Xu, Wenyue (Schlumberger-Doll Research Center) | Prioul, Romain (Schlumberger-Doll Research Center) | Berard, Thomas (Schlumberger-Doll Research Center) | Weng, Xiaowei (Schlumberger) | Kresse, Olga (Schlumberger)
This work introduces a new set of energy-balance-based criteria for the vertical growth of a plain-strain planar hydraulic fracture across a horizontally laminated reservoir formation with heterogenous layer properties and multiple weak interfaces. Combined with Coulomb's friction law for mechanical balance along sliding interfaces, these criteria were built into a novel semi-analytical model of fractional fracture height growth. The model was then applied to investigate the growth of hydraulic fractures in an idealized symmetric three-layer rock formation, with the upper and lower layers acting as barriers to the growth. Preliminary modeling results show how the vertical growth of a hydraulic fracture is influenced by the various mechanical/energy barriers. Three primary types of barrier behaviors are identified. A stress barrier leads to gradually increasing fluid pressure when the barrier layer is crossed. A toughness/modulus barrier, on the other hand, results in an immediate sharp increase in fluid pressure followed by gradual decline in pressure. The effect of individual sliding interfaces is similar to that of a toughness/modulus barrier. The cumulative effect becomes more important when multiple closely spaced interfaces are present. A formation layer containing multiple closely spaced weak interfaces behaves collectively much like a stress barrier.
Ball sealers are commonly applied in fracturing and acidizing treatments for diverting treatment fluid to the desired zones by plugging perforations. It has proven that injecting ball sealers is a low-cost and efficient method for diversion. To predict the effectiveness of ball sealers, an improved ball sealer seating model is developed by introducing the maximum seating efficiency and random functions to capture the stochastic nature of ball-sealer plugging. The new model can predict ball sealer performance with different ball densities in vertical, deviated and horizontal wells.
The traditional ball sealer model was originally designed for vertical wells, where ball sealers with different densities have similar behavior. However, for deviated and horizontal wells, the seating of buoyant and dense balls is more complicated. Buoyant balls tend to plug the perforations at the top of wellbore, and dense balls tend to plug the perforations at the bottom of wellbore. Thus, the traditional ball sealer model cannot be applied in these wells. A maximum seating efficiency for each ball is introduced in the new model, which is obtained by correlations based on experimental results. To describe the stochastic nature of ball sealer seating on perforations, a random number is assigned to each ball sealer, and a range is assigned to each perforation based on the ratio between flow rates through the perforations and flow rate in the wellbore.
With the improved model, it can predict seating efficiency of ball sealers for all types of well with buoyant, neutral and dense balls. The results are showing that the seating efficiency of ball sealers predicted by the model can match the experimental results, which validates the model. Based on the simulation results, when ball sealers with mixed densities are pumped into deviated or horizontal wells, the seating efficiency is better than pumping ball sealers with only one density. For vertical wells, the benefit of mixing densities is minimal.
Commercial hydraulic fracture software are generally based on the planar 3D model (PL3D) or the Pseudo 3D model (P3D). The PL3D model is more accurate but very CPU intensive, which makes it unsuitable to simulate complex hydraulic fracture networks due to the interaction with natural fractures. The pseudo-3D model is faster but considers separately the vertical and horizontal propagation of the fractures. In recent years, a sophisticated P3D-based complex hydraulic fracture network model, the Unconventional Fracture model (UFM) was developed that can simulate complex physical mechanisms such as interaction with natural fractures, stress shadow and the proppant placement in the complex fracture network. However, one of the main challenges still faced by such simulators is to accurately predict the height growth in formations with heterogeneous mechanical and stress properties.
One assumption of the P3D model is the fracture is being initiated and extended in a lower stress layer than the adjacent layers. In practice, the assumption is not always satisfied, leading to inaccurate or unstable height growth. A comparison of these two models through examples illustrates the situations where P3D model works well and where it does not, and the difficult compromise between accuracy and computational efficiency.
A better compromise can be achieved through a new Stacked Height Growth model (SHG). This model is an enhancement to the P3D model, consisting of multiple rows of elements vertically stacked, to more precisely account for the effect of vertical stress heterogeneity, and to allow multiple horizontal propagation fronts. The width profile and stress intensity factors at the top and bottom depend on the stress and pressure profiles along the stack of elements. The theoretical background of the model is presented. Comparison shows good agreement with the PL3D model for cases that the P3D model cannot accurately simulate, and at a fraction of the computational cost of PL3D. A particularly interesting feature of the SHG model is its flexibility to transition smoothly from a 1D cell-based model such as the P3D model, to a fine scale 2D gridding in the fracture plane. This feature greatly facilitates the implementation of other modeling features into the fracturing simulator.
