Petroleum exploration in deep to ultra-deep sea areas is important to offshore reserve and production increase in China. In view of complicated structural setting and high cost of production, offshore petroleum exploration and development with high quality and high efficiency rely on geophysical techniques to obtain high-graded reserves of scale and rapid production increase without drilling many wells. Thanks to theoretical innovation and technical application in the latest years, some medium- and large-sized fields, e.g. Lingshui 17-2, Lingshui 25-1, Baodao 21-1, Kaiping 11-4, and Kaiping 18-1, have been successfully discovered in deep to ultra-deep areas, the South China Sea. These discoveries have also advanced geophysical techniques. We use several case studies to illustrate the challenges and solutions in deep-sea prospecting. Varying layout acquisition is a proven solution to the complexities caused by submarine relief, sea water depth, and abrupt tectonic change to obtain high-quality seismic data. With respect to heterogeneous water velocity, multiple and ghost reflections, complex wave field induced by irregular sea bottom, and inaccurate velocity in a complex structural zone, such broadband techniques as high-precision velocity analysis through full waveform inversion and viscoelastic prestack depth migration are employed to improve the resolution and signal-to-noise ratio of seismic data. Rock physical experiments, seismic response analysis, and rock physical modeling are established to improve reservoir prediction, fluid detection, prestack source rock prediction, and non-bright spot gas identification. The article concludes with the thoughts and suggestions for the future technical development of abyssal petroleum geophysical exploration in China.

Great achievements have been made in deep oil and gas exploration in the Tarim Basin, especially the blocks represented by Tahe and Shunbei, where high-yield oil and gas reservoirs have been encountered in deep–ultradeep strata (˃7 000 m). However, the key technologies for deep–ultradeep oil and gas exploration have not been systematically reviewed. In essence, the exploration targets of middle-shallow and deep-ultra-deep oil and gas are high-quality oil and gas reservoirs. Either shallow–medium or deep–ultradeep oil and gas exploration, essentially, targets high-quality reservoirs. In the latter, however, the wave-medium interaction is more complex, necessitating the changes in excitation, reception and shot-receiver arrangement. Moreover, the concepts and methods of deep–ultradeep oil and gas seismic exploration need to be raised to a new level. This paper takes the in-phase stacking of seismic wavelets reflected/scattered from various shot-receiver pairs at the same reflection/scattering point in the underground medium as the core concept of deep–ultradeep seismic wave imaging. Firstly, the changes in the physical mechanism for the detection of deep–ultradeep oil and gas reservoirs are analyzed. Then, the new challenges posed by these changes to seismic data acquisition techniques are discussed, and accordingly the seismic data acquisition techniques suitable for detecting deep–ultradeep oil and gas reservoirs. Specifically, in view of seismic data preprocessing, a noise suppression technique for data with strong noise and weak signals is proposed; in view of seismic wave imaging, a velocity modeling methodology required for imaging medium, deep and ultradeep geological bodies is proposed. Finally, centered on the fidelity and high-resolution imaging gathers, the equivalence of data processing in the prestack data domain and the prestack imaging domain is highlighted, and a reasonable methodology of high-precision seismic wave imaging for deep–ultradeep oil and gas reservoirs is formed, allowing the imaging results to fit into the characterization and evaluation of deep–ultradeep oil and gas reservoirs.

The Dongdaohaizi Sag in the Junggar Basin has a great exploration potential, but it has been less explored. In this paper, selected deep wells and newly acquired 2D and 3D seismic data within the Dongdaohaizi Sag were used to produce synthetic seismic records and perform horizon calibration, so as to identify the characteristics of seismic wave groups from medium–deep strata in the sag and clarify the exploration potential of the study area. On the basis of log characterization and synthetic seismic record calibration, four clear reflection interfaces, T₀, T₁, T₂, and T₃, were identified at the top and bottom and in the interior of the medium–deep Permian system. The characteristics of main seismic wave groups from the medium–deep strata in the sag were summarized. Specifically, T₀ is the interface between the Permian and the Carboniferous, and it is the key marker bed for regional correlation in the basin. T₁ is the interface between the Permian Jiangjunmiao Formation and the Pingdiquan Formation with source rocks developed. T₂ is the interface between the Pingdiquan Formation and the Upper Wuerhe Formation. T₃ is the interface between the Triassic and the Permian, and it represents the response of the regional unconformity between the Late Permian and the Triassic. Based on the detailed interpretation of seismic profiles, the characteristics and geological significance of seismic wave groups from the medium–deep strata in the Dongdaohaizi Sag were revealed. It is found that the seismic profile of the Upper Wuerhe Formation (between T₂ and T₃), the key target horizon, is composed of 4–6 wave peaks and troughs, with moderate continuity, as well as the transverse attenuation of the strong event energy or frequency change, which reflects the characteristics of lithologic sand bodies deposited under the background of lake transgression. The "onlapped" wave group characteristics of the Upper Wuerhe Formation in the northwestern part of the sag reflect the existence of a large provenance system, making this area a favorable target for future risk exploration.

