Method, device, equipment and medium for cross-modal association of data in oil and gas field database

Abstract

The present application relates to the technical field of oil and gas field databases, and discloses a method, apparatus, device, and medium for cross-modal association of data in an oil and gas field database, comprising: receiving a pre-input original seismic waveform time series, and generating a time-frequency spectrum through Fourier transform; receiving a pre-input core scan image set, extracting local binary pattern texture features from the core scan image set, and obtaining a spatially continuous three-dimensional core digital model based on the local binary pattern texture features; injecting the time-frequency spectrum and the three-dimensional core digital model into a pre-generated hierarchical semantic framework to obtain a cross-modal unified semantic vector; constructing a comparative learning framework for the frequency spectrum and well logging text, and generating a spatiotemporally aligned semantic feature matrix based on the comparative learning framework; and obtaining a dynamic association map of well logging and seismic waveforms based on the cross-modal unified semantic vector and the semantic feature matrix.

Description

Technical Field

[0001] The present application relates to the technical field of oil and gas field databases, and in particular to a method, apparatus, device, and medium for cross-modal association of data in an oil and gas field database. Background Art

[0002] In the exploration and development of oil and gas fields, seismic waveform data, core image data, and well logging data are three key multimodal data sources. Seismic waveform time series can be transformed into time-frequency spectrograms through Fourier transform, reflecting the time-frequency characteristics of the formation structure. Core scan images can be used to construct three-dimensional digital core models using texture feature extraction methods such as local binary patterns (LBP) to characterize the microstructure of reservoir rocks. Well logging data, on the other hand, records formation physical properties at different depths in text form. Effectively linking these multimodal data sets is a core technical challenge in improving the accuracy of oil and gas reservoir interpretation.

[0003] In existing technologies, most independent feature extraction strategies are adopted (such as Fourier transform of seismic data, LBP feature extraction of core images, and statistical analysis methods of logging data), which leads to a lack of uniformity in the semantic representation space of different modal data and makes it difficult to establish accurate cross-modal mapping relationships. Summary of the Invention

[0004] One or more embodiments of this specification provide a method, apparatus, device, and medium for cross-modal association of data in an oil and gas field database, which are used to solve the technical problems raised in the background technology.

[0005] One or more embodiments of this specification adopt the following technical solutions:

[0006] One or more embodiments of this specification provide a method for cross-modal data association in an oil and gas field database, the method comprising:

[0007] Receive the pre-input original seismic waveform time series and generate a time-frequency spectrum through Fourier transform;

[0008] receiving a pre-input core scan image set, extracting local binary pattern texture features from the core scan image set, and obtaining a spatially continuous three-dimensional core digital model based on the local binary pattern texture features;

[0009] Injecting the time-frequency spectrum and the three-dimensional core digital model into a pre-generated hierarchical semantic framework to obtain a cross-modal unified semantic vector;

[0010] Constructing a comparative learning framework between the spectrogram and the well logging text, and generating a spatiotemporally aligned semantic feature matrix based on the comparative learning framework;

[0011] Based on the cross-modal unified semantic vector and the semantic feature matrix, a dynamic correlation map of well logging and seismic waveforms is obtained.

[0012] It should be noted that the embodiments of this specification have the following beneficial effects through the above content:

[0013] Improved cross-modal semantic unity: By integrating seismic time-frequency spectra and 3D core digital models into a hierarchical semantic framework, a multimodal semantic space mapping of seismic waveforms, core images, and well logging text is achieved. This approach overcomes the limitations of traditional independent feature extraction strategies and establishes deep semantic associations across modal data in a unified semantic vector space, effectively bridging the semantic gap between the microscopic time-frequency characteristics of seismic waveforms and the macroscopic structure of core textures.

[0014] Optimizing spatiotemporal alignment accuracy: A contrastive learning framework is used to jointly model spectrograms and well logging text. By dynamically sensing the nonlinear relationship between the temporal dimension of seismic waveforms and the spatial dimension of well logging depth, this method achieves spatiotemporal consistency in cross-modal features. This mechanism significantly improves the feature misalignment caused by the spatiotemporal dimensionality splitting of traditional methods, providing a highly accurate semantic feature matrix for subsequent dynamic association.

[0015] Innovation in the dynamic association mechanism: By integrating a unified semantic vector with a spatiotemporal alignment feature matrix, a dynamically updateable association graph is constructed. This graph can adapt to the real-time updates of logging data during the drilling process, overcoming the limitations of traditional static matching mechanisms that fail to reflect dynamic reservoir changes and providing real-time data support for development plan adjustments.

[0016] A breakthrough in 3D modeling efficiency: During the core digital model construction phase, an algorithm optimizing the spatial continuity of local binary pattern texture features significantly reduces the processing complexity of large-scale core scan image sets while maintaining accurate microstructural representation. This approach effectively addresses the efficiency bottleneck caused by feature redundancy in traditional 3D modeling.

[0017] Enhanced engineering application value: Through the deep correlation and dynamic visualization of multimodal data, seismic waveform anomaly areas, core pore structure characteristics and logging physical parameters can be cross-validated, significantly improving the accuracy of reservoir fluid identification and the rationality of well trajectory design, providing core technical support for intelligent exploration and development of oil and gas fields.

[0018] Furthermore, the receiving of the pre-input original seismic waveform time series and generating a time-frequency spectrum through Fourier transform includes:

[0019] The original seismic waveform time series is received, and a short-time Fourier transform algorithm is used to set a specified window to generate a time-frequency spectrum matrix including time, frequency and amplitude;

[0020] Obtain the time-to-depth conversion model based on well logging acoustic time difference data:

[0021] The time-spectrogram matrix is ​​input into a time-depth conversion model to generate a time-spectrogram of depth and time.

[0022] It should be noted that the embodiments of this specification have the following beneficial effects through the above content:

[0023] Improved time-frequency feature resolution: By combining the short-time Fourier transform algorithm with a configurable window function, the local time-varying characteristics of the seismic waveform are accurately captured while maintaining frequency resolution. The generated three-dimensional time-frequency spectrum matrix (time-frequency-amplitude) achieves a fine-grained representation of microscopic stratigraphic structural characteristics, providing a basic feature expression with greater temporal and spatial resolution for subsequent cross-modal correlation.

[0024] Optimizing depth-domain alignment accuracy: A time-to-depth conversion model is constructed based on well-logging acoustic time-difference data, mapping the traditional time-domain seismic spectrum to a depth coordinate system. This establishes a unified depth reference for seismic waveforms and well-logging data. This conversion mechanism effectively resolves the spatial misalignment of cross-modal data caused by nonlinear time-depth relationships, significantly improving the accuracy of matching seismic signatures with true formation depths.

[0025] Multi-dimensional feature fusion enhancement: The generated depth-time dual-domain time-frequency spectrum simultaneously preserves the temporal evolution of the original waveform and the depth distribution characteristics after transformation, forming a composite feature expression that is both time-sensitive and spatially interpretable. This multi-dimensional feature fusion mechanism provides a unified spatial anchor for cross-scale correlation of seismic data with core and well logging data.

[0026] Intelligent data processing process: Through the automated parameter mapping of the time-depth conversion model, the traditional experience-dependent method of manually calibrating the time-depth relationship of earthquakes is replaced. While reducing subjective errors, it realizes batch depth calibration of large-scale seismic data and significantly improves the engineering implementation efficiency of the data preprocessing link.

[0027] Furthermore, the receiving of a pre-input core scan image, extracting local binary pattern texture features from the core scan image, and obtaining a spatially continuous three-dimensional core digital model based on the local binary pattern texture features includes:

[0028] The receiver is a core scan image set;

[0029] Calculate the LBP value for each pixel point of each core scan image and generate an LBP coding map;

[0030] Counting the LBP histogram features of the LBP coding image to characterize the local texture distribution of the core;

[0031] Based on the LBP histogram similarity of adjacent slices, the spatial correspondence between slices is established;

[0032] The two-dimensional slices are stacked into a three-dimensional voxel grid based on the spatial correspondence to obtain the three-dimensional core digital model.

[0033] It should be noted that the embodiments of this specification have the following beneficial effects through the above content:

[0034] Enhanced microstructural characterization capabilities: By extracting pixel-by-pixel texture features from local binary pattern (LBP) encoding maps, the spatial heterogeneity of microgeological features, such as pore structure and mineral grain distribution, in core scan images is fully preserved. The LBP histogram statistical mechanism effectively captures the differences in texture patterns at different lithologic interfaces, significantly improving the digital model's accuracy in representing the complex microstructure of reservoir rocks.

