The future of simulation: genuine integration of hydraulic fracturing and reservoir modeling
Reality has tight coupling between hydraulic fracturing and the reservoir. Your simulator should too.
We've built a 'complete product' around our technology - prioritizing your ability to drive business value
Engineers and geoscientists
Why use ResFrac?
A physics-based simulator will deliver the most impactful results
ResFrac’s simulator is physics-based and solves the governing equations of fracture mechanics, proppant transport, fluid transport and geomechanics. Physics-based simulators uncover why things happen so that you can apply learnings to your next pad. Physics-based simulators are complimentary with data driven analyses and the other tools in your toolbox.
You should integrate hydraulic fracturing simulation with reservoir simulation
Processes such as parent/child interaction, refracs, EOR, and DFITs involve hydraulic fracture opening and propagation simultaneous with multiphase/reservoir flow. They can’t be modeled without a coupled approach. Even for basic fracturing/production simulations, separating the models is awkward, sacrifices physical realism, and can lead to the wrong answer.
You need a ‘true’ hydraulic fracturing simulator
Sometimes, so-called ‘coupled hydraulic fracturing and reservoir simulators’ aren’t really hydraulic fracturing simulators. They use conventional reservoir simulators to model hydraulic fractures as high permeability slabs of rock. A ‘true’ fracture simulator meshes the cracks as cracks and uses fracture mechanics and equations for transport in a crack. This delivers more accurate results and better optimization of well performance.
Get up to speed quickly with the ResFrac
A to Z Guide and step-by-step video tutorials
Our training materials cover the benefits of numerical modeling, modeling workflow best practices, how to setup and run simulations, how to perform model calibration, and the soft skills needed to execute a modeling project from start to finish.
Technical documentation and training materials
Download our Technical Write-up
Take a Training Course
The #1 most common application of ResFrac is to optimize frac design and well placement in shale. We optimize: cluster spacing, well spacing (vertical and horizontal), well landing depth, proppant mass, fluid volume, and perforation design (limited entry). In addition, we help design mitigation strategies for parent/child issues, design refracs, optimize diverter placement and timing, and design enhanced oil recovery in shale.
The term ‘Enhanced Geothermal Systems’ (EGS) refers to the use of hydraulic stimulation to improve production for geothermal energy production. Because ResFrac couples wellbore flow, hydraulic fracturing, and fluid and heat transport, it is very well suited for optimizing these systems. We have several clients in this space. ResFrac includes correlations enabling the simulation of the water/steam transition and ultrahigh temperatures, including supercritical water.
Even though we have traditionally focused in shale, ResFrac has also been applied on projects in conventional formations. Actually, it is very well-suited for conventional applications because it inherently includes a full 3D multiphase leakoff calculation, which in high permeability formations, is much more accurate that the conventional Carter leakoff approach.
Finally, ResFrac has been applied to investigate specialized topics such as diagnostic fracture injection tests and offset fiber optic measurements.
Most modeling studies should follow this general workflow:
- ‘Kickoff’ – Establish scope and objectives; plan the workflow
- ‘Initial simulation’ – Gather relevant data and set up an initial simulation; communicate back to stakeholder to check communication; present a list of ‘key observations to match’
- ‘List of key observations’ – Review all available field data, and create a list of key observations that characterize the dataset. Prior to starting model calibration, create a plan for what you plan to vary to achieve a match, based on your hypotheses about what is happening in the data.
- ‘History matching – Calibrate to field data
- ‘Sensitivities’ – Run numerical experiments on the calibrated model
In our case studies section (link to library, case study subsection), you can find examples where we help companies optimize frac design to achieve 20% increase or more in production or ROI. We can’t predict in advance – we run an analysis to optimize ROI, and then compare with the current design, and see what the difference is! Longer term, we can drive even more value by attacking problems such as mitigation of parent/child issues, and design of enhanced oil recovery.
Numerical modeling capabilities
ResFrac fully integrates a 3D ‘true’ hydraulic fracturing simulator, a reservoir simulator, and a wellbore simulator. This allows us to simulate the entire life-cycle of the well, from fracturing to long-term production. The simulation is continuous – solving all equations in all elements in every timestep – and always using the same consistent mesh. There is no distinction in the simulation between ‘fracturing’ and ‘production.’ All physics are active at all times. This enables us to simulate complex processes that involve simultaneous fracturing and reservoir processes – such as frac hits, enhanced oil recovery, and diagnostic fracture injection tests. Conventionally, hydraulic fracturing and reservoir simulation have been performed with two separate codes. This creates awkward handoffs, and makes it impossible to solve problems that involve both physics simultaneously. We avoid these problems with our integrated approach.
