Multi-scale lithofacies identification and sedimentary microfacies reconstruction method and system

Abstract

The invention belongs to the technical field of geological research, and discloses a multi-scale lithofacies identification and sedimentary microfacies reconstruction method and system, and the method comprises the steps: data collection and preprocessing: collecting core data, slice data, logging data and seismic data of a research region; multi-scale lithofacies feature extraction, wherein the multi-scale lithofacies feature extraction comprises micro-scale feature extraction, medium-scale feature extraction and macro-scale feature extraction; constructing and training a multi-scale lithofacies recognition model, constructing the multi-scale lithofacies recognition model by adopting a deep learning algorithm, and taking the extracted microscopic, mesoscopic and macroscopic lithofacies features as input data; multi-scale lithofacies identification: extracting multi-scale lithofacies features from the preprocessed multi-source data; and sedimentary microfacies reconstruction: establishing a lithofacies-sedimentary microfacies conversion relation in combination with a sedimentology principle and a geological knowledge base. According to the method, multi-source data such as rock cores, slices, logging and earthquakes are integrated, lithofacies features are extracted from microcosmic, mesoscopic and macroscopic multiple scales, the features and changes of lithofacies can be comprehensively and accurately reflected, and compared with a traditional single-scale method, the accuracy and reliability of lithofacies recognition are improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of geological research, but not limited to, and particularly relates to a multi-scale lithofacies identification and sedimentary microfacies reconstruction method and system. BACKGROUND

[0002] In geological research, lithofacies identification and sedimentary microfacies reconstruction are key links to understand the depositional environment in geological history and predict the distribution of reservoirs. Traditional lithofacies identification methods mainly rely on manual means such as core observation and thin section analysis. These methods are not only low in efficiency and high in cost, but also greatly affected by human factors, making it difficult to achieve large-scale and rapid lithofacies identification. With the development of geophysical technology, seismic data and logging data are widely used in lithofacies identification. However, existing technologies often only use single-scale data information, making it difficult to fully reflect the multi-scale characteristics of lithofacies. In terms of sedimentary microfacies reconstruction, existing methods are usually based on limited geological data and empirical models, lacking effective fusion and comprehensive analysis of multi-source data, resulting in low accuracy of reconstruction results and inability to accurately depict the complex spatial distribution of sedimentary microfacies. Therefore, there is an urgent need for a method and system that can comprehensively utilize multi-scale data to achieve high-precision lithofacies identification and sedimentary microfacies reconstruction.

[0003] Xu et al. in the study "Joint use of multi-seismic information for lithofacies prediction via supervised convolutional neural networks" in 2022 input angle-dependent pre-stacked traces, impedance inversion bodies and various seismic attributes into a convolutional neural network to predict the lithofacies of clastic reservoirs in thin coal seams. The results show that the fusion model significantly improves the thin layer identification accuracy and overall classification performance compared to single post-stack or inversion data. Since this scheme has tried to integrate multi-source seismic information within the same scale, it can be considered as the closest prior art to the multi-scale lithofacies identification concept in the present invention.

[0004] However, the above method still faces key bottlenecks: first, it only relies on seismic scale data, lacks cross-scale fusion with core CT, thin section images and logging curves, and cannot capture micro features such as particle level morphology and pore structure, limiting the fine depiction of reservoir heterogeneity; second, the output stops at lithofacies classification, without establishing a lithofacies-sedimentary microfacies conversion and interpolation reconstruction mechanism, making it difficult to generate continuous visual microfacies distribution bodies, and thin layer coal or sheet sand may still be missed or boundary drift. Such methodological limits have also been repeatedly pointed out in other deep learning logging log studies. SUMMARY

[0005] The present application provides a multi-scale lithofacies identification and sedimentary microfacies reconstruction method in view of the problems of the prior art.

