- Copyright © 2004 Society of Exploration Geophysicists
A novel approach to perform multiwell facies analysis was undertaken in the clastic reservoirs of Betty Field, Malaysia. The objective was to integrate the lithofacies information from core with petrophysical evaluation through the use of a neural network system. This provided a facies database that was used to estimate lithofacies in adjacent uncored wells at the 3D facies modeling stage.
A facies model was created for the key well based on petrophysical results calibrated with core description of the lithofacies. Effective porosity, permeability, and clay-volume calculated from petrophysical measurements and evaluation were used as inputs in the neural network application. The neural network model was then used to estimate facies in uncored wells in the field. The results were tied to five dominant core-derived lithofacies, namely high quality sand, laminated sand, poor quality sand, heterolithic shale, and massive shale. After multiple iterations, the network with the least error in training and cross-validation was selected for facies estimation.
This approach integrates core observations with petrophysical evaluation in a pragmatic yet robust facies framework. A strong relationship was revealed between the estimated lithofacies and the porosity and permeability across the field as demonstrated by predictable ranges of reservoir properties for each lithofacies.
Reservoir modeling using the stratigraphic framework obtained in this study has benefited greatly from the facies estimation and its propagation across the field. Reservoir properties are heavily influenced by lateral and vertical stacking of lithofacies units. This study provided control on the spatial and temporal facies variation and uncertainties, which …