ABSTRACT: Permeability Estimation Using a Neural Network: A Case Study from the Roberts Unit, Wasson field, Yoakum County, Texas
OSBORNE, DEBRA A., Texaco Exploration and Production Inc., Midland, TX
Accurately estimating reservoir permeability is vital to reservoir simulation. The best method for determining reservoir permeability is to model core-derived permeability data. However, most Permian basin oil fields lack sufficient core coverage for core-based models. Therefore, the common method has been to develop linear relationships between core-derived porosity and permeability, then apply these relationships to porosity logs from noncored wells. This method has limitations because the linear relationships commonly are poor.
Neural network technology provides an alternative method for determining reservoir permeability. Neural networks estimate permeability based on the relationships between many reservoir characteristics, not just between porosity and permeability.
Data from five cored wells in the San Andres (Upper Permian) reservoir of the Roberts unit were loaded into a neural network designed to predict permeabilities. This neural network had 1 hidden layer with 30 processing elements and used a sigmoid activation function. The network was trained in 3.1 million iterations using the geographic location of the cored well, the depth, the core porosity, and the specific reservoir flow unit as inputs, and the difference between the core-derived permeability and the linear-regression-derived permeability as the output. A 0.81 correlation coefficient was calculated for the neural-network-derived permeability values. This compares to a 0.44 correlation coefficient for the linear-regression-derived permeability values. Neural-network-derived permeab lities in noncored wells are consistent with production data, whereas linear regression-derived permeabilities are inconsistent.
AAPG Search and Discovery Article #91018©1992 AAPG Southwest Section Meeting, Midland, Texas, April 21-24, 1992 (2009)