This study is an application of data fusion techniques, especially fuzzy theory, in determining oil producing zones through four nearby wells, located on an oil field in south west of Iran. Two fusing techniques, used here are based on Bayesian and fuzzy theories. At first, two Bayesian classifiers are being constructed by training in two different wells; then a fuzzy operator, called Sugeno discrete integral, is used to fuse outputs of two mentioned Bayesian classifiers. Finally, it is concluded that using fuzzy classifier fusion improves not only certainty and confidence of decision making, but also generalization ability of determining productive zones.
Abstract:Exploration specialists conventionally utilize a cut-off-based method to find productive zones inside the oil wells. Using conventional method, pay zones are separated crisply from non-pay zones by applying cut-off values on some petrophysical features.In this paper, a Bayesian technique is developed to find productive zones (net pays), and Bayesian Network is used to select the most appropriate input features for this newly developed method. So, two Bayesian methods were developed: the first one with conventional pay determination inputs (shale percent, porosity and water saturation), the other with two inputs, selected by Bayesian Network (porosity and water saturation). Two developed Bayesian methods are applied on well log dataset of two wells: one well is dedicated for training and testing Bayesian methods, the other for checking generalization ability of the proposed methods. Outputs of two presented methods were compared with the results of conventional cut-offbased method and production test results (i.e. a direct procedure to check validation of proposed methods).Results show that the most prominent advantage of developed Bayesian method is determination of net pays fuzzily with no need to identify cut-offs, in addition to higher precision of classification: nearly 30% improvement in precision of determining net pays of first well (training well), and about 50% improvement in precision of determining productive zones through the generalizing well.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.