Relationship between fractures, fault zones, stress, and reservoir productivity in the Suban gas field, Sumatra, Indonesia

Transcription

1 Relationship between fractures, fault zones, stress, and reservoir productivity in the Suban gas field, Sumatra, Indonesia Peter Hennings, Patricia Allwardt, Pijush Paul, Chris Zahm, Ray Reid Jr., Hugh Alley, Roland Kirschner, Bob Lee, and Elliott Hough ABSTRACT It is becoming widely recognized that a relationship exists between stress, stress heterogeneity, and the permeability of subsurface fractures and faults. We present an analysis of the South Sumatra Suban gas field, developed mainly in fractured carbonate and crystalline basement, where active deformation has partitioned the reservoir into distinct structural and stress domains. These domains have differing geomechanical and structural attributes that control the permeability architecture of the field. The field is a composite of Paleogene extensional elements that have been modified by Neogene contraction to produce basement-rooted forced folds and neoformed thrusts. Reservoirscale faults were interpreted in detail along the western flank of the field and reveal a classic oblique-compressional geometry. Bulk reservoir performance is governed by the local stress architecture that acts on existing faults and their fracture damage zones to alter their permeability and, hence, their access to distributed gas. Reservoir potential is most enhanced in areas that have large numbers of fractures with high ratios of shear to normal stress. This occurs in areas of the field that are in a strike-slip stress style. Comparatively, reservoir potential is lower in areas of the field that are in a thrust-fault stress style where fewer fractures with high shear-to-normal stress ratios exist. Achieving the highest well productivity relies on tapping into critically stressed faults and their associated fracture Copyright The American Association of Petroleum Geologists. All rights reserved. Manuscript received May 3, 2009; provisional acceptance June 18, 2009; revised manuscript received July 6, 2011; final acceptance August 16, DOI: / AUTHORS Peter Hennings ConocoPhillips Subsurface Technology, PR2014, 600 N. Dairy Ashford, Houston, Texas; peter.hennings@conocophillips.com Peter Hennings received his B.S. and M.S. degrees in geology from Texas A&M University and his Ph.D. in geology from the University of Texas. He has held various technical and supervisory positions in Mobil Research Company, Phillips Petroleum Company, and Conoco- Phillips. His research and application focus in these positions includes structure and tectonics, seismic interpretation, reservoir description, geomechanics, and fracture characterization. He is currently the manager of the Structure and Geomechanics Group in ConocoPhillips Subsurface Technology. He is an AAPG distinguished lecturer, a Geological Society of America honorary fellow, and is an adjunct professor at the University of Wyoming and consulting professor at Stanford University. Patricia Allwardt ConocoPhillips Subsurface Technology, PR2014, 600 N. Dairy Ashford, Houston, Texas; patricia.f.allwardt@conocophillips.com Patricia Allwardt received her B.S. degree in earth and planetary sciences from Harvard University and her Ph.D. in structural geology and geomechanics from Stanford University. During work on this paper, Tricia was a member of the Subsurface Technology Organization at ConocoPhillips focused primarily on integrating structural analysis, fracture characterization, and geomechanics into reservoir performance. Tricia is currently working in ConocoPhillips Gulf of Mexico Exploration Group. Pijush Paul ConocoPhillips Subsurface Technology, PR2014, 600 N. Dairy Ashford, Houston, Texas; pijush.k.paul@conocophillips.com Pijush Paul works in the structure and geomechanics team of ConocoPhillips Subsurface Technology Group in Houston. He leads the team s computational geomechanics program. His other projects focus on providing geomechanical models of reservoirs for completion and production optimization. He holds a Ph.D. in geomechanics and an M.S. degree in petroleum engineering from Stanford University, an MTech degree in applied geophysics from the Indian Institute of Technology, and a B.S. degree in AAPG Bulletin, v. 96, no. 4 (April 2012), pp

