PhD Thesis defense of Laurez Maya

Abstract

When exploiting renewable resources, such as geothermal energy, the injection of fluids into underground reservoirs can drastically impact the permeability of the porous medium near the injection wells. Fine suspended particles (colloids), whether initially present in the injected fluids or detached from the porous matrix by the pressure gradient, are transported, aggregated, irreversibly or reversibly deposited,
and/or lead to pore clogging. The consequence of this pore-clogging (filtration, bridging, or particle aggregation) on permeability results in a drastic decrease in injectivity in the wells, potentially leading to their abandonment. Studying clogging phenomena is crucial to better control injectivity and propose effective unclogging solutions to maintain well exploitation. Thus, this work aims to understand the evolution of the permeability of a porous medium during the injection of a suspension, to predict the injectivity drop, and to optimize injection processes through numerical models. At the scale of exploitation sites (macroscopic), classical approaches for modeling particle transport and clogging rely on heuristic parameters and restrictive assumptions that limit their predictive capabilities. Notably, the consideration of electrochemical effects on particle deposition, aggregation, and detachment, as well as their feedback on flow, can be improved. This thesis aims to provide a solution for modeling colloidal transport in porous media. The strategy adopted is based on a cascade modeling approach across spatio-temporal scales of the porous medium. First, we focus on microscopic scales (molecular, pore, and pore network scale) where particle-fluid and fluid-matrix interfaces, the sites of hydromechanical and electrochemical phenomena controlling clogging mechanisms, are well described. We have developed and validated a new numerical approach to simulate colloidal transport at the pore scale. It is based on an Euler-Lagrange method of the CFD-DEM type, where the fluid is described by a continuous phase and particle transport is represented by a discrete phase (individual tracking). In particular, our approach overcomes the classical limitations on the size of computational cells relative to the size of particles. The developed model serves as a foundation for studying the predominance of physicochemical variables on clogging (infiltration velocity, particle concentration, solution pH and salinity, pore and particle size, etc.). Subsequently, and in a logic of upscaling, we no longer consider particles as discrete elements but as a concentration field. To do this, we revisit the theory of colloidal deposition around a cylinder to analytically determine macroscopic deposition kinetic laws. Finally, the numerical model is used to simulate particle retention in porous media. It captures the three main clogging mechanisms (size exclusion, arch formation, and aggregation). It allows for determining porosity-permeability relationships and retention kinetics depending on flow regimes, solution chemistry, and suspension properties represented by appropriate dimensionless numbers. The advances brought by this work improve the understanding of clogging mechanisms and guide the development of models on larger scales.