Intelligent Classification of Carbonate Reservoir Quality Using Multisource Geophysical Logging and Seismic Data

Reservoir quality evaluation is an integral part of oil–gas exploration and development in fracture-vug carbonate formations. Varieties of geophysical data with different advantages are used in reservoir evaluation. These multisource data include conventional logs (CLs), borehole imaging logs (BILs), borehole acoustic reflection imaging logs (BARILs), and even seismic data. Obviously, combining these data can get richer formation information, which is more reliable than single-source data. In this work, the fusion dataset was constructed by calculating the principal components of CLs, BILs, and BARILs. The fusion dataset describes the fracture-vug information in different depths of investigation. Then, the agglomerative hierarchical clustering was used to analyze the fusion data and evaluate the reservoir quality in the borehole. Finally, taking the borehole-side seismic data and borehole evaluation results as training data, an intelligent model was trained based on the random forest algorithm. Also, the intelligent model can be used for reservoir quality classification in a 3-D space. In the case study, the evaluation results of reservoir quality are consistent with the validation data, and the accuracy of the final intelligent model is 85.15%.