Gates2007.pdf

Steam-Injection Strategy and Energetics of Steam-Assisted Gravity Drainage Ian D. Gates, U. of Calgary; Joseph Kenny and Ivan L. Hernandez-Hdez, ATECH Application Technology Ltd.; and Gary L. Bunio, Paramount Resources Ltd.

Summary Steam-assisted gravity drainage (SAGD) is being operated by several operators in Athabasca and Cold Lake reservoirs in Central and Northern Alberta, Canada. In this process, steam, injected into a horizontal well, flows outward, then contacts and loses its latent heat to bitumen at the edge of a depletion chamber. As a consequence, the viscosity of bitumen falls, its mobility rises, and it flows under gravity toward a horizontal production well located several meters below and parallel to the injection well. Despite many pilots and commercial operations, it remains unclear how to optimally operate SAGD. This is especially the case in reservoirs with a top-gas zone in which pilot data are nearly nonexistent. In this study, a steam-chamber operating strategy is determined that leads to optimum oil recovery for a minimum cumulative steamto-oil ratio (SOR) in a top-gas reservoir. These findings were established from extensive reservoir-simulation runs that were based on a detailed geostatistically generated static reservoir model. The strategy devised uses a high initial chamber injection rate and pressure prior to chamber contact with the top gas. Subsequent to breakthrough of the chamber into the gas-cap zone, the chamber injection rates are lowered to balance pressures with the top gas and avoid (or at least minimize) convective heat losses of steam to the top-gas zone. The results are also analyzed by examining the energetics of SAGD. Introduction A cross-section of the SAGD process is displayed in Fig. 1. Steam is injected into the formation through a horizontal well. In Fig. 1, the wells are portrayed as points that extend into the page. Around and above the injection well, a steam-depletion chamber grows. At the edge of the chamber, heated bitumen and (steam) condensate flow under the action of gravity to a production well typically placed between 5 and 10 m below and substantially parallel to the injection well. Usually, the production well is located several meters above the base of pay. In industrial practice (Singhal et al. 1998; Komery et al. 1999), injection and production well lengths are typically between 500 and 1000 m. Because the steam chamber operates at saturation conditions, the injection pressure controls the operating temperature of SAGD. SAGD has been piloted extensively in Athabasca and Cold Lake reservoirs in Alberta (Komery et al. 1999; Butler 1997; Kisman and Yeung 1995; Ito and Suzuki 1999; Ito et al. 2004; Edmunds and Chhina 2001; Suggett et al. 2000; Siu et al. 1991; AED 2004) and is being used as a commercial technology to recover bitumen in several Athabasca reservoirs (Yee and Stroich 2004). These pilots and commercial operations have demonstrated that SAGD is technically effective, but it has not been fully established whether its operating conditions are at optimum values. This is especially the case in reservoirs in contact with gas or water zones where the optimum operating strategy remains unclear. The variability of the cumulative injected-steam (expressed in cold water equivalents, or CWE) to produced-oil ratio (cSOR) shows that some SAGD well pairs operate fairly efficiently (with cSOR be-

Copyright © 2007 Society of Petroleum Engineers This paper (SPE 97742) was first presented at the 2005 SPE International Thermal Operations and Heavy Oil Symposium, Calgary, 1–3 November, and revised for publication. Original manuscript received for review 29 August 2005. Revised manuscript received 12 August 2006. Paper peer approved 10 October 2006.

February 2007 SPE Reservoir Evaluation & Engineering

tween 2 and 3), whereas others operate at much greater cSOR (up to 10 and higher) (Yee and Stroich 2004). Higher cSOR means that more steam is being used per unit volume bitumen produced. The higher the steam usage, the greater the amount of natural gas combusted, and the less economic the process. One key control variable in SAGD is the temperature difference between the injected steam and the produced fluids. This value, known as the subcool, is typically maintained in a form of steamtrap control between 15 and 30°C (Ito and Suzuki 1999). The subcool is being used as a surrogate variable instead of the height of liquid above the production well. The liquid pool above the production well prevents flow of injected steam directly from the injection well to the production well, thus promoting injected steam to the outer regions of the depletion chamber and enabling delivery of its latent heat to the bitumen. The value of the optimum steamtrap subcool temperature difference and how the operating pressure impacts the optimum subcool value remains unclear. It also remains unclear how the subcool should be controlled in heterogeneous reservoirs that have top gas. Improvements to the SAGD operating strategy have been studied by many authors (Kisman and Yeung 1995; Ito and Suzuki 1999; Ito et al. 2004; Edmunds and Chhina 2001; Suggett et al. 2000; Siu et al. 1991; AED 2004; Yee and Stroich 2004; Gates and Chakrabarty 2005). Some papers have looked at the impact of operating pressure on the performance of SAGD. Kisman and Yeung (1995), by simulating SAGD in the Burnt Lake oil sands, showed that the lower the operating pressure, the lower the oil rate and cSOR. In Hangingstone reservoir, Ito and Suzuki (1999) and Ito et al. (2004) showed in a series of sensitivity simulations that the optimal subcool lies between 30 and 40°C. Edmunds and Chhina (2001) showed that the economics of SAGD are more sensitive to the cSOR than the oil production rate and that, if the operating pressure were constant, then the optimum operating pressure could be as low as 400 kPa for typical McMurray reservoirs. Gates and Chakrabarty (2005) used a genetic algorithm together with a commercial thermal reservoir simulator to optimize the cSOR by altering the steam-injection strategy in a generic McMurray reservoir. They found that the cSOR can be improved by operating SAGD with a profile of steam-injection pressures throughout the life of the process over that of constant injection pressure. This study was limited to 2D and is therefore an idealization of the reservoir description. There are have been limited studies of SAGD in top-gas reservoirs (Good et al. 1997; Nasr et al. 2003; Law et al. 2003a, 2003b). Good et al. (1997), in a sensitivity study to assess the performance of SAGD in top-gas bitumen reservoirs, found that the larger the gas cap, the worse the economics of the process and the lower the ultimate bitumen recovery. They also examined the impact of overburden thickness, net pay, vertical and horizontal permeabilities, and initial oil viscosity on performance. The simulation model was 2D. In many of the sensitivity cases, significant oscillations of the fluid rates and well bottomhole pressures are evident in the simulation results. Law et al. (2003a) showed in laboratory-scale experiments that oil moved from the oil-rich zone into the top-gas zone and that the higher the pressure difference between the oil and top-gas zones, the greater the oil and steam flow into the top-gas zone and the higher the SOR. Law et al. (2003b) conducted field-scale reservoir simulations of SAGD in a reservoir with a top-water thief zone and showed that the higher 19

Fig. 1—Cross section of the SAGD process.

