Reservoir sedimentary microfacies depicting method based on artificial intelligence image recognition

Abstract

The invention provides a reservoir sedimentary microfacies depicting method based on artificial intelligence image recognition, and belongs to the technical field of oil and gas reservoir evaluation, and the method comprises the following steps: collecting and sorting different types of sedimentary microfacies mode patterns, and constructing a sedimentary microfacies mode pattern library; carrying out image preprocessing operation; after preprocessing, key feature parameters are extracted from the image; training the features by using a deep learning algorithm to obtain an identification rule and a decision rule, and constructing a sedimentary microfacies diagram identification model; evaluating and verifying the trained model, and adjusting and optimizing the model according to feedback; and inputting limited constraint data into a sedimentary microfacies diagram recognition model, and optimizing and matching a microfacies mode through a related matching algorithm and a correction rule, so as to calculate and draw a microfacies plane diagram and a profile diagram which conform to a sedimentary mode. According to the method, the reservoir sedimentary microfacies can be automatically, efficiently and accurately identified and depicted, the working efficiency is greatly improved, the accuracy is improved, and the labor cost is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of oil and gas reservoir evaluation, and in particular to a reservoir sedimentary microfacies characterization method based on artificial intelligence image recognition. Background Art

[0002] Sedimentary microfacies is an important parameter for describing the sedimentary environment and characteristics of reservoirs, and is of great significance for the development of oil and gas reservoirs. Traditional sedimentary microfacies analysis mainly relies on core observation, thin section analysis and other methods, and manually completes the division and delineation of sedimentary microfacies based on geological prior knowledge and experience. These methods are not only labor-intensive, time-consuming, and costly, but the results are also greatly affected by human factors, making it difficult to achieve large-scale and efficient reservoir sedimentary characterization and evaluation.

[0003] At present, artificial intelligence methods have achieved some applications and results in the intelligent characterization of single-well sedimentary microfacies, but there has been no research and application in the characterization of the spatial distribution of sedimentary microfacies based on deep learning image recognition based on sedimentary pattern constraints. Summary of the invention

[0004] In view of this, the present invention aims to propose a reservoir sedimentary microfacies characterization method based on artificial intelligence image recognition, which can automatically, efficiently and accurately identify and characterize reservoir sedimentary microfacies.

[0005] To achieve the above object, the technical solution of the present invention is implemented as follows: a reservoir sedimentary microfacies characterization method based on artificial intelligence image recognition comprises the following steps:

[0006] Step 1: Based on the existing geological prior knowledge of sedimentary patterns, different types of sedimentary microfacies patterns are collected and sorted, including sedimentary microfacies plane maps and well-connected phase cross-section maps, to construct a sedimentary microfacies pattern library;

[0007] Step 2, obtaining image data in the sedimentary microfacies pattern library and preprocessing the image, the preprocessing includes denoising, smoothing, transformation, enhancement and standardization;

[0008] Step 3, extracting key feature parameters from the preprocessed image, including but not limited to morphology, scale, combination relationship, superposition relationship, distribution direction, and change law;

[0009] Step 4: Use deep learning algorithms to train the extracted image features, obtain recognition rules and decision rules, and build a sedimentary microfacies recognition model;

[0010] Step 5: Evaluate and verify the trained model, compare expert analysis with existing data, evaluate the accuracy of image recognition results, and adjust and optimize the model based on feedback to ensure the accuracy and reliability of the model;

[0011] Step 6: Input the limited constraint data into the sedimentary microfacies identification model. Based on the relevant matching algorithm and correction rules, the model intelligently optimizes the matching microfacies pattern and calculates and draws sedimentary microfacies plan and cross-sectional diagrams that fully match the data and conform to the laws of the sedimentary pattern.

[0012] Furthermore, step 1 includes:

[0013] Arrange and classify the collected sedimentary microfacies patterns, including sedimentary microfacies plane maps and well-connected facies cross-section maps;

[0014] Classify sedimentary microfacies model maps according to different sedimentary environments, sedimentary systems and geological periods, and systematically establish a sedimentary microfacies model library by comprehensively applying knowledge and methods of multiple disciplines such as geology, geostatistics and geological information technology to ensure the quality and application effect of the sedimentary microfacies model library;

[0015] The sedimentary microfacies pattern library is updated regularly to absorb new research results and practical experience.