[0035] 3D spatial continuity assurance: Spatial correspondence established based on the similarity of LBP histograms of adjacent slices overcomes the limitations of traditional 2D image stacking methods that rely on manual alignment. This method automatically constructs topological connections between slices through the statistical correlation of texture features, ensuring the vertical and lateral structural continuity of the 3D voxel grid and faithfully reproducing the sedimentary sequence characteristics of the subsurface rock formations.

[0036] Multi-scale feature fusion optimization: The generation process of the 3D voxel grid simultaneously integrates the microscopic texture information encoded by pixel-level LBP encoding with the macroscopic statistical characteristics of the slice-level histogram, forming a multi-scale fusion digital representation system for the core. This fusion mechanism enables the model to reflect both millimeter-level pore structure details and the spatial distribution of meter-level lithologic units.

[0037] Automated modeling process upgrades: Algorithmic processing of the entire process, from LBP feature extraction to 3D reconstruction, replaces the traditional empirical approach of manually annotating slice correspondences. This automated process not only avoids the introduction of subjective errors but also significantly improves the processing efficiency of large-scale core scan image sets, providing standardized data generation capabilities for the development of oilfield digital twin systems.

[0038] Furthermore, the step of injecting the time-frequency spectrum and the three-dimensional core digital model into a pre-generated hierarchical semantic framework to obtain a cross-modal unified semantic vector includes:

[0039] defining a template as a hierarchical semantic framework, wherein the hierarchical semantic framework includes a cross-attention mechanism between the time-spectrogram and the three-dimensional core digital model;

[0040] Obtaining interactive features based on a cross-attention mechanism between the time-spectrogram and the three-dimensional core digital model;

[0041] The interaction features are input into the fully connected layer and reduced to a cross-modal unified semantic vector.

[0042] It should be noted that the embodiments of this specification have the following beneficial effects through the above content:

[0043] Deeply optimize cross-modal feature interaction: By constructing a cross-attention mechanism between the time-frequency spectrum and the 3D core model, we achieve dynamic interactive perception of seismic waveform time-frequency features and core microtexture characteristics. This mechanism overcomes the limitations of traditional methods that simply concatenate multimodal features. It establishes implicit association rules between seismic wave propagation characteristics and rock physical properties within the feature space, significantly improving the precision of cross-modal semantic alignment.

[0044] Hierarchical semantic representation enhancement: The hierarchical semantic framework uses a hierarchical feature abstraction mechanism to integrate the macroscopic stratigraphic interface characteristics of seismic data with the microscopic pore structure characteristics of core models at multiple scales. This hierarchical and progressive semantic integration approach preserves the unique feature representations of each modality while constructing a cross-scale geological semantic association network, enhancing the geological interpretability of the unified semantic vector.

[0045] Improved semantic vector space compactness: Fully connected layers are used to perform targeted dimensionality reduction on high-dimensional interactive features, eliminating redundant seismic time-frequency noise and core texture features while preserving key cross-modal semantic information. The resulting compact semantic vector not only reduces subsequent computational complexity but also enhances the separability of different geological units through orthogonalization of the feature space.

[0046] A breakthrough in the interpretability of geological patterns: The construction of the cross-attention weight matrix reveals the physical correlation between seismic wave attenuation and physical parameters such as rock permeability. This interpretable feature interaction mechanism provides a new theoretical path for reservoir parameter inversion, breaking through the limitations of traditional black-box models in oil and gas geology research.

[0047] Furthermore, generating a spatiotemporally aligned semantic feature matrix based on the contrastive learning framework includes:

[0048] Based on the attention mechanism of the contrastive learning framework, the spectrum graph and the well logging text are mutually corrected to obtain fusion features;

[0049] The fused features are compressed into a spatiotemporally aligned semantic feature matrix.

[0050] It should be noted that the embodiments of this specification have the following beneficial effects through the above content:

[0051] A breakthrough in cross-modal dynamic correction capabilities: A cross-attention mechanism within a contrastive learning framework enables bidirectional feature correction between spectrograms and well logging text. This mechanism enables the temporal evolution of seismic waveforms to correct for depth interpretation biases in well logging text. Simultaneously, the spatial constraints of logging parameters feed back into the physical meaning of the spectrogram, forming a self-optimizing feature interaction system with spatiotemporal consistency.

[0052] Improved spatiotemporal coupling accuracy: During the feature fusion process, a dynamic allocation strategy for attention weights automatically captures the nonlinear mapping relationship between seismic signal propagation time and logging depth space. This technology overcomes the limitations of traditional linear interpolation methods and significantly improves the spatiotemporal matching accuracy of key geological elements such as identifying the top and bottom interfaces of oil and gas reservoirs and locating formation pinch-out points.

[0053] An innovative semantic feature matrix compaction approach uses a feature compression algorithm to orthogonally reduce the dimensionality of high-dimensional fusion features, eliminating redundant information between modalities while preserving the correlation between seismic waveform anomaly patterns and logging parameter mutation characteristics. The resulting semantic matrix combines low dimensionality with high information entropy, providing a lightweight and highly discriminative feature base for subsequent dynamic correlation map construction.

[0054] Enhanced confidence in geological interpretation: By comparing the attention correction trajectory formed during the learning process, we can reversely analyze the physical interaction mechanism between seismic attributes and well logging responses. This interpretable correction mechanism provides a transparent decision-making basis for reservoir hydrocarbon content identification, effectively overcoming the "black box" risk of traditional end-to-end models.

[0055] Furthermore, the dynamic correlation map between well logging and seismic waveforms is obtained based on the cross-modal unified semantic vector and the semantic feature matrix, including:

[0056] horizontally concatenating the cross-modal unified semantic vector and the semantic feature matrix to generate a fusion feature matrix;

[0057] Defining well logging nodes and seismic nodes in the fusion feature matrix, and determining similarity between the well logging nodes and the seismic nodes;

[0058] A dynamic correlation map of well logging and seismic waveforms is obtained based on the similarity.

[0059] It should be noted that the embodiments of this specification have the following beneficial effects through the above content:

[0060] Enhanced complementarity of multimodal features: By horizontally splicing a cross-modal unified semantic vector with a spatiotemporal aligned semantic matrix, we achieve triple information fusion of seismic waveform time-frequency characteristics, core microstructural characteristics, and well logging physical parameters. This fusion mechanism transcends the limitations of single-modal representation, cross-validating seismic signal propagation patterns with rock physical properties and significantly improving the ability to jointly discriminate reservoir sensitivity characteristics.

[0061] A breakthrough in dynamic correlation topology adaptability: A node connection mechanism based on similarity calculations automatically detects real-time correlations between changes in logging parameters and seismic waveform characteristics. This technology transcends the rigid constraints of traditional fixed-threshold association rules, forming an intelligent correlation network that can be dynamically reconstructed with incremental updates of drilling data, effectively adapting to the dynamic evolution of geological understanding during oil and gas reservoir development.

[0062] Improved accuracy in correlating geological anomalies: By defining heterogeneous logging and seismic nodes in the fused feature space, high-dimensional feature similarity metrics are used to accurately capture hidden correlation patterns between anomalous segments of well logging curves and areas of seismic waveform distortion. This technology significantly improves correlation accuracy for key geological interpretation tasks such as fault identification and gas anomaly detection, addressing the technical pain point of traditional methods' insensitivity to weak signal correlation.

[0063] Upgraded visualization support for engineering decision-making: Dynamic correlation maps visualize node connection strength and topology, directly presenting the spatial matching relationship between seismic reflection interfaces and well logging interpretation conclusions. This interactive graphical representation provides a transparent and intelligent interface for engineering decisions such as real-time well trajectory adjustment and reservoir sweet spot delineation.

[0064] Furthermore, obtaining a dynamic correlation map between well logging and seismic waveforms based on the similarity includes:

[0065] Filter the node pairs corresponding to the similarity greater than a preset threshold, where the node pairs are associated well logging nodes and seismic nodes:

[0066] The dynamic association graph is generated based on the node pairs.

[0067] It should be noted that the embodiments of this specification have the following beneficial effects through the above content:

[0068] Innovation in dynamic adaptive association mechanisms: This technology dynamically optimizes the association between well logging and seismic nodes through an intelligent screening mechanism with preset thresholds. This technology breaks through the rigid constraints of traditional fixed association rules and automatically adjusts the association strength threshold based on the actual geological scenario, ensuring that the map topology evolves adaptively with data quality and geological complexity.

