Arafath Md Yeasin, Haque Akm Eahsanul, Siddiqui Numair Ahmed, Venkateshwaran B, Ali Sohag
Department of Geoscience, Universiti Teknologi PETRONAS, Bandar Seri Iskandar 32610, Perak, Malaysia.
Centre for Subsurface Imaging, Institute of Hydrocarbon Recovery, Universiti Teknologi PETRONAS, Seri Iskandar 32610, Malaysia.
ACS Omega. 2025 Feb 5;10(6):5430-5448. doi: 10.1021/acsomega.4c07242. eCollection 2025 Feb 18.
Establishing a potential site characterization for carbon dioxide (CO) storage in geological formations anticipates the appropriate reservoir properties, such as porosity, permeability, and so forth. Well logs and seismic data were utilized to determine key reservoir properties, including volume of shale, porosity, permeability, and water saturation. These properties were cross validated with core data sets to ensure accuracy. To enhance permeability estimation, sophisticated machine learning (ML) methods were employed, categorizing permeability into five classes ranging from extremely good (0) to very low (4). Two ML models, Naïve Bayes (NB) and multilayer perceptron (MLP), were applied to predict permeability. The MLP model outperformed the NB model, achieving 99% training accuracy and 93% testing accuracy, compared to 78 and 73%, respectively, for the NB model. The resulting comprehensive permeability model revealed the distribution across three stratigraphic layers: the B100 zone exhibited extremely low permeability, suitable as a caprock, while the D35-1 and D35-2 zones demonstrated excellent permeability, indicating potential as CO storage reservoirs. The "X" field reservoir, located at depths exceeding 1300 m, meets the depth requirements (1000-1500 m) for CO storage. Our integrated approach, combining empirical and ML-based calculations with core data and well logs, proved effective in characterizing the reservoir. The lithological model defined nonreservoir sections between the clay and silt lines, identifying important caprocks and interbedded shale/clay intervals. Seismic profiling confirmed the B100 zone as a continuous caprock overlying the D group reservoir zone, crucial for preventing upward CO migration. This comprehensive analysis supports the potential of the "X" field in the Malay Basin as a viable site for CO storage, contributing to the ongoing efforts in carbon capture and storage research.
建立地质构造中二氧化碳（CO₂）储存的潜在场地特征需要预测合适的储层特性，如孔隙度、渗透率等。利用测井曲线和地震数据来确定关键储层特性，包括页岩体积、孔隙度、渗透率和含水饱和度。这些特性与岩心数据集进行交叉验证以确保准确性。为了提高渗透率估计，采用了复杂的机器学习（ML）方法，将渗透率分为从极好（0）到极低（4）的五个等级。应用了两种ML模型，朴素贝叶斯（NB）和多层感知器（MLP）来预测渗透率。MLP模型优于NB模型，训练准确率达到99%，测试准确率达到93%，而NB模型的训练和测试准确率分别为78%和73%。由此产生的综合渗透率模型揭示了三个地层的分布情况：B100区渗透率极低，适合作为盖层，而D35 - 1和D35 - 2区渗透率极好，表明有作为CO₂储存储层的潜力。位于深度超过1300米的“X”油田储层满足CO₂储存的深度要求（1000 - 1500米）。我们将基于经验和ML的计算与岩心数据和测井曲线相结合的综合方法，在表征储层方面被证明是有效的。岩性模型定义了粘土和粉砂线之间的非储层段，识别出重要的盖层以及页岩/粘土互层间隔。地震剖面证实B100区是覆盖D组储层区的连续盖层，对于防止CO₂向上运移至关重要。这一综合分析支持了马来盆地“X”油田作为CO₂储存可行场地的潜力，为正在进行的碳捕获与储存研究做出了贡献。
