Study on CO2 Injection-Enhanced Gas Extraction Based on COMSOL Multiphysics Numerical Simulation and SVM–MLP Machine Learning

As energy demand rises, coal seam gas extraction using CO2 injection is gaining attention for its ability to enhance coal seam permeability and promote gas desorption. This study combines a COMSOL Multiphysics numerical model with an SVM–MLP machine learning model to analyze how the coal seam permeability, gas pressure, and CO2 injection pressure affect gas extraction. The results indicate that coal seam permeability is the dominant factor, showing a strong correlation of 0.8171 with the amount of gas extraction. When permeability increases from 0.756 × 10–16 to 75.6 × 10–16 m2, the growth rate of gas extraction rises from −0.261 to 63.543%. Under conditions of high permeability, increasing the CO2 injection pressure significantly enhances gas flow and release. The SVM–MLP model demonstrates excellent prediction performance, achieving R2 values of 0.9996 in training and 0.9968 in prediction. This model effectively integrates the global optimization capability of SVM with the nonlinear fitting ability of MLP, providing precise predictions for gas extraction optimization and offering valuable insights for the practical application of the CO2 injection technology.