2 geology and physics from St. Xavier College, India. Chris Zahm ConocoPhillips Subsurface Technology, PR2014, 600 N. Dairy Ashford, Houston, Texas; present address: University of Texas, Bureau of Economic Geology, Austin, Texas; Chris Zahm received his B.S. degree in geology from the University of Wisconsin, his M.S. degree in geology from the University of Texas at Austin, and his Ph.D. from the Colorado School of Mines. He was employed by ConocoPhillips Subsurface Technology where he worked on reservoir structural geology projects before joining the Bureau of Economic Geology at the University of Texas at Austin. He is currently a research associate as part of the BEG-Reservoir Characterization Research Laboratory (RCRL) Industrial Associates program and adjunct professor in Jackson School of Geosciences. Ray Reid Jr. ConocoPhillips Subsurface Technology, PR2014, 600 N. Dairy Ashford, Houston, Texas; ray.r.reidjr@conocophillips.com Ray Reid Jr., is a senior petrophysical analyst in ConocoPhillips Subsurface Technology. Ray joined Phillips Petroleum Company in 1979 where he supported geophysical acquisition and exploration. In 2003, Ray joined ConocoPhillips. Since 2003, Ray has led the petrophysical image processing and interpretation function for ConocoPhillips. Hugh Alley ConocoPhillips Indonesia Inc., Ltd., Jakarta, Indonesia Hugh Alley received his B.Sc. and M.Sc. degrees in geology from the University of Manitoba in 1975 and 1982, respectively. He has held various technical and supervisory positions in the Exploration and Development departments of ConocoPhillips, Gulf Canada Resources, Maxus Energy, and Amoco. Hugh is currently a principal geologist supporting unconventional shale oil and shale gas exploration in ConocoPhillips Canada. Roland Kirschner ConocoPhillips Indonesia Inc., Ltd., Jakarta, Indonesia Roland Kirschner is an exploration geologist with ConocoPhillips in Perth, Australia. He received his M.S. degree in geology from Louisiana State University. Since joining Phillips Petroleum in 1999, he has held various technical positions within Phillips and, later, within ConocoPhillips both in damage zones. Two wellbores have been drilled based on this concept, and each shows a three- to seven-fold improvement in flow potential. INTRODUCTION The relationship between stress, stress variability, active faults, and the permeability of natural fracture systems in the subsurface is becoming widely recognized. Barton et al. (1995) presented initial evidence for elevated fluid flow associated with potentially active faults in boreholes in crystalline rock. Zoback (2007) expands this discussion to consider faults at a variety of scales in different geologic environments. Tamagawa and Pollard (2008) discuss a fractured basement gas reservoir where fracture-controlled well performance is significantly impacted by stress fields perturbed by faults. Also, active deformation, as manifested in faulting, can significantly alter the regional pattern of horizontal stress. Castillo and Zoback (1994) discuss how recent seismic movement along faults changes the local stress configuration at the scale of oil field structures in southern California. Active deformation related to fault movement generates fractures and results in local stress-field perturbations, both of which affect permeability in the vicinity of faults. The purpose of this article is to present an analysis of subsurface data from a fractured gas field that strongly reinforces these themes and shows how integration of these concepts can be used to assess reservoir potential and drill wells with higher productivities. GEOLOGIC BACKGROUND Suban field is located along the southwestern edge of the South Sumatra Basin in south-central Sumatra. The field produces wet gas from compressionally uplifted fractured crystalline and metamorphic basement and overlying clastic and reefal carbonate rocks (stratigraphic units [SUs] 1, 2, and 3; Figure 1). The island of Sumatra owes its present-day complex tectonic architecture to northeast-directed oblique subduction of the Indo-Australian plate along its southwestern margin. Significant crustal decoupling and strain partitioning occur along this zone with the fore-arc region southwest of the Sumatran fault moving in a northerly direction, along with the subducting Indo-Australian plate (Figure 1) (Milsom, 2005). This oblique lateral movement interacts with the Sunda craton along the 754 Suban Gas Field, Sumatra, Indonesia

3 dextral transcurrent Sumatran fault and Barisan Mountains transpressional belt that run the length of Sumatra. Estimates of lateral displacement along the Sumatran fault range greatly, but 150 km (93 mi) is a widely cited value (McCarthy and Elders, 1997). Modern contractional deformation along the Barisan Mountains spreads to the northeast, involving a wide swath adjacent to the core of the range and includes our area of study. Global positioning system observations along this zone indicate complex tectonic movements, with northeast to northwest azimuths over a 150-km (93-mi) width as measured orthogonally to the Sumatran fault (Barber and Crow, 2005; Milsom, 2005). The South Sumatra Basin is one of five basins that developed northeast of the present-day Sumatra volcanic arc, Barisan mountain chain, and Sumatran fault. The basins formed as rift systems in the early Cenozoic in a terrain floored by Mesozoic granitic, volcanic, and metasedimentary rocks. The dominant extensional fault fabric in the South Sumatra Basin trends northeast-southwest, with a subordinate trend to the northnorthwest south-southeast. The main phase of basin development occurred in the late Eocene to the early Oligocene (Barber and Crow, 2005). This first phase of sedimentation occurred as infilling of structural depressions by clastic debris eroded from exposed basement blocks (SU2) (Figure 1) (De Smet and Barber, 2005). Thermal subsidence followed rifting in the late Oligocene to the early Miocene and allowed a marine incursion that introduced fine-grained marine sequences and reefal buildups on high-standing blocks (SU3). Continued subsidence drowned the carbonate system and resulted in deposition of organic-rich deep-water shales and marls that later became the gas-prone hydrocarbon source rocks and top seals of the system (SU4). Northeast-directed compression and tectonic inversion began in the mid-miocene, and terrestrial sediments prograded from the southwest, resulting in deposition of SU5. Stratigraphic unit 6 (SU6) spans the related transition from marginal marine to erosional emergence and contains coal beds and terrestrial clastic deposits. South Sumatra has a complex history of volcanicity that spans the Cenozoic Era and continues today (Crow, 2005; Gasparon, 2005). Stratigraphic unit 6 contains numerous tuff and volcaniclastic beds. The degree of recent strike-slip dissection in the region of Suban field is unclear, but Pulunggono (1986) documents several presumed lateral surface offsets with the southwestern part of the basin. Figure 1 (AA ) shows preserved extension to the northeast and increased inversion and contraction to the southwest. The complex architecture of Suban field was unraveled using a prestack depth-migrated seismic volume acquired in domestic and overseas locations. Besides a fascination with fractured reservoirs systems, his main interests focus on the analysis and modeling of the sedimentology and stratigraphy of shallow-water to deep-marine clastic reservoir systems. Bob Lee ConocoPhillips Indonesia Inc., Ltd., Jakarta, Indonesia Bob Y. Q. Lee received his B.S. degree in chemical engineering from the University of Saskatchewan in He has held reservoir engineering positions with various companies including ConocoPhillips Indonesia and is currently with InterOil. His technical interest is to integrate classical and analytical reservoir engineering techniques with the modern-day workflow of reservoir characterization and simulation. Elliott Hough ConocoPhillips Indonesia Inc., Ltd., Jakarta, Indonesia Elliott Hough received his B.S. degree in mechanical engineering from Colorado State University in He has held various technical and supervisory reservoir engineering positions in Phillips Petroleum Company and ConocoPhillips. Elliott is currently a principal reservoir engineer, supporting unconventional shale oil and shale gas exploration in ConocoPhillips L48 Exploration. ACKNOWLEDGEMENTS We thank ConocoPhillips Technology, Conoco- Phillips Indonesia Inc., Badan Pelaksana Kegiatan Usaha Hulu Minyak dan Gas Bumi (BPMIGAS), and partners Talisman Energy Inc. and PT Pertamina (Persero) for permission to publish the data and our findings. We thank Alan P. Morris, Laird B. Thompson, and especially Mark Zoback for their thorough reviews of the manuscript that greatly improved its organization and technical message. Badley Geoscience TrapTester was used for structural interpretation and fault stress modeling, GeoMechanics International SFIB and MohrFracs were used for wellbore stress and fracture characterization, and Paradigm GOCAD was used for structural modeling and data integration. The information contained herein is for information purposes only, and no representation or warranty is provided as to its content and accuracy. The AAPG Editor thanks the following reviewers for their work on this paper: Alan P. Morris, Laird B. Thompson, and Mark D. Zoback. Hennings et al. 755