[0006] The present application is implemented as a multi-scale lithofacies identification and sedimentary microfacies reconstruction method, which comprises: S1: data acquisition and preprocessing, collecting core data, thin section data, logging data and seismic data of the study area; performing data cleaning, missing value filling and outlier processing on the collected data, digitizing the core and thin section data, performing depth correction and standardization processing on the logging data, and performing denoising and deconvolution processing on the seismic data; S2: multi-scale lithofacies feature extraction, including microscale feature extraction, mesoscale feature extraction and macroscale feature extraction; S3: multi-scale lithofacies identification model construction and training, a multi-scale lithofacies identification model is constructed using a deep learning algorithm, the extracted microscale, mesoscale and macroscale lithofacies features are used as input data, and known lithofacies categories are used as output labels, and the model is trained and optimized through a large number of sample data; S4: multi-scale lithofacies identification, after the multi-source data after preprocessing are extracted for multi-scale lithofacies features, the trained multi-scale lithofacies identification model is inputted, and the lithofacies identification result of the study area is obtained; S5: sedimentary microfacies reconstruction, according to the identified lithofacies result, a lithofacies-sedimentary microfacies conversion relationship is established in combination with sedimentology principles and a geological knowledge base; spatial interpolation is performed on the lithofacies identification result by using a spatial interpolation algorithm to obtain continuous lithofacies distribution data, which is then converted into sedimentary microfacies data, so that the reconstruction of the sedimentary microfacies of the study area is realized, and the reconstruction result is visually displayed. Further, in S1, the core data includes core photos, lithology descriptions, porosity and permeability parameters; the thin section data includes mineral composition and particle structure information; the logging data includes natural gamma, resistivity and acoustic travel time logging curve data; and the seismic data is a three-dimensional seismic data volume.

[0007] Further, the microscale feature extraction specifically comprises: based on the thin section data, using image analysis technology to extract microscale lithofacies features such as mineral composition proportion, particle size distribution and particle shape parameters; The mesoscale feature extraction specifically comprises: combining the core data and the logging data to extract mesoscale lithofacies features such as lithology combination, sedimentary structure and logging facies features; The macroscale feature extraction specifically comprises: extracting seismic attribute features such as amplitude, frequency, phase and waveform from the seismic data as macroscale lithofacies features.

[0008] Further, in the micro-scale feature extraction, the particle morphology parameters include roundness and sphericity, different mineral phases and particle structures are identified by image segmentation and classification, and then the related feature parameters are calculated. Further, in the meso-scale feature extraction, the meso-facies features are extracted according to the core lithology description and the shape, amplitude and other features of the logging curve by establishing a logging facies-lithology conversion relationship. Further, in the macro-scale feature extraction, three-dimensional seismic data are calculated by using seismic attribute extraction technology, and different seismic attribute bodies are obtained to analyze the spatial distribution characteristics. Another object of the present application is to provide a multi-scale facies identification and sedimentary microfacies reconstruction system based on the multi-scale facies identification and sedimentary microfacies reconstruction method. The data acquisition and preprocessing module is used for acquiring multi-source data and preprocessing the data, including data cleaning, digitization, correction and standardization operations. The multi-scale feature extraction module is connected with the data acquisition and preprocessing module, and the facies features are extracted from the micro, meso and macro scales, and the functions of image analysis, core-logging data correlation analysis and seismic attribute extraction are integrated. The model construction and training module is connected with the multi-scale feature extraction module, and a multi-scale facies identification model is constructed by using a deep learning algorithm, and the model is trained and optimized. The facies identification module is connected with the model construction and training module, and the extracted multi-scale facies features are input into the trained model to realize facies identification. The sedimentary microfacies reconstruction module is connected with the facies identification module, and according to the facies identification result, the sedimentary microfacies is reconstructed and visually displayed by combining the principles of sedimentology and spatial interpolation algorithm. Another object of the present application is to provide a computer device, which comprises a memory and a processor, the memory stores a computer program, and the computer program is executed by the processor to make the processor execute the steps of the multi-scale facies identification and sedimentary microfacies reconstruction method.

[0009] Another object of the present application is to provide a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to make the processor execute the steps of the multi-scale facies identification and sedimentary microfacies reconstruction method.

[0010] Another object of the present application is to provide an information data processing terminal for realizing the multi-scale facies identification and sedimentary microfacies reconstruction system.

