3D Modeling, Horizontal Drilling
Give New Life to Aging Fields
by Christopher C. Phillips, Donald D. Clarke and Linji Y. An
Christopher C. Phillips is chief geologist with Tidelands Oil Production Company. He has been employed by the company since 1981 in various areas of development geology and engineering. Phillips holds a B.S. in geology from California State University, Long Beach.
Donald D. Clarke is division engineer of geology and subsidence with the Long Beach Department of Oil Properties, and has been employed by the city of Long Beach since 1981. He holds a B.S. in geology from California State University, Northridge.
Linji Y. An is a research associate at the University of Southern California, and a consultant for INTERA Inc. An holds a B.S. in geology from Chengdu Institute of Geology, an M.S. in geochemistry from the Graduate School of Academia Sinica, China, and a Ph.D. in earth sciences from USC.
Posted with permission from the American Oil & Gas Reporter (September 1996)
LONG BEACH, CA.—Advanced reservoir modeling and characterization, and state-of-the-art stratigraphic bit control technology can breathe new life in mature and thin-zone production areas when otherwise independent technologies are integrated to create a synergy between geological knowledge and drilling capability.
A project aimed at increasing heavy oil reserves in the Wilmington Field in Southern California through advanced reservoir characterization and thermal production technologies called for the project team to merge data from old and new wells, and create a three-dimensional deterministic model to allow cross sections to be cut along the horizontal well course and be used for geosteering. Once the wells were precisely placed, new completion techniques were applied to optimize steam-assisted gravity drainage.
Planning and drilling horizontal wells in the 64-year-old field requires excellent data control. With as much as 29 feet of total subsidence, defining geological features depends on the ability to correct directional well survey data through time.
The project was launched in 1995 as a U.S. Department of Energy Class III reservoir cost-share project (DE-FC22 95BC14939) aimed at extending thermal recovery in the Fault Block II-A Tar zone. Project partners include Tidelands Oil Production Company, the City of Long Beach Department of Oil Properties, the University of Southern California, and David K. Davies and Associates.
To help offset the high up-front capital costs of a new steam flood, management and engineering strategies include:
- Capitalizing on accurate reservoir description using 3D modeling and visualization;
- Consolidating drill sites;
- Drilling horizontal wells to reach under local industry; and
- Initiating steam-assisted gravity drainage.
Historically, steam flooding of the Fault Block II-A Tar zone has been relatively inefficient because of producibility problems common in slope and basin clastic (SBC) reservoirs. Inadequate characterization of the heterogeneous turbidite sands, high-permeability ″thief″ zones, and non-uniform distribution of remaining low-gravity (13 degree API) crude oil have contributed to poor sweep efficiency, high steam-to-oil ratios, and early steam breakthrough.
Operational problems related to steam breakthrough, high reservoir pressure, and unconsolidated formation sands have caused premature well and downhole equipment failures. In aggregate, these constraints have resulted in increased operating costs and decreased recoverable reserves. To revitalize the field, a development program has been devised that combines detailed reservoir modeling with advanced horizontal techniques.
Defining the Geology
Staff geologists collected all old well data and constructed a comprehensive digital data base in order to develop a three-dimensional, deterministic geological model. The intent was to rigorously define the geology so that horizontal wells could be placed with great accuracy in the heavy oil sands to facilitate steam-assisted gravity drainage. Eighteen horizons were defined and stratigraphy was correlated in more than 600 wells. Most of the stratigraphic section is represented by the log shown as Figure 1. The primary correlation used the spontaneous potential (SP) and induction curves from wells drilled between 1937 and 1994.
Figure 1—SP and induction curves were used for detailed reservoir characterization work in the West Wilmington Field, as represented in this type section. The goal was to correlate units that are in hydrologic continuity.
Reservoir characterization was achieved by correlating logs from vertical and directional wells, and integrating them with cross sections, from the 3D model. The characterization included the discovery of a paleochannel (Figure 2) and onlap sands (Figure 3), such as the DIA, DIC, DIE and DIF subsands. The target Dl sand was divided into seven subsands over the entire block. Based on shale and sand relationships seen on the SP and induction log curves, the Dl, DIA, DIB, DIC, DID, DIE and DIF subsands were defined.