The exploration and development of biogenic reef oil and gas reservoirs in the margin zones on both sides of the Kaijiang–Liangping trough in the Sichuan Basin have achieved great success, but research on the bioclastic beach beneath the biogenic reefs remains limited. The reef beach reservoir is deep, with a small thickness, and it changes rapidly in both longitudinal and transverse directions, making it difficult to accurately predict the reservoir and develop benefits. In order to solve the above problems, reef beach reservoirs of Changxing Formation in Yuanba West District were studied. Based on the fine division of strata and the study of palaeogeomorphology, it was concluded that the sedimentary model of the reef beach reservoirs of Changxing Formation was formed in the early stage of Changxing Formation, and the biogenic reef was inherited and developed in the middle and late stages of Changxing Formation, thus forming the unique sedimentary characteristics of “upper reef and lower beach” in the longitudinal direction. Based on a deep understanding of the sedimentary model of the reef beach reservoir, a method for identifying the reef beach reservoir by combining well seismic analysis was proposed, successfully guiding the drilling of Well Y203H_C in the study area and achieving significant progress in the evaluation of reef beach development. The research results indicate that: ① The early stage of the Changxing Formation in the Yuanba West District exhibits a distal steep carbonate gentle slope sedimentary model, and the reef beach reservoir is mainly developed along the steep zone. ② The application of mixed phase deconvolution and F-X denoising techniques has improved the quality of seismic data, and waveform indication inversion has been used to enhance the accuracy of longitudinal reservoir prediction. Combined with GR waveform indication simulation technology, it effectively avoids carbonaceous mudstone traps and achieves quantitative prediction of thin reef beach reservoirs. ③ The research results have successfully guided the deployment of the first reef beach development evaluation well, and the drilling results have confirmed the reliability of the geological understanding and reservoir prediction mentioned above. It is concluded that the sedimentary model and prediction method of thin reef beach reservoirs are helpful for the exploration and development of reef beach reservoirs in the Sichuan Basin.

The new direct push storage logging technology has strong applicability and high well control safety, effectively solving the logging data acquisition problem under complex well conditions such as high temperature and high pressure, easy blowout, and frequent lost circulation in ultra-deep wells of Fuman Oilfield. However, the well type of Fuman Oilfield, dominated by highly deviated and horizontal wells, makes the instrument measurement not centered and the motion speed of the instrument not constant, which greatly affects the quality of direct push storage slowness, and the reservoir grading evaluation method mainly based on acoustic waves is incomplete, seriously affecting the evaluation of logging reservoirs. Based on the characteristics of direct push storage acoustic instruments, a method for extracting the direct push storage slowness of the P-wave and S-wave and processing the relative amplitude of the mode wave was developed. A reflection imaging processing method near the wellbore in horizontal wells was established, achieving effective evaluation of fault-controlled carbonate reservoirs in the wellbore and around the wellbore. The distance between the wellbore was calculated, and 3D imaging of the reflectors was performed. The effectiveness of the reflectors near the sidetrack wellbore was comprehensively evaluated. Those technologies have been applied on a large scale in the ultra-deep fault-controlled carbonate rocks of Fuman Oilfield, supporting the comprehensive effectiveness evaluation of reservoirs in the wellbore, around the wellbore, and near the wellbore. The logging interpretation accuracy rate has reached 96.7%, and the application effect is significant.

