Effects of Well Placement using Multi- Segmented Wells in a Full Field Thermal Model for SAGD: Athabasca Oil Sands

Size: px
Start display at page:

Download "Effects of Well Placement using Multi- Segmented Wells in a Full Field Thermal Model for SAGD: Athabasca Oil Sands"

Transcription

1 PAPER Effects of Well Placement using Multi- Segmented Wells in a Full Field Thermal Model for SAGD: Athabasca Oil Sands F. AKRAM Schlumberger Canada Limited This paper has been selected for presentation and publication in the Proceedings for the World Heavy Oil Congress All papers selected will become the property of WHOC. The right to publish is retained by the WHOC s Publications Committee. The authors agree to assign the right to publish the above-titled paper to WHOC, who have conveyed non-exclusive right to the Petroleum Society to publish, if it is selected. Abstract Standing at 2.5 trillion barrels, Canada has the largest portion of the world s ultra-heavy oil and bitumen resources 1. While shallow heavy oil reserves are extracted from pit mines, deeper reserves can only be extracted through wells. Production requires Steam Assisted Gravity Drainage (SAGD) and Cyclic Steam Simulation (CSS) methods 2. The optimal placement of wells defines the propagation of steam within the reservoir and the resulting flow of crude towards the producers. A full field thermal model was developed using Petrel* Reservoir Engineering to simulate the SAGD recovery process for Athabasca Oil Sands. This study was conducted to observe the behavior of the SAGD process under a full field environment with multiple pairs using an advanced well model. This model divides the wellbore into multiple segments, where each segment acts as an individual connection to the reservoir. This method allows for the simulation of complex multi-phase flow effects such as counter-flow in slowly flowing horizontal wells, fluid fall-back, variable well-bore storage and friction 3. Sensitivity analysis on parameters, such as well spacing (inter-well and horizontal) and steam injection rates combined with the complex well model allowed a true simulation of a SAGD process in a full field environment. Software programs including Eclipse * Parallel optimization and the John Appleyard Linear Solver (JALS) * were used to run multiple simulations in less time. This paper describes the process and the sensitivity analysis used to design various SAGD models. The impact of using Multi-Segmented Wells and the effects of a full field model on the steam chambers are also discussed 4. Introduction Reservoir studies including reservoir simulation for SAGD have been conducted for the past few years to understand the effect of steam injection, and the consequent formation of a steam chamber, on crude recovery 5. The dominant issue throughout this period has been the limitation on thermal simulation calculations imposed by computing power. Due to their complex nature, thermal simulations require significantly more iterations, processing power, and computer memory when compared to conventional black oil simulations. At the same time, computing limitations often prevent the use of a small grid block size in simulations. To minimize artifacts associated with simulation, a small grid block size is required to model gravity drainage appropriately. The combination of these factors results in reservoir studies that are done either on a single well pair or * Mark of Schlumberger

