Stochastic Multiscale Electrofacies Classification of Well Logs: A Case Study From North Slope, Alaska

Abstract

Lithofacies classification is commonly used as a stepping stone to property propagation and upscaling. The classification workflows used in the industry are mostly deterministic and they do not incorporate uncertainty analysis. Thus, geomodeling workflows are forced to incorporate well data as hard data. This study presents a stochastic multi-scale workflow for electrofacies classification. The workflow presented here extends existence workflows by using a combination of statistical methods including unsupervised neural network self-organizing maps, principal component analysis, k-means clustering, and hierarchal clustering. In this workflow, data are normalized and principal component analysis is applied to extract independent variables. Self-organizing maps or k-means clustering are used to create the initial classification space. Hierarchal clustering is used to study the relationship between the nodes in the classification space and build the upscaling tree. Measurements, be it from the same training dataset or from another dataset assuming stationary, is then classified stochastically based on a distance measurement between the model classification space and the data. Different realizations are created. The multi-scale visualization of realizations allows for the identification of important lithological changes and studying small scale vertical heterogeneity at the same time. The developed workflow is applied on the Shublik Formation, North Slope, Alaska. Five statistically significant electrofacies are identified. Results are validated using conventional geologic description of the core. There is a general agreement between core-derived lithofacies and well-log derived electrofacies. Planned extensions include the inclusion of spatial component to the classification through the use of hidden Markov chains, and optimizing parameters automatically.

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