| The injection of gas, such as carbon dioxide, into oil reservoirs can be used as a means of enhanced recovery. If the injection pressure is sufficiently high, then the local displacement efficiency can also be high. That is, where gas contacts resident oil, nearly all of the oil can be displaced towards producing wells. However, due to the heterogeneity of reservoirs and the low viscosity of gas, the fraction of the reservoir contacted by gas may not be high. The performance of gas injection processes depends on this balance between local displacement efficiency and global sweep efficiency and both should be accurately predicted by numerical simulations.; This dissertation develops numerical techniques that improve our ability to simulate gas injection processes at the field scale. A multicomponent or compositional fluid description is required in order to model the transfer of components between phases that determines local displacement efficiency. Even in one dimension (1D), the simulation of gas/oil displacements is challenging due to the strong, nonlinear coupling of the component advection equations and the weak hyperbolicity of the overall transport system. We show that the component-wise extension of a third order, upwind essentially non-oscillatory finite difference scheme outperforms the first order upwind scheme and total variation diminishing schemes.; In multiple dimensions we use a streamline-based simulation method to extend our 1D results. The accuracy of streamline-based methods depends not only on the 1D solver used to move fluids along streamlines, but also on the techniques used to map compositions between the background grid and the set of streamlines used for advection. We present a second order method to map compositions to streamlines that uses a piecewise linear reconstruction of compositions on the background grid. We also suggest using an unstructured interpolation algorithm based on block Kriging to map compositions from streamlines to the background grid. The use of Kriging allows for greater flexibility and adaptivity in forming the streamline grid relative to standard approaches. Several examples demonstrate the ability of our method to predict the global sweep of gas injection processes. |