the pressure difference between the steam chamber and the topwater zone, the lower the production rate and recovery and the higher the cSOR. Nasr et al. (2003) compared results from SAGD elemental physical experiments in top-gas and top-water model reservoirs (3 D sand in cylinder arrangement of dimensions: 30 cm in diameter by 60 cm tall). They showed that the rate of heat transfer was less with the top-gas reservoir. This is explained by the lower thermal conductivity of the top gas compared to top water. The oil recovery was lower in the top-gas experiment than that in the top-water one. This was found to be caused by greater flow of oil into the top-gas zone (16% of initial oil) than what flowed in the top-water zone (10% of initial oil). The influence of capillary pressure, especially because this could be important over the dimensions of the physical model (and not at field scale), was not explored in the paper. In this study, a sequence of SAGD simulations were conducted on a reservoir with a detailed heterogeneous geological description obtained from geostatistical analysis of log and core data. After more than 100 simulations, an optimized steam-injection strategy was devised that produced a reasonable cSOR for design of a SAGD operation in McMurray reservoir with a top-gas zone. In addition, the energetics of SAGD are examined by evaluating the flowing steam quality in the steam chamber. Geological Model To allow the assessment of geological variation and uncertainty on the SAGD process, a detailed, statistically based static geological description was prepared. The overall domain of the parent geo-

20

logical model covers 32 sections in the vicinity of Paramount Resources’ oil sands lease at Surmount described by Robinson et al. (2005). Petrophysical interpretations and picks for seven horizons were available at 33 wells within the parent geological domain and were mapped from the well inputs by using a local B-spline algorithm in Roxar’s IRAP RMS geostatistical package (Roxar 2004). Of the 33 wells, 14 of them were within the subdomain that was targeted for reservoir simulation. From these 14 wells, facies logs were prepared to populate the geological model (Roxar 2004). A map showing the well locations and an example of a type log are displayed in Robinson et al. (2005). A stacked rectangular grid was used to build the geostatistical model. In the east-to-west direction, 136 columns, each 50 m wide, were used, giving a total width equal to 6800 m. In the north-tosouth direction, 259 rows, each 50 m wide, were used to give a total length of 12 950 m. In the vertical direction, 95 layers of variable thickness, most of them under 1 m thickness, were used. The total cell count for the geological grid was 3,346,280. The facies logs distinguished eight rock types that were identified as follows: • Facies 1: Shoreface sand. • Facies 2: Muddy marine sand. • Facies 3: Mudstone. • Facies 4: Mud-dominated heterolithic strata. • Facies 5: Sand-dominated heterolithic strata. • Facies 6: Sandstone. • Facies 7: Breccia. • Facies 8: Mudstone filled. Petrophysical interpretations provided logs of Sg, So, Sw, ␾, and kh for the 33 wells within the geological-model parent domain. From a detailed review and synthesis of core, core analysis, logs, and a history match of a nearby McMurray SAGD pilot, correlations of permeability vs. porosity and vertical to horizontal permeability ratio (kv/kh) were established for each facies in the model. The correlations are listed in Table 1. The facies-specific permeability/ porosity transforms were used to generate well permeability logs. For each facies, the vertical to horizontal permeability ratio for the model was set to 0.65, 0.055, 0.055, 0.055, 0.300, 0.500, 0.500, and 0.005 for facies 1, 2, 3, 4, 5, 6, 7, and 8, respectively. As described in Robinson et al. (2005), the facies and petrophysical well logs were entered into Roxar’s IRAP RMS geostatistical modeling package and upscaled to populate the gridblocks intersected by the wells with the following suite of parameters: facies, Sg, So, Sw, ␾, kh, and kv. The grid was kept relatively fine in the vertical direction, and the distribution of layer thickness ensured that the blocked well data represented the actual log data

February 2007 SPE Reservoir Evaluation & Engineering

central elevations of the model. The porosity and horizontal and vertical permeability distributions are displayed in Robinson et al. (2005). Fig. 2—Facies distribution in the geological model.

sufficiently. To populate the facies at gridblocks away from the wellbores, a sequential indicator simulation (Computer Modelling Group 2004), conditioned on the facies vertical proportion curve, was performed. Any number of equiprobable facies realizations may be generated from this data set. The facies distribution through the reservoir for one realization is displayed in Fig. 2. After review by team geologists, the geological model was considered to be a reasonable reflection of the geological environment. To populate the permeabilities, porosities, and saturation distributions in the regions between the well locations, sequential Gaussian simulation (SGS) (Roxar 2004) was used. The SGS that was used to populate gas, oil, and water saturations in the model was run independently to ensure that the underlying statistics describing these parameters were honored in the model. The saturations were then normalized to ensure that Sg+So+Sw⳱1 at every gridblock. This adjustment changed the fluid saturations by less than 0.005 saturation units. To obtain porosity and permeability distributions between the well locations, an SGS was conducted with upper and lower bound cutoffs applied to the wellblock porosity and permeability distributions. Tables 2 and 3 list the minimum and maximum cutoffs, average values, and standard deviation of the porosity and permeability distributions. As described in Robinson et al. (2005), 100 equiprobable facies, porosity, permeability, and saturation realizations were constructed. A comparison revealed that there were small variations in volumes among the 100 realizations, and a realization near the center of the population was chosen to construct the working geological model from which a reservoir-simulation model could be extracted. The gas- and oil-saturation distributions within the model are shown in Fig. 3. The gas saturation is concentrated in the upper marine sands, whereas the highest oil saturations are located in the

February 2007 SPE Reservoir Evaluation & Engineering

Reservoir-Simulation Model The next step of the workflow was to upscale, extract, and import the geological model into the reservoir simulator. The reservoir simulation of the SAGD process was conducted with Computer Modelling Group’s (2004) thermal reservoir simulator STARSTM. As presented in Robinson et al. (2005), an upscaled geological description of Section 30 was extracted consisting of 516,096 cells from the original geological parent model. From this upscaled model, within the CMG preprocessor, a subdomain able to accommodate two 750-m well pairs, with 500 m in the east/west (X⳱lateral) direction by 950 m in the north/south (Y⳱downwell) direction, was extracted. In the downwell direction, the subdomain was tessellated into 12 gridblocks. The total number of gridblocks equals 74,592. Fig. 4 displays a grid of the reservoir-simulation model two-thirds of the way down the wells; the left well pair is referred to as LP, whereas the right well pair is referred to as RP. Fig. 4 displays a cross section of the reservoir-simulation grid that shows the locations of the left and right well pairs as well as wells that were inserted into the gas cap to mimic the continuity of the gas cap beyond the model. All wells are modeled as source/ sink wells; that is, no discretized wellbore model was used. The gas-cap wells were constrained to constant bottomhole pressure equal to the gas-cap pressure (they will be referred to as pressuremaintenance wells). The grid blocks in the east/west direction were refined to ensure that the maximum gridblock width (for 50 m on each side of the left and right well pairs) was 3.845 m. The interwell-pair spacing is 200 m. Fig. 5 displays porosity and horizontal permeability distributions of the reservoir-simulation model at planes along the well pairs and two lateral cross sections. The black lines in the downwell planes indicate the locations of the well pairs. In the central region, the porosity ranges from 0.27 to 0.38, and it is improving from south to north. Similarly, the horizontal permeability mainly lies between 2 and 6 D in the central regions of the reservoir, with better permeability in the north.

21

Fig. 4—Cross section of the dual-well-pair reservoir-simulation model grid.

