智能地球物理丨《石油地球物理勘探》精选双语文章|地震|石油地球物理勘探|重磁

围绕智能地球物理研究主题，精选5篇发表在《石油地球物理勘探》上的文章，欢迎阅读！

精选文章Selected Articles

基于深度学习的高效构造解释技术研发及工业化应用

Research and industrial application of efficient structural interpretation technology based on deep learning

【摘要】深度学习为传统地震资料解释技术的发展带来了强劲动力，由此催生了大量的智能化解释技术。但目前的研究成果很少能实现规模化生产应用。为此，面向中、低信噪比地震资料，着力基于深度学习的智能化解释技术工业化落地。在开发智能化软件研发平台的基础上，形成了具有较强资料适应能力的智能化层位解释与断层检测技术，在大规模连片资料层位解释、复杂断块精细描述中发挥了重要作用。与传统的自动解释技术相比，该技术的层位解释效率提高8～20倍，断层识别精度相对于相干、曲率属性等得到显著提高。目前已能完全取代原有的自动解释功能，基本实现了构造解释的智能化转型。

【Abstract】Deep learning technology has given a strong impetus to the development of interpretation technology for traditional seismic data, which has spawned a large number of intelligent interpretation technologies. However, limited research achievements can be applied in large-scale production. This study focuses on the industrial implementation of intelligent interpretation technologies based on deep learning for low-to-medium signal-to-noise ratio (SNR) data. Upon the development of the intelligent software development platform, an intelligent horizon interpretation and fault detection technology with strong data adaptability is formed, which plays an important role in the horizon interpretation of large-scale continuous survey data and the fine description of complex fault blocks. The efficiency of horizon interpretation by this method is increased by a factor of 9–21 compared with that of the traditional automatic interpretation techniques, and the accuracy of fault identification is significantly improved in comparison with that of classic techniques such as coherence and curvature. It can completely replace the original automatic interpretation modules and achieve the intelligent transformation of structural interpretation.

改进的整体嵌套边缘检测地震断层识别技术

Seismic fault interpretation based on improved holistically-nested edge detection

【摘要】断层解释的精度和效率对油气藏的勘探与开发非常重要。传统的断层解释方法多以人工为主，其依赖解释人员的经验且耗时较长；常规自动断层解释方法主要是分析地震数据的不连续性，往往涉及多个参数，因而断层解释精度多依赖选取的参数。近年来，随着深度学习技术的发展，非线性卷积神经网络能够描述地震数据中的不连续特征。为此，引入深度学习中的边缘检测技术，即整体嵌套边缘检测(Holistically-Nested Edge Detection, HED)网络，并根据地震数据和断层特点对网络结构进行改进和优化，提出适用于地震断层智能解释的改进HED（Improved HED，IHED）网络。主要步骤包括：(1)将原始二维HED网络推广至三维，搭建三维HED网络；(2)根据HED网络的多尺度特点，调整三维HED网络构架；(3)利用三维合成地震数据及其标签数据训练得到三维IHED模型，将该模型用于实际地震数据进行断层智能解释。与相干体算法和U-Net模型相比，三维IHED模型对断层预测的准确性更高，连续性更好。该方法为地震断层智能识别提供了一条可靠途径。

【Abstract】The accuracy and efficiency of fault interpretation greatly affect the exploration and development of oil and gas reservoirs. The traditional manual fault interpretation method relies on the experience of interpreters and takes a long time; the conventional automatic fault interpretation method mainly interprets faults by discontinuity analysis of seismic data and often contains multiple parameters, and thus its accuracy in fault interpretation mostly depends on the selected parameters. With the development of deep learning in recent years, convolutional neural networks (CNNs) with nonlinear properties can also describe the discontinuous characteristics of seismic data. Therefore, an edge detection technology in deep learning, i.e., the holistically-nested edge detection (HED) network, is introduced in this study, and the network is improved and optimized on the basis of the characteristics of seismic data and seismic faults, which leads to the improved HED (IHED) network suitable for intelligent seismic fault interpretation. The main steps are as follows: (1) The original two-dimensional (2D) HED network is extended to a three-dimensional (3D) version, and thus a 3D HED network is constructed; (2) the architecture of the 3D HED network is adjusted considering the multi-scale property of the network; (3) the 3D HED network is trained with 3D synthetic seismic data and corresponding label data for a 3D IHED model, and then the 3D IHED model is applied to field data for seismic fault interpretation. Compared with the coherence cube algorithm and U-Net model, the 3D IHED model features higher accuracy in the prediction of faults and better continuity. The proposed model provides an efficient and reliable new idea for intelligent fault interpretation.

