As crucial storage spaces for underground hydrocarbon resources, karst caves are essential for the exploitation and development of carbonate-prone reservoirs. The precise characterization of karst caves enables efficient resource management and enhances the effectiveness of oil and gas exploration and development. Due to noise interference and discontinuities in geologic strata, traditional methods that rely on seismic properties to detect karst caves are highly subjective, prone to multiple solutions, and require extensive manual intervention, resulting in low development efficiency. With the growing demand for exploration, overcoming the limitations of traditional methods through more efficient and precise karst cave identification technologies has emerged as a critical focus in carbonate-prone reservoir research. To address these limitations in 3D karst cave identification, we introduce a variant architecture of TransUnet, which integrates transformer and dual attention mechanisms with a convolutional neural network while upgrading the convolutional module from two dimensions to three dimensions to describe better the location and boundary characteristics of the karst caves in field seismic data. Compared with traditional U-shaped architectures, the model offers a faster convergence rate (epochs <10) during training and fewer false predictions during validation with open-source synthetic data. Further applications of physical models and field seismic data provide more accurate details on the positions and boundaries of karst caves, demonstrating their significant potential to improve the accuracy and resolution of karst cave detection and effectively enhance the characterization of carbonate reservoirs.

The accurate imaging of shallow marine subsurface structures is important for oil/gas exploration and geotechnical applications. Guided waves from a dense ocean-bottom seismometer acquisition program over the Qiuyue areas of the East China Sea were utilized to estimate P- and S-wave velocities. A data-driven dispersion spectra inversion network (DSINet) that directly maps the dispersion spectra to both P- and S-wave velocity (Vp and Vs) models is presented. DSINet was trained with synthetic inputs that consisted of S- and P-wave velocity models with their corresponding theoretical dispersion spectra. Transfer learning was then introduced to enhance the applicability of the network to the dispersion spectra generated from seismic shot gathers. A well-trained DSINet was used on synthetic data. The results enable a detailed assessment of the network performance. Subsequently, the field data from the Qiuyue demonstrated that the DSINet has the ability to retrieve plausible subsurface P- and S-wave velocity structures. The improvements in the migration image demonstrate that the guided-wave dispersion spectra can reconstruct reliable Vp and Vs results. Agreements between the predicted data and the well information highlight the potential of our proposed approach as an alternative method for generating Vp and Vs models and can provide a site-specific Vp-Vs relationship.

Fluid factor plays a crucial role in seismic fluid discrimination, but it is influenced by the complex petrophysical properties of reservoir rocks, which also leads to the statistical diversity of reservoir parameters. In this study, the commonly used Russell fluid factor is first investigated using the inclusion-based rock-physics theory. Rock-physics analysis reveals that the Russell fluid factor is constrained by lithology and a dry rock parameter that is difficult to determine for real reservoirs. To address these limitations, a new fluid factor is developed. Theoretical analysis and field data application illustrate its wide applicability. In addition, to reduce the prediction bias caused by the complex petrophysical properties, a generalized mixture prior considering the lithology contribution is developed. This innovative approach enables the characterization of the lithology-based probabilistic statistical distributions of different reservoir parameters. Field data tests, such as well-logging and seismic data, demonstrate the advantages of our approach, which is helpful in identifying the hydrocarbon-bearing zones in complex reservoirs.

Seismic facies analysis plays a key role in stratigraphy and has developed rapidly in the context of deep learning methods. However, seismic facies analysis cannot only be regarded as an image segmentation or classification task because this would exclude prior knowledge of geophysicists on seismic interpretation. Seismic parameters summarized by geophysicists were included in the seismic facies analysis model. Seismic parameters were used to calibrate the different seismic facies units, and well logging data and sequence stratigraphy were integrated to perform geological interpretation and to establish a training dataset. A multi-head seismic parameter classification (MSPC) model was then established to divide the seismic facies units. A seismic facies geological interpretation (SFGI) model was constructed for the geological interpretation of seismic facies units, and an end-to-end seismic facies analysis model was designed based on the MSPC model and the SFGI model. The proposed model effectively improves the accuracy of the seismic interpretation, solves the intra-facies noise problem, and is easy to train and use.

