基于深度学习的地震速度反演方法研究

许旺; XU Wang; 王非翊; WANG Feiyi; 张志强; ZHANG Zhiqiang; 刘恭利; LIU Gongli; 段新意; DUAN Xinyi

doi:DOI:10.48014/cpngr.20230425002

基于深度学习的地震速度反演方法研究

Research on Seismic Velocity Inversion Method Based on Deep Learning

中国石油天然气研究 2023年第2卷第2期页码[15-24] 下载全文[2.5MB] [HTML]

摘要

获得地震速度的方法包括偏移速度分析、层析速度反演和全波形反演, 但这些方法都有共同的问题: 随着地震数据量的增大, 处理数据获得地震速度的时间成倍增加, 且后两种方法对初始地震速度依赖性较强。针对以上方法所存在的问题, 本文改进了一种基于深度学习的地震速度反演方法。同时提出了一种随机生成大量速度模型的方法, 该方法生成的速度模型拥有与真实地下构造相近的地质特征 (起伏层、断层、异常体等) 。利用生成的速度模型和波动方程进行正演, 可以实现数据集的高效建立。本文改进的深度学习反演方法基本原理为: 通过卷积神经网络提取训练数据的特征信息, 经过大数据训练, 获得地震炮记录和地震速度的非线性映射关系。在反演阶段, 将地震炮记录输入训练好的网络, 可以快速地反演出地震速度。为了使该网络充分发挥处理地震数据的优势, 作者通过数值模拟的方式获得了优势网络结构, 并取得了令人满意的反演结果, 最后本文通过对比实验, 验证了该方法的优势和适用性。

Abstract

Methods for obtaining seismic velocity include offset velocity analysis, tomography velocity inversion and full waveform. inversion, but these methods all share common problems: As the amount of seismic data increases, the time required to process the data to obtain the seismic velocity increases exponentially, and the latter two methods are more dependent on the initial seismic velocity. To address the problems of the above methods, this paper improves a seismic velocity inversion method based on deep learning. At the same time, a method that can generate a large number of velocity models randomly with geological features (undulation layer, faults, anomalies, etc. ) similar to those of real subsurface structures is also proposed. The generated velocity model and fluctuation equation for forward modeling are used to perform. the forward modeling, which allows for the efficient establishment of data set. The basic principle of the improved deep learning inversion method in this paper is as follows: the characteristic information of the training data is extracted by convolution neural network, which is trained with large data to obtain a nonlinear mapping relationship between seismic record and seismic velocity. In the inversion stage, the seismic velocity can be inversed quickly by inputting the seismic records into the trained network. In order to make the network give full play to the advantages of processing seismic data, the authors obtained the dominant network structure by means of numerical simulation, and achieved satisfactory inversion results. Finally, through comparative experiments, this paper verifies the advantages and applicability of the method.

