--> Abstract: Neural Networks: Prediction of Carbonate Lithology and Permeability from Wireline Logs in a Miocene Buildup, Offshore Sarawak, by F. V. Abbots; #90982 (1994).
[First Hit]

Datapages, Inc.Print this page

Abstract: Neural Networks: Prediction of Carbonate Lithology and Permeability from Wireline Logs in a Miocene Buildup, Offshore Sarawak

Frances Vivien Abbots

A gas-bearing Miocene carbonate field is used as a case study to demonstrate the impact of neural network technology (Previous HitnonlinearNext Hit complex-pattern recognition) in carbonate reservoir modeling. In the Sarawak Miocene buildups, as in many carbonate reservoirs, most porosity is of diagenetic origin, and poroperm character is determined by dominant pore type (i.e., moldic, intergranular, intercrystalline). A fundamental prerequisite to modeling geometries of depositional/diagenetic bodies in 3-D is the ability to recognize these lithofacies (and their permeability values) from wireline logs in wells where core is unavailable; a task that has often proved elusive, as in this case, using conventional programming techniques. Neural networks appear to have solved these problems by their ability o "learn" the relationship between the core-defined lithofacies and permeability measurements from conventional petrophysical log suites.

Neural networks can recognize Previous HitnonlinearTop relationships, and they do not rely on predefined algorithms, thereby avoiding the limitation of fitting scattered data to linear trends. On this Miocene training set, lithofacies classifications were "learnt" to an accuracy of over 90%. The neural network solution for permeability prediction from logs halved the root-mean square error of the predictions obtained by conventional techniques. These results were obtained using standard log suites, making the technique especially suited to field reviews and regional studies.

Application of the trained networks to log suites of noncored wells in the field gives geological control by providing lithology and permeability profiles. This allows detailed interwell correlation of diagenetic rock types and their lateral porosity-permeability trends, forming the basis for 3-D reservoir modeling.

AAPG Search and Discovery Article #90982©1994 AAPG International Conference and Exhibition, Kuala Lumpur, Malaysia, August 21-24, 1994