2 with large grid block sizes that are inadequate to truly simulate the gravity effect. Reservoir simulation can yield a lot of information about field development and appropriate use of this technology helps us to see more than what we are accustomed to. With advanced visualization and the ability to run large reservoir simulation models over a shorter period of time, we can run sensitivity analysis on multiple parameters as presented in this document for optimal well placement and well completion design. This results in more rigorous reservoir studies yielding better results which help us to make better business decisions. In this study, we used multiple grid block sizes to study the impact on production and on Steam-Oil Ratio (SOR). We simulated 6 well pairs over two completion scenarios; full and half tubing. Individual well pair simulations were compared with the full field model, where the full field model refers to simulation of all well pairs together. The use of a multisegmented well (MSW) model provided a more accurate production profile 6. We also tested several completion scenarios and varied horizontal spacing between the well pairs. This study exploits a hypothetical heavy oil reservoir **. The well pairs were drilled with a horizontal separation of 80 meters. The injector and the producer were spaced with a 5 meter vertical separation between them. The simulation was run for a period of five (5) years with four months of preheating. The virgin reservoir pressure was 16 bar, which was increased to 30 bar for the first year, and then lowered to 25 bar for the remaining 5 years. The reservoir temperature was 13 C and the injected steam was kept at a temperature of 258 C. At the production wells, the steam trap constraint was used with 7 C subcooling 5. The multiple sensitivity runs along with their resulting impact on field development plan are described in the sections that follow. Schlumberger products Petrel * Reservoir Engineering and ECLIPSE * Thermal were used for modeling and simulation. Geological Model To study the impact of well placement and completion design on a SAGD process, a detailed statistical geological model was prepared using the principles described by the Center for Computational Geostatistics 7 for Athabasca Oil Sands. A rectangular grid, measuring 940 meters long by 650 meters wide, was used to build the model. The choice of grid block size dictated the number of grid blocks in either direction. There were five rock types introduced in the model as follows: Facies 1: Mud Facies 2: Sandy Mud Facies 3: Muddy Sand Facies 4: Fine Sand Facies 5: Sand The facies were distributed using Sequential Gaussian Simulation (SGS) 8. Facies distribution in the geological model is shown in Table 1. The oil saturation, porosity and permeability distribution were computed by assigning the individual facies a constant value shown in Table 1. Any number of probable facies distribution models may be generated using SGS. The realization that was used for sensitivity analysis is shown in the Figure 1 and the oil saturation distribution is shown in Figure 2. All saturations were normalized by S o +S w =1. ** Personal communication with D. Law Edmonton, Schlumberger Limited * Mark of Schlumberger Simulation Runs Simulation runs were set up after preparing the geological model. Use of Petrel Reservoir Engineering * eliminated the need for export and import; instead the same geological model was simulated under various sensitivity parameters that are discussed in this document. The simulations were run as a deadoil case 2 with oil molecular weight of 500. Varying Grid Block Sizes Three grids were chosen for simulation with different grid block sizes: 1. 5x50x1 2. 2x50x1 3. 2x25x1 Block sizes are indicated in i,j,k format where: i is distance of the cross section of the well, j is the length of the well bore, and k is the thickness of layers. Grid orientation is shown in Figure 3. The cumulative production, cumulative steam injection and the resulting SOR for the three different grid block size simulations can be seen in Figure 4, 5 and 6 respectively. As depicted in Figure 4 and 5, increasing the grid block size along the length of the well bore (j) had minimal impact on production however increasing the grid block size along the cross-section of the well bore (i) had a significant impact on cumulative production; production estimates decreased by mmbbl. Similarly, the steam injection requirements were less in the larger grid block models when compared to the smaller grid block model. The impact on resulting SOR in the simulation was interesting. The SOR was found to decrease with the decrease in grid block size; 2.05 for the 5x50 model compared with 1.88 for the 2x25 model (Figure 6). Steam distribution for the 2x50 and the 5x50 simulations is shown in Figure 7. The results show that the steam chamber was more gradually connected in the 2x50 model when compared to the 5x50 model for all 6 well pairs. The simulations were run on an 8-processor cluster. Using multi CPU clusters was essential for building accurate thermal simulation models. Being able to simulate smaller grid block sizes and to visualize the distribution of steam chamber gave us more insight and allowed us to get more information to make better business decisions while preserving the geological heterogeneity of the reservoir. Testing Well Completion Designs The well pairs used in the simulation were 850 meters long with the last 610 meters of the wells entirely perforated. Three different scenarios were tested: 1. Full tubing design 2. Half tubing design 3. Half tubing design with 5 infill producer wells In all three scenarios, both injector and producer wells were fully cased. The producers also had tubing to the toe of the well. Packers were placed at the top of the perforation with the oil flowing alongside the tubing to the toe and then inside the tubing. In the full tubing completion design, the injectors had full tubing. In the half tubing design, the injectors had tubing half way through the perforation. The different completion designs for the producer and injector wells are shown in Figure 8. Along the length of the well bore, the well was segmented every 50 meters. In the third scenario, the half tubing design was used, but 5 infill producers were added. These producers were placed between the 6 well pairs 40 meters apart to recover incremental oil. The producers were put on production 2 years into the 2

