With the advent of high-resolution methods to predict hydraulic fracture geometry and subsequent production forecasting, characterization of productive shale volume and evaluating completion design economics through science-based forward modeling becomes possible. However, operationalizing a simulation-based workflow to optimize design to keep up with the field operation schedule remains the biggest challenge owing to the slow model-to-design turnaround cycle. The objective of this project is to apply the ensemble learning-based model concept to this issue and, for the purpose of completion design, we summarize the numerical-model-centric unconventional workflow as a process that ultimately models production from a well pad (of multiple horizontal laterals) as a function of completion design parameters. After the development and validation and analysis of the surrogate model is completed, the model can be used in the predictive mode to respond to the "what if" questions that are raised by the reservoir/completion management team.
Hydraulic-fracture initiation and propagation in the presence of multiple layers with different mechanical and flow properties are investigated experimentally using a novel fracturing cell. Mixtures of plaster, clay, and hydrostone are used to cast sheet-like and porous test specimens in layers with different configurations and properties. The layered specimens are hydraulically fractured under varying far-field differential stress. Fracture growth is recorded using a high-resolution digital camera. Key frames are subsequently analyzed using digital image correlation (DIC) to reveal microcracks, measure strains, and show other features such as shear-failure events that are difficult to detect with the naked eye.
The problem of a hydraulic fracture induced in a soft layer bounded by harder layers is considered. We demonstrate numerous laboratory experiments that reveal a clear tendency for induced fractures to avoid harder bounding layers. This is seen as fracture deflection or kinking away from the harder layers, fracture curving between the harder bounding layers, and fracture tilt from the maximum far-field stress direction. These observations appear to be more pronounced as the contrast in Young’s modulus and fracture toughness between the layers increases and/or the far-field differential stress decreases. Moreover, when a fracture is induced in a relatively thin layer, the fracture avoids the harder bounding layers by starting and propagating parallel to the bounding interfaces. Fracture propagation parallel to the bounding layers is also observed in relatively wide layers when the far-field stress is isotropic or very low.
A fracture approaching a dipping, harder layer tends to curve away from the hard layer by kinking toward the high side of the interface. Nonplanar fracture trajectories are observed even in homogeneous materials when the far-field differential stress is relatively low. Furthermore, various other fracture behaviors in layered specimens are demonstrated and discussed, such as fracture offsetting at material interfaces, fracture branching and complex fracture trajectories, and shear failure of weakly bonded interfaces.
Micro-seismic data suggest that complex fracture networks are formed frequently in unconventional reservoirs due to the interaction of hydraulic fractures (HF) with natural fractures (NF). Understanding this interaction is critical for optimizing fracturing design. It is generally accepted that under certain conditions, a propagating HF can cause shear failure of a NF before intersecting with it. This fact is not accounted for in the development of the existing fracture interaction criteria. The goal of this study is to account for these dynamic interactions and present new criteria that define the conditions under which a HF will cross, kink, branch, or turn along a NF.
We have used our peridynamics-based poroelastic fracturing simulator in this study, which solves for rock displacements and fluid pressure in a fully coupled and implicit manner. Shear failure of the NF is modeled using a Mohr-Coulomb failure criterion. The frictional force on the NF surface is modeled implicitly. The stress distribution around the HF is monitored as the NF approaches it. Considering the effects of shear failure, different propagation behavior such as turning, and crossing are characterized as a function of in-situ stress ratio, angle of approach, NF characteristics, and matrix permeability. It should be noted that the peridynamics model used in this study does not require a crossing criterion as an input, rather it can predict the interaction behavior based on local poroelastic stresses.
