Permeability characterization of Schrader Bluff Sands using artificial neural networks

• Main Author: Deshpande, Aditya U.
• Format: Thesis
• Language: English
• Published: 2009
• Subjects: Oil sands; Oil reservoir engineering; Heavy oil; Petroleum reserves; Fairbanks; Milne Point; Alaska North Slope; north slope; Alaska
• Online Access: http://hdl.handle.net/11122/12800

Thesis (M.S.) University of Alaska Fairbanks, 2009 "Permeability is a fundamental and often difficult to predict property of any reservoir. This is especially true for unconsolidated formations where any type of physical permeability measurement is difficult. This study develops a more detailed picture of reservoir permeability by generating continuous predicted permeability logs for the Schrader Bluff sands. Schrader Bluff sands are a medium heavy oil reservoir currently produced from the Milne Point oil field on the Alaska North Slope (ANS). A total of about 400 ft of core samples from two Milne Point wells were analyzed using a probe permeameter. These data were then integrated with available permeability data and used along with electric well log data for training an Artificial Neural Network to obtain continuous predicted permeability logs. The predicted data were then used to make Modified Lorenz Plots to study the flow unit behavior and to identify possible flow units. A similar dual approach that includes both probe permeameter measured and predicted permeability data can be used for flow unit characterization in other reservoirs with deficient permeability datasets. This approach would be especially useful for permeability characterization of unconsolidated or semiconsolidated reservoirs"--Leaf iii 1. Introduction -- 2. Background -- 2.1. Introduction to the Schrader Bluff oil pool -- 2.2. Flow unit behavior -- 2.3. Challenges associated with measuring the permeability of the Schrader Bluff sands -- 2.4. Methods of permeability measurements -- 2.5. Direct measurement using a probe permeameter -- 2.6. Permeability prediction using strict mathematical relationships with well logs -- 2.7. Multiple variable regression (MVR) methods for calculating permeability -- 2.8. Using articial neural networks (ANN) to predict permeability -- 3. Methods -- 4. Results -- 4.1. Selection of well logs for analysis -- 4.2. Neural network model training testing and application -- 4.3. Final correlation (R²) values -- ...