河流相储层地震属性优选与融合方法综述  

Review on Optimization and Fusion of Seismic Attributes for Fluvial Reservoir Characterization

作　　者：岳大力[1,2] 李伟[1,2] 杜玉山[3] 胡光义[4] 王文枫 王武荣 王政 鲜本忠 Yue Dali;Li Wei;Du Yushan;Hu Guangyi;Wang Wenfeng;Wang Wurong;Wang Zheng;Xian Benzhong(State Key Laboratory of Petroleum Resources and Prospecting,China University of Petroleum,Beijing 102249,China;College of Geosciences,China University of Petroleum,Beijing 102249,China;Shengli Oil Field Company Ltd.,SINOPEC,Dongying 257015,China;CNOOC Research Institute Co.Ltd.,Beijing 100028,China;Kunlun Digital Technology Co.Ltd.,Beijing 100040,China)

机构地区：[1]中国石油大学(北京)油气资源与探测国家重点实验室,北京102249 [2]中国石油大学(北京)地球科学学院,北京102249 [3]中国石油化工股份有限公司胜利油田分公司,山东东营257015 [4]中海油研究总院有限责任公司,北京100028 [5]昆仑数智科技有限责任公司,北京100040

出　　处：《地球科学》2022年第11期3929-3943,共15页Earth Science

基　　金：国家自然科学基金项目(Nos.42272186,42202109);中国博士后基金项目(Nos.BX20220351,2022M713458)。

摘　　要：地震属性分析已广泛应用于河流相砂体预测并取得良好效果.地震属性分析技术主要包括属性提取、属性优选与属性融合,总结了河流相砂体预测中常见的地震属性提取方式、优选及融合方法,分析了由围岩干扰、地震分辨率限制导致的属性提取与分析误区,阐述了不同属性优选与融合方法的优缺点、适用条件与发展前景.总体而言,基于线性模型的地震属性融合提升效果较差,适用于少井区域;基于非线性模型的属性融合效果较好,但仅适用于钻井较多的地区,如油气开发阶段;无监督智能属性融合可应用于无井或少井区域,是未来无井或少井条件下属性融合的重要发展趋势之一.同时,重点阐述了新提出的分频属性融合与降低围岩干扰的属性融合方法.Seismic attribute analyses have been applied widely in hydrocarbon exploration and development of fluvial reservoirs,and obtained good results.The analysis procedure of seismic attributes mainly includes the extraction,the optimization and the fusion of attributes.In this paper it summarizes the main methods of attribute extraction,optimization and fusion,and evaluates their advantages,disadvantages and application conditions.Besides,the common misunderstanding genetically related to seismic resolution and interference of neighboring zone in attribute extraction is also analyzed.Generally,fusion methods of seismic attributes using linear models cannot significantly improve the results,and are suitable for areas with several or a few wells;fusion methods with intelligent models(mainly for supervised learning)are commonly suitable for the areas with dozens of wells,such as areas within oil development stage.Fusion methods based on unsupervised learning are suitable for areas with few and even no wells,which have an optimistic development prospect since they can make full use of the seismic information,and are not limited by wells.In addition,the new fusion methods of frequency-decomposed attributes,and of attributes from target and neighboring zones are also summarized in this paper.

关 键 词：河流相储层 地震属性 属性优选 属性融合 机器学习 沉积学 

分 类 号：P631[天文地球—地质矿产勘探]