Layer | Fill | Outline |
---|
Map layers
Theme | Visible | Selectable | Appearance | Zoom Range (now: 0) |
---|
Fill | Stroke |
---|---|
Collaborating Authors
Reservoir Description and Dynamics
Deep learning with soft attention mechanism for small-scale ground roll attenuation
Yang, Liuqing (China University of Petroleum (Beijing), China University of Petroleum (Beijing)) | Fomel, Sergey (The University of Texas at Austin) | Wang, Shoudong (China University of Petroleum (Beijing), China University of Petroleum (Beijing)) | Chen, Xiaohong (China University of Petroleum (Beijing), China University of Petroleum (Beijing)) | Chen, Yangkang (The University of Texas at Austin)
ABSTRACT Ground roll is a type of coherent noise with low frequency, low velocity, and high amplitude, which masks useful signals and decreases the quality of subsequent seismic data processing. It is a challenge for traditional signal processing methods to separate useful signals effectively when the ground roll and useful reflected signals overlap seriously in the low-frequency band. We develop a supervised-learning-based framework with soft attention residual learning mechanisms for suppressing the ground roll noise. To reduce the cost of manual labeling, the 2D patching technique is used to segment large-scale seismic data into a large number of small-scale patches for training. Our network includes a multibranch attention block that uses multiple branches with different kernel sizes to extract waveform features at different scales from input noisy patches. Then, we use the soft attention mechanism to select and fuse the feature maps of different branches. Our network can achieve encouraging ground roll attenuation performance by using a small number of training samples, which is demonstrated by synthetic and field data examples. Compared with one traditional method and two advanced deep-learning frameworks, our network has better abilities in preserving low-frequency useful signals and removing ground roll.
- Asia > China (0.68)
- North America > United States > Texas (0.28)
- Reservoir Description and Dynamics > Reservoir Characterization > Seismic processing and interpretation (1.00)
- Data Science & Engineering Analytics > Information Management and Systems > Neural networks (1.00)
Deep learning with soft attention mechanism for small-scale ground roll attenuation
Yang, Liuqing (China University of Petroleum (Beijing), China University of Petroleum (Beijing)) | Fomel, Sergey (The University of Texas at Austin) | Wang, Shoudong (China University of Petroleum (Beijing), China University of Petroleum (Beijing)) | Chen, Xiaohong (China University of Petroleum (Beijing), China University of Petroleum (Beijing)) | Chen, Yangkang (The University of Texas at Austin)
Ground roll is a type of coherent noise with low frequency, low velocity, and high amplitude, which masks useful signals and decreases the quality of subsequent seismic data processing. It is a challenge for traditional signal processing methods to separate useful signals effectively when the ground roll and useful reflected signals overlap seriously in the low-frequency band. We develop a supervised-learning-based framework with soft attention residual learning mechanisms for suppressing the ground roll noise. To reduce the cost of manual labeling, we use the 2D patching technique to segment large-scale seismic data into a large number of small-scale patches for training. The proposed network includes a multi-branch attention (MBA) block, which uses multiple branches with different kernel sizes to extract waveform features at different scales from input noisy patches. Then, we use the soft attention mechanism to select and fuse the feature maps of different branches. The proposed network can achieve encouraging ground roll attenuation performance by using a small number of training samples, which is demonstrated by both synthetic and field data examples. In addition, we compare the proposed network with one traditional method and two advanced deep-learning frameworks. The denoising results show that the proposed network has better abilities in preserving low-frequency useful signals and removing ground roll.
