Permeability and porosity prediction using logging data in a heterogeneous dolomite reservoir: An integrated approach

The accurate prediction of permeability and porosity is an important foundation for high-quality reservoir identification and geological modelling. However, the strong heterogeneity, complex lithology and diagenesis in carbonates have brought great challenges to the accurate evaluation of reservoir permeability and porosity. In this study, 253 core samples with logging data from the Lower Cambrian dolomite reservoir in the Tarim Basin were collected to establish reliable methods for permeability and porosity prediction. Five typical permeability porosity correlations and six machine learning methods were applied to the core data and logging data to evaluate the applicability or prediction performance of different methods. By comparison, the flow zone indicator (FZI) and FZI* combined with discrete rock type (DRT) analysis in the five conventional models had a stronger ability to classify samples, and each DRT/DRT* had a special permeability-porosity relationship. Hence, an integrated indirect permeability prediction method was developed by combining the petrophysical rock typing methods (FZI or FZI*) with the PSO-SVM algorithm that hybridized the particle swarm optimization (PSO) and support vector machines (SVM). Compared with the direct machine learning prediction methods, the proposed integrated approach greatly improved the permeability prediction accuracy with the highest R-2 of 0.869, indicating that the combination of the conventional permeability model and machine learning algorithm had the potential to reduce the influence of heterogeneity on permeability prediction. The superior performance of the integrated approach in permeability and porosity prediction lays a good theoretical foundation for the identification of high-quality dolomite reservoirs.

Automated summary

This study established reliable methods for permeability and porosity prediction in Lower Cambrian dolomite reservoirs in the Tarim Basin by collecting core samples with logging data. Through comparing different methods, it was found that combining the flow zone indicator (FZI) and discrete rock type (DRT) analysis had a stronger ability to classify samples. An integrated indirect permeability prediction method was developed using a hybrid of particle swarm optimization (PSO) and support vector machines (SVM), which greatly improved accuracy. The integrated approach shows potential in reducing the impact of heterogeneity on permeability prediction.