A probabilistic approach to improved geological knowledge and reduced exploration risks using hydrocarbon migration modelling

Abstract

This paper elaborates a methodology for assessing the consequences of uncertainties in the source and reservoir models on hydrocarbon phase distributions in prospects by improving the a-priori probability distributions. The uncertainties of geological systems can be described by probabilistic distributions of quantified geological variables, such as source rock and reservoir thicknesses. A Monte Carlo simulation of hydrocarbon generation, expulsion, migration and entrapment is carried out with probabilistic distributions defining important input variables. A weight for each simulation run is computed from a misfit function that describes how well the modelled oil and gas columns fit the observed columns in a set of calibration wells. These weights are used to compile a-posteriori probability distributions of the input variables. The resulting probability distributions, therefore, describe the uncertainties of the geological system after the calibration of the migration model. In a demonstration example from the Tampen Spur area of the North Sea, the most sensitive geological input distributions become significantly narrower, thus limiting the range of possible outcomes. Other distributions become more uneven, with peaks and lows. The changes in the variable distributions have been quantified and variables can, therefore, be ranked with respect to their impact on changes in probabilities for key outcomes. A large number of simulation runs is required for the proposed method to be effective because many simulation runs result in too large a misfit. A statistical fit of a second-order function to the misfit values allows for the effective screening of input variables, thus reducing the computing time, in the demonstration example, to 15% of the unscreened times. The result of the analysis is that the effect of geological uncertainties on output values is effectively reduced and that a quantitative learning has been obtained. This knowledge can be applied in other hydrocarbon generation and migration studies within the study area, allowing for more rapid and more accurate estimates of oil and gas resources before drilling.