Most of the projection onto convex sets （POCS） algorithms widely used for seismic signal reconstruction use linear or exponential threshold models，which have high computational efficiency，but have poor reconstruction effect due to the difficulty in eliminating the noise caused by the leakage of missing signals. Therefore，this paper proposes a POCS seismic signal reconstruction method based on the regional threshold model，which transforms the numerical threshold into a regional threshold，and the regional filtering window is iteratively updated as the threshold. The core idea is to reserve the transform coefficients of effective signal as much as possible by selecting a rectangular or a sectorial region of fixed size as a threshold according to a certain law in each POCS reconstruction iteration based on the frequency-wavenumber （F-K） spectrum distribution range of missing seismic signal in the spatiotemporal domain，and reserving and zeroing the transform coefficients inside and outside the region respectively. The rectangular and sectorial threshold models for POCS reconstruction of seismic signals are thus constructed. The numerical results demonstrated that compared with the POCS reconstruction of the exponential threshold model in the F-K domain，the regional threshold model in the F-K domain has a higher reconstruction accuracy for continuous missing signals. Compared with the sectorial region threshold model，the reconstruction accuracy and computational efficiency of the rectangular region threshold model are slightly higher. Compared with the exponential threshold model reconstructed by POCS in the curvelet domain，the reconstruction accuracy of the regional threshold model in the F-K domain is similar，but the computational efficiency is increased by about 90%.

 

Microseismic monitoring is a common means to guide hydraulic fracturing operations and evaluate fracturing effects in shale gas extraction. The signal collected by ground monitoring has weak energy and low signal-to-noise ratio， which makes it difficult to identify microseismic events， and seriously affects the accuracy of positioning. Aiming at the ground microseismic monitoring data with low signal-to-noise ratio， a new noise cancellation method is proposed by combining synchronoussqueezing S-transform， spectral decomposition and τ-p transform. Firstly， the time difference correction is carried out on the monitoring data， and the in-phase axis of the microseismic signal is leveled. Then， the synchrosqueezing S-transform was applied to decompose the leveled data to obtain single-frequency slices. Then， the τ-p transform is performed on each single-frequency slice， and the microseismic signal position is obtained according to the results of the τ-p transform. Finally， noise cancellation is completed in the time-frequency domain according to the position of the signal. The processing results of ground microseismic monitoring data with low signal-to-noise ratio show that the new method can obtain ideal noise cancellation results.

The challenge faced by seismic velocity inversion methods based on deep neural networks is that the weak semantic mapping correspondence between seismic data in the time domain and model information in the spatial domain leads to a high degree of multiplicity.Additionally， neural networks lack effective guidance in mapping seismic data to velocity models， making them susceptible to noise interference and thus affecting inversion accuracy.Therefore， a seismic velocity inversion method based on feature enhancement U-Net is proposed.Firstly， by integrating the features of multi-shot seismic data， the spatial relationship between the seismic time series signal input to the network and the corresponding velocity model becomes more apparent.Subsequently， based on the concept of multi-scale feature fusion， modules with convolutional kernels of varying sizes are designed to bolster the network’s capacity for learning effective features.Next， attention gates are used to guide the network and enhance the features that the network focuses on.Finally， based on the bottleneck residual and pre-activation， a pre-activation bottleneck residual is incorporated into the network， to avoid gradient disappearance and network degradation.The experiment shows that this method has higher accuracy in seismic velocity inversion and performs well in noise testing.It has a certain generalization ability.

At present， CNPC’s oil and gas exploration and development are expanding from the uplift area to the basin slope， lake basin center， and basin periphery. The target layer is expanding from medium and shallow layers to deep and ultra-deep layers， and the object is expanding from conventional oil and gas to unconventional oil and gas. The exploration and development are faced with more complicated exploration areas， more hidden targets， and more difficult operations. In response to the problems of seismic exploration in highly difficult exploration areas on land， CNPC has strengthened technical research on seismic acquisition， processing， and interpretation and formed a series of technologies such as ‘acquisition of wide azimuth， broadband， high density， and high fold’， processing of ‘high resolution and high fidelity’， true surface imaging processing， and quantitative prediction of pre-stack reservoirs， which have supported continuous breakthroughs in exploration of highly difficult exploration areas and long-term stable growth of reserves， and achieved stable and rising oil and gas production. The above achievements benefit from the change in the concept of geophysical exploration technology， which brings the new concept of oil and gas exploration，‘oil and gas are not only in the mind of geologists but also hidden in high-quality seismic data’. In addition， thanks to the grasp of the breakthrough of seismic exploration technology on land， the technical idea of ‘the target is located at deep layers； the problem emerges at the near-surface； the velocity serves as the key， and the refinement remains as the core’ is proposed. Benefiting from the innovation and transformation of management ideas， a complete geophysical exploration technology management system has been established， which greatly improves the quality of seismic data in each exploration area. The breakthrough in understanding and management innovation will guide the development of onshore oil and gas seismic exploration technology to the all-directional， high-density， all-digital， multi-parameter， automated， intelligent， full-waveform， full-wave field， and other directions.

