On the Inference of Gas Diffusion Coefficient in Organic Matter of Shale Gas Reservoirs

Abstract

Abstract Gas production from shale-gas reservoirs constitutes the largest portion of total gas production. The US shale reservoirs are tight and inherently heterogeneous with abundant presence of kerogenic material. Modeling fluid flow in shale reservoirs is complex and still an active field of research. The complexity arises from different flow physics such as pressure flow and diffusion. Many of the field performance forecasts constantly underestimate production from these reservoirs because most of the current models ignore important governing physics. This study provides new insights on diffusion in organic matter, in an effort to correct a main source of underestimation of gas production in shale gas models. In an earlier study, we developed for the first time a detailed diffusion model and showed how pore size distribution and specific surface area of the pores in organic matter can significantly influence gas production. An important parameter controlling the rate of gas release is the diffusion coefficient of gas diffusing into organic matter which appears in the flow equations. One of the methods of estimating the diffusion coefficient is based on analysis of gas uptake into shale samples in a closed chamber. The coefficient is extracted by comparing experimental observations to the solution of diffusion equation in the domain of pore/kerogen interface. If the mathematical representation of the organic matter is inaccurate, the diffusion coefficient will be inaccurate as well, regardless of lab-measurements accuracy. The values reported in the literature are based on the slab-shaped mathematical representation of organic matter, assuming a single scale for diffusion characteristic length. In this study, we implement a multi-scale diffusion model to estimate gas diffusion coefficient in organic matter. The previously reported evaluations are on the order of 10−20 m2/s. Reanalysis of the same set of experimental data using our detailed model suggests the interpretation of the coefficient is largely dependent on the diffusion-length scales being considered. We show that diffusion occurs over multiple time scales and the coefficient could be as much as four orders of magnitude higher than reported. The developed diffusion model is a robust and practical mathematical model and can be implemented in reservoir simulators. The findings of this study shed some light on why production forecasts constantly underestimate gas production from shale gas reservoirs.