波动方程正演引导的深度学习地震波形反演

段友祥, 崔乐乐, 孙歧峰, 杜启振

石油地球物理勘探 ›› 2023, Vol. 58 ›› Issue (3) : 485-494.

PDF(3750 KB)
PDF(3750 KB)
石油地球物理勘探 ›› 2023, Vol. 58 ›› Issue (3) : 485-494. DOI: 10.13810/j.cnki.issn.1000-7210.2023.03.001
智能地球物理

波动方程正演引导的深度学习地震波形反演

  • 段友祥1, 崔乐乐1, 孙歧峰1, 杜启振2
作者信息 +

Deep learning seismic waveform inversion based on the forward modeling guidance of wave equation

  • DUAN Youxiang1, CUI Lele1, SUN Qifeng1, DU Qizhen2
Author information +
文章历史 +

摘要

物理驱动的全波形反演方法计算成本高，数据驱动的深度学习反演方法对标记数据集的依赖性强。为了在有限的数据条件下获得更好的反演结果，结合数据驱动与物理驱动，提出了波动方程正演引导的深度学习地震波形反演方法。首先，利用地震数据应用神经网络重建速度模型，对网络预测的速度模型进行正演建模，通过最小化速度模型的误差及地震数据的误差训练网络；其次，使用有限差分法将二阶偏微分波动方程近似为可微算子，使正演过程能够传递梯度，并根据梯度方向动态调整地震数据损失的权重。实验结果表明，该方法能在一定程度上降低数据驱动方法对标记数据集的依赖性，可得到更准确的速度模型，且具有较强的鲁棒性。

Abstract

Physics-driven full-waveform inversion method has high computational overhead, while data-driven deep learning inversion method has a high dependence on the marker dataset. In order to obtain better inversion results with limited data, a deep learning seismic waveform inversion method based on the forward modeling guidance of the wave equation is proposed, which integrates data-driven and physics-driven methods. Firstly, a neural network is applied to reconstruct the velocity model based on seismic data, and the velocity model predicted by the network is modeled by forward simulation. Then the network is trained by minimizing the error of the velocity model and seismic data. Secondly, a finite difference method is adopted to approximate the second-order partial differential wave equation as a differentiable operator. The forward modeling process enables gradient transfer, and the weights of seismic data loss are dynamically adjusted according to the direction of the gradients. The experimental results show that the method can reduce the dependence of the data-driven method on the marker dataset to a certain extent, obtain more accurate velocity models, and have stronger robustness.

关键词

深度学习 / 地震反演 / 速度模型建立 / 正演建模 / 波动方程

Key words

deep learning / seismic inversion / velocity-model building / forward modeling / wave equation

引用本文

EndNote

Ris (Procite)

Bibtex

导出引用
段友祥, 崔乐乐, 孙歧峰, 杜启振. 波动方程正演引导的深度学习地震波形反演[J]. 石油地球物理勘探, 2023, 58(3): 485-494 https://doi.org/10.13810/j.cnki.issn.1000-7210.2023.03.001
DUAN Youxiang, CUI Lele, SUN Qifeng, DU Qizhen. Deep learning seismic waveform inversion based on the forward modeling guidance of wave equation[J]. Oil Geophysical Prospecting, 2023, 58(3): 485-494 https://doi.org/10.13810/j.cnki.issn.1000-7210.2023.03.001
中图分类号： P631   