[0011] In combination with the above technical solutions and the technical problems solved, the technical solutions to be protected by the present application have the following advantages and positive effects: 1. The present application can comprehensively and accurately reflect the characteristics and changes of lithofacies by integrating multi-source data such as cores, thin sections, logging and seismic, and extracting lithofacies characteristics from micro, meso and macro scales, thereby improving the accuracy and reliability of lithofacies identification compared with traditional single-scale methods; 2. The deep learning algorithm is used to construct a multi-scale lithofacies identification model, which can automatically learn the complex relationship between multi-scale lithofacies characteristics and lithofacies categories, reduce the interference of human factors, and improve the efficiency and accuracy of lithofacies identification; 3. In the process of sedimentary microfacies reconstruction, the spatial interpolation algorithm is used in combination with sedimentology principles and geological knowledge base to accurately reconstruct the spatial distribution of sedimentary microfacies, thereby providing more accurate basic data for geological research and resource exploration; 4. The system provided by the present application realizes the automation and intelligentization of multi-scale lithofacies identification and sedimentary microfacies reconstruction, reduces labor costs, improves work efficiency, and has a wide application prospect. BRIEF DESCRIPTION OF DRAWINGS

[0012] Figure 1 is a multi-scale lithofacies identification and sedimentary microfacies reconstruction method flowchart provided by an embodiment of the present application.

[0013] Figure 2 is a multi-scale lithofacies feature extraction schematic diagram provided by an embodiment of the present application.

[0014] Figure 3 is a multi-scale lithofacies identification and sedimentary microfacies reconstruction system structure diagram provided by an embodiment of the present application.

[0015] In the figure: 1, data acquisition and preprocessing module; 2, multi-scale feature extraction module; 3, model construction and training module; 4, lithofacies identification module; 5, sedimentary microfacies reconstruction module. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in combination with embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0017] As shown in Figure 1 , an embodiment of the present application provides a multi-scale lithofacies identification and sedimentary microfacies reconstruction method, which comprises: S1: data collection and preprocessing, collecting core data, thin section data, logging data and seismic data of the study area; data cleaning, missing value filling and outlier processing are performed on the collected data; core and thin section data are digitized; depth correction and standardization processing are performed on logging data; denoising and deconvolution processing are performed on seismic data; S2: multi-scale lithofacies feature extraction, including micro-scale feature extraction, meso-scale feature extraction and macro-scale feature extraction; S3: multi-scale lithofacies recognition model construction and training, a multi-scale lithofacies recognition model is constructed using a deep learning algorithm, the extracted micro, meso and macro lithofacies features are used as input data, and the known lithofacies categories are used as output labels; the model is trained and optimized through a large number of sample data; S4: multi-scale lithofacies recognition, after the multi-source data after preprocessing are extracted for multi-scale lithofacies features, the trained multi-scale lithofacies recognition model is inputted, and the lithofacies recognition result of the study area is obtained; S5: sedimentary microfacies reconstruction, according to the recognized lithofacies result, a lithofacies-sedimentary microfacies conversion relationship is established combining sedimentology principles and a geological knowledge base; spatial interpolation is performed on the lithofacies recognition result using a spatial interpolation algorithm to obtain continuous lithofacies distribution data, which is then converted into sedimentary microfacies data to realize reconstruction of the sedimentary microfacies of the study area, and the reconstructed result is visualized and displayed. In the S1, the core data includes core photos, lithology description, porosity and permeability parameters; the thin section data includes mineral composition and particle structure information; the logging data includes natural gamma ray, resistivity and acoustic travel time logging curve data; and the seismic data is a three-dimensional seismic data volume.

[0018] As shown in Figure 2 The micro-scale feature extraction specifically includes: based on the thin section data, image analysis technology is used to extract micro-lithofacies features such as mineral composition ratio, particle size distribution and particle shape parameters; The meso-scale feature extraction specifically includes: combining the core data and the logging data, meso-lithofacies features such as lithology combination, sedimentary structure and logging facies features are extracted; The macro-scale feature extraction specifically includes: extracting seismic attribute features such as amplitude, frequency, phase and waveform from the seismic data as macro-lithofacies features.

[0019] In the micro-scale feature extraction, the particle shape parameters include roundness and sphericity, different mineral phases and particle structures are identified by thin section image segmentation and classification, and then the related feature parameters are calculated. In the meso-scale feature extraction, a logging facies-lithology conversion relationship is established, and meso-lithofacies features are extracted according to the shape, amplitude and other features of the core lithology description and logging curves. In the macro-scale feature extraction, a seismic attribute extraction technique is used to calculate the three-dimensional seismic data body to obtain different seismic attribute bodies to analyze the spatial distribution characteristics. Particular fields of application or related products of the invention.