Scholte waves propagate along the seafloor fluid-solid interface and exhibit typical dispersion characteristics. The dispersion properties of Scholte waves can be utilized for the inversion of the shear wave velocity of the shallow seabed， making it an effective tool for seabed shear wave velocity modeling. It is of critical importance to establish the theoretical model of Scholte wave dispersion. In this study， a horizontal layered model of seawater-seabed elasticity is developed based on a real seawater-seabed environment. The dispersion equation and displacement equation for Scholte waves in this model are derived using the continuity conditions of boundary stress and displacement. The influence of seawater depth and seabed property parameters on the dispersion characteristics of Scholte waves is analyzed. Experimental results using a 6-layer seawater–seabed model show that： ①Regardless of the seabed is hard or soft， Scholte waves exhibit distinct dispersion characteristics. In the case of a hard seabed， the energy of Scholte waves is mainly concentrated in the seawater， with the fundamental mode having the weakest energy and the second mode having the strongest energy. As the seawater depth increases， the dispersion characteristics of Scholte waves weaken. For the soft seabed model， the energy of Scholte waves on the solid seabed significantly increases， and the seawater depth has little effect on Scholte wave dispersion characteristics. ②Compared with deep-water environments， Scholte waves have stronger energy and more pronounced dispersion characteristics in shallow water. Therefore， utilizing Scholte waves for inverting the shallow seabed shear wave velocity in shallow water environments yields higher accuracy and reliability. Finally， based on an actual seabed elastic geological model of a working area in the East China Sea， multi-component seismic data are simulated， and dispersion curves are extracted. A comparison analysis between the theoretically calculated dispersion curves and the extracted dispersion curves shows a good agreement， confirming the accuracy of our theoretical approach.

Amplitude-versus-angle （AVA） inversion plays an important role in the prediction of reservoir elastic parameters. The angles of the AVA equation （physical prior knowledge） are not easily determined， and the ill-posed nature of the large sparse matrix will make the conventional prestack inversion procedure unstable. For this reason， a prestack decoupled stepwise inversion method harmonized with physical and data prior knowledge is proposed. Firstly， a non-sparse forward framework is established based on physical knowledge to increase the stability of parameter inversion and lay the foundation for decoupled stepwise inversion. Next， this paper takes well data as data prior knowledge and harmonizes it with physical and data prior knowledge to decouple the prestack seismic data for more accurate prestack seismic attributes. Finally， inversion of decoupled prestack seismic attributes is performed to get the reservoir elastic parameters. This method can correct the inversion procedure by well logging prior information， which can avoid the errors caused by the inaccurate angles in the physical prior knowledge. The actual data test results show that the inversion results of the proposed method have higher accuracy compared with the industrial AVA inversion method. In addition， the inversion results of this method have higher accuracy， in which the accuracy of the lame parameter， shear modulus， and density is improved by 14.1%， 13.6%， and 11.9%， respectively.

Fracture prediction is a crucial aspect of unconventional reservoir forecasting.The fracture density not only reflects the degree of fracture development but also serves as a significant parameter influencing fracture porosity and permeability.In this study， a fracture density inversion method is proposed， which uses weighted sine and cosine components of Fourier coefficients of P-wave reflection amplitude.Drawing upon the linear sliding crack equivalent medium theory， relationships between Fourier coefficients of the HTI （Horizontal Transversely Isotropic） medium P-wave reflection amplitude and fracture density are derived separately for oil-bearing and gas-bearing conditions.Additionally， the second-order Fourier coefficients are employed to predict the azimuthal angle of the fracture symmetry axis.The research investigates the variation patterns of the signal-to-noise ratio of the second-order Fourier coefficients’sine and cosine components with respect to the azimuthal angle of the fracture symmetry axis when seismic data contain noise.An inversion method for fracture density is then proposed using weighted sine and cosine components of second-order Fourier coefficients.Model simulation results indicate that the proposed fracture density inversion method exhibits strong noise resistance and stability.Application of the method to actual seismic data demonstrates that the predicted fracture density aligns with well log information， thus validating the effectiveness of the approach.

A set of key techniques of seismic exploration is proposed for the identification of deep carbonate fracture-cavity bodies and small faults in the Yingxiongling structural belt of the Qaidam Basin. First of all， from the aspect of data acquisition， the imaging accuracy of small faults can be improved by expanding the observation orientation， using a small combination of excitation and reception， and adopting an even arrangement of source and receiver points. From the aspect of seismic data processing， OVT domain multi-dimensional fidelity noise suppression technology and thick surface Q compensation processing technology can be used to improve the signal-to-noise ratio and fidelity and effectively increase the bandwidth and resolution of seismic data. From the aspect of seismic data interpretation， the sub-azimuth data volume and artificial intelligent fault identification technology can be used to identify small faults effectively，and the anisotropic strength attribute can be used to identify the fracture-cavity body effectively.The techniques can be used for reference in other similar areas.

As Shunbei oil and gas area and Luntan 1 well in the Tarim Basin obtain industrial oil and gas flows at a depth of more than 8200 m， carbonate rock exploration is rapidly moving towards the deep and ultra-deep fields， posing a severe challenge to seismic exploration technology. This article mainly analyzes the research progress and problems faced in the theory of ultra-deep complex wave field seismic imaging. In terms of key technologies for predicting ultra-deep reservoirs， the current status of small fault identification through seismic data structure characterization and quantitative prediction methods for pore structure based on digital cores has been analyzed. From the perspective of geological exploration requirements， this paper proposes the development trend and key research directions of deep and ultra-deep carbonate reservoir and fluid prediction technology， so as to provide a reference for the theoretical and technical research of marine carbonate rock seismic exploration， and the following understandings are obtained： ① For ultra-deep seismic data with low signal-to-noise ratios， Q-stack depth migration and TTI medium RTM technology have achieved certain results in carbonate reservoir imaging. The key research directions include interlayer multi-wave suppression based on wave theory， anisotropic Q-RTM， least squares Q-RTM， and anisotropic omnidirectional angle domain imaging technology. ② The seismic data-based prediction technology of deep and ultra-deep heterogeneous carbonate reservoirs has the problems of weak theoretical methods and low prediction accuracy， so it is urgent to strengthen the exploration of theoretical methods and technical breakthroughs. ③ The deep integration of seismic petrophysical experiment and reservoir geology， the refined seismic data-based prediction technology of sensitive reservoir properties based on the wave characteristics of two-phase media （frequency， dispersion， and attenuation）， artificially intelligent and quantitative prediction of carbonate reservoirs， and fluid detection technology are important development directions. The development trend of “reliable deep seismic data， multi-disciplinary and high-precision characterization of reservoirs， and deep learning artificial intelligence” is obvious.

