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.

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.

To achieve low frequency sweep with conventional vibrators, the sweep length of the sweep signal needs to be extended, which leads to the decrease of production efficiency and the increase of acquisition cost. Therefore, the frequency blending simultaneous sweep method of vibrators is proposed. In this method, a long sweep signal is split into several short sweep signals for frequency blending simultaneous sweep, which realizes the transformation of vibroseis sequential single frequency sweep to multi-frequency simultaneous sweep. There are two simultaneous sweep methods. One is array simultaneous sweep with high SNR data and low blended noise. The other is independent simultaneous sweep with higher efficiency and slightly higher blended noise. Field tests have shown that the frequency blending simultaneous sweep not only has high production efficiency, but also low blended noise. Special deblending is not necessary in data processing, and the quality of seismic data can be greatly improved. It has been proven that the frequency blending simultaneous sweep method is a cost-effective solution for conventional vibrators low frequency sweep.

In two-dimensional oil and gas exploration, the wide spacing between seismic traces in field data acqui-sition is often constrained by factors such as field construction conditions and cost-effectiveness. Such seismic records are prone to spatial aliasing, leading to a decrease in the signal-to-noise ratio of their frequency velocity spectra. To address this issue, this paper proposes utilizing the excellent time-frequency analysis characteristics of wavelet transform to perform wavelet-based inter-traces interpolation on densely sampled records. This is done to enhance the accuracy of high-frequency information of Rayleigh surface waves in P-SV converted wave seismic data, thereby improving the accuracy of inverting near-surface shear wave velocity structures and enhancing the static correction accuracy of P-SV converted waves. Verification calculations on theoretical models and actual data demonstrate that the wavelet-based interline interpolation method exhibits anti-aliasing properties, resulting in clearer inflection points in the dispersion curves of Rayleigh surface waves and a more concentrated energy in the high-frequency range. The accuracy of the inverted shear wave velocity is higher, achieving good results.

With the increasing complexity of the structure and the surface geological conditions of the exploration target,the problems of irregular and incomplete data often occur in the process of seismic data acquisition,which brings serious difficulties to the follow-up data processing. To solve this problem, this paper proposes a seismic data reconstruction method based on the XGBoost algorithm. From the perspective of local learning,this method selects a certain number of adjacent seismic traces around the randomly missing seismic traces as a reference. By constructing the regression model between the trace numbers,sampling point numbers and their values of the reference seismic traces,the missing seismic trace data can be accurately learned and reconstructed. In order to fully evaluate the performance of the proposed method,the experiments are performed on simulated data with different missing seismic traces,and the reconstruction methods such as U-net convolutional neural network and Curvelet algorithm based on projection onto convex sets are compared. The experimental results show that the reconstruction method based on the XGBoost algorithm presents high accuracy in the reconstruction of randomly missing seismic data. The actual data processing results show that this method can provide high-precision regular shot gather for the follow-up seismic data processing.

The finite difference （FD） method is widely used to solve different types of seismic wave equations.It’s crucial to simultaneously improve the FD accuracy for seismic wavefield simulation in the temporal domain and the spatial domain. To promote the accuracy and flexibility of the existing high-order FD method in the spatio-tempral domain, this paper proposes an improved spatio-tempral high-order FD method based on rectangular/cuboid grid elements and solves the 2D and 3D acoustic equations, respectively.The improved FD method combines off-axis grid nodes and traditional axial nodes to jointly approximate partial derivatives and further applies plane wave theory and Taylor series expansion method to compute the high-order FD coefficients of the improved stencils. Numerical accuracy analysis indicates that the improved FD method presents higher numerical accuracy and stability than the traditional one.Several computational examples by models demonstrate that under the identical conditions of model parameters, the proposed FD method can generate higher temporal-domain simulation accuracy while keeping satisfactory spatial-domain simulation accuracy.In addition, the improved stencil can adopt different grid spacing along different spatial directions, which effectively increases the flexibility of wavefield simulation and can provide an effective wavefield extrapolation tool for seismic imaging and inversion.

