Abstract
Nuclear magnetic resonance (NMR) logs in long, high-angle wells drilled across a single reservoir often consist of a few recurring features which are difficult to identify visually in the NMR T2 distribution logs. This paper presents a fast data clustering method called NMR factor analysis (NMRFA) that categorizes massive LWD NMR datasets into a few poro-fluid facies, which enables a quick, data-driven geological interpretation using single or multiple wells.
NMRFA is a data reduction technique that aims to describe and interpret variability in NMR spectral logs using a small number of unique components. These components are clustered statistically to distinct groups called poro-fluid facies, which reflect combinations of pore volume, pore size, and fluid NMR properties. While drilling the well, the NMRFA is applied to the real-time T2 distribution measurements to quickly associate new observations of rock quality or heavy oil with stratigraphic understanding to assist well placement decisions. Joint interpretation of the NMR-based para-fluid facies logs from multiple laterals underpins data-driven updates of reservoir-scale geological facies maps.
After testing the method on two data sets already calibrated to core and formation testing measurements, the NMRFA technique was used in three laterals drilled back-to-back in a thin carbonate where pore size variations and possible presence of heavy oil were expected. In the first lateral, the best NMR facies were mostly observed in the first part of the well, followed by poorer poro-fluid facies that indicated the presence of heavy oil. At the end of the well, the facies analysis indicated mostly tight rocks. In the second lateral, the NMR log suggested a more homogeneous facies distribution than in the first well, with medium T2 and medium porosity. Tar or heavy oil were not encountered in this well. With the facies trends in the first two wells, the third lateral targeted a likely occurrence of the high-quality facies seen in the top section of the first well. After crossing a tight zone, this well encountered an excellent NMR facies of high porosity and very long T2 with no indications of heavy oil. The well was successfully navigated within the favorable facies whose quality matched or exceeded the best rocks encountered by the first two laterals.
NMRFA analysis was used for poro-fluid facies evaluation of LWD NMR datasets in real-time settings for the first time to better categorize complex reservoir types in horizontal wells. The robust, data-driven analysis method and its intuitive log-based and structural visualizations helped well placement decisions and enhanced structural learnings in three extended reach laterals drilled in a complex carbonate reservoir.
