--> Combining Sequence Stratigraphy with Artificial Neural Networks to Enhance Regional Correlation and Determination of Reservoir Quality in the “Mississippian Limestone” of the Mid-Continent, USA
[First Hit]

AAPG Annual Convention and Exhibition

Datapages, Inc.Print this page

Previous HitCombiningNext Hit Previous HitSequenceNext Hit Stratigraphy with Artificial Neural Networks to Enhance Regional Correlation and Determination of Reservoir Quality in the “Mississippian Limestone” of the Mid-Continent, USA

Elizabeth Elium1, G. Michael Grammer1 and Matthew J. Pranter2

1 Oklahoma State University, Stillwater, Oklahoma, United States
2 University of Oklahoma, Norman, Oklahoma, United States

April 5, 2017 Wednesday 8:30 A.M. AAPG Annual Convention and Exhibition, Houston, Texas

Abstract

The “Mississippian Limestone” of the U.S. Mid-Continent region is a complex, highly variable, mixed carbonate and siliciclastic system that serves as an important unconventional hydrocarbon reservoir. Recent studies have focused on developing and applying a Previous HitsequenceNext Hit stratigraphic framework to enhance the understanding of depositional facies and reservoir architecture, and to enhance regional and sub-regional correlation. Throughout the region, the system is characterized by a 3-fold hierarchy of probable 2nd, 3rd and 4th order sequences and high frequency sequences. The third order sequences have proven to be the most reliable for correlation purposes due to distinct wireline Previous HitlogNext Hit signatures that have been confirmed by ground-truthing the logs to multiple cores and tying to facies stacking patterns. An integrated approach utilizing Previous HitwellNext Hit logs, Previous HitsequenceNext Hit stratigraphy, core Previous HitanalysisNext Hit, and 2D modeling is used to correlate and assess the reservoir quality of the “Mississippian Limestone” at both regional and sub-regional scales. Nine cored wells are used to create an artificial neural network (ANN) which is tested with k-means clustering methods to create a lithofacies Previous HitlogNext Hit. The ANN is tested and trained to determine the lithofacies present based on Previous HitwellNext Hit Previous HitlogNext Hit signatures alone, in an attempt to pick up small scale changes in this heterogeneous carbonate system. This lithofacies Previous HitlogNext Hit is applied to non-cored wells throughout the study area, which spans several counties, in order to predict the lithofacies in areas without core data. 2D geostatistical models are created from this data in order to correlate Previous HitsequenceTop stratigraphic frameworks, facies changes, and reservoir distribution in the Mississippian Limestone across the region.

AAPG Datapages/Search and Discovery Article #90291 ©2017 AAPG Annual Convention and Exhibition, Houston, Texas, April 2-5, 2017