Brittleness is an important index to evaluate the fracturing ability of tight reservoir and is of great significance to oil and gas exploration and development. Due to the low exploration degree， large area and few exploration wells， it is difficult to describe and predict the brittleness of the dolomitic tight reservoir in the middle and lower part of Permian in the eastern Junggar basin. This paper carries out the prestack inversion and brittleness prediction of the reservoir by using the angle domain data， according to the petrophysics experiment. Prestack inversion includes the following methods. AVO inversion is applied to obtain the P+G and P×G attributes. Y（Young's modulus，E） P（Poisson's ratio，σ） D（density，ρ） and（L（λ）M（μ）R（ρ） inversions are applied to obtain the reflection coefficient of elastic parameters logging constrained inversion is utilized to get the actual value of multiple elastic parameters. Then， the brittleness is calculated with various characterization formulas， and its physical meaning and application effects are analyzed. The practical application shows that： ①There is a good fitting relationship between shear wave velocity and P-wave velocity in the study area， and Young's modulus and Poisson's ratio are distinguished from mudstone and non-mudstone. ②The dolomitic rock has the characteristic of high velocity， and the reservoir shows a double sweet spot structure. ③Using three small-angle data bodies to directly invert the reflection coefficient of elastic parameters， the YPD inversion method has a better reservoir identification effect and high resolution than LMR inversion. Especially in the density reflection coefficient data， obvious fan and channel alluvial features can be seen. ④There is a certain difference between the brittleness characterization results and the characteristics of elastic parameters. The favorable areas of the dolomitic rock characterized by brittleness are distributed in bands， while the elastic parameters basically distributed in continuous patches. ⑤ρE/σ is the best brittleness characterization formula in the study area， which can provide a reference for the prediction of brittleness in dolomitic tight reservoirs.

Quantum computing has revolutionary advances in computing due to its powerful computational capabilities.It provides a new tool for the computation of complex problems and has been applied in many disciplines and fields.In recent years， quantum computing has been applied in geophysics.The application of quantum algorithms and quantum computers provides technical support for revealing the internal structure of the earth and probing deep resources.Quantum computing can improve computation efficiency and has great potential in geophysics.To this end， the principle of quantum computing is systematically analyzed， and the development status of the quantum algorithm is summarized.The existing achievements in the fields of geophysical data acquisition， wave field simulation， and inverse problem solving are also summarized.The superiority of the quantum algorithm is verified by establishing the theoretical model and conducting the inversion.Finally， the possible research directions of quantum computing in the future are prospected.

Rendering traditional modeling methods are inadequate for fluvial reservoir characterization due to the reservoir's rapid lateral variations and strong heterogeneity. To address this, we initially employ steerable pyramid technology to decompose and reconstruct seismic data, enhancing the geological laws crucial for sedimentary facies characterization. Subsequently, we apply Bayesian-sequential Gaussian seismic-constraint modeling to the reservoir, establishing constraints between logging data and seismic attributes to enhance the model's vertical resolution. Modeling practices in the fluvial reservoir of the Sha-2 member in Shengli Oilfield demonstrate that, after the steerable pyramid process, seismic data, lateral characterization improves, with effective sand body boundary identification. Compared to traditional Sequential Gaussian methods, the Bayesian-sequential Gaussian approach achieves higher vertical resolution, with an 86% match between sand body thickness and the actual value, making it more effective for sand body identification. This method provides guidance for remaining oil recovery.

As degrees of oil field exploration and development deepen， most target reservoirs are found to be thin， small， deep， and fractured. High-precision identification of thin interbedding has become the focus of re-servoir prediction. At present， time-frequency analysis methods and spectral decomposition techniques commonly used in thin interbedding identification are limited by seismic data， and the processing results fail to meet the requirements of accurate characterization of thin interbedding. Due to the nonlinear characteristics of geophysical inversion， the results of high-resolution inversion have multiple solutions. In this paper， a thin interbedding identification method based on the similarity of seismic waves is proposed. Firstly， the idea of a waveform library is used to construct the waveform library of well bypass-sensitive logging curves. Secondly， the improved Manhattan distance combined linear correlation coefficient method is then used to calculate the waveform similarity. Finally the waveform similarity is taken as the only driver to establish the mathematical relationship between seismic data and high-resolution logging data， until the high-resolution processing profile is obtained. This method makes full use of the high transverse resolution of seismic data and the high longitudinal resolution of logging data， effectively identifies thin interbedding， and greatly reduces the multi-solution of inversion results. The feasibility of this method is verified by model test. The actual data processing results show that the proposed method significantly improves the longitudinal resolution and provides technical support for the identification of thin interbedding within a tuned scale.

