Layer | Fill | Outline |
---|
Map layers
Theme | Visible | Selectable | Appearance | Zoom Range (now: 0) |
---|
Fill | Stroke |
---|---|
Collaborating Authors
Machine Learning
Submarine cable positioning using residual convolutional neural network based on magnetic features
Liu, Yutao (Chinese Academy of Sciences) | Wu, Yuquan (Chinese Academy of Sciences) | Huang, Liang (Zhejiang University) | Yang, Lei (Zhejiang Institute of Marine Geology Survey) | Kuang, Jianxun (Zhejiang Qiming Offshore Power Co., Ltd) | Yu, Wenjie (Zhejiang Qiming Offshore Power Co., Ltd) | Wang, Jianqiang (Zhejiang Institute of Hydrogeology and Engineering Geology) | Xu, Zhe (Zhejiang Engineering Survey and Design Institute Group Co., Ltd) | Li, Gang (Zhejiang University, Zhejiang University)
Accurate positioning is important for improving the efficiency of repairing submarine cables and reducing the related repairing cost. The magnetic anomaly produced by a submarine cable can be used to estimate its vertical and horizontal positions. A novel approach using magnetic data for estimating the position of submarine cables based on the 1-D residual convolutional neural network (RCNN) is investigated. Infinitely long ferromagnetic cylinder models with different parameters are used to generate datasets for the model training and testing. Tests on noisy synthetic datasets show that the developed 1-D RCNN method can capture detailed features related to the magnetic source position information, which are more accurate than the conventional Euler method in estimating the position of submarine cables. The developed 1-D RCNN method has also been successfully applied to processing field data. Furthermore, the processing workflow of our 1-D RCNN method is less noise sensitive compared with the conventional Euler method. The proposed 1-D RCNN method and its workflow open a new window for estimating the position of submarine cables using magnetic data.
On cost-efficient parallel iterative solvers for 3-D frequency-domain seismic multi-source viscoelastic anisotropic wave modeling
Ma, Guoqi (Khalifa University of Science and Technology) | Zhou, Bing (Khalifa University of Science and Technology) | Riahi, Mohamed Kamel (Khalifa University of Science and Technology) | Zemerly, Jamal (Khalifa University of Science and Technology) | Xu, Liu (King Fahd University of Petroleum and Minerals)
Solving large sparse linear systems in 3-D frequency-domain seismic wave modeling, especially in viscoelastic anisotropic media, poses significant challenges due to the increasing number of discrete moduli and nonzero elements in the linear system matrix. The computational load surpasses that of acoustic or viscoacoustic media, making it even more challenging when dealing with multi-source problems. Popular scientific tools for solving a linear system like MUMPS, STRUMPACK, and PETSc can be utilized, but their applicability to our specific problem has not been comprehensively evaluated. Our study aims at tackling the challenges in solving large sparse, complex-valued symmetric linear systems with multiple right-hand-side vectors for 3-D frequency-domain seismic wave modeling. We have leveraged preconditioned conjugate gradient iterative algorithms as the foundation for our research, introducing two highly cost-effective parallel iterative solvers: the Parallel Symmetric Successive Over-Relaxation Conjugate Gradient (P-SSORCG) and the Parallel Incomplete Cholesky Conjugate Gradient (P-ICCG). These novel solvers were subjected to a comprehensive comparative analysis against well-established scientific tools, including MUMPS, STRUMPACK, and PETSc, in the context of 3-D frequency-domain seismic wave modeling. We show their promising performances in a practical 3-D SEG/EAGE overthrust model and demonstrate that the grouped P-SSORCG offers an efficient alternative to parallel direct solvers, particularly in situations where computational resources are limited.
