Reservoir porosity and permeability prediction from petrographic data using artificial neural network: A case study from Saudi Arabia. 9th Middle East Geosciences Conference, GEO 2010.

Abstract

Understanding reservoir heterogeneity is essential for the assessment and the prediction of reservoir properties and quality. This study investigates the prediction of the reservoir petrophysical properties of the Ordovician Upper Dibsiyah Member of the Wajid Sandstone in southwest Saudi Arabia. The artificial neural networks (ANNS) technique was used to study pattern recognition and correlation among the petrographic thin section data such as grain size, sorting, matrix and cementation percentages, and petrophysical properties of the reservoir such as porosity, permeability and lithofacies. For this purpose, artificial intelligence techniques were designed and developed and these are the multilayer perception (MLP) and the general regression neural network (GRNN). The good agreement between core data and predicted values by neural networks demonstrated a successful implementation and validation of the network’s ability to map a complex non-linear relationship between petrographic data, permeability and porosity. The GRNN technique provides better prediction of the reservoir properties than that obtained from the use of the MLP technique.