This paper focuses on the low-permeability sandstones of the upper Es4 member in the Boxing Sag,where pore structures in core samples were categorized into four types(Ⅰ,Ⅱ,Ⅲ and Ⅳ)based on poroperm analysis,mercury injec-tion capillary pressure,nuclear magnetic resonance,and petrophysical property analysis.Using this classification,a logging identification model for pore structures was established by calibrating logging samples with core samples and applying the sup-port vector machine(SVM)algorithm with an F-score feature selection.The model's performance was validated using 215 test samples.Based on the logging identification of pore structures from single wells,a three-dimensional geological model of pore structure types for key oil-bearing layers in the G oilfield was developed using sequential Gaussian simulation,enabling the prediction of favorable porosity and permeability zones.The results indicate that there are only 22 misjudgments in 215 test samples,all of which were among adjacent samples,achieving a prediction accuracy of 89.77%.Compared with the ear-ly statistical methods and neural network algorithms,the pore structure prediction model established with SVM-assisted F-score algorithm exhibits superior predictive performance.
关键词
渤海湾盆地/低渗透砂岩/孔隙结构/支持向量机
Key words
Bohai Bay basin/low-permeability sandstones/pore structure/support vector machine
维普
万方数据