PRST - Python Reservoir Simulation Toolbox

Abstract

In this thesis I have developed a prototype for a software framework for the simulation of flow in porous media, based on the MATLAB Reservoir Simulation Toolkit (MRST). The software -- Python Reservoir Simulation Toolkit -- is written in the Python programming language. First, the equations for single-phase flow in porous media are developed and discretized using the finite-volume method. %TODO dsicertte? The nonlinear systems of equations that arise are solved using Newton's method, with Jacobian matrices found using automatic differentiation.

Automatic differentiation is a method for finding gradients and Jacobians of programatically defined functions. It is based on decomposing a function into elementary operations such as addition and multiplication, and then using the chain rule repeatedly.

Then, MATLAB and Python's capabilities are compared. Software unit testing is introduced as a tool to ensure the correctness of PRST functions compared to the equivalent MRST functions. Unstructured grids are introduced -- and several tools for visualizing unstructured grids are investigated. After that, an example of how compilation and just-in-time compilation can be used to improve Python performance is shown, using Cython and Numba. Finally, a flow simulator example from MRST is converted to use PRST, and the visualization capability and performance is compared.