Interferences between multiple industrial-scale CO2 injections in a deep saline aquifer

This thesis studies the regional-scale response of an aquifer system to a massive CO$_2$ injection. Industrial-scale CO$_2$ injections into deep saline aquifers affect natural groundwater systems by generating short-term to medium-term pressure gradient perturbations. To evaluate contamination risks and interference risks between injection projects or other uses of underground space, modelling studies become necessary. The geological parameters of underground formations are also to be taken into consideration as they certainly influence the injection reponse. But, saline aquifers are generally poorly-characterized which adds uncertainties to an already complex system. This thesis aims to explore uncertainties in pressure perturbations and CO$_2$ migration predictions, and their consequences in terms of CO$_2$ storage feasibility studies. Firstly, modelling and geological uncertainties have been tested on 2D conceptual models. This step, based on simpler models than 3D ones, allows a fast uncertainties discrimination and save computational time. Hundreds of stochastic realizations are generated to define the influence of permeability spatial variability. To limit the number of flow simulations, selection procedures of realizations are applied and tested. Selections are derived from fast-calculations methods called "'proxy-response"'. Secondly, once these methods have been 2D tested and validated, and once a number of uncertainties have been eliminated, these methods and related ones are applied to the underground system 3D modelling. The 3D models have been built based on available data from an existing Paris Basin hydrogeological model. Several injection scenarios have been considered and tested. Permeability spatial variability and pore compressibility are the two main parameters chosen to evaluate the injection response. This last step allows a better definition of interference risks between the major uncertainties from geological parameters and injection-related physical parameters.