Published 1994 as part of Computer Applications 3
Copyright © 1994 The American Association of Petroleum Geologists.   All Rights Reserved.
 

This work investigates the utility of applying fractal geometry for describing the spatial correlation structure of rock properties. It also provides a proper framework within which to apply a stochastic conditional simulation technique to generate equiprobable distributions of rock property.

A normalization technique is developed to remove the seasonal effects found in the well-bore data. Where the seasonal effects are unusually strong, as in the case of one carbonate reservoir, removal of the local means will result in a narrower distribution of the Hurst (H) exponent values. Theoreti-cal and analytical variogram plots of the results of these estimations indicate that the box-counting method gives slightly improved estimates over rescaled range (R/S) analysis.

Conditional simulation runs using simulated annealing are performed for three fields. The results indicate that a combination of the greedy algorithm and a bilinear interpolation technique to generate the initial pattern decreases computer usage by as much as 85%. More important, the simulations also indicate that an fGn (fractional Gaussian noise) horizontal correlation structure gives better predictions of interwell distributions in a carbonate environment. In contrast, in a sandstone field, the use of an fBm (fractional Brownian motion) horizontal correlation structure gives more reliable results.