Experimental Investigation of Gas Diffusion in Coal – Comparison Between Crushed and Intact Core Samples

Abstract

Abstract To forecast gas production from coal and other dual-porosity reservoirs, the reservoir specific diffusion properties must be estimated. Diffusion properties can be determined in the laboratory as part of adsorption or desorption isotherm measurements. Typically, such experiments are performed on crushed coal rather than intact core samples, which can reduce the experimental run time from months or weeks to days. We investigated the relationship between crushed coal and intact core samples by performing two sets of diffusion experiments on two Bowen Basin coals of different rank, Coal#1 (low volatile bituminous) and Coal#2 (medium volatile bituminous). Core samples were analysed in the Triaxial Multicomponent Gas Rig in CSIRO's Unconventional Gas Laboratory. Measurements on crushed coal with a particle size range of 0.6 – 1 mm and 0.5 – 1 mm prepared from off-cuts of Coal#1 and Coal#2 respectively were performed in CSIRO's Isotherm Rig. The diffusion kinetics of the two coals were measured with respect to three gases: methane, carbon dioxide, and nitrogen. These gases are commonly present in coal seams and are associated with primary and enhanced coal seam gas production. The observations from the adsorption and desorption experiments on crushed and intact core sample highlight significant differences in diffusion kinetics between the different coal forms. For the crushed coals, the diffusion process was significantly accelerated and the effective diffusivities were typically an order of magnitude higher than for the core samples. The diffusion kinetics on crushed and core sample may be relatable through the diffusion path length, which requires further investigation. Both crushed and core samples adsorb CO2 preferentially. The differences between the slower diffusing CH4 and N2 were small, though we found that the sorption preference towards CH4 and N2 was different for the core and the crushed coal samples. The observed difference could be caused by sample heterogeneity, as the crushed coal is prepared from off-cuts rather than the coal cores themselves. The observed diffusion behaviour can be fitted by a bidisperse model. Two fractions, a ‘fast’ and a ‘slow’ fraction, were necessary to match the experimental data. While Coal#2 is characterised by a partitioning of approximately 70% ‘fast’ fraction and 30% ‘slow’ fraction, for Coal#1 the fractions were not uniform across the different gases and sample forms.