How fracture interference impacts Woodford shale gas production
Integrated digital workflows are becoming increasingly important in understanding complex, unconventional reservoirs and basins. Bringing disciplines together in a single-project environment is central to maximizing productivity through better knowledge sharing, increased collaboration, and a standardized approach. The end goal is always the same—to allow asset teams to make more informed decisions through an optimal understanding of risk. A recent example of how a multidisciplinary digital workflow shaped understanding and developmental planning can be seen in the Woodford shale of the Anadarko basin in Oklahoma.
The Woodford shale was deposited during the late Devonian, along the western margin of the North American Craton. Regional studies indicate that the Cherokee Platform and Nemaha Ridge were positive features at the time, segmenting the Devonian shelf into proto-Anadarko and Arkoma basins, and disrupting deep marine upwelling conditions, resulting in regional facies changes.
The workflow was created by Devon Energy and Schlumberger, to integrate geology, geophysics, petrophysics, geomechanics, stimulation, completion, production, and reservoir engineering. The team wanted to describe the mechanics of the hydraulic fracture stimulation interference and its impact on gas production and drivers, well performance, and recovery in horizontal wells.
The project’s integrated multidisciplinary workflow, built in the Petrel E&P software platform, combined data and interpretations from geology, geophysics, petrophysics, geomechanics, stimulation, completion, production, and reservoir engineering. The workflow allowed for significant productivity increases through simultaneous task completion. While petrophysics validation was undertaken with core data, geomechanical interpretation was started, and the 3D geological model was built, based on well and seismic data.
The seismic response of the top of the Woodford shale is a decrease in acoustic impedance from the overlying Osage, while the base displays a strong peak on amplitude volume, where the shale overlaps underlying Hunton limestone. Faulting is interpreted on the amplitude volumes, and dips are generally near-vertical and predominately strike northerly or easterly. Three different seismic signature zones were observed, and corresponding depth tops from well data were tied to the seismic to produce four horizons: Woodford, middle Woodford, lower Woodford and Hunton. These zones were used as boundaries to extract average velocity interval ratios to inform the petrophysical model.
Calibrated petrophysical parameters were then added to the model. The process then shifted to geomechanics and hydraulic fracture simulation. Several iterations were completed to refine an accurate and representative mechanical earth model (MEM), in relation to existing measurements and data. Based on the results from hydraulic fracture geometry and geomechanical iterations, production history matching using analytical and numerical reservoir simulators was then performed. Initial modeling was undertaken on a single-well basis. The process proved repeatable in single wells on opposite sides of a major geologic feature. Once satisfied with single-well results, the team modeled multiple-well situations to explain existing production and well interference conditions, and provide a basis for developing operational solutions.
A petrophysical model was developed to match core data for total organic carbon (TOC), effective porosity, effective water saturation, X-ray diffraction (XRD) results, and permeability with the standard density/neutron, resistivity, and gamma ray logs, Fig. 1. Clay volume was then calculated and matched to measurements from cored wells. The lithology of the Woodford is mostly clay and quartz, modeled with a multi-mineral deterministic approach, and tied back to core XRD results. A correction was made for Kerogen content and added to the total lithology. Core water saturation results indicated that the formation is at sub-irreducible saturation and would not produce water.
A porosity transform equation was used to estimate permeability, based on bound and free fluid porosity, and tied back to the core measurements. Reserves were then calculated using the Langmuir isotherms collected from pilot wells. The model was then finalized and used to generate field-wide data and map sweet spots in the area. A 1.4-million-cell structural model was then built, using fault and well data, and refined with geosteering data from horizontal wells. Well logs were upscaled, and the model was populated with relevant petrophysical properties.
MEMs were computed for select vertical wells around the study area, to provide the earth stress and mechanical property information for hydraulic fracture modeling. The MEMs integrate elastic properties from advanced dipole sonic logs, mechanical testing from full core, and observations from offset wells to create a representative model of the minimum and maximum horizontal stresses. The calibrated geomechanics model showed the difference between the minimum and maximum principal stresses to be about 1,500 psi for the Woodford shale, and 2,200 psi for the Hunton limestone. The maximum principal stress was found to strike slightly north of east/west. Diagnostic fracture injections tests (DFIT) from three wells in multiple Woodford layers, and a lower boundary layer, provided closure pressure-calibration points, data for small-volume, fracture net-pressure development, and fluid efficiency information. Production data from wells indicated interference between wells, as offset wells were being completed near older producers, Fig. 2. Integrating this information with DFIT, microseismic, bottomhole pressure measurements, Digital Acoustic Signals (DAS), Distributed Temperature Signals (DTS) and chemical tracers in our workflow aided in calibrating hydraulic fracture geometries.
