【作者】 李浩；

【导师】 郭振威；

【作者基本信息】 中南大学 ， 地球探测与信息技术， 2023， 硕士

【摘要】 海洋可控源电磁法是近二十年来发展最为迅速的海洋地球物理勘探方法,因其成本低、便捷、高效及对地下高阻体敏感等特性被广泛应用于海洋油气资源的勘探与储层评价中。相较于重磁勘探等方法,海洋可控源电磁法拥有仅次于海洋地震勘探的成像分辨率,同时对于地震勘探成像中不易区分的目标体,海洋可控源电磁法可以起到补充解释作用。海洋可控源电磁反演具有强烈的不适定性,因此正则化技术被引入增强反演结果的稳定性。然而,传统的吉洪诺夫正则化、聚焦反演、全变差正则化等正则化技术很难兼顾反演的稳定性与准确度,而开展联合反演又对计算和存储能力提出了更高的要求。因此本文提出地震图像结构张量引导的海洋可控源电磁反演方法,在不涉及地震数据及正演建模的前提下,直接利用先验地震图像的特征信息改进正则化约束,提高单一电磁数据的成像精度。本文通过数值模拟实验证明了:(1)图像引导反演可显著提升传统吉洪诺夫正则化反演和聚焦反演的成像分辨率与准确度;(2)强化网格v方向的模型惩罚可以保证成像目标的一致性与连续性,而放松网格u方向的模型惩罚可以拓展了模型梯度的搜索空间,有利于锐化边界;(3)如地震偏移图像、速度图像等可以反映地下地质结构的模型参数或图像的结果都可以作为引导图像,且图像引导反演的计算消耗与传统的单一海洋可控源电磁数据反演没有明显差异。本文提出的图像引导反演的本质是利用先验图像改进非结构化网格下的模型梯度的计算方式,故该方法适用于所有含模型梯度的正则化约束反演算法。同时,因其不涉及第二种勘探数据及其正向建模,在多地球物理数据的正则化反演中有良好的应用前景。图71幅,表4个,参考文献182篇更多还原

【Abstract】 Marine controlled-source electromagnetic(CSEM)surveying is the most rapidly developed marine geophysical exploration technique in the past two decades.It is widely used in the exploration and reservoir evaluation of marine oil and gas resources for its low cost,convenience,efficiency,and sensitivity to high electrical resistivity.Compared with gravity and magnetic exploration methods,the marine CSEM method has an imaging resolution second only to marine seismic exploration.It can provide complementary interpretations for targets that are difficult to distinguish in seismic imaging.The marine CSEM inversion is strongly ill-posed,so regularization techniques are introduced to enhance the stability of the inversion results.However,the traditional regularization techniques such as Tikhonov regularization,focused inversion,and total variation regularization are difficult to balance the stability and accuracy of the inversion results.While joint inversion requires higher computational cost and memory.Therefore,a novel marine CSEM regularization inversion method guided by the structure tensor of seismic image is proposed to improve the resolution and accuracy of individual marine CSEM data inversion.The new approach directly utilizes the characteristic information of the prior seismic image to modify the regularization constraints,which doesn’t involve the real seismic data or forward modeling in the inversion process.This paper demonstrates through numerical simulations that:(1)image-guided inversion could improve the imaging resolution and accuracy of traditional Tikhonov regularization inversion and focusing inversion;(2)strengthening the model penalty in direction of the cells could promise the consistency and continuity of the imaging targets,while relaxing the model penalty in direction of the cells expands the search space of the model gradient,which is beneficial for yielding the clear boundaries;(3)model parameters or images that could reflect underground geological structures,such as seismic migration images and velocity images,can be used as guiding images.The computational cost of the image-guided inversion is not notably different from that of the conventional individual marine CSEM data inversion.The essence of the image-guided inversion proposed in this paper is to use prior images to adjust the calculation of the model gradient based on the unstructured mesh.Therefore,this approach could be applied in all regularization inversion methods that involve the constraints of the model gradient.Furthermore,it exhibits promising potential in the regularization inversion of multiple geophysical datasets since it does not call for the use of the second type of exploration data or forward modeling.更多还原