The conventional method of migration-based geometry design is accurate and illustrative for target analysis. However, routine migration cannot be accomplished without high-performance computers and a lot of computation time for wavefield modeling. To improve the time-efficiency and accuracy of assessing seismic geometry design, particularly for a survey with complicated surface and subsurface conditions, we develop a seismic geometry evaluation method based on fast migration. This method involves point spread function computation and convolution in a higher-dimensional space. Specifically, the Hessian approximate operator is derived quickly from a preset model and a recording geometry, and then convolves with the reflectivities of the model (indicating the information of large wave numbers) in a higher-dimensional space, for subsurface imaging with high efficiency. Moreover, the imaging effects at varying folds, bin sizes, aspect ratios and trace densities are comparatively analyzed. The proposed method has been successfully applied to the geometry evaluation in southwest surveys with steeply dipping structures, northwest surveys with fractured-vuggy reservoirs, and the South China Sea with complex fault blocks.

To address the problems of spatial discontinuity, blurred edges, and loss of structural details in seismic data reconstruction, a new algorithm is proposed based on multi-scale feature fusion and generative adversarial network (MSF-GAN). The algorithm designs a multi-scale feature fusion generator in the GAN for effective seismic feature extraction and multi-scale fusion. A feature splicing module is designed in the generator for adaptively adding masks to seismic data, so as to splice the reconstructed data and the original intact data and improve computational efficiency. A multi-dimensional adversarial discriminator is designed in the GAN to improve reconstruction accuracy. Furthermore, a hybrid loss function integrating Smooth L₁ reconstruction loss and adversarial loss is proposed to update the generator and improve reconstruction reliability. Public data and seismic data from Daqing oilfield are reconstructed to validate the algorithm in different scenarios: continuous data loss, random data loss, and regular data loss. MSF-GAN performs better than orthogonal matching pursuit, projection onto convex sets, and spectrally normalized generative adversarial network in structural details and spatial continuity.

With the large-scale application of "wide azimuth, wide frequency band, and high density" acquisition technology, seismic data now contain increasingly rich subsurface media information, making it difficult for conventional imaging algorithms to meet the requirements of high-precision imaging from wide-azimuth seismic data. To address this challenge, this study proposes a gradient-constrained full-azimuth reverse time migration (RTM) imaging technique that enables efficient utilization of subsurface effective information and precise imaging of complex structures. First, the technique constructs constraint equations by incorporating spatial and temporal gradients of the wavefield, solving for the wavefield propagation vector through iterative minimization of an energy objective functional. This effectively addresses the issues of local computational instability and inaccuracy inherent in conventional wavefront vector methods. Subsequently, it accurately extracts subsurface azimuth and reflection angle information using local propagation angle geometry. Finally, imaging values are normalized based on these angle parameters to achieve mapping of full-azimuth angle gathers. Theoretical model tests and real data applications show that the proposed technique can fully utilize the propagation angle information of the wavefield to produce high-precision imaging profiles and high-quality full-azimuth angle gathers, thereby providing strong support for tasks such as migration velocity analysis and reservoir description.

Imaging of high-angle and sub-salt structures is one of the key and challenging aspects in current oil and gas exploration research. The data acquisition method of surface excitation and vertical well reception can be used to obtain information about high-steep structures. In addition, cross-well seismic data contains even richer information about high-steep structures. However, in actual acquisition, cross-well seismic surveys are costly. To achieve the accurate imaging of high-angle structures in complex anisotropic media, this paper presents an interferometric method. The VSP interferometry formula for VTI media is derived based on the reciprocity theorem. The principle of cross correlation is used to eliminate seismic waves with overlapping propagation paths in two shot records and generate the responses of a virtual source, which simulates a recording geometry moving closer to the targets for more information about high-angle structures. A comparative study is performed to check whether or not virtual source data are consistent with the VSP data acquired using the source and geophones in the borehole. The final image is the reverse time migration results of virtual source data. Numerical tests show that anisotropic migration is superior to isotropic migration in continuous imaging of high-angle structures with reduced noises. The proposed VSP interferometric method is a feasible solution to VTI media imaging and may effectively improve the imaging accuracy of high-angle and sub-salt structures.

