Abstract Conventional shaped-charge perforation completion for low-permeability oil and gas reservoirs faces persistent challenges, namely compaction damage zones, high seepage resistance, and reduced productivity. To address these issues, two novel perforation techniques are proposed: self-cleaning perforation and after-effect body perforation. Analytical productivity prediction models for both methods are developed on the basis of the Joshi model, and their reliability is rigorously verified. Furthermore, dynamic numerical simulations of the perforation process are carried out, and the simulation results are validated against surface experiments. With the maximization of the productivity ratio as the optimization objective, the applicability and sensitivity of each technique to reservoir physical parameters are systematically investigated through orthogonal experimental design, employing level k values and range coefficient R. The results indicate that the maximum deviation between the productivity prediction model and actual field data is 6.73%, while the discrepancy between dynamic simulations and experimental measurements is 7.74%. In terms of applicability, self-cleaning perforation exhibits favorable performance under porosity conditions of 15%-25%, confining pressure of 10-20 MPa, and elastic modulus of 2.4-3.6 GPa; in contrast, after-effect body perforation demonstrates superior performance at porosity of 5%-15%, confining pressure of 20-40 MPa, and elastic modulus of 3.6-14.5 GPa. Both techniques perform comparably when compressive strength ranges from 20 to 60 MPa. Sensitivity analysis of reservoir physical parameters reveals the following order of influence for self-cleaning perforation: confining pressure > compressive strength > porosity > elastic modulus; for after-effect body perforation, the order is: compressive strength > porosity > confining pressure > elastic modulus. This study provides theoretical guidance for the application of these two techniques in low-permeability reservoirs and holds significant implications for oil and gas well stimulation.
Abstract High-resolution characterization of subsurface geophysical parameters is fundamental for hydrocarbon exploration, mineral prospecting, and geological structure analysis. Full waveform inversion can potentially estimate high-precision parameter models by solving a wave-equation-constrained optimization problem when initialized with an accurate starting model. While Implicit Full Waveform Inversion (IFWI) employs multilayer perceptrons (MLPs) to reparameterize velocity models and reduces dependence on initial model accuracy, the inherent low-frequency bias and limited high-frequency representation capacity of MLPs result in blurred boundaries and poor robustness to noisy data. To further improve the performance, we propose a Boundary-Enhanced IFWI method (BEIFWI) that integrates implicit neural representations with a Sobel-integrated convolutional neural network. Specifically, the Sobel layer is introduced to enhance boundary features in the inversion results, effectively mitigating the boundary blurring inherent in standard IFWI. Furthermore, a residual connection mechanism is employed to preserve shallow-layer details while optimizing deep-layer representations, thereby stabilizing the representation process and minimizing the least-squares seismic-data misfit during network training. Experimental results on the two-dimensional Marmousi and SEG/EAGE Salt models, as well as the Chevron 2014 blind test, show that BEIFWI obtains inversion results with sharper boundaries. Even under challenging scenarios, for severely noisy seismic data and low-frequency deficient records, BEIFWI can stably estimate velocity models, thereby highlighting its superior noise resistance and inversion accuracy improvement.
Abstract Understanding and characterizing seismic facies is essential for reconstructing depositional settings, identifying stratigraphic boundaries, and supporting reservoir evaluation in both exploration and development contexts. In this study, we present a spatially constrained self-organizing map (SOM) framework for seismic facies analysis that integrates multi-attribute information while incorporating local spatial consistency during training. Unlike conventional SOM approaches that rely solely on attribute similarity, the proposed method introduces a spatial regularization term into the best matching unit (BMU) assignment, promoting consistent mapping among neighboring samples while preserving discontinuities associated with structural or stratigraphic boundaries. The spatial constraint is implemented through a local weighting function that reflects structural continuity; in this implementation, seismic coherence is used as an indicator, although the formulation is not restricted to a specific attribute. Following SOM training, the attribute manifold is further represented using a geodesic-based one-dimensional ordering that captures intrinsic relationships among SOM nodes, providing a continuous representation of facies variability and improving interpretability. Application to a deep-water 3D seismic dataset demonstrates that the proposed framework reduces local instability in facies mapping, enhances spatial continuity, and preserves boundary sharpness compared with conventional SOM-based workflows.
Abstract To address the challenges of least-squares reverse time migration (LSRTM), where the convergence speed and gradient stability are difficult to balance, and traditional optimization algorithms are not well aligned with physical characteristics, the rational design of optimization methods is crucial for enhancing the computational efficiency and inversion accuracy of LSRTM. Mainstream optimization methods have certain limitations: limited-memory Broyden-Fletcher-Goldfarb-Shanno (L-BFGS) achieves fast convergence but incurs substantial memory usage and lacks iterative robustness. Hybrid conjugate gradient (HCG) methods typically combine two conjugate coefficient formulae, feature strong stability and low memory demand; however, they exhibit slow convergence and limited consideration of the physical essence of LSRTM. To overcome these drawbacks, this paper proposes a wavefield-energy adaptively constrained triple hybrid conjugate gradient (WE-ATCG) method. It adopts wavefield energy as a physical metric to determine weighting factors, and dynamically integrates the unique strengths of three conjugate coefficients: Fletcher-Reeves (FR) ensures the stable convergence of shallow stratas, Polak-Ribière+ (PRP+) enhances restart optimization, and Dai-Yuan (DY) accelerates the convergence speed. This method retains the traditional HCG with the advantages of high stability and low memory requirements while improving the convergence of quasi-Newton methods, and simultaneously achieves adaptive gradient balancing between the shallow and deep strata. The WE-ATCG method was applied to LSRTM, and its inversion performance was systematically evaluated through acoustic LSRTM model testing and field data validation. The results demonstrate that the WE-ATCG effectively synergizes the advantages of the three conjugate coefficients, enhances the gradient-restart capability, and improves the convergence efficiency and iteration stability of the LSRTM.
Abstract As a core product of tectonic evolution, fractures play a key role in hydrocarbon accumulation and development. They serve as primary migration pathways, provide secondary reservoir space, and form lateral or top–bottom seals to regulate hydrocarbon enrichment and distribution. Although fracture connectivity evaluation methods have become increasingly mature, quantitative studies on static fracture conductivity remain relatively underdeveloped. This parameter directly determines transmission efficiency but faces severe challenges in data acquisition and scale conversion, particularly in deep carbonate reservoirs. This paper proposes a method for evaluating static fracture conductivity in carbonate reservoirs based on seismic attributes: first, attributes such as ant-tracking volume, likelihood volume, and coherence volume are selected and fused using a feedforward neural network to generate a comprehensive attribute volume; second, a seed-growing algorithm is employed for automated segmentation of complex strike-slip fractures; third, local fracture widths are statistically determined via a moving window approach; and finally, fracture conductivity is calculated based on the cubic law under the parallel-plate model assumption. Application in the AD4 well area of the Tahe Oilfield, Tarim Basin, China, reveals strong spatial heterogeneity in conductivity, with wide fracture zones forming high-contrast seepage networks. Interwell connectivity validation achieves an 83% match rate, with only localized influences from geological heterogeneity. This work provides a practical workflow for optimizing well placement and enhancing recovery in deep carbonate reservoirs.
