Abstract
Facies interpretations’ fundamental inputs into three-dimensional (3D) petroleum systems models (PSM) form the basis for modeling calculations and are integral to understanding both conventional and unconventional reservoirs. Building facies models is time consuming and involves a lengthy petrophysical evaluation, and difficulties arise when data are inconsistent temporally and spatially. Machine-learning algorithms (ML), 3D basin facies models, PSM, and geostatistics have been used to constrain important properties at the local reservoir scale to produce reliable models. The final model forms show how unconventional reservoir characterization can be performed more effectively in the presence of sparse data. Here, vital reservoir properties (i.e., brittleness, maturity, porosity, pressure, and organic enrichment), often unavailable from local data, can be calculated.
Introduction
Natural gas production for 2017 in the US was the second highest level ever recorded (EIA 2018). The increases since 2005 have principally been the result of horizontal drilling and hydraulic fracturing; therefore, the US currently produces approximately all the natural gas it demands and is now a natural gas exporter. The road to this remarkable success has been a fact-finding mission and lengthy process in an attempt to precisely understand what constitutes an economically viable resource. Although results are mixed, much has been learned during the process.
Currently, the US has an evident understanding of shale reservoirs and many years of data acquisition to perform in-depth analysis to identify more accurate “sweet spots” for well placement. By comparison, other countries are just beginning shale exploration and development and, as a consequence, have to leverage learnings from the North American experience. Nevertheless, some regions in Europe have strict regulations, and the opportunity for large-scale exploratory drilling is limited (Godec and Spisto 2016).
The challenge for these countries is generating insight into the reservoir without the luxury of drilling vast numbers of exploratory wells.
Shale reservoirs are complex, and finding the optimal drill site and landing zone with appropriate socio-economic factors and shale rock quality is fundamental to securing economic success. However, requirements necessary to understand the physics of the reservoir are often lacking. Although geophysical and geomechanical logs [e.g., minimum horizontal stress, Poisson's ratio, Young's modulus, total organic carbon (TOC), etc.] (Eshkalak et al. 2014) are sometimes gathered and analyzed along with seismic data, it is far too common to identify a paucity of these important data. Further, spatial distribution of wells can be irregular or sparse, and quantity can be insufficient to provide comprehensive insight into formation and reservoir properties (Casey et al. 2018), presenting a significant challenge when building reliable reservoir models.