Horizon tracking is basic and important work in seismic data interpretation, but the accuracy of conventional methods makes it difficult to meet the requirements of actual production. Therefore, this paper proposes a method for high-precision multi-set horizon tracking based on U²-Net. Firstly, a method is designed to produce filling labels, which traverse each pixel of seismic data to determine the location of each pixel and divide a horizon region for it. The method can automatically search for the adjacent horizons for the fault-crossing horizons, thus realizing the production of filling labels for the seismic reflection horizons and unconformities under the complex conditions of partial horizons and faults. Then, a U²-Net model is used to train the F3 data and the M-zone seismic data by utilizing the filling labels. Compared with the U-Net+PPM model, the U²-Net model has higher prediction accuracy, better stability, stronger generalization, and shorter training time. In addition, its accuracy and Mean Intersection over Union both exceed 95% in predicting the seismic reflection horizons in complex areas. The proposed method can be better adapted to the horizon tracking of seismic data with low signal-to-noise ratio.

There are abundant oil and gas resources entrapped in the carbonate reservoirs of the Ordos Basin. However， exploration results showed that the Majiagou Formation in the Daniudi Gas Field had developed multiple kinds of faults with small fault throws due to complex origins， which brings many challenges to its exploration and development. To address these challenges， it is crucial to optimize the sensitive elastic parameters for reservoir prediction. Therefore， the relationship between seismic attributes and the velocity ratio of compressional to shear waves （v_P/v_S） in the reservoir has been established， based on the analysis of elastic-sensitive parameters in the Daniudi Gas Field. Then， a prediction method for the v_P/v_S based on the XGBoost algorithm and multiple seismic attributes is proposed. To further improve the performance and generalization ability of the model， the hyperparameters of the XGBoost algorithm are optimized by Bayesian algorithm. This approach aims to find the optimal combination of hyperparameters， ensuring improved performance of the model on both training and testing datasets. The XGBoost algorithm is applied to the Marmousi 2 model for predicting shear wave velocity， achieving a correlation coefficient between predicted and actual values exceeding 0.88. With root mean squared error and mean absolute percentage error below 6.55×10^-7 and 4% respectively， the accuracy and reliability of the proposed method are demonstrated. The method applied in the Daniudi Gas Field of the Ordos Basin has successfully identified gas-bearing reservoirs， and the results are consistent with actual dril-ling data. Both theoretical model and practical data indicate that XGBoost， as a powerful machine learning algorithm， exhibits high accuracy， which can provide an effective approach for directly predicting v_P/v_S from post-stack seismic attributes.

Gravity inversion is one of the important means to obtain the spatial structure and physical properties of underground geological bodies through surface information， and each gravity gradient component represents different geological body information. Gravity inversion combined with gravity gradient components can better reflect the shape and distribution of underground abnormal bodies. In this paper， a neural network-based algorithm for gravity full tensor data inversion is proposed. The U-Rnet network is applied to three-dimensional gravity full tensor data inversion. In order to test the effectiveness of the algorithm， six representative models are used for simulation experiments， and inversion results with clear boundaries and sparsity are obtained. Firstly， by comparing the inversion results of L₂ and Tversky loss functions， it is found that the inversion results corresponding to Tversky loss functions can clearly represent the boundary position of the model. Then， by comparing the inversion results of different gradient tensor combinations， the results of four tests show diffe-rent inversion accuracy on three directions （x， y， z）， and the test 4 shows the lowest fitting error. Finally， the proposed method is applied to the FTG data of Vinton Salt Dome in Texas， USA， and the inversion results are consistent with the real geological information.

Faults identification of seismic data plays an important role in oil and gas exploration. At present, the application of machine learning and deep learning techniques has enhanced faults identification precision and efficacy. However, the fracture prediction outcomes remain challenging to meet the production needs. Therefore, a Transformer-based seismic fault identification method, namely 3D SwinTrans-U-Net, is proposed. The network consists of Swin Transformer module and convolution module. Among the aforementioned modules, the Swin Transformer module employs the attention mechanism of the Transformer to extract global information, and transforms global attention computing into window attention computing, resulting in less computational complexity compared to the Transformer. The convolution module, with the property of inductive bias, avoids the Swin Transformer's defect of weak inductive bias. Finally, the U-Net structure is utilized to combine the Swin Transformer and the convolutional layer. Thus, the structure can achieve deep and shallow information fusion, relevant feature extraction, as well as full learning of global and local dependency information. All these improve computational efficiency while ensuring fault identification accuracy, and enable end-to-end seismic fault deep learning. Synthetic and field seismic data tests have proven that the 3D SwinTrans-U-Net network can further improve the accuracy of fault identification.

In conventional seismic data processing, velocity analysis relies heavily on manually picking the velocity spectrum. However, this method is time-consuming, labor-intensive, and restricts the efficiency and accuracy of large-scale 3D seismic data processing. To address this issue, this paper propose an automatic velocity spectrum picking method using a dual-path convolutional neural network (DCNN). Firstly, this paper utilize a convolutional neural network combined with an attention mechanism as the main network, extract the features of energy clusters from velocity spectrum data, and realize the automatic picking of velocity. Secondly, this paper train a main neural network to integrate the information of velocity and the hidden representation of the uncorrected CMP gather input by another convolutional neural network (auxiliary network) through feature fusion and feature transformation before outputting the time-velocity sequence, reconstructing the corrected CMP gather. Finally, the process of CMP gather dynamic correction is simulated through the auxiliary network, and the accuracy of speed picking is optimized using dynamic correction. Model and real data tests demonstrate that, after incorporating dynamic correction information through the auxiliary neural network, the proposed velocity spectrum picking method achieves a higher accuracy than a single CNN in velocity picking.

