Elsevier

Journal of Computational Physics

Volume 346, 1 October 2017, Pages 514-538
Journal of Computational Physics

Fully implicit mixed-hybrid finite-element discretization for general purpose subsurface reservoir simulation

https://doi.org/10.1016/j.jcp.2017.06.034Get rights and content

Abstract

We present a new fully-implicit, mixed-hybrid, finite-element (MHFE) discretization scheme for general-purpose compositional reservoir simulation. The locally conservative scheme solves the coupled momentum and mass balance equations simultaneously, and the fluid system is modeled using a cubic equation-of-state. We introduce a new conservative flux approach for the mass balance equations for this fully-implicit approach. We discuss the nonlinear solution procedure for the proposed approach, and we present extensive numerical tests to demonstrate the convergence and accuracy of the MHFE method using tetrahedral elements. We also compare the method to other advanced discretization schemes for unstructured meshes and tensor permeability. Finally, we illustrate the applicability and robustness of the method for highly heterogeneous reservoirs with unstructured grids.

Introduction

Management of subsurface resources, including water aquifers and oil/gas reservoirs relies on numerical reservoir simulation. Flow simulation entails discretization of the equations and constitutive relations that describe multi-component, multiphase flow in subsurface porous formations. The reservoir characterization model (RCM) that serves as input to the flow simulator can be quite complex, both in terms of the grid geometry and the formation properties (e.g., permeability). Each gridblock is assigned material properties and, in most cases, the properties are different in each block. For highly-heterogeneous fields, comprised of many stratigraphic compartments, the number of gridblocks can be extremely large, and the computational domain can be highly unstructured. Therefore, accurate modeling of the local flow dynamics, especially the fluxes between the gridblocks is of great importance. Moreover, the highly nonlinear nature of the coupled conservation equations and constitutive relations make honoring local conservation a fundamental requirement.

Traditionally, a two-point flux approximation (TPFA) technique is employed to approximate the phase flux in the Darcy equation [1]. As the name implies, two points are used to approximate the flux at the interface between the control-volumes (gridblocks). The TPFA scheme requires the mesh of the domain to be structured, such that the interfaces are orthogonal to the interface between control-volumes, and the grid is aligned with the principal directions of any anisotropic material property, e.g., the permeability. There is strong interest in the community to use high-resolution, geometrically complex RCMs that represent the large-scale subsurface formations as accurately as possible. Accommodating three-dimensional unstructured grids with full-tensor properties is a difficult – practically impossible – task using TPFA. Multi-point flux-approximation (MPFA) discretization schemes have been developed to address issues related to nonorthogonal grids and full-tensor heterogeneous permeability [2], [3]. Similar to TPFA schemes, divergence-free conditions are imposed on the control-volume to approximate the flux. MPFA schemes – subject to limitations – allow for using unstructured grids and anisotropic material properties [4]. Several variations of the method have been developed [5], [6] with corresponding convergence studies [7], [8], [9]. In practice, the ‘MPFA-O’ method is the most widely used. More recently, nonlinear flux approximation schemes have been developed [10], [11], [12].

Another flux approximation scheme for unstructured meshes is the mixed-finite-element method (MFEM), where the mass and momentum equations are coupled and solved simultaneously [13]. This method is locally conservative, and it can accommodate high-order approximations and anisotropic material properties. However, in its original form, MFEM leads to an algebraic system of saddle-point type, whereby the linearized matrix is indefinite. An efficient variant for the method was developed to overcome the challenges associated with saddle-point problems, in which Lagrange multipliers for the primary variables are placed on the elements interfaces [14]. This approach is called the mixed-hybrid finite-element (MHFE) method. Here, we present a new fully-implicit method for the MHFE formulation of general-purpose reservoir simulation. Another approach to overcome the saddle-point problem is to eliminate the momentum balance equation from the linear system and replace it with a particular MPFA scheme in the mass balance equation using special quadrature rules [15]; this elimination procedure can be considered a special case for the MPFA-O method [16], [17]. Previous MHFE-based efforts include single-phase flow [18]; [19] used the MHFE method for modeling two-phase incompressible fluids; [20] extended the method for compositional equation-of-state models, and gravity was included by [21]. However, all these schemes were designed for specific problems and have not been applied for general-purpose reservoir simulation where all the main physics are modeled and simulated in-conjunction. The existing literature on MHFE has employed explicit, or partially explicit, temporal discretization. In these schemes, the governing equations for the system are decoupled, and the explicit treatment greatly restricts the time steps that can be used. More recently, sequential temporal discretization has been applied for MHFE without capillary pressure [22].