4 Figure 1. Tectonic setting, regional structural configuration, and simplified stratigraphic units of Suban gas field. The location of regional seismic cross section AA is shown in the inset map. The location of field-specific cross section BB is shown in Figure 2. TWT = two-way traveltime (BB, Figure 1) and data from 11 wellbores (Figure 2). The uplift that forms the field has approximately 3 km ( 1.9 mi) of local structural relief and can be divided into three lateral domains based on morphology and genesis. The northeast structural domain contains a pronounced structural culmination produced by a system of arcuate northeast-directed thrust faults that may have reactivated a preexisting normal fault system, although synrift sediments are not encountered. In the northern area of the field, a direct expression exists in the seismic data that constrains the trajectory of the master fault of this system at depth (BB, Figure 1). Overall, the eastern anticline is a northeast-vergent fault-propagation fold. Wells 2, 3, and 5 were drilled in this domain. The field center contains additional thrust faults coring northeast-vergent folds. Faults in this do756 Suban Gas Field, Sumatra, Indonesia main have displacements on the scale of tens of meters. Wells 6, 7, and 9 drilled into this domain. The southwestern structural domain of the field is characterized by a doubly plunging linear anticlinal uplift produced over a single southwest-dipping thrust fault with a maximum of 700 m (2297 ft) of throw (Figure 2). The maximum structural relief on the anticline is approximately 800 m ( 2625 ft). The western limb of the anticline is also the western flank of the overall field. Directly southwest of Suban field is a deep and tight syncline (AA, Figure 1). Southwest- and northeast-vergent thrust faults emanate out of the syncline on both its limbs, thus indicating a synclinal crowding genesis (e.g., Mitra, 2002) for the western anticline. We contend in this paper that the main thrust fault that cores the western anticline and forms the southwestern structural domain of the field has a significant

5 component of dextral strike slip in addition to the dip slip. The main thrust faults that form both the eastern and western anticlinal folds of Suban field alter the topographic surface; this, insights from regional tectonic indicators, and our wellbore stress analysis and geomechanical modeling discussed below cause us to propose that the geologic processes that formed the field are active today. Suban field was discovered in 1998, and production began in Estimates of original gas in place exceed 6 tcf, the gas column is approximately 1400 m ( 4593 ft), and the aquifer below the gas column is normally pressured. An extensive well test program from surface and bottom-hole measurements indicates that the field is in widespread pressure communication across all reservoir layers and most structural domains through an extensive network of faults and natural fractures. The pre- Cenozoic basement has no primary conventional porosity, and 100% of its permeability comes from fractures of all scales. Stratigraphic unit 2 has primary porosities of 8 to 14% and permeabilities of 0 to 8 md. Stratigraphic unit 3 has primary porosities of 4 to 8% and permeabilities of 0.5 to 5 md. FRACTURES, STRESS, AND RESERVOIR PERFORMANCE The Suban field reservoir has been evaluated using 6.3 km (3.9 mi) of wellbore and petrophysical data distributed across 11 wells and includes image logs, production logs, and other data that are valuable for fracture and stress characterization. One hundred sixty meters of full core sampled from stratigraphic intervals SU1 to SU3 was also obtained. Ten wells in the field have subsurface data allowing detailed fracture and stress characterization. The data were studied using a consistent approach by thesameteamtomitigatethesubjectivityandambiguity that is introduced when multiple data analysts work in a disconnected time frame. Well-log and drilling data consist of electrical image (Formation MicroImager [FMI]), production (PLT), photoelectric absorption (PE), mud loss (ML), and total mud gas (TG). The data set used for reservoir performance analysis is not discussed here, but it consists of 26 single-well production tests, 4 interference tests, and frequent wellhead and bottomhole pressure tests made during production. Fracture Characterization and Hydraulic Productivity Stratigraphic units in the gas column are ubiquitously fractured on all scales. Formation Micro- Imager data were analyzed visually for the presence of wellbore-crossing, planar, and continuous features that are diagnostic of natural fractures. The fracture width and appearance on the wellbore image logs were used to categorize fractures into three petrophysical types: 1. Strong-resistivity contrast; large, multiple or complex aperture; fault zone or fracture cluster 2. Moderate-resistivity contrast, well-defined aperture, significant fracture 3. Low-resistivity contrast with poorly defined aperture and minor fracture, presumably mineralized Table 1 summarizes the results of the fracture interpretation. A data integration step was conducted in an attempt to ascertain the fraction and location of total natural fractures from the wellbore image interpretation that are important to well deliverability and reservoir performance. This analysis compared the occurrence of FMI fracture types 1 to 3 with ML, PE, PLT, and TG variations. In this context, significant and discrete ML events signify the presence of permeable fractures. The PLT log measures the rate of influx of gas into the wellbore where localized high velocities are indicative of flow from fractures. The PE, the photoelectric effect from gamma-ray absorption, is a measure of the invasion of barite-rich drilling mud into fractures with aperture. The TG curve measures the gas content of the circulating drilling mud. Rapid changes in gas concentration are indicative of permeable fractures producing gas into the mud stream. We show results from well D2 as an example of the characterization method that was applied to all wells in the field that have appropriate data (Figure 3). As discussed in Nelson (2001), we contend that thorough and Hennings et al. 757