Seismic data acquired in the Yingxiongling structural belt feature complex wave fields, severe scattered noises, and weak signals. The low-velocity layer and strong anisotropy are a huge challenge to the precise imaging of deep sub-salt structures. Owing to the occurrence of extremely thick evaporites and four large fault zones, it is hard to accurately image steeply dipping formations and faults without distortion. This paper proposes an integrated scheme of acquisition and processing for the precise imaging of deep steep structures and foot walls. The key geometry parameters for imaging are clarified based on forward modeling and the comparative study of field and modelled data. A target-oriented scheme for near-surface survey and the techniques of near-surface velocity modeling are established based on the quantitative analysis of shallow velocity influence on deep structure imaging. Guided by the velocity modeling from the "true" surface, a prestack depth migration velocity model for the precise imaging of foot walls is finally obtained through tomographic inversion of partial offset data and multi-azimuth grid tomography. The research suggests that: (1) wide azimuth and large offset can improve the reliability and accuracy of deep fault and basement imaging in the study area; (2) constrained tomographic inversion of partial offset data is a better solution to the modeling of the extremely thick near-surface layer with improved accuracy in mountainous areas; and (3) multi-azimuth grid tomography based on wide-azimuth seismic data can obtain a reasonable anisotropic velocity field of evaporites and foot walls for improved imaging of complex highly steep structures and foot walls.

The microseismic data containing diverse noises with extremely low signal-to-noise ratio cannot be effectively processed using traditional denoising methods. This paper proposes a deep convolutional auto-encoder network with optimized loss function constraints fused with residual attention (RADNet). The proposed method employs a deep convolutional auto-encoder structure for local feature extraction from noisy data and fuses global features to assign weights to different features using the attention mechanism. The optimized loss function is used to guide network training, and denoised signals are finally reconstructed based on the residual network. The application of RADNet and other denoising methods to simulated and real data demonstrates that RADNet improves peak signal-to-noise ratio (PSNR) by 2.783 and 8.099 dB and structural similarity (SSIM) by 0.031 and 0.065, respectively, compared to denoising convolutional neural network (DnCNN) and deep convolutional auto-encoder networks. Furthermore, RADNet decreases mean square error (MSE) and better preserves effective signals and texture details in microseismic waveform.

The organic reef of Changxing Formation in the LHC area of northeastern Sichuan Basin, as the main production layer, has shifted from the exploration stage to the development stage after several years of development. In addition, an accurate assessment of their development potential is crucial and needs to be achieved through geological modeling. However, deep organic reefs have blurred boundaries and unclear internal structural responses due to insufficient seismic imaging resolution. Moreover, there is a significant difference in porosity between organic reefs and non-reefs under the same impedance background, making traditional modeling methods fail to achieve accurate characterization of reservoir parameters. To solve this problem, a facies controlled modeling method based on seismic data processed by a steerable pyramid method for geological feature enhancement was proposed. First, the steerable pyramid method decomposed seismic data into multiple scales and multiple directions and reconstructed the data for geological feature enhancement, improving the accuracy of organic reef identification. Secondly, according to the seismic response characteristics of the organic reef and the restoration results of palaeogeomorphology, the three-dimensional favorable facies zone model of the organic reef was established. Finally, the inversion results of the reservoir were extracted into the geological model, and Bayesian facies controlled modeling was carried out to improve the consistency of geological rules with in-well and seismic information and realize the fine characterization of reef porosity. The application test shows that the porosity distribution of the facies controlled model is point-like, which is more consistent with the distribution characteristics of the organic reef and has a good agreement with the logging results, which proves that the proposed method is reliable. This method can facilitate the precise calculation and description of reservoir reserves in the future.

Deep learning-based fault detection methods typically rely on pre-interpreted fault labels for network training. However, practical applications face significant challenges due to the labor-intensive nature of manual fault interpretation and insufficient labels for the effective training of a deep learning model. The utilization of synthetic labels is an alternative. However, structural and lithologic variations in different regions lead to notable discrepancies between synthetic and field seismic data, which severely limit the generalization of a synthetic data-trained model for fault identification. To address these limitations, this study proposes a transfer learning approach employing a domain-adaptive neural network for cross-domain fault detection. A domain discriminator is designed in the U-Net architecture to simultaneously process both synthetic and real seismic data. A feature extractor is designed to learn domain-invariant representations from both data types, while the domain discriminator engages in adversarial training to distinguish the data source of extracted features. Through iterative optimization, the network eventually achieves domain confusion when the discriminator cannot reliably identify the origin of input features. At this equilibrium state, the model effectively captures the shared characteristics between synthetic and real seismic data, enabling the robust generalization to field data for accurate fault detection. Comparative experiments using seismic data from Netherlands North Sea F3 block and western China demonstrate the enhanced accuracy of our method compared to conventional deep learning approaches and its good performance of addressing domain shifts in seismic interpretation.

