A Quick Look Kriging Technique for Reservoir Characterisation

Abstract

Abstract Correct descriptions of reservoir rock heterogeneity provide useful input for simulation models and are essential for efficient reservoir management. This paper discusses a novel kriging technique that uses sparse irregularly spaced permeability data to generate permeability values at other points within chosen grid blocks. A uniquely designed computer program, expressed in FORTRAN provides solution to the sequence of algorithms required for the solution of the kriging system in the absence of more sophisticated kriging models. Results obtained from the model allows for relatively cheap and reasonably accurate characterization of heterogeneous reservoirs. Introduction The processes of deposition, tectonism and environmental sequences occurring over millions of years are responsible for the heterogeneous nature of reservoir rocks. Optimum success in hydrocarbon recovery depends largely on adequate knowledge of the reservoir rock parameters. The task of quantifying the porous media heterogeneity, therefore, is the major challenge for the reservoir engineer. A common obstacle encountered by the reservoir engineer, especially at an early stage of field development, is almost always the lack of sufficient control data (from cores, logs, seismic, etc.) that is necessary to correctly predict the internal architecture of the reservoir and estimate its petrophysical properties. The use of oversimplified geologic models based on data from a limited number of widely spaced wells is probably one of the most important reasons for the failures in predicting field performance. Stochastic or geostatistical models are capable of expressing the nature of a property's variation and its disposition in a useful mathematical form and also of providing detailed distributions of physical properties. These distributions can honour the available data, some statistical functions (like the histogram) and the variogram of the reservoir. Probabilistic (geostatistical or stochastic) modeling is used to take advantage of this capacity and still quantify the uncertainty due to the lack of information. Stochastic techniques, therefore, seek to generate synthetic variability in a certain phenomenon or parameter which ‘fills in the gaps’ between observations and descriptive needs in terms of detail. Some geostatistical or stochastic techniques of estimation include Kriging, Co-Kriging, Conditional simulation, Fractal techniques and Transition probability (Markov) procedures. Any of these can be adopted as a model for characterising the spatial structure of property variations.