[0016] Furthermore, in step 2, when preprocessing the image, it is necessary to select appropriate methods and parameters according to the specific situation to perform denoising, smoothing, transformation, enhancement and standardization on the image, such as median filtering or Gaussian filtering denoising, mean filtering or Gaussian filtering smoothing, image geometric transformation, histogram equalization or contrast stretching enhancement, grayscale standardization, edge detection, and evaluate and adjust the preprocessing effect to ensure the accuracy of subsequent image recognition.

[0017] Furthermore, in step 3, when extracting key characteristic parameters, it is necessary to select appropriate methods and indicators in combination with the actual characteristics of the sedimentary microfacies and the purpose of analysis. At the same time, for complex images and features, it is necessary to combine multiple methods for comprehensive analysis and extraction, including using edge detection and morphological operations to extract morphological features such as sediment shape, size, and contour; extracting scale features of sediment length, width, and thickness through pixel counting and contour area calculation; extracting combination relationship features between sedimentary microfacies through image segmentation and connectivity analysis; extracting superposition relationship features between sedimentary microfacies through texture analysis, edge detection, and morphological operations; extracting distribution direction and trend features of sedimentary microfacies through directional filtering and Hough transform; and extracting the change law and spatial distribution characteristics of sedimentary microfacies through statistical analysis, spatial analysis, and image feature description.

[0018] Furthermore, step 4 includes:

[0019] Design and build a model architecture suitable for sedimentary microfacies image recognition, design and adjust the model architecture according to the requirements of the task, and train it based on the sedimentary microfacies images that have been preprocessed and feature extracted;

[0020] The model parameters are updated through the back-propagation algorithm. During the training process, common optimizers and loss functions are used to minimize the prediction error of the model. Recognition rules and decision rules are obtained based on the training, thus constructing a sedimentary microfacies recognition model.

[0021] Furthermore, step 5 includes:

[0022] Evaluate the trained model using an independent test set that contains data that the model has not seen before.

[0023] Compare the model's predictions with expert analysis to determine the model's error patterns and error types. Based on the model's evaluation metrics and expert analysis, identify the model's strengths and areas for improvement. Record the model's performance in different categories and under different conditions.

[0024] Optimize the model’s hyperparameters based on evaluation results and expert feedback;

[0025] Use data augmentation techniques to increase data diversity and improve the generalization ability of the model;

[0026] Carry out model structure optimization and change the model architecture to further improve model performance;

[0027] Repeat verification and adjustment until the model reaches a satisfactory performance level, and regularly verify and re-evaluate the model to ensure its stability and reliability in long-term operation to cope with different automatic identification and analysis tasks of sedimentary microfacies images.

[0028] Furthermore, step 6 includes:

[0029] Collect and prepare constraint data such as single well microfacies profiles, seismic attribute planes, and seismic attribute profiles, and preprocess the data, including removing noise, filling missing values, and standardizing to ensure data quality and consistency;

[0030] Input the constraint data into the sedimentary microfacies map recognition model, and use the relevant matching algorithm to match it in the microfacies pattern library. The algorithm includes a traditional pattern matching method or a feature extraction and matching algorithm in deep learning;

[0031] The model performs intelligent optimization and application of correction rules based on the matching results to ensure that the matching results conform to the laws and constraints of the sedimentation pattern. The correction rules are based on expert knowledge and experience, or on a data-driven approach.

[0032] Generate sedimentary microfacies plane and cross-sectional images based on the optimized matching results;

[0033] The results were compared and discussed by geological experts to verify the accuracy and reliability of the model. Based on the evaluation and verification results, the model was optimized and adjusted to improve its performance and applicability.

[0034] Compared with the prior art, the reservoir sedimentary microfacies characterization method based on artificial intelligence image recognition described in the present invention has the following advantages: the present invention realizes a new reservoir sedimentary microfacies characterization method by introducing advanced artificial intelligence technology, which changes the traditional sedimentary facies research method and provides an efficient and accurate reservoir evaluation technical solution, specifically including:

[0035] (1) The technical process starts with the construction of a sedimentary microfacies model library, which has the advantages of making full use of prior knowledge and knowledge integration;

[0036] (2) Image preprocessing operations and evaluation adjustments are more targeted and flexible, effectively improving image quality;

[0037] (3) The key elements of the deep learning model architecture design and training process are described in detail. The model construction and training process are more suitable for the sedimentary microfacies image recognition task, which effectively improves the model performance. The model has strong self-learning ability, strong complex image data processing ability, and high recognition accuracy and efficiency.

