Predict permeability from wireline logs using neural networks
- Texaco Exploration and Production Inc., Denver, CO (United States)
Artificial neural networks offer engineers an alternative to traditional regression techniques used in analyzing permeability from well logs. Additionally, neutral networks provide accuracy, consistency and improved overall quality to reservoir management. A comparative study was conducted at Texaco`s Stockyard Creek field, T154N-R99W, Williams County, N.D., using standard statistical techniques, neural networks and core data to assess permeability in the Mississippian Mission Canyon formation from the well log data in the field. Stockyard Creek field is a structural/stratigraphic trap where the oil is emplaced within porous dolomites that grade into tight limestones. For this study, the reservoir at Stockyard Creek field was characterized by core analysis, petrographic analysis and the SEM examination to determine crystal size, pore, diameter, pore throat diameter, porosity type and depositional environment. This study illustrates the advantage of neural networks over statistical regression techniques when analyzing formation permeability from well logs. The neural network solutions to this complex problem takes the evaluation beyond regression in that not only does the neural network accurately produce permeability, but it also maps the physical model of the reservoir in its connection weight pattern.
- OSTI ID:
- 49297
- Journal Information:
- Petroleum Engineer International, Journal Name: Petroleum Engineer International Journal Issue: 5 Vol. 68; ISSN 0164-8322; ISSN PEEID4
- Country of Publication:
- United States
- Language:
- English
Similar Records
Modeling of capillary pressure behavior using standard open hole wireline log data: Demonstrated on carbonates from the Middle East
Quantification of porosity and permeability reduction due to calcite cementation using computer-assisted petrographic image analysis techniques