[0069] Improved Correlation Noise Suppression: A node pair selection strategy based on similarity thresholds effectively filters out spurious correlation signals caused by data acquisition noise or cross-modal semantic bias. This mechanism significantly enhances the geological credibility of node connections in correlation maps, preventing random perturbations in seismic waveforms from being misinterpreted as valid reservoir responses.

[0070] Enhanced Focus on Key Geological Targets: The threshold screening process essentially creates a targeted enhancement of strongly correlated features, automatically focusing the map on areas of high correlation between logging parameter abrupt changes and seismic waveform anomalies. This focusing effect significantly improves the sensitivity of identifying key geological targets such as subtle reservoir boundaries and small faults.

[0071] A breakthrough in interpretability for engineering implementation: A preset threshold mechanism provides a transparent basis for decision-making in generating correlation maps. Geological engineers can intuitively control the balance between sparsity and reliability of the correlation network by adjusting the threshold. This explainable interaction significantly enhances technicians' trust in and acceptance of AI-generated results.

[0072] One or more embodiments of this specification provide a device for cross-modal association of data in an oil and gas field database, including:

[0073] A spectrum generating unit receives a pre-input original seismic waveform time series and generates a time spectrum through Fourier transform;

[0074] a core digital model determination unit, receiving a pre-input core scan image set, extracting local binary pattern texture features from the core scan image set, and obtaining a spatially continuous three-dimensional core digital model based on the local binary pattern texture features;

[0075] a semantic vector determination unit, which injects the time-frequency spectrum and the three-dimensional core digital model into a pre-generated hierarchical semantic framework to obtain a cross-modal unified semantic vector;

[0076] A feature matrix generation unit constructs a comparative learning framework between the spectrogram and the well logging text, and generates a spatiotemporally aligned semantic feature matrix based on the comparative learning framework;

[0077] The correlation map determining unit obtains a dynamic correlation map of well logging and seismic waveforms based on the cross-modal unified semantic vector and the semantic feature matrix.

[0078] One or more embodiments of this specification provide a device for cross-modal data association in an oil and gas field database, including:

[0079] at least one processor; and,

[0080] a memory communicatively connected to the at least one processor; wherein,

[0081] The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:

[0082] Receive the pre-input original seismic waveform time series and generate a time-frequency spectrum through Fourier transform;

[0083] receiving a pre-input core scan image set, extracting local binary pattern texture features from the core scan image set, and obtaining a spatially continuous three-dimensional core digital model based on the local binary pattern texture features;

[0084] Injecting the time-frequency spectrum and the three-dimensional core digital model into a pre-generated hierarchical semantic framework to obtain a cross-modal unified semantic vector;

[0085] Constructing a comparative learning framework between the spectrogram and the well logging text, and generating a spatiotemporally aligned semantic feature matrix based on the comparative learning framework;

[0086] Based on the cross-modal unified semantic vector and the semantic feature matrix, a dynamic correlation map of well logging and seismic waveforms is obtained.

[0087] One or more embodiments of this specification provide a non-volatile computer storage medium storing computer-executable instructions. When executed by a computer, the computer-executable instructions can achieve:

[0088] Receive the pre-input original seismic waveform time series and generate a time-frequency spectrum through Fourier transform;

[0089] receiving a pre-input core scan image set, extracting local binary pattern texture features from the core scan image set, and obtaining a spatially continuous three-dimensional core digital model based on the local binary pattern texture features;

[0090] Injecting the time-frequency spectrum and the three-dimensional core digital model into a pre-generated hierarchical semantic framework to obtain a cross-modal unified semantic vector;

[0091] Constructing a comparative learning framework between the spectrogram and the well logging text, and generating a spatiotemporally aligned semantic feature matrix based on the comparative learning framework;

[0092] Based on the cross-modal unified semantic vector and the semantic feature matrix, a dynamic correlation map of well logging and seismic waveforms is obtained.

[0093] At least one of the above technical solutions adopted in the embodiments of this specification can achieve the following beneficial effects:

[0094] Improved cross-modal semantic unity: By integrating seismic time-frequency spectra and 3D core digital models into a hierarchical semantic framework, a multimodal semantic space mapping of seismic waveforms, core images, and well logging text is achieved. This approach overcomes the limitations of traditional independent feature extraction strategies and establishes deep semantic associations across modal data in a unified semantic vector space, effectively bridging the semantic gap between the microscopic time-frequency characteristics of seismic waveforms and the macroscopic structure of core textures.

[0095] Optimizing spatiotemporal alignment accuracy: A contrastive learning framework is used to jointly model spectrograms and well logging text. By dynamically sensing the nonlinear relationship between the temporal dimension of seismic waveforms and the spatial dimension of well logging depth, this method achieves spatiotemporal consistency in cross-modal features. This mechanism significantly improves the feature misalignment caused by the spatiotemporal dimensionality splitting of traditional methods, providing a highly accurate semantic feature matrix for subsequent dynamic association.

[0096] Innovation in the dynamic association mechanism: By integrating a unified semantic vector with a spatiotemporal alignment feature matrix, a dynamically updateable association graph is constructed. This graph can adapt to the real-time updates of logging data during the drilling process, overcoming the limitations of traditional static matching mechanisms that fail to reflect dynamic reservoir changes and providing real-time data support for development plan adjustments.

[0097] A breakthrough in 3D modeling efficiency: During the core digital model construction phase, an algorithm optimizing the spatial continuity of local binary pattern texture features significantly reduces the processing complexity of large-scale core scan image sets while maintaining accurate microstructural representation. This approach effectively addresses the efficiency bottleneck caused by feature redundancy in traditional 3D modeling.

[0098] Enhanced engineering application value: Through the deep correlation and dynamic visualization of multimodal data, seismic waveform anomaly areas, core pore structure characteristics and logging physical parameters can be cross-validated, significantly improving the accuracy of reservoir fluid identification and the rationality of well trajectory design, providing core technical support for intelligent exploration and development of oil and gas fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0099] In order to more clearly illustrate the embodiments of this specification or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are only some of the embodiments described in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without inventive work. In the drawings:

[0100] Figure 1 A flowchart of a method for cross-modal association of data in an oil and gas field database provided in one or more embodiments of this specification;

[0101] Figure 2 A schematic diagram of the structure of a device for cross-modal association of data in an oil and gas field database provided in one or more embodiments of this specification;

[0102] Figure 3 A schematic diagram of the structure of a cross-modal data association device in an oil and gas field database provided in one or more embodiments of this specification. DETAILED DESCRIPTION

[0103] The embodiments of this specification provide a method, apparatus, device, and medium for cross-modal association of data in an oil and gas field database.

[0104] To help those skilled in the art better understand the technical solutions in this specification, the following will provide a clear and complete description of the technical solutions in the embodiments of this specification, in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of this specification, not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this specification without creative work should fall within the scope of protection of this specification.

[0105] Figure 1 This is a flow chart of a method for cross-modal data association in an oil and gas field database, provided in one or more embodiments of this specification. This process can be executed by a cross-modal data association system. Certain input parameters or intermediate results in the process can be manually adjusted to help improve accuracy.

[0106] The method steps of the embodiment of this specification are as follows:

[0107] S101, receiving a pre-input original seismic waveform time series, and generating a time-frequency spectrum through Fourier transform.

[0108] In the examples of this specification, the following specific implementation schemes can be used:

[0109] Seismic data preprocessing: De-noise and normalize the input seismic waveform time series to eliminate high-frequency interference signals introduced by the acquisition equipment.

[0110] Sliding window framing: Overlapping Hamming windows are used to frame continuous seismic signals to ensure local stability of time-frequency analysis.

[0111] Short-time Fourier transform: Performs spectrum calculation frame by frame to generate a three-dimensional time-frequency spectrum matrix containing time, frequency, and amplitude.

[0112] Time-depth conversion: Integrate logging acoustic time difference data and map the time axis to the real formation depth coordinate through the depth calibration network.

[0113] S102 , receiving a pre-input core scanning image set, extracting local binary pattern texture features from the core scanning image set, and obtaining a spatially continuous three-dimensional core digital model based on the local binary pattern texture features.

[0114] In the examples of this specification, the following specific implementation schemes can be used:

[0115] Core image enhancement: A histogram equalization algorithm is used to eliminate brightness differences in scanned images and improve the visual recognition of pore structures.

[0116] Multi-scale LBP extraction: Rotation-invariant local binary patterns are calculated layer by layer in the range of 3×3 to 15×15 pixels to capture rock texture features of different grain sizes.

