Framework for Constructing Clastic Reservoir Simulation Models

Abstract

Summary. This paper surveys practical methods for reservoir modeling. To eliminate unnecessary jargon and to promote synergy, a simple classification system of sand-body architecture is proposed that helps relate geology to fluid flow. Guidelines are given for treating reservoir heterogeneities and upscaling properties to gridblock-scale averages. Introduction The main objective of this paper is to give guidelines for the preparation of simulation input for clastic reservoirs. Reservoir simulation efforts range from material-balance-type analysis to very detailed 3D simulation with as many as 50,000 gridblocks. Early simulation efforts attempted to estimate the drive mechanism and recovery efficiency for a given development scheme and to present preliminary ideas on well position, spacing, and the completion interval. In a fairly simple reservoir, about five wells may supply sufficient data for a 3D simulation, and radial models around these wells can provide useful insight into the relationship between core permeability measurements and actual regional permeability distribution. This will require geological modeling and can extend to modeling of sedimentary structure. For successful detailed 3D simulation, a reservoir geological model and an excellent understanding of the permeability distribution and continuity are necessary. Reservoir Heterogeneities The main difficulty in geological modeling is to take proper account of the various largeto small-scale heterogeneities. These can be conveniently subdivided into seven basic types ranging from large-scale faults, external and internal sand-body characteristics, to microscopic features, with fractures as the seventh class. Each type of heterogeneity influences fluid flow and hence sweep and recovery efficiencies (Table 1). In this paper, we focus on the modeling of reservoir architecture and permeability within genetic sand-body units. Faults and fractures can be very important in reservoir modeling, but these types have been discussed recently in some excellent books and papers. Thus, in this paper, the discussion is restricted to some remarks on major features and influences with references to useful literature. Sealing faults constitute no-flow boundaries in simulation models and therefore often form the grid boundaries. Thus, major faults tend to have a strong influence on grid orientation. Faults that are semisealing or nonsealing can be incorporated into modern simulators by inserting special gridblock connections with estimated transmissibilities. Transient pressure testing can be carried out to arrive at realistic tmnsmissibilities of fault zones. Small-scale fractures are difficult to incorporate into models. Tight cemented fractures can be detected on borehole-wall imaging logs or in cores. They may reduce permeability significantly and occasionally may cause severe small-scale compartmentalization of the reservoir. Well testing will reveal the degree of reservoir deterioration caused by the tight fractures. Open fractures are comparatively rare in most clastic reservoirs. In tight sandstones, they may play an important role, especially in gas reservoirs. Reservoir simulation of these so-called dual-porosity reservoirs is difficult because of the problem of quantifying the degree of capillary contacts across fractures. A recent paper presents methods to model block-to-block interaction. Welltest analysis can show the presence of an open-fracture system, but the results may be ambiguous because certain layered systems have a pressure response similar to dualporosity reservoirs. For this reason, one generally has to carry out core studies to analyze the natural fracture system and the fracture spacing. General Approach to Reservoir Modeling Any scheme for geological reservoir modeling starts with a data base from which one first derives a depositional model through a core/log facies analysis (Fig. 1). Regional information is usually a major help, not only in reaching conclusions, but also in planning the data acquisition. With the facies model and the facies pattern identified in each well, a correlation can be attempted. At this stage, one needs a statistical data base of the geometry of each sand- or shale-body type detailed seismics may be of use, especially if high-quality 3D surveys are available. The detail of the correlation that can be obtained is a function of well spacing and reservoir complexity. If the well spacing is large or if the reservoir architecture is very complex, a detailed, deterministic correlation may not be possible. In that case, we can apply statistical methods to develop likely configurations of the reservoir architecture. JPT P. 1248⁁