A true hydraulic fracturing simulator meshes cracks as cracks. Alternatively, models may attempt to approximate hydraulic fractures as regions of high permeability in a conventional reservoir simulator. This approach sacrifices realism. A true fracturing simulator is needed to describe the mechanics of stress shadowing and crack propagation, proppant transport, and fracture reopening and closure.
ResFrac solves for mass balance on fluid components, water solutes, and proppant types, energy balance, and momentum balance in the wellbore, and also solves the equations of continuum mechanics for deformation in the formation (quasi-static stress equilibrium, compatibility, and Hook’s law for linear elastic deformation). The user has options to turn on or off different physics, depending on application. For example, they can run thermal or isothermal simulations. There is no limit to the number of water solutes, proppants, or fluid components that you can define.
Fluid, proppant, and heat transport are calculated using the finite volume method.
ResFrac simulates multiphase flow, non-Darcy pressure drop, and non-Newtonian fluid rheology. Fracture conductivity is a strongly nonlinear function of crack aperture, which is calculated continuously as part of the mechanical calculations. Different equations are used when the crack is mechanically open, mechanically closed with proppant, and mechanically closed without proppant. The constitutive relations smoothly transition between these limiting cases. The proppant transport algorithm includes correlations for a large number of key processes, including the effect of fluid viscosity, particle density, and particle size, hindered settling, clustered settling, the effect of proppant on slurry viscosity and density, size exclusion from limited aperture, and jamming. As the proppant volume fraction approaches 0.66, the proppant becomes immobilized, but fluid is able to continue to flow through the packed bed of particles. The conductivity of the fracture elements is a nonlinear function of aperture, normal stress, fluid pressure, the amount of proppant present, the properties of the proppant, and certain formation properties. Fracture conductivity is calculated implicitly in every element during every timestep. Typically, we run simulations in which the propped fracture conductivity is much less stress sensitive than the unpropped conductivity.
ResFrac supports the black oil model, the modified black oil model, and the compositional model. In the black oil and modified black oil model, correlations are used to calculate unsaturated properties. Alternatively, the user may specify a ‘table of tables’ for unsaturated properties. The compositional model uses the Peng Robinson equation of state, with industry best-practice algorithms from “Thermodynamic Models: Fundamentals and Computational Aspects” by Michelsen and Mollerup. ResFrac includes correlations enabling the simulation of the water/steam phase transition and can accurately calculate water properties at ultrahigh temperatures, including supercritical.
Fracture stress shadowing is calculated using the boundary element method. The boundary element method does not utilize the volumetric mesh of the surrounding rock, and so works naturally with the simulator’s nonconforming mesh strategy. We use modifications of the standard technique in order to avoid problems caused by closely spaced fracture elements and to approximate the effects of heterogeneity and anisotropy. For crack propagation, we developed a multilayer scheme called MuLTipEl. It tracks the position of the crack front within each tip element and uses tip asymptotic solutions to enhance accuracy. It uses linear elastic fracture mechanics. Stresses induced by pressure and temperature changes in the formation are calculated using a finite volume based technique.
All equations are solved in every equation in every timestep, using a fully coupled scheme. Timestepping is performed with the adaptive implicit method.
The formation can be meshed with either a rectilinear grid or with a corner point grid. We provide tools to import grids from geocellular modeling software. The fracture mesh and the wellbore elements and nonconforming with the matrix mesh, which means that their vertices do not have to coincide. This is very convenient because it avoids the need for complex remeshing as the fractures propagate (which would not even be feasible in a conventional corner point grid). Our method of calculating matrix-fracture fluid exchange is related to the EDFM technique. However, the EDFM method requires an extremely refined mesh to avoid inaccuracy in very low permeability formations .This is not practical for many problems. To address this challenge, we perform a multiscale strategy called the ‘1D submesh method’ to locally generate a fine scale mesh for each fracture-matrix element pair.
Wellbore boundary conditions can be defined at the wellhead, as bottomhole constraints (which removes the wellbore elements from the system of equations), or a ‘MD’ constraints that place the boundary condition within the well (such as at the pump intake depth of an ESP). The user specifies a series of boundary conditions for each well. Each condition includes both a rate and pressure constraint. For example, you may specify injection rate and also specify a maximum injection pressure. The constraints are applied implicitly. If a constraint is violated during a timestep, the timestep is discarded, the well is switched from rate to pressure control (or vice-versa), and then the timestep is repeated. At the edges of the model, there are options to specify constant pressure or no-flow boundary conditions.
We often run simulations of individual stages, rather than full wells. The model width is equal to the width of the stages included in the model, with no-flow boundary conditions along the edges in order to avoid ‘double counting’ production from adjacent stages.
Fracture lengths are much greater than producing fracture lengths, which creates challenges for modeling the effect of adjacent wells from beyond the model boundary in the ‘along strike’ direction. This is handled by making the simulation domain relatively long during the fracturing simulation (to allow the hydraulic fractures to grow unrestricted), but then deactivating the simulation domain halfway along the distance to closest ‘unmodeled’ well to avoid double counting production.
