Semester
Spring
Date of Graduation
2002
Document Type
Dissertation
Degree Type
PhD
College
Statler College of Engineering and Mineral Resources
Department
Petroleum and Natural Gas Engineering
Committee Chair
Khashayar Aminian.
Abstract
The Appalachian Basin has numerous abandoned or marginally producing oilfields having significant recoverable oil remaining in place. Typically production records and reservoir data is not available. Presented is a new methodology, applicable to any field having limited records, designed for reservoir characterization via flow unit identification.;This methodology utilizes limited core permeability data from a few wells as a key to predicting flow units within a field when only log-based data is available. Primary software tools used include NeuroShell 2, an artificial neural network (ANN) program, and the Boast98 numerical simulator.;Various techniques for flow unit identification including graphic approaches using the permeability-porosity relationship within a given flow unit and ANN (Kohonen) analysis are utilized as a part of the methodology developed. The core data and flow units are utilized in neural network models designed to predict flow units and permeability field-wide using only electric well logs. Field wide prediction of flow units and permeability are products of the study.;The study field selected was the Jacksonburg-Stringtown field. The producing horizon is the Upper Devonian Gordon sandstone. Discovered in 1895, waterflood operations were commenced in 1981. The characterization study utilizes core data and electric logs (gamma ray-density) from six core wells for flow unit identification.;Two dual five spots were selected in the field for verification of this new methodology. Reservoir simulation analysis utilizing the predicted flow units and permeability was completed for comparison to actual production records. A close match was achieved. As another comparison step a single layer simulation model for the two patterns was generated. The single layer model included the same inputs as the flow unit model except for thickness and average permeability. The simulation model utilizing flow units was a far more accurate prediction method.
Recommended Citation
Thomas, Benjamin Hale, "Flow unit prediction with limited permeability data using artificial neural network analysis" (2002). Graduate Theses, Dissertations, and Problem Reports. 2431.
https://researchrepository.wvu.edu/etd/2431