For example, the SHG model can be used to simulate interaction of hydraulic fractures with Multi-layer Discrete Fracture Networks (MDFN), to better model naturally fractured reservoirs. The SHG model can also be used to model offsets of hydraulic fractures through weak interfaces called ledges. The model can capture the influence of these discontinuities on the proppant placement. Another example, is how a fine 2D gridding using the SHG model can predict proppant placement as accurately as a PL3D model. These extensions of the SHG model have been implemented into the UFM model and are illustrated in this paper through several examples.
Microseismic source mechanisms in shales often show similar double-couple events with one vertical nodal plane aligned with the hydraulic fracture and the other nodal plane oriented horizontal. We interpret the horizontal nodal plane as the fault plane where the shearing represents bedding-plane slip driven by opening of a connected, vertical hydraulic fracture. Using precisely mapped microseismic source locations from a Barnett shale hydraulic-fracture completion, Rutledge et al. (2015) proposed a model of step-over or jog structures along bedding to describe bedding plane slip and sense-of-slip reversals driven by vertical fracture opening. However, common step-over features will usually be too small to create a detectable microseismic signal, and if large, will terminate vertical growth. Chuprakov and Prioul (2015) describe a model of hydraulic-fracture growth through weak bedding interfaces where tensile crack growth can arrest at interfaces encountered. Fluid invasion and pressure can weaken the interface which can then shear from the added stress of adjacent crack opening. Under certain conditions, with subsequent net pressure recovery and gain, tensile fracture propagation can continue across the interface. Invoking this model, at least qualitatively, makes a more consistent explanation of the continuity and time-space development of the microseismic structures, as well as explaining the observed patterns of shear-fracture first motions.
Presentation Date: Tuesday, October 18, 2016
Start Time: 10:20:00 AM
Presentation Type: ORAL
As is well known, hydraulically fractured horizontal wells have been extremely successful in the development of low permeability reservoirs throughout the world. The vast majority of these completions employ cased and cemented wellbores drilled approximately in the direction of minimum horizontal stress. Multiple, relatively short perforation clusters are included within each frac stage along the lateral. This efficiently creates many hydraulic fractures propagating orthogonal to the well, but it does not insure that each perf cluster is effectively stimulated.
Many efforts have been made to improve the effectiveness of horizontal completions. This has mainly focused on using lateral measurements to place perforation clusters in rock of similar stress so they are more likely to be successfully stimulated. But this ignores the impact of formation initiation pressure and tectonics on fracture initiation. In addition, the number, dimensions and orientation of the perforations in each cluster can greatly influence the effectiveness of the stimulation at each initiation point.
To address these issues a near-wellbore fracture initiation calculator has been developed that predicts whether a fracture will initiate at a perforation, the minimum initiation pressure, the fracture initiation location and orientation at each perforation, and the injection rate into each perforation. These parameters are a function of the casing size and orientation, the mechanical properties of the rock and cement, the principle effective stresses, and properties of the perforations.
A series of sensitivities have been performed to quantify the impact of injection rate, tectonic setting, stress variation between clusters, and perforation properties on hydraulic fracture creation, orientation and complexity at each perf cluster. The sensitivities demonstrate that fractures may not initiate at many clusters and that within an active cluster some perforations may not be accepting fluid. Incorporating the results from this model enables engineers to design completions that insure all perforation clusters are effectively stimulated and near-well fracture complexity is minimized.
This methodology does not just look at a single perforation, or cluster. Instead, it accounts for the stress variation between multiple perforation clusters within a frac stage, in addition to perforation orientation, dimensions and eccentricity, to predict the likelihood that each perforation cluster will be stimulated. By employing this methodology one can better design a perforating system and optimize perforation placement within a lateral to insure hydraulic fractures are created at all perforation clusters.
An improved 1D-averaged model for wormhole propagation was developed that considers acid spending and leakoff along wormholes. The 1D-averaged type of model is one of several types that have been developed for predicting wormhole propagation during matrix acidizing treatment in carbonates; other models include discrete models, Darcy-scale 2D and 3D continuum models, and multi-pore-scale network models. Among these models, the 1D-averaged model is commonly applied in acidizing simulation software because of its simplicity.
Most of 1D-averaged models are based on the pore volume to breakthrough correlations developed from core flooding experiments on different rock samples and acid systems. It is also generally assumed that the effect of acid spending is already included in the experimental results. Although this is a reasonable assumption when the wormhole length is shorter than the core length (typically 6 in.), once wormhole penetration is longer than 6 in., the acid spending becomes more significant while acid travels along wormholes due to the reaction between acid and the side walls of wormholes. Meanwhile, acid leak-off through wormhole walls becomes an important factor for longer wormholes. Not considering acid spending and leakoff in relatively longer wormholes may result in overestimation of wormhole penetration depth.
