MODEL TESTING OF METHANE HYDRATE OCCURRENCE ON THE NORTH SLOPE OF ALASKA WITH ARTIFICIAL NEURAL NETWORKS

Mary M. Poulton¹, Robert R. Casavant¹, Charles E. Glass¹, and Bo Zhao^2
1 Department of Mining and Geological Engineering, The University of Arizona, Tucson, AZ 85721
² Department of Geosciences, University of Houston, Houston, TX 77204

A multi-faceted research program at the University of Arizona (UA) is focused on a detailed and comprehensive characterization of heterogeneous gas hydrate and free-gas bearing reservoirs on the central North Slope of Alaska. In addition to geological and geophysical characterizations, research objectives also include the assessment of resource volumes, fluid distribution, and other geological and reservoir engineering inputs that may lead to commercial evaluation of this potential unconventional energy resource. The current area of interest (AOI) includes all of the Milne Point Unit (MPU), a large portion of the Kuparuk River Unit (KRU), and the western quarter of the Prudhoe Bay Unit (PBU) on the North Slope of Alaska. The analysis includes well log data from 90 wells across the AOI and a 3-D seismic volume over MPU.

One component of the UA research is the application of artificial neural network analysis (ANN) in characterizing and predicting potential gas hydrate and free-gas resources. A neural network is able to analyze seismic waveform characteristics that represent a horizon and form robust templates that can be used to match waveforms through a seismic volume^1,2. It has been shown that a neural network can also extract subtle patterns from our well log data sets. This may help characterize the heterogeneity of gas hydrate zones across the AOI and, when integrated with the seismic analysis, may lead to more accurate reservoir modeling.

Neural networks have been investigated to identify and map gas hydrate-bearing facies within the MPU seismic volume by analyzing the morphology of wavelets within a specified horizon. A preliminary analyses resulted in an initial model for methane hydrate formation in the MPU using a self-organizing map (SOM)³. An unsupervised (untrained) classification was performed using three seismic attributes: instantaneous frequency, amplitude acceleration, and dominant frequency extracted from 3D seismic data. The classification results of the seismic attributes showed that the SOM classification of interpreted gas hydrate-bearing zones correlated in several areas with gas hydrate-bearing zones identified in well logs. The dominant frequency attribute produced the most consistent results for tracking layers of suspected methane hydrate. In general, zones identified as possible gas hydrate-bearing layers were characterized by relatively high dominant frequency. Early SOM classifications were completed before satisfactory time-depth corrections for the seismic data, full stratigraphic analyses, chronostratigraphic sequencing, and fault pattern analyses.

Neural networks will also be used to develop a classification scheme derived from well log data and other petrophysical analyses⁴ to aid in identification of gas hydrate in stratigraphically complex areas in the MPU. The results will be used to test a regional tectonic model5 that may relate to emplacement of gas hydrate and free-gas resources.

Acknowledgements and Disclaimer:

The University of Arizona contribution is part of a larger collaborative program that includes researchers from the University of Alaska Fairbanks and the U.S. Geological Survey. BP Exploration (Alaska), Inc. provides overall project coordination and provided data for the characterization and modeling efforts. Interpretation and processing software was made available through support from the University Grants Program of Landmark Graphics Corporation and from GeoPlus Corporation. This research was funded by the Department of Energy (Award # DE-FC-01NT41332). The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.

References:

1. Poulton, M., (Ed.), 2001, Computational Neural Networks for Geophysical Data Processing: Pergamon, Amsterdam, 335p.

2. Poulton, M., 2002, Neural networks as an intelligence amplification tool: A review of applications: Geophysics, vol. 67, no. 3, pp. 979-993.

3. Zhao, B., 2003, Classifying Seismic Attributes In The Milne Point Unit, North Slope of Alaska: MS Thesis, University of Arizona, Tucson, Arizona.

4. Glass, C. E. 2003, Estimating pore fluid concentrations using acoustic and electrical log attributes, Interim Report, UA Gas Hydrate Project.

5. Casavant, R. R., 2001, Morphotectonic Investigation of the Arctic Alaska Terrane: Implications to Basement Architecture, Basin Evolution, Neotectonics and Natural Resource Management: Ph.D thesis, University of Arizona, 457 p.