The P-wave to S-wave velocity ratio（v_P/v_S） is a vital tool for gas reservoir identification，reservoir characterization，and lithology recognition. At present，the P-wave velocity and S-wave velocity are mainly obtained through the inversion of the reflection coefficient approximate equation，and then the v_P/v_S is calculated. However， this indirect calculation method creates a cumulative error. To obtain the v_P/v_S directly from pre-stack seismic data，this paper proposes a new generalized elastic impedance equation and further derives an approximate equation of the P-wave reflection coefficient，which is linked to the v_P/v_S，P-wave velocity，and density. To achieve high-precision inversion results，this paper proposes a prestack seismic inversion method based on the sparse constraint of the L_p quasi-norm utilizing the derived approximation equation of reflection coefficient，which is solved by the alternating direction multiplier algorithm. The proposed direct inversion method is applied to theoretical models and practical data and compared with the indirect inversion method. The results demonstrate that the direct inversion method exhibits higher inversion accuracy and clearer boundary characterization of gas-bearing reservoirs.

Five-dimensional （5D） seismic data can better analyze the changes in attributes such as travel time， speed， amplitude， frequency， and phase of seismic waves propagating in anisotropic media with azimuth angle. Moreover， the offset information is related to the scale， stratigraphic lithology， and fluid composition of the target geological body， while the azimuth angle information is related to the development characteristics of stratigraphic faults and fractures. Therefore， this paper proposes a technique for volcanic structure characterization and lithofacies identification based on 5D seismic data. By considering that the underground structure responds more obviously in the direction perpendicular to the strike direction， the paper constructs an azimuth analysis window to extract the dominant azimuth information and uses dip imaging to enhance the processed seismic data， so as to predict the volcanic structure. By using changes in dip and azimuth angles， the paper calculates the similarity of adjacent channels， improves the lateral signal-to-noise ratio of seismic data， and clarifies the macroscopic distribution range of volcanic structures. By defining the azimuth time window and combining the seismic trace inverse distance weighting algorithm， the paper extracts the most sensitive information about faults at each azimuth angle， improves the accuracy of volcanic structure characterization， and obtains a clearer volcanic morphology. Combined with kernel principal component analysis （KPCA）， the dominant attributes are fused to predict the favorable lithofacies of volcanic rocks. The proposed method accurately predicts the favorable zones of the third phase of volcanic rocks in the KM1 well area of the Junggar Basin， laying the foundation for the exploration and development of volcanic rock reservoirs in this area.

As the random noise in seismic data is highly irregular, conventional denoising methods often fail to achieve satisfactory results, thereby hindering subsequent interpretation and analysis of seismic data. Therefor, a seismic signal denoising method based on the diffusion model is proposed. The forward diffusion process of this method involves adding a certain degree of noise to the seismic data, transforming it into noisy seismic data with a large amount of isotropic Gaussian noise. Then, the trained diffusion model is usedtoreconstruct the noisy data and improve its signal-to-noise ratio. The prediction network component is an improved U-Net network, which incorporates attention modules and ResNet modules. These modules elevate the network's attention on important regions and mitigate gradient disappearance in deep networks. Both theoretical and practical data have verified the effectiveness of the proposed method. In terms of denoising effect, it significantly surpasses traditional denoising methods such as FX filtering and SVD, and it also outperforms classic deep lear- ning networks like CNN and GAN. This approach effectively preserves valid signals, thereby significantly enhancing seismic dataquality.

Horizon tracking is a very important step in seismic data processing and interpretation. The existing automatic horizon tracking technology often has a poor tracking effect when encountering faults. Therefore, a pure data-driven multihorizon tracking method using dynamic time warping (DTW) is proposed based on the waveform similarity between adjacent seismic channels. Firstly, the characteristic values of the seismic channels are extracted and the seismic channels are divided according to the characteristics of troughs, peaks, or zero crossings so that the tracked horizon positions strictly follow the above characteristics. In this way, a high accuracy of the tracking results can be ensured. Secondly, the DTW method is used to calculate the characteristic value sequences divided into two adjacent seismic channels to obtain similar paths. Finally, all the similar characteristic value point pairs, extracted from the similar paths, are divided according to the reference horizon to obtain each set of horizons. The results of model data calculation and actual data tests show that the proposed method can quickly track the positions of each set of horizons in the target area, and it effectively overcomes the hindering effect of faults on horizon tracking. It has a certain generalizability to seismic data in different geological conditions and has a certain application value.

The Y target in the ultra deep water area of Qiongdongnan Basin is a high-level delta underwater distributary channel sedimentary reservoir controlled by multi-level fault steps. The gas and water layers revealed by drilling data are both shown as “bright spots” in seismic data and have Class Ⅲ AVO features with amplitude enhancement along offset. Conventional oil and gas detection results based on amplitude information have multiple solutions. Therefore, a seismic full information hydrocarbon detection technology that fully utilizes the amplitude, frequency, and phase information of seismic data is proposed. Firstly, this paper comprehensively analyzes the post-stack and pre-stack seismic response characteristics of sandstone reservoirs and influencing factors via seismic forward modeling. It is pointed out that poro-sity and gas layer thickness are the main reasons affecting fluid identification by using amplitude information and proves that thick gas-bearing sandstones can be identified based on amplitude information. Secondly, the seismic reflection spectrum analysis shows that the attenuation of seismic waves caused by hydrocarbon layers is the main reason for the decrease in the main frequency and bandwidth of gas reservoirs. Finally, numerical simulation shows that both thick gas sand and thin gas sand under the background of water layers exhibit obvious absorption and phase anomalies, so the frequency and phase information related to seismic wave attenuation can be used for oil and gas detection. The application in the research area shows the proposed method can effectively distinguish between gas and water layers. The detection results are consistent with the actual drilling results, and provide a reliable basis for drilling implementation as well as effectively reduce exploration risks.

