Prediction of Fluid Compositional Heterogeneities in Accumulations Using Temis Suite^® Enhanced with Local Grid Refinement (LGR): Example from the Jurassic of Northern Kuwait

F. Monnier¹, F. Lorant², P.-Y. Chenet¹, J.-M. Laigle¹, and A. Al-Khamiss³
¹Beicip-Franlab, 232, Avenue Napoléon Bonaparte, P.O. BOX 213, 92502 Rueil-Malmaison, France
²IFP, 1-4 Avenue de Bois Préau, 92852 Rueil-Malmaison Cedex, France
³Kuwait Oil Company (KOC), AhmadI, Kuwait

Classical basin simulation is designed to account for large scale processes, both in time and space, associated to petroleum systems. Using this technology to also predict fine fluid composition heterogeneities in reservoirs meshes has been a growing ambition for the past ten years, though it has never really been addressed operationally. There are few prerequisites that need to be dealt with before undertaking this type of operation. First, it is necessary to handle a geochemical model accounting for oil and gas compositions and properties through time, based on a compositional description of hydrocarbon fluids. Secondly, as a result of the previous issue, it is necessary to hold a calculator able to simulate realistic compositional fluid flows in space and time, taking all driving forces into account, i.e., hydrodynamics, buoyancy and capillarity. Thirdly, in order to focus the simulation on reservoir scale heterogeneities, the calculator must also have the ability to solve fluid flow equations continuously from the large size grid of a regional model to the fine meshes of the hydrocarbons accumulations

Improvements in compositional kinetic modeling of oil and gas generation and cracking have been reported in a recent series of papers (Behar et al. 2008a and b). Compared to previous IFP scheme, this new compositional model contains two main evolutions. The first one concerns the description of kerogen conversion process. Classically it is assumed that hydrocarbons are directly generated from kerogen cracking through a set of parallel, independent reactions. Behar et al. showed that this kind of mechanism is not suitable for marine and coaly source rocks at least, in which sedimentary organic matter rather converts firstly into heavy bitumen, water, CO₂, and mature kerogen. Most of the oil potential actually results from secondary cracking of this heavy bitumen within the source rock. The gas potential is mainly associated to late cracking of mature kerogen (Lorant and Behar, 2001). Therefore, oil and gas generation does not result from direct decomposition processes, but from consecutive stages during organic matter transformation. The second feature of the new kinetic model is its capacity to predict changes in reactivity sequences and fluid properties as a function of the temperature domain. Of course the model account for the relative stoichiometry in bulk gas derived from oil cracking, which is inversely related to heating rates (i.e. increased stoichiometry with decreasing heating rate). Behar et al. (2008b) showed that this phenomenon results from the reversal of reactivity between aromatics and saturates in oil from low to high temperature conditions. Hence, coupled to equation of state, the new kinetic scheme significantly enhances the capacity of basin simulation to predict real fluid compositions resulting from primary and secondary cracking processes, and consequently, GOR and API gravity.

In addition to enhancing petroleum fluid parameterisation, significant efforts were made during the past ten years to develop a new Temis simulator capable of computing efficiently and realistically pressure, temperature and fluid flow in both 2D and 3D using a finite volume spatial discretization scheme (Schneider et al. 2000; Schneider and Wolf, 2000). Recently new numerical techniques, which considerably improve the convergence of the Newton algorithm both in terms of robustness and CPU time, have been implemented (Agelas et al., 2007). These improvements speed up considerably compositional simulations using generalized Darcy 2-phase flow (water + hydrocarbons), thus enabling efficient fluid composition and pressure predictions in the millions of cells of 3D models.

This simulator uses fixed structured meshes over the whole block, thus precluding applications of the basin model to account for local high resolution fluid properties. For that latter purpose, refining a grid locally is of course possible by generating a tartan in x,y directions and a z-layering over the entire domain of the block. However, the resulting 3D Scottish grid would still contain a lot of cells in areas where refinement is neither necessary nor wanted, leading to a waste of computation time.

To overcome this situation, and speed up computation time while keeping a good physical model (Darcy 2-phase flow) and local high resolution, IFP developed a new calculator that handles true 3D local grid refinements (LGR). LGR technology allows ascribing fine lateral and vertical meshes in selected areas within the block, where information is available and where local property distributions need to be predicted, i.e. in accumulation areas. Assuming an initial high resolution geological model (Figure 1), one defines areas of interest in the 3D block, i.e. local portions of the basin where hydrocarbons are likely accumulated and/or where data are available. One possible workflow would consist of keeping the original grid resolution in local areas while upscaling the 3D grid anywhere else in the block (Figure 1). So starting from one geo-model, a basin - LGR study handles one coarse grid (i.e. the regional block), plus several imbricated Cartesian grids refined in all three spatial directions with several levels of refinements. For instance, there may be as many refined grids as there are potential hydrocarbon accumulations in the basin. Then the calculation is launched, as for any conventional basin simulation.

In this paper, we report a case study of Northern Kuwait, where LGR technique coupled to compositional 3D Darcy flow modeling was applied to predict the distribution of hydrocarbon composition and properties in local reservoir areas.

The original model covers a surface of 70 x 130 km² with a mesh resolution of 500 m x 500 m, and contained 31 layers stacked over 8.5 km thickness. The petroleum system addressed in this work includes a series of thin Late Jurassic reservoir and marine source rock layers capped by evaporites and a subjacent Early Jurassic reservoir fed downward from the Late Jurassic source rocks. Downward hydrocarbon migration occurred along a system of active faults during Cretaceous time, leading to hydrocarbons accumulations within the Early Jurassic carbonate reservoir sealed at the top by a shaly limestone layer.
In order to select areas for LGR within the carbonate reservoir, a preliminary study was conducted using a ray-tracing algorithm. Fine local grids were generated on the potential traps so determined (Figure 2). Finally, full 3D compositional migration was simulated over the entire coarse block plus refined blocks together.

Fine compositional variations within refined areas are accounted for by the model, including API and GOR distributions that would never be assessed using standard kinetic modeling in a classical basin grid. Local fluid property variations result from both migration/accumulation of generated hydrocarbons and secondary cracking along the migration path and within the carbonate reservoir.

This work demonstrates the efficiency and precision of coupling LGR to advanced kinetic modeling, and offers interesting bridging possibilities between basin and reservoir modeling.

References

Agelas L. Faille I., Wolf S., Requena S. (2007) High performance computing for Darcy compositional single phase fluid flow simulations. AAPG Hedberg Conference Basin Modeling Perspectives: Innovative Developments and Novel Applications. May 6-9, 2007, The Hague, The Netherlands.

Behar F., Lorant F., Lewan M.D. (2008) Role of NSO compounds during primary cracking of a Type II kerogen and a Type III lignite. Organic Geochemistry, 39, 1-22.

Behar F., Lorant F., Mazeas L. (2008) Elaboration of a new compositional kinetic schema for oil cracking. Organic Geochemistry, 39, 764-782.

Schneider F., Wolf S., Faille I., Pot D. (2000) A 3D basin model for hydrocarbon potential evaluation: application to Congo offshore. Oil & Gas Science and Technology – Rev. IFP, 55, 1, 3-13.

Schneider F., Wolf S. (2000) Quantitative HC potential evaluation using 3D basin modelling: application to Franklin structure, Central Graben, North Sea, UK. Marine and Petroleum Geology, 17, 841-856.

Figure 1. Workflow for the generation of local refined grids (LGR) in a 3D basin model.