Seismic full waveform inversion (FWI) faces cycle-skipping problem when the seismic data lack low-frequency information and the initial velocity model is far from the true model. To address this challenge, we proposed a neural network-based weighted envelope correlation inversion with a learnable envelope power, which is referred to as FWI-NN-WECP. In this method, we constructed the loss function using an envelope-based misfit to enhance low-frequency components during the inversion. We also introduced a learnable envelope-power parameter to control the effective bandwidth of the data envelope. In addition, we applied a regularization term to further constrain the evolution of the power parameter, thereby guiding the inversion to progressively focus from shallow to deep parts of the velocity model. Synthetic data experiments demonstrated that the FWI-NN-WECP method significantly mitigates cycle skipping and successfully reconstructs complex geological structures, even when low-frequency data (below 5 Hz) are absent and the initial model is inaccurate. The field data experiment confirmed the accuracy of the inverted velocity models and demonstrated the potential of the proposed method to mitigate cycle-skipping through by using neural networks and a carefully designed loss function.

Accurate estimation of S-wave velocity from well logs is relevant for processing seismic data from hydrocarbon reservoirs, CO2 storage sites and geothermal systems. Standard methods rely on empirical relationships and basic rock physics models (RPMs), such as the Gassmann equation, which are often inaccurate because they cannot account for the complex factors that influence the relationship between S-wave velocity and other properties. Machine learning techniques, including deep neural networks (DNNs), have been shown to be effective methods for indirect estimation of S-wave logs and other rock properties. In this work, we develop a novel prediction method that combines RPMs (Xu-White and Biot-Rayleigh (BR) models) and DNNs based on pseudo-labels, i.e., a class of algorithms that use model outputs to assign labels to unlabelled samples, which are then used as labelled samples during training. The BR model takes into account the volume ratio of inclusions and it is treated as a spatially varying parameter, allowing a better description of the pore microstructure. Data from five boreholes demonstrate the applicability of the proposed model. A single well is used for training, while the remaining four wells are used as a blind test. Compared to DNN without physical poroelasticity models, the method driven by models show a significant improvement in prediction accuracy. By incorporating pseudo-labels into the BR model, the prediction performance of the network model is improved. The method proposed in this work is validated and applied in other work areas.

Electrical surveys are widely employed for the exploration of geothermal energy, a clean and renewable resource that can help to support the sustainable development of our society. Efforts have been made to increase the understanding of the temperature effects on the electrical properties of reservoir rocks that will help for the interpretation of electrical survey data. However, there has been no valid rock physics model for the electrical conductivity of sandstones with varying temperature that can practically aid the quantitative interpretation of electrical survey data for geothermal energy. We have developed an approach for modeling the temperature dependent electrical conductivity of clean Berea sandstones. Using theoretically inverted values for the aspect ratio of the insulating grains, and for the aspect ratio of the temperature induced variation in the porosity of the microcracks, in combination with the calculated temperature dependent brine conductivity, the modeled electrical conductivity with varying temperature compares satisfactorily with a published dataset on five Berea sandstones, with relative errors of better than ±5% for all cases. The results demonstrate the applicability of the developed approach for modeling the temperature dependent electrical conductivity of sandstones with a minor amount of clay minerals, and thus pave the way to the improved quantitative interpretation of electrical survey data for the exploration of geothermal energy.

