Characterizing Fluvial reservoir using topological descriptors: application to CO2 storage.

@inproceedings{DahrabouGM2015,
 abstract = { In order to prevent the release of large quantities of CO2 into the atmosphere, innovative strategies must be applied. Among the techniques currently available, carbon capture and storage (CCS) is proposed. It is a potential means of mitigating the contribution of fossil fuel emissions to global warming and ocean acidification. Fluvial saline aquifers are favorite targeted reservoirs for CO2 storage. Despite fluvial reservoirs are known to be very heterogeneous, these heterogeneities were rarely taken into account in CO2 reservoir models. Moreover, few data are generally available, which leads to wide uncertainties. In classical petroleum reservoir studies, several stochastic models are generated to perform risk assessment and deal with uncertainties. However, flow simulation is still only applied on some of these models as flow simulators are time consuming. To overcome this issue, the proposed strategy is to simulate flows only on some stochastic models that a priori represent very different reservoir 3D structures in order to quickly span the range of uncertainties. This is particularly crucial since huge uncertainties exist, hence in CO2 storage context. The selection of models is based on the computation of geometrical, topological and process-based descriptors, used to define distances between generated models. Then, multivariate statistics are performed for selecting a given number of representative models dedicated to flow simulation. The work presented in this paper belongs to this research field but it focuses on finding topological descriptors. Two kinds of topological studies are addressed in this paper. On one hand, the question is to determine if the spectral analysis of graphs is able to discriminate different complexity degrees of the reservoir rock network. Skeletons of the reservoir rock volume are computed and converted into graphs that are finally studied through a spectral analysis of the Laplacian connectivity matrix. Even if spectral analysis can be performed in 3D, the computation of a skeleton in 3D is a very difficult task. Moreover, skeletons in 3D are not only composed of linear components but also 2D ones. Then, only 2D reservoir rock networks are studied for this part. On the other hand, the Betti numbers are scrutenized to determine if they can be used for describing differences on reservoir geobody network or connectivity. Betti numbers are stemming from the Morse theory and have been recently recognized as powerful topological indices. These indices are particularly interesting as they can be computed in the cases of volume studies. The presented works are focused on fluvial reservoir and the aim is then to quantify the impact of fluvial heterogeneities and their spatial distribution on the connectivity of the reservoir geobodies and therefore, on CO2 storage capacities. To achieve this, representative models of different stacking scenarios of channels and internal heterogeneities are generated using different simulation methods. For the spectral analysis, only 2D synthetic simulations are used. Betti numbers are computed on each simulation and statistically analysed. Moreover, hydraulic behavior will be also studied to check if relationships exist between static (reservoir rock network) and dynamic (flow path) topology. The results of these studies will be presented and discussed. },
 author = { Dahrabou, A. AND Viseur, Sophie AND Gonzalez Lorenzo, A. AND Mari, J-L. AND Bac, Alexandra AND Rohmer, J. AND Audigane, Pascale },
 booktitle = { 35th Gocad Meeting - 2015 RING Meeting },
 publisher = { ASGA },
 title = { Characterizing Fluvial reservoir using topological descriptors: application to CO2 storage. },
 year = { 2015 }
}