基于自注意力机制深度学习的重磁数据网格化和滤波方法

Gridding and filtering method of gravity and magnetic data based on self-attention deep learning

【摘要】重磁数据网格化和滤波结果直接影响解释结果，为此设计了合理的深度学习网络结构以实现高精度重磁数据网格化和滤波处理。建立基于自注意力机制深度学习的网格化方法，使用自注意力机制层对二维位置编码进行处理，得到融合了全局与局部信息的位置编码向量，再将位置信息与异常信息融合输出节点异常，从而降低数据的失真性。针对重磁数据噪声具有随机性、条带状的特点，首先采用卷积神经网络进行噪声分类，针对条带状噪声和随机噪声分别采用自注意力机制神经网络和卷积自编码器进行去除，可获得质量较高的基础数据。模型试验表明，深度学习的网格化结构相对常规方法更接近真实结果，所开发的滤波方法能很好地实现不同类型噪声的去除，为后续反演提供更准确的基础数据。将基于深度学习的网格化和滤波方法用于实际磁场数据的处理，获得了较好的结果，证明该方法具有较强的实用性。

【Abstract】The gridding and filtering of gravity and magnetic data directly influence the result of data processing. This paper designs a rational deep learning model to improve the accuracy of gridding and filtering. The gridding method based on self-attention deep learning is constructed, and the self-attention mechanism layer is utilized to process the two-dimensional position embeddings. In this way, the vector of position embeddings is obtained with global and local information integrated. Then, the position information and anomaly information are fused to output node anomaly to alleviate distortion. For the random and stripe noise of gravity and magnetic data, a convolutional neural network is first employed to classify noise. The stripe noise is filtered by self-attention convolutional neural network and the random noise by convolutional autoencoder to produce high-quality basic data. The model experiment shows that the gridded structure of deep learning is closer to the real result than that of the traditional method. The proposed filtering method can remove various noise, providing more accurate basic data for the following inversion. The application of the gridding and filtering method based on deep learning to practical magnetic data achieves good results, which proves that it has strong feasibility and practicability

应用平稳小波变换与深度残差网络压制地震随机噪声

Seismic random noise attenuation based on stationary wavelet transform and deep residual neural network

【摘要】常规去噪方法众多，但每种方法都受某种假设或条件限制。另外，常规去噪方法中一些优化问题具有多个局部极值，导致算法可能收敛到局部最优解而非全局最优解。为此，提出了一种基于平稳小波变换与深度残差网络的地震随机噪声压制方法。采用残差网络（ResNet）的拓扑结构，结合平稳小波变换压制地震数据噪声。残差模块有效避免了网络过深引起的梯度消失或计算消耗但损失函数趋于饱和的问题。另外，小波变换是一种高效的特征提取方法，可获得信号低频和不同方向高频特征信息，分区域学习信号或噪声的特征。首先，对Train400数据集中的每幅图片旋转不同角度以增加训练集数据量，经过旋转变换后加入高斯噪声。然后，对每幅图片进行1级平稳Haar小波分解，得到训练数据集；通过训练提取信号中噪声的小波变换高、低频信息，在此基础上通过直连通道，从含噪数据的小波分解中减去学习到的噪声的小波分解，得到去噪信号的小波分解。最后，通过逆平稳小波变换得到去噪信号。合成信号和实际地震数据去噪试验表明，所提方法能较好地压制地震随机噪声，去噪信号的信噪比、峰值信噪比均较高。