6 758 Suban Gas Field, Sumatra, Indonesia

7 Figure 2. Structure map of the top of the Suban reservoir interval with warm colors representing structural highs (depth subsea, contours in meters). The surface represents the top of the productive reservoir that consists of stratigraphic units 1 and 2 (SU1 and SU2) but also includes the Batu Raja reefal units of SU3a on the southwestern side of the field (the dashed line shows the presumed eastern limit of the Batu Raja reefal unit). Map highlights major faults identified from seismic and well interpretation. Stereonets show equal area lower hemisphere plots of poles to the fracture types indicated from the wellbore analysis. Primary and secondary wellbore stress directions are also indicated in the stereonets. Reservoir pressure interference communication times are indicated for three well pairs. Table 1. Tabulation of Well Performance and Fracture Characterization by Well* Well D R 2 * Hennings et al. 759 Well performance (AOF, bcf/day) Wellbore-reservoir contact length (m) Total type 1 3 fractures Productive fractures Productive fractures w/o well Productive fractures, critically stressed (m 0.6) Type1+2fractures,criticallystressed(m 0.5) Type1+2fractures,criticallystressed(m 0.6) Type1+2fractures,criticallystressed(m 0.7) Type fractures, not critically stressed (m < 0.6) Ratio of type to critically stressed fractures (m 0.6) (%) *Of special significance is the coefficient of determination (R 2 ) relating the various fracture classes to the respective well performance.

8 760 Suban Gas Field, Sumatra, Indonesia

9 Figure 3. Example of the wellbore data compilation, fracture and stress characterization, and geomechanical analysis conducted for all wells in the study (well D2 shown). Included in the compilation are petrophysical curves, lithology, drilling, and production indicators of fracture permeability, fracture interpretation, stress orientation and magnitude, and geomechanical analysis. MD = measured depth; TVD = total vertical depth; RHOB = density (in grams per cubic centimeter); DT = sonic traveltime (in microseconds per foot);gr = gamma ray (API); PE = photoelectric effect (API); TG = total gas (API); PLT = production log (%); SH = maximum horizontal stress; Sh = minimum horizontal stress; Sv = vertical stress; Pp = pore pressure; SU = stratigraphic unit. consistent integration and visual compilation of all data relating to fracture occurrence, permeability signature, and geomechanical character are essential in characterizing the hydraulic nature of fracture systems. On the left of Figure 3, we show typical log data and the lithology as determined from drill cuttings and core. Indicators of potential fracture hydraulic importance (ML, PE, PLT, and TG) are shown on the right of the lithology. Fractures picked from the image data are shown as tracks of cumulative occurrence as well as fracture intensity. These curves are shown with Terzaghi correction applied to compensate for one-dimensional sampling bias (Terzaghi, 1965; Peacock et al., 2003). Fracture intensities, reported as the number intersecting the wellbore per 5 m (16.4 ft), generally increase downward. Stratigraphic unit 2 and SU3a have fracture intensities in the range of 1 to 7 per 5 m and an average of 4 per 5 m. Fracture intensities in basement range from 1 to 15 per 5 m, with an average of 8 per 5 m. We assume that fractures identified from the FMI log may significantly contribute to production if they are associated with anomalies in one or more of the ML, PE, PLT, and TG data sets. In contrast to wellbore image data that can locate fractures with centimeter accuracy, these indicators of hydraulic contribution operate at a resolution of 1 to tens of meters and are, therefore, quite general. For this reason, we choose to discuss fracture intensities at the scale of 5-m sampling windows as measured along the wellbore. An additional limitation is that ML, PE, and TG data are all influenced by drilling and mud-handling techniques that are difficult to interpret quantitatively. The result of our data integration is a semiquantitative discrimination of hydraulically significant fractures (productive fractures) from the total population of fracture types 1 to 2 (see cumulative fracture occurrence and fracture intensity curves, Figure 3). The number of productive fractures ranges from a high of 143 in well 4 (17.9% of the total fractures in that well) to a low of 3 in well 5 (0.3% of total fractures) (Table 1). Although they most certainly contribute to or dominate distributed gas storage in SU1, type 3 fractures are not considered in our geomechanical analysis because of our interpretation that they Hennings et al. 761

