获取高质量的地下流体样品对于二氧化碳地质利用与封存（CCUS）场地的CO₂泄漏监测具有重要意义。基于气体推动原理的地下流体取样器具有取样精度高、地层扰动小、易与井下监测技术集成从而实现地下流体原位连续监测等多方面优势，已广泛应用于多个CCUS场地中。然而，在实际应用中取样器存在取样量不稳定、受地下水水位波动影响大等问题。为解决上述关键问题，作者研发并搭建一套模拟井筒内地下流体取样可视化实验装置，通过开展不同液面高度下地下流体的取样实验，探究气体推动式取样器的取样过程及取样机制。实验结果显示：气体推动式取样技术在取样过程对井筒内液体无干扰，证明具有被动取样的特点，对环境影响程度低。取样量大小主要取决于井筒内液面高度和储流容器大小，可通过液面高度及管路尺寸精确计算得出理论取样量。不同液面高度下实际取样量与理论计算量基本一致，其误差随液面高度的增加而降低，最大误差小于6.6%。取样时间与平均取样速率主要取决于注气压力，并与注气压力呈现幂函数关系，拟合相关系数大于99%。上述实验结果将有助于进一步理解气体推动式取样器的取样机理，对于取样器结构优化以及指导现场取样效果具有重要意义，可进一步提高CCUS以及其他污染场地环境风险预警能力。

Obtaining underground fluid samples with high quality is important for CO₂ leakage monitoring in CO₂ geological utilization and storage (CCUS). The gas-operated underground fluid sampler has the advantages of high sampling accuracy, low formation disturbance, and in-situ monitoring through integrating downhole monitoring technologies, and has been widely applied in several CCUS applications. However, during practical application processes, sampling volume of the sampler is unstable and greatly affected by groundwater level. In order to solve these problems, a visualized experimental device to simulate subsurface fluid sampling in a wellbore was developed. A set of fluid sampling experiments at different liquid heights were carried out to explore sampling process and sampling mechanism of the gas-operated sampler. The results show that there is no disturbance to the liquid height during the sampling process, indicating the gas-operated sampling technology is characterized by passive sampling and low interference to external environment. The sampling volume depends on the size of the storage container and the liquid height in the wellbore. The sampling volume can be calculated accurately by the liquid height and the pipeline size. The measured sampling volumes under different liquid heights are basically consistent with the theoretical values. The error between the measured volume and calculated one decreases with the increasing liquid heights. The maximum error is less than 6.6%. Both the sampling time and average sampling rate mainly depend on the injection pressure and present a power function relationship with the injection pressure with a correlation coefficient of greater than 99%. The above experimental results can help to understand the sampling mechanism of gas-driven sampler and are of great significance to feedback and guide sampling performance in field by optimizing the structure. In addition, it can improve the ability of warning environmental risks in CCUS and contaminated sites.