Improved numerical simulation of non-thermal enhanced heavy oil recovery

The dependence on unconventional resources such as heavy oil is on the rise due to geometric increase in demand for energy and the decline of production from mature conventional oil reservoirs. Heavy oil reservoirs contain oil that has some limited mobility under reservoir conditions and only a small fraction of the oil-in-place can be recovered by primary technique which involve harnessing the internal reservoir energy. The remaining oil after the primary depletion is still mostly continuous and present a valuable target for enhanced recovery. However, most of these reservoirs are relatively thin, making them poor candidates for thermal methods, in addition to associated high energy requirement and adverse environmental effects of the heating process. Therefore, any incremental oil recovery must be through non-thermal methods, such as waterflooding, chemical and gas injection. These methods however suffer from adverse mobility ratio which significantly affect the efficiency of the displacement process. The simulation of these processes for the purpose of reservoir prediction and performance is a herculean task due to the complex physics of instability and compositional effect taking place that is not fully understood. In this thesis, the results of improved numerical simulation techniques of non-thermal heavy oil recovery were presented, demonstrating the viability of the techniques as simulation methods heavy oil non-thermal enhanced heavy oil recovery (EHOR). Several displacement mechanisms were identified through the simulation of the secondary and tertiary processes that contributed to significant incremental heavy oil recovery. A systematic lumping scheme of the heavy oil components into pseudo-components based on the behaviour of the produced oil was proposed. A new methodology for the estimation of relative permeability from displacement with instability and compositional effect using a two-dimensional (2D), high-resolution model to effectively capture the finger, and a versatile, three-parameter function (L.E.T correlation) was demonstrated. A semianalytical approach through a combination of theoretical and an empirical prediction method based on the famous works of Koval, and Todd and Longstaff on viscous fingering was employed for the verification of the estimated relative permeability. Lastly, a multiscale approach to history matching, for the estimation of unstable relative permeability that is computationally more efficient, was proposed. It involves the history matching of a set of coarse grid models to predict the fine-scale relative permeability. In this approach, fine-scale information was resolved without direct solution of the global fine-scale problem. The results showed that the time required to estimate relative permeability using the multiscale approach was only about 35% required to estimate the same relative permeability using a single high-resolution model. The memory requirement for the approach was also about 50% required for simulation of the single high-resolution model. Therefore, the lower memory size and computations required in the multiscale approach mean that a less powerful computer can be used to estimate the relative permeability curves for unstable displacements with accuracy similar to that obtained using a high-resolution model approach