【Abstract】There are many conventional denoising methods, but each of them is limited by certain assumptions or conditions. In addition, multiple local extrema in some optimization problems may cause the denoising algorithm to converge to a local optimal solution instead of a global one. For this reason, a random noise suppression method based on the stationary wavelet transform (SWT) and deep residual neural network (WaveResNet) is proposed. It combined the topology structure of the residual neural network (ResNet) with SWT. The residual module effectively avoids the vanishing gradient or computational consumption with loss function saturation caused by the deep network. In addition, the wavelet transform is an efficient feature extraction method, which can obtain the low-frequency and high-frequency feature information in different directions and learn the features of signals or noise in different regions. First, each picture in the Train400 data set is rotated by different angles to increase the amount of data in the training set, after which Gaussian noise is added. Then, the level-1 stationary Haar wavelet decomposition is performed on each picture to gain a training data set. The high- and low-frequency information in the wavelet transform domain is extracted through training. On this basis, through the direct channel, the wavelet decomposition of the learned noise is subtracted from that of the noisy data, which leads to the wavelet decomposition of the denoised signal. Finally, the denoised signal is obtained through the inverse SWT (ISWT). Experiments of synthetic signals and field seismic data show that the proposed method can satisfactorily suppress seismic random noise, and the signal-to-noise ratio (SNR) and peak SNR of the denoised signal are higher than those of conventional methods.

基于深度卷积神经网络的地震数据溶洞识别

Identification of Karst caves in seismic data based on deep convolutional neural network

【摘要】溶洞识别对于缝洞型油气藏的勘探与开发具有重要意义。传统溶洞识别方法多解性强且效率低，因此将具有强特征学习能力、高泛化性的深度学习方法引入溶洞识别中，但溶洞的地震波场响应特征复杂、异常体尺寸较小、训练样本难以获取等导致深度学习在识别溶洞时仍具挑战性。为此，提出一套识别地震数据溶洞的“两步法”深度学习方法：首先通过U-Net模型识别地震剖面上的“串珠状”异常反射；再根据“串珠状”异常识别结果对地震数据进行小范围截取，输入深度残差网络中，实现对实际溶洞轮廓的预测。对于实际溶洞预测训练数据难以获取这一问题，采用波动方程正演模拟的方法制作具有准确标签的溶洞地震数据。实际地震数据的应用表明，该方法对于溶洞识别准确性高，抗噪能力强，可以极大地节约人工解释成本。

【Abstract】Karst cave identification is significant for the exploration and development of fracture-cavity oil and gas reservoirs, but conventional identification methods are multi-solution and inefficient. Therefore, a deep learning method with strong feature learning and generalization capabilities is introduced into karst cave identification. However, it is still a challenging task to identify karst caves by deep learning due to the complex response characteristics of karst caves to the seismic wavefield, the small sizes of anomalies, and the difficulties in obtaining training samples. Faced with these problems, we propose a “two-step” deep learning method for identifying karst caves in seismic data. Specifically, the U-Net model is used to identify the “bead-shaped” anomalous reflection on a seismic profile. Then, according to the identification results of the “bead-shaped” anomalies, seismic data are cropped into small patches and input into the deep residual network to implement the prediction of the actual karst cave contour. Considering the difficulties in obtaining training data for actual karst cave prediction, we propose implementing wave-equation forward modeling to generate seismic karst cave data with accurate labels. The application of field seismic data shows that the method is accurate in karst cave identification with strong noise resistance, and it can greatly save the cost of manual interpretation.

期刊推荐Recommend Journal

《石油地球物理勘探》面向国内外的石油勘探、地矿、煤田、计算机数字处理等的教学、科研、生产等行业和单位，及时传播物探技术信息，推广新技术、新经验，促进物探领域的科技进步。

主要报道内容：

（1）各种地球物理勘探方法（地震勘探、重磁勘探、电法勘探、井中地球物理测试）的新理论、新技术、新工艺和新经验；

（2）各种地球物理勘探方法的应用新成果及典型实例；

（3）地球物理数据处理新方法；

（4）物探技术方面软件开发与应用；

（5）地球物理的评述、讨论和论坛；

（6）我国知名地球物理学家的生平事迹介绍；

（7）国内外地球物理技术发展动态及学术活动。

联系我们Contact Us

中国知网“中文精品学术期刊外文版数字出版工程”（简称JTP）自2015年启动，已与400余种学术期刊合作出版了4万余篇双语对照论文，积累了丰富的学术翻译/英语加工/学术推广经验。形成了集双语出版、主题电子书出版、双语讲座视频制作、期刊英文内容编校加工、资讯编译、海外推广为一体的全方位服务体系，全面助力期刊提升国际影响力。

联系人：张老师13661148416；010-62969002-8173；zxd6974@cnki.net

扫码二维码

添加张老师微信

了解更多双语信息