Conditioning surface-based geological models to well data using artificial neural networks

Abstract Surface-based modelling provides a computationally efficient approach for generating geometrically realistic representations of heterogeneity in reservoir models. However, conditioning Surface-Based Geological Models (SBGMs) to well data can be challenging because it is an ill-posed inverse problem with spatially distributed parameters. To aid fast and conditioning, we use here SBGMs that model geometries using parametric, grid-free surfaces require few parameters represent even geological architectures. A neural network trained learn the underlying process by learning relationship between parametrized SBGM inputs resulting facies identified at locations. condition these observed data, achieved replacing forward pre-trained optimizing back-propagation technique applied training network. An analysis uncertainties associated conditioned realisations demonstrates applicability evaluating spatial variations away from control modelling. This geologically plausible models are calibrated could also extended other techniques such as object- process-based