Publications

Predicting permeability in uncored wells based on core data is very important in building 3D geological models for hydrocarbon in-place assessment and flow simulation. Traditionally, permeability has been computed from porosity-permeability equations, which are derived empirically from routine core analysis data. Recently, advanced statistical techniques have become popular, in which permeability models can be built in wells with sufficient core coverage using routine core analysis and wireline log data. Models built using an advanced statistical technique called the k-nearest neighbor clustering or Multi-Resolution Graph-based Clustering is presented in this paper. The model is then used to predict permeability in nearby uncored wells across the same reservoir. Routine core analysis and wireline log data from a carbonate reservoir are used to demonstrate this technique. The model is successfully applied to predict permeability in control wells (wells with core data but not used in building the model) and uncored wells. This paper will describe the methodology and show the permeability prediction results.