Simulation of Complex Composite EOR Processes at Lab and Field Scale

Abstract

Abstract Experiments on injection of composite EOR fluids, simultaneously or sequentially, have been validated by small scale simulation, with cell sizes in the millimeter-range. The injected fluids have been a combination of low-salinity brine, surfactant, and polymer. This presentation addresses the challenges of extending the lab scale results to larger models, up to field size. The complex injection scheme itself is not straightforward to model in any simulator, and novel techniques have been developed to handle this. The main mechanisms that must be addressed are concentration dependent properties for all fluids involved and any combination of these, e.g. relative permeability and viscosity. The methods used have been adapted to both a black oil simulator, and a composition based simulator. As the basic modeling principles for black oil and compositional simulators are per se very different, the way of handling fluid descriptions from the user view point are naturally quite different for the two approaches. Hence also adaption to the composite processes necessarily differs. History matching (validating) experiments at lab scale was successfully done with both simulators, with qualitative equal though not identical results. When attempting to generalize results to larger models, two issues have been addressed; extension, where the size of the simulation model was increased without changing the grid resolution (cell size), and upscaling, where model size was unchanged while grid cell sizes were increased. Ideally, simulated results should be identical at different scales and grid resolutions, at least within standard grid resolution errors. Both simulators were well-behaved with respect to upscaling, while the compositional simulator handled extension better than the black oil simulator. Moreover, the black oil simulator was found to be very sensitive to smoothness of relative permeability curves. As example, using a relative permeability curve that was determined by history matching of a lab experiment was not successful. One magnitude that had to be focused on was simulated pressure. Simulated injection pressure and resulting average pressure during polymer injection was not in agreement with field observations, probably because the simulators do not include all the complex mechanisms of polymer flood. Hence maintaining reservoir pressure within a reasonable range during simulation turned out to be a challenge, especially for the black oil simulations. In conclusion it was found that with the appropriate approach, robust and reliable results could be obtained at scales less than some critical value, consistent with similar results obtained for pure water flood simulations. Compositional simulations were apparently more reliably scaled than black oil simulations. An important finding was that results were extremely sensitive to correct and consistent definition of fluid descriptions, and users must be aware of different approach angles for compositional and black oil simulators in this respect.