Physics-Based Probabilistic Permeability Prediction in Thin-Layered Reservoirs: Transport Theory, Dielectric Dispersion Logging and Core-to-Log Bayesian Statistics

Accurate permeability prediction is probably the most challenging issue in reservoir characterization, and, at the same time, it is one of the most desired targets. Conventionally, a permeability profile is inferred from core-calibrated algorithms applied at the well location to different openhole logs. However, as the models are not exact, and the assumptions are not always satisfied, the uncertainty attached to the results is high. This uncertainty is even higher in thin-layered reservoirs characterized by bed thicknesses well below the resolution capabilities of standard logging tools. In this respect, the paper deals with a novel physics-based probabilistic methodology for high-resolution permeability estimation. This relies on core data and dielectric dispersion logging (DDL)—its cm-scale vertical resolution and the related fit-for-purpose petrophysical model make the DDL tool response suitable to capture and describe the permeability heterogeneity of these very thin-laminated scenarios. The approach is presented through a study performed on several wells drilled and cored into gas-bearing distal turbidite reservoirs. Their exploration and production are particularly challenging due to the presence of thinly bedded sections with individual laminations sometimes thinner than 1 cm. DDL can be a useful tool to characterize these thin-layered scenarios since the response of rock to electromagnetic (EM) fields at different frequencies is a good indicator of its components and microscopic structure. A dedicated interpretation model has been developed to invert DDL data and obtain high-resolution (cm-scale) estimates of shallow water volume fraction, shallow water salinity, and a textural parameter that is correlated to cation exchange capacity (CEC). Permeability is then computed from a physics-based analytical model taking advantage of transport phenomena in composite materials, effective medium theory, dielectric dispersion modeling, and selected core data (i.e., grain-size distribution and CEC). All the input parameters of the proposed methodology come from DDL outcomes embedded into a core-to-log Bayesian framework. This generates a high-resolution permeability profile, together with the associated uncertainty. The match with core measurements in key wells proves to be very accurate. The prediction capability has also been demonstrated on dozens of wells by comparing DDL-based permeability to wireline formation tester (WFT) results (considering the differences in scale and lateral extent of the techniques). It is worth noting that the major strength of the method relies on the fact that no additional calibrations and/or adjustable parameters are needed. Given the high vertical resolution of the DDL tool, it is one of the few methods suitable for thin-layered reservoirs.