分频智能反演协同三维岩相建模的单层砂体预测  

Prediction of single-layer sand body by spectral-decomposition intelligent inversion collaborating 3D lithofacies modeling

作　　者：王武荣 岳大力[1,2] 李伟[1,2] 吴胜和[1,2] 芦凤明[3] 续一简[1,2] Wang Wurong;Yue Dali;Li Wei;Wu Shenghe;Lu Fengming;Xu Yijian(State Key Laboratory of Petroleum Resources and Engineering,China University of Petroleum(Beijing),Beijing 102249;College of Geosciences,China University of Petroleum(Beijing),Beijing 102249;Research Institute of Exploration and Development,Dagang Oilfield Company,CNPC,Tianjin 300280)

机构地区：[1]中国石油大学(北京)油气资源与工程全国重点实验室,北京102249 [2]中国石油大学(北京)地球科学学院,北京102249 [3]中国石油大港油田分公司勘探开发研究院,天津300280

出　　处：《地质科学》2024年第6期1694-1707,共14页Chinese Journal of Geology(Scientia Geologica Sinica)

基　　金：国家自然科学基金项目(编号:42302128,42272186);中国石油大学(北京)科研基金项目(编号:2462023YJRC039);中国石油天然气集团有限公司--中国石油大学(北京)战略合作科技专项(编号:ZLZX2020-02)资助。

摘　　要：陆相断陷湖盆重力流沉积横向相变快、非均性强,常规地震储层预测技术通常仅能识别到复合砂体分布,对于单层砂体的井间预测仍较困难。为此,针对港中油田南三断块沙河街组远岸水下扇沉积,采用分频智能反演协同三维岩相建模的方法,旨在定量预测单层砂体分布。首先,对地震资料进行分频处理,优选分频反演参数,采用支持向量机(SVM)方法进行分频智能反演;然后,建立反演属性与岩相(砂岩与泥岩)概率的相关关系,以砂岩和泥岩三维概率模型作为软数据,采用协同指示克里金方法建立三维岩相模型,刻画单层砂体分布;最后,结合单井相解释和分频智能反演剖面进行连井相分析,采用“厚度约束、模式拟合、多维互动”的方法预测单层砂体与沉积微相分布。研究结果表明:目的层Es31-4和Es31-5小层发育辫状水道、朵叶体、席状砂和水道间泥4种沉积微相类型,自下而上整体表现为砂体逐渐退积的过程;Es31-5-2单层朵叶砂体发育广泛,朵叶体之间侧向拼接或局部被带状泥岩侧向隔挡,朵叶体上部发育条带状辫状水道。该方法将分频智能反演与三维地质建模相结合,在井点处忠实于井数据,在井间充分挖掘地震信息,明显提高了砂体预测精度,实现了单层砂体定量精细刻画,可为相似研究区储层精细刻画提供借鉴。The gravity-flow deposits in continental rift lacustrine basin exhibit rapid lateral changes for sedimentary facies and strong heterogeneity.Conventional seismic technologies for reservoir prediction can usually only identify composite sand body distributions,making inter-well sand body predictions for single layers still challenging.Therefore,targeting the offshore submarine fan deposits of the Shahejie Formation in the southern third block of the Gangzhong oilfield,a method combining spectral-decomposition intelligent inversion with three-dimensional lithofacies modeling is adopted to quantitatively predict the distribution of sand body in the single layer.Firstly,the seismic volume is processed through spectral decomposition,with optimal parameters for spectral-decomposition inversion selected,and the spectral-decomposition intelligent inversion is performed using the Support Vector Machines(SVM)method.Then,the relationship between inversion attributes and the probabilities of lithofacies(including sandstone and mudstone)is established.Using the three-dimensional probability models of sandstone and mudstone as soft data,a three-dimensional lithofacies model is constructed using the co-kriging method,and the sand body distribution in single layers can be obtained from the lithofacies model.Finally,by combining sedimentary facies interpretations of single wells with spectral-decomposition intelligent inversion profiles,inter-well facies analysis is conducted,and the sand body distribution and sedimentary microfacies distribution of single layers is predicted using methods of“thickness constraint,pattern fitting,and multidimensional interaction”.The results show that the target layers Es31-4 and Es31-5 develop four types of sedimentary microfacies,i.e.,braided channels,lobe bodies,sheet sands and inter-channel mudstones,which display the gradual retrogradation process of sand body from bottom to top;the lobe sand bodies are extensively developed in single layer Es31-5-2,with lateral superposition between lobe

关 键 词：砂体预测 分频反演 机器学习 岩相建模 港中油田 

分 类 号：P618.13[天文地球—矿床学]