[0038] (4) The model is comprehensively evaluated and continuously optimized to ensure that the model performance is continuously improved and adapted to different tasks, making the model more reliable;

[0039] (5) Compared with only integrating sand body related data to characterize planar sedimentary microfacies, the present invention adds model optimization and in-depth expert participation;

[0040] (6) Utilizing the powerful feature extraction capability of convolutional neural networks (CNN), the convolution layer can automatically extract local features in the image, such as morphological features in sedimentary microfacies patterns. For complex geological structures and sedimentary features in sedimentary microfacies patterns, CNN can capture feature information at different scales through convolution kernels of different sizes. For example, in identifying the boundaries and morphological changes of sedimentary bodies, the convolution layer can focus on the features of local areas, which helps to more accurately characterize the detailed features of sedimentary microfacies. Traditional methods may require manual design of feature extraction rules, which are difficult to adapt to complex and changeable sedimentary microfacies images.

[0041] (7) Convolutional neural network (CNN) implements parameter sharing to reduce the amount of calculation. The parameter sharing mechanism in CNN enables the same features at different locations to be calculated using the same parameters when processing images. In the present invention, for image data such as sedimentary microfacies pattern maps with certain regularity and repetitive structures, parameter sharing can greatly reduce the number of model parameters, thereby reducing the calculation cost and improving the calculation efficiency. Compared with other model structures such as fully connected networks, CNN can complete training and prediction tasks faster when processing large-scale sedimentary microfacies image data;

[0042] (8) Improved efficiency: The method of the present invention greatly reduces the time required for microphase analysis through automated processing, effectively improving processing efficiency;

[0043] (9) Improved accuracy: The method of the present invention utilizes the powerful feature extraction and classification capabilities of deep learning to reduce human errors and improve analysis accuracy;

[0044] (10) Cost reduction: The sedimentary microfacies analysis by the method of the present invention effectively reduces the need for manual analysis and reduces long-term operating costs;

[0045] (11) Wide applicability: The method of the present invention is applicable to different geological backgrounds and complex reservoirs and has wide applicability. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:

[0047] Figure 1 It is a flow chart of a reservoir sedimentary microfacies characterization method based on artificial intelligence image recognition according to an embodiment of the present invention;

[0048] Figure 2 It is a braided river delta depositional model (Figure (a) is a steep slope type, and Figure (b) is a gentle slope type);

[0049] Figure 3 This is a flow chart of the convolutional neural network (CNN) image segmentation process;

[0050] Figure 4 is the input constraint data map (wherein, map (a) is the root mean square amplitude seismic attribute map of the 7th layer of the W oil group, map (b) is the relative impedance seismic attribute map of the 7th layer of the W oil group, and map (c) is the Vp / Vs inversion seismic attribute map of the 7th layer of the W oil group);

[0051] Figure 5 This is the sedimentary microfacies diagram of the 7th sublayer of the W oil group. DETAILED DESCRIPTION

[0052] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0053] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0054] The present invention introduces advanced artificial intelligence technology to realize a new reservoir sedimentary microfacies characterization method, which changes the traditional sedimentary facies research method and provides an efficient and accurate reservoir evaluation technical solution, as follows:

[0055] like Figure 1-Figure 5 As shown, the present invention is a reservoir sedimentary microfacies characterization method based on artificial intelligence image recognition. The technical process of the method described in the present invention starts from the construction of a sedimentary microfacies model library, and has the advantages of making full use of prior knowledge and knowledge integration; image preprocessing operations and evaluation adjustments are more targeted and flexible, and effectively improve image quality; the key elements of the deep learning model architecture design and training process are described in detail, and the model construction and training process are more in line with the sedimentary microfacies image recognition task, which effectively improves the model performance, and the model has a strong self-learning ability, strong complex image data processing ability, high recognition accuracy and efficiency; the model is comprehensively evaluated and continuously optimized to ensure that the model performance is continuously improved and adapted to different tasks, making the model more reliable; compared with only comprehensive sand body related data to characterize planar sedimentary microfacies, the method described in the present invention adds model optimization and expert in-depth participation. Specifically includes the following steps:

