Sca 2001-18.pdf

Nuclear magnetic resonance (NMR) relaxometry tests, carried out on partially saturated porous media, convey information on the distribution of fluids within the pore space. Proper interpretation of this information requires a means to isolate the effects of pore structure (pore and throat shape and size distributions, spatial correlations, etc.) wettability and saturation history. A pore network simulator of NMR relaxation that can clarify the effect of these factors on observed T2 distributions is developed in this work. The pore network model can be calibrated with pore structure parameters (pore volume and throat size distributions, pore connectivity) determined from analysis of stochastic 3D replicas of the pore space generated from thin-section image information. The presence of irregular and, possibly, fractal pore surfaces is accounted for through a "late pore filling" model calibrated to match mercury porosimetry data. The network model is then used to simulate the water distribution and proton magnetization decay under conditions of drainage and imbibition. The results obtained reveal: (a) conditions under which diffusive coupling between pores has a significant effect on the decay spectra, (b) contributions to the magnetization decay spectra from water trapped at the irregular pore surfaces, and (c) contributions to the magnetization decay spectra from (i) water in large pores which are inaccessible by the non-wetting phase during drainage and (ii) water in pores containing oil trapped during imbibition. INTRODUCTION With the increasing use of NMR techniques in the field and laboratory, there is a need to clarify how the NMR response of partially saturated reservoir rock depends on pore structure parameters, wettability and displacement history. Pore network simulation has been used extensively in the past to study the relationship of the aforementioned factors on capillary pressure, electrical resistivity and relative permeability. In recent years, pore network simulation has been endowed with predictive power, as a result of advances in (a) 3D stochastic reconstruction from thin-section image statistical information (Ioannidis et al., 1999; Liang et al., 2000; Bekri et al., 2000) and (b) grain deposition/diagenesis modeling (Bakke and Øren, 1997). If pore network models are to be further developed as tools for petrophysical evaluation, then their extension to include predictions of the NMR response of reservoir rock should be considered. This task was recently undertaken by Chang et al. (2000) and is further pursued in the present contribution.