遗传和模拟退火混合寻优一步法三维CRS叠加  

An improved one-step 3D CRS stacking method based on hybrid optimization algorithm

作　　者：孙小东[1,2] 侯孟睿 任丽娟 王伟奇 李振春[1] SUN Xiaodong;HOU Mengrui;REN Lijuan;WANG Weiqi;LI Zhenchun(Key Laboratory of Deep Oil&gas,China University of Petroleum(East China),Qingdao,Shandong 266580,China;Shandong Provincial Key Laboratory of Reservoir Geology,China University of Petroleum(East China),Qingdao,Shandong 266580,China;Nanhai Western Petroleum Research Institute,Zhanjiang Branch,CNOOC,Zhanjiang,Guangdong 524000,China)

机构地区：[1]中国石油大学(华东)深层油气重点实验室,山东青岛266580 [2]中国石油大学(华东)山东省油藏地质重点实验室,山东青岛266580 [3]中海石油(中国)有限公司湛江分公司南海西部石油研究院,广东湛江524000

出　　处：《石油地球物理勘探》2021年第2期283-288,I0009,共7页Oil Geophysical Prospecting

基　　金：国家自然科学基金项目“基于局部角度域扩展模型的非线性地震反演方法”(41574098)、“基于三角网格剖分模型的高斯波包立体层析成像”(41630964);中央高校基本科研业务费专项资金项目“深层粘弹介质矢量波联合逆时偏移方法研究”(18CX02059A)、深层油气重点实验室开发基金项目“三维共反射面元叠加提高深层数据信噪比”(20CX02111A);中国石化地球物理重点实验室开放基金项目“低信噪比数据三维共反射面元叠加”(wtyjy-wx2018-01-07);山东省自然科学基金面上项目“深层粘滞性介质弹性波逆时偏移”(ZR2020MD048);中石油重大科技项目“塔里木盆地深层复杂高陡构造与碳酸盐岩储层地震速度建模及成像关键技术”(ZD2019-183-003)联合资助。

摘　　要：共反射面元(Common Reflection Surface, CRS)叠加充分利用了菲涅耳带半径内的地震信息,在不降低分辨率的前提下最大程度地提高了信噪比,是低信噪比地震资料成像的有力手段。同时,CRS叠加考虑了地下反射层的倾角和局部曲率,具有更高的成像精度。常规的三维CRS叠加,多步依次求取的八参数精度较低,影响了最终叠加效果。为此,提出了一步法三维CRS叠加的实现策略。结合遗传算法快速搜索的优势和模拟退火算法的全局收敛特性,采用两种算法混合的多种群分层式并行寻优算法,即顶层采用遗传算法中的热槽法产生初始种群,中层执行并行的遗传迭代算法实现种群进化,底层采用模拟退火算法实现全局寻优。该CRS叠加方法及混合算法的设计方案明显降低了参数的优化成本并提高了精度。实际资料的测试验证了混合寻优一步法三维CRS叠加的实用性。Common reflection surface(CRS)stacking makes full use of seismic data within the range of a Fresnel radius,and maximizes the signal-to-noise ratio without reducing the resolution.It is a powerful imaging means of the seismic data with low signal-to-noise ratio.At the same time,CRS stacking takes into account the inclination and local curvature of the underground reflector,so the imaging accuracy is higher.In conventional three-dimensional CRS stacking,the accuracy of the eight parameters obtained sequentially in multiple steps is low,and affects the final stacking effect.We propose a one-step three-dimensional CRS stacking strategy which combines the rapid search of genetic algorithm and the global convergence of simulated annealing algorithm.It uses a multi-group hierarchical hybrid parallel optimization algorithm that mixes the two algorithms,that is,a thermal slot method is used to generate initial population in the top layer,the middle layer executes a parallel genetic iterative algorithm to achieve population evolution,and the bottom layer uses a simulated annealing algorithm to achieve global optimization,and the genetic and simulated annealing hybrid algorithm is used for hierarchical parallel computing.The CRS stacking method and the design of the hybrid algorithm significantly improve the optimization cost and accuracy of parameters.Tests on real data have verified the practicability of the onestep 3D CRS stacking method.

关 键 词：共反射面元叠加 参数耦合 遗传算法 模拟退火 参数寻优 

分 类 号：P631[天文地球—地质矿产勘探]