基于先验模型约束的最小二乘逆时偏移方法  

Regularized least-squares reverse time migration with prior model

作　　者：李振春[1] 李闯 黄建平[1] 王蓉蓉[2] 

机构地区：[1]中国石油大学(华东)地球科学与技术学院,山东青岛266580 [2]中国石油大学(华东)信息与控制工程学院,山东青岛266580

出　　处：《石油地球物理勘探》2016年第4期738-744,5,共7页Oil Geophysical Prospecting

基　　金：国家"973"课题(2014CB239006;2011CB202402);国家自然科学基金(41104069;41274124);中国石化地球物理重点实验室开放基金(33550006-15-FW2099-0033);中央高校基本科研业务费专项资金(16CX06046A)联合资助

摘　　要：由于反演问题的不适定性,最小二乘逆时偏移(LSRTM)收敛速度缓慢,甚至陷入局部极值,无法收敛。另一方面,由于观测系统及地层吸收衰减的影响,往往造成地下照明盲区,LSRTM无法恢复照明盲区的构造。为此,通过测井数据构建先验反射系数模型,将其作为正则化项加入LSRTM约束反演过程,发展了基于先验模型约束的最小二乘逆时偏移算法(RLSRTM),并通过动态的正则化参数及对正则化项的预处理改善了约束效果。在实现算法的基础上对稀疏Marmousi模型进行了成像测试,计算结果表明:(1)常规LSRTM能够压制逆时偏移(RTM)结果中的成像噪声,补偿深部反射同相轴能量,但是对于照明不均匀现象的补偿效果有限,并且照明盲区的构造信息无法恢复;(2)正则化项预处理RLSRTM能够进一步恢复LSRTM照明补偿不足区域的构造信息,尤其对深部的背斜构造边界及其他地层刻画得更加清楚,也可以恢复部分照明盲区的构造信息;(3)无正则化项预处理RLSRTM的部分构造边界模糊,甚至出现假构造。因此正则化项梯度预处理RLSRTM能够加快收敛速度,改善弱照明及照明盲区的成像分辨率和保幅性,可以保证反演结果稳定,防止在反射率模型中引入极值,相比LSRTM,RLSRTM对低信噪比炮记录具有更好的适应性。Least-squares reverse time migration (LSRTM) converges slowly, and sometimes drops into local extremum because of the ill-posedness of the inversion problem. On the other hand, the influence of irregular geometry and the absorption of underground layers result in some blank areas of illumination, in which the structures cannot be imaged by LSRTM. To solve these problems, the paper presents the theory of regularized least-squares reverse time migration (RLSRTM) with prior model. The prior model is constructed from logging data and incorporated into LSRTM as the regularization constraint. And dynamic regularization parameters and preconditioned regularization-term gradients are proposed to ensure better constraint. Based on numerical tests on a sparse Marmousi model, the following observation are obtained: LSRTM can suppress the migration artifacts and compensate energy in deep part compared with reverse time migration (RTM), but the compensation to uneven illumination is limited and the structures in blank illumination areas cannot be imaged; RLSRTM with preconditioned regularization-term gradient can further compensate energy of structures with poor illumination, produce more clear images of the boundary of anticlines and other layers in deep part, and even recover some information in blank illumination areas; RLSRTM without preconditioned regularization-term gradient produces images with some blurry boundaries and false structures. Therefore, RLSRTM with preconditioned regularization-term gradient can accelerate the convergence, improve the resolution and amplitude preservation of the images, ensure the stability of the inversion, and has reduced sensitivity to low signal-to-noise ratio shot data compared with LSRTM and RLSRTM without preconditioned regularization-term gradient. © 2016, Editorial Department OIL GEOPHYSICAL PROSPECTING. All right reserved.

关 键 词：最小二乘逆时偏移 照明 先验模型 约束 正则化项 

分 类 号：P631[天文地球—地质矿产勘探]