Investigation of Petrophysical-Properties Heterogeneity for Electrofacies Classification in Carbonate Reservoirs

Abstract

Electrofacies classification for rock typing in complex carbonate reservoirs is very challenging due to high degrees of heterogeneity in lithology, mineralogy, and pore structure. In extreme cases, multivariate pattern recognition and classification methods such as discriminant function analysis, principal component analysis, and cluster analysis using conventional well logs are insufficient for electrofacies classification, particularly in carbonate reservoirs with “low dynamic” log curves. This study investigates how heterogeneity of petrophysical properties can improve electrofacies classification for rock typing in carbonate reservoirs. It utilizes statistical measures of heterogeneity such as Coefficient of Variation, Lorentz Coefficient, and Dykstra-Parson Coefficient to quantify variability in petrophysical properties. Within this investigation, the heterogeneity logs based on wireline-log data including gamma ray, density, neutron, sonic, and photoelectric factor are calculated over set intervals of 3 ft (∼1m) through Mississippian limestone of the mid-continent. The heterogeneity logs and five predictor variables from logs, gamma ray; deep resistivity; photoelectric factor; density-neutron separation; and average density-neutron porosity, are linked to core lithofacies to train a Multi-Resolution Graph-Based Clustering model for predicting lithofacies in wells without core. The result is compared with the same clustering and classification technique which utilizes only the five predictor variables logs.

AAPG Datapages/Search and Discovery Article #90259 ©2016 AAPG Annual Convention and Exhibition, Calgary, Alberta, Canada, June 19-22, 2016