INTERNATIONAL CONGRESS ON RECENT ADVANCES IN SCIENCES AND TECHNOLOGY - Kuala Lumpur - Malaysia (2019-02-20)

Facies characterization and prediction of carbonate reservoir in Central Luconia, Offshore Sarawak
Reservoir facies characterization and prediction are crucial steps in reservoir modeling. In this research, core study and neural network techniques have been applied to predict reservoir facies for Miocene carbonate buildup in the Central Luconia Province, Offshore Sarawak. Neural network needs to be trained with defined reservoir facies types before it can predict facies in areas with sparse data. Therefore, characterization of the carbonate reservoirs using 2D seismic data, well logs, cores and thin sections, was carried out prior to running the neural network program. The geometry of the carbonate buildup was clearly defined using seismic facies interpretation and showed asymmetrical morphology with internal architecture that directly manifested a clear depositional facies distribution pattern in the carbonate system. Based on vertical and lateral variation on the seismic reflectors three depositional sequences were interpreted in the buildup. A northeast to southwest apparent palaeo-wind margins was interpreted from the steeper platform slope facing to the northeast direction. The palaeo-wind margin is apparently associated with better reservoir quality facies distribution. However, from the core description four groups of depositional facies of the buildup were identified. These are Wackestone-Packstone, Floatstone-Rudstone, Bindstone-Bafflestone and Framestone. Textural analysis complimented by identification of diagnostic foraminifera in the petrographic analysis suggested depositional environment ranging from back reef, internal lagoon, and reef front to fore reef within an isolated carbonate platform setting. The results from core, petrographic studies and cross plots of measured porosity and permeability values showed an associate to five lithofacies for the carbonate buildup. These lithofacies are chalky-mouldic limestone, mixed carbonate (dolomitic limestone and calcitic dolomite), argillaceous limestone, tight limestone, and dolomite. The lithofacies were combined with well logs and seismic data to train the neural network software for prediction of carbonate lithofacies in uncored intervals. Supervised neural network approach was applied for the lithofacies prediction. Among the different prediction methods of the supervised neural network the result from back-propagation approach was able to produce acceptable result with some restrictions in selecting appropriate combination of well log data. The results from training of the well log data were good enough to be used for further work in uncored intervals of the platform. Prediction of lithofacies from the training of seismic data for reservoir facies prediction may enhanced using inverted seismic section data to provide additional constrain in training stage of the neural network.
Assoc. Prof. Chow Weng Sum, Assoc. Prof. Wan Ismail Wan Yosuff