Abstract:Due to the lacking low frequency information and the low resolution of seismic data for deep or ultra-deep land layers, the accurate interpretation of subsequent seismic data is affected. Model-dri-ven low frequency compensation methods have dependence on strict assumptions and inflexible parameter adjustment. The convolutional neural network (CNN) has limited feature extraction ability for subtle changes and no obvious gradient changes, and the network is easy to fall into local optimum, resulting in low frequency undercompensation or low compensation accuracy. Therefore, a low frequency compensation method for pre-stack seismic data combining improved CNN and double constrained loss function is proposed. On the premise of not increasing the computational complexity of CNN, residual blocks network units that can directly learn residual features between input and output are added to solve gradient disappearance. Additionally, batch normalization is adopted to make the network more sensitive to subtle changes, to improve network training efficiency. Since the gradient changes are not obvious, the network convergence is premature. To address the problem, this paper takes the difference and the correlation between the network output and original seismic record as optimization objectives and establishes the loss function by the weighted sum of the mean square error and Pearson distance to calculate the compensation error under double constraints. Finally, the gradient changes become more evident and ensure that the local optimal can be jumped out during gradient descent, so as to improve the low frequency compensation accuracy. The synthetic data and the low frequency compensation results of the real pre-stack seismic data in X area of western China verify the feasibility and effectiveness of the proposed method. Compared with the low frequency compensation method based on CNN and that based on deconvolution combined with broadband Yu low-pass filter, the proposed method can compensate the low frequency components without destroying the original medium and high frequency information.
戴永寿, 高倩倩, 孙伟峰, 万勇, 吴莎莎. 结合改进CNN和双约束损失函数的叠前地震数据低频补偿方法[J]. 石油地球物理勘探, 2022, 57(6): 1287-1295.
DAI Yongshou, GAO Qianqian, SUN Weifeng, WAN Yong, WU Shasha. Low frequency compensation of pre-stack seismic data based on improved CNN and double constrained loss function. Oil Geophysical Prospecting, 2022, 57(6): 1287-1295.
杨文采,于常青. 深层油气地球物理勘探基础研究[J]. 地球物理学进展,2007,22(4): 1238-1242.YANG Wencai,YU Changqing. On basic research problems in applied geophysics for deep oil and gas fields[J]. Progress in Geophysics,2007,22(4): 1238-1242.
[2]
滕吉文,司芗,王玉辰. 我国化石能源勘探、开发潜能与未来[J]. 石油物探,2021,60(1): 1-12.TENG Jiwen,SI Xiang,WANG Yuchen. Potential and future of fossil fuel exploration and development in China[J]. Geophysical Prospecting for Petroleum,2021,60(1): 1-12.
[3]
戴晓峰,徐右平,甘利灯,等. 川中深层—超深层多次波识别和压制技术——以高石梯—磨溪连片三维区为例[J]. 石油地球物理勘探,2019,54(1): 54-64.DAI Xiaofeng,XU Youping,GAN Lideng,et al. Deep & ultra-deep multiple suppression in Central Sichuan: an example of Gaoshiti-Moxi[J]. Oil Geophy-sical Prospecting,2019,54(1): 54-64.
[4]
曲寿利. 面向深层复杂地质体油气勘探的地震一体化技术[J]. 石油物探,2021,60(6): 879-892.QU Shouli. An integrated seismic technology for oil and gas exploration in a deep complex geological body[J]. Geophysical Prospecting for Petroleum,2021,60(6): 879-892.
[5]
毛博,韩立国. 基于相似性重构低频数据的金属矿频域全波形反演[J]. 地球物理学报,2019,62(10): 4010-4019.MAO Bo,HAN Liguo. Full waveform inversion in the frequency domain of low-frequency seismic data based on similarity reconstruction for exploration of deep metallic ores[J]. Chinese Journal of Geophy-sics,2019,62(10): 4010-4019.
[6]
YANG J,ZHU H. Low-frequency compensation and its application in full-waveform inversion[C]. SEG Technical Program Expanded Abstracts,2018,37: 1304-1308.
[7]
韩立国,张莹,韩利,等. 基于压缩感知和稀疏反演的地震数据低频补偿[J]. 吉林大学学报(地球科学版),2012,42(增刊3): 259-264.HAN Liguo,ZHANG Ying,HAN Li,et al. Compressed sensing and sparse inversion based low-frequency information compensation of seismic data[J]. Journal of Jilin University(Earth Science Edition),2012,42(S3): 259-264.
