With the rise of global temperatures, climate change has become one of the major challenges facing humanity. Carbon dioxide is one of the main gases causing the greenhouse effect, and an increase in its emissions directly leads to global warming. Therefore, the development of efficient CO 2 storage technology is one of the important strategies to mitigate climate change. By prediction of the amount of CO 2 buried, the CO 2 storage potential under different geological conditions can be assessed to optimize storage schemes and select suitable storage sites. In this study, 184 real CO 2 storage data points were collected. After processing the data and conducting feature analysis, reliable training and testing sets were constructed. Random forest and XGBoost machine learning models were selected to predict CO 2 storage capacity, and grid search along with K-fold cross-validation methods were used to optimize the model parameters, further enhancing algorithm performance. The results indicate that the CO 2 storage capacity is closely related to effective formation thickness and porosity, with correlation coefficients exceeding 50%. Compared to the random forest model, the XGBoost model achieved a regression coefficient of 0.95 on the test set, demonstrating high prediction accuracy and making it an effective tool for uncertainty assessment in carbon storage projects.