10 are partially mineralized, are a manifestation of igneous or metamorphic fabric, or have no hydraulically contributing aperture. As discussed below, the total number of fractures that we have identified as being hydraulically productive in a given wellbore does not have as strong a correlation to well performance. However, the number of critically stressed fractures does have a strong correlation. Stress Characterization Moos and Zoback (1990), Zoback et al. (2003), and Zoback (2007) describe methods for using wellbore image data and additional information to constrain the magnitudes and directions of the horizontal principal stresses. The technique uses observations of tensile and compressional failures of the wellbore walls integrated with estimations of rock strength, overburden stress, and mud fluid pressure. Tensile failures are manifest as small fractures in the borehole wall that strike in the direction of the maximum horizontal stress (S H ) and occur where the circumferential hoop stress exceeds the tensile strength of the rock. Compressive failures are manifested as enlargements of the borehole ( breakouts) caused by shear fractures and spalling in the orientation of the minimum horizontal stress (S h ). In our study, we use FMI data to identify tensile failures and breakouts in the reservoir interval. With the exception of wellbores D2 and 11, which had only tensile fractures with which to constrain the S H azimuth, all other wellbores had tensile fractures and breakouts. Vertical stress gradients (S v ) were estimated from an integration of density logs from shallow levels through the reservoir interval. Pore pressure (P p ) was interpreted from drillstem test data. Rock strength in the reservoir was estimated from sonic logs using the empirical relationship of Hickman and Zoback (2003) that relates unconfined compressive strength (UCS) to compressional sonic velocity of intact rock. This resulted in UCS ranging between 190 and 215 MPa (27,557 31,183 psi). The relationship of Hoek and Brown (1980) was used to reduce the calculated UCS to account for fractures. This step was constrained by a UCS laboratory test on core from well 6 that failed at 109 MPa (15,809 psi) along preexisting type 3 fractures. Fieldwide, our estimates for UCS range between 100 and 210 MPa (14,504 30,458 psi), with a median value of 160 MPa (23,206 psi). Generalizing our rock strength analysis, we calculate mean estimates of UCS of 140 MPa (20,305 psi) at well 4 in the southwestern structural domain and 170MPa(24,656psi)atwell6inthecentraldomain. Figure 3 illustrates the results of the D2 wellbore stress analysis that again serves as our example. In the case of D2, a consistent S H azimuth of approximately 128 from 2247 to 2383 m ( ft) exists. At approximately 2400 m ( 7870 ft), an abrupt shift in azimuth to 164 for 24 m (79 ft) exists that coincides with a structural or lithologic boundary. As discussed in a subsequent section, we believe that this zone is a small but seismically mappable fault. Tensile fractures disappear approximately 30 m ( 98 ft) downhole from this boundary, but the S H azimuth remains at 164 until the cluster of tensile fractures at approximately 2490 m ( 8170 ft). We interpret that the 164 azimuth is constant over this interval based on analysis for well 11, which is 500 m (1640 ft) from D2 and has the same S H azimuth at this depth. The complexity observed at approximately 2490 m ( 8170 ft) and below is interpreted to coincide with other small faults. Based on the relative magnitudes of principal stresses, the region of well D2 is in a strikeslip stress state (S H > S v > S h ) and the S H -S h gradient ranges from 12 to 16 kpa/m ( psi/ ft). A slightly reduced gradient of S H is interpreted in the depth interval of approximately 2400 to 2490 m ( ft) because of the smaller magnitude of horizontal stress difference that is required to create observed tensile fractures with an azimuth of approximately 164 compared with approximately 128. In summary, well D2 is subject to two distinct stress states, represented by a variation in both the S H azimuth and the horizontal stress difference. Because of the structural complexity observed in the field, we choose to report our stress results as gradients, allowing the reader to calculate the absolute stress magnitude at any desired depth in the reservoir interval. The wellbore-based stress analysis at Suban reveals significant variability in the orientation and 762 Suban Gas Field, Sumatra, Indonesia

11 architecture of current stress throughout the field (Figure 2). We believe that this variability is partitioned by the structural domains and varies with depth as associated with reservoir-scale features. Stress analysis data from all wells in the southwestern structural domain (wells D2, 4, 8, 11) indicate a primary strike-slip stress state and S H azimuth that roughly parallel the trend of the structural grain of the domain. This represents an approximately 90 rotation from the regional northeast north-northeast S H trend identified in the northeastern structural domain of the field and regionally throughout the South Sumatra Basin (Heidbach et al., 2007). Wells D2 and 4 each have secondary northerly trending S H azimuths that are related to minor faulting. With the exception of the bottom interval of well 7, all other wells in the field (2, 3, 5, 6, 7, and 9) are in a thrust-fault stress state. Although the dominant S H azimuth in the thrust-fault domain is north-northeast, wells 5 and 9 have a northwestern S H azimuth that is similar to the S H azimuth of the southwestern structural domain. Several wells exhibit a variable S H azimuth with depth. The bottom 12 m (39 ft) of well 7 and a 50-m (164-ft)- thick interval in well 2 also have a northwestern S H azimuth, and well 3 has the most complex local stress heterogeneity, with the S H azimuth rotating from northeast to northwest at least six times over the reservoir interval. Fractures and Stress Using our wellbore stress model, we assess the proximity of each type 1 to 2 fractures to the critical stress state by calculating the ratio of shear stress to normal stress resolved on each fracture surface the slip tendency of Morris et al. (1996). Based on the laboratory friction experiments of Byerlee (1978) and the analyses of Townend and Zoback (2000) and Zoback and Townend (2001), who studied stress states of fractures in crystalline rocks in settings of active deformation, we consider the coefficient of sliding friction (m) of fractures of 0.6 to represent the potential onset of stress-induced slip that may produce dilation and, therefore, enhance permeability. Fractures with m 0.6 are considered critically stressed. For example, well D2 has 45 productive fractures and 91 critically stressed fractures (Table 1). Ten of the productive fractures are also critically stressed (Figure 3, right track). We graphically plot the results of our productive and critically stressed fracture analysis in stereonet form in Figure 2 to illustrate how our characterized fractures group spatially. For the analysis of stress acting on individual fractures, we used the appropriate depth-specific stress state from the wellbore geomechanical solutions. The resulting variability in the orientation of critically stressed fractures can be seen when comparing wells, such as 11, which has a uniform stress architecture with depth, and wells D2, 2, 3, 4, and 7, which have variation in stress orientation and magnitude. Stress state is the most significant control on the spatial patterns. Wells in the northeastern and central domains of the field (thrust-fault stress states) have critically stressed fractures with dips approximately 20 to 40 (fracture poles near the stereonet center) and variable strikes. Well 6, as an example, has 117 critically stressed fractures, including 3 that are hydraulically productive, with a dominant northwestern strike. Wells in the southwestern domain (strike-slip stress states) have critically stressed fractures with dips approximately 65 to 90 and a strong propensity for north-northeastern to northwestern strikes. This knowledge was used when planning wells 10 and 11, such that their deviated paths transected as many of the suspected productive fractures as possible. For example, well 11 has 2 welldefined clusters of critically stressed fractures with northerly strikes and steep dips, 10 productive fractures, and 11 minor fault zones. Relationship to Reservoir Performance Panels A and B of Figure 4 show a tabulation of our wellbore fracture characterization compared with absolute open flow (AOF) estimates from each well. The AOF is the maximum production rate (performance) a well can theoretically deliver while flowing against zero pressure. We find some surprising relationships in comparing the various fracture characterizations with well performance using Hennings et al. 763

