistent in Fig. 22 Dimensionless ‘sureltemperature fit. saturation vapor-pres: reference temperature, taken here to be 32°F), and the effective C,A7, where AT is the difference between steam temperature and the initial reservoir temperature, 7;, taken to be 85°F. In practice, the effective C,,A7 is obtained from the difference between the water enthalpies at T, and 7. The table also includes the calculated model pressures with corresponding steam temperatures (obtained from steam tables) and other quantities to be determined. Determination of Model Temperature Sealing For hot-water scaling, the model temperature sealing is arbitrary. One can select the upper and lower model temperatures quite freely as long as a model oil can be found with the appropriate Viscosity/temperature relation. In. scaling steam- recovery methods, one has to treat the problem in a more sophisticated manner. The dimensionless pressure/temperature relation for saturated steam is a constraint on the scaling that determines the op= timum temperature sealing. In general, to obtain the best match between the model and prototype oil-viscosity curves, the model temperature range should be made as large as possible. Usually, this also aids in matching the pressure/temperature relation for saturated steam. ‘Consequently, in the first iteration, the initial model temperature should be taken as low as possible. As a practical limitation, the minimum initial model temperature is about 38°F. Cooling below this point leads to difficulties because of localized freezing of model lines and problems with frosting of the refrigeration system itself. Consequently, the initial model temperature is taken to be (7) = 38°F. -(B7) To scale the temperature, corresponding peratures at only one other point are required. Usually, itis best (0 take a value from the middle of the pressure range for finding the temperature difference ratio because most of the oil will be produced when these temperatures are significant. Further, the lower part of the pressure/temperature relation for saturated steam is difficult to fit and itis, tem- im best to try to fit the middle and the upper parts of the relation, ‘Using midrange prototype and model temperatures in Table 8 corresponding to a pressure of 300 psia yields the following value for the temperature ratio: (ADp _ (T-T))p __ 417-85 (Dg (T=T)) yy 129.738 3.62. ++ (B8) This ratio must be constant over the temperature range to have the proper proportion of energy stored as internal energy. If so, from Eqs. B-7 and B-8, for any prototype temperature, Ty = 0.271 p+ 14.5. (B9) ‘The model values calculated from Eq. B-9, where Tp=(T,) p, are shown in the third column of Table 8, and are designated T,, to denote that they are the calculated values. Check of Pressure/Temperature Saturation Fit ‘The calculated values, T,., do not correspond exactly with the saturation values, 7. In fact, there is no reason that the dimensionless pressure drop and the dimensionless temperature difference for steam at the saturated condition should correspond at all values. From Table 8 it can be seen that for our example case the values match very well for all but the bottom row of the table, corresponding to 50 psia. This is typical of attempts to obtain matches at this low production pressure, and this effect is shown, graphically in Fig. 22. Because of the limitations of minimum model temperatures and pressures, little can be done to improve the overall match. By choosing to match the lower temperatures at the expense of the upper temperatures, even larger errors are introduced. For example, selecting a AT ratio based on the lowest temperature, 7, =101.7°F, would provide a good match only at this point. At other points, the temperatures would be off by as much as 18.9°F. We conclude that for our example problem, better scaling is obtained by allowing the error to occur at the low temperature. Further, it should be noted that the portion of the pressure/temperature saturation relation that is difficult to match is the part pertaining to a prototype pressure of about 100 psia or less. If the minimum prototype pressure (in general, the production pressure) is greater than 100 psia, there usually is no problem in making the match, Calculation of Steam Quality Required The model steam quality is determined from Parameter Il in Table 3. If cap and base rock heating predominates, then f,yy is given by Cw fsby pas, tar (S) (eae (Seas) eC) Ces ( a Gerard )J-3. --e10 SOCIETY OF PETROLEUM ENGINEERS JOURNAL 0.24, ASp = 0.70, fap = 0-80, dng = 0.29, ASy = 0.85, ppCp = pmCys and pepCcp = Pe Cem then for the bottom entry of Table 8, $4.3 (0.80)(924) fu acelin”) | 0.24)0.70) 13 (0.29)(0.85) The other fyyy Values calculated from Eq. B-I1 are shown in Table 8. Calculation of Model Viscosity It generally is important to scale the oil production after steam breakthrough as well as possible. To match the pressure gradient in the steam zone and the oil, the oil viscosity must be sealed according to Parameter II in Table 3, so that How _ (Easy (ese) ( (212) (Bat Mor usp) Spow? Por ‘The values corresponding to a prototype pressure of 400 psia are 117. ....