The model is validated against analytical crossing criteria derived using Linear Elastic Fracture Mechanics (LEFM) by ignoring shear slippage prior to intersection and poroelasticity in our model. Recent experimental observations that show an increase in approach angle before intersection of a HF with a NF are also used to test the model. Shear failure of the NF before intersection results in relaxation of the stresses locally. This in turn leads to the HF bending towards the NF. Though these effects are found to be important in low permeability rocks (100 nD), they are more pronounced in high permeability rocks (10 mD). In high permeability rocks, poroelastic effects are much more significant, leading to greater stress relaxation and thus a near-orthogonal modified approach angle. When stress relaxation due to prior shear slippage of the NF is considered, the HF is more likely to turn along the NF. For low angles of approach and low stress ratios (1.0-1.1 for low permeability rocks and 1.0-1.2 for high permeability rocks), the crossing criteria derived in this study are considerably different from those derived using LEFM. However, for near-orthogonal angles of approach or high stress ratios, the crossing criteria do not change much.
The crossing criteria derived here can serve as direct inputs for discrete fracture network models simulating the growth of complex fracture networks (
During hydraulic fracturing, the interaction of hydraulic fractures with natural fractures can result in the formation of complex fracture networks. In the past these interactions have been captured in hydraulic fracturing models using crossing criteria developed based on two-dimensional geometries. In this work, we investigate the interaction of hydraulic fractures and natural fractures in three-dimensions and demonstrate that there can be significant differences in the observed interactions.
A hydraulic fracturing simulator is presented that solves the coupled fluid flow and geomechanics problem for three-dimensional fractures. The simulator captures the physics of fracture growth and the intersection of hydraulic fracture with pre-existing discrete fracture network. The model employs a robust algorithm to account for the stress relaxation due to the slippage of natural fractures. The displacement of failed natural fracture elements is calculated rigorously. The model allows the partial failure of three-dimensional natural fractures and accurately calculates the stresses acting on the plane of the natural fracture.
It is shown that a natural fracture inclined at an angle to an approaching hydraulic fracture experiences compression in one region (due to the stress shadow of the growing hydraulic fracture) and tension in other regions (in front of the approaching hydraulic fracture tip). The generated stresses can fail the natural fracture partially. The failure of the natural fracture relaxes the stresses around it, which can modify the direction of propagation of the approaching hydraulic fracture. In addition, if the elliptical front of the hydraulic fracture crosses an intact planar natural fracture, the three-dimensional geometry results in a line of intersection (between natural fracture and hydraulic fracture). This can lead to failure of the natural fracture even after the elliptical front has partially crossed the natural fracture. Such an interaction can allow the hydraulic fracture to both cross the natural fracture and activate (or dilate) it. These effects cannot be captured by two-dimensional simulations. This work improves our understanding of the interaction between hydraulic fractures and natural fractures. The novel results provide new insights into the mechanisms responsible for the complexity that is often observed in hydraulic fractures.
The most important factors governing hydraulic fracture propagation are completions and treatment design, in-situ stresses, and reservoir heterogeneity at different length scales (including natural fractures and bedding planes). However, it has been recognized that in depleted reservoirs, stress changes arising due to reservoir drainage significantly affect the growth of fractures and attract them towards the depleted regions. Using a poroelastic hydraulic fracturing simulator based on the theory of peridynamics, stress reorientation due to production and its sensitivity on Biot's constant, pressure drawdown and reservoir fluid type is studied. It is shown that a tensile region is created between the fractures of a producing parent well. Consistent with previous studies, it is verified that a fracture from a newly drilled child well grows asymmetrically towards the depleted regions. When the child well lateral is not landed centrally between the parent wells and there is unequal depletion of these wells, asymmetry in the geometry of the child well fracture may further be accentuated. Similar observations can be made when production is from a gas reservoir. This asymmetric fracture growth leads to parts of the undepleted reservoir remaining unstimulated. Re-pressurization of the parent well fractures is shown to revert the stress state closer to the in-situ conditions, thereby reducing the attraction of the child well fracture towards depleted regions resulting in a better stimulation treatment.