[0020] As Figure 3 shown, the embodiment of the present application provides a multi-scale lithofacies identification and sedimentary microfacies reconstruction system based on the multi-scale lithofacies identification and sedimentary microfacies reconstruction method, which specifically comprises: A data acquisition and preprocessing module 1 is used to acquire multi-source data and pre-process the data, including data cleaning, digitization, correction and standardization operations; A multi-scale feature extraction module 2 is connected with the data acquisition and preprocessing module 1, and extracts lithofacies features from micro, meso and macro scales, integrating image analysis, core-logging data correlation analysis and seismic attribute extraction functions; A model construction and training module 3 is connected with the multi-scale feature extraction module 2, and uses a deep learning algorithm to construct a multi-scale lithofacies identification model, and trains and optimizes the model; A lithofacies identification module 4 is connected with the model construction and training module 3, and inputs the extracted multi-scale lithofacies features into the trained model to realize lithofacies identification; A sedimentary microfacies reconstruction module 5 is connected with the lithofacies identification module 4, and according to the lithofacies identification result, combines sedimentology principles and spatial interpolation algorithms to realize reconstruction and visual display of sedimentary microfacies. The present experiment aims to verify the lithofacies identification accuracy of TriFusion-MSFE multi-scale feature extraction algorithm and GeoVisionNet model in hidden sand bodies, and the spatial consistency of SedRecoStudio on sedimentary microfacies reconstruction results. Three exploratory wells and one development well with a size of 5 km x 6 km on the eastern margin of the Ordos Basin are selected as the research objects, and the strata span focuses on the Lower Shihezi Formation, including delta front and sheet sand deposition.

[0021] Hyperspectral-CT joint scanning is performed on the cores of the four wells to obtain a 0.2 mm resolution spectral cube and a 0.1 mm pixel CT body; 240 thin sections are digitized by a high-resolution scanner, and the logging curves include five natural gamma, resistivity, acoustic time difference, density, and caliper; the three-dimensional seismic body covers an area of 36 km², with a sampling interval of 2 ms and an azimuth of 60°-240°. All data are subjected to missing value interpolation, depth synchronization and amplitude equalization, and finally written into a unified HDF5 database.

[0022] GeoVisionNet is trained with 70% data (random wells, random traces). TriFusion-MSFE has 12 layers of residual convolution clusters, 4 layers of bidirectional GRU, and 8 layers of 3D-DenseBlock in three pathways, respectively. The fusion layer adopts 8 heads of attention. The training batch is 5000, and the initial learning rate is 1e-4 with cosine annealing decay. The remaining 30% data is reserved as a blind test set, and any inter-well data leakage is disabled.

[0023] The rock facies recognition accuracy is mainly measured by pixel-by-pixel F1-score, single-well recognition accuracy, and layer consistency coefficient. The sedimentary microfacies reconstruction accuracy is measured by root mean square error (RMSE) of microfacies profile to real geological section and plane superimposed area ratio. The improved rate is compared with the baseline of traditional well logging clustering-seismic facies analysis method.

[0024] The experimental hardware platform includes 4 × A100-80 GB GPU, Intel Xeon 6348 dual-channel CPU, and 1 TB DDR4 RAM. The training duration, single-well inference delay, and edge device (Jetson Orin NX) inference frame rate are recorded to evaluate the feasibility and resource consumption of the algorithm in cloud and on-site deployment.

[0025] If the rock facies F1-score of the blind test set is ≥ 0.87 and the microfacies reconstruction RMSE is ≤ 40 m, it is considered to pass. At the same time, the GPU inference delay is required to be < 120 ms, and the edge inference frame rate is required to be ≥ 2 Hz. In view of the possible inter-well attribute drift, a transfer learning patch is preset. For the artifacts caused by high density of minerals in CT scans, an adaptive denoising module is introduced. After completion, the model weight, comparison report, and source data meta-information are output, and the next stage of scene demonstration is entered. It should be noted that the embodiments of the present application can be realized by hardware, software or a combination of software and hardware. The hardware part can be realized by using special logic; the software part can be stored in the memory and executed by the appropriate instruction execution system, such as microprocessor or specially designed hardware. Those skilled in the art can understand that the above devices and methods can be realized by computer executable instructions and / or included in processor control code, such as carrier medium, such as magnetic disk, CD or DVD-ROM, programmable memory, such as read-only memory (firmware), or data carrier, such as optical or electronic signal carrier. The devices and modules of the present application can be realized by hardware circuit, such as very large scale integrated circuit or gate array, semiconductor, such as logic chip, transistor, etc., or programmable hardware device, such as field programmable gate array, programmable logic device, etc., or by software executed by various types of processors, or by a combination of the above hardware circuit and software, such as firmware.