[0117] Inter-layer topological association: The similarity of LBP histogram sequences of adjacent slices is analyzed through the dynamic time warping algorithm to construct a three-dimensional spatial adjacency graph.

[0118] Voxel model reconstruction: Based on the Markov random field to optimize the inter-layer connection relationship, the marching cube algorithm is used to generate a continuous three-dimensional mesh model.

[0119] S103: injecting the time-frequency spectrum and the three-dimensional core digital model into a pre-generated hierarchical semantic framework to obtain a cross-modal unified semantic vector.

[0120] In the examples of this specification, the following specific implementation schemes can be used:

[0121] Semantic framework initialization: Construct a three-level fusion architecture that includes low-level signal characteristics, mid-level physical attributes, and high-level geological semantics.

[0122] Bidirectional attention injection: A multi-head cross-attention module is deployed in the middle layer to establish a correlation mapping between seismic frequency band energy and core porosity.

[0123] Gated feature aggregation: Semantic expressions at different levels are integrated through residual connections, and an adaptive weight distribution strategy is used to balance modal contributions.

[0124] Semantic space compression: Use variational autoencoders to perform nonlinear dimensionality reduction on fused features to generate low-dimensional dense cross-modal semantic vectors.

[0125] S104: constructing a comparative learning framework between the spectrogram and the well logging text, and generating a spatiotemporally aligned semantic feature matrix based on the comparative learning framework.

[0126] In the examples of this specification, the following specific implementation schemes can be used:

[0127] Dual-modal encoder construction: Design parallel convolutional neural networks and Transformer architectures to extract the time-frequency features of spectrograms and the semantics of logging text, respectively.

[0128] Deformable Attention Alignment: Deploy an attention module with adjustable receptive field in the feature space to capture the nonlinear relationship between seismic wave propagation delay and logging depth offset.

[0129] Contrastive loss optimization: A hard example mining strategy is used to construct positive and negative sample pairs, and the cosine similarity loss function is used to drive feature space alignment.

[0130] Matrix orthogonalization: Perform principal component analysis on the optimized feature matrix to eliminate redundant dimensions and improve the linear separability of cross-modal features.

[0131] S105 , obtaining a dynamic correlation map of well logging and seismic waveforms based on the cross-modal unified semantic vector and the semantic feature matrix.

[0132] In the examples of this specification, the following specific implementation schemes can be used:

[0133] Multimodal feature concatenation: Concatenate semantic vectors and feature matrices along the channel dimension to construct a joint representation space and perform whitening.

[0134] Heterogeneous node definition: well logging nodes are divided according to the sampling interval of logging depth, and seismic nodes are defined according to the spatial distribution of seismic gathers.

[0135] Dynamic similarity calculation: Adopts adaptive kernel function to calculate the association strength between nodes in real time, and realizes incremental update of association weight through sliding window mechanism.

[0136] Knowledge graph visualization: Generates dynamic topology graphs based on the force-directed layout algorithm, providing node attribute query and associated path backtracking functions.

[0137] It should be noted that the embodiments of this specification have the following beneficial effects through the above content:

[0138] Improved cross-modal semantic unity: By integrating seismic time-frequency spectra and 3D core digital models into a hierarchical semantic framework, a multimodal semantic space mapping of seismic waveforms, core images, and well logging text is achieved. This approach overcomes the limitations of traditional independent feature extraction strategies and establishes deep semantic associations across modal data in a unified semantic vector space, effectively bridging the semantic gap between the microscopic time-frequency characteristics of seismic waveforms and the macroscopic structure of core textures.

[0139] Optimizing spatiotemporal alignment accuracy: A contrastive learning framework is used to jointly model spectrograms and well logging text. By dynamically sensing the nonlinear relationship between the temporal dimension of seismic waveforms and the spatial dimension of well logging depth, this method achieves spatiotemporal consistency in cross-modal features. This mechanism significantly improves the feature misalignment caused by the spatiotemporal dimensionality splitting of traditional methods, providing a highly accurate semantic feature matrix for subsequent dynamic association.

[0140] Innovation in the dynamic association mechanism: By integrating a unified semantic vector with a spatiotemporal alignment feature matrix, a dynamically updateable association graph is constructed. This graph can adapt to the real-time updates of logging data during the drilling process, overcoming the limitations of traditional static matching mechanisms that fail to reflect dynamic reservoir changes and providing real-time data support for development plan adjustments.

[0141] A breakthrough in 3D modeling efficiency: During the core digital model construction phase, an algorithm optimizing the spatial continuity of local binary pattern texture features significantly reduces the processing complexity of large-scale core scan image sets while maintaining accurate microstructural representation. This approach effectively addresses the efficiency bottleneck caused by feature redundancy in traditional 3D modeling.

[0142] Enhanced engineering application value: Through the deep correlation and dynamic visualization of multimodal data, seismic waveform anomaly areas, core pore structure characteristics and logging physical parameters can be cross-validated, significantly improving the accuracy of reservoir fluid identification and the rationality of well trajectory design, providing core technical support for intelligent exploration and development of oil and gas fields.

[0143] Furthermore, when receiving the pre-input original seismic waveform time series and generating the time-frequency spectrum through Fourier transform, the original seismic waveform time series can be received, a short-time Fourier transform algorithm can be used, a specified window can be set, and a time-frequency spectrum matrix containing time, frequency and amplitude can be generated; a time-depth conversion model established based on logging acoustic wave time difference data can be obtained: the time-frequency spectrum matrix can be input into the time-depth conversion model to generate a time-frequency spectrum of depth and time.

[0144] In the examples of this specification, the following specific implementation schemes can be used:

[0145] Step 1: Generate time-frequency spectrum matrix

[0146] Seismic data preprocessing: baseline correction and outlier removal are performed on the input raw seismic waveform time series to eliminate DC offset and sudden interference introduced by the acquisition system;

[0147] Sliding window framing: Uses a Hamming window function with adjustable length to perform overlapping framing on continuous waveforms, balancing time resolution and spectrum leakage suppression.

[0148] Short-time Fourier transform execution: Calculate the spectrum distribution of the seismic signal in each time window frame by frame to generate a three-dimensional matrix containing timestamps, frequency components and amplitude intensities;

[0149] Time-frequency graph normalization: Logarithmic compression and normalization of spectrum amplitude values ​​are performed to enhance the visual recognition of weak reflection signals.

[0150] Step 2: Time-depth domain spectrum generation

[0151] Time-depth conversion model loading: Call the pre-trained depth conversion network, which establishes a time-depth mapping relationship based on the logging acoustic time difference data;

[0152] Data format adaptation: Convert the time axis data of the time-frequency spectrum matrix into the tensor format required by the model input, preserving the frequency and amplitude dimension information;

[0153] Depth coordinate prediction: The depth value of the formation corresponding to each timestamp is obtained through time-depth conversion model inference, and a time-depth mapping lookup table is established;

[0154] Dual-domain spectrum synthesis: The original time axis and the predicted depth axis are aligned and fused to generate a two-dimensional spectrum profile with both time and depth annotated.

[0155] It should be noted that the embodiments of this specification have the following beneficial effects through the above content:

[0156] Improved time-frequency feature resolution: By combining the short-time Fourier transform algorithm with a configurable window function, the local time-varying characteristics of the seismic waveform are accurately captured while maintaining frequency resolution. The generated three-dimensional time-frequency spectrum matrix (time-frequency-amplitude) achieves a fine-grained representation of microscopic stratigraphic structural characteristics, providing a basic feature expression with greater temporal and spatial resolution for subsequent cross-modal correlation.

[0157] Optimizing depth-domain alignment accuracy: A time-to-depth conversion model is constructed based on well-logging acoustic time-difference data, mapping the traditional time-domain seismic spectrum to a depth coordinate system. This establishes a unified depth reference for seismic waveforms and well-logging data. This conversion mechanism effectively resolves the spatial misalignment of cross-modal data caused by nonlinear time-depth relationships, significantly improving the accuracy of matching seismic signatures with true formation depths.

[0158] Multi-dimensional feature fusion enhancement: The generated depth-time dual-domain time-frequency spectrum simultaneously preserves the temporal evolution of the original waveform and the depth distribution characteristics after transformation, forming a composite feature expression that is both time-sensitive and spatially interpretable. This multi-dimensional feature fusion mechanism provides a unified spatial anchor for cross-scale correlation of seismic data with core and well logging data.

[0159] Intelligent data processing process: Through the automated parameter mapping of the time-depth conversion model, the traditional experience-dependent method of manually calibrating the time-depth relationship of earthquakes is replaced. While reducing subjective errors, it realizes batch depth calibration of large-scale seismic data and significantly improves the engineering implementation efficiency of the data preprocessing link.

