The estimation of fluid properties and rock composition is an integral part of formation evaluation. Unconventional resources such as organic-rich shales are geologically complex, due to great variability in rock composition and post-depositional diagenetic processes. In addition, low porosities, such as those usually encountered in unconventional reservoirs, are not well defined from conventional logs due to the large influence of the rock lithological components and high uncertainties of the log measurements. The errors in estimated porosity not only relate to measurement errors, but also to the incomplete knowledge of matrix parameters. Consequently, a reliable method that integrates mineral and fluid petrophysical models is needed to determine the key formation properties in whole-rock characterization.
It is well known that the most important petrophysical parameter obtained from nuclear magnetic resonance (NMR) logging data is a lithology independent porosity. The ill-defined problem in NMR inversion requires proper regularization methods to stabilize the inversion results. The incorporation of conventional logs as penalty constraints in the NMR inversion process can lead to an improved quantification of fluid typing and formation porosity. However, the accuracy of the fluid model is also influenced by the quality of the estimated matrix properties. Geochemical logs based on pulsed neutron technology have been successfully applied to determine elements for identifying lithology and mineralogy. A probabilistic method incorporating geochemical and conventional log measurements can reliably estimate the matrix mineral composition and porosity. However, a good understanding of the fluid properties is also important for accurate results from the mineral model. The fluid and the mineral model complement each other. Consequently the integration of the two models using NMR, geochemical, and conventional logs can provide an improved whole-rock characterization.
In this work, a unified workflow that includes petrophysical relations and rock physics models is proposed. Probabilistic mineralogy inversion using geochemical and conventional logs solves for matrix properties that are used as inputs for the fluid inversion model. Simultaneous inversion using NMR echo trains and conventional logs determines pore fluid composition and formation porosity that are used as fluid parameters to improve the results in the mineral model. By using both inversion models jointly, a set of petrophysical properties that account for all available log measurements can be determined for the whole rock, and the integrated workflow can potentially reduce the uncertainty in the well log interpretation that usually applies a limited set of logs. A field example from an Argentina well in the Vaca Muerta unconventional play is presented and discussed to show the superior interpretation proficiency of the proposed methodology.