Carbon sequestration in deep, unmineable coal seams is a viable strategy for carbon reduction. However, the impact of CO₂ on coal mechanical performance poses safety concerns for a reservoir. This study proposes an integrated learning methodology that leverages experimental data involving CO₂ immersion in various phases to evaluate the mechanical performance of coal seams during carbon sequestration. The approach integrates support vector regression (SVR) through the bagging method and employs a novel algorithm to optimize SVR. The model systematically assesses seven key factors, including coal rank, sample size, saturation medium, saturation time, saturation pressure, saturation temperature, and loading rate, to understand their influence on mechanical performance. The study identified saturation temperature, coal rank, and the saturated medium as pivotal elements affecting coal seam weakening. Evaluation metrics such as squared correlation coefficient (R²), mean absolute error, and root mean square error were employed for performance comparison between the polynomial model and the integrated model. The results demonstrate the superior performance of the integrated model, with R² of 0.98, emphasizing its effectiveness in predicting coal seam strength weakening during carbon sequestration. These insights contribute to safety assessment of coalbed carbon sequestration practices.

在深层、不可开采的煤层中固碳是一种可行的碳减排策略。然而，CO ₂对煤机械性能的影响给水库带来了安全隐患。本研究提出了一种综合学习方法，利用涉及各个阶段的CO ₂浸泡的实验数据来评估碳封存过程中煤层的机械性能。该方法通过 bagging 方法集成支持向量回归（SVR），并采用一种新颖的算法来优化 SVR。该模型系统评估了煤阶、样本量、饱和介质、饱和时间、饱和压力、饱和温度和加载速率等七个关键因素，以了解它们对力学性能的影响。研究确定饱和温度、煤阶和饱和介质是影响煤层弱化的关键因素。采用平方相关系数（R ²）、平均绝对误差和均方根误差等评价指标来对多项式模型和集成模型进行性能比较。结果证明了集成模型的优越性能，R ²为0.98，强调了其在预测固碳过程中煤层强度减弱的有效性。这些见解有助于煤层碳封存实践的安全评估。