The thin-slab approximation and the screen approximation are two dual-domain methods for implementing the De Wolf approximation,and are widely used in seismic wave numerical simulation and migration imaging. There are several computational problems in implementing the two algorithms, such as the space aliasing in the F-K domain, background velocity selection,thin slab thickness division, the calculation of background Green’s function in laterally heterogeneous media,and how to deal with lateral heterogeneity. How-ever, there have been no reports on systematic discussions on the aforementioned computational problems. In response to this situation, this paper first reviews the specific realization of the thin slab approximation and the screen approximation method. Secondly, a systematic study is conducted on the aforementioned computational problems in five specific aspects. ①The causes of spatial aliasing in the F-K domain are clarified, and the spatial aliasing problem is completely solved by use of vertical and horizontal amplitude attenuation factors. ②The selection of background velocity is studied, and the key to selecting background velocity is to ensure that the overall velocity disturbance inside the thin slab is relatively small as much as possible. ③Forward record and computational efficiency of the thin slab approximation method and the screen approximation method are compared under different thickness conditions of the thin slab, and bases for the division of the thin slab are discussed. ④The reasons for the instability of the background Green’s function are studied,as well as the computation error in the case of laterally heterogeneous media. ⑤The influence of the lateral heterogeneity of the media on the accuracy of the thin-slab approximation and the screen approximation is discussed, and several methods are proposed to improve the algorithm operator’s adaptability to the lateral changes of the media. Finally, the discussion results are tested by use of a three-dimensional simple layered model and a two-dimensional actual velocity model. The research results, to some extent, can provide theoretical basis for the application of the thin-slab approximation method and the screen approximation method in complex media.

For the new acoustic approximation, the velocity of qSV wave in all directions is set to 0, which eliminating residual qSV waves present in traditional acoustic approximations during the simulation process, so it is naturally able to simulate pure qP wave.The spatial approximation method using propagation direction vectors instead of wave number vectors can be used to solve the qP wave dispersion based on the new acoustic app-roximation in the time-space domain by using finite difference methods, but there may be some theoretical errors. In order to improve the simulation accuracy of pure qP wave, this paper by introducing intermediate variable, the fractional spatial wavenumber domain mixed operator in qP wave dispersion relationship based on the new acoustic approximation is transformed into a linear equation for solving intermediate variables. This converts the complex qP wave dispersion relationship into a simpler form of time-space domain pure qP wave equation group, and the derivation process does not adopt any approximations. Therefore, this equation can accurately simulate the propagation law of qP waves based on the new acoustic approximation, theoretically with higher accuracy, and achieve pure qP wave forward simulation in two-dimensional VTI media through finite diffe-rence method.The numerical results also demonstrate that the new pure qP wave equation has higher accuracy.

For models with known velocities, traditional pre-stack depth migration imaging methods aim to quickly obtain high-precision profiles. The pre-stack depth migration imaging method based on the ima-ging operator can quickly obtain the migration results without the forward modeling and migration processing of the wave field, but the imaging operator is constructed according to the scattering vector of a single scattering point, and can only accurately simulate the local area near the scattering point. In order to improve the authenticity of the direct simulated imaging results and take into account the advantages of high computational efficiency of direct simulated imaging, the coarse point diffusion operator is established under the premise of the coarse mesh discrete original velocity model, and different interpolation methods are tested on the coarse point diffusion operator in the wavenumber domain to test the influence of different scales of coarse mesh and different interpolation methods on the simulated imaging results. On this basis, a multi-operator cooperative simulated imaging is proposed, in which multiple local migration imaging windows are set up, scattering points are arranged in each window and the ray path is calculated, and the imaging operator for controlling the local imaging is constructed in the partition, so that the simulated imaging results can better reflect the influence of the acquisition geometry on the theoretical imaging results of different regions. This method not only retains the high computational efficiency of fast imaging, but also obtains the imaging operator that controls local imaging, which can simulate the theoretical simulated imaging results of different regions under the same acquisition geometry and detect the influence of the acquisition geometry on global imaging, which can directly demonstrate the control of the acquisition geometry on theoretical imaging in different regions.

