AAPG Middle East Region Geoscience Technology Workshop:
3rd Edition Carbonate Reservoirs of the Middle East

Datapages, Inc.Print this page

Chancing Methods to Predict Porosity in a Middle Eastern Carbonate Reservoir from Full-Function Machine-Learning Neural Networks, Seismic Attributes and Inversions


Integrated reservoir management reduces uncertainty in hydrocarbon reserves estimates. Specifically, integrated seismic reservoir characterization of carbonate rocks potentially thwarts uncertainties arising from reservoir heterogeneity owed to a complex geological history accumulated under a protracted regimen that incorporates biochemical origins, facies, texture, chemical weathering, mineralization, stylolitization, karst, diagenesis, (paleo)stresses, and tectonics. Therefore, for the same 3D seismic survey over a Middle East carbonate reservoir, four independent porosity prediction workflows are featured that increasingly reconcile 1.) geophysical (seismic attributes), 2.) geological (lithostratigraphy vs. sequence stratigraphy), and, 3.) engineering components (flow units) to address reservoir complexity in terms of paleo-depositional environment, stratigraphic architecture and a property distribution that honors flow zonation. In the first workflow, trace-based reflectivity waveform changes serve to classify reservoir porosity in terms of amplitude and phase to be qualitatively calibrated to seismic attribute expression of spectral decompositions, multi-trace attributes, well logs, as well as a modern-day environmental analog of deposition. The second workflow directly predicts reservoir earth impedance and porosity by combining well and seismic data applying a new-age, i.e., full-function, non-linear neural network, that is additionally augmented by regularization to stabilize predictive operators for more unique results. A third workflow transforms earth impedances from model-based, deterministic inversion to porosity using rock physics. Finally, the fourth workflow aims to provide an answer to the prerequisite “How to bridge static and dynamic data?” by introducing fine-scale reservoir flow units into the prior model of a seismic stochastic inversion, because definition of flow units creates the proper geodynamic “bricks” for implementing a more accurate and reliable dynamic simulation. Pros and cons of each method in terms of practitioner’s criteria, such as accuracy of correlation, speed, ease-of-use, availability, etc. are subsequently discussed and ranked.