Quantitative Evaluation of Sand Body Connectivity Based on the Support Vector Machine Algorithm: A Case Study of the Putaohua Oil Reservoir in the Daqing Oil Field, Songliao Basin, China

Abstract

Taking the braided river reservoir of Pu-I Member of the Putaohua Oil Reservoir in the Daqing Lamadian Oil Field as an example, this study integrates data from field outcrops, well logging, and cores. Based on a precise characterization of the sand body structure, the contact relationships of the braided river reservoir sand bodies were systematically summarized. Three connectivity patterns of the braided river reservoir sand body in the lateral, longitudinal, and internal directions were established. The support vector machine (SVM) method was employed to quantitatively predict the connectivity of the reservoir sand bodies. Research findings indicate that by categorically optimizing the evaluation parameters of sand body connectivity and applying the SVM algorithm, the connectivity of sand bodies can be rapidly and accurately evaluated. Through mutual validation of dynamic and static data, the prediction accuracy reached 88%, compared with 81% for BP neural networks and 79% for fuzzy comprehensive evaluations. On this basis, a target-based geological modeling approach was adopted to establish a single sand body model controlled by 3rd to 4th level configuration interfaces. Leveraging the characterization of interlayers, the quantitative evaluation results of sand body connectivity obtained using the SVM method were utilized as deterministic data to assign conductivities to sand bodies across different zones and categories, thereby guiding the refined numerical simulation of oil reservoirs. This approach achieved the quantitative characterization and simulation of sand body connectivity coupled with interlayers and conductivities, and the numerical simulation results better reflect actual production conditions. These outcomes provide a new technical foundation for optimizing and adjusting oil field development in subsequent stages.