Figure 2
Figure 3—Visualization of the D1F Sand Onlapping the D2 Shale from the South.
A waterflood was initiated in the West Wilmington Field in 1959. The stratigraphic marker depths for pre-waterflood wells were adjusted downward to account for post-drilling subsidence. This adjustment was necessary to facilitate intersecting the newer wells with the target horizons at the predicted depths. A computer program inherited from the previous field operator applied a single subsidence correction so that all field data would be compatible, no matter when a particular well was drilled and logged.
Building the 3D Model
To create a geological model accurate enough for horizontal drilling, all field data had to be verified and faults and horizons precisely defined. To optimize the process, a geological data base and a 3D geological mapping and modeling package were used. Accurate fault depiction was critical, because faults define the boundaries used for geologic modeling and reservoir simulation. Therefore, a data set of fault picks was gathered, with each data point carrying a well name, x, y, z coordinates, stratigraphic section discrepancy (missing or repeat section), and fault name.
It was convenient to compare fault geometries by using the 3D software to create 3D fault plane models. The data points for each fault plane were checked for accuracy by posting the data on the fault surface (Figure 4). Irregularities in the fault surface as a result of data problems could be identified easily in the 3D Viewer. Although most of the fault picks were accurate, all needed verification.
Figure 4—3D Fault Plane Model of Fault Picks on the Ford Fault.
Vintage surveys (those taken in the dog house instead of down hole, for example) were responsible for most of the question able fault picks. Once obvious errors were removed, the faults were refined by performing an intersecting well search and back interpolating the nearest survey point onto the fault plane. It was possible to identify additional points, and also define regions of ″no throw.″ Defining fault extent is important; it enables an accurate geological model to be built that defines reservoir compartments.
In mid-1995, a detailed, ten-layer 3D geological model was constructed to plan nine new wells in Fault Block II (Figure 5), including five observation wells, two of which were cored. After the first three observation wells—OB2-001, OB2-002 and OB2-004—were drilled, project partners noticed that the stratigraphic markers for OB2-001 and OB2-002 appeared to have anomalous vertical subsea depths. They believed the wells′ proximity to the Ford Fault caused the structural misfit.
Figure 5—This 3D deterministic model shows the relationships among the observation wells and the horizontal wellsdrilled in the pilot area.
Everything associated with the error was checked, including drilling and surveying measurements. It was perplexing when the first well drilled (OB2-004) agreed within one foot, but subsequent wells were off by as many as 18 feet. Finally, when the old datum maps were compared with the 1995 maps, it became clear that the problem was in the data. The new maps had not been properly corrected for subsidence.
Because of the large number of old wells and inadequate subsidence correction, contouring the horizon data resulted in maps more representative of a 1940 than a 1995 structure. The reason for these discrepancies was clear after drilling the first core hole (OB2-003)—the markers were as much as 17 feet too deep with respect to the 3D model. The directional data needed to be corrected for subsidence through time in order to define current structure and enable newly-drilled horizontal wells to be accurately placed. Although the subsidence software used did make these corrections, they were in error.
Appropriate subsidence corrections were estimated manually and applied to the second core hole. Results were excellent, and the core point was accurately predicted within one foot. Before kicking off an initial horizontal well (a rig had been contracted to spud within one month), field data had to be corrected, a new geological model built, cross sections created, and all structure maps reconstructed. Needless to say, it was a challenge to complete these steps without having to put the rig on standby status.
Correcting for Subsidence
The subsidence at the West Wilmington Field occurred gradually. Ground elevation dropped because internal thickness between horizons was reduced as oil was withdrawn. Therefore, not only did surface elevation for each well have to be corrected, but the difference between markers needed regular adjustment because of internal compaction.
After studying the data, it was determined that two corrections were required: normalizing existing wellheads to a consistent datum, and artificially moving the surveys down to account for the varying amounts of subsidence. The severity of this problem can be seen in Figure 6, which shows the horizontal well courses overlaid on the subsidence bowl contours (to tal subsidence). Total subsidence of 1222 feet affects the location of the stratigraphic horizons. The Dl sand is about 60 feet thick in the producing intervals, so an error of 20 feet could be disastrous either the zone would be missed, or 30 percent of the producing sands would be lost.