In the top-gas zone, the maximum gas saturation is approximately 0.82, with the remainder of the pore space containing water. Initially, the pressure of the gas cap is about 1050 kPa (the pressure equaled 936 and 1631 kPa at the top and bottom of the reservoir model, respectively). The pressure-maintenance wells located in the gas cap are set to produce fluids if the pressure exceeds 1075 kPa in order to mimic a gas-cap zone that extends beyond the

Fig. 3—Gas- and oil-saturation distributions in the geological model.

Fig. 6 shows cross sections of the gas, oil, and water saturations in the reservoir-simulation model. Consistent with the geological model, the gas cap, typically 3 to 4 m thick, is located throughout the model.

Fig. 5—Dual-well-pair reservoir-simulation model: porosity and horizontal permeability distributions (cross sections along well pairs and two lateral locations downwell). 22

Fig. 6—Dual-well-pair reservoir-simulation model: gas, oil, and water saturations (cross sections along well pairs and two lateral locations downwell). February 2007 SPE Reservoir Evaluation & Engineering

model domain. It is anticipated that in the case with an established depletion chamber in contact with the top-gas zone, if the steaminjection pressure is too large, steam will be diverted from the steam chamber into the top-gas zone, raising the pressure there. Then, steam and gas will flow out into the gas cap and through the gas-cap pressure-maintenance wells. The oil-saturation distribution displayed in Fig. 6 reveals that there is a central region of the reservoir with relatively high oil saturation. The average thickness of the oil-rich zone (>0.7 oil saturation) is approximately 20 m. In some parts of the reservoir, it is as high as 26 m, and in others it drops as low as 14 m. The production wells of each well pair are located just a few meters above the bottom of the oil-rich zones. The injection wells are located 5 m above the production wells. The water-saturation distribution shown in Fig. 6 reveals that there are relatively high water-saturation regions at the top and bottom of the model. The region at the top is above the gas cap and consists of tight muddy marine sand reservoir rock (Facies 2). The porosity and permeability of this facies are both low (see Tables 2 and 3). The high water saturation at the bottom of the reservoir is in Facies 6, and the water is relatively mobile given the permeability of this part of the reservoir. Table 4 summarizes additional heat-loss parameters, fluid, and rock/fluid properties. The reservoir-simulation model fluid components consisted of bitumen, water, and solution gas. The bubblepoint pressure in the model was taken to be 1000 kPa. Because most of the model is above this pressure, the initial solution gas/oil ratio was constant at 4.2 m3/m3 throughout the oil layer. The K-value relationship used in the model is listed in Table 4. The bitumen viscosity was similar to Mehrotra and Svrcek’s (1986) correlation for Athabasca bitumen. The oil/water and gas/liquid relative permeability curves were obtained from a detailed history match of a McMurray SAGD pilot. The endpoints are listed in Table 4. Well Constraints. At the injection wells, the steam-injection pressure is constrained to a maximum bottomhole pressure. The steam quality at sandface equaled 0.8. This is typical of values found after heat losses in the injection wellbore are taken into account. At the production wells, the CMG (Computer Modelling Group 2004) steam-trap constraint was used with a 5°C setting. The CMG steam-trap control algorithm does not impose the temperature difference between injection and production wells. Rather, a 5°C subcool in CMG’s algorithm means that the bottomhole pressure

in the wellblock is set corresponding to the pressure 5°C above the saturation temperature of the gridblock. This means that no live steam can be produced from the well. The subcool temperature difference often referred to from field data is the difference between the steam injection and the produced-fluid temperature. Model Initialization. To model steam circulation, line heaters were positioned in the locations of the injection and production wells of each well pair. The heating rate corresponded to the heat delivered by 250 m3/d CWE of 0.8 quality steam. In the location of the production wells, the wells were opened with bottomhole pressure equal to the initial reservoir pressure. The reason for this is to relieve pressure buildup caused by thermal expansion of the fluids near the wellbore. Similarly, temporary production wells were inserted into the same locations as the steam injectors so that pressure was relieved at the injection-well locations as the reservoir fluids near the wellbore heat up. The circulation period lasted 3 months. When the well pairs were changed to SAGD mode, the line heaters were turned off, the temporary production wells positioned in the injection-well locations were removed, and steam injection commenced at the target rate or pressure. Results: Optimization of Performance: From Dual-Well-Pair to Single-Well-Pair Models The simulation runs of the dual-well-pair model took up to 20 hours to complete a 12-year forecast with a 2.45-GHz dualprocessor computer workstation with parallel-enabled STARS. It was recognized early in the study that to carry out a large number of simulations would be prohibitive and would not be possible on the five available workstations. To test the sensitivity to grid, the grid near the wellbores (10 m on each side, 8 m above, and 4 m below) was tripled in the crosswell (X) direction (one gridblock refined to three gridblocks) and each gridblock in the downwell (Y) direction was subdivided into five smaller gridblocks. The results (both production rates from the wells and temperature and pressure in locations around the wellbores) were found to change by less than 1%. To obtain simulation run times in reasonable run times without compromising the geological description and physics of the SAGD process, two single-well-pair models, identified as right pair (RP) and left pair (LP), were created. Each single-well-pair model consists of half the dual-well-pair model. Each half has a 250 m by 950 m areal footprint. There are 37,296 blocks in each of the single-well-pair models, and 12-year forecast runs lasted under 9 hours on a 2.45-GHz dual processor workstation with parallelenabled STARS. It was decided that first, the operating strategy of each single-well-pair model would be optimized, and second, the individually optimized operating strategies would be applied in the dual-well-pair model. It was recognized that because of interwellpair communication, the results of the single-well-pair operating strategies would have to be adjusted once introduced into the dualwell-pair model. However, it was expected that given the presence of the gas cap, the operating strategy would have to be gentle on the reservoir (otherwise, excessive steam would be lost to the gas cap), and, therefore, communication issues would not be too hard to resolve. The same circulation preheat strategy as was described for the dual-well-pair model was used for the single-well-pair models. Many cases were run or partially run to improve the cSOR after 6 and 12 years of SAGD operation. These runs were in sequence and in parallel, and the operating strategy was modified after carefully reviewing and analyzing the results of each run to further improve the cSOR. Constant Pressure Injection. First, the results for constant steaminjection pressure at 2000 kPa will be presented. Figs. 7 and 8 show plots of the injection rate and pressure and production rates, cumulative volumes produced, and cSOR for the LP well pair, respectively. Figs. 9 and 10 show the same plots for the RP well pair. The results in Figs. 8 and 10 reveal high cSOR profiles that are the result of excessive steam losses from the steam chamber to the

February 2007 SPE Reservoir Evaluation & Engineering

23

Fig. 7—Steam-injection strategy in the LP well pair operated at constant 2000-kPa injection pressure.

