Seismic Inversion Research Articles

Integrating machine learning with simultaneous quantile regression for uncertainty estimation

Accurate velocity models are crucial for understanding subsurface structures, which is vital for various subsurface activities including hydrocarbon exploration, development, and CO2 storage. The application of data-driven techniques based on machine learning (ML) to velocity model building has garnered significant attention due to their computational efficiency. However, most existing ML implementations provide only a single prediction for a given input, which may not adequately reflect the distribution of the testing data. The simultaneous quantile regression method can estimate the entire conditional distribution of the target variable given the input using pinball loss. In this study, we incorporate this technique into seismic inversion and test the proposed method on synthetic Kimberlina data. The uncertainty map is then calculated pixel by pixel from a particular prediction interval around the median. We also introduce an innovative data-augmentation method that leverages uncertainty to enhance prediction accuracy. The results validate the reliability of the calculated uncertainty in the proposed InvNet_UQ model. The method remains robust, even if the testing data are distorted due to problems in the field data acquisition. Another test demonstrates the effectiveness of the developed data-augmentation method in increasing the spatial resolution of the estimated velocity field and in reducing the prediction error.

Geophysical-guided Wasserstein cycle-consistent generative adversarial networks for seismic impedance inversion

Geophysical-guided Wasserstein cycle-consistent generative adversarial networks for seismic impedance inversion

Volumetric carbon storage capacity estimation at Mississippi Canyon Block 118 in the Gulf of Mexico using Post-Stack Seismic Inversion

Volumetric carbon storage capacity estimation at Mississippi Canyon Block 118 in the Gulf of Mexico using Post-Stack Seismic Inversion

Improving the Seismic Impedance Inversion by Fully Convolutional Neural Network

Applying deep neural networks (DNNs) to broadband seismic wave impedance inversion is challenging, especially in generalizing from synthetic to field data, which limits the exploitation of their nonlinear mapping capabilities. While many research studies are about advanced and enhanced architectures of DNNs, this article explores how variations in input data affect DNNs and consequently enhance their generalizability and inversion performance. This study introduces a novel data pre-processing strategy based on histogram equalization and an iterative testing strategy. By employing a U-Net architecture within a fully convolutional neural network (FCN) exclusively trained on synthetic and monochrome data, including post-stack profile, and 1D linear background impedance profiles, we successfully achieve broadband impedance inversion for both new synthetic data and marine seismic data by integrating imaging profiles with background impedance profiles. Notably, the proposed method is applied to reverse time migration (RTM) data from the Ceduna sub-basin, located in offshore southern Australia, significantly expanding the wavenumber bandwidth of the available data. This demonstrates its generalizability and improved inversion performance. Our findings offer new insights into the challenges of seismic data fusion and promote the utilization of deep neural networks for practical seismic inversion and outcomes improvement.

CO2 characterization using seismic inversion based on global optimization techniques for enhanced reservoir understanding: a comparative study

CO2 characterization using seismic inversion based on global optimization techniques for enhanced reservoir understanding: a comparative study

Petroelastic modeling of well logging and seismic inversion data combining rock physics templates and conventional petrophysical analysis for 1D, 2D, and 3D supervised reservoir characterization of the Stybarrow Field in Exmouth sub-basin inner Carnarvon Basin, Australia

Petroelastic modeling of well logging and seismic inversion data combining rock physics templates and conventional petrophysical analysis for 1D, 2D, and 3D supervised reservoir characterization of the Stybarrow Field in Exmouth sub-basin inner Carnarvon Basin, Australia

Reducing the uncertainty in the distribution of cm-scale rock properties in the near well-bore region

Rock properties at cm-scale impact geological carbon storage by enhancing capillary and mineral trapping. Hence, it is important to accurately capture their distribution in geo-models which are used for numerically estimating the fate of injected CO2. However, there could be high variability in the cm-scale distribution of rock properties even close to wells which is not captured with traditional workflows. This study explores the impact of grid cell resolution, seismic inversion and placement of an additional well in proximity to CO2 injection well on improving the representation of cm-scale lithological heterogeneity in the near well bore region in geological models. We utilize wireline and seismic data from Parasequence-2 of the Paaratte Formation, Otway Basin, Australia, which is a shallow to coastal marine deltaic deposition comprising a high degree of lithological heterogeneity and a prospective unit for pilot scale geological carbon storage operations. The data was used to prepare a suite of reservoir models capturing the impact of above factors on the plausible distributions of facies, porosity and permeability in the formation. The analysis suggests that smaller grid cell size (1 m × 1 m × 0.3 m) compared to the typical industry standard (10 m × 10 m × 2 m) significantly improves the representation of cm-scale rock properties. Additionally, stochastic seismic inversion could play an important role in capturing rock property distribution even for smaller CO2 storage sites used for pilot scale injection operations. Further, we show that the placement of an additional well only 116-m away from the CO2 injection well can drastically improve the probability in the distribution of cm-scale rock properties in reservoir models.