Figure 6—Shows horizontal well paths on top of total surface subsidence contours (in feet). Figure 7, below, displays well test results showing D sands to be in hydrologic continuity and separate from upper and lower (T and F) reservoirs.
Tidelands also discovered another needed correction. The subsidence program was making only one correction, and it was inaccurate. Thus, sample data sets were created, the erroneous correction was backed out of the data, and the first two corrections were incorporated, including a new data field-kelly bushing elevation (KBE) datum-for wellhead datum normalization. Horizons were gridded from the corrected data and compared to new observation wells. Old and new data fit with good consistency, and the subsidence program was recoded to account for the corrections.
Subsidence-related problems can adversely affect productive clastic reservoirs with long development histories. Operators attempting horizontal well development in such fields should carefully evaluate and correct field data for intra-reservoir compaction. So-called ″bad data″ can thereby become very accurate and useful.
Drilling, Coring and Logging
Three observation wells and two core hole/observation wells were drilled in the steam-drive pilot area in the summer of 1995. Four wells are for monitoring reservoir temperatures and the fifth was placed to determine post-steam oil saturations and mineral alterations to the formation rocks.
Core recovery through the T and D subzones was excellent, with more than 99 percent of the planned 517-foot core interval recovered. A well testing tool (Repeat Formation Tester) run in observation well OB2-001 showed that reservoir pressures within the Dl subzones were in vertical pressure communication and differed from the subzones directly above and below (Figure 7). This was significant, because it meant a horizontal well could be placed in the bottom of the Dl sands and oil could be efficiently produced from all the Dl subsands via gravity steam-drive.
After completing the observation wells, four horizontal wells were drilled in the pilot area—two producers and two steam injectors—with the goal of placing the horizontal section near the base of the DID sand (Figure 8). Area strata are not horizontal; they dip 6–9 degrees to the west.
Figure 8—Eighteen layers provided the detail needed to drill the West Wilmington horizontal wells. This cross-section, taken from the 3D deterministic model, shows four of the seven mapped faults, logs alongside four vertical wells, and horizontal well UP955 with its perforations in the D1 sand.
A geologist was on location during drilling to monitor progress and determine bit location with respect to the geological section being cut. Logging while drilling and measurement while drilling were used for directional⁄geological control (geosteering). The induction resistivity tool was 30 feet behind the bit, and the MWD 62 feet behind the bit. The MWD survey was plotted on the geological cross section, and bit position in the geological section was located by extrapolating 62 feet ahead of survey points. Once bit position was known, the LWD resistivity and gamma ray values were compared to offset logs and cross sections to verify that the horizontal section was in the correct location.
The D2 shale is at the base of the DID sand in the area where the horizontal wells were drilled.
The offset logs from the observation wells showed good 10 ohm-m resistivity at the base of the DID sand. The resistivity measurements of the horizontal wells appeared to be 1.2 times higher than those in similar sections in observation wells. When drilling horizontally, the induction curves separate as the logging sensors (Compensated Dual-Resistivity tool) approach the D2 shale. As shown in Figure 9, at 10 ohm-m, the depth of investigation for the attenuation deep resistivity (Rad) is about 43 inches, and the depth of investigation for the phase shift shallow resistivity (Rps) is about 24 inches.
Figure 9—Note: Chart provided by ″Logging While Drilling,″ Schlumberger Educational Services.
Curves Separate
If drilling subparallel to bedding near a two- to four-foot shale⁄sand boundary, the curves separate because of the Rad measuring the resistivity of the adjacent horizon. This can be seen between the measured depths of 3,850 and 3,960 on the cross section and log for well UP955 (Figures 8 and 10). In this case, the Rad tool is reading the resistivity of the top of the sand and the low-resistivity shale above, while the Rps is measuring only the higher resistivity of the upper DID sand.
Figure 10—This LWD log from UP955 shows separation of the resistivity curves at about 3,900 feet, which indicates the BHA is approaching bed boundary, and must be steered away from the top of the sand for perfect placement.
The well plan was adjusted to follow the base of the DID sand. Theoretically, the lower a well is located in the DID sand, the more recoverable reserves are available for gravity drainage. Assuming none of the reserves below the well bore are recoverable, every foot of pay between the well bore and the base of the DID sand is equivalent to 15,876 stock tank barrels of lost reserves.
