Abstract The uncertainty of reservoir and operating parameters challenges the accuracy of risk assessment, as well as the efficiency of optimization in carbon capture and storage (CCS) operation. To quantitatively analyze the role of uncertainty parameters on the response of CO2 injection, the effects of geomechanical and hydrogeological parameters on CCS are investigated using the approaches of distance correlation and machine learning support vector regression (SVR). In addition, a risk factor is introduced as a combination of brittleness and stress increment to assess the potential risk of caprock integrity. Using quantitative analysis, the order of importance of the parameters on the fluid pressure in the reservoir and the caprock, and the formation deformation at the ground surface and the bottom and top of the caprock are obtained. Compared to formation deformation, the pressure change can provide more valuable information regarding the assessment of the integrity of the caprock. The trained SVR surrogate model based on SVR can predict both the pressure change as well as formation deformation with reliable accuracy.

中文翻译：

---

基于距离相关和支持向量回归的 CO2 封存参数不确定性分析

摘要 储层和操作参数的不确定性对风险评估的准确性以及碳捕集与封存（CCS）操作的优化效率提出了挑战。为了定量分析不确定性参数对 CO2 注入响应的作用，使用距离相关和机器学习支持向量回归 (SVR) 方法研究地质力学和水文地质参数对 CCS 的影响。此外，引入风险因子作为脆性和应力增量的组合来评估盖层完整性的潜在风险。通过定量分析，得到各参数对储层和盖层流体压力、地表和盖层底部和顶部地层变形的重要性大小顺序。与地层变形相比，压力变化可以为评估盖层完整性提供更有价值的信息。训练好的基于 SVR 的 SVR 代理模型能够以可靠的精度预测压力变化和地层变形。