12 Figure 4. (A) Well performance measure, estimated absolute open flow (in billion cubic feet per day). (B) Plots of flow performance of select well groupings versus a selection of fracture characterization data from Table 1. R 2 = coefficient of determination. linear regression (Table 1, right column; R 2 =coefficient of determination). As indicated by values of R 2, no clear relationship exists between wellborereservoir contact length and well performance. A positive but weak relationship exists between the total number of fractures (types 1 3) and well performance (R 2 = 0.26). If well 11 is excluded from the regression, then we obtain a moderate relationship between the identified productive fractures and the performance of the remainder of the wells (R 2 = 0.56). This is an admissible consideration because the identification of productive fractures in well 11 was hindered by the absence of PLT data and the use of managed pressure drilling that mitigated mud losses. The best correlation between well productivity and our fracture characterization data is obtained by considering the total number of critically stressed fractures transected by the wellbores. We assess this correlation using a range of m values to account for uncertainty in our knowledge of the frictional coefficient of the fractures in the reservoir. From this simple linear-fit analysis, well performance is clearly most closely controlled by the number of critically stressed fractures transected by the wellbore (R 2 = 0.93) when m 0.6. Our results also reinforce our assumption that a m 0.6, as compared with m 0.5 and m 0.7, captures the most important fraction of contributing fractures. It naturally follows that the ratio of all type 1 to 2 fractures to critically stressed fractures also correlates strongly. In Figure 5, we summarize the stress state and magnitudes derived from each well following Moos and Zoback (1990). We use stress gradient to generalize the results for application in a depthindependent context. We find two general wellbore stress groupings in Suban field. Wells D2, 8, 11, and the lower part of 7 are squarely in a strikeslip stress state. Stress gradients are relatively low in this (A) group with stress differences ranging between 12 and 16 kpa/m ( psi/ft). Wells 4 and 9 plot in the upper area of the group on the transition between strike-slip and thrust-faulting stress regimes. Stress gradients in group B are higher than in A with differences of approximately 20 kpa/m (0.88 psi/ft). Wells 2, 3, 5, 6, and the upper parts of 7 are in a distinct thrust-fault stress style (group B), with the highest gradients in the field and stress differences of approximately 21 kpa/m (0.93 psi/ft). With the exception of well 8, which is in a down-flank position, wells with the highest performance occur in stress group A along the crest of the anticline in the southwestern domain of the field where the hydraulic character benefits from 764 Suban Gas Field, Sumatra, Indonesia

13 stitute the hydraulic backbone of the southwestern domain. RESERVOIR-SCALE FAULTS AS THE HYDRAULIC BACKBONE Fault Interpretation Figure 5. Maximum horizontal stress (S H ) versus minimum horizontal stress (S h ) gradient plot summarizing the wellbore stress state for wells in the Suban field. The shaded polygon represents possible stress states (total stress) and related faulting style, and the ellipses show the results of our wellbore stress analysis. The wells can be generalized to occur in two regions (A and B) that are controlled by structural domains. The outside boundary of the polygon is constrained by m =0.8.Them =0.6 boundary is also indicated. S v = vertical stress; m = coefficient of sliding friction. having high numbers of critically stressed fractures. Although no fracture or stress data were collected in well 10, its location and strong performance place it in this group. The analysis described thus far constitutes our fracture performance model for Suban field, which is based exclusively on wellbore data. Exploiting the bulk behavior of the field to maximize well performance requires a deeper understanding of the geologic controls on fracture formation. For this goal, we have closely examined the nature and hydraulic implications of the reservoirscale faults in the southwestern structural domain of the field. We conclude that these faults and the fracture damage zones that surround them con- The prestack depth-migrated three-dimensional seismic volume was interpreted in great detail over the southwestern structural domain of the field to ascertain the extent and character of reservoirscale faults for integration with our fracture-based reservoir performance model (Figure 6). The criteria for fault interpretation were consistent with fault offset observable over at least three consecutive inlines or crosslines and a geologically reasonable fault surface geometry and throw variation. The interpretation yielded 27 seismically mappable faults, all with reverse separation, with map lengths ranging between 50 m (164 ft) and 1.2 km (0.8 mi), and strikes that are subparallel to the master fault of the southwestern domain (Figure 7). The faults are concentrated in a 1 8 km (0.6 5 mi) area along the crest of the anticline that forms the southwestern domain. The faults can be divided into two sets based on strike: a north-northwest set of 10 faults and a northwest set of 17 faults. The faults dip between 55 and 80 and are predominantly parallel to the master fault, although several have antithetic dips. Maximum fault dip slip ranges between 8 and 180 m ( ft) with a linear regression of dip slip to map length of 0.08 and a coefficient of determination of 0.9. For most of the faults, the upper tip cuts through SU1 to SU3 and into SU4. We are uncertain about the downward extent and trajectory of most of the faults because of poor seismic reflectivity at depth in SU1; therefore, we have taken a conservative approach and interpreted only fault surfaces for which we have direct seismic evidence (Figure 6). Our fracture model reinforces our interpretation of the reservoir-scale faults. Well D2, for example, had significant drilling mud loss events at 2122, 2303, and 2451 m (6962, 7556, and 8041 ft) measured depth, each generally coinciding with Hennings et al. 765