FRACTURE MODELING AND SIMULATION
The Woodford is highly laminated, with many lithology and stress changes across a vertical interval of up to 280 ft in the study area. Initial runs over coarse vertical grid divisions were not satisfactory for this complexity, so a fine-grid, 3D fracture simulator was used. The resulting simulation improved agreement with maximum hydraulic fracture height well data. Initial simulations on this fine grid had to accommodate over 600 ft of Woodford and boundary layers, making runs very time-consuming.
Once key parameters and matches were obtained, the vertical grid was upscaled to 20-ft divisions. Several iterations were made on locating best boundaries and average stresses in the up-scaled model, to agree with the finer grid models. Eventually, a reliable model was constructed that ran in hours rather than days, using allocated injection rates and volumes from perforation-entry analysis.
Simulation results indicated very large hydraulic fracture half-lengths, extending up to 4,000 ft. Effective propped fracture lengths, with a conductivity cutoff at 5 Md-ft, were limited to less than 0.1 of hydraulic length, at just under 400 ft. Propped fracture length was largely limited by inefficient transport in large-geometry, induced low-fluid velocity, and thin, slick-water treatment fluids. Propped fracture height was limited to about half the interval inhibited by thin fracture development in high-stress layers, and rapid settling, keeping most or, in some cases, all proppant at or below lateral placement layers.
Overall, the geomechanical model, material balance, simulated and observed surface and downhole pressure, fracture simulations, and microseismic data agreed on dominantly planar fracture geometry of large dimensions. The formation is naturally fractured and confined to certain beds, but image logs indicated that these fractures are parallel, or close to parallel, to the induced hydraulic fractures and did not add
Induced fracture geometries on long spacings (greater than 4,000 ft) communicated hydraulically with existing, productive, well fracture systems. On tighter well spacing, induced fractures crossed existing producer wellbores. Offset well stimulation fluids severely inhibited production, even with aggressive lift, by loading the existing well fracture systems. Depleted reservoir pressure around the producers prevented fluids unloading from the fracture system, permanently damaging productivity, either through relative permeability, or capillary pressure changes in the fracture system.
In situations where the producing wells had significant reservoir pressure depletion, mechanical stresses were lower in proximity to the producing wells, Fig. 3. As offset wells were stimulated, fractures developed asymmetrically, biased toward producing wells. Depleted stress-grid, asymmetric-fracture simulations indicated that infill-well propped geometries were only slightly changed, but fracture fluid volume propagating toward the producing well was larger than toward the non-depleted region on the opposite side of the treatment well.
With the large geometries created in the original and infill wells, hydraulic communication between all local wells was inevitable. Multiple infill wells were drilled through dozens of existing hydraulic fractures. Other root causes of producing well fracture system flooding included large hydraulic geometry extension, caused by high fracture-fluid efficiency, and low perforation-cluster efficiencies. Simulations indicated that as little as three casing volumes of fluid was enough to extend hydraulic fracture half-length more than three well spacings. When simulating entire treatment volumes, hydraulic half-lengths extended across six to eight current well spacings. Reservoir pressure depletion in existing producers prevented older wells from recovering fracture load water from
An analytical approach was applied to understand the impact of uncertainty in key reservoir matching parameters, such as fracture area, effective permeability and fracture height. Analytical modeling was performed on one of the study wells (W4). The well history showed a straight-line portion, indicative of the internal, linear-transient flow. During this flow period, all the transverse fractures drained independently. There were also uncertainties highlighted, relating to the unique fracture area and permeability. Multiple combinations of fracture area (Af) and effective permeability (k) were possible, as long as the overall Af√k was conserved. The project sought to narrow this uncertainty and validate it with the integrated workflow results from geomechanical modeling.
Well performance was calibrated, using multiple combinations of fracture area and effective matrix permeability by conserving the overall Af√k. By performing the analytical modeling on a study well, permeability and fracture half-length uncertainty was narrowed. Initial analytical modeling and geomechanical results were then fed back into the numerical modeling workflow for optimal future results.