3 simulation. Cumulative oil production and SOR are compared for the three scenarios in Figure 9 and 10. In the simulation, the half tubing design was predicted to produce 1.61 mmbbl more oil over a period of 5 years compared to full tubing design. The extra infill producers were predicted to produce 0.3 mmbbl extra oil. Assuming one barrel of synthetic oil sells for $60.00, the undiscounted revenue for the full tubing design was estimated at $340 million and the half tubing design was estimated at $437 million. The use of half tubing in the well completion design resulted in two benefits: first, tubing costs were half that of a full tubing completion, and second, the design recovered more oil. The addition of infill producers in the simulation added another $18 million to the undiscounted revenue. If we assume a cost of one million dollars per infill well, there would still be a margin of 13 million dollars. The placement of infill producers can be visualized in Figure 11. Despite the estimate of additional production shown in the simulation, the placement of infill producers requires further study. These wells do not seem to produce significant amounts of oil until the steam chamber has reached them (Figure 11). Further optimization may result in even greater recovery and decreased SOR. Yet, even without optimization, the third scenario still has the better SOR among the three at These well completion design scenarios could only be evaluated when we performed a full field simulation run. When we ran individual well pair runs and compared them with the full field design, the results were quite different. Comparison of Individual Well Pair Runs and their Cumulative with Full Field Run In order to test the validity of the results from the full field simulation, we ran individual well pair runs, added their results and compared them with the full field simulation run (Figure 12). The simulation showed a difference of 0.39 million barrels of oil when the cumulative individual well pair production was compared with the full field model. This difference occurs because of a simulation artifact known as double dipping. The double dipping effect is essentially counting production of the same oil twice. Oil between two well pairs is produced once by the first pair and then again by the second pair. As the results are cumulated, the amount can be significant. The double dipping effect is only a concern if there is communication between well pairs. If the wells don t communicate over the simulation period, the individual well pairs are adequate and the results can be cumulated. If the well pairs do communicate in a full field environment, it is imperative to run a full field simulation. More accurate prediction of cumulative production will allow better planning of surface facilities. Varying Well Pair Spacing We tested the impact on cumulative production by moving well pair 1 and 6, 120 meters apart from well pair 2 and 5 respectively from original spacing of 80 meters (Figure 13). The difference in cumulative production in a full field model for pair 1 for the two well pair positions is shown in Figure meter spacing placement predicted less oil production because of the clay that was present on the west of the new placement position (Figure 13). When we examined the oil rate for production well 1 in the original placement the well showed that it was in decline mode. For the new placement, production well 1 showed a steady rate. It was expected to predict more cumulative oil for longer simulation period than the original placement (Figure 15). Our next step was to increase the perforation for both injectors and producers for all well pairs, from 610 meters from the toe to 800 meters from the toe with well pairs 1 and 6 under the new horizontal placement. The production profile for the new placement field design was compared with the original field design and the results were plotted (Figure 16). In this scenario the steam traveled farther down towards the heel. The simulation predicted greater oil recovery along the length of the well bore, but the communication of steam chamber took longer so estimated production was less for the simulation period. The SOR value was almost unchanged. Figure 17 shows steam chamber distribution and Figure 18 shows the impact on SOR. When examining this over an extended simulation period, the new design predicted more cumulative oil production since production rates are steady compared to the old design, where production is in decline. SOR decreases as simulation continues beyond 5 years for the new design (not shown). The simulation showed that communication between the steam chambers for all well pairs was not complete (Figure 20), while it was complete for the old design (Figure 19) at the end of simulation. However, this communication would be completed over time which would ultimately result in higher recovery. As a result, we will see reduced SOR with better sweep efficiency. As expected, well placements and completion designs have a huge impact on the overall recovery of the field. These effects can be studied accurately only with full field simulation models. Conclusions 1. The investment in heavy oil reservoirs is enormous and with the high cost of production, it is paramount to do reservoir studies that are as true to reality as possible. 2. Varying the grid block size resulted in different cumulative field oil production estimates. Decreasing the grid block size increases the number of grid blocks in a simulation run, however it simulates the physics of gravity drainage better. This difference is noted only if MSW is used. 3. Two completion designs were tested. The half tubing design proved to be the better completion design in this reservoir simulation. This design required less tubing resulting in lower completion costs, was more efficient in steam distribution, and estimated higher production. 4. Adding infill producer wells increased recovery. However, these infill wells need to be optimally planned to offset their cost with increased production. 5. The practice of simulating standalone well pairs and cumulating their results proved to be inadequate. The results were optimistic compared to the full field runs which took into account the communication of steam chambers across the entire field. 6. Our full field simulation approach allows for accurate estimation of different well placement and completion scenarios. Acknowledgement The author thanks David Law, Schlumberger Data and Consulting Services (DCS) and Paul Naccache, Schlumberger Information Solutions (SIS) for their advice on this research. He also thanks his colleagues in Calgary for their support. 3

