A New Approach to Shale Management in Field-Scale Models

Abstract

Virtually every sandstone reservoir contains significant amounts of shale. Although the sand/shale sequence in the wells may be known, a reliable reservoir description must define quantitatively the lateral continuity of the shale, the shale distribution in unsampled areas, and the effects of the shales on transport properties. This paper presents statistical techniques by which the lateral continuity and spatial disposition of shales can be estimated. These techniques rely on shale statistics from the wells and outcrop statistics from appropriate depositional environments. The resulting shale description can be conditioned to conform with the observations of the wells. The subsequent discretizing of the reservoir into gridblocks for numerical simulation yields large elemental units that usually contain both sand and shale. The paper concludes by showing how the permeability and paper concludes by showing how the permeability and porosity may be estimated from the disposition of shales porosity may be estimated from the disposition of shales within each block. Introduction Shales either divide reservoirs into separate hydraulic units, or, if the shales are discontinuous, set a tortuous environment in which fluid transport occurs. While continuous correlatable shales are handled with ease in reservoir simulation models, discontinuous shales present a problem. This study is concerned with such present a problem. This study is concerned with such uncorrelatable discontinuous shales in sandstone reservoirs. The primary source of information about shales is, as with virtually all reservoir properties, the wells. An empirical distribution of shale thicknesses can he inferred from an observed sand/shale sequence. Unless the well spacing is extremely dense, nothing is revealed about the shale's lateral dimensions. A secondary source of information is visual studies on shale dimensions in outcrops of ancient sediments. Once the depositional environment is identified by the geologist, outcrop statistics from a depositional environment similar to the one to be simulated are used for the assignment of lateral shale dimensions. Since we are concerned here with uncorrelatable or stochastic shales, the shales are assumed to be randomly distributed in space. This means that the shale dimensions and the coordinates of the shale centers are random, independent statistical events. This paper describes a shale management procedure that handles sampling of the empirical statistics, generates shale coordinates, and maintains control such that conditioning at the wells is achieved. Conditioning means that the fraction of shale and the vertical succession of sand and shale observed in a well are identical to that in a synthetic description. Finally, when a user-defined grid system is superimposed on the resulting description, the procedure calculates effective porosities and permeabilities for each gridblock. Definition Four scales of averaging volume can be recognized for porous media averages: microscopic (the scale of only a porous media averages: microscopic (the scale of only a few pores), macroscopic (the size of conventional core plugs), megascopic (the size of large gridblocks in field plugs), megascopic (the size of large gridblocks in field models), and gigascopic (total formation or regional scale). These concepts are illustrated in Fig. 1a. As discussed by Bear, porous-media transport equations usually are based on the continuum approach. This means that the actual porous medium is replaced by a fictitious continuum. Physical properties and dependent variables are averages over elemental physical volumes constituting the continuum. The suggested distinction among four scales is necessary because measurements on one scale are not necessarily applicable on another scale (Fig. 1b). In particular, measurements on cores, by which intrinsic sand properties are found, cannot be used readily for megascopic blocks, which include gross textural differences such as sand and shales. This discrepancy is particularly noticeable in numerical reservoir simulation particularly noticeable in numerical reservoir simulation where fluid flow equations are formulated on a macroscale but are frequently solved on a megascale. Shales are generally recognized by geologists as fine-grained, indurated sedimentary rocks with finely laminated structure. They are believed to be a lithification product of muddy sediment, of any origin, and the finely laminated texture is related to orientation of micaceous clay mineral constituents. For our purposes, the term "shale" includes shale laminae, shale streaks, and massive shales. Two shale types are distinguished. The definitions are based on whether the dimensions and spatial disposition of the shales are known. Stochastic shales cannot be correlated between wells and appear to be scattered randomly within the sand matrix (Fig. 2a). Deterministic shales are continuous between observation points. There is no uncertainty associated with their existence and lateral continuity (Fig. 2b). Most reservoirs are hybrid with respect to shales in that both deterministic and stochastic shales coexist; however, the occurrence of stochastic shales clearly becomes more prevalent with large well spacings. SPEJ P. 447

The application of outcrop analogues in geological modelling: a review, present status and future outlook

“…Early stochastic modelling of shales (Haldorsen & Lake 1984;Begg & King 1985;Halderson & Chang 1986) and the modelling of fluvial channels as geobodies (Fielding & Crane 1987;Haldorsen & Damsleth 1990;Holden et al 1999) led to a requirement for data to condition models. The SAFARI project from the mid-1980s to the early 1990s Keogh et al 2007) was undertaken by a consortium of Norwegian Oil Companies, universities and the Norwegian Petroleum Directorate.…”

Three-dimensional modeling of clinoforms in shallow-marine reservoirs: Part 2. Impact on fluid flow and hydrocarbon recovery in fluvial-dominated deltaic reservoirs

“…Where mudstone beds are laterally extensive, effective vertical permeability approaches zero. The spatial distribution of erosion at the base of sandstone beds is poorly understood but, where present, increases connectivity between sandstone beds that are otherwise isolated, thus increasing the effective vertical permeability (Haldorsen and Lake, 1984;Begg and Chang, 1985;Begg and King, 1985). We consider both scenarios (2200 bbl/day) over 10 yr FC sandbodies may be present or absent FC sandbodies are always present *Four factors (factors 1-4) investigate uncertain parameters in reservoir characterization, which correspond to three lengthscales of heterogeneity in the hierarchy described for fluvial-dominated deltaic reservoirs ( Figure 1B-D).…”

Bayesian data assimilation for stochastic multiscale models of transport in porous media.

“…Binary models of conductivity are widely applied in subsurface flow through porous media particularly for representation of permeability patterns in fluvial deposits (e.g., [15,21,22,23,24]). Additionally, fractured media are often characterized using a binary permeability model where the fractures represent strongly anisotropic, high conductivity inclusions embedded within a matrix of low conductance.…”