Logging Facies Classification and Permeability Evaluation: Multi-Resolution Graph Based Clustering

Abstract

Abstract In the process of reservoir parameters evaluation, it is common to use PoroPerm index relationship model to calculate the permeability. Due to the reservoir heterogeneity, the modeling error of this method can be relatively large. The errors in the calculated permeability directly impact the productivity estimation and detailed study of reservoir in the oilfield development and production stage. Therefore, improving the accuracy of the reservoir permeability modeling has been a major challenge in logging research. In this paper, we propose a new method to improve the accuracy of reservoir permeability modeling using the Multi-Resolution Graph based Clustering. Specifically, we calculate the Neighboring Index (NI) and adapt the graph based multi-resolution clustering algorithm to classify the logging data. The Kernel Representative Index (KRI) is calculated for the clustering results and the optimal logging facies are identified accordingly. Integrating with the information such as well logging data and core thin section image, the correspondence is established between logging facies from the target areas and their lithology. Then, a permeability evaluation model is constructed using the logging facies as the constraints. The logging facies classification method using Multi-Resolution Graph based Clustering overcomes the shortcomings of traditional clustering algorithm using Graph theory or the k-Nearest Neighbor. The method of this paper can efficiently classify the logging data and identify the optimal facies using the NI and KRI derived from the normalized Euclidian distances of logging data. This method also improves the resolution of the logging facies classification. Logging facies are comprehensive integrations of reservoir lithology, physical property and pore structure. Proper logging facies classification can show sedimentary environments and reservoir heterogeneity characteristics. Our methods establish a correspondence between the target block well logging data and their lithology through analyzing the logging data combined with core thin section. It can be applied in further and detailed logging study. Based on the correspondence of logging facies and lithology, we have built different permeability evaluation models under the constraints of logging facies. This model has been successfully applied to evaluate the unconsolidated sandstone permeability in PengLai 25-6 oilfield. Compared with conventional PoroPerm index relationship model, the calculated permeability using our methods has a much higher consistency with the core analysis permeability. The accuracy of permeability evaluation improved significantly and provided reliable permeability parameters for precise estimation of reservoir.