The finite difference method based on a standard staggered grid is widely used in the forward simulation of surface waves. The single use of the Rayleigh or the Love wave simulation method is no longer able to meet the surface wave simulation needs in the context of transversely isotropic （TI） media. Therefore, a two-dimensional three-component surface wave simulation method based on a rotated staggered grid （RSG） for TI media is proposed. Firstly, the first-order velocity-stress two-dimensional three-component wave equation is adopted to combine the P-SV wave equation and SH wave equation. Next, the RSG finite difference method is combined with multi-axis perfectly matched layer technology to implement a two-dimensional three-component surface wave simulation including Rayleigh wave simulation and Love wave simulation. Then, qualitatively and quantitatively compare the effect and accuracy of the RSG finite difference method and the standard staggered grid（SSG） finite difference method in the simulation of surface waves in two-dimensional isotropic homogeneous half-space media, in three aspects of wavefront snapshots, waveform curves, and dispersion curves. Finally, this method is applied to the two-dimensional three-component surface wave simulation in homogeneous half-space TI media （including VTI, HTI, and TTI） and two-layer velocity-increasing TTI media. Through wavefield snapshots, seismic records, and dispersion energy diagram, the characteristics of surface waves in TI media and the effect of anisotropic parameters on surface wave dispersion characteristics are analyzed. The experimental analysis realizes the simulation of Rayleigh wave and Love wave with the same equation, which proves the applicability of the proposed method in surface wave simulation practice, and provides an important reference for full-wave seismic wave simulation and even full-wave field inversion in two-dimensional cases. Moreover, it is of great significance for further understanding the propagation characteristics of surface waves in the anisotropic media.

The resolution of reverse time migration （RTM） results highly depends on the accuracy of the seismic wave simulation. In this paper,the symplectic stereo-modeling method （SSM） is used to solve the visco-acoustic wave equation in the Birkhoffian system and RTM with amplitude compensation is conducted. First,a Birkhoffian system is constructed from the visco-acoustic wave equation. For the system,the second-order symplectic separable Runge-Kutta method is used to advance the time,the stereo-modeling method （STEM） and the conventional finite difference method are used to discrete the spatial operators,respectively,and two methods are obtained correspondingly：SSM and conventional symplectic method （CSM）. Then a series of numerical properties of SSM and CSM are analyzed,including numerical dispersion analysis,comparison with analytical solution,efficiency comparison, and long-term calculation stability test,etc. The results show that the maximum numerical dispersion error of SSM is approximately 9%,while that of CSM method is approximately 26%. SSM exhibits high calculation efficiency and stability for long-time calculation, whose calculation efficiency is approximately twice that of CSM. Finally,the forward numerical simulation and RTM experiments are conducted through the application of three models. The results show that visco-acoustic RTM imaging amplitude is higher than acoustic V-RTM （attenuated data）,and it can obtain imaging amplitude similar to acoustic A-RTM （un-attenuated data）. For reservoir model,SSM has higher imaging accuracy and smaller numerical dispersion than CSM.

The multi-wave and multi-component Gaussian beam migration method based on the elastic wave equation presents high computational efficiency and accurate imaging, and is appropriate to be applied to complex geological structures. At present, it is mostly applied to terrestrial three-component data. However, for seabed seismic data, there exist obvious adaptability issues： ①Incomplete seabed boundary conditions lead to incomplete separation of P waves from S waves, and the up-going and down-going waves will be aliased in the recording manner of the seabed, which results in false imaging. ②The general anisotropy in underground media brings kinematic information errors, which leads to problems such as inaccurate reflection wave homing, poor convergence effect of diffraction wave, and non-focusing energy in the stage of final imaging. Therefore, an elastic Gaussian beam migration imaging method is proposed, by which the anisotropy characteristics are taken into account and the four-component seismic data of the seabed including the water pressure component are directly used. Firstly, provided that the receiving interface of the seabed is isotropic, the wave field is extended based on the elastic wave equation and the complete boundary of the seabed, and the waveform separation matrix of the four-component data is obtained. Then, by joint use of anisotropic ray tracing and dynamic parameter approximation, the elastic wave Gaussian beam migration of VTI media is realized.The numerical results of the 2D layered model and the Gullfaks region model in the North Sea show that the Gaussian beam migration method adopted in this paper can suppress the imaging illusion caused by incomplete boundary conditions. It can automatically separate PP and PS waves in the seabed anisotropic media to achieve accurate imaging of converted waves, and can better recover and migrate seismic waves to improve the imaging resolution. Thus, the proposed method is especially suitable for seismic exploration of complex seabed under the conditions of large offset and wide azimuth.