DOI	DOI:10.48014/cpngr.20230425002
文章类型	研究性论文
收稿日期	2023-04-25
接收日期	2023-05-20
出版日期	2023-06-28
关键词	地震速度, 深度学习, 卷积神经网络, 数据集, 反演问题
Keywords	Seismic velocity, deep learning, convolutional neural networks, data set, inversion problems
作者	许旺¹, 王非翊^2,*, 张志强¹, 刘恭利¹, 段新意¹
Author	XU Wang¹, WANG Feiyi^2,*, ZHANG Zhiqiang¹, LIU Gongli¹, DUAN Xinyi¹
所在单位	1. 中海石油 (中国) 有限公司天津分公司, 天津 300459 2. 北京大学地球与空间科学学院, 北京 100871
Company	1. China National Offshore Oil (China) Corporation, Tianjin Branch, Tianjin 300459, China 2. School of Earth and Space Sciences, Peking University, Beijing 100871, China
浏览量	390
下载量	198
参考文献	[1] Hinton G E, Osindero S, Teh Y W. A fast learning algorithm for deep belief nets[J]. Neural computation, 2006, 18(7): 1527-1554. https://doi.org/10.1162/neco.2006.18.7.1527 [2] LeCun Y, Bengio Y, Hinton G. Deep learning[J]. Nature, 2015, 521(7553): 436-444. https://doi.org/10.1038/nature14539 [3] Hinton G E, Srivastava N, Krizhevsky A, et al. Improving neural networks by preventing co-adaptation of feature detectors[ J]. arXiv preprint, 2012, arXiv: 1207. 0580. https://doi.org/10.48550/arXiv.1207.0580 [4] Deng L, Yu D. Deep learning: methods and applications[J]. Foundations & Trends in Signal Processing, 2014, 7(3): 197-387. http://dx.doi.org/10.1561/2000000039 [5] Wang F, Wu X, and Wang H. Seismic horizon identification using semi-supervised learning with virtual adversarial training[J]. IEEE TGRS, 2022, 60: 1-11. https://doi.org/10.1109/TGRS.2022.3154439 [6] Bi Z, Wu X, Geng Z, et al. Deep relative geologic time: a deep learning method for simultaneously interpreting 3D seismic horizons and faults[J]. JGR, Solid Earth, 2021, 126(9): 1-24. https://doi.org/10.1029/2021JB021882 [7] Gao H, Wu X, and Liu G. ChannelSeg3D: channel simulation and deep learning for channel interpretation in 3D seismic images[J]. Geophysics, 2021, 86(4): IM73-IM83. https://doi.org/10.1190/geo2020-0572.1 [8] Röth G, Tarantola A. Neural networks and inversion of seismic data[J]. Journal of Geophysical Research: Solid Earth, 1994, 99(B4): 6753-6768. https://doi.org/10.1029/93JB01563 [9] Nath S K, Chakraborty S, Singh S K, et al. Velocity inversion in cross-hole seismic tomography by counterpropagation neural network, genetic algorithm and evolutionary programming techniques[J]. Geophysical Journal International, 1999, 138(1): 108-124. https://doi.org/10.1046/j.1365-246x.1999.00835.x [10] Zhu J Y, Park T, Isola P, et al. Unpaired image-to-image translation using cycle-consistent adversarial networks[ A]. Proceedings of the IEEE international conference on computer vision[C]. United States: IEEE, 2017: 2223-2232. https://doi.org/10.48550/arXiv.1703.10593 [11] Araya-Polo M, Jennings J, Adler A, et al. Deep-learning tomography[J]. The Leading Edge, 2018, 37(1): 58-66. https://doi.org/10.1190/tle37010058.1 [12] Wu Y, Lin Y, Zhou Z. Inversion Net: Accurate and efficient seismic waveform inversion with convolutional neural networks[M]. United States: Society of Exploration Geophysicists, 2018. https://doi.org/10.1190/segam2018-2998603.1 [13] Yuan C, Zhang X, Jia X, et al. Time-lapse velocity imaging via deep learning[J]. Geophysical Journal Interna-tional, 2020, 220(2): 1228-1241. https://doi.org/10.1093/gji/ggz511 [14] Yang F, Ma J. Deep-learning inversion: A next-generation seismic velocity model building method[J]. Geophysics, 2019, 84(4): R583-R599. https://doi.org/10.1190/geo2018-0249.1 [15] Ren Y, Nie L, Yang S, et al. Building complex seismic velocity models for deep learning inversion[J]. IEEE Access, 2021, 9: 63767-63778. https://doi.org/10.1109/ACCESS.2021.3051159 [16] Liu B, Yang S, Ren Y, et al. Deep-learning seismic fullwaveform inversion for realistic structural models[J]. Geophysics, 2021, 86(1): R31-R44. https://doi.org/10.1190/geo2019-0435.1 [17] 朱军. 二维地震正演模拟方法技术研究[D]. 西安: 长安大学, 2007. [18] Hornik K, Stinchcombe M, White H. Multilayer feedforward networks are universal approximators[J]. Neural Networks, 1989, 2(5): 359-366. https://doi.org/10.1016/0893-6080(89)90020-8 [19] Ronneberger O, Fischer P, Brox T. U-net: Convolutional networks for biomedical image segmentation[A]// International Conference on Medical image computing and computer-assisted intervention [C]. Germany: Springer, 2015: 234-241. https://doi.org/10.1007/978-3-319-24574-4_28
引用本文	许旺, 王非翊, 张志强, 等. 基于深度学习的地震速度反演方法研究[J]. 中国石油天然气研究, 2023, 2(2): 15-24.
Citation	XU Wang, WANG Feiyi, ZHANG zhiqiang, et al. Research on seismic velocity inversion method based on deep learning[J]. Chinese Petroleum and Natural Gas Research, 2023, 2(2): 15-24.