14 Figure 6. Example of the seismic expression of the reservoir-scale faults that were constrained by detailed threedimensional seismic structural interpretation and correlated to all wells in the southwestern structural domain. Seismic profile is in depth and corresponds to the well 11 profile in Figure 7. SU = stratigraphic unit. reservoir-scale faults, and clusters of hydraulically significant fractures (Figure 7). Well 11 mirrors this behavior and equally reinforces the interpretation. The trajectory of well 8 is parallel to one of the reservoir-scale faults, approaching it at depth. In this well, we observe an overall increase in fracture intensity as the well approaches the fault. Combining our reservoir-scale fault interpretation with our wellbore fracture model provides us with a multitiered concept for the southwestern domain that consists of productive fractures of tectonic origin distributed throughout the reservoir volume and clusters of especially productive fractures in halos around reservoir-scale faults. This interpretation is supported by well to well pressure interference analysis. Figure 2 shows interference data for three well pairs in the field. The time required for a pressure pulse to travel between wells 4 and D2, spaced 4 km (2.5 mi) apart, in the southwestern domain is 12 hr. This is in contrast to the measured 20.8 days for communication between wells 4 and 6 that lie 3 km (1.9 mi) apart along a northeast-southwestern azimuth. Connectivity betweenwells2and5intheeasterndomain(13days) is also greatly reduced compared with the southwestern domain. Clearly, the northwest-southeast fracture connectivity along the crest of the southwestern domain is greatly accentuated compared with elsewhere in the field. Clearly, from the geometry of the fault-related folding in the hanging wall of the southwestern structural domain, significant dip slip has occurred (Figure 6).The reservoir-scale faults are most densely developed adjacent and southeast of a bend in the master fault where it strikes more to the northwest compared with its principal north-northwestern 766 Suban Gas Field, Sumatra, Indonesia

15 Figure 7. Three-dimensional structural model of the southwestern domain, viewed obliquely from the south. The model highlights the seismic interpretation of reservoir-scale faults along the crest of the anticline that represent the permeability backbone of the domain. Cross sections are sliced to show the trajectory and structural context of wells in the southwestern domain. Highlighted along the wellbores are locations of significant drilling fluid loss signifying faulted or heavily fractured zones. Fracture intensity is also shown as log tracks along the wellbores. strike (Figure 7). This pattern and style of faulting are compatible with expectations of deformation that are concentrated against a restraining bend in a dextral strike slip or transpressional system (Gamond, 1987; Sylvester, 1988; McClay and Massimo, 2001). This interpretation is strengthened by the regional existence of dextral movement along northwest-striking faults in the South Sumatra Basin (Pulunggono, 1986) and the northwesttrending present-day wellbore measurements of S H for the southwestern structural domain of the field. Combining these observations leads us to propose an oblique-slip origin for the southwestern domain with slip on the master fault, anticlinal folding, and reservoir-scale faulting forming together. A less likely scenario would be dip slip along the master fault, accompanied by most of the folding and reservoir-scale faulting, followed by strike-slip movement and subsequent accentuation of deformation. Fault Stress Model A final modeling step in our analysis addresses how the current stress state in the southwestern structural domain might affect the reservoir-scale faults. The approach taken is similar to our fracture stress analysis, whereby the normal and shear stress acting on each triangular element of a fault is determined. We assume m = 0.6 and stress orientations and gradients generalized from wells 4, 8, 11, and Hennings et al. 767

16 Figure 8. Fault stress analysis. (A) Oblique view and down of the master fault of the southwestern domain. The reservoir-scale faults are colored by their slip likelihood. Green colors represent parts of the fault surfaces that have pore pressure in excess of what is required to cause slip. Blue areas require additional pore pressure to cause slip. S H = maximum horizontal stress; S h = minimum horizontal stress; S v = vertical stress; P p = pore pressure; S H az = azimuth of maximum horizontal stress; m = coefficient of sliding friction. (B) Cumulative distribution of slip likelihood results for the reservoir-scale faults in the southwestern domain, indicating that fault strike is a primary control on slip likelihood, with the northwest-striking fault family being more prone to slip. D2 (see Figure 5, stress state A). To visualize and discuss the results, we show the additional (+) or excess ( ) pore pressure required to reach the critical stress state at which failure occurs for each fault element. This parameter is referred to as slip likelihood (Figure 8A). This allows us to use a single scalar measure with which to compare the slip likelihood of our population of reservoir-scale faults. At a depth of 2.5 km (1.6 mi), the reservoir pore pressure is approximately 30 MPa (4350 psi). The analysis shows that the pore pressure required to activate slip ranges between a pore pressure of 22 and 40 MPa( psi) or slip likelihood of 8 to10mpa( 1160 to 1450 psi) (Figure 8B). The predicted slip behavior is mainly dependent on the strike and dip of each individual fault segment, but because the faults are generally planar, they tend to act uniformly across their surface area. Northwest-striking faults generally have higher slip likelihoods (lower required P p ), meaning that the existing pore pressure is close to or more than sufficient for making their slip geologically likely. The north-northwest striking faults generally require 2 to10 MPa ( psi) additional pore pressure to become potentially active. By surface area, approximately 40% of all the faults are potentially active. This value increases to 90% if an additional 5 MPa (725 psi) of pore pressure was somehow added to the system or if other geomechanical parameters were changed slightly. Well D2 drilled through 2 faults each with slip likelihoods between 3 and 4 MPa( psi). Within surface location and drilling constraints, wells 10 and 11 specifically targeted, and subsequently penetrated, parts of the reservoir faults with higher slip likelihoods. As discussed below, these wells have performance 768 Suban Gas Field, Sumatra, Indonesia