Full waveform inversion （FWI） can establish high-precision velocity models of subsurface media. Due to geophone limitations, towed-streamer data often lacks reliable low-frequency components. Consequently, FWI based on towed-streamer data often encounters the cycle-skipping problem, resulting in the objective function converging to a local extremum. Ocean bottom node （OBN） seismic data contains more low-frequency components compared to towed-streamer data. Combining OBN data with towed-streamer data in waveform inversion can compensate for the low-frequency deficiency in towed-streamer data. However, the source wavelet of towed-streamer data is different from that of OBN data, and the joint waveform inversion is more severely affected by the source wavelet. This paper proposes a joint waveform inversion method for towed-streamer and OBN data that does not depend on source wavelets. When constructing the objective function, the simulated and observed data are respectively convolved with the average traces of the other, solving the problem of wavelet error in joint waveform inversion and achieving the effect of not relying on the source wavelet. The feasibility and effectiveness of the proposed method are verified through model testing using the Marmousi model synthesized towed-streamer data and OBN data.

Inversion for logging-while-drilling extra-deep azimuth electromagnetic measurement is an important technique to characterize formation parameter information. The inversion method for logging-while-drilling electromagnetic wave measurement based on regularization （physics-driven） is widely used in field interpretation, but it needs to utilize forward modeling for many times in the iterative process, which takes a long time to calculate and fails to obtain real-time inversion results. Therefore, an efficient inversion method is urgently needed for real-time inversion of logging-while-drilling electromagnetic data. In recent years, a deep learning （data-driven） inversion algorithm of azimuthal logging-while-drilling electromagnetic wave measurement has attracted widespread attention in the field of oil and gas exploration, but the algorithm relies too much on data, overlooking the Maxwell theory during the training process. Consequently, the effect of deep learning inversion is not good when the data set is not complete. In this paper, a hybrid inversion method coupling physics-driven and data-driven methods is proposed for two-dimensional anisotropic formations： a network is trained using randomly generated datasets comprising models with and without faults, based on the data from logging-while-drilling extra-deep azimuthal electromagnetic wave measurement； and model predictions are then made using the trained network. Compared with traditional deep learning methods, the prediction accuracy of the proposed method is significantly improved. Test results also show that, under the influence of different test noises, the physics-driven deep learning inversion method achieves favorable outcomes for resistivity models, exhibiting strong robustness and enhanced generalization ability.

Influenced by multiple tectonic movements, the foreland thrust-fold belt at the southern margin of the Junggar Basin exhibits segmented east-west sections and north-south zones, with deformation characteristics overlaid by vertical tectonic superposition. However, differing understandings of the tectonic deformation mechanism and styles in this area have hindered the depth of oil and gas exploration. To study the tectonic deformation mechanism and process of this area since the Neogene, this study utilizes high-precision seismic, dril-ling, and rock mechanics data. Based on actual geological conditions, it focuses on the number, strength, and thickness variations of detachment layers, combining factors such as their vertical combinations, lateral distribution ranges, syn-sedimentary processes, and pre-existing structures. Consequently, ten sets of models are designed, and comparative experiments are conducted using discrete element numerical simulation. The experimental results indicate that the strength, thickness, and combination of detachment layers control the vertical superposition relationships and structural styles of the thrust-fold belt, while the distribution of detachment layers and syn-sedimentary processes control the lateral deformation range. Pre-existing structures affect the later-stage inherited structural development. On this basis, multi-factor combination simulation experiments are carried out in sections and compared with actual seismic profiles, reconstructing the deformation processes of the southern margin of the Junggar Basin and revealing the tectonic deformation mechanism since the Neogene. This mechanism involves pre-existing faults, ancient uplifts, the vertical superposition relationship of three sets of slip layers with different nature, and syn-sedimentary processes, which control the formation and evolution of the structures in the western segment. Two sets of vertical superimposed slip layers, which are strong and thin at the bottom as well as weak and thick at the top, control the formation and evolution of the structures in the central segment. Pre-existing faults and a single set of weaker slip layers control the formation and evolution of the structures in the eastern segment. This method can provide a reference for similar areas with complex structural deformation.

