Advances in NMR Fluid Typing Assist in the Petrophysical Evaluation of a Carbonate Well Drilled With Oil-Base Mud

Abstract

Nuclear magnetic resonance (NMR) is the key petrophysical evaluation technology in Brazil’s deep water Aptian stromatolytic carbonate reservoirs. As such, it is extremely important to obtain accurate fluid typing volumetric fractions of water, oil, and oil-based-mud filtrate (OBMF).

A common means of fluid typing is based on using T2-Diffusivity- (T2D) maps. However, when fluid properties are very similar, the T2D maps fail to resolve them, which reduces the quality of the evaluation. Other attempts using 1D spectra ―either T1 or T2 distributions― are normally not considered and typically evaluated via cutoff.

In this work, we present a solution for fluid typing based on blind-source separation (BSS), a statistical machine-learning technique. Specifically, we separate sources (fluid components) blindly based on their statistical independence, which is the only requirement imposed on the NMR input data. Considering the three fluids mentioned above, this requirement is reasonably fulfilled.

Results show that determining BSS based on independent component analysis (ICA) is a powerful technique to retrieve the fluid components of the NMR signal. For T1 or T2 spectra, we are able to retrieve the fluid components. Results were also validated by simulations and against lab data.

With respect to the OBMF component, its presence indicates good permeability. Thus, we elaborate on an OBMF-based permeability equation that implicitly honors the presence of unconnected vuggy porosity and is independent of cutoff values. Finally, the OBMF-based permeability can be used as a rock-quality indicator to support depth-selection decisions for pressure tests and fluid sampling.

INTRODUCTION

Achieving a good petrophysical evaluation fluid typing in Brazil’s deepwater Aptian stromatolytic carbonate reservoirs —part of the Presalt trend— is a very tough, challenging task: the resistivity log responses are influenced by both lithology and fluids, and the NMR logs —key for the interpretation— frequently present overlapping of reservoir fluids and oil-base mud filtrate. In addition, the poorly connected vuggy porosity, derived from diagenetic processes, raises questions about the reliability of applying T2 cutoffs for determining free- and bound-fluid volumes, essential variables in the Timur-Coates permeability estimation (Trevizan et. al, 2014), which usually supports the adequate selection of depths for pressure tests and sampling points.