LAUSR

Abstract The pore pressure in the coal seam during gas injection affects the coalbed methane (CBM) recovery and increases the risk of coal mine disasters. However, the limitations of conventional experimental methods hinder the direct characterization of gas pressure evolution during CO2 or N2 injection. In this study, a large-scale coal matrix was used to conduct experiments on different gases, focusing on the time–space dependence of the gas pressure during gas injection and adsorption. Finally, the gas pressure was predicted using a decision tree model based on a light gradient boosting machine (LGBM). The results showed that in the early stage of gas injection, it is difficult to increase the gas pressure in the coal matrix. As the gas pressure increases, the injection of CO2 or N2 becomes increasingly difficult. In the later stage, a low-volume gas injection will lead to higher gas pressure. Because of the different adsorption capacity of CO2 or N2, the decay rate of N2 injection is considerably lower than CO2 injection. The direction parallel to the inlet end produces the highest pressure difference. The size of the CBM seam can affect the pore pressure. As the gas adsorption capacity decreases, the pressure difference in each direction gradually decreases. Further, the LGBM can efficiently predict the pressure changes during gas injection and determines the time–space dependence of the pressure for different gases. The proposed method based on the results and the LGBM model facilitates the understanding of gas migration and pore pressure changes during gas-induced CBM extraction.