Fracture-fracture stress shadow calculations are performed assuming the cracks are embedded in an infinite-sized elastic domain. The effect of fractures from adjacent stages are included as special ‘external fracture’ constraints.
Poroelastic and thermoelastic calculations are performed with symmetric boundary conditions. They assume that the adjacent rock is experiencing the same deformation as in the problem domain. This is useful for the ‘sector’ scale models that we often perform. If not performing a sector-scale model, then the model boundaries should be far from the zone of deformation, in order to approximate an infinite domain (which is the same approach used with conventional constant displacement or stress boundary conditions).
There is no limit on the number of wells that you can include in the model. Practically, we rarely include more than 12.
Overall modeling approach
The traditional paradigm in shale has been that fracturing creates a branching, dendritic network of fractures. In recent years, this conceptual model has been contradicted by field data, such as: (a) fiber showing linear, narrow fracture propagation, (b) offset fiber strain responses that resemble the analytical prediction for opening-mode cracks and do not show significant off-azimuth propagation, opening, or shear, (c) core-through showing consistent fracture orientation, and (d) offset pressure gauges showing strong correlation of pressure gauges with propped hydraulic fractures. In our own experience, we consistently find that production data can be matched with DFIT-derived permeability and linear hydraulic fractures.
We believe that planar fracture modeling is not only more practical than DFN; it is more realistic. We provide more discussion of this topic in SPE 199726 and SPE 204172 and the video “The case for planar fracture modeling.”
Yes, you can specify preexisting fractures. We commonly use these to represent seismic-scale faults. The preexisting fracture capability has occasionally been used in ResFrac for ‘discrete fracture network’ modeling, though this is not the typical recommended workflow. Also, we occasionally use preexisting fractures to represent the fractures around legacy wells, if we do not have enough information about the original completion to simulate the original fracturing treatment.
We have honed a step-by-step calibration procedure, as outlined in the ResFrac A to Z Guide. We start by making a list of ‘key observations’ to match. These include observations related to fracture length and height, perforation efficiency, and production. Offset fiber, sealed wellbore pressure monitoring, frac hits, and microseismic are used to infer fracture geometry. We build RTA plots of production data (to normalize for the effect of variable BHP), and use the trends to identify physically what is happening in the reservoir. Relative permeability curves are varied to match GOR and water cut trends. Interference tests and geochemistry (if available) are used to calibrate propped length and height. DFITs (if available) are used to estimate stress, pore pressure, and permeability. Fiber and downhole imaging are used to calibrate perf efficiency and/or erosion. A variety of other data are used. We have even gotten to work with core-through!
As discussed in SPE 199726, the amount of fracture asymmetry is determined by a competition between fracture stress shadowing and viscous pressure drop. Viscous pressure drop tends to make fractures more symmetrical, because cracks tips further from the wellbore are exposed to lower pressure than crack tips closer to the wellbore, and so tend to propagate less rapidly (causing an evening-out of fracture length). However, stress shadowing tends to create asymmetric propagation. Cracks try to propagate out of each other’s way, which leads them to go left, right, up, and down. The relative strength of viscous pressure drop and stress shadowing depends on the effective fracture toughness, fracture spacing, and the viscosity of the injection fluid. Effective toughness tends to vary by formation, and so the degree of asymmetry may also vary by formation. Note that even though individual fracture strands within a stage may be asymmetrical, the overall geometry of a collection of fractures may be mostly symmetrical. Microseismic lacks the resolution to differentiate individual asymmetrical strands within a stage, where the collection of fractures overall propagates in every direction.
We usually perform pad-scale simulations, involving 2-12 wells. However, we usually run ‘sector’ models in which we simulate only a section of the wells.
In some applications, we do simulate the entire lateral. However, with modern designs, wells may have 500 or more perforation clusters. And often, we are performing simulations that include every well in a pad, or a few pads. Thus, full well simulations would require including 1000s of fractures. This would not be computationally tractable in a 3D fracturing simulator, unless we greatly approximated the fracture geometries and the physics.
We believe that with the right approach, sector models can very effectively answer the great majority of shale design questions. First, they should be ‘history matched’ against data from a ‘typical’ stage (or the typical distribution of observations from stages), rather than by picking out one or a few particular stages. Individual stages may show considerable random variability. Precise matching to stage-to-stage data leads to a model overfit. Second, if there are different sections of the well with different frac design or different geologic properties, consider running multiple separate models to represent those sections.
In some applications, such as a bullheaded refrac, we do need to simulate the full well. In this case, we usually are restricted to considering only one well in the model.
Yes, ResFrac can easily simulate gas injection, either with the black oil, modified black oil, or compositional model.