[0160] Furthermore, when receiving a pre-input core scan image, extracting local binary pattern texture features from the core scan image, and obtaining a spatially continuous three-dimensional core digital model based on the local binary pattern texture features, a core scan image set can be first received; the LBP value of each pixel point of each core scan image is calculated to generate an LBP coding image; the LBP histogram features of the LBP coding image are counted to characterize the local texture distribution of the core; based on the similarity of the LBP histograms of adjacent slices, a spatial correspondence relationship between slices is established; and based on the spatial correspondence relationship, the two-dimensional slices are stacked into a three-dimensional voxel grid to obtain the three-dimensional core digital model.

[0161] In the examples of this specification, the following specific implementation schemes can be used:

[0162] Step 1: Core image preprocessing

[0163] Image quality verification: Perform resolution detection and distortion correction on the input core scan image set, and remove abnormal slices with scanning artifacts or blurring and distortion;

[0164] Lighting equalization processing: Adaptive histogram equalization algorithm is used to eliminate brightness differences between adjacent slices to ensure consistency in texture feature extraction;

[0165] Image segmentation: Divide large-scale core scans into overlapping sub-regions to improve the accuracy of local texture feature extraction.

[0166] Step 2: LBP coding map generation

[0167] Multi-scale LBP operator configuration: Defines rotation-invariant LBP operators with different neighborhood radii in the range of 3×3 to 15×15 pixels;

[0168] Pixel-by-pixel feature calculation: Perform circular neighborhood sampling on each pixel point, compare the grayscale values ​​of the central pixel with the neighboring pixels to generate a binary code;

[0169] Coding map optimization: Isolated coding noise is eliminated through morphological closing operations, retaining effective texture patterns reflecting pore structure and mineral boundaries.

[0170] Step 3: Texture feature statistical modeling

[0171] Local histogram construction: Divide the LBP coding map into grid cells and count the distribution frequencies of different coding modes in each cell;

[0172] Feature vector fusion: Aggregate local histograms along the radial and circumferential dimensions of the core slice to generate a global feature vector that represents the texture distribution of the entire slice;

[0173] Feature dimensionality reduction: Principal component analysis is used to compress feature dimensions and retain key texture patterns that distinguish different lithologies.

[0174] Step 4: Establishing spatial relationships between layers

[0175] Similarity dynamic matching: The similarity of adjacent slice histogram sequences is calculated using a dynamic time warping algorithm to capture the texture gradient caused by sedimentary cycles;

[0176] Misalignment compensation calibration: Automatic correction of slice spatial offset caused by core sampling tilt based on cross-correlation peak detection;

[0177] Adjacency matrix generation: Construct a weighted graph structure that represents the three-dimensional topological connection relationship between slices, and the weight value is positively correlated with the histogram similarity.

[0178] Step 5: 3D model reconstruction

[0179] Voxel space initialization: Define the resolution and spatial range of the 3D voxel grid based on the core diameter and slice spacing;

[0180] Inter-slice interpolation optimization: A Markov random field model is used to optimize the voxel filling strategy between adjacent slices to eliminate stair-step artifacts;

[0181] Surface smoothing: Anisotropic diffusion filtering is used to enhance the continuity of pore-skeleton boundaries and generate high-fidelity 3D core models.

[0182] It should be noted that the embodiments of this specification have the following beneficial effects through the above content:

[0183] Enhanced microstructural characterization capabilities: By extracting pixel-by-pixel texture features from local binary pattern (LBP) encoding maps, the spatial heterogeneity of microgeological features, such as pore structure and mineral grain distribution, in core scan images is fully preserved. The LBP histogram statistical mechanism effectively captures the differences in texture patterns at different lithologic interfaces, significantly improving the digital model's accuracy in representing the complex microstructure of reservoir rocks.

[0184] 3D spatial continuity assurance: Spatial correspondence established based on the similarity of LBP histograms of adjacent slices overcomes the limitations of traditional 2D image stacking methods that rely on manual alignment. This method automatically constructs topological connections between slices through the statistical correlation of texture features, ensuring the vertical and lateral structural continuity of the 3D voxel grid and faithfully reproducing the sedimentary sequence characteristics of the subsurface rock formations.

[0185] Multi-scale feature fusion optimization: The generation process of the 3D voxel grid simultaneously integrates the microscopic texture information encoded by pixel-level LBP encoding with the macroscopic statistical characteristics of the slice-level histogram, forming a multi-scale fusion digital representation system for the core. This fusion mechanism enables the model to reflect both millimeter-level pore structure details and the spatial distribution of meter-level lithologic units.

[0186] Automated modeling process upgrades: Algorithmic processing of the entire process, from LBP feature extraction to 3D reconstruction, replaces the traditional empirical approach of manually annotating slice correspondences. This automated process not only avoids the introduction of subjective errors but also significantly improves the processing efficiency of large-scale core scan image sets, providing standardized data generation capabilities for the development of oilfield digital twin systems.

[0187] Furthermore, when the time-frequency spectrum graph and the three-dimensional rock core digital model are injected into a pre-generated hierarchical semantic framework to obtain a cross-modal unified semantic vector, a template can be defined as a hierarchical semantic framework, and the hierarchical semantic framework includes a cross-attention mechanism between the time-frequency spectrum graph and the three-dimensional rock core digital model; based on the cross-attention mechanism between the time-frequency spectrum graph and the three-dimensional rock core digital model, interaction features are obtained; and the interaction features are input into a fully connected layer to reduce the dimension to a cross-modal unified semantic vector.

[0188] In the examples of this specification, the following specific implementation schemes can be used:

[0189] Step 1: Building a hierarchical semantic framework

[0190] Multi-level feature abstraction architecture design: Construct a three-level hierarchical processing architecture that includes low-level signal features, mid-level physical attributes, and high-level geological semantics;

[0191] Modal feature space alignment: perform frequency band energy normalization on the time-spectrogram matrix, perform isotropic voxel resampling on the 3D core model, and unify the spatial resolution benchmark;

[0192] Cross-attention module configuration: A multi-head cross-attention mechanism is deployed in the middle layer of the framework, and a bidirectional correlation channel is set up between the seismic frequency band dimension and the core space dimension.

[0193] Step 2: Cross-modal interaction feature extraction

[0194] Dynamic calculation of attention weights: The correlation weight matrix between the frequency band energy of the time spectrum graph and the core pore distribution is established through the query-key matching mechanism;

[0195] Feature bidirectional projection fusion: Project the core texture feature vector into the seismic frequency domain, and simultaneously map the time-frequency feature into the core three-dimensional space to form bidirectional feature enhancement;

[0196] Residual feature compensation: Integrate original modal features and attention-weighted features through jump connections to retain key detail information in the cross-modal interaction process.

[0197] Step 3: Unified semantic vector generation

[0198] Feature channel cascade: splicing time-frequency-core interaction features and high-level semantic features output by the hierarchical framework along the feature dimension;

[0199] Adaptive feature compression: Using a bottleneck fully connected layer structure, it gradually reduces the dimensionality to filter cross-modal common features and suppress modality-specific noise;

[0200] Semantic space regularization: An orthogonality constraint loss is imposed on the output vector to ensure that the semantic vectors of different geological units are distinguishable in a low-dimensional space.

[0201] It should be noted that the embodiments of this specification have the following beneficial effects through the above content:

[0202] Deeply optimize cross-modal feature interaction: By constructing a cross-attention mechanism between the time-frequency spectrum and the 3D core model, we achieve dynamic interactive perception of seismic waveform time-frequency features and core microtexture characteristics. This mechanism overcomes the limitations of traditional methods that simply concatenate multimodal features. It establishes implicit association rules between seismic wave propagation characteristics and rock physical properties within the feature space, significantly improving the precision of cross-modal semantic alignment.

[0203] Hierarchical semantic representation enhancement: The hierarchical semantic framework uses a hierarchical feature abstraction mechanism to integrate the macroscopic stratigraphic interface characteristics of seismic data with the microscopic pore structure characteristics of core models at multiple scales. This hierarchical and progressive semantic integration approach preserves the unique feature representations of each modality while constructing a cross-scale geological semantic association network, enhancing the geological interpretability of the unified semantic vector.

[0204] Improved semantic vector space compactness: Fully connected layers are used to perform targeted dimensionality reduction on high-dimensional interactive features, eliminating redundant seismic time-frequency noise and core texture features while preserving key cross-modal semantic information. The resulting compact semantic vector not only reduces subsequent computational complexity but also enhances the separability of different geological units through orthogonalization of the feature space.