4 NOMENCLATURE SAGD = Steam Assisted Gravity Drainage CSS = Cyclic Steam Simulation JALS = John Appleyard Linear Solver MSW = Multi-segmented Well SOR = Steam-Oil Ratio SGS = Sequential Gaussian Simulation S o = Oil Saturation S w = Water Saturation k (md) = Permeability (in millidarcy) φ = Porosity % = Percentage mmbbl = Millions of Barrels of Oil REFERENCES 1. Canadian Petroleum Communications Foundation,, Downloaded 14 March PRATZ, M., Thermal Recovery. Monograph Series, SPE, Richardson, Texas. Vol ECLIPSE 300 Reference Manual, Schlumberger, OBALLA, V., and BUCHANAN, W.L., Single Horizontal Well in Thermal Recovery Processes; SPE Paper 37115, GATES, I.D., KENNY, J., and HERNANDEZ-HDEZ, I.L., Steam-Injection Strategy and Energetics of Steam- Assisted Gravity Drainage; SPE/PS-CIM/CHOA 97742, HOLMES, J.A., BARKVE, T., and LUND, Ø., Application of a Multisegment Well Model to Simulate Flow in Advanced Wells; SPE Paper 50646, DEUTSCH, C.V., and MCLENNAN, J.A., Guide to SAGD (Steam Assisted Gravity Drainage) Reservoir Characterization Using Geostatistics; Centre for Computational Excellence (CCG), Guidebook Series Vol. 3, University of Alberta, April Petrel User Manual, Schlumberger, FIGURES Figure 1: Realization of the facies model used for simulation. TABLES Facies Statistics Type Name % fraction S o % k (md) φ % 1 Mud Sandy Mud Muddy Sand Fine Sand Sand Figure 2: Distribution of Oil Saturation for the simulation model. Table 1: Statistics for different facies type 4

5 Figure 3: Grid orientation used for simulation. Figure 5: Field steam injection cumulative for varying grid block sizes. Figure 4: Field oil production cumulative for varying grid block sizes. Figure 6: Field cumulative steam-oil ratio for varying grid block sizes. 5

6 Figure 7: Temperature distribution for two cases of grid block size. Figure 9: Comparison of cumulative oil production between the three completion designs. Figure 8: Completion designs; half tubing on the left and full tubing on the right. Figure 10: Comparison of cumulative SOR for the three completion designs. 6

7 Figure 11: Infill producers and the distribution of steam chamber at the end of 5 years and 9 months. Figure 13: New placement of well pair 1 and 6, 120 meter apart whereas rest are at original 80 meter spacing. Figure 12: Comparison of individual 6 well pair runs and their cumulative with full field simulation run. Figure 14: Comparison of cumulative oil production for well pair 1 under original and new placement. 7

8 Figure 15: Comparison of oil production rate for production well 1 under the original (80m) and new placement (120m). Figure 17: Steam chamber distribution for the old and new design. The variation is quite visible. Much better sweep efficiency for the new design. Figure 16: Comparison of cumulative oil production for the entire field between old spacing and completion with new well spacing and completion. Figure 18: Comparison of SOR for the old and new field design. 8

9 May 1, 2007 May 1, 2007 Jan 1, 2008 Jan 1, 2008 Jan 1, 2009 Jan 1, 2009 Jan 1, 2010 Jan 1, 2010 Jan 1, 2011 Jan 1, 2011 Jan 1, 2012 Jan 1, 2012 Figure 19: Evolution of steam chamber for the old design. All wells are 80 meters apart with 610m long perforation from the toe. Figure 20: Evolution of steam chamber for the new design. Well pairs 1 and 6 are 120 meters apart with new perforation of 800 meters from the toe. 9