Modelling of CO2 Injection in Fluvial Sedimentary Heterogeneous Reservoirs to Assess the Impact of Geological Heterogeneities on CO2 Storage Capacity and Performance

International audienceThe integration of sedimentary heterogeneity in flow simulations is still a challenging issue in reservoir engineering modelling. Yet in the CCS context, the discussion on how the heterogeneities may impact the flow performances is at the beginning. Fluvial reservoirs consist of complex heterogeneous sedimentary bodies with varying connectivity, controlled by the sedimentary history of the system. The main difficulty is to handle two different scales when evaluating the storage capacity performance of such reservoirs: (i) large scale of the pressure footprint (∼10km) and (ii) small scale of the sedimentary heterogeneity (∼m). This induces the generation of reservoir grids containing several 100 000 of gridblocks making the dynamic flow simulation difficult to handle. In addition to that, the high level of uncertainty requires the generation of several models to cover a large spectrum of equi-probable solution. The present work focuses on fluvial reservoir performances using a stochastic algorithm to reproduce deterministically realistic architectural models with high resolution heterogeneities. Each geological model is split into two identical architectural models that differ in their sedimentary fill. One considers the reservoir bodies (fluvial belts) as homogeneous stacked point bar sand bodies (1st order heterogeneity, Model B), while the second type contains also flow barriers (shale oxbow lakes; 2nd order heterogeneity, Model A). To perform the CO2 injection two codes were used: TOUGH2-MP (integrated finite volume approach with massive parallel implementation) and 3DSL (streamline black oil simulator on a single processor). Two sets of 50 realizations were assessed (Models A and B). The streamline simulations enable quick ranking of dynamic capacity estimate at the scale of the geological model while only eight models could be conducted in a reasonable CPU time framework with TOUGH2-MP. This is explained by the high detailed characterisation of the fluid properties coupled with the flow which lowers substantially the speed of calculations.The study reveals that heterogeneities affect the storage capacity as well as the injectivity of the well. For a reservoir formation with a typical size of 23km x 25 km x 60m, the capacities vary between 2.5 and 11Mt. The presence of oxbow lakes induces a loss of capacity that varies between 1.2Mt (23%) and 1Mt (12%) after 10 years, and 30 years, respectively