In our model, we adopted an empirical correlation to determine the leak-off velocity profile along wormholes. Then, we developed a numerical model to simulate the non-uniform acid concentration along wormholes. Both the empirical correlation for leak-off velocity profile and the numerical model for acid concentration are validated by comparing to published materials and an independent computational fluid dynamics (CFD) software. As expected, wormhole penetration predicted by the new model shows shorter wormholes than the traditional 1D-averaged model.
We applied the new model to a field case and showed its capability of reasonably matching the downhole pressure during acidizing treatments and PLT data before and after acidizing treatments, which demonstrates the applicability of the model to job design and interpretation in the field.
Yu, Xin (Schlumberger) | Rutledge, Jim (Schlumberger) | Leaney, Scott (Schlumberger) | Sun, Jianchun (Schlumberger) | Pankaj, Piyush (Schlumberger) | Weng, Xiaowei (Schlumberger) | Onda, Hitoshi (Schlumberger) | Donovan, Michael (Schlumberger) | Nielsen, Jacob (Schlumberger) | Duhault, John (LightStream Resources)
Geometries of fractures, both natural and hydraulically induced, are commonly represented by a discrete fracture network (DFN) in reservoir simulations. Although microseismic data is the best diagnostic tool for imaging the volume of rock fractured, its incompleteness in representing the total deformation presents a challenge in interpreting the complex fracture networks from hydraulic fracturing. While microseismicity under-represents most of the tensile fracture volume created, the hydraulic fracturing modeling software can, in principle, generate the full hydraulically induced fracture network. Unlike other simulation methods that assume simple planar geometry, the unconventional fracture model (UFM) simulates complex fracture structure, but requires a reliable description of natural fractures in the formations with determined locations of the fracture patches as the input. The DFNs derived from measurements such as well bore image logs and seismic properties have large uncertainties due to gross differences in scale of measurement compared with the hydraulic fracture process. Microseismic data can be used as direct evidence of hydraulically induced fracture propagation to calibrate the fracture model, as a constraint to the DFN used in UFM to recover information such as the effective surface area, aperture and retained permeability.
This study presents a workflow to use the microseismic data including source mechanism information in UFM for complex fracture calibration and simulation. A dual well monitoring data set from 8-stages of a 30 stage sliding-sleeve hydraulic fracturing treatment in the Cardium formation, west central Alberta, Canada is used as a case study to demonstrate the workflow. First, moment tensor inversion is conducted to estimate the fault plane solutions. Second, a DFN is extracted from the microseismic event locations and fault plane solutions. Third, rheology and geomechanics models generated from the well logs from the offset wells are calibrated by applying the DFN from the second step as the input into UFM to match the microseismic geometry and treating pressure. Last, the simulated complex fracture networks with full information, such as aperture and conductivity, are input to reservoir simulation for production simulation and reservoir management. This approach is compared with a traditional P3D fracture model. The results shows that using the DFN extracted from the microseismic data reduces the uncertainty of the models input into the hydraulic fracturing modeling and results in better pressure and geometry fitting.
The interaction of hydraulic fractures with the pre-existing natural fractures may play a major role in increasing productivity from unconventional formations. When a hydraulic fracture meets a natural fracture, the hydraulic fracture can cross the natural fracture or be arrested. If the natural fracture is permeable, fracturing fluid can leak from the hydraulic fracture into the natural fracture causing elevation of pore pressure in the natural fracture and reducing the effective normal stress acting on the natural fracture, which could then lead to shear failure or slippage along the natural fracture plane. Shear-slip causes dilation, potentially increasing fracture conductivity and enhancing fluid flow deeper into the natural fracture. The conductivity of unpropped shear-induced fractures can play an important role in enhancing the productivity from ultralow-permeability formations like shale. In this paper, we first evaluate analytically the shear-slip condition and its propagation along a natural fracture under remote normal and shear stresses, when it is exposed to the fluid pressure in a hydraulic fracture. Analytical approximations under some limiting conditions are considered. A rigorous 2D numerical model based on coupling between fluid flow and rock deformation using displacement discontinuity method and fluid flow in the fracture is then described. The results of numerical simulations are presented to illustrate the effect of rock stress anisotropy, initial natural fracture conductivity, and fluid properties on the evolution of the fluid and slip fronts along the natural fracture and the associated permeability enhancement.
