International audienceDetermining productive zones has always been a challenge for petrophysicists. On the other hand, Artificial Neural Networks are powerful tools in solving identification problems. In this paper, pay zone determination is defined as an identification problem, and is tried to solve it by trained Neural Networks. Proposed methodology is applied on two datasets: one belongs to carbonate reservoir of Mishrif, the other belongs to sandy Burgan reservoir. The results showed high precision in classifying productive zones in predefined classes with Classification Correctness Rate of more than 85% in both geological conditions. Applicability of proposed pay zone determination procedure in carbonate environment is a great advantage of developed methodology. Fuzzified output, being independent of core tests and verification with well tests results are of other advantages of Neural Network-based method of pay zone detection

International audienceNet pay detection is a key stage in reservoir characterization for several purposes: reserve estimation, reservoir modeling and simulation, production planning, etc. Determining productive zones always is simultaneous with some amount of uncertainty due to lack of enough data, insufficiency of knowledge and wild-nature of petroleum reservoirs. It becomes even more challenging in carbonates, because of their highly heterogeneous environment. Conventionally, evaluating net pays is done by applying petrophysical cutoffs on well-logs, which results in crisp classification of pay or non-pay zones. In addition, cutoff based method is developed in sandstones, and does not provide suitable results in carbonates at all. Proposed methodology of this work, Dempster-Shafer Theory, is a generalization of Bayesian Theory of conditional probabilities. Net pays are studied in two oil reservoirs by this theory; one of them is carbonate reservoir of Mishrif, the other is sandy Burgan reservoir. For validation, results 2 are compared to well tests and output of conventional cutoff method. The advantages of using Dempster-Shafer Theory, comparing to conventional cutoff based method in studying net pays is: to have a continuous fuzzy output, based on geological facts, with high generalization ability and more compatibility with well test data