CT Scan and Neural Network Technology for Construction of Detailed Distribution of Residual Oil Saturation During Waterflooding

We present an integrated approach to imaging the progress of air displacement by spontaneous imbibition of oil into sandstone. We combine Computerized Tomography (CT) scanning and neural network image processing. The main aspects of our approach are I) visualization of the distribution of oil and air saturation by CT, 11) interpretation of CT scans using neural networks, and III) reconstruction of 3-D images of oil saturation from the CT scans with a neural network model. The neural networks developed here construct 3-D images of fluid distribution at any time and/or location within the core. One neural network model interpolates between the CT images for a given position at different time levels and extrapolates beyond the interval of time during which the images were collected. Likewise, the network interpolates spatially between images at a given time, After interpolation and extrapolation, other network models have been developed to reconstruct the three-dimensional distribution of oil in the core. Excellent agreement between the actual images and the neural network predictions is found. Introduction An increasing global demand for energy and simultaneous depletion of conventional hydrocarbon reserves impose a formidable challenge for efficient recovery from nonconventional rock systems, such as naturally hctured reservoirs. Fractured petroleum reservoirs provide over 20 ?ZO of the world oil reserves [ 1]. Examples of prolific fmctured reservoirs are: the Monterey Shales in California (estimated tens of billions of barrels of oil-in-place); the California Diatomites (estimated fifteen billion barrels of oil-in-place); the West Texas Carbonates; the North Sea Chalks; and the Asmari Limestones in Iran. Hydrocahon recovery from naturally fractured reservoirs is not yet fully understood. This is mainly due to the lack of a complete understanding of multiphase flow through fractured porous media. Twoor three-phase flow in a fractured reservoir depends on the combined nonlinear effects of hydraulic connectivity and physiochemical properties of fractures, relative permeabilities to multiphase flow in the fractures, rock-matrix nature, matrix block size, capillary forces and fracture closure stress. The nonlinear interplay of all these factors determines the ultimate hydrocarbon recovery from fractured reservoirs. In contrast, most of the published data have been produced in controlled experiments that have focused on one or more of the above factors considered in isolation. These data are then upscaled in numerical simulators to model the coupled nonlinear behavior of fractured reservoirs. As a result, current numerical simulation models of fractured reservoirs lack firm predictive capability and must be tuned for each field case with the available data. Thus, itmight be helpful to undertake a systematic experimental and theoretical study of joint effects of all the factors governing multiphase fluid flow in a fractured porous rock. Of course, such a study is beyond the scope of this paper. Nevertheless, we have undertaken a study to evaluate the influence of four major factors on hydrocarbon recovery. These are: fracture configuration, rock-matrix block size, nettability characteristics of the rock, and fluid flow rates. This paper reports our progress on a scoping study of spontaneous imbibition of a hydrocarbon (kerosene) into a single air-filled block of rock matrix (Berea sandstone). Our experiments are a preamble to a more difficult study of the most important production mechanism in fractured reservoirs during 2 CT SCAN AND NEURAL NETWORK TECHNOLOGY FOR CONSTRUCTION SPE 35737 TURATIO N waterflooding, i.e., counter-current imbibition of water to displace oil and gas from the matrix. Here, we want to understand the pattern of imbibition from the distribution of fluid saturations and to design a neural network model of insitu fluid saturations obtained directly from a CT scanner, The model is then used to generate three-dimensional time-lapse images of kerosene imbibition. Finally, we intend to incorporate the experimental results into our integrated-finite difference simulator, M2NOTS (Multiphase Multicomponent Non Isothermal Organics Transport Simulator [2]), to allow for a more realistic simulation of multiphase flow through fractures. The mathematical formulation of M2NOTS does not rely on a global coordinate system; therefore, it naturally extends the method of Multiple Interacting Continua [3] for modeling flow in fractured media to multiphase and multicomponent systems. In this project, we use high resolution X-ray computerized tomography to obtain images of the cross-sectional distribution of kerosene and air in Berea sandstone cores as a function of time. Scans perpendicular to the axis of the cote were made using a high resolution EMI 5005 (second generation) CT scanner. Each CT slice consists of a series of volume elements (voxels). Every voxel has its own characteristic attenuation, and can be mapped into a 2-D image matrix of picture elements (pixels). Using standard computer software, the 2-D fluid distributions at specific times ad locations m visualized for each CT slice. CT is a fast, nondestructive imaging technique for determining in-situ fluid saturation with excellent 3-D resolution. Using this technique, attenuation differences as small as 0.17’0with a cross sectional resolution of less than 1 mm~ can be realized. For extrapolating and interpolating between different slices obtained, neural network models were developed. Neural networks are very useful in modeling nonlinear, complex, and multi-dimensional data and find wide application in analyzing experimental, industrial, and field data. Neural networks, unlike regression analysis, do not require specification of a structural relationship between the input ad output data and they can be trained easily by using sufficient data from the system under study. In addition, neural networks have the ability to infer general rules and extract typical patterns from specific examples. These properties give the neural networks the ability to interpolate between typical patterns or data and generalize their learning in order to extrapolate to a region beyond their training domains. Principles of CT Imaging Various visualization methods have been used for fluid saturation determination during laboratory core fled experiments [4]. Some of the more common ones in use are transparent models [5], resistivity [6], microwave attenuation [7], NMR, MRI, X-ray, and gamma ray attenuation [8]. While most of these methods provide only average saturation and impose restrictions on experimental techniques, CT is a very fast and accurate technique with few restrictions on experimental conditions and offers fine spatial resolution [9]. Earlier investigators [1016] have illustrated the importance of computerized X-ray tomography as a powerful tool for petroleum industry researchers, To obtain a CT slice of an object, an X-ray source is collimated to provide a thin beam which is received by an array of crystal detectors, X-ray photons which strike these crystals cause them to fluoresce with an intensity proportional to the number of photons received. When a body is placed in the beam between the source and detector array, only those photons that are not absorbed by the body reach the detectors. Fig. la illustrates the principles of X-ray tomography. The values attained when the detectors are read represent the beam attenuation by an object placed in the path of the X-rays. The detectors are in a stationary array surrounding the object. The X-ray bmms are always directed through the object aperture as the source moves around it in a circular path. The detectors are rmd at small rotational intervals and the resulting data are stored in a computer. This rotational excursion is called a pass and the total data acquired during this pass are termed a slice. After all readings for a slice have been acquirwl and stored in a computer, a cross-sectional image or matrix of attenuation coefficients P(X, y) is created. Radon [17] established the mathematical foundation for image reconstruction from projection data, The basic synthetic unit is the volume element or voxel. The CT slice is composed of many voxels, each with its own characteristic attenuation, which are displayed as a 2-D image matrix of picture elements (pixels), shown in Fig. la. CT measures linear attenuation coefficients /.L, which are defined by Beer’s law: -(p / p)px I/Z. = e ......................(1) where 1{, is the source X-ray intensity, I is the intensity measured by the detectors, w is the linear attenuation coefficient, p is the density of the medium, p/p is the linear mass attenuation coefficient, and x is the thickness of the material, If several materials are placed in the path of the X-ray beams, Beer’s law can be generalized as: Illo=e ‘XPi f Pi )Pxi ............... (2) where i is the material considered. If the object contains a mixture of components, the overall mass attenuation coefficient of the mixture is given by: Pm, =~o 1 . . . . . . . . . . . . . . . . . . . . . . . . (3) where S, is the saturation of the phase i, i.e., Si is the volume