ResFrac does not simply use constant flow or constant pressure boundary conditions at each cluster. This would lead to significant inaccuracy. Perforation pressure drop between the well and the fracture is calculated dynamically and implicitly in each timestep. The perforation pressure drop calculation follows the standard equation: related to perforation diameter to the fourth power, related to the flow rate to the second power, etc. Perforation diameter adjusts dynamically to reflect perforation erosion as proppant flows through the perforations. Additionally, the well itself is meshed, and the full set of governing equations is solved in each wellbore element in each timestep. The distribution of fluid pressure, proppant concentration, etc. is not uniform along the well.
No, ResFrac simulations are performed on the cloud (we use Microsoft Azure). The simulations are built and visualized locally, but all computations are done on cloud servers. ResFrac has a locally installed user interface for Microsoft Windows that is used to set up simulations, view results, and manage upload/download with the cloud. The locally installed ResFrac user interface does not include the simulation engine.
You need to be connected to the Internet when submitting jobs to run in the cloud, and to download results. You don’t need to be connected for the job to run on the cloud after it is submitted. Therefore, you can set up simulations and view previously downloaded results offline, connect to the Internet to start the simulations, disconnect while the simulations are running (such as for a flight), and then reconnect later to download results.
ResFrac can be used for a variety of applications, requiring different amounts of data. On the one hand, if performing proof of concept, or generic design simulations, you can run ResFrac without any actual data. On the other hand, to perform a confident optimization of frac design in a particular area, you do need to have some data on the reservoir. Generally, better data collection improves the model calibration, and the confidence in the model.
Model inputs include the basic parameters of both a fracturing and a reservoir simulator: porosity, permeability, saturation, stress, modulus, etc. Properties may be defined by layer, on an element-by-element basis, or with a corner point grid. For many model inputs, our UI provides recommended ‘typical’ values. It is also necessary to specify wellbore geometry and perforation information. For most applications, we recommend calibrating against fracturing and production data from at least one well. This includes: the frac design and production rates versus time.
The user-interface provides a variety of ‘wizards’ to automate tasks involved in setting up a simulation, and to help fill gaps in data availability. For example, we provide a wizard to generate a black oil table from a correlation based on initial GOR and other information. We also provide a variety of preview plots to graph tabular model input data, and to view a 3D image of the problem setup. The user-interface includes embedded multimedia help content, including pictures, equations, and movies.
We provide a custom-built, fully featured visualization tool. It provides the option to show multiple panels simultaneously – with either line plots or 3D images. This allows you to visualize multiple properties simultaneously, to see how they interact, and sync up the visualization with tabular data in the line plots. You can visualize and plots dozens of different properties, create slices through the matrix, link or uncouple the camera perspectives in different panels, and toggle the visibility of different plot attributes. Importantly, you can save a ‘layout’ to make it easy to reopen a visualization later, or to repeat the same visualization setup on multiple different simulations. We provide a ‘multiplot’ tool so that you can automatically creates figures from multiple simulations using the same layout, or to make line plots comparing the results from multiple simulations.
We recently released an automated sensitivity analysis tool. From within the user-interface, it is easy to set up a series of sensitivity analysis simulations, and then the tool automatically submits them to the server and downloads the results. We provide post processing tools to compare the results of the sensitivity analysis. We are currently doing internal testing of automated history matching tools, and expect to release later in 2021. We also expect to release automated frac design optimization later in 2021.
Recent content from the ResFrac blog
In this blog post, we review a refracturing and economic optimization case study. The model and history match are loosely based on “SPE Data Repository Well #1,” a publicly available refracturing and production dataset from the Eagle Ford shale. This post extends the analysis that we presented at a recent SPE workshop, “What New for PTA and RTA”.
Commentary on Four New DFIT Papers: (a) Direct In-Situ Measurements of Fracture Opening/Closing from the EGS Collab Project; (b) Comparison of Stress Measurement Techniques from the Bedretto Project; (c) a Statistical Summary of 62 DFITs Interpretations Across Nine Shale Plays; and (d) A Different Perspective: An Article Advocating the Use of the Tangent Method
This post provides commentary on recent four papers on diagnostic fracture injection testing (DFIT). The first paper uses in-situ deformation measurements to directly observe fractures opening and closing during fracture injection-falloff tests (Guglielmi et al., 2022). The second compares various stress measurement techniques in a series of fracture/injection tests from the Bedretto project (Bröker and Ma, 2022). The third statistically reviews results from applying the interpretation procedure from McClure et al. (2019) to 62 DFITs across nine different shale plays (McClure et al., 2022). The fourth is an op-ed written in JPT (Journal of Petroleum Technology) by an advocate of the tangent method for estimating DFIT closure stress (Buijs, 2021; 2022). This article presupposes that the reader already has familiarity with these topics. If you would like more background, please refer to McClure et al. (2019).