A fault-fracture body is an important mode of occurrence in the deep carbonate reservoirs of the Fuman Oilfield in the Tarim Basin. For such a body, there are problems such as unclear concepts, uncertain geological models, and the lack of suitable seismic characterization. Therefore, by studying the outcrops, wells, seismic data and production dynamics data, this paper clarifies the connotation of fault-fracture bodies and establishes four geological development modes for ultra-deep carbonate oil and gas reservoirs with fault-fracture bodies. It also develops a smart technology for depicting internal seismic structures for ultra-deep carbonate fault-fracture bodies. The research results show that：①Fault-fracture bodies possess three dimensions of connotation. First, they mainly develop in the ultra-deep compact limestone strata in carbonate intracratonic platforms. Second, under the action of fracturing and faulting, they exhibit great internal porosity-permeability sto-rage performance and diverse types of storage space （including fault-fracture intergranular pores, cavities formed by the shearing and torsional action, and medium-high angle fractures）. Third, they exhibit favorable characteristics for the source connection of oil reservoirs and the migration of oil and gas, and oil reservoirs can extend to heights of kilometers, making them favorable locations for oil and gas accumulation in ultra-deep compact limestone reservoirs. ②Fault-fracture bodies develop six types of fault structural elements, which form four internal structure modes, including a single sliding surface, a pressure bulge core belt, a shearing and torsional cavity, and a lattice-like fracture network. ③ Conditional adversarial deep learning networks can effectively depict the internal structure of fault-fracture bodies, and the predicted results have geological significance and are highly consistent with actual conditions. This achievement has been well applied in actual production, and it has practical and promotional values for the efficient exploration and development of fault-fracture bodies.

Predicting the scale and determining the position of Ordovician carbonate fracture-cavity bodies in the Tarim Basin is very challenging, which affects the target design and the implementation of fracturing for oil and gas exploration. This paper proposes a method for the spatial positioning of fracture-cavity bodies based on frequency-segmented scale constraints, which effectively improves prediction accuracy. Firstly, based on the seismic data volumes of pre-stack frequency segmentation, this paper determines the tuning frequency of fracture-cavity bodies by the Gauss function to obtain scale parameters. Then the scale parameters are used to scan the threshold of the impedance curve of central fracture-cavity bodies. When the time thickness of the curve with low impedance matches the scale parameters, the constraints and calibration of the impedance threshold are completed, thereby achieving scale prediction and spatial positioning. The reliability and accuracy of the proposed method are verified through forward modeling. Practical applications show that the prediction results can better describe the scale features and spatial positions of fracture-cavity bodies, and provide a clearer relationship between well paths and fracture-cavity bodies. This method provides effective support for fracturing plans and oil and gas exploration.

The fast Fourier transform moving average （FFT-MA） method is a flexible and efficient stochastic modeling method,which is of great importance in some aspects such as high-resolution modeling of subsurface media,non-stationary modeling of complex media and uncertainty evaluation. Accurately constructing the spatial structure model is the key to generating a reasonable stochastic model by the FFT-MA method. However,in the previous research on the FFT-MA method,no effective method for accurately constructing the spatial structure model has been proposed. Therefore,an effective estimation method for spatial structure model is proposed. Based on the idea of inversion,by minimizing the spatial structure difference between the stochastic model and the logging data as well as the seismic data, the vertical autocorrelation length and the lateral autocorrelation length are estimated respectively. In order to optimize the method’s estimation performance for the spatial structure model,the edge-preserving regularization is introduced in the inversion process of the vertical autocorrelation length to enhance the stability of the inversion. In addition,seismic constraints are introduced into model optimization process to improve the stability of the stochastic model. Experimental results show that this method can stably estimate the non-stationary spatial structure model of underground media and thus helps establish a high-resolution stochastic model that accurately describes the non-stationary spatial correlation characteristics of complex reservoirs. Compared with the stochastic modeling method based on sequential Gaussian co-simulation, the FFT-MA stochastic modeling method with an optimized spatial structure model can effectively present various complex geological structures, by which complex reservoir modeling can be achieved.