A transient electromagnetic vector finite-element forward-modeling method based on 3D anisotropic resistivity constrained by well-logging and seismic data is developed. This method integrates detailed stratigraphic division provided by well-logging data and structural information from seismic interpretation to establish a structural model that reflects the actual geologic conditions. Based on the structural model, a kriging interpolation method is applied to the logging resistivity data to obtain the anisotropic resistivity distribution information of each layer in the reservoir, thereby establishing a 3D anisotropic resistivity model. To ensure the smoothness of the model, we use a variogram function based on a spherical model to fit the discrete interpolation points. After the resistivity model is spatially discretized by locally refined unstructured tetrahedral meshes and temporally discretized by a nonuniform time step, the large sparse matrix equation is solved using the PARDISO direct solver to obtain the forward response. We apply this method to layered models of isotropic and vertical transversely isotropic (VTI) media and verify the accuracy of the forward simulation method by comparative analysis with quasi-analytical solutions. Doubling the nonuniform time step in our forward simulations does not affect the accuracy of the solution. Applying our method to China’s D oilfield, the electromagnetic response at observation points generally matches traditional methods well. Comparative analysis of the forward results between our method and the model without geologic data or homogeneous resistivity shows that we achieve better fitting with the field measurement data and provide more detailed electrical properties.

Geologic carbon storage (GCS) must be safe and profitable. To achieve these goals for gigaton-scale GCS operations, decision-making in the presence of uncertainty is required. Geophysical monitoring methods can inform such decisions, given their sensitivity to the spatiotemporal changes in the subsurface during and after injection. We investigate a novel framework for the optimal control of GCS operations using geophysical monitoring. We refer to this decision-making tool as “geophysical control” and develop sequential decision-making models trained using digital twins of GCS operations and the corresponding geophysical monitoring signals. In particular, we obtain these models via deep reinforcement learning (DRL) and specifically focus on two types of uncertainty: geophysical noise and uncertainty in the subsurface petrophysical model. Our objective is to demonstrate how each source of stochasticity affects the decision-making process when one seeks to maximize profit while minimizing the risk of induced seismicity through an optimal policy that determines the annual target CO2 injection rate. We train a suite of DRL agents with different geophysical observations (surface time-lapse gravity, surface seismic amplitude-variation-with-offset (AVO), and combined gravity and AVO surveys), different signal-to-noise ratio levels, and with/without petrophysical model uncertainties. A comparison of the learning behavior of these independent DRL agents shows that (1) the DRL framework has the capacity to learn optimal CO2 injection policies; (2) training performance degrades with increasing geophysical noise (especially more in the seismic AVO case); and (3) the combination of AVO and gravity enhances decision-making, especially in the presence of geophysical noise. Our results show that the use of multigeophysical measurements and the incorporation of subsurface model uncertainties are critical in developing robust injection control agents using DRL.

Efficient and accurate traveltime calculations of seismic waves have important applications in traveltime tomography for initial velocity model building, Kirchhoff depth migration, earthquake location, etc. Because anisotropy significantly affects the traveltimes of seismic waves, neglecting it would result in inaccurate imaging and inversion of the underground structure. Therefore, it is necessary to consider anisotropy in the traveltime calculations. The fast sweeping method (FSM) has important applications in computing the anisotropic first-arrival traveltime. The conventional method, which solves a transformed traveltime quartic equation combined with the FSM, is well suited for general anisotropic media and does not rely on approximations to Christoffel’s equation. However, this method has the following problems: (1) the form of the traveltime quartic equation is highly complex and exhibits multiple solutions, requiring additional effort to find solution intervals and determine which solution meets the criteria, and (2) Solving high-order polynomial equations may encounter numerical instability, especially when coefficients undergo minor variations or rounding errors occur for 3D problems. To address the stability and efficiency issues in anisotropic traveltime calculation, we analyze the characteristics of the constructed triangular-pyramid local solver and the quartic coupled slowness equation for the qP and qSV waves in tilted transversely isotropic (TTI) media. We observe that the decomposed slowness equation yields only one or no solution in the triangular-pyramid local solver and satisfies monotonicity conditions. Therefore, we develop the use of the Newton method to solve the factorized slowness equation efficiently, thus addressing stability and computational efficiency concerns inherent in the conventional approach. For the qSH wave, its slowness equation is quadratic and simple to solve. Our method provides an efficient and stable procedure for the traveltime calculations of qP, qSV, and qSH waves in 3D general TTI media. Numerical examples verify the efficiency and accuracy of our method.