- Asia > China (0.67)
- North America > United States > Texas (0.27)
- Reservoir Description and Dynamics > Reservoir Characterization > Seismic processing and interpretation (1.00)
- Data Science & Engineering Analytics > Information Management and Systems > Neural networks (1.00)
SLKNet: An Attention-based Deep Learning Framework for Downhole Distributed Acoustic Sensing Data Denoising
Yang, Liuqing (China University of Petroleum (Beijing), China University of Petroleum (Beijing)) | Fomel, Sergey (The University of Texas at Austin) | Wang, Shoudong (China University of Petroleum (Beijing), China University of Petroleum (Beijing)) | Chen, Xiaohong (China University of Petroleum (Beijing), China University of Petroleum (Beijing)) | Chen, Yunfeng (Zhejiang University) | Chen, Yangkang (The University of Texas at Austin)
Distributed acoustic sensing (DAS) is a novel and fast-developing seismic acquisition technology, which enjoys many advantages compared with traditional geophones. However, DAS data often suffer from severe and diverse types of noise with varying amplitudes, resulting in a low signal-to-noise ratio (SNR) and making the extraction of hidden signals a challenging task. Therefore, exploring a high-efficiency and high-generalization denoising method is crucial for improving the SNR of DAS data and subsequent processing. We propose a dense connection network with the kernel-wise attention mechanism to denoise complex and diverse noise (e.g., high-amplitude erratic, high-frequency, random, and horizontal noise) on real DAS datasets. We use an integrated denoising framework that is suitable for attenuating DAS noise to generate labels for network training. The proposed network consists of five types of blocks, i.e., convolutional, dense, transition down, transition up, and selective kernel blocks. In particular, the selective kernel block is used to fuse multi-scale features by weighting, thereby improving the denoising accuracy. The computational efficiency and denoising performance are further augmented by employing a patching method to segment the DAS data and generate many small-scale patches. The proposed network is trained on a small DAS data set and tested on the synthetic and field data from vastly different geographic areas. The comparisons of our network with three state-of-the-art deep-learning-based benchmark models demonstrate more robust performance and superior signal extraction ability.
- Asia > China (0.92)
- North America > United States > Texas (0.67)
- Reservoir Description and Dynamics > Reservoir Characterization (1.00)
- Production and Well Operations > Well & Reservoir Surveillance and Monitoring > Production logging (1.00)
- Data Science & Engineering Analytics > Information Management and Systems > Neural networks (1.00)
Warped P-SV wavelet distortion correction using a time-frequency adaptive shaping filter
Geng, Weiheng (China University of Petroleum-Beijing) | Chen, Xiaohong (China University of Petroleum-Beijing) | Li, Jingye (China University of Petroleum-Beijing) | Wang, Jianhua (CNOOC Research Institute Co., Ltd) | Wu, Fan (China University of Petroleum-Beijing) | Tang, Wei (China University of Petroleum-Beijing) | Zhang, Junjie (China University of Petroleum-Beijing)
ABSTRACT To perform joint PP/PS amplitude-variation-with-angle inversion or attribute analysis, we usually map converted-wave (PS) data to the compressional wave (PP) time domain to guarantee that the reflection events from similar reflectors have the same two-way traveltimes. Mapping PS data to the PP time domain leads to a nonstationary characteristic of the PS data. Nonstationary here means varying spectral content of the wavelet in PS data, which will degrade inversion results. We develop a zero-phase time-frequency adaptive shaping filter to filter the nonstationary PS data to be stationary. This adaptive filter can correct the distorted wavelets in the PS data without introducing numerical noise and artifacts. In addition, we use the dynamic time warping algorithm to estimate the time shifts between the PP and PS data and the local seismic attribute estimation method to estimate the subsurface ratios from the time shifts. The estimated ratios will be used to calculate the time- and location-varying wavelets by the Fourier scaling theorem, and then the adaptive shaping filter is constructed. Based on the synthetic and field data examples, we determine the better performance of the adaptive shaping filter over the conventional shaping filter in attenuating numerical noise and artifacts.