The Ordovician deep fractured-vuggy reservoirs in Well Fuyuan-1 of the Fuman Oilfield in the Tarim Basin exhibit strong internal reservoir heterogeneity. During the drilling process, drilling fluid leakage mainly occurs in fracture-vuggy reservoirs, while the well sections that are emptied primarily consist of vuggy-cave reser-voirs. It is necessary to spatially characterize fracture and fault systems of different scales with specificity. Therefore, based on previous research, this study uses the heterogeneity and high-energy characteristics of the internal cavities in the strike-slip faults in seismic profiles as a starting point and proposes a method for extrac-ting seismic features and tracking energy ridges in deep fracture-vuggy reservoirs. An improved greedy algorithm with azimuth sector constraints is used to track the ridges with heterogeneous energy in fractured-vuggy reservoirs, to identify the features of core band structures or grid-like structures in fracture and fault systems. The specific steps are as follows. ① Seismic data preprocessing. Preprocessing is performed based on seismic data, including stratigraphic dip angle scanning and structural layer interpretation. ② Heterogeneous energy ridge tracking. Heterogeneous seismic information is extracted to determine the effective root mean square amplitude attribute range and data truncation based on precise well-seismic calibration results. Heterogeneous energy ridges are then identified based on threshold truncation tracking data. ③ Drilling verification and analysis. Based on the precise calibration results, the correlation between heterogeneous energy ridge data and drilling, logging, and cumulative fluid production data is analyzed to validate the fault identification results at different scales. The application example shows that the identification results of heterogeneous energy ridges in the deep fractured-vuggy reservoirs of Well Fuyuan-1 show a 98.5% match with emptying, mud loss, and other factors, which can effectively reveal the spatial development position of high-quality fractured-vuggy reservoirs and support well deployment.

The central part of the Sichuan Basin is a core production area for tight gas, where the Jurassic Shaximiao Formation is an important formation for the exploration and development of tight gas reservoirs in recent years. The channel sand body of the Shaximiao Formation exhibits rapid lateral changes, complex vertical overlapping relationships, and strong reservoir heterogeneity, which leads to low accuracy of the inversion method for post-stack wave impedance lithology. Meanwhile, the calculation process of pre-stack lithology inversion is complex and time-consuming and cannot meet the needs of rapid increase in reservoirs and production. Therefore, the pre-stack fast inversion technology based on lithologic data reconstruction is proposed in this paper. Firstly, by combining both P-wave and S-wave impedance with geological prior information, the lithology impedance logging curves are reconstructed, the lithology identification threshold is determined and the lithology impedance analytical formulas are established. Secondly, the AVO attribute analysis is carried out to derive the lithologic impedance reflection data expression characterized by intercept P and gradient G, and to reconstruct the lithologic impedance reflection data which characterizes a more prominent seismic response. Finally, high-precision three-dimensional lithofacies are quickly obtained through post-stack inversion, which is used as a three-dimensional spatial constraint for subsequent pre-stack geostatistical inversion. The application results of the technologyin a 13,000 km² contiguous three-dimensional area in the Central Sichuan Jinqiu Block show that this technology can effectively improve the stability and reliability of the inversion process while reducing time consumption. The inversion results achieve higher precision and resolution. The match rate between reservoir characterization and actual drilling exceeds 90%, and the success rate for its implementation in the exploration wells reaches 100%. Thus, the application of the technology promotes a high-efficiency exploration and development of the tight gas reservoirs in the Shaximiao Formation.

The surface and underground geological conditions in the Yingzhong area of Qaidam basin are complex, making it difficult to precisely delineate the intersalt faults in the Lower Ganchaigou Formation with conventional data of prestack depth migration. This study utilizes optimized prestack depth migration data based on multi-directional grid tomography to accurately identify the complex faults in that area. The distribution characteristics of the complex intersalt fault system are also studied with attribute analysis using azimuthal dip scanning. This study provides a basis for the delineation and selection of favorable zones in the Yingzhong area. The results are as follows. ①The optimized prestack depth migration data based on multi-directional grid tomography solves the problems of inaccurate occurrence imaging and unclear fault imaging of the lower Ganchaigou Formation in the Yingzhong area. The reinterpretation of the new seismic data reveals that in the “included angle area” formed by the intersection of Fault XI and Fault IV, there are several small detachment faults in the inter salt strata of the lower Ganchaigou Formation, which belong to different fault systems compared to the imbricate thrust faults developed in the sub-salt strata. ②Based on the new prestack depth migration data, the azimuthal dip scanning attributes are extracted, which reveals that these small inter salt detachment faults are distributed in an inverse “S” shape in the map. This is consistent with the regional compressive tectonic stress in the NNE direction, indicating that they are formed during the intense compression and stabilization stage from the late Pliocene to the Quaternary. ③The intersalt faults in the lower Ganchaigou Formation have strong control over local structures. Six favorable traps are identified in the “included angle area” and several wells are very productive, indicating that the intersalt faults are key favorable zones for future exploration.

