Lithofacies classification using Bayes theorem method

Case study Western Desert, Egypt

Abstract

This paper shows the availability of using the Bayesian classification method to predict class membership probabilities in one of the deep tight reservoirs in Western Desert, Egypt. The workflow of our project that using the Bayesian method used the deterministic petrophysical results of three training wells to train the data and extract the classifiers. The classified data were modeled using Gaussian distribution for each lithofacies. The used wells were acquired from a deep Jurassic gas reservoir in the Western Desert of Egypt. The fitting between actual and modeled data has been reached by minimizing the L2 norm. Besides, a cross-validation process was used for validating the resulted classifiers. Finally, the Bayesian classification method can predict the GWC with an accuracy of 4 m. To avoid probability interference caused by the compacted shale more data should be added to the initial model.