Robust chemical solver for fully-implicit simulations

The outcome of water-based Enhanced Oil Recovery (EOR) processes, such as polymer and surfactant injection, highly depends on the chemical composition of the displacing fluid in the reservoir. A robust and efficient solver for geochemistry is therefore crucial to design successful EOR strategies. In this talk, we present the implementation of a new geochemistry solver as a module of the open source MRST (Matlab Reservoir Simulation Toolbox). The solver includes non-isothermal aqueous speciation, solid and gas phase equilibrium, and full surface complexation chemistry. Automatic differentiation is used to assemble the linearized equations and enables to easily set up different systems and models. The residual equations are obtained after a log-log transformation. Such a transformation is likely to bring robustness to the solver by increasing the convexity of the governing equations. An other important ingredient is a-priori bounds on the concentrations which when incorporated in the solver, improve the convergence of the Newton iterations. The solver can be easily combined within MRST with a transport solver. Numerical simulation of low salinity water injection will be presented to illustrate the solver capabilities.