基于模型的薄层结构智能反演试验分析  

Experimental analysis of intelligent inversion for thin-layer structure based on model

作　　者：杨娜霞 赵东凤 郭淑文 熊金良[4] 李国发[1,2] YANG Naxia;ZHAO Dongfeng;GUO Shuwen;XIONG Jinliang;LI Guofa(State Key Laboratory of Petroleum Resources and Prospecting,China University of Petroleum,Beijing 102249,China;Key Laboratory of Geophysical Exploration of China National Petroleum Corporation,China University of Petroleum,Beijing 102249,China;Research Institute of Exploration and Development,Dagang Oil Field,PetroChina,Tianjin 300280,China;Dagang Oil Field Branch,PetroChina,Tianjin 300280,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)CNPC物探重点实验室,北京102249 [3]中国石油天然气股份有限公司大港油田勘探开发研究院,天津300280 [4]中国石油天然气股份有限公司大港油田分公司,天津300280

出　　处：《石油物探》2023年第3期419-430,共12页Geophysical Prospecting For Petroleum

基　　金：国家自然科学基金面上项目(41874141);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(ZLZX2020-03)共同资助

摘　　要：受地震数据有效频带的限制,常规的地震反演方法很难对薄层结构进行准确刻画和描述。基于机器学习的地震反演方法是近年来用于薄层结构预测的新技术。为此,基于BLSTM-Net神经网络模型,针对薄层空间结构的预测问题开展了简单及复杂陆相沉积模型的阻抗反演试验分析。首先,构建简单的薄互层模型,开展基于BLSTM-Net模型智能反演与基于测井约束的常规地震反演方法的试验对比,同时对BLSTM-Net模型的抗噪性进行测试;然后,构建典型的陆相沉积复杂薄互层地质模型,对反演结果的可靠性及其对地震频带的依赖性进行试验分析;最后,对比分析BLSTM-Net神经网络模型与稀疏脉冲反演对弱反射的恢复和保护能力。模型试验结果表明,基于BLSTM-Net模型的反演方法较常规反演方法具有更强的薄层结构预测能力,且对弱反射具有更好的保护作用,具有更大幅度提高实际地震数据薄层刻画精度的理论优势和技术潜力。Owing to the effective frequency band of seismic data,it is challenging to accurately describe and characterize thin-layer structures using conventional seismic inversion methods.Seismic inversion has therefore been combined with machine learning in recent years to predict thin-layer structures.Hence,the impedance inversion experimental analysis of simple and complex continental deposition models was carried out for the prediction of thin-layer spatial structure based on the BLSTM-Net network.First,a simple thin interbed model was used to compare the BLSTM-Net model with an inversion method constrained by well logging data.Simultaneously,the anti-noise ability of the BLSTM-Net model was tested.Then,a complex thin interbed geological model was constructed with typical continental sedimentary deposits;the reliability of the inversion results and their dependence on the seismic frequency band was analyzed.Finally,the ability of the BLSTM-Net model to recover weak seismic reflections was compared with that of sparse spike inversions.The results of the model tests showed that the BLSTM-Net model could predict thin-layer structures better than conventional inversion methods.The BLSTM-Net model performed better than sparse spike inversions in capturing weak reflection signals.The theoretical advantages and technical potential of the BLSTM-Net model significantly improve characterization and accuracy of thin-layer structures based on seismic data.

关 键 词：储层表征 薄层结构 BLSTM-Net模型 阻抗反演 智能反演 分辨率 

分 类 号：P631[天文地球—地质矿产勘探]