In the pilot area where the four horizontal wells were drilled, only the Dl, DIB and DID sands were initially modeled. Horizontal continuity was verified for the DID sand by the consistent LWD readings. The fault sub-blocks are down-dropped to the east in Fault Block 11, and the known faults were observed at their predicted locations. The exception to this was well UP956. A change in resistivity from 12–13 ohms to 19–20 ohms was unexpectedly observed about 850 feet from total depth suggesting that the hole was as much as 10 feet higher in the stratigraphic section, and encountered an unmapped fault of small displacement with an oil bank next to the fault.
While the goal of staying in the lower DID sand was not achieved in well UP955, the experience gained drilling the well allowed an offset to be drilled with incredible efficiency. Fault Block V horizontal well T5I-002 was drilled in late 1995, and was a major success. Accurate stratigraphic control of the bit was achieved (the well bore was within five feet of the lower shale), and only six days were needed to drill the well at rates up to 600 feet an hour (Figure 11). Achieving these results required not only the modeling and subsidence correction efforts, but also good communication between the LWD⁄MWD operator, mud engineer, geologist and drilling supervisor.
Figure 11—Cross-section Parallel to T51-002.
Completion Techniques
A steam-assisted gravity drainage technique tested extensively in Canada′s heavy oil fields was adapted to operate the West Wilmington horizontal steam-drive wells. In West Wilmington, the technique involves completing the last 600 feet of the horizontal wells in the most up-dip section of the reservoir. The horizontal segments of the Fault Block II pilot wells average 1,300 feet in length, and were drilled from west to east at angles ranging from 96 to 99 degrees to compensate for reservoir dip.
The objective is to concentrate the steam up-dip in a smaller area to take advantage of gravity segregation of the steam, which promotes early development of a steam chest. As the steam chest grows to envelop the initially-completed intervals, more perforations will be opened down-dip and the up-dip perforations may be plugged.
The four pilot horizontal wells in Fault Block II were completed using an alkaline steam⁄hot water sand consolidation technique to test its effectiveness. Between 80,000 and 100,000 cold water equivalent (CWE) barrels of 80 percent quality steam were injected into each well to provide sand consolidation and initial cyclic steam stimulation. Good results are expected since the Dl sands are in hydrostatic communication, and have vertical permeability barriers above and below.
These wells were selectively perforated over the last 600 feet to optimize the productive interval. Based on geologic maps and LWD results, 11 or 12 perforations were selected for each injection well. All faults cut were assumed to be sealing, so perforations were selected to provide the best coverage across the faults. Perforations were shot outside the D2 shale sections (lower red horizon on Figures 8 and 12). The limited-entry perforations allow better distribution of the steam across the horizontal section and aid in sand control.
Figure 12—Cross-section Parallel with UP956.
Sand production is controlled in the conventional wells producing from the Tar zone by expensive open-hole gravel pack, slotted-liner completions. The sanding problem is a general one, unrelated to performing thermal operations. Although the West Wilmington pilot remains under investigation, if the alkaline water⁄steam consolidation technique proves out as expected, it could have wide applicability. This technology could provide a viable and cost-effective completion alternative, especially in areas where steam is readily available.
What once took geologic and drafting teams years to accomplish is now possible in a short time with new computing technology and computer-assisted techniques. Even when catastrophic data problems exist, geological models, maps and cross sections can be completed rapidly and ahead of drilling rig schedules.
The computing system used in the West Wilmington project includes EarthVision 3.0 geospatial modeling software from Dynamic Graphics, Inc.; the NEWILMA data base software application inherited from the previous operator; Ingres™ data base management system from Computer Associates; and IRIS INDIGO with a UNIX 5.2 operating system and 50 megahertz IP20 processor from Silicon Graphics.
Editor′s Note: The preceding was adapted from presentations at the Association of American Petroleum Geologists′ annual convention, held May 20–22 in San Diego. The authors thank Terry Smith, Iraj Ershaghi, Xen Colazas, James Hemphill, Jim Quay, Mark Kapelke and Scott Hara for project support. They also acknowledge Mike Henry, Rudy Payan, Steve Siegwein, Juan Santillan, Nancy Dunn and Keith Jones for their roles in the project, as well as Paul L. White at Dynamic Graphics, Inc. for DGI′s ongoing assistance.