Fig. 8—Production rates, cumulative volumes, and cSOR of the LP well pair operated at constant 2000-kPa injection pressure.

gas cap. Because steam rates are high, very little latent heat is being delivered to the bitumen, so bitumen rates are low. Optimized Steam Injection. With manual optimization, improved operating strategies were determined for both LP and RP well pairs. The strategies for both well pairs are similar and comprise a high steam-injection pressure until each steam chamber establishes contact with the top gas. High injection pressure implies a relatively high saturation temperature that leads to favorable bitumen viscosities in the early stages of SAGD. After the top gas is encountered, the top-gas pressure dictates the steam-chamber operating pressure, which, according to the optimized strategy, is maintained constant thereafter at or just below the gas-cap pressure. This ensures that convective losses of steam after breakthrough to the top gas are avoided or at least minimized. This operating strategy is consistent with the results of Law et al. (2003a, 2003b) and Gates and Chakrabarty (2005). In uniform steam-injection pressure simulations, the northern end of the LP reservoir was depleted more rapidly than the southern part of the reservoir. This is because the geology along the LP well pair has a large contrast in reservoir quality going from toe to heel. The toe section of the injection well has better vertical permeability than the heel section and, consequently, breakthrough to the top-gas zone happens quickly and thus makes favorable thermal management of the steam chamber more problematic. On the other hand, the reservoir quality along the well pair of the RP model is more uniform and makes the breakthrough time more uniform along the injection well, which makes favorable thermal management before and after breakthrough to the gas cap easier. To promote more-uniform formation of the depletion chamber in

the reservoir, steam injection into the toe of the LP and RP wells was stopped periodically. Without steam-placement control, injected steam takes the path of least resistance to a better-quality reservoir at the toe and bypasses a reservoir of poorer quality at the heel. By introducing control of steam placement, steam is concentrated at the heel section at higher pressures for a longer time, so the steam chamber grows more uniformly at the heel. Also, injection rates and pressures at the toe section are lowered to delay the onset of the chamber contacting the top-gas zone. Robinson et al. (2005) discuss steam-placement control in more detail. Strategies to control steam placement might include a combination of packers and chokes along the horizontal section and/or the use of limited-entry perforations as described by Boone et al. (1998), as well as any inflow-control devices. Flexibility will be key in the design of the injection string because the geology along the injector will dictate the optimal steam-placement requirements. Pressure and temperature monitoring along the steam chamber will be essential to effective management. Fig. 11 displays the injection-well constraints of the optimized operating strategy over the 12-year forecast period for the LP well pair. The plot reveals that over the first 2 years, the injection constraint was a maximum bottomhole-pressure limitation, and beyond that, the constraint was the steam-injection rate that was lowered in steps until the end of the forecast. The production rates and associated cumulative volumes of gas, oil and water, and cSOR over the 12-year forecast period for the LP well pair are presented in Fig. 12. In the first year of the process, the steam-injection pressure is relatively high at 1800 kPa. The injection pressure is then lowered

Fig. 9—Steam-injection strategy in the RP well pair operated at constant 2000-kPa injection pressure.

Fig. 10—Production rates, cumulative volumes, and cSOR of the RP well pair operated at constant 2000-kPa injection pressure.

24

February 2007 SPE Reservoir Evaluation & Engineering

Fig. 11—Steam-injection strategy in the optimized LP well pair.

to 1100 kPa, and thereafter, in the rate-constrained period, the pressure remains near 1000 kPa. The maximum steam-injection rate is short-lived at 175 CWE m3/d. Over the majority of the process, the steam-injection rate ranges from 100 to 150 CWE m3/d. Fig. 12 shows that the cSOR settles after an initial transient period after the onset of SAGD to approximately 2.6 m3/m3. The shape of the bitumen production-rate profile is similar to SAGD field data from Athabasca reservoirs. The field rate hovers just above 50 m3/d for most of the life of the process, and after 8 years of SAGD operation, it begins to decline. With the exception of the first year, throughout the process, more water is produced than injected. This indicates that water from the formation is being produced. Fig. 13 displays the injection-well constraints of the optimized operating strategy for the RP well pair. The production rates and cumulative volumes of oil and water and cSOR over the 12-year forecast period for the RP well pair are presented in Fig. 14. Similar to the LP well pair, in the RP well pair over the first 2 years, the injection constraint was a maximum bottomholepressure limitation, and beyond that, the constraint was the steaminjection rate, which was adjusted in a downward trend until the end of the forecast. During most of the life of the RP well pair, the pressure was roughly 1000 kPa. The rate profiles for the RP well pair presented in Fig. 14 are similar to LP well pair. The cSOR profile passes through an initial transient period and then evolves to a uniform value of approximately 2.4 m3/m3. The bitumen-rate profile has a typical shape and hovers at over 60 m3/d for most of the life of the well pair. As the steaminjection rate declines, the bitumen rate also falls. Similar to the LP well pair, throughout the majority of the life of the RP well pair, more water is produced than is injected into the reservoir. This means that formation water is being produced from the reservoir.

Fig. 13—Steam-injection strategy in the optimized RP well pair. February 2007 SPE Reservoir Evaluation & Engineering

Fig. 12—Production rates, cumulative volumes, and cSOR of the optimized strategy in the LP well pair.

Compared to the constant pressure-injection cases described above, the cSOR of the optimized cases are significantly improved. In the constant-pressure case, the cSOR after 6 and 12 years are 188 and 533 m3/m3 in the LP well pair and 96 and 259 m3/m3 in the RP well pair, respectively. The normalized average bitumen production rates (normalized against the well length⳱750 m) of the constant-pressure cases are 0.0077 and 0 m3/d/m in the LP well pair and 0.025 and 0.00067 m3/d/m in the RP well pair at 6 and 12 years, respectively. In the optimized LP well pair, the cSOR after 6 and 12 years equals 2.6 m3/m3 at both times. The normalized average bitumen production rates at 6 and 12 years are 0.068 and 0.067 m3/d/m, respectively. In the optimized RP well pair, the cSOR is 2.4 m3/m3 at both 6 and 12 years. The normalized average bitumen rate is 0.071 and 0.069 m3/d/m at 6 and 12 years, respectively. The reason for improved performance in the optimized strategies is that after the steam chamber contacts the top-gas zone, the injection pressure drops to values below that of the gas-cap pressure. As a consequence, steam does not invade and penetrate the gas cap and is not lost from the depletion chamber. This implies that the steam’s latent heat is more directly transferred to the bitumen at the edges of the chamber than lost to the gas zone and overburden. Also, as the pressure falls, the saturation temperature falls, and some of the invested heat in the formation and overburden rock is harvested back to the chamber fluids. Figs. 15 through 17 show the sequence of temperature, oil saturation, and flowing steam-quality cross sections roughly twothirds downwell for the LP well pair, respectively. The LP well

Fig. 14—Production rates, cumulative volumes, and cSOR of the optimized strategy in the RP well pair. 25

Fig. 15—Temperature distribution in a section two-thirds downwell of the LP well pair (optimized strategy).

