Compositional reservoir simulation-based reactive-transport formulations, with application to CO2 storage in sandstone and ultramafic formations

Advisors

Louis Durlofsky, primary advisor
Hamdi Tchelepi, primary advisor
Katharine Maher, advisor

Abstract

Chemical reactions are important in many simulation applications, including geological carbon storage. The incorporation of chemical reaction treatment in general compositional reservoir simulators is thus necessary to enable this modeling. In this work, we develop robust numerical schemes for modeling CO₂ sequestration. All of the methods developed are implemented into Stanford's Automatic Differentiation-based General Purpose Research Simulator (AD-GPRS). We first address a special case of crossing thermodynamic phase boundaries, i.e., aqueous phase disappearance and reappearance in the context of CO₂ sequestration. A specialized treatment for handling aqueous-phase components when the aqueous phase disappears (or reappears) is introduced under the natural set of variables. This variable set includes pressure, phase saturations, and phase compositions. We demonstrate the robustness of our fully-implicit natural-variable formulation for carbon storage simulations, even when the aqueous phase disappears in multiple grid blocks. We also propose a novel reactive transport formulation based on overall-composition variables. This formulation effectively treats the aqueous phase disappearance phenomenon, because the overall-composition variables are valid for all fluid-phase combinations. Overall-composition variables, however, suffer from the high cost of thermodynamic calculations in two-phase grid blocks. This motivates the development of a hybrid numerical scheme which takes advantage of the favorable features of both the natural and overall-composition variable formulations. Simulation results for CO₂ sequestration scenarios with the three formulations demonstrate the stability of these schemes. A comparison of the numerical performance of these treatments suggests that the use of natural variables in general offers enhanced computational efficiency compared to overall-composition variables. Under the natural-variable formulation, however, one of the special treatments proposed in this work should be considered for grid blocks with single-phase gas. We next investigate the use of ultramafic rocks for geological carbon storage. These rocks are highly reactive and offer considerable CO₂ storage capacity. We begin by analyzing a weathering system in this type of rock, where our AD-GPRS implementation is validated against field observations. We then simulate idealized carbon storage projects in an ultramafic reservoir. The general features and patterns of carbonation are identified and discussed. This type of rock offers nearly complete conversion of the injected CO₂ to mineral forms in many cases, enhancing storage security. Sensitivity analyses are conducted to examine the impact of various reservoir properties and operation parameters on carbonation efficiency. We demonstrate that well control scenarios can be designed to improve the carbonation process substantially by providing a more effective distribution of the injected CO₂ in the formation.