Missing traces reconstruction is a key step for seismic data processing. In recent years，various seismic data reconstruction methods based on deep learning theory have been proposed. However，normal convolution operation can only capture local dependencies and make insufficient use of global information. Moreover，the operation of pooling also results in the loss of feature map information，which destroys detailed features of seismic reflections. Therefore， a seismic data reconstruction method based on wavelet channel attention network is proposed. The Haar wavelet transform effectively extracts multi-scale characteristics and avoids the loss of information during the up-sampling process. Efficient channel attention modules are introduced to model the correlations between feature maps of different channels，which can make full use of the global information. Experimental results on synthetic and field datasets illustrate that the wavelet channel attention network can produce more accurate reconstruction results than some representative deep learning methods.

Horizon tracking is a key step in seismic data interpretation.It is typically performed manually by interpreters in a human-computer interaction manner, which results in low efficiency. Convolutional neural network （CNN） can establish a nonlinear mapping relationship between seismic data and training labels to achieve horizon tracking. However, since it is difficult to obtain manually interpreted results, models trained merely with a few labels tend to have relatively poor generalization capability. Therefore, a semi-supervised horizon tracking method based on a convolutional neural network is proposed to transform horizon tracking into image segmentation between horizons and faults. First, the unlabeled data is trained by the autoencoder. Then a small amount of labeled data is used for supervised learning after part of the parameters are transferred to the supervised learning network. Finally, the seismic data of the whole working area is predicted, and the edge of the segmentation result is extracted as the horizon tracking result. The test results of both synthetic data and the real data show that compared with the supervised learning horizon tracking method, the proposed method pre-sents less error segmentation and smaller errors between the horizon extracted from the segmentation edge and the artificial horizon interpretation results, and thus has better generalization capability.

In-situ stress seismic data-based prediction approaches of shale reservoirs usually apply the elastic parameters to calculate differential horizontal stress ratios （DHSR）. However， there are shortcomings. First， the anisotropy parameters （fracture weakness） are included in the prediction formula and cannot be solved easily， making it difficult to predict in-situ stress； second， parameters such as Young's modulus and Poisson's ratio in the prediction formula are obtained by indirect inversion， and the accuracy is low， which is difficult to meet the requirements of shale gas geoengineering integration. Therefore， an in-situ seismic data-based prediction approach of shale reservoirs based on fracture density inversion is proposed. For enhancing the prediction accuracy， a longitudinal wave azimuthal anisotropy AVO formula based on Young's modulus， Poisson's ratio， and fracture density is established to directly invert elastic parameters， and the DHSR formula expressed by Poisson's ratio and fracture density is derived. The anisotropic petrophysical modeling of shale reservoirs is carried out according to well data， and pre-stack azimuthal anisotropy inversion is performed. The inverted Poisson's ratio and fracture density are applied to the estimated DHSR. The in-situ stress properties of shale reservoirs can be evaluated by the estimated DHSR. A real example verifies that smaller DHSR indicates that more orthogonal and complex fracture networks will be generated by hydraulic fracturing in different directions， which is beneficial for modifying the physical properties and seepage channels of the reservoir. A larger volume of reservoir reconstruction is more conducive to reservoir fracturing. In addition， larger DHSR means that hydraulic fracturing will generate non-orthogonal plane fractures parallel to the maximum horizontal principal stress， forming isolated fractures that are not beneficial for volume transformation. Meanwhile， the estimated DHSR result agrees with existing well logging in-situ stress calculation， fracturing monitoring， and production testing， and it fits with the geology recognition， revealing that the approach is effective and reliable.

Sequential stochastic simulation is generally used to characterize reservoir heterogeneity both in the iterative stochastic inversion and linear Bayesian stochastic inversion. Most sequential simulation methods rely on variograms or training images to describe the spatial correlation of model parameters. In addition， the simulation results are required to be calculated point by point， which makes parallel computing difficult and reduces computational efficiency. Therefore， a conditioned fast Fourier transform moving average （FFT-MA） is introduced into the linear inversion framework， and a prestack seismic stochastic inversion method based on spatial co-simulation is proposed. Firstly， the posterior probability distribution of elastic parameters is obtained by integrating seismic data and low-frequency well logging information under the Bayesian framework. Then， the probability field is generated according to the FFT-MA algorithm. Well logging data is taken as conditional data to conduct Bayesian posterior probability field co-simulation. High-resolution prestack stochastic inversion results of elastic parameters constrained by well logging and seismic data are thus obtained. No iteration and update of model parameters is required by the method， which greatly improves the computational efficiency of stochastic inversion. Finally， the validity of the proposed method is demonstrated by numerical model examples and practical data application cases. The numerical model examples show that the proposed method has significant advantages over conventional methods in terms of high-resolution reservoir prediction and computational efficiency. Small-scale reservoir characterizations can be explored stably and accurately. The practical data application cases show that the high-resolution inversion results obtained by the proposed method match well with well logging data. The practicality of stochastic inversion in characterizing quantitatively thin reservoirs is greatly improved.

