Propagating Interval Uncertainties In Supervised Pattern Recognition For Reservoir Characterization

摘要

In the petroleum exploration/production context,characterizing the reservoir quality, identifying the main rocktypes, or predicting their spatial variations is a challenge forthe industry. To achieve this purpose, supervised patternrecognition methods, as discriminant analysis are widely used.These methods aim at calibrating, when possible, arelationship between field features -for example a set ofborehole measurements, or a set of seismic attributes- and apredefined set of classes -for example, different rock types-. Ithas yet the major drawback not to take into account theuncertainties on the measurement arrays, which may causedrastic misinterpretations of reservoir characteristics. Themethodology developed is an extension of the standardparametric approach to discriminant analysis. The calibrationphase follows the same main steps as the standard algorithm,except that all the processed quantities are intervals. Thedifferent interval computations are based on intervalarithmetic. Eventually, any imprecise object is assigned to asubset of classes, consistent with the measurements and theiruncertainties. As a result, the computed reservoir qualitymodel is less precise, but more realistic, taking into accountdata and its uncertainties.