Monitoring stored CO2 in carbon capture and storage projects is crucial for ensuring safety and effectiveness. We introduce DeepNRMS, a novel noise-robust method that effectively handles time-lapse noise in seismic images. The DeepNRMS leverages unsupervised deep learning to acquire knowledge of time-lapse noise characteristics from pre-injection surveys. By utilizing this learned knowledge, our approach accurately discerns CO2-induced subtle signals from the high-amplitude time-lapse noise, ensuring fidelity in monitoring while reducing costs by enabling sparse acquisition. We evaluate our method using synthetic data and field data acquired in the Aquistore project. In the synthetic experiments, we simulate time-lapse noise by incorporating random near-surface effects in the elastic properties of the subsurface model. We train our neural networks exclusively on pre-injection seismic images and subsequently predict CO2 locations from post-injection seismic images. In the field data analysis from Aquistore, the images from pre-injection surveys are utilized to train the neural networks with the characteristics of time-lapse noise, followed by identifying CO2 plumes within two post-injection surveys. The outcomes demonstrate the improved accuracy achieved by the DeepNRMS, effectively addressing the strong time-lapse noise.
- 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)
- Reservoir Description and Dynamics > Storage Reservoir Engineering > CO2 capture and sequestration (1.00)
- Reservoir Description and Dynamics > Reservoir Characterization > Seismic processing and interpretation (1.00)
- Reservoir Description and Dynamics > Formation Evaluation & Management > Seismic (four dimensional) monitoring (1.00)
- (2 more...)
Multichannel deconvolution with a high-frequency structural regularization
Wang, Pengfei (China University of Petroleum) | Zhao, Dongfeng (China University of Petroleum) | Niu, Yue (National Engineering Research Center for Oil and Gas Exploration Computer Software) | Li, Guofa (China University of Petroleum) | Gu, Weiwei (China University of Petroleum)
The resolution of seismic data determines the ability to characterize stratigraphic features from observed seismic record. Sparse spike inversion (SSI) as an important processing method can effectively improve the band-limited property of the seismic data. However, the approch ignores the spatial information along seismic traces, which causes the unreliability of the reconstructed high-resolution data. In this article, we have developed a high-frequency structure constrained multichannel deconvolution (HFSC-MD) to alleviate this issue. This method allows the cost function to incorporate high-frequency spatial information in the form of prediction-error filter (PEF), to regularize the components of the result beyond the original frequency. The PEF also called high-frequency structural characterization operator (HFRSC operator), is estimated from the mapping relationship of low and high-frequency components. We adopt the alternating direction method of multipliers (ADMM) to solve the cost function in HFSC-MD. Synthetic and field data demonstrate that the proposed method recovers more reliable high-resolution data, and enriches the reflective structures.
- Geophysics > Seismic Surveying > Seismic Processing (1.00)
- Geophysics > Seismic Surveying > Seismic Modeling > Velocity Modeling > Seismic Inversion (0.34)
- Information Technology > Artificial Intelligence > Representation & Reasoning (0.54)
- Information Technology > Artificial Intelligence > Machine Learning (0.46)
Accurate production forecasting plays a pivotal role in understanding and effectively developing reservoirs. Numerous research shows that machine learning could be used to achieve fast and precise production predictions. In this three-part article, we use long short-term memory (LSTM), a machine learning technique, to predict oil, gas, and water production using real field data. This first part discusses the mathematics behind the LSTM and part two and three focus on its practical implementation. Recurrent neural network (RNN) and its variants, namely LSTM and gated recurrent unit (GRU), are specifically tailored for time-dependent data.
Simultaneous source acquisition has become common over the past few decades for marine seismic surveys because of the increased efficiency of seismic acquisition by limiting the time, reducing the cost, and having less environmental impact than conventional single-source (or unblended) acquisition surveys. For simultaneous source acquisition, seismic sources at different locations are fired with time delays, and the recorded data are referred to as the blended data. The air-water interface (or free surface) creates strong multiples and ghost reflections for blended seismic data. The multiples and/or ghost reflections caused by a source in the blended data overlap with the primary reflections of another source, thus creating a strong interference between the primary and multiple events of different sources. We develop a convolutional neural network (CNN) method to attenuate free-surface multiples and remove ghost reflections simultaneously from the blended seismic data. The CNN-based solution that we develop operates on single traces and is not sensitive to the missing near-offset traces, missing traces, and irregular/sparse acquisition parameters (e.g.,for ocean-bottom node acquisition and time-lapse monitoring studies). We illustrate the efficacy of our free-surface multiple attenuation and seismic deghosting method by presenting synthetic and field data applications. The numerical experiments demonstrate that our CNN-based approach for simultaneously attenuating free-surface multiples and removing ghost reflections can be applied to the blended data without the deblending step. Although the interference of primaries and multiples from different shots in the blended data makes free-surface multiple attenuation harder than the unblended data, we determine that our CNN-based method effectively attenuates free-surface multiples in the blended synthetic and field data even when the delay time for the blending is different in the training data than in the data to which the CNN is applied.