Hydraulic fracture modeling suggested that, for a given stage, two dominant fractures were usually created, and that one was longer than the other. There were also indications that the hydraulic fractures were pinching out at the Woodford B and C interval boundary. Fracture geometry parameters served as preliminary input data for the reservoir simulation model. The production history matched the model of a section with nine horizontal wells, and has two fractures per stage, with productive half-lengths of 350 ft and 300 ft, and a productive interval from Woodford C
to E, Fig. 4.
Production modeling provided a valuable insight on the extent of pressure depletion, which can alter in-situ stress significantly (Fig. 3), weakening the rock in the depleted region, and impact new offset well completions, as the fractures from new offset wells may grow preferentially toward the weakened rock. The depletion profile was then integrated into the hydraulic fracture modeling to evaluate the severity of depletion on the offset infill wells completion strategy.
One significant hypothesis that evolved during the analysis was that hydraulic fracture productivity impairment could be due to depletion in and around fractures, which get flooded with fracture fluids from new offset wells. Due to the depleted state of the fractures, they are unable to lift the fluid toward the wellbore, cutting off the flow pathway for gas within the system. The team also developed a hypothesis to explain the damage mechanism witnessed in the Woodford development.
For example, in a high well-density area, there were 11 horizontal wells, of which nine were in a single section with approximately 590-ft well spacing. Hydraulic fracture modeling showed which wells could be detrimental to offset well producers and might cause long-term production impairment, as illustrated in Fig. 2. The first well in this area began production in September 2008. Seven months later, a second well was completed 5,200 ft in a westerly direction, resulting in a 13% reduction in daily gas rate for the first well. Both these wells continued production for 17 months, at which point a third well was completed 3,000 ft west of the second well. Production in the second well declined 31% instantaneously and was attributed to fracture interference between the second and third wells. Eight infill horizontal wells were completed later in the vicinity, with a further eight deployed near the first well, which had a major impact on parent wells’ production.
Hydraulic fractures were included in the flow model, using a local grid refinement (LGR) approach, via the ECLIPSE reservoir simulator. A snapshot of pressure was taken from the simulation, just before the infill well campaign, to provide an estimate of depletion around parent wells. Pressure depletion profiles were then fed back to the hydraulic fracture modeling workflow, to estimate fracture geometries for infill wells. Hydraulic fracture modeling under the depletion scenario was then undertaken for all infill wells, allowing for estimation of the appropriate fracture geometries. These geometries were then included in the production model via the
The team concluded that hydraulic fracture system damage was most likely due to stimulation fluid flooding into depleted fracture systems around parent wells, resulting in a reduction in the effective fracture area. Productive half-lengths for multiple wells were found to be around 300–350 ft., without significant effect from asymmetric fractures in infill wells. The team also discovered that stage or lateral landing points may impact vertical connectivity. For example, landing into Woodford B offers a better chance at establishing a high productive fracture height. However, fluid system, proppant type, injection rates and stress variability in zones will eventually control productive fracture height.
A number of key mechanical drivers for hydraulic fracture interference were noted, including excessive hydraulic fracture half-lengths, low perforation efficiency, high fluid efficiency in very low matrix permeability and limited or no fracture complexity, and large hydraulic fracture heights. Damage to the hydraulic fracture system mechanism is most likely, due to stimulation fluid flooding into depleted fracture systems around the parent well, resulting in a reduction in the effective fracture area. Ineffective fracture lengths and heights were found to exacerbate negative gas production by disconnecting a major portion of the productive contact area.
The team proposed a number of options to mitigate the impact of fracture interference on gas production, including refracturing the existing producing well with an energized fluid, distributed along the entire lateral in the fracture system, while incorporating a biodegradable diversion agent. This procedure was judged to be simpler and more direct than offset foam solutions, as well as more cost-effective. Improvements to infill well perforation efficiency, effective fracture heights, and propped-to-unpropped fracture ratios were also discussed, in addition to the importance of optimal reservoir
The team now plans to use the 3D model, on an ongoing basis, in well-performance forecasts for future completion and stimulation designs, and well performance and recovery simulations for field development planning.
This article contains information from URTeC paper 1923397, presented at the Unconventional Resources Technology Conference, Denver, Colo., Aug. 25-27, 2014.
- How to improve design and engineering for upstream projects (September 2023)
- First oil (August 2023)
- Machine learning-assisted induced seismicity characterization of the Ellenburger formation, Midland basin (August 2023)
- Executive viewpoint (July 2023)
- Utilizing electronic data captured at the bit improves PDC design and drilling performance (July 2023)
- Oil and gas in the Capitals (June 2023)