MRST

MRST
Name	MRST
Genre	Numerical simulation toolbox
Developer	SINTEF, University of Oslo, ConocoPhillips
Initial release	2006
Programming language	MATLAB
Operating system	Linux, Windows, macOS
License	GPL

MRST

MRST is a MATLAB-based simulation toolbox designed for reservoir simulation, numerical methods, and porous media research. It provides modular components for discretization, upscaling, multiphase flow, rock physics, and optimization, enabling researchers and industry practitioners from ConocoPhillips, Equinor, TotalEnergies, Shell and academic groups at University of Oslo and SINTEF to prototype workflows bridging field development, history matching, and uncertainty quantification. MRST's architecture links algorithmic building blocks to applications in enhanced oil recovery, CO2 storage, geothermal energy, and hydrology, supporting collaborations with institutions such as Stanford University, Imperial College London, ETH Zurich and Lawrence Berkeley National Laboratory.

Overview

MRST bundles discretization schemes, linear solvers, and utility modules under a consistent MATLAB API. Core components include structured and unstructured grid representations borrowed from practices at Society of Petroleum Engineers, finite-volume and mimetic discretizations influenced by work at Los Alamos National Laboratory and Oxford University, and robust nonlinear solvers comparable to implementations at Schlumberger and Kongsberg Gruppen. The toolbox emphasizes rapid prototyping tied to data-management formats used by Eclipse (software), PVT data standards, and field-scale datasets typical of studies presented at the SPE Annual Technical Conference and Exhibition.

History

MRST originated in the mid-2000s at SINTEF and the University of Oslo to transfer academic developments in discretization and upscaling into practical simulation workflows used by ConocoPhillips and other industry partners. Early releases integrated schemes from the mimetic finite-difference community associated with University of Texas at Austin and algorithmic contributions from researchers linked to Los Alamos National Laboratory and Imperial College London. Over successive versions MRST incorporated modules for compositional simulation influenced by methods from Stanford University and TotalEnergies Research & Technology, and expanded support for multiphysics coupling pursued in collaborations with ETH Zurich and Lawrence Berkeley National Laboratory. Community workshops at venues such as the SPE Reservoir Simulation Symposium and academic meetings at NTNU contributed to its dissemination.

Applications

Researchers and engineers apply MRST to problems in hydrocarbon recovery, CO2 sequestration, geothermal reservoir management, and groundwater remediation. Studies using MRST have modelled EOR workflows like polymer flooding and chemical injection referencing case studies from North Sea fields and Gulf of Mexico reservoirs, and CO2 injection scenarios tied to projects such as Sleipner and In Salah CO2 Storage. MRST has been used to evaluate well placement and optimization problems similar to those addressed by Schlumberger Eclipse workflows, to design monitoring strategies analogous to work by Equinor for subsea projects, and to prototype workflows for coupled thermal-hydraulic simulations relevant to Geysers Geothermal Field studies and Iberian Peninsula geothermal assessments.

Methodology

The toolbox implements finite-volume, two-point flux approximation, multipoint flux approximation, and mimetic discretizations drawing on mathematical advances from Oxford University, Princeton University, and Université Paris-Sud. Nonlinear solvers use Newton–Raphson strategies aligned with techniques from Lawrence Livermore National Laboratory and robust line-search and trust-region logic similar to methods reported by Stanford University. Upscaling and multiscale strategies incorporate ideas from multiscale finite-volume methods developed at University of Texas at Austin and Imperial College London, while adaptive timestep control and sensitivity-based parameter estimation echo algorithms from Los Alamos National Laboratory and Sandia National Laboratories.

Software Implementations

MRST is implemented in MATLAB with performance-critical kernels optionally implemented in compiled languages and integrated via Mex interfaces as practiced by groups at Argonne National Laboratory. The codebase provides plugins to import and export formats used by Eclipse (software), Petrel, and common data standards used by Norwegian Petroleum Directorate datasets. Parallelization strategies in MRST employ shared-memory paradigms compatible with multicore Intel and AMD processors and mirror approaches used in toolkits from MathWorks and Scilab communities.

Performance and Validation

Validation of MRST against industrial simulators has been reported in benchmark studies comparing MRST results to Eclipse (software), CMG and in academic test cases such as the SPE Comparative Solution Projects used at SPE events. Performance profiling identifies bottlenecks in linear solvers and matrix assembly similar to observations in development reports from Sandia National Laboratories and improvements rely on solver libraries and preconditioners akin to those in PETSc and Trilinos. Case studies show MRST reproducing displacement fronts, pressure responses, and tracer breakthrough curves consistent with field data from Statfjord, Ekofisk, and other North Sea operations.

Criticisms and Limitations

Critics note that MRST's dependence on MATLAB imposes licensing and scalability constraints compared with native implementations in languages promoted by Argonne National Laboratory and Lawrence Livermore National Laboratory. The toolbox's performance for very large field models can lag optimized commercial simulators such as Eclipse (software) and CMG without substantial low-level optimization or external solver integration. Additionally, adaptation to reservoir characterization workflows originating from vendors like Schlumberger and interoperability with proprietary datasets require custom adapters commonly developed in collaborations with organizations like Equinor and TotalEnergies.

Category:Numerical software