Log interpretation of carbonate rocks based on petrophysical facies constraints

The complex pore structure of carbonate reservoirs hinders the correlation between porosity and permeability. In view of the sedimentation, diagenesis, testing, and production characteristics of carbonate reservoirs in the study area, combined with the current trends and advances in well log interpretation techniques for carbonate reservoirs, a log interpretation technology route of “geological information constraint + deep learning” was developed. The principal component analysis (PCA) was employed to establish lithology identification criteria with an accuracy of 91%. The Bayesian stepwise discriminant method was used to construct a sedimentary microfacies identification method with an accuracy of 90.5%. Based on production data, the main lithologies and sedimentary microfacies of effective reservoirs were determined, and 10 petrophysical facies with effective reservoir characteristics were identified. Constrained by petrophysical facies, the mean interpretation error of porosity compared to core analysis results is 2.7%, and the ratio of interpreted permeability to core analysis is within one order of magnitude, averaging 3.6. The research results demonstrate that deep learning algorithms can uncover the correlation in carbonate reservoir well logging data. Integrating geological and production data and selecting appropriate machine learning algorithms can significantly improve the accuracy of well log interpretation for carbonate reservoirs. [Display omitted] •A log interpretation technology route developed for carbonate reservoirs.•Principal component analysis (PCA) employed to establish lithological identification criteria with an accuracy of 91%.•Bayesian stepwise discriminant method used for sedimentary microfacies identification with an accuracy of 90.5%.•The mean porosity interpretation error compared with core analysis.