Improved upscaling scheme for steam assisted gravity drainage (SAGD) and semi-analytical modeling of the SAGD rising phase

Steam assisted gravity drainage (SAGD) process commonly applied for heavy oil and bitumen recovery consists of two production phases: a steam rising phase and a spreading phase. Extensive research has been done on modeling the SAGD spreading phase, but fewer analytical/semi-analytical models exist for the unstable rising phase. This thesis presents a semi-analytical method, MS-SAGD, to model the SAGD rising phase. In addition, an improved upscaling technique that takes into account the unique flow geometry observed during SAGD is presented that enables more accurate predictions of oil recovery rates in heterogeneous reservoirs during both phases. The MS-SAGD semi analytical method, based on the Myhill and Stegemeier frontal advance model for steam drive processes, tracks the growth of the steam chamber as a function of time. Two different oil production rate models are proposed and the comparison of results from flow and transport simulations is presented. Model 1 is similar to Butler’s approach using the rising steam finger theory. Model 2 is obtained by modifying Butler's spreading phase model and applying it to the rising phase. Both models use the outputs of the MS-SAGD model to estimate the oil production rates during the SAGD rising phase. The application of the MS-SAGD model is extended to heterogeneous reservoirs by treating the heated volume estimated by the original MS-SAGD model as an effective heated volume. In addition, the homogeneous permeability in the proposed oil production rate model is replaced with an upscaled effective permeability that is a function of time. The improved upscaling technique is based on a global approach that minimizes the differences between the fine scale and upscaled model pressure solutions. Sources and sinks by means of wells are used in the upscaling to simulate the convergent flow pattern observed during the SAGD process. The proposed models outperform existing analytical/semi-analytical methods and are in good agreement with the results obtained from CMG-STARS reservoir simulation. Both oil production rate models perform comparatively well, producing similar results in terms of cumulative oil production. However, Model 2 performs better than Model 1 in describing the overall behavior of the oil production observed in the reservoir simulation and is thus a better model for the SAGD rising phase.