Fig. 16—Oil-saturation distribution in a section two-thirds downwell of the LP well pair (optimized strategy).

pair and gas-cap pressure-maintenance wells are displayed. SAGD mode starts at 2005-04-01 after circulation. The temperature and oil- and gas-saturation distributions are displayed in Figs. 18 through 20 for the RP well pair. The flowing steam quality is determined by calculating the volume of mobile water in the vapor phase, converting it to a mass of water in the vapor phase, and dividing it by the total mass of mobile water (both vapor and liquid) in the gridblock. This is a novel way to visualize the steam chamber in SAGD, where heat transfer is viewed as a change in the flowing steam quality. Because temperature and pressure are nearly constant in SAGD, the

flowing steam quality provides a means to examine convective heat transfer in the steam chamber. At the gridblock above the injection wellblock, the flowing steam quality is higher than the injected steam quality. The reason for this is because there is a separation effect as the liquid water phase in the injection stream falls under gravity to the region below the injection well, whereas the vapor rises into the steam chamber. In the region between the wells, as a consequence of this mechanism, the flowing steam quality is relatively low because the liquid water content, derived from injected liquid water, is relatively high in this region. Above the injection well, the flowing

26

February 2007 SPE Reservoir Evaluation & Engineering

Fig. 17—Flowing steam-quality distribution in a section twothirds downwell of the LP well pair (optimized strategy).

steam quality forms a nearly uniform radial profile moving away from the injection well. Fig. 18 displays the temperature distributions of the RP well pair roughly two-thirds downwell. A comparison with Fig. 15 shows that the RP well pair operates at slightly lower temperature than the LP well pair and that the thermal zone grows more in the vertical direction in the LP well pair than it does in the RP well pair. Fig. 20 shows how the RP well pair gas-cap zone and steam chamber interact as the process evolves. The visualizations, especially at 2008-04-01, reveal that the gas zone has thickened just above the chamber, most likely because heated oil just above the February 2007 SPE Reservoir Evaluation & Engineering

Fig. 18—Temperature distribution in a section two-thirds downwell of the RP well pair (optimized strategy).

chamber is flowing downward under gravity and gas has fingered up to the gas zone from the chamber. At this particular section of the reservoir, the chamber reaches the top-gas zone between 3 and 4 years after SAGD starts. Results From the Dual-Well-Pair Model After the individual LP and RP well pairs were optimized, the two operating strategies were integrated into the dual-well-pair model. Fig. 21 displays a comparison between the production rates and cSOR calculated from the LP well pair in the dual-well-pair model vs. the single LP well-pair model. 27

Fig. 19—Oil-saturation distribution in a section two-thirds downwell of the RP well pair (optimized strategy).

Fig. 20—Gas-saturation distribution in a section two-thirds downwell of the RP well pair (optimized strategy).

Fig. 22 shows a comparison between the production rates and cSOR calculated from the RP well pair in the dual-well-pair model vs. the single RP well-pair model. The results show that communication occurs between the two well pairs after approximately 1 year of SAGD mode. This communication is in the form of the pressure fields interacting from the two well pairs. However, fluid communication between the well pairs is small, so differences of the rate profiles between the singlewell-pair and dual-well-pair models are not significant. The results of the SAGD models demonstrate how the steam chambers grow to the top-gas zone as SAGD proceeds in each well pair. Because of geological differences, the steam chamber inter-

acts and reaches the top-gas zone first with the LP well pair compared to the RP well pair. The oil production rates from each of the well pairs are similar, but the RP well pair has slightly lower cSOR than that in the LP well pair. The reason for this is explained by the earlier interaction of the left steam chamber with the top-gas zone. After the steam chamber reaches the top-gas zone, steam (i.e., latent heat) is delivered to the top gas instead of being delivered entirely to bitumen. The gas-saturation distributions in Fig. 23 reveal that the RP well-pair chamber, up to 2010, was growing faster in the lateral (horizontal) direction than in the vertical direction. A comparison of the cSOR profiles in Figs. 20 and 21 shows that the cSOR of the

28

February 2007 SPE Reservoir Evaluation & Engineering

Fig. 21—Production rates and cSOR of the single LP wellpair model vs. the result for the LP well pair from the dual-wellpair model.

RP well pair was slightly lower than that of the LP well pair. This comparison reflects the higher thermal efficiency achieved when breakthrough to the top gas is delayed. These results indicate that after the chamber is near the top of the oil pay, it is advantageous to lower the rate of vertical growth of the chamber to prevent penetration into the top-gas zone. At this point, it would be advantageous to promote, if possible, lateral growth of the steam chamber. Steamtrap Control and Subcool As described above, at the production wells, a 5°C subcool in the CMG algorithm means that the bottomhole pressure in the wellblock is set corresponding to the pressure 5°C above the saturation temperature of the gridblock, and, therefore, no live steam can be produced from the well. In the field, the steamtrap subcool temperature difference is defined as the difference between the steaminjection temperature and the produced-fluids temperature. Fig. 24 displays the temperature difference between the injection and production wells for the LP and RP well pairs in a section roughly two-thirds downwell of the dual-well-pair model. In the first year and a half of SAGD, the subcools of both well pairs oscillate and achieve high values. From Figs. 21 and 22, the cSOR achieves its maximum values in this time interval. To recall, the optimized operating strategy had relatively high-pressure steam injection in the first year at 1800 kPa and was lowered to approximately 1100 kPa thereafter. After 1 year of SAGD, the subcool in the LP well pair climbed to over 60°C for a couple of months. In the same time interval, the RP well pair subcool

Fig. 23—Gas-saturation distribution in a section two-thirds downwell of the dual-well-pair model (optimized strategy obtained from single-well-pair models).

Fig. 22—Production rates and cSOR of the single RP wellpair model vs. the result for the RP well pair from the dual-wellpair model. February 2007 SPE Reservoir Evaluation & Engineering

dropped to zero for several months. When the injection pressure was lowered after 1 year to 1100 kPa, the LP well pair responded by a jump in liquid production, which then led to cooler liquid being produced from the LP producer and, consequently, a higher subcool. After the system stabilized, the subcool reached the steady value between 20 and 30°C, which persisted throughout the remainder of the process. In the RP well pair, after the injection pressure was reduced to 1200 kPa, liquid production rates did not rise immediately but initially remained roughly constant, and, consequently, the produced-fluids temperature remained roughly the 29

Fig. 24—Subcool temperature difference in the LP and RP well pairs obtained from the dual-well-pair model.

same. However, because the injection pressure was reduced, its temperature also fell, and the subcool became nearly zero. Beyond approximately 2 years, the subcool of the RP well pair also stabilized between 20 and 30°C. The differences in subcool behavior revealed by the simulation reflect the difference between the geology at each of the well pairs and its impact on the growth of the steam chambers. From the gas-saturation distributions displayed in Fig. 25, the LP well pair has a smaller steam chamber than that in the RP well pair after 1 year of SAGD. Energetics of SAGD As has been described above, the optimum steam-chamber operating strategy used in all the cases we have reviewed has a common methodology. This methodology calls for maintaining a high steam-chamber pressure early in the SAGD process. The higher steam-chamber pressures lead to faster chamber growth and higher chamber temperatures. This situation in turn leads to a favorably higher oil production profile. In general, the higher the chamber temperature, the higher the oil production. Eventually, the steam chamber will contact the top gas, after which the steam-chamber operating pressure is dropped in line with the prevailing top-gas pressure. In this section, the relative roles of vertical and horizontal heat transfer are investigated. The analysis here will focus on the LP well-pair model. Figs. 26 through 28 display the temperature, pressure, and flowing steam quality along a vertical plane intersecting the wells roughly twothirds downwell. The injection well is located at 37 m, whereas the production well is located at approximately 42 m (the distance equal to 0 m is located at the top of the reservoir). The plots in Figs. 26 through 28 are good representatives of the profiles at other locations along the well pair. Beyond 2008, the profiles at each time overlay each other, suggesting that the steam chamber is under quasisteady-state conditions. Fig. 26 shows that the temperature across the steam chamber is roughly constant at 177°C. The temperature profiles above the chamber edge indicate conductive heat transfer into the reservoir rock directly above the steam chamber. In the period between 2006 and 2008, the steam chamber grows roughly 20 m in the vertical direction, indicating a vertical rise rate of approximately 2.7 cm/d. Given the temperature profile and due to the chamber being at saturation conditions, beyond 2008, the pressure is largely constant across the steam chamber. Thus, the temperature and pressure gradients in the steam chamber are very small. Fig. 28 displays the flowing steam-quality profiles. At the start of SAGD model in 2005-04-01, the quality profile exhibits two peaks, one at the injection well elevation and the other at the production well location. The reason for these two peaks is that circulation was occurring in both wells and it was sufficiently hot that there was steam in the reservoir at the well locations. After a quasisteady state has evolved, beyond about 2008, the vertical quality profile has a roughly constant slope equal to 0.016 quality 30