17 estimates that are three to seven times higher than any other wells previously drilled in the Suban field. DISCUSSION Stress Variability In the absence of ample wellbore pressure leak-off and laboratory rock strength data, integrating observations from wellbore images to develop a model of stress magnitude is an iterative process. Uncertainty in the calculated stress magnitude relates mainly to the assumption of m. For wellbores in our study that have both tensile fractures and breakouts, we find that using a m < 0.6 is inappropriate because it provides a solution window for stress magnitude that is too narrow to explain the wellbore deformation features observed. Our approach is to assume a m value of 0.8 as the upper limit of frictional strength for the Suban fractured reservoir rock that allows us to compare wells across the field (Figure 5). Published accounts of stress rotations within individual fields generally show variations in S H azimuth of 20 to 30 as compared with regional trends (Barton and Zoback, 1994). Tamagawa and Pollard (2008) document one well in their fractured granite gas field that has a 90 rotation in S H and conclude that the most significant stress rotation occurs in the tip regions of active faults. In the Suban field, we find that stress perturbations occur at two scales. Local rotations in S H azimuth and alterations in stress gradient occur commonly in our data set and can be related to proximity of reservoir-scale faults. The more fundamental observation is that S H azimuth, stress style, and stress gradient magnitude are strongly partitioned into structural domains, which in turn are controlled by the larger faults that form the overall structure of the field. We do not propose a definitive mechanism to explain the variations in stress between the structural domains, but we offer discussion on four scenarios. The first scenario is that the overall uplift that formed the field is undergoing complex deformation produced by a superposition of regionally driven thrust and strike-slip elements that naturally partition the field into distinct stress domains. The second is that local folding of the southwestern structural domain above its master fault is causing a reorganization of the principal stress axes whereby outer arc stretching counteracts the regional S H, thereby reducing it to the least compressive principal stress. This scenario is strengthened in the observationthatwells2and3inmoderatelyfolded areas of the (thrusting) central and eastern domains also show some tendency toward this behavior in parts of their wellbores. The third scenario is coseismic stress relaxation. Here, the master fault of the southwestern structural domain has recently experienced slip and associated reductions in stress gradient and reorientation in azimuth in a fashion similar to that described by Healy et al. (2004) in their analysis of the 1980 El Asnam earthquake rupture. Consistent with this scenario is our observation that stress gradients are lower in the southwesternstructuraldomainthaninthecentraland northeastern domains; however, a greater degree of fracturing and overall deformation in this domain could produce the same effect. The fourth scenario follows Tamagawa and Pollard (2008) in proposing that stress orientations change in the vicinity of active faults, especially their tip zones, although they do not also document wholesale reorganization of the stress architecture as we have observed at Suban field. These four scenarios explaining stress heterogeneity are not unique, and it is probable that some combinations of these or other factors act in concert. Application to Reservoir Productivity and Drilling The concept of potentially active faults and damage zones rich with critically stressed fractures forms the basis for our exploitation strategy for future development drilling at Suban field. This concept was tested in the southwestern domain by wells 10 and 11. Wellbore fracture characterization and productivity data was collected in well 11. Figure 4B shows a regression of reservoir productivity for all wells in the field before drilling wells 10 and 11, which suggests a positive relationship between the number of critically stressed fractures and well performance Hennings et al. 769

18 Figure 9. Generalized fault and fracture (effective) stress summary for southwestern and central/northeastern structural domains illustrated with three-dimensional Mohr-Coulomb plots for frictional faulting and lower hemisphere stereonets, both contoured for slip likelihood. The diagrams are shown for a depth of 2500 m (8200 ft). Poles to the three fault surfaces intersected by wellbore 11 are also plotted in A. S H = maximum horizontal stress; S h = minimum horizontal stress; S v = vertical stress; m = coefficient of sliding friction. (R 2 = 0.56). Considering only the wells in the southwestern domain, all in a strike-slip stress state, suggests a stronger positive relationship locally (R 2 = 0.63). The engineering and geologic characterization of strong fracture-dominated permeability and preferential northwest-southeast connectivity along critically stressed faults inspired locating the new wellsinaclusteraroundwelld2tominimizethe surface footprint. The well trajectories were planned to deviate to the northeast in the reservoir to maximize the intersection of reservoir-scale faults and their associated fracture damage zones. The wells were drilled with slightly larger diameters compared with previous wells and used a downhole valve to manage mud pressure, thereby minimizing mud losses and potential degradation of flow potential. Drilling data for well 10 indicates that it encountered three heavily fractured fault zones (Figure 7). The drill string instantaneously dropped approximately 1 m ( 3.28 ft) while traversing the second of these fault zones, demonstrating that it has significant aperture at this location. Drilling was terminated after crossing the third fault zone, which was specifically targeted for its stress character, when mud losses and gas influx exceeded operating limits. Well 11 also encountered three zones interpreted as faults. Tests of the 10 and 11 wells indicate that the parts of the reservoir penetrated have extremely favorable bulk properties that we attribute to the permeability architecture of the intersected faults and their associated fractures. Tapping into the system in this manner 770 Suban Gas Field, Sumatra, Indonesia