Microporosity Quantification and Flow Unit Prediction Using Neural Networks: A Case Study from Malaysia

Abstract

AbstractThis paper aims to quantify microporosity and access its impact on reservoir quality properties, to identify Hydraulic Flow Units (HFU) using core porosity and permeability data, to predict HFU and accurately estimate permeability in un-cored intervals/wells. Mercury Injection Capillary Pressure (MICP), Nuclear Magnetic Resonance (NMR) at high pressure (T2 distribution), and Water-Oil relative permeability curve data were used to quantify microporosity. The median pore-throat size was estimated through Windland's R35 equation and modified Kozeny-Carmen equations led to an accurate determination of Flow Zone Indicators (FZI) in each cored interval. Subsequently, Neural Networks were used to predict HFU in un-cored intervals using log data, this model was then extrapolated to estimate permeability in un-cored intervals. Microporosity estimations from MICP are relatively higher compared to NMR and Water-Oil Relative Permeability estimations. Estimations from MICP range from 25-69% while 11-32% is the range of estimations from the other methods. In conjunction to MICP data and clay volume assessment, two distinct zones were identified. The top intervals are characterized by high microporosity levels and steep-convex capillary pressure curves while the bottom intervals have lower microporosity levels and steep-concave capillary pressure curves. Analysis indicated that there are six (6) flow units and that the correlation coefficient (R2) between FZIcore and FZIlog is 99.89% and 91.89% for Kcore and KLog thus validating the model to be used for predictions in un-cored intervals. Through Windland's R35 equation it was concluded that there are two dominant pore throat sizes, meso and macro with ranges between 0.72 – 2.53 μm.