A comparison between deterministic and stochastic fault seal techniques

Abstract

Abstract Traditionally, the analysis of fault seal has been purely deterministic or a combination of deterministic and stochastic methods. In a deterministic model, prediction of the locations of reservoir overlaps is made from the static model of the reservoir horizon and fault geometry. The principal aim is to map faulted reservoir overlaps and determine their sealing character. This is usually performed using a predictive algorithm such as the shale gouge ratio (SGR) that relates the shale content of the formations that have moved past a point on the fault zone to the sealing capacity of the fault rock. Deterministic fault seal studies are sensitive to the uncertainties associated with mapping of horizons in proximity to faults and the inherent uncertainty in a static fault interpretation in both position and fault zone complexity. Uncertainty in the static structure model can be addressed by convolving uncertainty in throw magnitude with juxtapositions at the fault. However, this does not address the uncertainty in the distribution of reservoirs on either side of the fault. With stochastic models multiple realizations of the stratigraphy can be tested. Stochastic models capture the uncertainty in the position of the reservoir at the fault by allowing multiple realizations of stacking geometries, where the principal assumption is that these stacked reservoir zones are laterally continuous covering the entire likely fill area. Despite the conceptual differences between these two approaches to fault seal analysis, comparison of the predictions they make on the Ling Gu field shows a surprising degree of conformity. The cut-off used to determine the number of sand and shale beds in the stochastic workflow appears to account for seal by fault zone materials, since a conservative cut-off implies fewer sand beds with lower probability of leak and correlates with more shale in the section and higher SGR values.