- Asia > China (1.00)
- North America > United States > Texas (0.28)
- Asia > China > Xinjiang Uyghur Autonomous Region > Junggar Basin (0.99)
- Asia > China > Sichuan > Sichuan Basin (0.99)
Porosity and permeability prediction using a transformer and periodic long short-term network
Yang, Liuqing (China University of Petroleum (Beijing), China University of Petroleum (Beijing)) | Fomel, Sergey (The University of Texas at Austin) | Wang, Shoudong (China University of Petroleum (Beijing), China University of Petroleum (Beijing)) | Chen, Xiaohong (China University of Petroleum (Beijing), China University of Petroleum (Beijing)) | Chen, Wei (Yangtze University) | Saad, Omar M. (NRIAG) | Chen, Yangkang (The University of Texas at Austin)
ABSTRACT Effective reservoir parameter prediction is important for subsurface characterization and understanding fluid migration. However, conventional methods for obtaining porosity and permeability are based on either core measurements or mathematical/petrophysical modeling, which are expensive or inefficient. In this study, we develop a reliable and low-cost deep learning (DL) framework for reservoir permeability and porosity prediction from real logging data at different regions. We leverage an advanced learning architecture (i.e.,ย the transformer model) and design a new regression network (RPTransformer) that is sensitive to the depth period change of the logging data. The RPTransformer is composed of 1D convolutional, long short-term memory (LSTM), and transformer layers. First, we use a 1D convolutional layer for the first layer of the network to extract significant features from the logging data. Then, the nonlinear mapping relationships between logging data and reservoir parameters are established using several LSTM layers with a period parameter. Afterward, we use the encoder in the vision transformer with the self-attention mechanism to further extract logging data features. The developed network is a data-driven supervised learning framework and indicates highly accurate and robust prediction results when applied to different geographic regions. To demonstrate the reliable prediction performance of our network, we compare it with several classic machine learning and state-of-the-art DL methods, e.g.,ย random forest, multilayer LSTM, and long short-term time-series network (LSTNet). More importantly, we find the generalization and uncertainty of the network in real-world applications through comprehensive numerical experiments.
- Asia (0.68)
- North America > United States > Texas (0.46)
- Africa > Middle East > Libya > Murzuq District (0.16)
- Geology > Structural Geology > Tectonics (0.47)
- Geology > Geological Subdiscipline > Economic Geology (0.34)
- Geophysics > Borehole Geophysics (1.00)
- Geophysics > Seismic Surveying > Passive Seismic Surveying (0.93)
- Geophysics > Seismic Surveying > Seismic Modeling > Velocity Modeling > Seismic Inversion (0.46)
- Oceania > Australia > Queensland > Surat Basin (0.99)
- Oceania > Australia > New South Wales > Surat Basin (0.99)
- North America > Canada > Nova Scotia > North Atlantic Ocean > Scotian Basin > Sable Basin > Sable Project > Venture Field (0.99)
- (6 more...)
- Reservoir Description and Dynamics > Reservoir Fluid Dynamics > Flow in porous media (1.00)
- Reservoir Description and Dynamics > Reservoir Characterization (1.00)
- Reservoir Description and Dynamics > Formation Evaluation & Management > Open hole/cased hole log analysis (1.00)
- Data Science & Engineering Analytics > Information Management and Systems > Artificial intelligence (1.00)
Denoising of distributed acoustic sensing data using supervised deep learning
Yang, Liuqing (China University of Petroleum (Beijing), China University of Petroleum (Beijing)) | Fomel, Sergey (The University of Texas at Austin) | Wang, Shoudong (China University of Petroleum (Beijing), China University of Petroleum (Beijing)) | Chen, Xiaohong (China University of Petroleum (Beijing), China University of Petroleum (Beijing)) | Chen, Wei (Yangtze University) | Saad, Omar M. (NRIAG) | Chen, Yangkang (The University of Texas at Austin)
ABSTRACT Distributed acoustic sensing (DAS) is an emerging technology for acquiring seismic data due to its high-density and low-cost advantages. Because of the harsh acquisition environment and other unexpected reasons, the seismic signals acquired in DAS are masked by various types of complex noise, which seriously decreases the signal-to-noise ratio of seismic data. We propose a fully convolutional neural network with dense and residual connections to attenuate complex noise in DAS data. The network is designed to learn features of useful reflection signals recorded from a large number of earthquake and microseismic events, aiming at obtaining an unprecedented generalization ability. First, we generate labels using an integrated framework that attenuates specific types of noise in real DAS data, where the integrated framework includes carefully designed band-pass, structure-oriented median, and dip filters. Then, we use the patching technique to segment the training samples into many small-scale patches to reduce computational cost and improve the extraction of essential features from large-scale passive seismic data. Finally, we use the well-trained network to estimate the heavily polluted hidden signals. Compared with two advanced deep-learning methods and a traditional denoising framework, our proposed method can more effectively attenuate strong and complex noise and recover weak hidden signals in synthetic and real DAS data tests.
