首页|基于地震频相融合的智能化高分辨率反演

基于地震频相融合的智能化高分辨率反演

扫码查看
目前的地震反演主要以模型驱动为主,对噪声较敏感,反演结果的分辨率往往较低,通常只使用地震数据的振幅信息,没有充分利用相位信息.研究表明,波阻抗和地震振幅、相位的相关性较强,因此联合应用振幅谱和相位谱可以有效减小反演结果的多解性.为此,结合深度学习与地震数据的频(振幅谱)、相(相位谱)信息,提出基于数据驱动的频相智能波阻抗反演,有效提高了反演结果的分辨率和精度.具体步骤为:①基于高分辨率时频分析提取地震道的频相信息;②应用图像处理技术融合频相信息;③结合频相信息和地层波阻抗制作数据标签对,用数据标签对训练优选的深度网络;④提取待反演三维地震数据的频相信息并输入训练好的深度网络,即可得到高分辨率波阻抗反演结果.与现有的深度学习地震反演方法相比,所提方法主要创新点在于首次在地震反演中同时应用了地震数据的频相信息.模型测试结果表明,频相信息的联合应用有效地减小了反演结果的多解性,提高了反演精度.实际应用结果表明,与传统的稀疏脉冲反演相比,所提方法提高了反演结果的纵向分辨率,益于进一步推广、应用.
Intelligent high-resolution seismic inversion based on frequency-phase fusion
The existing seismic inversion methodsare mainly model-driven and sensitive to noise,and the resolu-tion of the inversion results is often low.In addition,they usually only use the amplitude information of seismic dataand fail to fully utilize the phase information.The results show that the correlation between wave impedance and seismic amplitude and phase is strong,so the joint application of amplitude spectrum and phase spectrum can effectively alleviate the multi-solution problem of inversion results.Therefore,this paper starts from the data-driven perspective to achieve intelligent wave impedance inversion based on frequency-phase fusion by combining deep learning with frequency(amplitude spectrum)and phase(phase spectrum)information of seis-mic data,which effectively improves the resolution and accuracy of inversion.The steps are as follows:①ex-tracting the frequency and phase information of seismic trace based on the inversion of high-resolution time-fre-quency analysis;②applying image processing technology to fuse frequency and phase information;③combi-ning frequency and phase information and formation wave impedance to make data label pairs and using the data label pairs to train the preferred deep network;④extracting the frequency and phase information of the three-di-mensional seismic data to be retrieved into the trained deep network,so as to obtain the high-resolution wave impedance inversion results.Compared with the existing data-driven deep learning seismic inversion methods,the innovation of the proposed method is as follows:the frequency and phase information are applied simultane-ously in seismic inversion for the first time.The model test results show that the joint application of frequency and phase information can effectively alleviate the multi-solution problem of inversion results and improve the in-version accuracy.The application results in the Dalujia area of Shengli Oilfield show that compared with the tra-ditional sparse pulse inversion,the proposed inversion method has significantly improved the longitudinal reso-lution,which is beneficial to further promotion and application.

frequency-phase informationdata-drivendeep learninghigh-resolution inversion

杨培杰

展开 >

中国石化胜利油田分公司勘探开发研究院,山东东营 257015

频相信息 数据驱动 深度学习 高分辨率反演

中国石化科技项目

P22161

2024

石油地球物理勘探
东方地球物理勘探有限责任公司

石油地球物理勘探

CSTPCD北大核心
影响因子:1.766
ISSN:1000-7210
年,卷(期):2024.59(3)