[0205] A breakthrough in the interpretability of geological patterns: The construction of the cross-attention weight matrix reveals the physical correlation between seismic wave attenuation and physical parameters such as rock permeability. This interpretable feature interaction mechanism provides a new theoretical path for reservoir parameter inversion, breaking through the limitations of traditional black-box models in oil and gas geology research.

[0206] Furthermore, when generating a spatiotemporally aligned semantic feature matrix based on the contrastive learning framework, the spectrum graph and the logging text can be mutually corrected based on the attention mechanism of the contrastive learning framework to obtain fused features; and the fused features are compressed into a spatiotemporally aligned semantic feature matrix.

[0207] In the examples of this specification, the following specific implementation schemes can be used:

[0208] Step 1: Cross-modal bidirectional correction

[0209] Construction of a dual-modal encoder: A convolutional neural network is designed to extract the time-frequency features of the spectrogram, while a Transformer architecture is used to encode the semantic information of the well logging text.

[0210] Deformable Attention Mechanism Deployment: Deploy an attention module with adjustable receptive field in the feature space to capture the nonlinear mapping relationship between seismic wave propagation delay and logging depth offset;

[0211] Feature mutual correction is performed: Through the cross-attention gating mechanism, the semantics of logging parameters are used to correct the frequency band energy distribution of the spectrogram, and the time-frequency anomaly features are used to enhance the keyword weight of the logging text;

[0212] Residual compensation fusion: The original modal features and the attention-corrected features are weightedly summed to retain the effective information increment during cross-modal interaction.

[0213] Step 2: Semantic Matrix Compression Alignment

[0214] Feature orthogonalization: Gram-Schmidt orthogonalization is performed on the fused features to eliminate redundant information dimensions between the spectrum and text modalities;

[0215] Spatiotemporal constraint compression: uses spatial pyramid pooling technology to compress feature dimensions, while simultaneously imposing temporal continuity loss and depth consistency loss;

[0216] Matrix optimization and calibration: Non-negative matrix factorization is used to extract spatiotemporal shared basis vectors and reconstruct a low-rank semantic matrix with spatiotemporal alignment characteristics.

[0217] Dynamic update mechanism: A sliding window strategy is deployed to achieve incremental updates of matrix parameters to adapt to the temporal changes of real-time data streams during drilling.

[0218] It should be noted that the embodiments of this specification have the following beneficial effects through the above content:

[0219] A breakthrough in cross-modal dynamic correction capabilities: A cross-attention mechanism within a contrastive learning framework enables bidirectional feature correction between spectrograms and well logging text. This mechanism enables the temporal evolution of seismic waveforms to correct for depth interpretation biases in well logging text. Simultaneously, the spatial constraints of logging parameters feed back into the physical meaning of the spectrogram, forming a self-optimizing feature interaction system with spatiotemporal consistency.

[0220] Improved spatiotemporal coupling accuracy: During the feature fusion process, a dynamic allocation strategy for attention weights automatically captures the nonlinear mapping relationship between seismic signal propagation time and logging depth space. This technology overcomes the limitations of traditional linear interpolation methods and significantly improves the spatiotemporal matching accuracy of key geological elements such as identifying the top and bottom interfaces of oil and gas reservoirs and locating formation pinch-out points.

[0221] An innovative semantic feature matrix compaction approach uses a feature compression algorithm to orthogonally reduce the dimensionality of high-dimensional fusion features, eliminating redundant information between modalities while preserving the correlation between seismic waveform anomaly patterns and logging parameter mutation characteristics. The resulting semantic matrix combines low dimensionality with high information entropy, providing a lightweight and highly discriminative feature base for subsequent dynamic correlation map construction.

[0222] Enhanced confidence in geological interpretation: By comparing the attention correction trajectory formed during the learning process, we can reversely analyze the physical interaction mechanism between seismic attributes and well logging responses. This interpretable correction mechanism provides a transparent decision-making basis for reservoir hydrocarbon content identification, effectively overcoming the "black box" risk of traditional end-to-end models.

[0223] Furthermore, when obtaining the dynamic association map of well logging and seismic waveforms based on the cross-modal unified semantic vector and the semantic feature matrix, the cross-modal unified semantic vector and the semantic feature matrix can be horizontally spliced ​​to generate a fusion feature matrix; well logging nodes and seismic nodes are defined in the fusion feature matrix, and the similarity between the well logging nodes and the seismic nodes is determined; and the dynamic association map of well logging and seismic waveforms is obtained based on the similarity.

[0224] In the examples of this specification, the following specific implementation schemes can be used:

[0225] Step 1: Multimodal feature fusion

[0226] Feature dimension alignment: Zero padding is performed on the cross-modal unified semantic vector to make its channel dimension consistent with the column width of the semantic feature matrix;

[0227] Horizontal splicing execution: concatenate semantic vectors along the columns of the feature matrix, use bilinear interpolation to eliminate scale differences, and generate a joint representation matrix;

[0228] Matrix whitening: Perform ZCA whitening operation on the fusion matrix to eliminate the linear correlation between different modal features.

[0229] Step 2: Heterogeneous node definition

[0230] Logging node division: The row vector of the fusion matrix is ​​cut according to the logging depth sampling interval. Each depth unit corresponds to a logging node, which carries physical properties such as porosity and permeability.

[0231] Seismic node extraction: Split the matrix column vectors along the spatial distribution of seismic gathers, and associate each gather node with seismic attributes such as time-frequency energy and reflection intensity;

[0232] Node attribute enhancement: Aggregate the features of adjacent nodes through the graph attention mechanism to enhance the discriminative expression of nodes in reservoir fluid identification.

[0233] Step 3: Dynamic similarity calculation

[0234] Multi-scale kernel function configuration: Design a hybrid metric function that combines Euclidean distance and cosine similarity to balance local details and global distribution characteristics;

[0235] Adaptive threshold setting: Dynamically adjust the similarity discrimination threshold based on the node feature distribution density to achieve strong correlation focusing in sparse areas and noise suppression in dense areas;

[0236] Incremental update mechanism: Deploy a sliding time window to track the latest logging data and trigger real-time iterative updates of association weights.

[0237] Step 4: Generate association graph

[0238] Topology optimization: Use force-directed layout algorithm to arrange node spatial positions, clustering strongly associated nodes and dispersing weakly associated nodes;

[0239] Dynamic visualization rendering: Integrates WebGL technology to achieve interactive display of graphs, supports pop-up query of node attributes and animation tracing of associated paths;

[0240] Abnormal correlation warning: Set up a correlation strength mutation detection module to trigger a sweet spot warning signal when the similarity between logging and seismic nodes changes suddenly.

[0241] It should be noted that the embodiments of this specification have the following beneficial effects through the above content:

[0242] Enhanced complementarity of multimodal features: By horizontally splicing a cross-modal unified semantic vector with a spatiotemporal aligned semantic matrix, we achieve triple information fusion of seismic waveform time-frequency characteristics, core microstructural characteristics, and well logging physical parameters. This fusion mechanism transcends the limitations of single-modal representation, cross-validating seismic signal propagation patterns with rock physical properties and significantly improving the ability to jointly discriminate reservoir sensitivity characteristics.

[0243] A breakthrough in dynamic correlation topology adaptability: A node connection mechanism based on similarity calculations automatically detects real-time correlations between changes in logging parameters and seismic waveform characteristics. This technology transcends the rigid constraints of traditional fixed-threshold association rules, forming an intelligent correlation network that can be dynamically reconstructed with incremental updates of drilling data, effectively adapting to the dynamic evolution of geological understanding during oil and gas reservoir development.

[0244] Improved accuracy in correlating geological anomalies: By defining heterogeneous logging and seismic nodes in the fused feature space, high-dimensional feature similarity metrics are used to accurately capture hidden correlation patterns between anomalous segments of well logging curves and areas of seismic waveform distortion. This technology significantly improves correlation accuracy for key geological interpretation tasks such as fault identification and gas anomaly detection, addressing the technical pain point of traditional methods' insensitivity to weak signal correlation.

[0245] Upgraded visualization support for engineering decision-making: Dynamic correlation maps visualize node connection strength and topology, directly presenting the spatial matching relationship between seismic reflection interfaces and well logging interpretation conclusions. This interactive graphical representation provides a transparent and intelligent interface for engineering decisions such as real-time well trajectory adjustment and reservoir sweet spot delineation.

