A New Workflow to Upscale and Propagate Saturation-Dependent Petrophysical Properties From Wireline Logs to 3-D Geocellular Models

Abstract

A workflow is presented which overcomes one of the more intractable problems of reservoir characterisation: upscaling and propagating saturation-dependent (dynamic) petrophysical properties from sparsely located and sampled 1D wells to every cell in a 3D geocellular reservoir model. Such properties, being array data, can only be propagated from secondary variables using an appropriate and consistent parameterisation. Advanced classification techniques are linked with a scale-independent parameterisation to implement the new workflow.

Classifications are developed using a Bayesian-based multivariate clustering technique which provides a probabilistic rock typing at each scale. The resulting multivariate model is then used to implement the propagation of static petrophysical properties away from regions where they were measured into unsampled regions in either the well or the geocellular model. The propagated properties are then upscaled using conventional methods.

The scale-independent parameterisation uses three Characteristic Length Variables (CLVs; Curtis, 2000 and 2015). The CLVs are calculated from the static and dynamic petrophysical properties at the fine scale and used to build a classification model. After upscaling the static properties, two of the CLVs are used in the model to predict the third at the coarser scale, thus facilitating the upscaling of the dynamic properties.

An example is presented for a siliciclastic reservoir in which both static and dynamic properties are moved from wireline log scale to the geocell scale in 1D and then into all geocells of a 3D geocellular model. The effects of the homogenisation of the properties at increasing scales is clearly seen in the CLVs derived from each level of upscaled data. The effect is also clearly evident in the capillary pressure curves developed at each scale. Saturation profiles with depth developed from the capillary pressure functions at both scales show a very good match, albeit with less rugosity at the geocell scale.

The new workflow provides a robust and consistent framework in which to implement the scale changes involved when integrating logs with geocellular models. The adoption of scale-independent CLVs permits consistent classification at any scale. The inherent value of the CLV multivariate model is realised when it is used to propagate petrophysical properties into the 3D volume, away from the necessarily limited set of well-derived data points used in their evaluation.

AAPG Datapages/Search and Discovery Article #90350 © 2019 AAPG Annual Convention and Exhibition, San Antonio, Texas, May 19-22, 2019