[8]
宋亚民,戴朝强,张丽萍,等. 流花地区巨厚灰岩层下伏构造落实方法研究[J]. 石油物探,2020,59(6): 978-987.SONG Yamin,DAI Chaoqiang,ZHANG Liping,et al. Confirming the structure of underlying a very thick limestone formation in the Liuhua area,China[J]. Geophysical Prospecting for Petroleum,2020,59(6): 978-987.
[9]
丁燕,杜启振,刘力辉,等. 基于压缩感知和宽带俞式低通整形滤波器的地震低频信息特征分析与补偿[J]. 地球物理学报,2019,62(6): 2267-2275.DING Yan,DU Qizhen,LIU Lihui,et al. Feature analysis and compensation of seismic low-frequency based on compressed sensing and broad-band Yu-type low-passing shaping filter[J]. Chinese Journal of Geophysics,2019,62(6): 2267-2275.
[10]
WU R S,LUO J R,WU B Y. Seismic envelope inversion and modulation signal model[J]. Geophysics,2014,79(3): WA13-WA24.
[11]
王鹏. 可控震源信号低频恢复和拓展研究[D]. 北京: 中国石油大学(北京),2016.
[12]
妥军军,王晓涛,窦强峰,等. 准噶尔盆地石炭系低频信号处理技术[J]. 石油地球物理勘探,2020,55(增刊1): 20-24.TUO Junjun,WANG Xiaotao,DOU Qiangfeng,et al. Research and application of the processing method for Carboniferous low-frequency signals,Junggar Basin[J]. Oil Geophysical Prospecting,2020,55(S1): 20-24.
[13]
张红军,魏程霖,张丹,等. 低频配套处理技术在岩性油气藏识别中的应用[J]. 石油地球物理勘探,2017,52(增刊2): 64-71.ZHANG Hongjun,WEI Chenglin,ZHANG Dan,et al. Application of low frequency matching processing technology in lithologic reservoirs[J]. Oil Geophysical Prospecting,2017,52(S2): 64-71.
OVCHARENKO O,KAZEI V,PETER D,et al. Neural network based low-frequency data extrapolation[C]. Workshop: Full Waveform inversion: What are we getting? Manama,Bahrain,2017.
[16]
SUN H Y,DEMANET L. Low frequency extrapolation with deep learning[C]. SEG Technical Program Expanded Abstracts,2018,37: 2011-2015.
[17]
毛博. 基于卷积神经网络的地震数据重构与模型构建研究[D]. 吉林长春: 吉林大学,2020.
[18]
OVCHARENKO O,KAZEI V,PLOTNITSKIY P,et al. Extrapolating low-frequency prestack land data with deep learning[C]. SEG Technical Program Expanded Abstracts,2020,39: 1546-1550.
[19]
WANG M X,XU S,ZHOU H B. Self-supervised learning for low frequency extension of seismic data[C]. SEG Technical Program Expanded Abstracts,2020,39: 1501-1505.
[20]
OVCHARENKO O,KAZEI V,KALITA M,et al. Deep learning for low-frequency extrapolation from multioffset seismic data[J]. Geophysics,2019,84(6): R989-R1001.
[21]
NAKAYAMA S,BLACQUIÈRE G. Machine-lear-ning-based data recovery and its contribution to seismic acquisition: simultaneous application of deblen-ding,trace reconstruction,and low-frequency extrapolation[J]. Geophysics,2020,86(2): P13-P24.
[22]
HE K,ZHANG X,REN S,et al. Identity mappings in deep residual networks[C]. European Conference on Computer Vision,2016,630-645.
[23]
HE K,ZHANG X,REN S,et al. Deep residual lear-ning for image recognition[C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition,2016,770-778.
[24]
张鹏,戴永寿,谭永成,等. 利用EMD和子波振幅谱与相位谱关系的时变子波提取方法[J]. 地球物理学报,2019,62(2): 680-696.ZHANG Peng,DAI Yongshou,TAN Yongcheng,et al. A time-varying wavelet extraction method using EMD and the relationship between wavelet amplitude and phase spectra[J]. Chinese Journal of Geophysics,2019,62(2): 680-696.
[25]
廖仪,刘巍,胡林,等. 地震保幅高低频拓展与多尺度贝叶斯融合反演[J]. 石油地球物理勘探,2021,56(6): 1330-1339.LIAO Yi,LIU Wei,HU Lin,et al. Research on high-and low-frequency expansion of seismic amplitude preserving and multi-scale Bayesian fusion inversion[J]. Oil Geophysical Prospecting,2021,56(6): 1330-1339.