(B1) (B-12) rr 70.082) / 0.0008 \/ 0861 \/ 1.0 felt = (Say) Contes) (aonres) Cos ) 7 19 The remainder of the values calculated are given in Table 8. Calculation of the Time Scale Because of the constraint on the scaling introduced by the heat capacity parameter, there are two possible choices in the selection of a time scale, as indicated by Parameter IV in Table 3. Neither of these is exact unless, of course, 6S is the same in the ‘model and prototype so that the criteria are essen- tially identical. Whenever the temperature distribution is more significant in the cap and base rock than in the reservoir, which frequently is the case in steam drives where the thermal efficiencies are less than 0.5, caprock diffusivity should be used for sealing the time. For steam soaks, the reservoir diffusivity generally should be used. For this example, the caprock and base rock heating is assumed to predominate, so that the time-scale ratio (! eo Ces me) (oat) (o The glass plates that often are © used to simulate caprock and base rock have a considerably lower thermal conductivity (0.5 Btu/hr-ft-"F) than typical prototype values, which are assumed to be 1.2 Btu/br-ft-"F. This lengthens the model time scale ‘over that in which reservoir-like materials such as limestone are used. If glass plates are used, glass beads also should be used in the model pack to provide scaled conduction effects in the reservoir. For our example with glass plates, (oC) = 25.9 Btu/eu fF and (B14) JUNE 1980 (0,¢,) p=25.9 Buu/eu ft- Then, the time scale is M #=( a3) (ass) (as) (525.960 min/s) =24.07 in (B-15) From Eq. B-15 it is seen that 24.07 minutes in the model will correspond with 1 year in the field prototype. (The conversion factor should not be confused with the time seale itself.) Calculation of Model Permeability The model permeability is scaled from Parameter V dees (sey) (B16) ce jzyy/sep is @ function of the temperature, this expression for ky /k p also is temperature dependent however, kyy/kp cannot be a function of the tem perature and a single value must be chosen In selecting a single representative value for the viscosity ratio, we recall that the actual value of this ratio is significant only after steam breakthrough, when steam pressures tend to be low. Further, its value is most significant in the vicinity of the production well. Consequently, the lower portion of Table 3 should be weighted most heavily; for our example, we take the value corresponding to a prototype pressure of 100 psia. If the late-time pressures are expected to be fairly high, but are unknown, some average value may be taken. Thus, using the values previously given and taking the prototype permeability to be 2 dat poo) Bp = (Gar) uaa) a9)" Te) + (gagtages) = 927 dacies. «..B17) Selection of Model 0 At this stage of the calculation, an attempt is made to 1d a model oil whose viscosity is the appropriate multiple of the prototype. The density of this oil is used for iterating with a new density ratio in Eq. B-4. If an appropriate oil cannot be found, the model temperature range may have to be changed to ii crease the temperature range. In making this change, ‘a compromise must be made between having the appropriate temperature distribution (because of the pressure/temperature relation at saturation) within the steam zone and having an appropriate viscosity ratio between model and prototype. Also, an upper. limit for permissible vacuum model pressures of about 12 psia must be considered, Calculation of Injection and Production Rate Seales Both the injection rates and production rates are sealed according to Parameter VI in Table 3: 1 02") (2)- - G18) wy, 1.0 1 30,29) 70.85 m7 (03) (335) Co3a) Coro 110.4 on? nin 1 aa (aa em? /min <7 B/D where the volumetric rates are at standard con- ditions. In assigning values on a per-well basis, the use of partial wells in a model must be taken into account. ++ (B19) Calculation of Water Proportion for Steam Quality To complete the scaling, the proportions of water and steam must be calculated to obtain correct model steam quality. The value depends on the enthalpy of the steam from the steam generator, the temperature of the quality-control water (generally room temperature), the heat loss from lines, and the Quality desired. The enthalpy rate of the two inlet streams that are mixed must equal the required enthalpy rate of the wet steam injected into the model, plus any heat losses. Thus, Woh + Wally = (Wes + Wa) Southey +CyOT) + Qhoss» where W,, and w, are the mass rates of flow of the superheated steam and the aqueous phase added for controlling quality, fi, and hg are the entering -(B-20) m4 superheated steam enthalpy and quality control water enthalpy, and jo, isthe rate of heat loss. For the Eq. B-20 example here, the entering superheated steam is assumed to be at 250°F, the quality control water 10 be at 75°F, and Qo, to be zero. Rearranging Eq. B-20 and” noting that Whur = Was Was Where Wy, is the total mass rate of injected water, yields fgg Seley ~CwAT= (toss! ¥0) ut ssh Fora model pressure of | psia, Wg _ 1173 =(0.117)1,041.6)—54.3-0 Ce 117343 =0.882. ... = (B21) (B22) Values of W/W; for other model pressures are shown in Table 8. SI Metric Conversion Factors acre x 4.046 873 E+03 bbl x 1.589 873 E-O1 Bu x 1.055 056 E+00 cp x 1.0" E-03 cut x 2.831 685 E-02 cuin, x 1,638 706 E+01 degree F (“F~32)/1.8 = ft x 3.048" B01 in. x 2.54 E+00 Ibm x 4.535 924 E-O1 mL x 1 Psi x 6.894 757 E+00 sqft x 9.290 304° E02 segmeehnPow Serene eset eS ft Sa SOCIETY OF PETROLEUM ENGINEERS JOURNAL

You might also like