[0246] Furthermore, when the dynamic correlation map of well logging and seismic waveforms is obtained based on the similarity, the node pairs corresponding to the similarity greater than the preset threshold are screened, and the node pairs are associated well logging nodes and seismic nodes: the dynamic correlation map is generated based on the node pairs.

[0247] In the examples of this specification, the following specific implementation schemes can be used:

[0248] Step 1: Intelligent threshold screening

[0249] Dynamic threshold initialization: Based on the spatial distribution density of node features, the kernel density estimation algorithm is used to automatically calculate the initial similarity threshold;

[0250] Multi-scale threshold optimization: Divide regional subsets into different geological units and calculate local optimal thresholds to adapt to reservoir heterogeneity characteristics;

[0251] Noise suppression processing: morphological opening operations are used to eliminate isolated low-similarity pseudo-correlations and retain continuous high-confidence node connections;

[0252] Focus on key areas: Implement threshold downsampling in sensitive areas such as sweet spots and fault zones to enhance the ability to capture weakly correlated signals.

[0253] Step 2: Heterogeneous Node Processing

[0254] Node attribute inheritance: inherit physical properties such as porosity and oil saturation for the selected logging nodes, and associate time-frequency energy, coherence volume and other attributes for the seismic nodes;

[0255] Composite node construction: Create virtual composite nodes for nodes with multiple associations to represent complex geological phenomena at the intersection of wells and seismic data;

[0256] Node index optimization: Establish a GeoHash-based spatial index and an inverted index based on feature vectors to support multi-dimensional fast retrieval.

[0257] Step 3: Dynamically construct the graph

[0258] Graph database initialization: Use the Neo4j graph database to build a node-relationship storage structure and define multi-dimensional edge attributes such as "association strength" and "spatial distance";

[0259] Incremental update mechanism: Deploy Kafka message queues to capture newly added node pairs in real time, triggering partial reconstruction of the graph topology.

[0260] Dynamic weight allocation: Set the associated weight attenuation coefficient according to the drilling timeliness to ensure that the map reflects the latest geological knowledge status;

[0261] Topology optimization execution: Spectral clustering algorithm is used to implement subgraph partitioning of high-density association areas to improve the interpretability of the graph structure.

[0262] Step 4: Visualization and Interaction

[0263] Multi-level rendering: Integrated LOD technology enables multi-scale visualization of the entire work area overview and local details, supporting mouse wheel zooming and dragging for roaming;

[0264] Dynamic focus tracking: Deploys a sight focus rendering optimization algorithm to highlight and enhance the display of related nodes around the current drilling trajectory;

[0265] Intelligent drilling function: supports right-clicking any node to trigger the associated path backtracking, visually displaying the geological basis chain of well-seismic association;

[0266] Abnormal correlation warning: When an unclosed correlation loop or a sudden change in strength edge is detected, a flashing alarm and log recording are automatically triggered.

[0267] Step 5: Graph maintenance and update

[0268] Version management: Uses a Git-like mechanism to record the graph evolution process, supporting comparative analysis and backtracking of any historical version;

[0269] Drift detection module: deploys a concept drift detection algorithm that triggers threshold recalibration when the node feature distribution changes significantly;

[0270] Self-learning optimization: Automatically adjust topology layout parameters through reinforcement learning mechanisms to continuously improve the cognitive friendliness of graph visualization.

[0271] It should be noted that the embodiments of this specification have the following beneficial effects through the above content:

[0272] Innovation in dynamic adaptive association mechanisms: This technology dynamically optimizes the association between well logging and seismic nodes through an intelligent screening mechanism with preset thresholds. This technology breaks through the rigid constraints of traditional fixed association rules and automatically adjusts the association strength threshold based on the actual geological scenario, ensuring that the map topology evolves adaptively with data quality and geological complexity.

[0273] Improved Correlation Noise Suppression: A node pair selection strategy based on similarity thresholds effectively filters out spurious correlation signals caused by data acquisition noise or cross-modal semantic bias. This mechanism significantly enhances the geological credibility of node connections in correlation maps, preventing random perturbations in seismic waveforms from being misinterpreted as valid reservoir responses.

[0274] Enhanced Focus on Key Geological Targets: The threshold screening process essentially creates a targeted enhancement of strongly correlated features, automatically focusing the map on areas of high correlation between logging parameter abrupt changes and seismic waveform anomalies. This focusing effect significantly improves the sensitivity of identifying key geological targets such as subtle reservoir boundaries and small faults.

[0275] A breakthrough in interpretability for engineering implementation: A preset threshold mechanism provides a transparent basis for decision-making in generating correlation maps. Geological engineers can intuitively control the balance between sparsity and reliability of the correlation network by adjusting the threshold. This explainable interaction significantly enhances technicians' trust in and acceptance of AI-generated results.

[0276] Figure 2 A structural schematic diagram of a data cross-modal association device in an oil and gas field database provided for one or more embodiments of this specification includes: a spectrum graph generation unit 201, a core digital model determination unit 202, a semantic vector determination unit 203, a feature matrix generation unit 204 and an association graph determination unit 205.

[0277] The spectrum generating unit 201 receives the original seismic waveform time series input in advance and generates a time spectrum through Fourier transform;

[0278] The core digital model determination unit 202 receives a pre-input core scan image set, extracts local binary pattern texture features from the core scan image set, and obtains a spatially continuous three-dimensional core digital model based on the local binary pattern texture features;

[0279] The semantic vector determining unit 203 injects the time-frequency spectrum and the three-dimensional core digital model into a pre-generated hierarchical semantic framework to obtain a cross-modal unified semantic vector;

[0280] A feature matrix generating unit 204 constructs a comparative learning framework between the spectrogram and the well logging text, and generates a spatiotemporally aligned semantic feature matrix based on the comparative learning framework;

[0281] The correlation map determining unit 205 obtains a dynamic correlation map between well logging and seismic waveforms based on the cross-modal unified semantic vector and the semantic feature matrix.

[0282] Figure 3 A schematic diagram of a cross-modal data association device in an oil and gas field database provided for one or more embodiments of this specification includes:

[0283] at least one processor; and,

[0284] a memory communicatively connected to the at least one processor; wherein,

[0285] The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:

[0286] Receive the pre-input original seismic waveform time series and generate a time-frequency spectrum through Fourier transform;

[0287] receiving a pre-input core scan image set, extracting local binary pattern texture features from the core scan image set, and obtaining a spatially continuous three-dimensional core digital model based on the local binary pattern texture features;

[0288] Injecting the time-frequency spectrum and the three-dimensional core digital model into a pre-generated hierarchical semantic framework to obtain a cross-modal unified semantic vector;

[0289] Constructing a comparative learning framework between the spectrogram and the well logging text, and generating a spatiotemporally aligned semantic feature matrix based on the comparative learning framework;

[0290] Based on the cross-modal unified semantic vector and the semantic feature matrix, a dynamic correlation map of well logging and seismic waveforms is obtained.

[0291] One or more embodiments of this specification provide a non-volatile computer storage medium storing computer-executable instructions. When executed by a computer, the computer-executable instructions can achieve:

[0292] Receive the pre-input original seismic waveform time series and generate a time-frequency spectrum through Fourier transform;

[0293] receiving a pre-input core scan image set, extracting local binary pattern texture features from the core scan image set, and obtaining a spatially continuous three-dimensional core digital model based on the local binary pattern texture features;

[0294] Injecting the time-frequency spectrum and the three-dimensional core digital model into a pre-generated hierarchical semantic framework to obtain a cross-modal unified semantic vector;

[0295] Constructing a comparative learning framework between the spectrogram and the well logging text, and generating a spatiotemporally aligned semantic feature matrix based on the comparative learning framework;

[0296] Based on the cross-modal unified semantic vector and the semantic feature matrix, a dynamic correlation map of well logging and seismic waveforms is obtained.

[0297] The various embodiments in this specification are described in a progressive manner. Similar portions between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from the other embodiments. In particular, the device, apparatus, and non-volatile computer storage medium embodiments are generally similar to the method embodiments, so their descriptions are relatively simplified. For relevant details, refer to the descriptions of the method embodiments.

[0298] The various embodiments in this specification are described in a progressive manner. Similar parts between the various embodiments can be referred to in conjunction with each other. Each embodiment focuses on the differences from other embodiments. In particular, the device embodiments are generally similar to the method embodiments, so the description is relatively simple. For relevant parts, refer to the description of the method embodiments.