[0026] The above merely provides the specific implementation of the present application, but the protection scope of the present application is not limited thereto, any modification, equivalent replacement and improvement within the technical range disclosed by the present application and within the spirit and principle of the present application should be covered within the protection scope of the present application.

Claims

1. A method for multi-scale lithofacies identification and sedimentary microfacies reconstruction, characterized in that, Execute in sequence: a. Acquire and preprocess multi-source data from core, thin section, well logging, and seismic sources; b. The TriFusion-MSFE multi-scale feature extraction algorithm was used to obtain micro, meso, and macro lithofacies features respectively; c. Input the features into the GeoVisionNet deep learning model and output the corresponding lithofacies category; d. Establish the relationship between lithofacies and sedimentary microfacies transformation, and use spatial interpolation algorithms to generate continuous lithofacies distribution bodies; e. Based on the distribution, the sedimentary microfacies reconstruction results are obtained and displayed in three dimensions.

2. A multi-scale lithofacies identification and sedimentary microfacies reconstruction system, characterized in that, include: The system includes modules for data acquisition and preprocessing, microscopic feature extraction, mesoscopic feature extraction, macroscopic feature extraction, model building and training, lithofacies identification, sedimentary microfacies reconstruction, and rendering and display. The model building and training module configures the GeoVisionNet deep learning model; Each module works in sequence to complete lithofacies identification and sedimentary microfacies reconstruction according to the method described in claim 1.

3. A computer-readable storage medium having instructions stored thereon, which, when executed by a processor, implement the multi-scale lithofacies identification and sedimentary microfacies reconstruction method as described in claim 1.

4. The method according to claim 1, characterized in that, The multi-source data mentioned in step a includes: core photographs, lithological descriptions, porosity data, permeability data, mineral composition images, grain structure images, natural gamma logging curves, resistivity logging curves, sonic transit time logging curves, and three-dimensional seismic data volumes. The above data are subjected to digitization, depth correction, standardization, denoising, and deconvolution processing, respectively.

5. The method according to claim 1, characterized in that, In step b, the TriFusion-MSFE multi-scale feature extraction algorithm extracts feature vectors in parallel through micro-level convolutional clusters, meso-level sequence encoders, and macro-level three-dimensional convolutional networks, and outputs joint feature representations using a multi-head self-attention mechanism in the fusion layer.

6. The method according to claim 1, characterized in that, The GeoVisionNet deep learning model in step c adopts an encoder-decoder structure. The encoder end is equipped with a micro residual convolutional path, a meso-gated recurrent path, and a macro-dense three-dimensional convolutional path. The decoder end fuses features through cross-scale jump connections and outputs a pixel-based lithofacies probability map.

7. The system according to claim 2, characterized in that, The micro-feature extraction module runs the TriFusion-MSFE multi-scale feature extraction algorithm of claim 5, and the model building and training module calls the GeoVisionNet deep learning model of claim 6.

8. The system according to claim 2, characterized in that, The data acquisition and preprocessing module includes: a high-resolution core imaging unit, a thin section digitization unit, a well logging synchronization unit, a seismic data preprocessing unit, a distributed cache node, and a graphics processing accelerator, to achieve real-time aggregation, format standardization, and high-speed transmission of multi-source data.

9. The method according to claim 1, characterized in that, In step d, the SedGNN interpolation algorithm based on graph neural network is used to map the lithofacies identification results to an irregular grid, construct node feature vectors and edge weight matrices, predict sedimentary microfacies properties through dynamic weight message passing, and output continuous sedimentary microfacies bodies.

10. The computer-readable storage medium according to claim 3, characterized in that, The instructions further enable the processor to invoke the SedRecoStudio rendering engine to perform hierarchical volume rendering, interactive profile display, and time-series playback on the sedimentary microfacies data obtained in claim 1.

Images