Abstract:
Seismic reflection is a key geophysical method for hydrocarbon prospecting, mapping geological features, primarily focusing on assessing reservoirs and their physical properties. Research was conducted on the Kadanwari Gas Field, situated in the Middle Indus Basin, characterized by an extensional regime dominated by horst and graben structures. The study involves a comprehensive approach, beginning with a 3D seismic structural interpretation aimed at mapping the designated block's geological structure and reservoir quality. Subsequently, a petrophysical interpretation is executed to pinpoint zones of interest within the target formations, specifically focusing on the G, F, and E sand intervals within the Lower Goru Formation. A model-based post-stack inversion algorithm was utilized to characterize the reservoir, and various attributes such as instantaneous phase, trace envelope, and spectral decomposition were applied to identify thin beds. Shear sonic velocity (Vs) is determined using both the Castagna equation and machine learning. Upon comparison and validation using synthetic AVA gathers it was
evident that the machine learning-driven multi-regression approach significantly improved the predictive accuracy of shear sonic velocity (Vs), yielding results that are 80% to 90% superior to an alternative method. Additionally, machine learning was
employed to perform facies modeling aimed at categorizing the challenging-to distinguish thin sand and shale layers within the Lower Goru Formation into three distinct groups: sand, shale, and shaly sand. Finally, the research focused on computing
geomechanical parameters, with a particular emphasis on automating the preconditioning of petrophysical logs using machine learning. This automation greatly facilitated the automatic detection of layer boundaries of sand and shale of Lower Goru
Formation. The comprehensive geophysical analysis of the Kadanwari area, empowered by advanced methodologies and machine learning, revealed the G and E sands of the Lower Goru Formation as promising reservoirs with significant hydrocarbon potential. Incorporating machine learning techniques, the study successfully deduced essential geomechanical parameters.vii