Concealed faults, which feature low grades, small throws, and small event distortions on seismic sections, are important to oil and gas exploration assessment, waterflood development, hydraulic fracturing, and gas storage safety. The concealed faults exhibiting differential extension associated with newly discovered intersected structural traps have garnered considerable attention. However, how to identify such faults with unclear genesis is a big challenge. This study focuses on Block X in central Bohai Bay Basin. Based on the geometric and kinematic characteristics of main faults and known concealed faults derived from seismic and well data, concealed faults are believed to commonly develop at the intersections between the segments of a main fault. A method known as “integrated geologic-geophysical interpretation” is proposed for predicting extensional concealed faults. This approach employs the “throw-distance curve” to pinpoint concealed faults, seismic attribute maps to indicate their strikes, and seismic profiles to verify their presence. The successful application of this method to Block X is referential to the identification of concealed faults with differential extension in similar areas.

Under single-well observation, the micro-seismic event location is usually using the time difference between P- and S-wave, and adjusted P-wave to calculate the distance and azimuth between the event and geophone, respectively. The real hydraulic fracturing monitoring data often contains lots of single S-wave micro-seismic events, but S-wave generally cannot be used for the calculation of azimuth, which brings difficulties for the location of single S-wave events. However, the focal mechanism inversion study indicates that the focal mechanism of micro-seismic events is closely related to the location, which provides the possibility of single S-wave events location under single-well observation. For the single S-wave events difficult location problem in real hydraulic fracturing data under single-well observation, based on ray tracing location and single-well focal mechanism inversion technology, this paper proposed a solution method by comparing the strike angle of focal mechanism between all feasible location and its nearest double phase micro-seismic event, whether meet similar or orthogonal as the constrain condition. Synthetic data test indicates that although this method localization error of the single S-wave event is greater than the double phase event, it is still within the acceptable range and is feasible. The real data application works well. The spatial distribution of micro-seismic events is consistent with the regional tectonic stress direction. The focal mechanism of events indicates that the rock fracture surface is mainly along the maximum or minimum horizontal stress direction, and the fractures are mostly high angles. This paper provides a feasible idea for single S-wave micro-seismic events location in single-well observation.

Lithology identification is an important foundation and critical part of well logging interpretation. However, the complexity of reservoir properties often leads to unavoidable inconsistency in cross-well lithology distribution and logging responses, impeding the robustness of cross-well lithology identification. This paper proposes several heterogeneous data representation features to reveal the invariant features of local reservoir description. Specifically, local topological information was first represented by employing graph representation technique in both vertical and lateral direction of well logging data. Then, three invariant features, namely structural tensor (ST), local binary pattern (LBP), and Hu moments (Hu), were obtained for robustly representing the local structure information of well logging data. Finally, the multi-view resampling strategy was adopted to address the distributional imbalance of log values and overlap of lithology in the original data domains, and meta-learning was employed to model the nonlinear relationship between the obtained heterogeneous features and target lithology information. Experiments were conducted using actual logging data from several wells in the Qijia depression of the Daqing Oilfield. The results indicate that the cross-well identification accuracy of the proposed method is more than 86%, proving it highly capable of solving the problems of inconsistent distribution of logging values and lithology between wells and unbalanced lithological data.

Increasing the polynomial order of basis functions is an effective strategy for enhancing the accuracy of finite element-based numerical simulations in geophysics. A conventional finite element method typically employs the basis functions with a uniform polynomial order across the entire computational domain. As a result, a higher order significantly increases computational cost. To address this challenge, a finite element method based on hierarchical basis functions is proposed. This method enables local p-refinement specifically for surface elements by using hierarchical basis functions and thereby improves the accuracy of numerical forward modeling without compromising computational efficiency. The validation through audiomagnetotelluric forward modeling on three typical geoelectric models demonstrates that the proposed method significantly enhances the accuracy of forward modeling with only a small increase in computation time, offering a novel approach for efficient numerical electromagnetic forward modeling.