[0299] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0300] In the embodiments provided in this application, it should be understood that the disclosed devices / network equipment and methods can be implemented in other ways. For example, the device / network equipment embodiments described above are merely illustrative. For example, the division of the modules or units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0301] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0302] In addition, the functional units in the various embodiments of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above units may be implemented in the form of hardware or software.

[0303] If the integrated module / unit is implemented as a software functional unit and sold or used as a standalone product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application can implement all or part of the process steps in the above-mentioned method embodiments by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal, and software distribution medium. It should be noted that the content of the computer-readable medium can be appropriately increased or decreased based on the requirements of legislation and patent practice in a jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media does not include electric carrier signals and telecommunication signals.

[0304] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the scope of protection of the present application.

Claims

1. A method for cross-modal data association in an oil and gas field database, characterized in that: include: Receive the pre-input original seismic waveform time series and generate a time-frequency spectrum through Fourier transform; receiving a pre-input core scanning image set, extracting local binary pattern texture features from the core scanning image set, and obtaining a spatially continuous three-dimensional core digital model based on the local binary pattern texture features; Injecting the time-frequency spectrum and the three-dimensional core digital model into a pre-generated hierarchical semantic framework to obtain a cross-modal unified semantic vector; Constructing a comparative learning framework between the time-frequency spectrogram and the well logging text, and generating a spatiotemporally aligned semantic feature matrix based on the comparative learning framework; Based on the cross-modal unified semantic vector and the semantic feature matrix, a dynamic correlation map of well logging and seismic waveforms is obtained; The step of injecting the time-frequency spectrum diagram and the three-dimensional core digital model into a pre-generated hierarchical semantic framework to obtain a cross-modal unified semantic vector includes: defining a template as a hierarchical semantic framework, wherein the hierarchical semantic framework includes a cross-attention mechanism between the time-spectrogram and the three-dimensional core digital model; Obtaining interactive features based on a cross-attention mechanism between the time-spectrogram and the three-dimensional core digital model; The interaction features are input into the fully connected layer and reduced to a cross-modal unified semantic vector.

2. The method according to claim 1, characterized in that The method of receiving a pre-input original seismic waveform time series and generating a time-frequency spectrum through Fourier transform includes: The original seismic waveform time series is received, and a short-time Fourier transform algorithm is used to set a specified window to generate a time-frequency spectrum matrix including time, frequency and amplitude; Obtain the time-to-depth conversion model based on well logging acoustic time difference data: The time-spectrogram matrix is ​​input into a time-depth conversion model to generate a time-spectrogram of depth and time.

3. The method according to claim 1, characterized in that The method of receiving a pre-input core scan image, extracting local binary pattern texture features from the core scan image, and obtaining a spatially continuous three-dimensional core digital model based on the local binary pattern texture features includes: receiving a core scanning image set; Calculate the LBP value for each pixel point of each core scan image and generate an LBP coding map; Counting the LBP histogram features of the LBP coding image to characterize the local texture distribution of the core; Based on the LBP histogram similarity of adjacent slices, the spatial correspondence between slices is established; The two-dimensional slices are stacked into a three-dimensional voxel grid based on the spatial correspondence to obtain the three-dimensional core digital model.

4. The method according to claim 1, wherein The generating of a spatiotemporal aligned semantic feature matrix based on the contrastive learning framework includes: Based on the attention mechanism of the contrastive learning framework, the spectrum graph and the well logging text are mutually corrected to obtain fusion features; The fused features are compressed into a spatiotemporally aligned semantic feature matrix.

5. The method according to claim 1, wherein The step of obtaining a dynamic correlation map between well logging and seismic waveforms based on the cross-modal unified semantic vector and the semantic feature matrix includes: horizontally concatenating the cross-modal unified semantic vector and the semantic feature matrix to generate a fusion feature matrix; The well logging nodes and seismic nodes are defined in the fusion feature matrix, and the similarity between the well logging nodes and the seismic nodes is determined. The row vectors of the fusion feature matrix are cut according to the sampling interval of the well logging depth. Each depth unit corresponds to a well logging node. The column vectors of the matrix are cut along the spatial distribution of the seismic gather. Each gather node is associated with the seismic attributes of time-frequency energy and reflection intensity. A dynamic correlation map of well logging and seismic waveforms is obtained based on the similarity.

6. The method according to claim 5, characterized in that The method of obtaining a dynamic correlation map of well logging and seismic waveforms based on the similarity includes: Filter the node pairs corresponding to the similarity greater than a preset threshold, where the node pairs are associated well logging nodes and seismic nodes: The dynamic association graph is generated based on the node pairs.

7. A device for cross-modal data association in an oil and gas field database, characterized in that: include: A spectrum generating unit receives a pre-input original seismic waveform time series and generates a time spectrum through Fourier transform; a core digital model determination unit, receiving a pre-input core scan image set, extracting local binary pattern texture features from the core scan image set, and obtaining a spatially continuous three-dimensional core digital model based on the local binary pattern texture features; a semantic vector determination unit, which injects the time-frequency spectrum and the three-dimensional core digital model into a pre-generated hierarchical semantic framework to obtain a cross-modal unified semantic vector; A feature matrix generation unit constructs a comparative learning framework between the time-frequency spectrum graph and the well logging text, and generates a spatiotemporally aligned semantic feature matrix based on the comparative learning framework; A correlation map determining unit, which obtains a dynamic correlation map between well logging and seismic waveforms based on the cross-modal unified semantic vector and the semantic feature matrix; The step of injecting the time-frequency spectrum diagram and the three-dimensional core digital model into a pre-generated hierarchical semantic framework to obtain a cross-modal unified semantic vector includes: defining a template as a hierarchical semantic framework, wherein the hierarchical semantic framework includes a cross-attention mechanism between the time-spectrogram and the three-dimensional core digital model; Obtaining interactive features based on a cross-attention mechanism between the time-spectrogram and the three-dimensional core digital model; The interaction features are input into the fully connected layer and reduced to a cross-modal unified semantic vector.

8. A device for cross-modal data association in an oil and gas field database, characterized in that: include: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to: Receive the pre-input original seismic waveform time series and generate a time-frequency spectrum through Fourier transform; receiving a pre-input core scanning image set, extracting local binary pattern texture features from the core scanning image set, and obtaining a spatially continuous three-dimensional core digital model based on the local binary pattern texture features; Injecting the time-frequency spectrum and the three-dimensional core digital model into a pre-generated hierarchical semantic framework to obtain a cross-modal unified semantic vector; Constructing a comparative learning framework between the time-frequency spectrogram and the well logging text, and generating a spatiotemporally aligned semantic feature matrix based on the comparative learning framework; Based on the cross-modal unified semantic vector and the semantic feature matrix, a dynamic correlation map of well logging and seismic waveforms is obtained; The step of injecting the time-frequency spectrum diagram and the three-dimensional core digital model into a pre-generated hierarchical semantic framework to obtain a cross-modal unified semantic vector includes: defining a template as a hierarchical semantic framework, wherein the hierarchical semantic framework includes a cross-attention mechanism between the time-spectrogram and the three-dimensional core digital model; Obtaining interactive features based on a cross-attention mechanism between the time-spectrogram and the three-dimensional core digital model; The interaction features are input into the fully connected layer and reduced to a cross-modal unified semantic vector.

9. A non-volatile computer storage medium, characterized in that The computer-executable instructions are stored, and when the computer-executable instructions are executed by a computer, they can achieve: Receive the pre-input original seismic waveform time series and generate a time-frequency spectrum through Fourier transform; receiving a pre-input core scanning image set, extracting local binary pattern texture features from the core scanning image set, and obtaining a spatially continuous three-dimensional core digital model based on the local binary pattern texture features; Injecting the time-frequency spectrum and the three-dimensional core digital model into a pre-generated hierarchical semantic framework to obtain a cross-modal unified semantic vector; Constructing a comparative learning framework between the time-frequency spectrogram and the well logging text, and generating a spatiotemporally aligned semantic feature matrix based on the comparative learning framework; Based on the cross-modal unified semantic vector and the semantic feature matrix, a dynamic correlation map of well logging and seismic waveforms is obtained; The step of injecting the time-frequency spectrum diagram and the three-dimensional core digital model into a pre-generated hierarchical semantic framework to obtain a cross-modal unified semantic vector includes: defining a template as a hierarchical semantic framework, wherein the hierarchical semantic framework includes a cross-attention mechanism between the time-spectrogram and the three-dimensional core digital model; Obtaining interactive features based on a cross-attention mechanism between the time-spectrogram and the three-dimensional core digital model; The interaction features are input into the fully connected layer and reduced to a cross-modal unified semantic vector.

Images