MRST-AD – an Open-Source Framework for Rapid Prototyping and Evaluation of Reservoir Simulation Problems

Krogstad, Stein; Lie, Knut–Andreas; Møyner, Olav; Nilsen, Halvor Møll; Raynaud, Xavier; Skaflestad, Bård

doi:10.2118/173317-ms

Cited by 120 publications

(77 citation statements)

References 33 publications

(39 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We believe that such solvers are particularly important to efficiently and robustly resolve the important stationary states of flat CO 2 interfaces. The fully-implicit solvers in MRST-co2lab are implemented using automatic differentiation [12], which makes it straightforward to compute gradients from adjoint equations. Having gradients is important to determine parameter sensitivities or if one wants to use rigorous mathematical optimization methods to propose plausible injection points and strategies.…”

Section: Choice Of Numerical Methodsmentioning

confidence: 99%

“…The four simulations were performed using a generalpurpose, fully-implicit black-oil solver from MRST [12]; more details will be given in the next section. Computational costs are reported in Table 1.…”

Section: Derivation Of the Basic Ve Modelmentioning

confidence: 99%

“…At the end of simulation, only 3 % has left our computational domain, and the mobile CO 2 has reached a region with a very low sloping angle and is thus almost immobilized. Figure 10 reports CPU time per step in a sequentially implicit simulator that utilizes a standard two-point pressure solver and a fully-implicit transport solver from the incomp module of MRST [10] and a fully-implicit blackoil solver based on automatic differentiation [12]. Both simulators have time-step control and will, if necessary, reduce the time steps to ensure convergence.…”

Section: Derivation Of the Basic Ve Modelmentioning

confidence: 99%

“…To contribute to this and to accelerate transfer of knowledge developed in academic research projects to end-users in industry and the public sector, we have started developing a community software called MRST-co2lab [4][5][6][7][8] implemented on top of the open-source Matlab Reservoir Simulation Toolbox (MRST) [9][10][11][12]. Both MRST and its CO 2 module can be freely downloaded and used under the GNU General Public License v3.0.…”

Section: Introductionmentioning

confidence: 99%

“…In particular, we point out in detail the approximations and numerical considerations needed to obtain flexible and efficient numerical formulations that resolve the main physical effect well within the accuracy normally available from input data. The overall formulation is implemented as an extension of existing black-oil solvers in MRST [12], which in turn have been implemented using automatic differentiation and hence enable simple computation of gradients and parameter sensitivities, e.g., through an adjoint formulation. This enables the users to easily perform sensitivity studies or formulate efficient strategies for rigorous mathematical optimization of large-scale injection strategies [6,7,48,49].…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Robust simulation of sharp-interface models for fast estimation of CO2 trapping capacity in large-scale aquifer systems

2015

Self Cite

View full text Add to dashboard Cite

Modeling geological carbon storage represents a new and substantial challenge for the subsurface geosciences. To increase understanding and make good engineering decisions, containment processes and large-scale storage operations must be simulated in a thousand-year perspective. Large differences in spatial and temporal scales make it prohibitively expensive to compute the fate of injected CO 2 using traditional 3D simulators. Instead, accurate forecast can be computed using simplified models that are adapted to the specific setting of the bouyancy-driven migration of the light fluid phase. This paper presents a family of vertically integrated models for studying the combined large-scale and long-term effects of structural, residual, and solubility trapping of CO 2 . The models are based on an assumption of a sharp interface separating CO 2 and brine and can provide a detailed inventory of the injected CO 2 volumes over periods of thousands of years within reasonable computational time. To be compatible with simulation tools used in industry, the models are formulated in a black-oil framework. The models are implemented in MRST-co2lab, which is an open community

show abstract

Section: Choice Of Numerical Methodsmentioning

confidence: 99%

Section: Derivation Of the Basic Ve Modelmentioning

confidence: 99%

Section: Derivation Of the Basic Ve Modelmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Robust simulation of sharp-interface models for fast estimation of CO2 trapping capacity in large-scale aquifer systems

2015

Self Cite

View full text Add to dashboard Cite

show abstract

Adaptive implicit finite element methods for multicomponent compressible flow in heterogeneous and fractured porous media

Moortgat

2017

Water Resources Research

View full text Add to dashboard Cite

This work presents adaptive implicit first‐order and second‐order discontinuous Galerkin (DG) methods for the transport of multicomponent compressible fluids in heterogeneous and fractured porous media, discretized by triangular, quadrilateral, and hexahedral grids. The adaptive implicit method (AIM) combines the advantages of purely explicit or implicit methods (in time). In grid cells with high fluxes or low pore volumes, the transport update is done implicitly to alleviate the Courant‐Friedrichs‐Lewy (CFL) time step constraints of the conditionally stable explicit approach. Grid cells with a large CFL condition are updated explicitly. Combined, this allows higher efficiency than explicit methods, but it reduces the “penalty” of implicit methods, which exhibit high numerical dispersion and are more computationally and storage expensive per time step. The advantages of AIM are modest for uniform grids and rock properties. However, in heterogeneous or fractured reservoirs explicit methods may become impractical, while a fully implicit approach introduces unnecessary numerical dispersion and is overkill for low‐permeability layers and matrix blocks. In such applications, AIM is shown to be significantly more efficient and accurate. The division between explicit and implicit grid cells is made adaptively in space and time. This allows for a high level of explicitness and can also adapt to high fluxes caused by, e.g., viscous and gravitational flow instabilities. Numerical examples demonstrate the powerful features of AIM to model, e.g., solute transport, carbon sequestration in saline aquifers, and miscible gas injection in fractured oil and gas reservoirs.

show abstract

Accelerating oil reservoir simulations using POD on the fly

Clainche

Varas

Vega

2016

Numerical Meth Engineering

View full text Add to dashboard Cite

A reduced order model (ROM) is presented for the long-term calculation of subsurface oil/water flows. As in several previous ROMs in the field, the Newton iterations in the full model (FM) equations, which are implicit in time, are projected onto a set of modes obtained by applying proper orthogonal decomposition (POD) to a set of snapshots computed by the FM itself. The novelty of the present ROM is that the POD modes are (i) first calculated from snapshots computed by the FM in a short initial stage, and then (ii) updated on the fly along the simulation itself, using new sets of snapshots computed by the FM in even shorter additional runs. Thus, the POD modes adapt themselves to the local dynamics along the simulation, instead of being completely calculated at the outset, which requires a computationally expensive preprocess. This strategy is robust and computationally efficient, which is tested in 10-and 30-year simulations for a realistic reservoir model taken from the SAIGUP project.ACCELERATING OIL RESERVOIR SIMULATIONS USING POD ON THE FLY Here, t is the time variable, is the porosity (void fraction of the rock) and, for j D o (oil) and w (water), m j accounts for sources and sinks associated with the injection and production wells, ¶ www.nr.no/en/SAIGUP obtained with the full model (blue) and the reduced-order model (ROM) (green) in the 30-year simulation.

show abstract

MRST-AD – an Open-Source Framework for Rapid Prototyping and Evaluation of Reservoir Simulation Problems

Cited by 120 publications

References 33 publications

Robust simulation of sharp-interface models for fast estimation of CO2 trapping capacity in large-scale aquifer systems

Robust simulation of sharp-interface models for fast estimation of CO2 trapping capacity in large-scale aquifer systems

Adaptive implicit finite element methods for multicomponent compressible flow in heterogeneous and fractured porous media

Accelerating oil reservoir simulations using POD on the fly

Contact Info

Product

Resources

About