JutulDarcy.jl: Darcy-scale and subsurface flow (CO2 sequestration, gas/H2 storage, oil/gas fields) using Jutul.jl developed by the Applied Computational Science group at SINTEF Digital.
JutulDarcy is a general purpose porous media simulator with high performance written in Julia. It is fully differentiable with respect to forces and discretization parameters.
- Immiscible multi-phase flow
- Black-oil type models with support for both dissolved vapor (Rs) and vaporized liquid (Rv)
- Equation-of-state compositional flow with up to three phases and any number of components
All solvers can incorporate general multisegment wells with rigorous mass balance, friction pressure loss, complex well limits and time-dependent controls.
- Written in pure Julia, with automatic differentiation and dynamic sparsity detection
- Support for sensitivities with respect to any model parameters using the adjoint method
- High performance assembly and linear solvers, with support for two-stage CPR BILU(0)-CPR Krylov solvers
- MPI support with domain decomposition and BoomerAMG-CPR solver with automatic METIS partitioning
- Unstructured grids and complex cases input from the Matlab Reservoir Simulation Toolbox (MRST) using the
jutulmodule - Experimental support for reading in and running .DATA files with corner point grids
- 3D visualization of grids and wells by loading a Makie.jl backend (requires Julia 1.9,
GLMakiefor interactivity) - Interactive plotting of well curves
The compositional simulator has been matched against commercial offerings, AD-GPRS and MRST. The blackoil simulator has been validated on the standard SPE benchmarks (SPE1, SPE9, ...).
| Name | Cells | Report steps | Preconditioner | Time [s] |
|---|---|---|---|---|
| SPE1CASE2 | 300 | 120 | block-ILU(0) | 0.30 |
| SPE9 | 9000 | 35 | block-ILU(0) | 3.41 |
| Egg | 18553 | 123 | CPR-block-ILU(0) | 8.60 |
Simulated with default julia, no threads. Runtime with single-threaded HYPRE on the full SPE10 model 2 model is about 20 minutes with default settings and strict mass-balance tolerances and less than 4 minutes for realization 1 of the OLYMPUS optimization benchmark challenge.
Jutul builds upon many of the excellent packages in the Julia ecosystem. Here are a few of them, and what they are used for:
- ForwardDiff.jl implements the Dual number class used throughout the code
- SparsityTracing.jl provides sparsity detection inside Jutul
- Krylov.jl provides the iterative linear solvers
- ILUZero.jl for ILU(0) preconditioners
- AlgebraicMultigrid.jl for AMG preconditioners
- HYPRE.jl for robust AMG preconditioners with MPI support
- PartitionedArrays.jl for MPI assembly and linear solve
- Tullio.jl for automatically optimized loops and Polyester.jl for lightweight threads
- TimerOutputs.jl and ProgressMeter.jl gives nice output to terminal
- Makie.jl is used for the visualization features
- MultiComponentFlash.jl provides many of the compositional features
...and many more, both directly in the Project.toml file and indirectly!
Install Julia and add the package to your environment of choice:
using Pkg
Pkg.add("CairoMakie")
Pkg.add("Jutul")
Pkg.add("JutulDarcy")You can then run any of the examples in the examples directory by including them.
The documentation is still work in progress, but contains a fair bit of useful information. In addition, see the examples folder for more information. Some functionality is also demonstrated in the tests.
Internals and undocumented functions are subject to change at this time. However, the main interface for the reservoir simulator itself seen in the examples should be fairly stable.