In this work, we employ the MHFE method for general-purpose compositional reservoir simulation using a fully-implicit method. The velocities are approximated with Raviart–Thomas vector basis functions [23], which ensure local conservation of momentum. The scalar variables: pressure, phase saturations, and component concentrations, are discretized using the locally conservative finite volume method. We employ a fully implicit scheme whereby the conservation equations are coupled, and the system is solved using the Newton–Raphson method. The nonlinear variables are spatially discretized on the elements and on element interfaces by the MHFE process. We also introduce a new continuous flux approach that guarantees that the momentum is conservative for the mass balance equations. This is the first fully-implicit temporal discretization and implementation for the MHFE for isothermal general-purpose reservoir simulation.

The MHFE framework is integrated into the Automatic Differentiation General Purpose Research Simulator (AD-GPRS) [24], [25]. This framework includes EOS-based multiphase multi-component formulation [26], [27], thermal-compositional capabilities [28], [29], fully coupled geomechanics [30], chemical reactions [31] and adjoint gradients, [32] which makes it suitable for solution of a wide range of subsurface problems. To validate the proposed discretization approach, we perform extensive numerical tests highlighting the convergence and accuracy for the different physical phenomena. We also compare the discretization to other advanced schemes. Moreover, we apply the method for problems with the emphasis on unstructured grids and tensor permeability.

The paper is organized as follows. In Section 2, the governing equations are discussed. In Section 3, the MHFE formulation is described, and in Section 4, the nonlinear solution is detailed. We present the numerical tests and examples in Sections 5 and 6. Finally, in Section 7, we present our conclusions.

Section snippets

Governing equations

In this section, we present the equations that govern the fluid flow and the thermodynamic phase equilibrium for isothermal compositional reservoirs.

Mixed hybrid finite element method

In this work, we use the MHFE method to discretize the conservation equations. In this method, we use vector basis functions of having a flux of one at interface and zero elsewhere (see Fig. 1).AjEWiEnjE={1ifi=j0ifij where WiE is the vector basis function of interface i, njE is the normal outward oriented vector on interface j and AjE is the area of interface j of element E. Another property of the mixed formulation is that W is constant over an element,VEWiE=1 where VE is volume of

Nonlinear solver

In this section, we discuss the procedure for the nonlinear solution for the unknowns described in the previous sections. We use the backward Euler method for the temporal discretization and the Newton–Raphson method to linearize the system, in which a Jacobian system is assembled and solved for each nonlinear iteration for our fully-implicit system:[JPPJPλJλPJλλ][yPyλ]=[RPRλ],p,Sα,xc,αyP,λp,λc,αyλ

Here, RP is residual for the thermodynamic equilibrium constraint and the mass balance

Numerical convergence

We present numerical tests which demonstrate the accuracy and convergence of the fully-implicit MHFE method for several problems. First, we test the convergence of the method to an existing analytical solution for a single-phase flow in heterogeneous domain in the presence of a barrier and a full permeability tensor. Then, we validate the applicability of the method for flow in the presence of strong capillarity and gravity forces, and we compare the results with analytical solution. In

Applied computational cases

In this section, we present applied computational cases to illustrate the capability and the robustness of our method for general purpose reservoir simulation. First, we test the capability of the method for modeling multiphase flow in discontinuous anisotropic domains. Then, we study the effect of geological faults for the migration of hydrocarbon components between reservoirs. Finally, we present a challenging heterogeneous case which is a modification of the SPE10 benchmark test [56] with

Concluding remarks

In this paper, we presented a fully implicit discretization scheme for the mixed hybrid finite element (MHFE) method for general-purpose compositional reservoir simulation. The scheme couples the momentum and mass balance equations and solves them simultaneously, and the mass balance is discretized using the finite volume method. We introduced a new flux approach for the mass balance equations that is conservative. The system was linearized using the Newton–Raphson method, and the non linear

Acknowledgements

We thank the Qatar Foundation – Research & Development for funding this research. We also thank the SUPRI-B Reservoir Simulation consortium at Stanford University.

References (59)

  • M.L. Michelsen

    The isothermal flash problem. Part II. Phase-split calculation

    Fluid Phase Equilib.

    (1982)
  • A.S. Abushaikha et al.

    Interface control volume finite element method for modelling multi-phase fluid flow in highly heterogeneous and fractured reservoirs

    J. Comput. Phys.

    (2015)
  • X. Wang et al.

