Intelligent Approach for Investigating Reservoir Heterogeneity Effect on Sonic Shear Wave

Abstract

Heterogeneity refers to a not uniform distribution of reservoir properties. To overcome the problem of heterogeneity, most reservoir studies split the reservoir into different zones. In general, this disparity affects all log tools. Sonic shear wave time (SSW) is a critical metric in geomechanical modeling that is strongly influenced by reservoir heterogeneity and the kind of porous fluid composition. To detect the effect of reservoir heterogeneity on SSW prediction, an artificial neural network (ANN) was applied as an intelligent technique. One Iraqi vertical well that penetrated the Asmari reservoir was selected for this study. It contains 2462 SSW measured points as well as the following seven log parameters: Gamma Ray, Caliper, Density, Neutron, Compressional sonic, and True resistivity log over measured depth. Based on formation assessment and available well data, the Asmari reservoir was classified into six zones (with different lithology and different fluid content): A, B1, B2, B3, B4, and C. To investigate the effect of lithology on SSW, two runs of ANN had been conducted in this study. Initially, we developed a single ANN for all 2462 measured points, while in the second, six ANNs were built, one for each zone. The optimum structure for all the developed ANNs was obtained with one hidden layer of 12 neurons (7-12-1). The statistical parameters used for comparison are average percent error (APE), absolute average percent error (AAPE), standard deviation (SD), mean square error (MSE), and correlation coefficient (R2). It was observed that these parameters are approximately close to each other for the developed seven ANNs. The R2 values of the seven ANNs are 0.98 for all zones, and 0.99, 0.99, 0.99, 0.99, 0.99 and 0.96 for each zone respectively. The insignificant differences of results can be explained by the fact that the log readings (i.e. inputs variables) are already reflected the effect of lithology. Therefore, we recommended using the ANN based on 2462 for predicting SSW to any lithology zone. A mathematical model for representing the suggested ANN to simplify the calculation.