参考文献

[1] TARANTOLA A.Inversion of seismic reflection data in the acoustic approximation[J].Geophysics,1984,49(8):1259-1266.
[2] 单蕊,卞爱飞,於文辉,等.利用叠前全波形反演进行储层预测[J].石油地球物理勘探,2011,46(4):629-633.
SHAN Rui,BIAN Aifei,YU Wenhui,et al.Pre-stack full waveform inversion for reservoir prediction[J].Oil Geophysical Prospecting,2011,46(4):629-633.
[3] LANDA E,TREITEL S.Seismic inversion:what it is,and what it is not[J].The Leading Edge,2016,35(3):277-279.
[4] LECUN Y,BENGIO Y,HINTON G.Deep learning[J].Nature,2015,521(7553):436-444.
[5] JIN K H,MCCANN M T,FROUSTEY E,et al.Deep convolutional neural network for inverse problems in imaging[J].IEEE Transactions on Image Processing,2017,26(9):4509-4522.
[6] LUCAS A,ILIADIS M,MOLINA R,et al.Using deep neural networks for inverse problems in imaging:beyond analytical methods[J].IEEE Signal Processing Magazine,2018,35(1):20-36.
[7] YAO H,DAI F,ZHANG S,et al.Dr2-Net:deep residual reconstruction network for image compressive sensing[J].Neurocomputing,2019,359:483-493.
[8] WALDELAND A U,JENSEN A C,GELIUS L J,et al.Convolutional neural networks for automated seismic interpretation[J].The Leading Edge,2018,37(7):529-537.
[9] 杨平,宋强功,詹仕凡,等.基于深度学习的高效构造解释技术研发及工业化应用[J].石油地球物理勘探,2022,57(6):1265-1275.
YANG Ping,SONG Qianggong,ZHAN Shifan,et al.Research and industrial application of efficient structural interpretation technology based on deep learning[J].Oil Geophysical Prospecting,2022,57(6):1265-1275.
[10] 闫星宇,顾汉明,罗红梅,等.基于改进深度学习方法的地震相智能识别[J].石油地球物理勘探,2020,55(6):1169-1171.
YAN Xingyu,GU Hanming,LUO Hongmei,et al.Intelligent seismic facies classification based on an improved deep learning method[J].Oil Geophysical Prospecting,2020,55(6):1169-1171.
[11] CHOPRA S,MARFURT K,SHARMA R.Unsupervised machine learning facies classification in the De-laware Basin and its comparison with supervised Bayesian facies classification[C].SEG Technical Program Expanded Abstracts,2019,38:2619-2623.
[12] YU S,MA J,WANG W.Deep learning for denoising[J].Geophysics,2019,84(6):V333-V350.
[13] 钟铁,陈云,董新桐,等.基于DBBCNN的沙漠区地震资料随机噪声衰减方法[J].石油地球物理勘探,2022,57(2):268-278.
ZHONG Tie,CHEN Yun,DONG Xintong,et al.Research on random noise attenuation method for seismic data from deserts based on DBBCNN[J].Oil Geophysical Prospecting,2022,57(2):268-278.
[14] ARAYA-POLO M,JENNINGS J,ADLER A,et al.Deep-learning tomography[J].The Leading Edge,2018,37(1):58-66.
[15] WANG W,YANG F,MA J.Velocity model building with a modified fully convolutional network[C].SEG Technical Program Expanded Abstracts,2018,37:2086-2090.
[16] WANG W,MA J.Velocity model building in a crosswell acquisition geometry with image-trained artificial neural networks[J].Geophysics,2020,85(2):U31-U46.
[17] WU Y,LIN Y.InversionNet:an efficient and accurate data-driven full waveform inversion[J].IEEE Transactions on Computational Imaging,2020,6:419-433.
[18] YANG F,MA J.Deep-learning inversion:a next-ge-neration seismic velocity model building method[J].Geophysics,2019,84(4):R583-R599.
[19] LI S,LIU B,REN Y,et al.Deep-learning inversion of seismic data[J].IEEE Transactions on Geoscience and Remote Sensing,2020,58(3):2135-2149.
[20] ARAYA-POLO M,FARRIS S,FLOREZ M.Deep learning-driven velocity model building workflow[J].The Leading Edge,2019,38(11):872a1-872a9.
[21] MOSSER L,DUBRULE O,BLUNT M J.Stochastic seismic waveform inversion using generative adversa-rial networks as a geological prior[J].Mathematical Geosciences,2020,52(1):53-79.
[22] BISWAS R,SEN M K,DAS V,et al.Pre-stack inversion using a physics-guided convolutional neural network[C].SEG Technical Program Expanded Abstracts,2019,38:4967-4971.
[23] HORNIK K,STINCHCOMBE M,WHITE H.Multilayer feedforward networks are universal approximators[J].Neural Networks,1989,2(5):359-366.
[24] STRIKWERDA J C.Finite Difference Schemes and Partial Differential Equations (2nd ed)[M].Society for Industrial and Applied Mathematics,Philadelphia,2004.
[25] RONNEBERGER O,FISCHER P,BROX T.U-Net:convolutional networks for biomedical image segmentation[C].Medical Image Computing and Computer-Assisted Intervention - MICCAI 2015,2015,234-241.
[26] LIN X,BAWEJA H S,KANTOR G,et al.Adaptive auxiliary task weighting for reinforcement learning[C].33rd Conference on Neural Information Proces-sing Systems (NeurIPS 2019),2019,1-12.
[27] PASALIC D,MCGARRY R.Convolutional perfectly matched layer for isotropic and anisotropic acoustic wave equations[C].SEG Technical Program Expan-ded Abstracts,2010,29:2925-2929.
[28] LOUBOUTIN M,LANGE M,LUPORINI F,et al.Devito (v3.1.0):an embedded domain-specific language for finite differences and geophysical exploration[J].Geoscientific Model Development,2019,12(3):1165-1187.

基金

本项研究受国家自然科学基金项目"弹性全波形反演与逆时偏移成像"(41930429)、中央高校基本科研业务费专项资金项目"面向深层油藏精细描述的地质知识库研究与开发"(20CX05017A)和中石油重大科技项目"面向深层油藏精细描述的地质知识库研究与开发"(ZD2019-183-006)联合资助。
PDF(3750 KB)

184

Accesses

0

Citation

Detail
段落导航
相关文章

作者中心

  • 在线投稿
  • 在线查稿
  • 投稿须知
  • 论文写作须知
  • 论文模板
  • 版权协议

    • 审稿中心

  • 主编审稿
  • 专家审稿
  • 申请成为审稿人

    • 在线期刊

  • 当期目录
  • 优先出版
  • 过刊浏览
  • 阅读排行
  • 下载排行
  • 引用排行

    • 编辑部

  • 关于我们
  • 编辑办公
  • 创刊50周年专题片

    • 联系我们

  • 地址：河北省涿州市甲秀路39号东方地球物理公司5号院科技楼A座415—419室
  • 邮编：072751
  • 电话：（0312)3822282 ,3821246
    • 传真：（0312)3822458
  • 网址： http://www.ogp-cn.com 或 http://www.ogp-cn.com.cn
    • 京ICP备13042042号　冀ICP备14016681号-4　版权所有 © 《石油地球物理勘探》编辑部

    /

    〈 〉