Artificial Neural Networks May Help Predicting Abnormal Pressures in the Anadarko Basin
Brooklyn College of the City University of New York, Brooklyn, NY
Many sedimentary basins throughout the world exhibit areas with abnormal pore-fluid pressures (higher or lower than normal or hydrostatic pressure). Predicting the presence as well as other parameters (depth, extension, magnitude, etc.) of such areas proves often to be a challenging task. Among other tools used by specialists to meet that challenge, the sonic log (DT) seems to be preferred due to its accuracy and sensitivity to changes in porosity or compaction produced by abnormal pore-fluid pressures. Unfortunately, the sonic log is not commonly recorded in oil and/or gas wells. We propose using artificial neural networks (ANN) to simulate a sonic log by employing more available logs, such as natural gamma (GR), deep resistivity (RD), and caliper logs (CAL) (the last one, only for quality control procedure). The operation of the ANN can be divided into three steps: (1) supervised training of the neural network; (2) verification of the model validity by applying it to additional sets of data that contain the inputs (GR and RD logs) and the target values (DT). During this step, the validity of the model can be evaluated in terms of relative errors and goodness of fitting between the “training well” (step one) and the “confirmation wells” (step two); (3) the model is applied to wells containing only the input curves (GR and RD) and the output is the simulated DT. This procedure was applied to predict the presence of overpressured zones in the Anadarko Basin, Oklahoma. The results are promising and encouraging for the future development of the method.