Abstract:

This study addresses the challenge of predicting the distribution of early Miocene reservoir rocks in the Northeast Bach Ho field, located in the Cuu Long basin, which has been a significant site for oil and gas exploration. The study aims to apply advanced seismic attribute analysis combined with the Principal Component Analysis (PCA) method to enhance the accuracy of reservoir distribution predictions in an area with limited wells data. The seismic attributes analyzed include Seismicity, RMS amplitude, Instantaneous frequency, Coverage, RAI, Instantaneous phase, Sweetness, and t* attenuation, which were used to determine seismic facies classification. Four principal components, PC0, PC1, PC2, and PC3, were selected after analyzing the seismic attributes, accounting for 85.44% of the total variance. These components were then used as inputs for training an Unsupervised Neural Network (UNN), a method particularly useful when well data are sparse or unavailable. The output of the trained network, combined with facies analysis and porosity data from well logs, was employed to predict reservoir distribution in the study area. The results showed that the sandstone deposits in the early Miocene sediments serve as potential reservoir rocks. These potential reservoirs are primarily located around the central part of the study area, with additional deposits in the eastern, northeastern, and western regions. The reservoirs are associated with river and deltaic sedimentary environments. The combination of PCA and UNN effectively reduced noise in the seismic data, leading to clearer identification of seismic facies classification and improved reservoir prediction. This integrated method is proven to be highly effective in improving oil and gas exploration strategies, particularly in regions with limited or no well data, and can support further assessment of the oil and gas potential of the Northeast Bach Ho field and the broader Cuu Long basin.