[0056] Step 1, construction of sedimentary microfacies pattern library: widely collect sedimentary microfacies pattern maps from different geological regions and different sedimentary environments (such as fluvial phase, delta phase, lacustrine phase, etc.), including the classic pattern maps in the published geological research literature, the pattern maps generated by the measured data in the actual exploration and development of oil fields, and these geological prior knowledge, and classify these pattern maps in detail according to the sedimentary environment and sedimentary system (such as meandering river sedimentary system, fan delta sedimentary system, etc.). For example, all pattern maps under fluvial phase sedimentary environment are classified into one category, and this category is further subdivided according to different river types (such as braided river, meandering river); in the process of sorting, geological knowledge is used to judge the sedimentary microfacies type, geostatistical methods are used to analyze the statistical laws of sedimentary characteristics, and geological information technology is used to digitize and store data; by establishing a database management system, the classified pattern maps are stored and constructed into a sedimentary microfacies pattern library, including sedimentary microfacies plane maps and well-connected phase profiles, and they are regularly updated to incorporate the latest research results and practical experience, such as adding newly discovered special sedimentary microfacies patterns to the corresponding categories in a timely manner.

[0057] The first advantage of this step is that it builds a knowledge base covering a variety of sedimentary conditions through multi-source data collection and multi-disciplinary knowledge integration and classification. Step 1 combined with subsequent steps provides a comprehensive and accurate reference standard for the characterization of the entire reservoir sedimentary microfacies. For example, in step 6, the model can find the most matching pattern in this library based on the input constraint data, thereby improving the accuracy of the final characterization result. The second advantage of this step is that it is classified and updated regularly according to multiple factors, so that the library can adapt to different geological conditions and research progress. When the geological conditions in the study area are complex and changeable, the rich library content and timely updated information can ensure that the model has sufficient reference basis and improve the applicability and reliability of the model.

[0058] Step 2, sedimentary microfacies pattern image preprocessing: obtain image data in the sedimentary microfacies pattern library, and perform preprocessing such as denoising, smoothing, transformation, enhancement and standardization on the image to improve the accuracy of subsequent image recognition;

[0059] For image denoising, the appropriate filtering method is selected according to the type of image noise. If the noise is salt and pepper noise, median filtering is preferred, which removes noise by replacing the central pixel value with the median value in the pixel neighborhood. If it is Gaussian noise, Gaussian filtering is used, which uses the Gaussian function to calculate the weighted average of the neighborhood pixels to smooth the image. During smoothing, mean filtering can be used to further remove small fluctuations in the image, and Gaussian filtering can be used again to retain image edge information while smoothing. Image geometric transformation can be used to correct the geometric distortion of the image, such as rotating a tilted image to a horizontal direction. Histogram equalization enhances image contrast by stretching the image histogram to make the details in the image clearer. Contrast stretching can enhance the contrast of specific areas by customizing the stretching range as needed. Grayscale normalization maps image pixel values ​​to a specific range for subsequent processing. Edge detection uses methods such as the Sobel operator and the Canny operator to highlight edge features in the image for subsequent feature extraction. By selecting a variety of targeted preprocessing methods, the image quality is improved by effectively removing noise, smoothing images, enhancing contrast and highlighting edges. Step 2 provides a high-quality data basis for subsequent steps. For example, in step 3, clear image edges help to more accurately extract the morphological features of the sedimentary body, and the high-quality image as a whole is conducive to extracting more accurate features such as scale and combination relationships; precise preprocessing operations enhance the recognizability of image features, and the combination of different preprocessing methods makes the sedimentary microfacies features in the image more obvious. Features such as morphology, scale and combination relationships can be better extracted and analyzed, thereby improving the recognition accuracy of the entire model.

[0060] Step 3, key feature parameter extraction: extract key feature parameters from the preprocessed image, including but not limited to morphology, scale, combination relationship, superposition relationship, distribution direction, change law, etc.;