Fig. 25—Oil-saturation distribution in a section two-thirds downwell of the dual-well-pair model (optimized strategy obtained from single-well-pair models).

units per meter. Beyond the steam chamber, the flow steam quality falls rapidly to zero. This is the location where the steam is releasing all of its latent heat to the oil sand at the edges of the steam chamber. Figs. 29 through 31 display the temperature, pressure, and flowing steam quality along a horizontal plane located 2 m above the injection well. The injection and production wells are located at 92 m. The plots in Figs. 29 through 31 are good representatives of the profiles at other locations across the domain. The temperature profiles in Fig. 29 demonstrate the lateral growth rate of the steam chamber. Beyond 2008, the lateral growth February 2007 SPE Reservoir Evaluation & Engineering

Fig. 26—Temperature through a vertical profile intersecting the LP well pair.

rate is on average 0.7 cm/d. This is roughly 3.8 times less than the initial vertical growth rate. The temperature in the steam chamber is roughly constant at 177°C. As expected, owing to saturation conditions in the steam chamber, Fig. 30 confirms that the pressure is also nearly constant within the steam chamber. Fig. 31 shows the flowing steam-quality profiles through a horizontal section 2 m above the injection well. In the steam chamber, the steam-quality profiles have an average slope roughly equal to 0.02 quality units per meter. As a means to analyze the quality slopes in the vertical and horizontal directions, the heat balance can be used. The heat balance is derived by considering a small-volume element of dimensions dx, dy, and dz, as shown in Fig. 32, and is given by the statement that the net heat into the element is equal to the heat flux, ˙ , lost from the element. Q The heat balance is given by:

Fig. 27—Pressure through a vertical profile intersecting the LP well pair.

qx = −kTH

where kTH is the thermal conductivity, T is the temperature, ␳V and ␳L are the vapor and liquid densities, uV and uL are the vapor- and liquid-phase (x-directed) velocities, and hV and hL are the vapor and liquid phase specific enthalpies. The first term on the right side of Eq. 3 is the conductive heat transfer term, whereas the second term on the right side is the convective term. The x-directed vaporand liquid-phase mass fluxes, mV⳱␳VuV and mL⳱␳LuL, can be substituted into Eq. 3 to give qx = −kTH

˙ dxdydzdt. . . . . . (1) + 关共qz + dqz兲dxdydt兴 + Q After dividing by the elemental dimensions, taking the limit as they approach zero, the result is: ⭸qx ⭸qy ⭸qz ˙ = 0. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (2) + + +Q ⭸x ⭸y ⭸z In the x direction, the heat flux in each direction is the sum of the conductive and convective terms:

Fig. 28—Flowing steam quality through a vertical profile intersecting the LP well pair. February 2007 SPE Reservoir Evaluation & Engineering

⭸T + 共mVxhV + mLxhL兲. . . . . . . . . . . . . . . . . . . . . . . . . (4) ⭸x

Often, the enthalpy is expressed as a function of temperature by using the specific heat capacity to express Eq. 3 as

共qxdydzdt兲 + 共qydxdzdt兲 + 共qzdxdydt兲

= 关共qx + dqx兲dydzdt兴 + 关共qy + dqy兲dxdzdt兴

⭸T + 共␳VuVhV + ␳LuLhL兲x , . . . . . . . . . . . . . . . . . . . . . . (3) ⭸x

qx = −kTH

⭸T + 关␳VuVcpV共T − Tref兲 + ␳LuLcpL共T − Tref兲兴x , . . . . . (5) ⭸x

which results in the more often found temperature-based convective heat transfer term in the differential heat balance. Given the definition of steam quality, f, the heat flux in the x direction given by Eq. 4 can be re-expressed as qx = −kTH

⭸T + 关fmxhV + 共1 − f兲mxhL兴, . . . . . . . . . . . . . . . . . . . . (6) ⭸x

qx = −kTH

⭸T + 共f ␭ + hL兲mx , . . . . . . . . . . . . . . . . . . . . . . . . . . . . (7) ⭸x

or

Fig. 29—Temperature through a horizontal profile 2 m above the LP well-pair injector. 31

Fig. 30—Pressure through a horizontal profile 2 m above the LP well-pair injector.

Fig. 31—Flowing steam quality through a horizontal profile 2 m above the LP well-pair injector.

where ␭ is the latent heat of vaporization, given by ␭⳱hV−hL, and mx is the total (vapor and liquid) mass flux. Similarly, the y- and z-directed heat fluxes are given by

the steam chamber—that is, the larger the heat transfer. On considering the steam-quality slopes in the vertical (0.016 m−1) and horizontal (0.02 m−1) directions determined from Figs. 28 and 31 after the injection pressure is reduced because the quality gradient is larger in the horizontal direction, more of the heat supplied is being directed toward expanding the chamber laterally as opposed to feeding overburden losses. To clarify, this is when the steaminjection pressure is reduced after it reaches the top-gas zone. This is a key result that indicates that the optimized steam-injection strategy provided less heat flux in the vertical direction than the horizontal direction after the top-gas zone was reached. This means that there was less heat transferred to the overburden and convective losses to the gas-cap zone, and most of it was directed to growth at the sides of the chamber. Thus, after the steam chamber reached the top-gas zone, the process was more thermally efficient than would have been the case if steam was convectively transported into the top-gas zone. In a 3D SAGD chamber, there are heat losses not only in the plane (lateral-vertical plane), but also in the downwell direction. If the temperature varies in the downwell direction, then there is also conductive heat transfer in the downwell direction. This will further impact the steam-quality variation at any lateral-vertical slice of the steam chamber.

qy = −kTH

⭸T + 共f ␭ + hL兲my . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (8) ⭸y

and qz = −kTH

⭸T + 共f ␭ + hL兲mz . . . . . . . . . . . . . . . . . . . . . . . . . . . . (9) ⭸z

After substituting Eqs. 7 through 9 into Eq. 2, the result is ⭸ ⭸ ⭸ 关共f ␭ + hL兲mx兴 + 关共f ␭ + hL兲my兴 + 关共f ␭ + hL兲mz兴 ⭸x ⭸y ⭸z ˙ + = −Q

