Intelligent approaches for prediction of compressional, shear and Stoneley wave velocities from conventional well log data: A case study from the Sarvak carbonate reservoir in the Abadan Plain (Southwestern Iran)

Abstract

Compressional, shear and Stoneley wave velocities ( V p, V s and V st, respectively) are important reservoir characteristics that have many applications in petrophysical, geophysical and geomechanical studies. In this study V p, V s and V st were predicted from well log data using genetic algorithms, fuzzy logic and neuro-fuzzy techniques in an Iranian carbonate reservoir (Sarvak Formation). A total of 3030 data points from the Sarvak carbonate reservoir which have V p, V s, V st and conventional well log data were used. These data were divided into two groups; one group included 2047 data points used for constructing intelligent models, and the other included 983 data points used for models testing. The measured mean squared errors (MSEs) of predicted V p in the test data, using genetic algorithms, fuzzy logic and neuro-fuzzy techniques, were 0.0296, 0.0148 and 0.029, respectively, for V s these errors were 0.0153, 0.0084 and 0.0184, respectively, and for V st they were 0.00035, 0.00020 and 0.00062, respectively. Despite different concepts in these intelligent techniques, the results (especially those from fuzzy logic) seem to be reliable.