Petrological and ultrasonic velocity changes of coals caused by thermal cycling of liquid carbon dioxide in coalbed methane recovery

Enhanced coalbed methane (CBM) recovery by injecting liquid CO2(LCO2) has been a promising research topic for decades. Although the phase-fracturing and blooding effect during injection have been studied, the non-isothermal impact of cryogenic fluids on the coal matrix, especially the effect of thermal cycling generated by the cyclic LCO2 injection process, has not been studied in detail. This paper reports the development of a high-pressure and low-temperature reaction system, which can simulate the effect of thermal cycling on three different coals of different ranks, and the use of an ultrasonic device to study the P-wave velocity (Vp) in different directions before and after the influencing thermal cycling. The results show that all the cores subjected to different thermal cycling fractured to diverse degrees due to their initial texture and rank, and more cycles and lower rank increased the final fracture results. Influenced by thermal cycling, two endpoints of the Vp confidence interval (confidence level of 95%) of the coals decreased, and the velocity range of confidence interval all increased, the change of which might be attribute to the rank and the companying thermal cycling effect, resulting in the changes of crack structure and discrete degree of the velocity. The velocity anisotropy coefficient (k) value of the coal increased as the cycles increased, indicating that thermal cycling alters the coal structure randomly by means of fatigue accumulation and grain bond deterioration. The Vp(z)/Vp(x) and Vp(z)/Vp(y) value changes indicated that the internal cracks were not generated uniformly in all directions, and the presence of cracks helped the transfer of heat flux, which caused some thermal stress concentration areas to occur, further facilitating crack propagation at the crack tips. Meanwhile, the Delta Vp% scatters of lignite, bituminite and anthracite cores were fitted as exponential, DoseResp and linear curves with cycles, respectively (R-2 > 0.94). The mean fractal dimension D of the coals all increased with increasing cycles, and were negatively correlated with coal rank and Delta Vp%, and positively correlated with Delta k. The results provide some useful theoretical guidance for field application.