The Main Controlling Factors of Coalbed Methane Productivity Based on Reservoir Structure—A Case Study of the Jiaozuo Block

In the process of CBM development, the fracturing effect has always been a major controlling factor for CBM productivity. The coal fragmentation degree is a special geological feature in the process of CBM development and research, and other types of reservoirs are not involved in this study. This paper addresses the problem of the inaccurate prediction of the reservoir fragmentation degree by studying the influence of the reservoir type and depth plane curvature on the reservoir fragmentation degree based on the coalbed characteristics of a block. It also studies the influence of faults on the reservoir fragmentation degree based on the reservoir geological characteristics and seismic inversion results. Combined with dynamic data on coalbed methane production, the influence of different geological characteristics on the productivity of coalbed methane wells is studied. The research results show that the reservoir fragmentation degree is mainly affected by the reservoir type. In the coal-forming period or after coal forming, the stronger the tectonic movement is, the higher the reservoir fragmentation degree is. Another manifestation of tectonic movement is faults. The effect of the reservoir fragmentation degree on production is negative. The better the reservoir fragmentation degree is, the worse the reconstruction effect of the coalbed methane well is, and the worse the later production effect is. At the same time, the faults generated by tectonic movement affect not only the reservoir fragmentation degree but also the water production of coalbed methane wells. The closer a well is to a fault, the greater the risk is of high water production and low gas production. Therefore, in the process of selecting a desert area, a complex reservoir fragmentation degree and areas with strong tectonic movement should be avoided. This study takes a structural control block as the research object to study the main controlling factors of coalbed methane reservoir productivity in complex structures. At present, there is no relevant research on this structure in terms of controlling productivity at home or abroad. The research in this paper can provide technical support for the development of similar CBM reservoirs. This method can guide the development of coalbed methane fields and lay a foundation for the selection of favorable coalbed methane reservoir areas.

Gabor-wavelet-activation implicit neural learning for full waveform inversion

Seismic full waveform inversion (FWI) is an efficient imaging method for estimating subsurface physical parameters. However, when the initial models are inaccurate or seismic data lack low frequencies, most traditional discrete grid-based FWI algorithms often encounter local minimum problems. Additionally, the inversion performance and computational cost are closely related to spatial resolution. Reparameterizing velocity models using neural networks (NNs) effectively mitigates local minimum issues. However, most NN-FWI approaches perform well when overparameterized and are limited to fixed grid-like representations. In contrast, implicit neural learning (INL) enables the representation of models at any resolution using a multilayer perception (MLP) network that learns a continuous function from discrete coordinates. To enhance the capability of INL, we developed a general Gabor-wavelet-activation INL approach and applied it to FWI, referred to as wavelet implicit neural learning FWI (WinFWI), using only the vertical particle-velocity. The constructed MLP was lightweight, which reduced the computational overhead. Numerical experiments on a synthetic block model and Marmousi2 model demonstrate the robustness of the proposed method to challenges such as parameter crosstalk. This was further validated using the Chevron 2014 blind test data. All comparisons show that our method is more general and robust than traditional and INL-based FWI algorithms. Moreover, additional feature visualization and numerical analysis illustrate the potential advantages of WinFWI in balancing computational cost and inversion accuracy.

Impact of Tectono-sedimentation Factor on Prospects of Oil and Gas Potential of Sakhalin Offshore of Okhotsk Oil and Gas Province Established through Stochastic Seismic Data Inversion and Constructed Digital Paleotectonic Model

Impact of Tectono-sedimentation Factor on Prospects of Oil and Gas Potential of Sakhalin Offshore of Okhotsk Oil and Gas Province Established through Stochastic Seismic Data Inversion and Constructed Digital Paleotectonic Model

Seismic inversion for CO2 volume monitoring and comprehensive evaluation of pore fluid properties: a case study

Seismic inversion for CO2 volume monitoring and comprehensive evaluation of pore fluid properties: a case study

Azimuthal Seismic Inversion for Fractured Reservoirs With Orthorhombic Symmetry Based on a Modified Reflection Coefficient Approximation

Azimuthal Seismic Inversion for Fractured Reservoirs With Orthorhombic Symmetry Based on a Modified Reflection Coefficient Approximation