- North America > United States > Illinois > Madison County (0.34)
- Europe > United Kingdom > North Sea (0.29)
- Europe > Norway > North Sea (0.29)
- North America > United States > Colorado (0.28)
Least-squares reverse-time migration of simultaneous source with deep-learning-based denoising
Wu, Bo (China University of Petroleum (Beijing), China University of Petroleum (Beijing), China University of Petroleum (Beijing)) | Yao, Gang (China University of Petroleum (Beijing), China University of Petroleum (Beijing), China University of Petroleum (Beijing)) | Ma, Xiao (China University of Petroleum (Beijing), China University of Petroleum (Beijing), China University of Petroleum (Beijing)) | Chen, Hanming (China University of Petroleum (Beijing), China University of Petroleum (Beijing), China University of Petroleum (Beijing)) | Wu, Di (China University of Petroleum (Beijing), China University of Petroleum (Beijing), China University of Petroleum (Beijing)) | Cao, Jingjie (Hebei Geo University)
Least-squares reverse-time migration (LSRTM) is currently one of the most advanced migration imaging techniques in the field of geophysics. It utilizes least-squares inversion to fit the observed data, resulting in high-resolution imaging results with more accurate amplitudes and better illumination compensation than conventional reverse-time migration (RTM). However, noise in the observed data and the Born approximation forward operator can result in high-wavenumber artifacts in the final imaging results. Moreover, iteratively solving LSRTM leads to one or two orders of computational cost higher than conventional RTM, making it challenging to apply extensively in industrial applications. Simultaneous source acquisition technology can reduce the computational cost of LSRTM by reducing the number of wavefield simulations. However, this technique can also cause high-wavenumber crosstalk artifacts in the migration results. To effectively remove the high-wavenumber artifacts caused by these mentioned issues, in this paper, we combine simultaneous source and deep-learning to speed up LSRTM, as well as, to suppress high-wavenumber artifacts. A deep-residual neural network (DR-Unet) is trained with synthetic samples, which are generated by adding field noise to synthesized noise-free migration images. Then, the trained DR-Unet is applied on the gradient of LSRTM to remove high-wavenumber artifacts in each iteration. Compared to directly applying DR-Unet denoising to LSRTM results, embedding DR-Unet denoising into the inversion process can better preserve weak reflectors and improve denoising effects. Finally, we tested the proposed LSRTM method on two synthetic datasets and a land dataset. The tests demonstrate that the proposed method can effectively remove high-wavenumber artifacts, improve imaging results, and accelerate convergence speed.
In recent years, large language models (LLMs) have become increasingly popular in natural language processing (NLP) research. The recent advancements in LLMs have revolutionized the field of natural language processing (NLP). LLMs are trained on massive amounts of text data and can generate human-like text with impressive accuracy. They have shown remarkable performance in various NLP tasks such as text classification, question answering, and language generation. Two of the most popular large language models are Generative Pre-trained Transformer 3 and 4 (GPT-3 & GPT-4), which were developed by OpenAI.
- Information Technology > Artificial Intelligence > Natural Language > Large Language Model (1.00)
- Information Technology > Artificial Intelligence > Natural Language > Chatbot (1.00)
- Information Technology > Artificial Intelligence > Machine Learning > Neural Networks > Deep Learning (1.00)
Join us for an SPE Live episode on "Innovation in Hydraulic Fracturing" with a preview of the SPE Hydraulic Fracturing Technology Conference and Exhibition on 6-8 February 2024, being held in the Woodlands, Texas – USA. A Q&A session will then follow with questions from the audience, answered by our guests.