Agrawal, Shivam (The University of Texas at Austin) | Ouchi, Hisanao (Japan Oil Engineering Co. Ltd.) | AlTammar, Murtadha J. (The University of Texas at Austin) | Sharma, Mukul M. (The University of Texas at Austin)
ABSTRACT: The effect of non-uniform pore pressure field and saturation on fracture propagation is investigated using a peridynamics-based hydraulic fracturing simulator. The model solves for rock displacement, fluid saturation, and fluid pressure both inside and outside the fracture in a fully-coupled manner. When fractures initiate from multiple injection points, they can propagate towards each other by opening against the maximum stress. Laboratory experiments conducted on synthetic rock samples show that saturating a porous rock with fluid before fracturing it decreases the breakdown pressure. Under low far-field stresses in the laboratory, fractures are attracted towards the high pore pressure region. The strength of this attraction depends on both the magnitude of the pressure and the pressure gradients. The simulation results are completely consistent with experiments and show why this effect is observed in the lab. These results highlight the importance of poroelasticity and non-planar fracture growth behavior in hydraulic fracture modeling.
Production of fluids from a reservoir reduces the pore pressure and creates pressure gradients in the rock. This modifies the stress state from its initial in-situ condition (Warpinski and Branagan, 1989; Wright et al., 1994). Numerous field studies have reported this phenomenon in the literature (Siebrits et al., 2000; Weng and Siebrits, 2007; Roussel and Sharma, 2012). In addition, laboratory experiments are often performed to understand the underlying mechanisms (Bruno and Nakagawa, 1991; Liu et al., 2008). Further insights into refracturing process are obtained by mathematical modeling and numerical simulations (Berchenko and Detournay, 1997; Wang et al., 2013; Agrawal and Sharma 2018).
Bruno and Nakagawa (1991) conducted fracturing experiments in the presence of a non-uniform pore pressure field and isotropic stresses. They showed that both mechanical and hydraulic fractures are attracted to the high pore pressure region. Their observations were justified by Berchenko and Detournay (1997) based on a deviation of the maximum stress trajectory towards the injector well. Recently, fracture propagation in the presence of different configurations of injection sources has been studied experimentally by (AlTammar et al., 2018). The purpose of this research is to simulate and explain their results using an effective stress law and to investigate the effect of pore pressure, fluid injection scheme, saturation conditions, and applied stress on fracture growth.
Most hydraulic fracturing models assume that the rock is homogeneous at a pore scale. However, reservoirs are highly heterogeneous at all length scales. Pore scale heterogeneity is evident from thin sections and scanning electron microscopic images (SEM). Heterogeneity at larger length scales is evident from logs and cores. In this paper, the effect of micro-scale (pore and core scale) heterogeneities caused by varying mineral composition and the presence of pores and microfractures, on fracture propagation has been investigated.
A model that solves the solid displacements and fluid pressures both inside and outside the fracture and allows for the creation and propagation of multiple fractures is presented. This peridynamics-based hydraulic fracturing model is used to model the growth of multiple, complex fractures in a heterogeneous rock. Thin section and SEM images of rocks are used to represent the geometry of the rock grains and the pores in several rock samples. Far-field stresses are then applied and a fluid induced fracture is propagated in the rock matrix.
The results of the model reveal that the stress distribution and the fracture geometry can be quite complex at the micro-scale. Fracture branching and turning is induced by variations in elastic moduli and stress concentration at the grain scale. The microstructure of the fracture is, therefore, determined by the geometry and distribution of mineral grains, their mechanical properties, and the initial stress anisotropy due to the co-existence of different mineral grains. A similar effect is observed at the core scale where differences in the microstructure of the rock can result in stress concentration at layer boundaries. For example, we show that the presence of a brittle mineral like calcite in the rock matrix causes fractures to branch at the mineral interface. Multiple fractures are shown to open, some that may not be in hydraulic contact with each other. As the fracture propagation continues, only the least tortuous path remains open. All other branches are bypassed hydraulically and are eventually closed. This fracture complexity occurs despite macroscopic stress anisotropy. Several examples of fracture propagation in rocks that are heterogeneous at a pore scale are provided to show that such fracture complexity should be expected in most lithologies.
These results clearly show that while we have traditionally represented fractures as planes perpendicular to the minimum horizontal stress, the fracture surfaces may indeed be much more complex due to existence of different minerals grains with widely different mechanical properties. Cracks can form away from the crack tip along planes of weakness. These damage zones resulting from strains induced by fracture propagation may explain the creation of the stimulated reservoir volume (regions of enhanced permeability) around fractures in shale reservoirs.