[0061] For morphological feature extraction, edge detection operators (such as Canny operators) are first used to detect the edges of the sediments, and then morphological operations (such as expansion, erosion, opening operation, closing operation, etc.) are used to optimize and analyze the edges; the expansion operation can fill small holes in the edges, the erosion operation can remove small protrusions on the edges, the opening operation is used to remove small isolated areas, and the closing operation can connect adjacent edge parts, so as to accurately extract the shape, size, contour and other morphological features of the sediments; in terms of scale feature extraction, the number of pixels of the sediments in the image is determined by pixel counting, and then converted into actual length, width, thickness and other dimensions according to the image resolution; for combined relationship features, image segmentation algorithms (such as threshold-based segmentation, region growing algorithm, etc.) are used to extract the image. Different sedimentary microfacies areas are separated, and then the connection and distribution relationship between different sedimentary microfacies are determined through connectivity analysis; when extracting the superposition relationship features, texture analysis (such as gray-level co-occurrence matrix method to analyze texture features), edge detection and morphological operations are combined to determine the superposition order and mode of different sedimentary microfacies in the vertical and plane directions; the distribution direction and trend characteristics use directional filtering to enhance the directional information in the image, and then the straight line or curve features are detected by Hough transform to determine the distribution direction of the sedimentary microfacies; the change law and spatial distribution characteristics are extracted by statistical analysis (such as calculating statistical indicators such as mean and variance), spatial analysis (such as analyzing the distribution density of sedimentary microfacies in different regions) and image feature description (such as describing the texture complexity of sedimentary microfacies). A variety of methods are used in combination to comprehensively and deeply explore the various characteristics of sedimentary microfacies. Step 3 is combined with the model evaluation optimization in step 5. The rich features extracted provide more judgment basis for the model, which helps the model to better learn and identify different sedimentary microfacies and improve the accuracy of the model. For example, accurate morphological features can help the model distinguish sedimentary bodies of different shapes, thereby accurately judging the type of sedimentary microfacies; meticulous and accurate feature extraction provides a basis for the precise characterization of sedimentary microfacies. The extracted features cover all aspects of sedimentary microfacies. These features are used in step 6 to match the constraint data and generate the final sedimentary microfacies plan and cross-sectional views, making the characterization results more accurate and reliable, and able to truly reflect the actual situation of the sedimentary microfacies.

[0062] Step 4, deep learning model construction and training: Use deep learning algorithms, such as convolutional neural networks (CNN), to train the extracted image features, obtain recognition rules and decision rules, and build a sedimentary microfacies recognition model. When designing the CNN model architecture, the number of convolutional layers, convolution kernel size, and step size are determined according to the characteristics of the sedimentary microfacies image. For example, for sedimentary microfacies patterns with rich image details, the number of convolutional layers can be increased and smaller convolution kernels can be used to extract more detailed features. The pooling layer uses the maximum pooling or average pooling method to reduce the data dimension and retain important features. The fully connected layer sets the appropriate number of neurons as needed to map the extracted features to the final classification or regression results. Training is performed based on preprocessed and feature-extracted sedimentary microfacies images, and the image data is divided into training set, validation set and test set. During the training process, the Adam optimizer is used to dynamically adjust the learning rate to accelerate the training convergence speed. The cross-entropy loss function is used to measure the difference between the model prediction value and the true label. The error is back-propagated from the output layer to the input layer through the back-propagation algorithm, and the weight and bias parameters of the model are updated. The training is continuously iterated until the performance of the model on the validation set is optimal, thereby constructing a sedimentary microfacies image recognition model. The reasonable design and training process of the CNN model architecture give the model powerful automatic feature learning capabilities. Step 4 is combined with steps 2 and 3, and training is performed using preprocessed high-quality images and extracted rich features, so that the model can automatically learn the complex features and patterns in the sedimentary microfacies images and improve recognition accuracy. For example, the model can learn the feature combination patterns of different sedimentary microfacies in the image, thereby accurately classifying them. The structural characteristics of CNN and the use of optimization algorithms improve data processing efficiency. When processing large-scale sedimentary microfacies image data, CNN's parameter sharing and other mechanisms reduce the amount of calculation. The Adam optimizer accelerates the training process, allowing the model to be quickly trained and predicted, which is suitable for large-scale data processing needs in actual engineering.

[0063] Step 5, model evaluation and optimization: Evaluate and verify the trained model, compare expert analysis with existing data, evaluate the accuracy of image recognition results, and adjust and optimize the model based on feedback to ensure the accuracy and reliability of the model. Through comprehensive evaluation and continuous optimization, ensure that the model performance continues to improve. Step 5 and step 4 are repeated in a cycle, and the model is adjusted according to the evaluation results so that the model can continuously adapt to different sedimentary microfacies image data to improve accuracy and reliability. For example, through data enhancement and structural optimization, the model can better handle new data situations and reduce the risk of overfitting. Regular verification and re-evaluation ensure the stability of the model in long-term operation. In practical applications, geological data and research needs may change over time. Regular verification can promptly detect changes in model performance, adjust the model to adapt to new situations, and ensure that the model can work effectively in different periods.