    Trust-region based solver for nonlinear transport in heterogeneous porous media

    J. Comput. Phys.

    (2013)
  • K. Lipnikov et al.

    The mimetic finite difference discretization of diffusion problem on unstructured polyhedral meshes

    J. Comput. Phys.

    (2006)
  • K. Schmid et al.

    Higher order FE–FV method on unstructured grids for transport and two-phase flow with variable viscosity in heterogeneous porous media

    J. Comput. Phys.

    (2013)
  • R. Worden et al.

    H2S-producing reactions in deep carbonate gas reservoirs: Khuff Formation, Abu Dhabi

    Chem. Geol.

    (1996)
  • K. Aziz et al.

    Petroleum Reservoir Simulation

    (1979)
  • I. Aavatsmark

    An introduction to multipoint flux approximations for quadrilateral grids

    Comput. Geosci.

    (2002)
  • M.G. Edwards

    Higher-resolution hyperbolic-coupled-elliptic flux-continuous CVD schemes on structured and unstructured grids in 3-D

    Int. J. Numer. Methods Fluids

    (2006)
  • I. Aavatsmark et al.

    A compact multipoint flux approximation method with improved robustness

    Numer. Methods Partial Differ. Equ.

    (2008)
  • Y. Cao et al.

    Geometrical interpretation of the multi-point flux approximation L-method

    Int. J. Numer. Methods Fluids

    (2009)
  • G.T. Eigestad et al.

    On the convergence of the multi-point flux approximation O-method: numerical experiments for discontinuous permeability

    Numer. Methods Partial Differ. Equ.

    (2005)
  • M. Pal et al.

    q-Families of CVD(MPFA) schemes on general elements: numerical convergence and the maximum principle

    Arch. Comput. Methods Eng.

    (2010)
  • K. Lipnikov et al.

    Minimal stencil finite volume scheme with the discrete maximum principle

    Russ. J. Numer. Anal. Math. Model.

    (2012)
  • K. Nikitin et al.

    A monotone nonlinear finite volume method for diffusion equations and multiphase flows

    Comput. Geosci.

    (2014)
  • K.M. Terekhov et al.

    Cell-centered nonlinear finite-volume methods for the heterogeneous anisotropic diffusion problem

    J. Comput. Phys.

    (2016)
  • F. Brezzi et al.

    Mixed and Hybrid Finite Element Methods

    (1991)
  • M.F. Wheeler et al.

    A multipoint flux mixed finite element method

    SIAM J. Numer. Anal.

    (2006)
  • S.F. Matringe et al.

    Mixed-finite-element and related-control-volume discretizations for reservoir simulation on three-dimensional unstructured grids. Society of petroleum engineers

  • Cited by (52)

    • An implementation of mimetic finite difference method for fractured reservoirs using a fully implicit approach and discrete fracture models

      2021, Journal of Computational Physics
      Citation Excerpt :

      Based on this, we will use the discrete fracture model to represent the fractures in this paper. Over the past two decades, the computational methods to characterize the fluid flow with discrete fracture models have mainly included the finite volume (FV), [18,21–25], the finite element (FE), [16,17,26,27], and the mixed finite element (MFE) methods [28–32]. Unlike the MFE and the FV methods, the velocity in the FE method is discontinuous across the elements and it may lead to inaccurate results [22].

    • A novel block non-symmetric preconditioner for mixed-hybrid finite-element-based Darcy flow simulations

      2021, Journal of Computational Physics
      Citation Excerpt :

      The reference model for the development of our preconditioner is the basic MHFE-FV discretized single-phase flow in porous media. This represents the first stage of a more comprehensive research project, where we plan to extend this preconditioning framework to a MHFE-FV multi-phase reservoir simulator based on the modeling approach described in [22]. The rest of the paper is organized as follows.

    • A multiscale mixed finite element method applied to the simulation of two-phase flows

      2021, Computer Methods in Applied Mechanics and Engineering
    • A fully implicit mimetic finite difference scheme for general purpose subsurface reservoir simulation with full tensor permeability

      2020, Journal of Computational Physics
      Citation Excerpt :

      For the mass balance equations, we employ the locally conservative single point upstream method. We also strongly impose the fluxes straddled between grids to accelerate the convergence for the nonlinear solver, as applied in [1]. Moreover, we validate the proposed discretization approach by performing extensive numerical tests for convergence and accuracy versus other discretization schemes.

    View all citing articles on Scopus
    View full text