Understand Petrophysical Properties of Real World-Scale Complex Conventional Rocks Economically With Emerging Technologies

Abstract

Reservoir characterization and evaluation are crucial for an economic oil and gas exploration and production. Detailed petrophysical properties such as porosity, permeability, formation factor, relative permeability and capillary pressure are desired for an accurate estimate hydrocarbon as well as for making subsequent decisions. Digital rock physics (DRP) has gain lots of interests as an approach to obtain detailed petrophysical properties with relatively short time and economic resource requirement. Nevertheless, current DRP technologies are limited to small rock samples and have difficulties dealing with multi-scale, heterogeneous nature of rocks. Recent advances in computer algorithms and hardware combined with DRP allow us to gain insights into complex rocks as well as bridging knowledge of petrophysical properties from smaller to larger-scale rocks. We developed a workflow for understanding petrophysical properties of large-scale rocks by using machine learning (ML) techniques with DRP on high performance computing (HPC) platform. At the desired largest-scale rock sample, an unsupervised ML technique is used to identify distinct fabrics within a digital representation of a rock sample. Rock volumes with similar fabric are assumed to follow similar petrophysical properties variation (i.e. trend, range). Consequently, if required, a ML technique recommends an optimal number of finer-scale sub-samples based on fabrics information. Petrophysical properties relations are then derived from the fine-scale sub-samples and populated back to the larger-scale rock sample. Depending on the complexity of the rock sample, the procedure can be repeated recursively to cover the range of scales present in the rock. The present workflow is applied to multi-porosity carbonate and sandstone rock samples. The three-dimensional digital representation samples are segmented into fabrics – both resolved and unresolved. Properties of each unresolved fabrics are derived from their finer-scale sub-samples and upscaled to the large sample. In this work, porosity, permeability and formation factor of the samples are extracted. A comparison between computed values and laboratory experiments shows good agreement. We present an innovative workflow that can be used to understand petrophysical properties of rock at real world-scale. The workflow combines emerging technologies in many industries to open a way for an economic and efficient oil and gas exploration and production.

AAPG Datapages/Search and Discovery Article #90259 ©2016 AAPG Annual Convention and Exhibition, Calgary, Alberta, Canada, June 19-22, 2016