In the last decade, following the success of horizontal drilling and multistage fracturing in the Barnett Shale, exploration and drilling activities in shale gas and shale oil reservoirs have skyrocketed in the US and abroad. Economic production from these reservoirs depends greatly on the effectiveness of hydraulic fracturing stimulation treatment. Microseismic measurements and other evidence suggest that creation of complex fracture networks during fracturing treatments may be a common occurrence in many unconventional reservoirs [1-3]. The created complexity is strongly influenced by the preexisting natural fractures and in-situ stresses in the formation. To optimize the fracture and completion design to maximize the production from these reservoirs, engineers must have a good understanding of the fracturing process and be able to simulate it to obtain information such as the induced overall fracture length and height, propped versus unpropped fracture surface areas, proppant distribution and its conductivity, and potential enhanced permeability through stimulation of the natural fractures.
The majority of planar hydraulic fracture models use two distinct approaches. The first one, referred to as the planar 3D model, is more accurate but also very CPU intensive. The second one is referred to pseudo-3D (P3D) model, and separately considers the vertical growth and horizontal propagation of the fractures. This approach is less CPU intensive, but requires the fracture being initiated in the lower stress layer. In practice, this assumption is not always verified, and the fracture height growth can become unstable. This paper presents a new model as an enhancement of the P3D, which consists of multiple rows of elements vertically stacked and connected. For each row of elements, the assumption of the fracture front being in the lower stress layer is satisfied locally. The width profile and stress intensity factor at the top and bottom of the fracture depend on the stress profile and the pressure profile along the stack of elements. This model predicts the fracture height more accurately than the P3D model, and gives results close to the ones from the full planar 3D model.
The rapid development of shale resources in the past decade has brought a focus on the process of hydraulic fracturing. Shale reservoirs tend to be characterized by a complex 3D stress field and vertically heterogeneous mechanical properties, which have always been challenging for hydraulic fracturing modeling and particularly for properly predicting the shape of an induced fracture . Most state-of-the-art planar fracture simulators use two distinct approaches. In the first one, referred to as the planar 3D model (PL3D), the fracture is assumed to be a plane and its entire footprint is discretized into elements. The equations governing fluid flow, elasticity, and mass balance are solved numerically, coupled with the fracture propagation rules. This approach is very accurate but also very CPU intensive . This type of model is mostly used when a large portion of the fracture propagates outside of the zone where the fracture was initiated and significant amount of vertical flow is expected. The second approach is based on the cell-based pseudo-3D (P3D) model , which separately considers the vertical growth and horizontal propagation of the fractures. In this approach, the width profile and fracture height are calculated based solely on the local pressure and local vertical stress profile. This approach is less CPU intensive, but relies on several assumptions including the fracture being initiated and its leading front propagating in the lower stress layer compared to the neighboring layers above and below. If this is not the case, the fracture height growth can become unstable, since it is not directly correlated to the global fracture mass balance as in the PL3D model, and this can lead to significant inaccuracy in the predicted fracture height growth.
There is accepted evidence that multistage fracturing of horizontal wells in shale reservoirs results in significant production variation from perforation cluster to perforation cluster. Typically, between 30 and 40% of the clusters do not significantly contribute to production while the majority of the production comes from only 20 to 30% of the clusters.
Based on numerical modeling, laboratory and field experiments, we investigate the process of simultaneously initiating and propagating several hydraulic fractures. In particular, we clarify the interplay between the impact of perforation friction and stress shadow on the stability of the propagation of multiple fractures. We show that a sufficiently large perforation pressure drop (limited entry) can counteract the stress interference between different growing fractures. We also discuss the robustness of the current design practices (cluster location, limited entry) in the presence of characterized stress heterogeneities.
Laboratory experiments carried out in organic shale blocks highlight the complexity of the fracture geometry in the near-wellbore region. Such complex fracture path results from local stress perturbations around the well and the perforations, as well as the rock fabric of these shales. The fracture complexity (i.e., the merging of multiple fractures and the reorientation towards the preferred far-field fracture plane) induces a strong nonlinear pressure drop on a scale of a few meters. Field experiments in horizontal wells show that this near-wellbore effect is larger in magnitude than perforation friction and is highly variable between clusters, without being predictable. Through a combination of field measurements and modeling, we show that such variability results in a very heterogeneous slurry rate distribution; and therefore, proppant intake between clusters during a stage, even in the presence of limited entry techniques. We also note that the distribution of proppant intake between clusters appears similar to published production log data.
We conclude that understanding and accounting for the complex fracture geometry in the near-wellbore is an important missing link to better engineered horizontal well multistage completions.