A research framework for seismic slip distribution inversion considering atmospheric effects and deformation field downsampling

ABSTRACT Traditional Interferometric Synthetic Aperture Radar (InSAR) methods have difficulties in capturing deformation information of small and medium-sized earthquakes due to small surface deformation and atmospheric interference. This study proposes a research framework for seismic slip distribution inversion considering atmospheric effects and deformation field downsampling. The research framework begins by generating multiple interferometric pairs, using the coherence thresholding method to select stable points as reference during phase unwrapping. This step includes averaging the unwrapped pairs post-phase superposition to mitigate atmospheric turbulence effects in the interferograms and effectively extract the faint coseismic deformation field. Additionally, the framework employs a saliency-based quadtree downsampling algorithm to differentiate between major and minor deformation zones, thereby streamlining the data associated with the deformation field. The Bayesian method is then used to determine the geometric parameters of the earthquake fault, with the steepest descent method (SDM) applied to invert the fault slip distribution. The methodology was tested using the March 20, 2020, Ms5.9 earthquake in Dingri County, Tibet, China as a case study. Results demonstrate that this approach reduces atmospheric influences in the deformation field and enhances the accuracy of fault slip inversion, showing high consistency with existing studies and validating the reliability of the proposed method.

Seismic characterization of carbonate stringers using machine learning techniques: an example from the western flank of South Oman Salt Basin

Carbonate stringers are defined as a slab of carbonate bodies encased inside salt. In Oman, the intra-salt carbonate stringers are a very common target, especially in South Oman Salt Basin (SOSB). These stringers contain a large amount of hydrocarbon resources. In this study, the behavior of carbonate stringer bodies hosted in Ara salt structures from the SOSB is investigated. Both synthetic and real seismic data are use. A geological model is created from the structural interpretation of a newly acquired and processed seismic data from the area. The synthetic model is used to generate different seismic and elastic quantities such as P-wave, S-wave, density, synthetic gathers, and stack seismic data. Different attributes are computed from the synthetic gathers before inverting them to acoustic impedance, shear impedance, and density. In order to help us detect carbonate stringers and predict their distribution,, namely, Artificial Neural Network (ANN), is deployed to combine all the computed attributes into one probability cube. The cube highlights only stringers detected by the trained ANN. The ANN demonstrated a promising capability of detection of the targeted stringers. The above experiment is repeated on real prestack data after converting them to impedances using prestack seismic inversion. Likewise, the ANN was able to delineate the stringers and predict the distribution in the 3-D data. After several tests, a final 3-D distribution model of the stringers was obtained. The cube has allowed us to find new potential zones that might be good prospects in any future drilling in the area.

Characteristics of Donghe Sandstone Transport Conductor in Hudson Oilfield, Tarim Basin and Its Oil and Gas Control Effect

The Hudson Oilfield’s Donghe sandstone reservoir has multiple independent oil–water systems. The complex oil–water relationship is mainly due to reservoir heterogeneity, especially the sealing and shielding effects of interlayers. This paper uses methods like well logging and test well data interpretation, seismic impedance inversion, and the simulation of oil and gas preferential migration paths. It analyzes and quantifies parameters related to sand layer oil content, physical properties, and the development degree of interbedded layers using small layers as units. Sandstone horizontal transportability has three types. Type I matches structural forms to form preferential migration paths. Type II transportability has enclosed sand bodies that obstruct oil and gas flow and change their migration paths. Unevenly distributed calcareous interbedded layers are the main controlling factor for vertical sand body transportability. Locally continuous cemented calcareous interbedded layers can form the base of a dome-shaped pinch-out lithologic trap in sandstone. The closure area of the trap is controlled by the pinch-out line of sandstone and the distribution of continuous interbedded layers together.

Seismic inversion based on principal component analysis and probabilistic neural network for prediction of porosity from post-stack seismic data

Seismic inversion based on principal component analysis and probabilistic neural network for prediction of porosity from post-stack seismic data

Seismic facies-controlled porosity prediction in a tight sandstone reservoir based on the XGBoost algorithm

