--> Abstract: Quantifying Karst-filling Anhydrite Distributions from Pore to Log Scales; San Andres Formation, Permian Basin Part 2: Petrophysical Deterministic and Statistical Prediction of Anhydrite, by H. F. El-Sobky, A. É. Csoma, R. M. Phelps, and R. M. Ostermeier; #120034 (2012)

Datapages, Inc.Print this page

Quantifying Karst-filling Anhydrite Distributions from Pore to Log Scales; San Andres Formation, Permian Basin Part 2: Petrophysical Deterministic and Statistical Prediction of Anhydrite

H. F. El-Sobky, A. É. Csoma, R. M. Phelps, and R. M. Ostermeier
ConocoPhillips, Houston, TX, USA

Accurate quantification and modeling of anhydrite distribution is crucial for proper hydrocarbon flow simulation and production. Depositional anhydrite and anhydrite cements may represent persistent flow barriers to significant accumulations of hydrocarbons, which might be the case in many parts of the Permian Basin. Petrophysical determination of the anhydrite volume is not an easy task especially when the well log resolution is greater than the anhydrite variation. Two main integrated techniques are investigated to quantify and estimate anhydrite distribution; a petrophysical deterministic technique and an advanced statistical technique.

Routine core analysis XRD data is used to build a deterministic model for anhydrite volumetric estimations. Four-component petrophysical model is found to be appropriate for defining the petrophysical endpoints, where total porosity, volumes of dolomite, clastics, and anhydrite are the four major components that have been selected. Three main petrophysical measurements are used to solve for these components, which are density, cross-sectional area and thermal neutron response. The validity of the 4-component petrophysical model is largely verified through reconstruction of the thermal neutron porosity log utilizing that model.

Core Image segmentation (CIS) analysis is used to detect the anhydrite abundance based on an object-oriented classification techniques for three cored wells. A number of advanced statistical techniques such as multivariate alternating conditional expectation (ACE) and modular neural network (MNNT), are used to estimate anhydrite from well logs and results are compared to the CIS output. Both techniques are considered as iterative nonlinear schemes for data pattern recognition. The results from these analyses clearly show that MNNT is superior to both ACE and deterministic petrophysical techniques to estimate anhydrite abundance in this Permian Basin.

 

AAPG Search and Discovery Article #120034©2012 AAPG Hedberg Conference Fundamental Controls on Flow in Carbonates, Saint-Cyr Sur Mer, Provence, France, July 8-13, 2012