Reservoir Quality Evaluation of Miocene Sediments Using Seismic Attributes and Neural Networks in Block 103 & 107, Song Hong Basin

Nguyen Thi Minh Hong

Abstract: This study evaluates the reservoir quality of Miocene sediments in Blocks 103 and 107 of the Song Hong Basin using an integrated approach that combines seismic attribute analysis, machine learning, and geological interpretation. Core data, wireline logs, and seismic inversion were utilized alongside committee-based neural network models to predict porosity and permeability. The results indicate that porosity values in the study area range from a few percent to 20%, with notable variation between stratigraphic zones. The committee modeling approach, using both average and weighted combinations, provided improved predictions over individual models. These findings contribute to a more reliable reservoir characterization and suggest promising zones for future gas exploration and drilling operations.

Keywords: porosity prediction, permeability modeling, Miocene reservoirs, seismic attributes, neural networks