Porosity is one of the most important properties for evaluating hydrocarbon reservoirs. There is a strong nonlinearity between seismic data and rock-physical properties. Machine-learning methods possess the powerful ability to capture complex nonlinear relationships between these two parameters, holding significant potential for porosity prediction in tight sandstone reservoirs. In this study, we develop a seismic facies-controlled porosity prediction method based on the extreme gradient boosting algorithm. Through the seismic meme inversion method, we obtain high-resolution P-impedance data. In the inversion process, lateral variations of seismic waveforms were used instead of the variogram function. We create labels for model training using porosity curves and borehole side P-impedance data, applying them to the extreme gradient boosting regressor. To demonstrate its feasibility, we apply the model to a real 3D case from an oil field in the Songliao Basin. Our application results demonstrate that this method can provide porosity predictions for tight sandstone reservoirs, maintaining consistency with seismic waveforms and adhering to seismic facies control patterns. The method uses only acoustic traveltime, density, gamma rays, spontaneous potential well-log data, and poststack 3D seismic data as inputs, allowing for quick porosity predictions in the field. Compared with the seismic facies-controlled inversion combined with support vector machine and multiLayer perceptron methods, the method developed in this study provides more reliable porosity predictions, offering insights and references for the exploration and development of tight sandstone reservoirs.

Mining-Induced Earthquake Risk Assessment and Control Strategy Based on Microseismic and Stress Monitoring: A Case Study of Chengyang Coal Mine

As large-scale depletion of shallow coal seams and increasing mining depths intensify, the frequency and intensity of mining-induced earthquake events have significantly risen. Due to the complex formation mechanisms of high-energy mining-induced earthquakes, precise identification and early warning cannot be achieved with a single monitoring method, posing severe challenges to coal mine safety. Therefore, this study conducts an in-depth risk analysis of two high-energy mining-induced earthquake events at the 3308 working face of Yangcheng Coal Mine, integrating microseismic monitoring, stress monitoring, and seismic source mechanism analysis. The results show that, by combining microseismic monitoring, seismic source mechanism inversion, and dynamic stress analysis, critical disaster-inducing factors such as fault activation, high-stress concentration zones, and remnant coal pillars were successfully identified, further revealing the roles these factors play in triggering mining-induced earthquakes. Through multi-dimensional data integration, especially the effective detection of the microseismic “silent period” as a key precursor signal before high-energy mining-induced earthquake events, a critical basis for early warning is provided. Additionally, by analyzing the spatiotemporal distribution patterns of different risk factors, high-risk areas within the mining region were identified and delineated, laying a foundation for formulating precise prevention and control strategies. The findings of this study are of significant importance for mining-induced earthquake risk management, providing effective assurance for safe production in coal mines and other mining environments with high seismic risks. The proposed analysis methods and control strategies also offer valuable insights for seismic risk management in other mining industries, ensuring safe operations and minimizing potential losses.

Reservoir-oriented migrated seismic image improvement and poststack seismic inversion using well-seismic mistie

In seismic exploration of subsurface hydrocarbon reservoirs, migrated seismic images are the foundation of structural interpretation and subsequent seismic inversion for reservoir properties. Seismic events in migrated images can be mispositioned due to several factors, such as seismic velocity inaccuracy. The mispositioned events lead to poor well-seismic tie and degrade the accuracy of subsequent inversion and reservoir prediction. Particularly, the mistie can be quite obvious for the mid- and mature-phases of the reservoir development after many wells drilled. We propose using the information from mistie between seismic images and synthetic traces to improve the accuracy of migrated seismic images and thus inversion results. Our method includes three major steps: (1) calculate time shifts between the synthetic traces and the colocated seismic trace using the path-limited dynamic time warping algorithm, (2) compute the time-shift volume for the entire migrated image using the seismic-driven modeling approach based on the time-shift traces at the well locations, and (3) apply the time-shift volume to update the migrated image. The improved seismic volume can be used for further processing, such as seismic inversion, which makes the inversion results more consistent with seismic and well data. We apply our method to synthetic and field data sets. The results demonstrate that the updated seismic image has an improved seismic well tie and an improved inversion result. Our method indicates a venue for a better honor of seismic and well data in reservoir characterization.

Seismic Foundation Model (SFM): a next generation deep learning model in geophysics

While computer science has seen remarkable advancements in foundation models, they remain underexplored in geoscience. Addressing this gap, we introduce a workflow to develop geophysical foundation models, including data preparation, model pre-training, and adaption to downstream tasks. From 192 globally collected 3-D seismic volumes, we create a carefully curated dataset of 2,286,422 2-D seismic images. Fully using these unlabeled images, we employ the self-supervised learning to pre-train a Transformer-based Seismic Foundation Model (SFM) for producing all-purpose seismic features that work across various tasks and surveys. Through experiments on seismic facies classification, geobody identification, interpolation, denoising, and inversion, our pre-trained model demonstrates versatility, generalization, scalability, and superior performance over baseline models. In conclusion, we provide a foundation model and vast dataset to advance AI in geophysics, addressing challenges (poor generalization, lacking labels, and repetitive training for task-specified models) of applying AI in geophysics and paving the way for future innovations in geoscience.