--> Application of a Training-Image Library to Fluvial Meandering Facies Models Using Multi-Point Statistics Conditioned on Analog-Based Forward Models

AAPG ACE 2018

Datapages, Inc.Print this page

Application of a Training-Image Library to Fluvial Meandering Facies Models Using Multi-Point Statistics Conditioned on Analog-Based Forward Models

Abstract

Meandering fluvial systems form highly compartmentalized hydrocarbon reservoirs. Variogram- and object-based modeling techniques commonly fail to reproduce the geometry, distribution and lithological heterogeneity of major geobodies (e.g., point-bar elements and sinuous channel-fill deposits, mud drapes).

A novel workflow for the generation of training images of fluvial meandering systems using Multi-Point Statistical techniques (MPS) has been developed. The aim is to produce a suite of models with higher geologic realism compared to outputs of traditional methods. The workflow includes the use of a library of training images in combination with tailor-made auxiliary-variable maps designed to handle non-stationarity. Training images with different levels of stationarity have been tested and included in a library to enable geomodelers to select the most suitable reservoir representation.

The training images are created using quantitative information derived from a relational database of geologic analogs (Fluvial Architecture Knowledge Transfer System; FAKTS), and a forward stratigraphic modeling tool which simulates fluvial meander-bend evolution and resulting point-bar facies organization (Point-Bar Sedimentary Architecture Numerical Deduction; PB-SAND). The devised training images incorporate fundamental features of the facies architecture of fluvial point-bar elements and larger meander belts composed of these and related elements.

The application of training images has been optimized to two MPS algorithms: SNESIM and DEESSE. To best model particular fluvial meandering successions, realizations have been performed whereby optimal reproduction of facies proportions, facies relationships, and architectural geometries is achieved, in part through incorporation of stationarity in the training images. The sensitivity of input parameters has been analyzed with multiple simulations across parameter space so as to define optimized modeling recipes for different fluvial systems, i.e., pairings of training images with sets of input parameters and auxiliary maps, and selections of appropriate MPS modeling algorithms. Modeling outcomes are compared quantitatively and qualitatively against corresponding facies models generated using variogram-based techniques.

Results show that MPS techniques benefit from training images based on forward modeling to deliver realistic realizations better able to incorporate the fundamental heterogeneities of fluvial meandering systems.