Abstract
This paper describes a methodology for delineating tar surface, incorporating it into a geological model, and the process for numerical modeling of oil viscosity variation with depth above the tar surface. The methodology integrates well log data and compositional fluid analysis to develop a mathematical model that mimics the oil's property variation with depth. While there are a good number of reservoirs that fit this description globally, there is a knowledge gap in literature regarding best practices for dealing with the peculiar challenges of such reservoirs. These challenges include; (i) how to delineate the top-of-tar across the field, (ii) modeling of Saturation Height Function (SHF) in a system where density and wettability is changing with depth, and (iii) the methodology for representing the depth-dependent oil properties (especially viscosity) in reservoir simulation.
Nuclear magnetic resonance (NMR) logs were used to predict fluid viscosity using a technique discussed by Hursan et al. (2016). Viscosity regions are identified at every well that has an NMR log, and these regions are mapped from well to well across the reservoir. Within each viscosity region, the analysis results of fluid samples collected from wells are used to develop mathematical models of fluid composition variation with depth. A reliable SHF model was achieved by incorporating depth-varying oil density and depth varying wettability into the calculation of J-Function. A compositional reservoir simulation was set-up, using the viscosity regions and the mathematical models describing composition variation with depth, for the respective regions. Using information obtained from literature as a starting point, residual oil saturation was modeled as a function of oil viscosity.
Original reservoir understanding places the top of non-movable oil (tar) at a constant fieldwide subsurface depth, which corresponds to the shallowest historical no-flow drillstem test (DST) depth. Mapping of the NMR viscosity regions across the field resulted in a sloping tar-oil contact (TOC), which resulted in an increase of movable hydrocarbon pore volume. The viscosity versus depth profile from the simulation model matched the observed data, and allow the simulation model better predict well performance. In addition, the simulation model results also matched the depth-variation of observed formation volume factor (FVF) and reservoir fluid density. Some wells that have measured viscosity data but no NMR logs were used as blind-test wells. The simulation model results also matched the measured viscosity at those blind-test wells. These good matches of the oil property variation with depth gave confidence, that the simulation model could be used as an efficient planning tool for ensuring that injectors are placed just-above the tar mat. The use of the simulation model for well planning could reduce the need for geosteering while drilling flank wells, leading to savings in financial costs.
This paper contains a generalized approach that can be used in static and dynamic modeling of reservoirs, where oil changes from light to medium to heavy oil, underlain by tar. It contains recommendations and guidelines to construct a reliable simulation model of such systems.