Prediction of petrophysical static rock type through nonlinear dimensionality reduction and mutual information

Carbonate reservoirs are characterized by sedimentation, differential diagenesis, and tectonic processes, resulting in significant differences in porosity and permeability among different rock types, with solid heterogeneity, leading to a diversity of reservoir rock types. Due to the heterogeneity caused by the diagenetic process, the precise identification of petrophysical static rock type has always been a tremendous challenge for reservoir engineers and geologists. At the same time, the precise determination of carbonate static rock types can help formulate more appropriate oil and gas extraction strategies and increase the oil and gas recovery rate. In the face of growing global energy demand, accurate prediction of static rock types has become critical. In this paper, we develop several novel methods by combining the Jackknife Mutual Information and nonlinear dimensionality reduction methods for identifying carbonate static rock types. The proposed methods are motivated by the nonlinear relationships between crucial variables and static rock types, aiming to enhance the performance of prediction. We examine the empirical performances of our methods and several other methods on the dataset of two coring wells from a specific carbonate reservoir. The performance of prediction for static rock type indicates that our novel methods provide satisfactory prediction accuracy.