- Reservoir Description and Dynamics > Reservoir Characterization > Seismic processing and interpretation (1.00)
- Production and Well Operations > Well & Reservoir Surveillance and Monitoring > Production logging (1.00)
- Data Science & Engineering Analytics > Information Management and Systems > Neural networks (1.00)
Sparse Radon transform in the mixed frequency-time domain with โ1-2 minimization
Geng, Weiheng (China University of Petroleum-Beijing) | Chen, Xiaohong (China University of Petroleum-Beijing) | Li, Jingye (China University of Petroleum-Beijing) | Ma, Jitao (China University of Petroleum-Beijing) | Tang, Wei (China University of Petroleum-Beijing) | Wu, Fan (China University of Petroleum-Beijing)
ABSTRACT Due to the finite acquisition aperture and sampling of seismic data, the Radon transform (RT) suffers from a smearing problem which reduces the resolution of the estimated model. In addition, inverting the RT is typically an ill-posed problem. To address these challenges, a sparse RT mixing the and norms of the RT coefficients in the mixed frequency-time domain is developed, and it is denoted as SRTL1-2. In most conventional sparse RTs, the sparse constraint term often is the norm of the Radon model. We prove that the sparsity effect of the minimization is better than that of the norm alone by comparing and analyzing their 2D distribution patterns and threshold functions. The difference of the convex functions algorithm and the alternating direction method of multipliers algorithm are modified by combining the forward and inverse Fourier transforms to solve the corresponding sparse inverse problem in the mixed frequency-time domain. Our method is compared with three RT methods, including a least-squares RT (LSRT), a frequency-domain sparse RT (FSRT), and a time-invariant RT in the mixed frequency-time domain based on an iterative 2D model shrinkage method (SRTIS). Furthermore, we modify the basis function in SRTL1-2 by including an orthogonal polynomial transform to fit the amplitude-variation-with-offset (AVO) signatures found in seismic data, and we denote this as high-order SRTL1-2. Compared to the SRTL1-2, the high-order SRTL1-2 performs better when processing seismic data with AVO signatures. Synthetic and real data examples indicate that our method has better performance than the LSRT, FSRT, and SRTIS in terms of attenuation of multiples, noise mitigation, and computational efficiency.
ABSTRACT The prediction of elastic parameters (i.e.,ย P-, S-wave velocity, and density) is one of the key tasks of seismic reservoir characterization. The amplitude-variation-with-offset/angle seismic inversion based on the exact Zoeppritz equation (EZE) or its approximations presupposes a single interface and ignores wave-propagation effects, resulting in low accuracy inversion results. The analytical solution of the 1D wave equation (i.e.,ย the reflectivity method [RM]) can simulate more of the totality of wave-propagation effects, including transmission losses and internal multiples, thus improving the accuracy of the inversion results. However, the RM-based inversion method is sensitive to noise and is usually performed trace-by-trace. When trace-by-trace-based inversion results are combined into a 2D profile, the lateral continuity of the final results is poor, which affects the subsequent interpretation and evaluation. To address these issues, the RM-based structure-oriented prestack waveform inversion (SORM) method is proposed to suppress the effects of data noise and improve the geologic reliability of the inversion results. This method adds an additional structure-oriented constraint term to the objective function, which facilitates the integration of the structural orientation into the inversion algorithm in the form of dips. We carry out the method on a synthetic model as well as on a field data set. A series of numerical tests indicate that the SORM gives significantly more accurate and geologically reliable results compared with inversion based on EZE or trace-by-trace RM.