The prediction accuracy of formation pressure is low when estimating effective pressure by obtaining vertical overburden pressure and pore pressure, due to the low accuracy of compaction curves fitted by acoustic travel time differences. The rock physical method only stays in the experimental stage or relies on an empirical model, failing to accurately predict the effective pressure in theory. The pore space stiffness theory provides a new approach for rock physical modeling considering pressure factors, but the study of effective pressure only remains in the stage of rock physical modeling and lacks methods to directly and stably predict effective pressure from seismic data. To predict effective pressure directly from seismic data, this paper proposes a method using pre-stack AVO inversion to predict effective pressure of sandstone reservoirs. Firstly, assuming that the elastic modulus ratio of a dry rock skeleton is constant, the rock physical relationship between shear modulus and effective pressure is established based on the pore space stiffness theory. Secondly, the linear transformation relationship between elastic parameters and physical property parameters is obtained by using Taylor expansion. Thirdly, the reflection coefficient equations of effective pressure, porosity, fluid modulus, matrix shear modulus and density are formulated, and a pre-stack seismic inversion method based on the Bayesian theory is constructed. Model testing and practical data applications show that the proposed method can stably predict the effective pressure in formations even with a low signal-to-noise ratio.

This paper proposes two-dimensional magnetotelluric inversion based on residual neural networks to improve the accuracy of conventional inversion with convolutional neural networks, which are affected by excessive layers. A large number of data sets are established through two-dimensional forward modeling of magnetotelluric data. The apparent resistivity and phase data in TE and TM modes are used as input to a four-channel network, and the corresponding geoelectric model is used as a label and output for supervised learning. The two-dimensional magnetotelluric inversion is achieved by utilizing residual neural networks. Based on the inversion results of geoelectric models with different noise levels, it is shown that residual networks can not only eliminate the problem of decreased accuracy caused by excessive layers but also exhibit strong noise resilience. Based on the inversion of the measured electromagnetic data in Jizhong depression, China, the resistivity distribution of deep carbonate rock is obtained, and the characteristics of the thermal storage structure in the working area are analyzed accordingly. Inversion results from theoretical models and measured data both show that the proposed method has excellent learning ability and noise resilience, and the inversion effect is stable and reliable.

The surface-borehole transient electromagnetic method （TEM） has been widely used because of its advantages such as high resistance to interference and strong response signals. In deep ore-prospecting and geological structure exploration, most oil and gas resources are stored in sedimentary rocks with obvious laminations,and the subsurface media exhibit anisotropic electrical conductivity. However, conventional inversion ima-ging techniques based on isotropic models are no longer applicable to the processing and interpretation of anisotropic data. To analyze the impact of conductivity anisotropy on the surface-borehole TEM, this paper studies electromagnetic responses in anisotropic media. According to the conductivity tensor of anisotropic media, this paper proposes a three-dimensional （3D） analysis model based on the finite element method （FEM）, using the PARDISO solver to solve problems. The transient electromagnetic responses in anisotropic media with different angles are simulated to analyze the influence of anisotropic media on the propagation and reception of electromagnetic waves. The characteristics of the transient electromagnetic responses in anisotropic media are also discussed. The results show that in vertical anisotropic media when the propagation direction of the electromagnetic wave is perpendicular to the direction of maximum conductivity, the electromagnetic response is the strongest； when the propagation direction of the electromagnetic wave is parallel to the direction of maximum conductivi-ty, the electromagnetic response is the weakest. In horizontal anisotropic media, the response of electromagnetic wave is the strongest when the electromagnetic wave is perpendicular to the direction of the minimum conductivity, and the weakest when the electromagnetic wave is parallel to the direction of the minimum conductivi-ty. In addition, this paper analyzes the electromagnetic responses of anisotropic strata and anomalies with different emission source directions during transient electromagnetic logging. The results show that the transient electromagnetic responses of anisotropic strata and anomalies are related to the direction of the emission sources. Specifically, for the principal axis anisotropic anomalies, the choice of emission source directions directly affects the determination of the anisotropic direction of anomalies. The conclusions of this paper provide valuable insights for the research on geophysical methods in oil and gas development.