The conventional total variation （TV） regularization model only considers the first-order derivative information in the horizontal and vertical directions. When dealing with prestack seismic data with curved reflection events，it can severely damage the amplitude information and cause “staircase effects” by suppressing the lateral gradient characteristics of the amplitude. The local dip information of seismic data is often applied to improve the amplitude-preserving ability of the TV model. However，the calculation of local dip information itself will be impacted by noise. To address this issue，this paper proposes a high-order TV regularization model to suppress random noise in prestack seismic data in the domain of normal moveout（NMO）. This method first transforms the prestack seismic data into the NMO domain，NMO is robust to noise and avoids the calculation of the local dip angle. In the NMO domain，the curved event is flattened，and then high-order TV denoising is performed. Finally，the prestack seismic data are restored through inverse NMO. Taking the second-order derivative as an example，a high-order TV regularization inversion denoising objective function is constructed，and a fast optimization method is derived under the split Bregman optimization framework. The processing results of synthetic seismic data and actual seismic data show that this method can not only effectively suppress random noise but also eliminate amplitude distortion caused by curved reflection events and “staircase effects”， improving the amplitude preservation performance of the TV denoising method.

With the deepening of oil and gas exploration and development， thin and lithologic reservoirs have gradually become important exploration targets， which also leads to higher requirements for the resolution of seismic data. This paper presents a high-resolution seismic data processing method of U-net based on an adaptive attention mechanism. This method first uses the U-net structure to learn the feature representation of seismic data， extracts the abstract features of seismic data through the encoder of the down-sampling process， and then reconstructs and refines the features through the decoder of the up-sampling process. The attention mechanism is introduced in the process of up-sampling， which is utilized to adjust the attention of the network to diffe-rent seismic features. Therefore， the network can capture more details and features of the seismic data more effectively. The experimental results of synthetic seismic records of the Marmousi model and real data show that the new network has less error and is more stable than the original U-net， as it can effectively improve the prediction accuracy and realize the high-resolution processing of seismic data.

 

In recent years， Tarim Oilfield has adopted the method of “overall deployment of the regional belt and centralized and efficient acquisition” in the desert area of the platform area in Tarim Basin， accelerating the full coverage of high-precision 3D seismic data in the area. Seismic acquisition faces many challenges such as heavy workload， short acquisition time window， and simultaneous construction by multiple teams. Blending caused by the close construction distance of multiple seismic teams will seriously affect the effective weak signals in deep layers， and extending the excitation time interval and the alternate acquisition will greatly reduce the acquisition production efficiency. To solve this problem， this paper conducts the qualitative analysis of numerical simulation and the quantitative analysis of large offset and long array test data in the work area and the 120 s long background record recorded in the field. It comprehensively establishes the time-space chart of data blending analysis and defines the types and mechanism of data blending in the desert area. Finally， the stimulation schemes of borehole shooting under different distance conditions are proposed， which have been successfully applied to the 2408 km² continuous 3D seismic acquisition of the desert in Area II of Fuman Oilfield. The practice indicates that the proposed method effectively avoids interference from blending， improves the daily efficiency of acquisition by about 35%， and ensures that the project is high-quality， efficient， and completed on schedule. The method consolidates the database for efficient oil and gas exploration and development， provides a reference for 3D seismic acquisition in large desert areas， and contributes technical ideas for improving synchronous excitation systems.

In onshore exploration, surface waves are one of the common interference, and the amplitude-preserving effect of surface wave suppression has a significant impact on subsequent data processing. The traditional surface wave suppression method is difficult to effectively suppress the surface wave while protecting the reflected wave. Based on the characteristics of surface waves, which have low speed, low frequency, and high amplitude, this paper proposes a time-frequency domain amplitude-preserving surface wave suppression method based on Gabor transform. Firstly, the seismic trace is divided into several segments using Gabor transform. Then, surface waves are identified in the seismic trace segments based on the difference in the amplitude energy between surface waves and effective waves, achieving a fine separation of surface waves and effective waves in the time domain. Finally, bandpass filtering is applied to the seismic trace segments containing surface waves in the frequency domain, and surface waves are suppressed based on the difference in frequency between surface waves and effective waves. The seismic trace after suppressing surface waves is obtained using Gabor inverse transform. Model tests and actual data processing show that this method effectively suppresses surface waves while protecting the information of effective waves to the greatest extent, improving the the amplitude preservation of surface wave processing.

The ultra-deep oil and gas reservoir gradually becomes an important growth point of explored reserves and production nationwide and worldwide， and the research and application of ultra-deep oil and gas reservoir seismic prospecting technology are of great significance. To this end， with the ultra-deep oil and gas reservoir in Tarim as the research target， the difficulties faced by the seismic and geological conditions and seismic prospecting of ultra-deep oil and gas reservoirs are analyzed. The formation， development， and application results and effects of the third-generation seismic prospecting technology for ultra-deep oil and gas reservoir are summarized， and the development strategy of this technology is prospected. The research suggests that the first generation of ultra-deep oil and gas reservoir seismic prospecting technology mainly based on wide-line and 3D geometry， high-velocity layer excitation， and pre-stack time migration has settled the low signal-to-noise ratio in ultra-deep seismic data. The second generation of ultra-deep oil and gas reservoir seismic technology mainly based on 3D， wide azimuth， higher-density geometry， and prestack depth migration has preliminarily resolved the inaccurate location of ultra-deep seismic data imaging. The third generation of ultra-deep reservoir seismic technology is mainly based on high-density and wide-azimuth 3D geometry， enhanced energy and noise reduction for seismic excitation， deep layer spectrum spread and fidelity， true surface TTI depth migration imaging，segmented fault interpretation and combination， OVT domain phase-controlled inversion and interpretation， etc. This techno-logy has basically overcome the high-precision imaging problem of ultra-deep seismic data and effectively supports the discovery and exploration of the Fuman ultra-deep ten-billion-ton oilfield and the Bozi-Dabei ultra-deep trillion-cubic-meter gas field. The energy enhancement and noise reduction， the spectrum spread and fidelity， the focus location， and the decoupling mapping will be prioritized in the future development of ultra-deep oil and gas reservoir seismic prospecting technology.

