Permeability prediction of submarine shallow sediments is significantly important for finding hydrate-bearing sands with commercial potential. Permeability could be more sensitive to hydrate occurrence than traditional velocity parameters, because the rapid decrease of permeability is probably caused by hydrates. The elastic characteristics of the hydrate-bearing sediments differ from the consolidated oil-gas reservoir. In this paper, two elastic limitations that the shallow sediments would be somewhere in between are assumed: 1) sediments are extremely soft; 2) sediments are consolidated. Derivations show that the difference between the bulk modulus in these two assumptions depends on the pore compressibility, and the difference was defined as the pore compressibility indicator (PCI). It has been theoretically proven that PCI is closely related to permeability. Logging data analysis indicates that sediments with high permeability own small PCI and overall large porosity. At the same time, the occurrence of hydrate significantly increases PCI, and even low-saturation hydrate can cause a significant decrease in permeability, so a successful permeability prediction enhances detection and identification of hydrates. Then, the relationship between the seismic reflections and PCI is studied, and a two-state seismic inversion is proposed to obtain PCI. Finally, to predict permeability, a neural network is constructed and well trained using PCI and porosity as inputs and permeability as the output. Error analysis and sensitivity analysis are then carried out. Field applications show that the predicted permeability is consistent with the logging data, and channels for fluid migration such as coarse-grained sands own high permeability, while providing favorable conditions for hydrate formation. In this study, we propose a new seismic approach for permeability prediction in unconsolidated sediments, which is a scientific and effective attempt to predict permeability with seismic data. It can improve the accuracy of hydrate identification and provide basic data for the evaluation of production potential.

海底浅层沉积物的渗透率预测对于寻找具有商业潜力的含水合物砂岩具有重要意义。渗透率可能比传统的速度参数对水合物的生成更为敏感，因为渗透率的快速下降可能是由水合物引起的。含水合物沉积物的弹性特征不同于固结油气藏。在本文中，假设浅层沉积物介于两者之间的两个弹性限制：1）沉积物非常柔软；2) 沉积物被固结。推导表明，这两个假设下体积模量的差异取决于孔隙压缩率，差异被定义为孔隙压缩率指标（PCI）。理论上已经证明PCI与渗透率密切相关。测井资料分析表明，高渗透性沉积物的PCI小，总体孔隙度大。同时，水合物的出现显着增加了PCI，即使是低饱和度的水合物也会导致渗透率显着下降，因此成功的渗透率预测增强了水合物的检测和识别。然后，研究了地震反射与PCI之间的关系，提出了一种二态地震反演以获得PCI。最后，为了预测渗透率，使用PCI构建和训练了一个神经网络孔隙度作为输入，渗透率作为输出。然后进行误差分析和敏感性分析。现场应用表明，预测渗透率与测井数据一致，粗粒砂等流体运移通道具有较高的渗透率，同时为水合物形成提供了有利条件。在这项研究中，我们提出了一种新的地震方法来预测松散沉积物的渗透率，这是利用地震数据预测渗透率的科学有效尝试。可以提高水合物识别的准确性，为生产潜力评价提供基础数据。