Cased Hole Solutions: Predicting Open Hole Logs Using Artifical Neural Networks

John F. Gegg
EnCana Oil and Gas USA, Denver, CO

Neural Net predicted open hole triple combo logs (density, neutron, resistivity) have been used successfully in Jonah for reservoir characterization, net pay determination, frac staging, and OGIP calculations. Drilling and Hole conditioning problems in the area create poor open hole logging conditions, many times resulting in the inability to reach TD (12,000 + feet) with logging tools. In wells where an open hole logging suite is obtained, the borehole conditions adversely affect tool measurements because of borehole rugosity and tool sticking issues. These issues have been the motivation to move toward a cased hole logging program. The benefits have been lower costs, reduced risks, and improved vertical resolution and logging data accuracy.

The process and interpretation workflow involves training and application wells. A training well has both open hole logging data and cased hole logging data. An application well has only cased hole data. An artificial neural network (ANN) model is developed for the training well, across a specific stratigraphic interval or geographic area, using the relationship between the open hole triple combo data and the cased hole data. Pulsed neutron logs work especially well for this because of the strong correlation between sigma, inelastic and capture count rates to resistivity, bulk density, and porosity respectively. Once developed, the ANN model can be applied to application wells that have only cased hole data. The ANN model results are checked against offsetting wells and normalized (if necessary) to insure the most accurate porosity and resistivity data possible. After analysis for frac staging, these data can then be integrated with other wells in the field for net pay and OGIP calculations and reservoir characterization.

AAPG Search and Discovery Article #90092©2009 AAPG Rocky Mountain Section, July 9-11, 2008, Denver, Colorado