Relying on the prior knowledge of geological experts， the traditional identification methods for sedimentary facies use seismic and logging data to conduct a qualitative analysis of sedimentary environments with the aid of the storage and computation capacity of computers. As sedimentary facies identification based on seismic data requires a lot of manual interpretation， the accuracy and efficiency are not ideal. How to characterize the geological characteristics of sedimentary microfacies from seismic data and realize the three-dimensional spatial characterization of sedimentary microfacies remains to be studied. In recent years， the knowledge graph has attracted wide attention in the field of geoscience， and the traditional identification method for sedimentary facies can be improved by constructing the knowledge graph as a constraint. However， it is an urgent technical problem to further integrate the knowledge graph， deep learning， and seismic identification technology of sedimentary facies to form a fine identification technology of sedimentary microfacies constrained by the knowledge graph. By introducing geological prior knowledge into the knowledge graph， this paper constructs a high-level semantic cognition system for complex underground sedimentary patterns. The knowledge graph is used for computer representation of geological prior knowledge， which can serve as constraint conditions and quality control measures to guide the identification and modeling of sedimentary microfacies. It ultimately forms an intelligent identification and modeling technology for sedimentary microfacies guided by the knowledge graph. After digitizing geological prior knowledge， the presented method characterizes the spatial distribution of carbonate microbial mound-beach complexes and multi-stage foreset bodies in the Dengying Formation of the central Sichuan Basin. The predicted results are in line with the geological condition of the target area. The proposed method is suitable for deep lithologic trap identification and well demonstration， providing an effective basis for reservoir prediction and has good industrial application value.

Due to the large error in conventional indirect prediction of fracture density using normal weakness and tangential weakness， this paper proposes a direct inversion method for fracture density in carbonate rocks （HTI medium）. Based on the inversion of amplitudes varying with offset and azimuth （AVAZ）， the paper introduces an inverse distance weighted method to optimize the transverse difference operator of traditional anisotropic total variational（ATV） multichannel inversion. The method fully utilizes the correlation between adjacent seismic tracks to further improve the horizontal continuity and stability of the inversion algorithm. Lateral constraints， low-frequency constraints， and sparse constraints are applied to construct the inversion objective function， and the alternate direction multiplier method （ADMM） is adopted to optimize the inversion objective function. With the two advantages of direct inversion of fracture density and inverse distance weighted interpolation， model tests are conducted on the Marmousi II model to verify the effectiveness and noise resistance of the proposed method. Then， the paper further verifies the practical feasibility of the proposed method by applying it to the actual data of carbonate rocks in YM block， Tarim Basin. The proposed method is more accurate and reliable for predicting fracture density and can be widely applied in similar areas.

Fault recognition is an essential step in seismic data interpretation. The development of deep learning has effectively improved the efficiency and accuracy of automatic fault recognition. However, in automatic fault recognition, it is still challenging to accurately capture subtle structures of faults and effectively resist noise interference. Thus, in this study, we propose a high-resolution fault recognition network model, AHRFaultSegNet, based on the HRNet network and decoupled self-attention mechanisms. The decoupling of self-attention mechanisms combines spatial attention and channel attention, replacing parallel convolution layers in HRNet. This reduces the computational amount of traditional self-attention mechanisms while allowing the model to calculate the relevance of input feature on a global scale, thus more accurately modeling non-local features. In decoupled self-attention, the residual connection is employed to preserve the original feature, speeding up model training and better maintaining detailed information. Experimental results demonstrate that the proposed network model outperforms other common automatic fault recognition network models in performance evaluation indexes such as Dice, Fmeasure, IoU, Precision, and Recall. Through fault recognition experiments on synthetic seismic data and actual seismic data, this method is proven to be effective in subtle fault structure identification and robust in noise resistance.

With the deepening of exploration, the a priori information available for regularization inversion becomes increasingly rich and accurate. A model constraint method is proposed based on the principle of mutual differences in the physical properties of inversion units, after an analysis of methods that introduce a priori rock physical property constraint, including parameter transformation functions, the fuzzy C-means (FCM) clustering model constraint and the alternating direction optimization of multipliers. The physical properties are introduced via the imposition of constraints on the values assigned to discrete inversion units, given that the essence of the model constraint based on the principle of mutual differences among inversion units is the finite number of geological units and that each discrete inversion unit can only belong to one lithology. A unified objective function is constructed through the combination of the aforementioned introduction of a priori physical properties with classical regularization inversion and is then optimized through the Gauss-Newton method. The mechanisms and mathematics essence underlying the introduction of rock physical properties through four distinct schemes are examined, and a strategy of employing multiple schemes to improve the inversion effect is proposed. Finally, the inversion effect of combining L₁ norm minimum structure model constraints and the introduction strategy of a priori physical property constraints is compared with that of combining L₂ norm minimum structure model constraints and the same strategy. The inversion results of the model and field data all demonstrate that leveraging a priori rock physical property information can effectively improve the inversion effect.