STOCHASTIC SIMULATION OF CATEGORICAL VARIABLES USING A CLASSIFICATIONALGORITHM AND SIMULATED ANNEALING

An important step of reservoir characterization is the stochastic mode ling of the geometry of lithofacies which control large-scale heteroge neities of petrophysical properties. Although multiple realizations ar e necessary to appreciate the uncertainty in the spatial distribution of facies, a common short cut consists of retaining the first realizat ion drawn. ntis paper presents an alternative to this potentially haza rdous selection: (1) a categorical map is generated by allocating a si ngle facies to each grid node according to the local probabilities of occurrence of the facies, and (2) the map then is post-processed using a steepest descent-type algorithm so as to improve reproduction of sp atial continuity and transition probabilities between facies. The proc edure is illustrated using a synthetic dataset. A waterflood simulatio n shows that retaining a single realization would yield, in average, l arger errors in production forecasts (water cuts and recovered oil) th an the single post-processed facies map.