冉 冊 冉 冊 冉 冊

⭸ ⭸ ⭸ ⭸T ⭸T ⭸T k + k + k ⭸x TH ⭸x ⭸y TH ⭸y ⭸z TH ⭸z . . . . . . . . . . . . . . . . . . . . . . . . . . (10)

Within a SAGD steam chamber, the temperature and pressure are nearly constant. This means that the heat-conduction terms in Eq. 10 are very small and that the latent heat of vaporization and liquid enthalpy are essentially constant. Applying constant temperature and pressure to Eq. 10 yields ⭸ ⭸ ⭸ ˙ 关共f ␭ + hL兲mx兴 + 关共f ␭ + hL兲my兴 + 关共f ␭ + hL兲mz兴 = −Q ⭸x ⭸y ⭸z . . . . . . . . . . . . . . . . . . . . . . . . . . (11) or, after rearrangement and canceling terms associated with mass continuity,

Enthalpy vs. Temperature Fig. 33 displays a steam saturation curve and the enthalpy-vs.temperature two-phase envelope. Within the envelope, both vapor and liquid exist. The lines within the two-phase region are constant steam-quality lines. The topmost line represents the 100% steam-

˙ ⭸ ⭸ ⭸ Q 共fmx兲 + 共fmy兲 + 共fmz兲 = − . . . . . . . . . . . . . . . . . . . . (12) ⭸x ⭸y ⭸z ␭ Eq. 12 reveals that the energy needed to account for the heat ˙ , is generated from a loss of steam quality in the domain. losses, Q For 1D flow, Eq. 12 simplifies to ˙ ⭸f Q . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (13) =− ⭸x ␭mx From observation, it can be seen that in a constant-temperature pressure domain where there are heat losses, the steam quality will fall with distance. In one dimension, if at x=x0, the steam quality is f0, then the steam-quality distribution is given by f = f0 −

˙ Q 共x − x0兲, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (14) ␭mx

which reveals (provided that the ratio between the heat losses and mass flow is constant) that the steam quality drops linearly with distance. Another way of interpreting Eq. 13 is that the larger the steamquality gradient, the larger the specific heat losses at the edges of 32

Fig. 32—Heat transfer into and out of a differential element. February 2007 SPE Reservoir Evaluation & Engineering

Fig. 34—Vertical flowing steam-quality profiles in the form of enthalpy vs. temperature for the LP well pair.

whereas the bottom parts of the curves represent the top of the vapor chamber where the quality transitions from that at steamchamber conditions to zero (where the plots cross the zeroquality line). After the chamber has matured (after approximately 3 years of operation), the maximum and minimum specific enthalpies of the steam (beyond the chamber edge—that is, above the dotted line in Fig. 34) are roughly constant throughout the chamber. The profiles in the edge region of the chamber (below the dotted line) become more vertical as the process evolves owing to continued conductive heating of the materials above the edge of the chamber. Fig. 33—Steam saturation curve and enthalpy-vs.-temperature diagram.

quality line, whereas the bottom-most line represents the 0% steam-quality line. Consider the situation in which the steam chamber exists at State A. The steam chamber is at nearly constant temperature and pressure and so sits at Point A throughout the chamber. If the steam quality is known at the steam injector, then the enthalpy of the injected steam is known. If the steam quality is 90%, then the injected steam’s enthalpy is given by the value at Point B. Because of heat losses to the outer regions of the steam chamber, the steam quality in the chamber falls, moving away from the injection well. Because the temperature and pressure are roughly constant throughout the chamber, the enthalpy at any point in the chamber will lie along the line from Point B to Point C. If the steam loses all of its latent heat, it reaches Point C and is liquid water. If further cooling occurs, then a reduction of the liquid water temperature results. As the pressure of the process evolves, given that the chamber is at saturation conditions, Point A will shift along the steam saturation curve. For example, if the injection pressure falls as the process evolves, Point A will move in the left direction along the saturation curve. As a consequence, Points B and C will also shift in the left direction. The extents of the vertical profiles of the flowing steam quality can be plotted on the enthalpy-vs.-temperature diagram, as shown in Fig. 34. This is a subregion of the diagram displayed in Fig. 33. The vertical portions of the profiles (above the dotted line in Fig. 34) represent the part of the chamber that is nearly all steam and is at saturation conditions. Here, the temperature is roughly constant, and heat transfer throughout the chamber is mainly due to convection and is reflected by the steam-quality variation. At the edges of the chamber (the portions of the plots below the dotted line), steam is losing its latent heat, the quality rapidly falls, and the specific enthalpy of the wet steam is relatively low. Also, partial pressure effects come into play because solution gas comes out of solution from the bitumen at the edges of the chamber. The top part of each profile represents the enthalpy of the steam near the injection well (highest quality and, consequently, enthalpy), February 2007 SPE Reservoir Evaluation & Engineering

Conclusions The flowing steam quality provides a novel method to visualize heat transfer within and at the boundaries of the steam chamber. This is useful because the pressure and temperature of the steam chamber are nearly constant. The flow steam-quality profiles provide a means to examine convective heat transfer in the reservoir. The optimized operating strategy for SAGD in reservoirs with top gas has high initial chamber injection rates and pressures prior to chamber contact with the top gas. Subsequent to breakthrough, the chamber injection rates are lowered to balance pressures with the top gas and so avoid or at the least minimize convective heat losses of the steam to the top-gas zone. The lower the top-gas pressure, the lower the chamber pressure and, therefore, the lower the chamber temperature. In addition to avoiding convective heat losses to top gas, the lowering of chamber pressure also reduces conductive heat losses to the over- and underburden and allows harvesting of heat stored in the chamber before breakthrough. The lower chamber temperature does, however, lead to higher bitumen viscosities. To optimize SAGD, steam conformance must be managed along the wellbore to ensure full contact of the steam chamber to the reservoir penetrated by the horizontal well. The optimized steam-injection strategy promotes heat transfer in the lateral direction over that in the vertical direction. This reduces heat losses to the overburden and convective losses to the top-gas zone. The amount of heat supplied by the injected steam should be only that required to mobilize the bitumen at the edges of the chamber, not to induce large conductive losses to the overburden. The dual-well-pair operating strategy was determined by optimizing the individual well pair reservoir-simulation models separately and then integrating the optimized operating strategies into the dual-well-pair model. Because of the existence of the top-gas zone, the strategies developed for each of the single-well-pair models were designed to be gentle and not promote extensive communication with the top-gas zone. For this reason, the two individually optimized operating strategies could be implemented together in the dual-well-pair model. The dual-well-pair model revealed that communication between the two well pairs did influence the performances of the well pairs slightly. 33

