Reservoir studies, including preparation of field development plan, are processes typically dominated by time constraints. In general, reservoir studies consist in multiple geoscience activities integrated to build a fine geological model that eventually leads to an upscaled numerical model suitable for history matching and forecast simulations. In the simulation stage, the quality and effectiveness of the activity is highly dependent on the computational efficiency of the numerical model. This is particularly true for complex, supergiant carbonate reservoirs. Often, even with today's simulators, upscaling is still needed and simplifications can be implemented to allow computationally intensive history matching and risk analysis workflows. This paper provides some real field examples where these issues were faced and successfully managed, without applying further simplifications to the geological concept of the model: principles of reservoir simulations and common sense reservoir engineering were used to adjust properties of the model and then speed-up numerical simulation. Tuning included a combination of various solutions, such as deactivating critical cells whenever possible, calibrating convergence and time stepping control, tweaking field management to prevent instability in the computation, optimization of number of cores and cells split among cores to optimize load balancing and scalability.
These solutions were used on two super-giant carbonate fields, a triple porosity (matrix, karst and fractures) undersaturated light oil reservoir and a supercritical gas and condensate reservoir. The former field was described using an upscaled model of about 700,000 active cells and a dual porosity - dual permeability formulation; the latter was described by a relatively coarse model of about 400 thousand active cells using a single porosity formulation. Large speed-up, up to five times with respect to reference simulations, was achieved without simplifying the geology and losing accuracy perceivably. Benefits were achieved for both conventional and high-resolution simulators.