Abstract: In view of China’s coal-based energy structure, CO₂ storage in coal seam is an important way to achieve negative carbon technology in China. Seismic exploration is an effective means to monitor CO₂ migration. However, the time-lapse seismic response usually shows the comprehensive change of multiple parameters (porosity, pore pressure, and fluid saturation) and is difficult to distinguish. To solve this problem, a simultaneous prediction method of multiple parameters based on deep learning was proposed. Firstly, the fluid replacement theory for CO₂ storage in coal seam was studied, and the calculation method of P- and S-wave velocities suitable for CO₂ storage in coal seam was obtained. Secondly, the pre-stack seismic responses before and after CO₂ injection were analyzed, and the contribution of CO₂ saturation, porosity, and pore pressure to the responses of different offset data was verified. Finally, a network architecture for simultaneous prediction of multiple parameters was built, and the method of obtaining network training data was described. The test results of synthetic and field data show that the network has good generalization ability and can obtain high-precision prediction results of reservoir multi-parameters. Compared with the traditional seismic inversion, the proposed method effectively improves the computational efficiency and provides an effective means for CO₂ geological storage monitoring.