地震分频迭代反演在薄层河道砂体预测中的应用  

Application of seismic frequency-divided iterative inversion in the prediction of thinly laminated channel sand bodies

作　　者：任宪军[1] 李钟 马应龙 董萍 田行达 REN Xian-Jun;LI Zhong;MA Ying-Long;DONG Ping;TIAN Xing-Da(Exploration and Development Research Institute,Sinopec Northeast Oil and Gas Company,Changchun 130062,China;Data Processing Center(Zhanjiang),Research Institute,Geophysical Branch,China Oilfield Services Limited,CNOOC,Zhanjiang 524057,China;Research Institute of Exploration Development,PetroChina Tarim Oilfield Company,Kuerle 841000,China;Engineering Technology Research Institute,No.3 Oil Production Plant of Huabei Oilfileld Company,Cangzhou 062450,China;Changjiang Geophysical Exploration&Testing Co.,Ltd.,Wuhan 430010,China)

机构地区：[1]中国石化东北油气分公司勘探开发研究院,吉林长春130062 [2]中海油田服务公司物探事业部研究院特普数据中心,广东湛江524057 [3]中石油塔里木油田勘探开发研究院,新疆库尔勒841000 [4]华北油田公司第三采油厂工程技术研究所,河北沧州062450 [5]长江地球物理探测(武汉)有限公司,湖北武汉430010

出　　处：《物探与化探》2023年第2期420-428,共9页Geophysical and Geochemical Exploration

基　　金：国家自然科学基金项目“裂缝性储层地震定量预测及流体识别方法研究”(41974124)。

摘　　要：龙凤山地区河道砂储层符合典型岩性油气藏特征,其砂体厚度薄、河道窄、岩性纵横向非均质性强,对5 m以下储层预测难度大。分频迭代反演充分利用全频段地震资料,对不同频段、尺度的地震信息逐级传递,优化反演结果。文中具体利用匹配追踪算法实现地震信号频段划分得到不同尺度地震数据体,在测井约束下,以低频大尺度的反演结果作为下一级频段反演的初始模型,调整反演结果;在反演过程中,通过相关算法自适应选取子波,增强反演准确性,基于贝叶斯理论自适应选取正则化参数,调节分辨率和稳定性关系达到最佳平衡,避免反演出现混沌现象。2019年该区新钻四口井钻遇营城组1-2-6和1-2-8小层的气层,同反演预测结果吻合。证明本文方法相较于常规分频反演而言,具有反演精度高、忠实于地震信息、频段应用充分的优点,可以有效提高薄层河道砂的识别能力,指导相关岩性油气藏的勘探开发。The channel sand reservoirs in the Longfengshan area have the characteristics of typical lithologic reservoirs.This area has thin sand bodies,narrow channels,and strong vertical and horizontal lithologic heterogeneity.It is difficult to predict the reservoirs at a depth of 5 m or greater.The frequency-divided iterative inversion can fully utilize the full-frequency band seismic data and transmit the seismic information of different frequency bands and scales step by step,thus optimizing the inversion results.In this study,the seismic signal frequency bands were divided using the matching pursuit algorithm to obtain seismic data volumes of different scales.Under the constraints of log data,the low-frequency,large-scale inversion results were used as the initial model for the next-order frequency band inversion,and the inversion results.During the inversion,wavelets were adaptively selected using the correlation algorithm to enhance the inversion accuracy.Regularization parameters were adaptively selected based on the Bayesian theory to adjust the relationship between resolution and stability to achieve the optimal balance and avoid chaos in inversion.In 2019,gas reservoirs in subzones 1-2-6 and 1-2-8 of the Yingcheng Formation were encountered in the drilling of four wells in the Longfengshan area.This result is consistent with the inversion prediction results.Therefore,compared with conventional frequency division inversion,the method proposed in this study has the advantages of high inversion accuracy,coincidence with seismic information,and full application of frequency bands.This method can effectively improve the identification performance of thinly laminated channel sand bodies and guide the exploration and development of related lithologic reservoirs.

关 键 词：薄层河道砂 匹配追踪 分频迭代反演 

分 类 号：P631.4[天文地球—地质矿产勘探]