- Geology > Geological Subdiscipline (0.46)
- Geology > Rock Type (0.46)
- Reservoir Description and Dynamics > Reservoir Characterization > Seismic processing and interpretation (1.00)
- Reservoir Description and Dynamics > Reservoir Characterization > Seismic modeling (1.00)
ABSTRACT Seismic image registration is crucial for the joint interpretation of multivintage seismic images in time-lapse reservoir monitoring. Time-shift analysis is a commonly used method to estimate the warping function by creating a time-shift map, where the energy of each time-shift point in the 3D map indicates the probability of a correct registration. We have adopted a new method to obtain a high-resolution time-shift analysis spectrum, which can help with manual and automatic picking. The time-shift scan map is obtained by trying different local shifts and calculating the local similarity attributes between the shifted and reference images. We adopt a high-resolution calculation of the time-shift scan map by applying the nonstationary model constraint in solving the local similarity attributes. The nonstationary model constraint ensures the time-shift scan map to be smooth in all physical dimensions, for example, time, local shift, and space. In addition, it permits variable smoothing strength across the whole volume, which enables the high resolution of the calculated time-shift scan map. We use an automatic-picking algorithm to demonstrate the accuracy of the high-resolution time-shift scan map and its positive influence on the time-lapse image registration. Synthetic (2D) and real (3D) time-lapse seismic images are used for demonstrating the better registration performance of the proposed method.
- Asia > China (0.68)
- North America > United States (0.46)
- Geophysics > Time-Lapse Surveying > Time-Lapse Seismic Surveying (1.00)
- Geophysics > Seismic Surveying > Seismic Processing (1.00)
- Geophysics > Seismic Surveying > Seismic Modeling > Velocity Modeling (0.93)
- Europe > Norway > North Sea > Central North Sea > Central Graben > Block 2/8 > Valhall Field > Tor Formation (0.99)
- Europe > Norway > North Sea > Central North Sea > Central Graben > Block 2/8 > Valhall Field > Hod Formation (0.99)
- Europe > Norway > North Sea > Central North Sea > Central Graben > Block 2/11 > Valhall Field > Tor Formation (0.99)
- (2 more...)
- Reservoir Description and Dynamics > Reservoir Characterization > Seismic processing and interpretation (1.00)
- Reservoir Description and Dynamics > Reservoir Characterization > Four-dimensional and four-component seismic (1.00)
- Reservoir Description and Dynamics > Formation Evaluation & Management > Seismic (four dimensional) monitoring (1.00)
Prestack waveform inversion based on analytical solution of the viscoelastic wave equation
Li, Yuanqiang (China University of Petroleum (Beijing), China University of Petroleum (Beijing)) | Li, Jingye (China University of Petroleum (Beijing), China University of Petroleum (Beijing)) | Chen, Xiaohong (China University of Petroleum (Beijing), China University of Petroleum (Beijing)) | Zhang, Jian (China University of Petroleum (Beijing), China University of Petroleum (Beijing)) | Bo, Xin (China University of Petroleum (Beijing), China University of Petroleum (Beijing))
ABSTRACT Amplitude-variation-with-offset (AVO) inversion is based on single interface reflectivity equations. It involves some restrictions, such as the small-angle approximation, including only primary reflections, and ignoring attenuation. To address these shortcomings, the analytical solution of the 1D viscoelastic wave equation is used as the forward modeling engine for prestack inversion. This method can conveniently handle the attenuation and generate the full wavefield response of a layered medium. To avoid numerical difficulties in the analytical solution, the compound matrix method is applied to rapidly obtain the analytical solution by loop vectorization. Unlike full-waveform inversion, the proposed prestack waveform inversion (PWI) can be performed in a target-oriented way and can be applied in reservoir study. Assuming that a Q value is known, PWI is applied to synthetic data to estimate elastic parameters including compressional wave (P-wave) and shear wave (S-wave) velocities and density. After validating our method on synthetic data, this method is applied to a reservoir characterization case study. The results indicate that the reflectivity calculated by our approach is more realistic than that computed by using single interface reflectivity equations. Attenuation is an integral effect on seismic reflection; therefore, the sensitivity of seismic reflection to P-and S-wave velocities and density is significantly greater than that to Q, and the seismic records are sensitive to the low-frequency trend of Q. Thus, we can invert for the three elastic parameters by applying the fixed low-frequency trend of Q. In terms of resolution and accuracy of synthetic and real inversion results, our approach performs superiorly compared to AVO inversion.
- Reservoir Description and Dynamics > Reservoir Characterization > Seismic processing and interpretation (1.00)
- Reservoir Description and Dynamics > Reservoir Characterization > Seismic modeling (1.00)