Nomenclature cpL ⳱ heat capacity of liquid cpV ⳱ heat capacity of vapor f ⳱ steam quality hL ⳱ liquid-phase specific enthalpies hV ⳱ vapor-phase specific enthalpies kTH ⳱ thermal conductivity mx ⳱ total (vapor and liquid) mass flux (x component) my ⳱ total (vapor and liquid) mass flux (y component) mz ⳱ total (vapor and liquid) mass flux (z component) qx ⳱ heat flux in x direction qy ⳱ heat flux in y direction qz ⳱ heat flux in z direction ˙ ⳱ heat flux lost from element Q T ⳱ temperature Tref ⳱ reference temperature uV ⳱ vapor-phase velocities (x component) uL ⳱ liquid-phase velocities (x component) ␭ ⳱ latent heat of vaporization ␳V ⳱ vapor density ␳L ⳱ liquid densities Acknowledgments The authors would like to acknowledge Paramount Resources Ltd. for permission to publish this study. The comments from the reviewers of this paper are also appreciated. References AED (Alberta Economic Development). 2004. Oil Sands Industry Update. Available at the Alberta Dept. of Energy website: http:// www.energy.gov.ab.ca/com/default.htm. Boone, T.J., Youck, D.G., and Sun, S. 1998. Targeted Steam Injection Using Horizontal Wells With Limited Entry Perforations. Paper SPE 50429 presented at the SPE International Conference on Horizontal Well Technology, Calgary, 1–4 November. DOI: 102118/50429-MS. Butler, R.M. 1997. Thermal Recovery of Oil and Bitumen. Calgary: GravDrain Inc. Computer Modelling Group (CMG). 2004. STARS Users Manual, Version 2004.10. Calgary: CMG Ltd. Edmunds, N. and Chhina, H. 2001. Economic Optimum Operating Pressure for SAGD Projects in Alberta. J. Cdn. Pet. Tech. 40 (12): 13. Gates, I.D. and Chakrabarty, N. 2005. Optimization of Steam-Assisted Gravity Drainage (SAGD) in Ideal McMurray Reservoir. Paper 2005193 presented at the Canadian Intl. Petroleum Conference, Calgary, 7–9 June. Good, W.K., Rezk, C., and Felty, B.D. 1997. Possible Effects of Gas Caps on SAGD Performance. Report to the Alberta Environment and Alberta Energy Utilities Board, Edmonton, Alberta, Canada, March. Ito, Y. and Suzuki, S. 1999. Numerical Simulation of the SAGD Process in the Hangingstone Oil Sands Reservoir. J. Cdn. Pet. Tech. 38 (9): 27– 35. Ito, Y., Hirata, T., and Ichikawa, M. 2004. The Effect of Operating Pressure on the Growth of the Steam Chamber Detected at the Hangingstone SAGD Project. J. Cdn. Pet. Tech. 43 (1): 47–53. Kisman, K.E. and Yeung, K.C. 1995. Numerical Study of the SAGD Process in the Burnt Lake Oil Sands Lease. Paper SPE 30276 presented at the International Heavy Oil Symposium, Calgary, 19–21 June. Komery, D.P., Luhning, R.W., and O’Rourke, J.C. 1999. Towards Commercialization of the UTF Project Using Surface Drilled Horizontal SAGD Wells. J. Cdn. Pet. Tech. 38 (9): 36–43. Law, D.H.S., Nasr, T.N., and Good, W.K. 2003a. Lab-Scale Numerical Simulation of SAGD Process in the Presence of Top Thief Zones: A Mechanistic Study. J. Cdn. Pet. Tech. 42 (3): 29–35.

34

Law, D.H.S., Nasr, T.N., and Good, W.K. 2003b. Field-Scale Numerical Simulation of SAGD Process with Top-Water Thief Zone. J. Cdn. Pet. Tech. 42 (8): 32–38. Mehrotra, A.K. and Svrcek, W.Y. 1986. Viscosity of Compressed Athabasca Bitumen. Cdn. J. Chem. Eng. 64: 844–847. Nasr, T.N., Law, D.H.S., Beaulieu, G., Golbeck, H., Korpany, G., and Good, W.K. 2003. SAGD Application in Gas Cap and Top Water Oil Reservoirs. J. Cdn. Pet. Tech. 42 (1): 32–38. Robinson, W., Kenny, J., Hernandez-Hdez, I.L., Bernal, A., and Chelak, R. 2005. Geostatistical Modeling Integral to Effective Design and Evaluation of SAGD Processes of an Athabasca Oil-sands Reservoir: A Case Study. Paper SPE 97743 presented at the SPE/PS-CIM/CHOA International Thermal Operations and Heavy Oil Symposium, Calgary, 1–3 November. DOI: 102118/97743-MS. Roxar. 2004. IRAP RMS User’s Manual. Stavanger: Roxar. Singhal, A.K., Ito, Y., and Kasraie, M. 1998. Screening and Design Criteria for Steam-Assisted Gravity Drainage (SAGD) Projects. Paper SPE 50410 presented at the SPE International Conference on Horizontal Well Technology, Calgary, 1–4 November. DOI: 102118/50410-MS. Siu, A.L., Nghiem, L.X., Gittins, S.D., Nzekwu, B.I., and Redford, D.A. 1991. Modeling Steam-Assisted Gravity Drainage Process in the UTF Pilot Project. Paper SPE 22895 presented at the SPE Annual Technical Conference and Exhibition, Dallas, 6–9 October. DOI: 102118/22895MS. Suggett, J., Gittins, S., and Youn, S. 2000. Christina Lake Thermal Project. Paper SPE 65520 presented at the SPE/CIM International Conference on Horizontal Well Technology, Calgary, 6–8 November. DOI: 102118/65520-MS. Yee, C.-T. and Stroich, A. 2004. Flue Gas Injection Into a Mature SAGD Steam Chamber at the Dover Project (Formerly UTF). J. Cdn. Pet. Tech. 43 (1): 54–61.

SI Metric Conversion Factors bar × 1.0* E+05 cycles/sec × 1.0* E+00 ft × 3.048* E−01 E−02 ft3 × 2.831 685 °F (°F−32)/1.8 in. × 2.54* E+00

⳱ ⳱ ⳱ ⳱ ⳱ ⳱

Pa Hz m m3 °C cm

*Conversion factor is exact.

Ian D. Gates, PhD, P.Eng., is an associate professor in the Dept. of Chemical and Petroleum Engineering at the U. of Calgary. He is also a principal investigator of Alberta Ingenuity for In Situ Energy, based at the U. of Calgary, where he is researching well placement and operating strategies for catalytic-airsteam recovery processes. His current research interests are in thermal and thermal-solvent processes for heavy-oil and bitumen reservoirs. Joseph Kenny, MSc, P.Eng., is President of ATECH Application Technology Ltd. and has more than 20 years of experience in the oil and gas industry, especially in well-test analysis, decline analysis, reservoir characterization, multiphase-flow analysis, compressor optimization and gathering systems, geostatistical analysis, and reservoir simulation. Ivan L. Hernandez-Hdez, M.Eng., is a reservoir engineer at ATECH Application Technology Ltd. and has more than 11 years of experience in reservoir engineering and simulation, optimization, oil and gas gathering systems, and geostatistical modeling. He is currently working on different SAGD strategies to improve bitumen recovery and reduce SOR. Gary L. Bunio, MSc, P.Eng., is Senior Vice President and COO of MGM Energy Corp. He has more than 25 years of experience in the oil and gas industry in the roles of researcher, manager, and innovator. He is currently working on innovative ways to bring Mackenzie Valley gas to market.

February 2007 SPE Reservoir Evaluation & Engineering