Abstract
Diffusion mixing is a dominant process in absence of convective mixing in various reservoir processes such as CO2 flooding of fractured reservoirs, heavy oil and bitumen recovery, solution gas drive processes, and gas re-dissolution process in depleted reservoir. In these processes, the diffusivity governs the rate and extent of mixing of light hydrocarbons/non-hydrocarbons with the oil that enhances the oil recovery through in-situ viscosities reduction. It is one of the key parameters for design and understanding of the displacement processes.
Due to its significance in various aspects of oil recovery processes, several experimental/theoretical studies are performed recently on the measurement of gas diffusivity in oils. Experimental work is most commonly based on the Pressure-decay concept due to its simplicity. The parameter estimation from these tests is based on the transient diffusion model, where the concentration/density gradient in gas phase is neglected or modelled with a transfer coefficient. However, these models are good for steady-state or late transient. Since the transfer coefficient concepts are validated only for steady-states (or near steady-state) in literature, its use in transient case must be investigated by analyzing the full diffusion model in both oil and gas phases. Therefore, in this article, the previous work [1] is extended and
the diffusion process in pressure-decay set-up is experimentally investigated
the transfer of gas from gas phase to the oil is captured properly through exact modeling based on the transient diffusion model in gas and oil phases coupled with continuity in state variable (using Henry's constant) and molar flux at the gas-oil interface;
the exact solution of detailed pressure-decay (transient diffusion) model is developed and, the early and late transient solutions are analyzed;
A robust inversion technique for parameter estimation is presented using exponentially decaying late transient data;
Most importantly, the inversion technique is based on a linear regression and numerical integration rather than a non-linear regression. This integral based linear representation avoids the multiple solutions and can be used with limited dataset and/or when noise in the experimental data is significant.