[0064] Step 6, reservoir sedimentary microfacies characterization: limited constraint data (such as single well microfacies profile, seismic attribute plane map, seismic attribute profile map, etc.) are input into the sedimentary microfacies map recognition model. Based on the relevant matching algorithm and correction rules, the model intelligently optimizes the matching microfacies pattern, and calculates and draws sedimentary microfacies plane maps and profile maps that fully match the data and conform to the laws of the sedimentary pattern. The comprehensive use of multiple constraint data and expert participation improve the accuracy of characterization. Step 6 is combined with step 1. The constraint data is matched in a rich model library. Experts compare and discuss to ensure that the results are in line with the actual geological situation, so that the final generated sedimentary microfacies plane map and profile map can accurately reflect the true characteristics of the reservoir sedimentary microfacies and reduce errors. The application of multiple matching algorithms and correction rules enhances the adaptability of the model. Under different geological regions and sedimentary conditions, the model can select appropriate matching algorithms and correction rules according to the constraint data to better adapt to complex and changeable geological conditions and improve the application effect of the model in different scenarios.

[0065] In step 1, it is necessary to sort and classify the collected sedimentary microfacies pattern maps, including sedimentary microfacies plane maps and well-connected phase profile maps; classify the sedimentary microfacies pattern maps according to different sedimentary environments, sedimentary systems, geological periods and other factors, and comprehensively use geology, geostatistics, geological information technology and other disciplines. Knowledge and methods, systematically establish a sedimentary microfacies pattern library to ensure the quality and application effect of the sedimentary microfacies pattern library; regularly update the sedimentary microfacies pattern library to absorb new research results and practical experience. In the description of the present invention, unless otherwise specified, the meaning of "multiple" is two or more.

[0066] In step 2, when preprocessing the image, it is necessary to select appropriate methods and parameters according to the specific situation, and perform preprocessing such as denoising, smoothing, transformation, enhancement and standardization on the image, such as median filtering or Gaussian filtering denoising, mean filtering or Gaussian filtering smoothing, image geometric transformation, histogram equalization or contrast stretching enhancement, grayscale standardization, edge detection, etc., and evaluate and adjust the preprocessing effect to ensure the accuracy of subsequent image recognition.

[0067] In step 3, when extracting key characteristic parameters, it is necessary to select appropriate methods and indicators in combination with the actual characteristics of sedimentary microfacies and the purpose of analysis. At the same time, for complex images and features, it is necessary to combine multiple methods for comprehensive analysis and extraction, including using edge detection, morphological operations (dilation, erosion, opening operation, closing operation, etc.) and other methods to extract morphological features such as sediment shape, size, and contour; extracting scale features such as sediment length, width, and thickness through pixel counting, contour area calculation and other methods; extracting combination relationship features between sedimentary microfacies through image segmentation and connectivity analysis; extracting superposition relationship features between sedimentary microfacies through texture analysis, edge detection and morphological operations; extracting distribution direction and trend features of sedimentary microfacies through directional filtering, Hough transform and other methods; and extracting the change law and spatial distribution characteristics of sedimentary microfacies through statistical analysis, spatial analysis and image feature description.

[0068] In step 4, a model architecture suitable for sedimentary microfacies image recognition is designed and constructed. For example, a CNN model architecture usually includes components such as convolutional layers, pooling layers, and fully connected layers. The model architecture is designed and adjusted according to the requirements of specific tasks, and training is performed based on sedimentary microfacies images that have been preprocessed and feature extracted. The model parameters are updated through a back-propagation algorithm. During the training process, common optimizers (such as Adam, SGD, etc.) and loss functions (such as cross-entropy loss function) can be used to minimize the prediction error of the model. Recognition rules and decision rules are obtained based on training, thereby constructing a sedimentary microfacies image recognition model.

[0069] Utilizing the powerful feature extraction capability of convolutional neural networks (CNN), the convolution layer can automatically extract local features in the image, such as morphological features in sedimentary microfacies pattern maps. For complex geological structures and sedimentary features in sedimentary microfacies pattern maps, CNN can capture feature information of different scales through convolution kernels of different sizes. For example, in identifying the boundaries and morphological changes of sedimentary bodies, the convolution layer can focus on the features of local areas, which helps to more accurately characterize the detailed features of sedimentary microfacies, while traditional methods may require manual design of feature extraction rules, which is difficult to adapt to complex and changeable sedimentary microfacies images. Convolutional neural networks (CNN) implement parameter sharing to reduce the amount of calculation. The parameter sharing mechanism in CNN enables the same features at different locations to be calculated using the same parameters when processing images. In this patent, for image data with certain regularity and repetitive structures such as sedimentary microfacies pattern maps, parameter sharing can greatly reduce the number of model parameters, thereby reducing computational costs and improving computational efficiency. Compared with other model structures such as fully connected networks, CNN can complete training and prediction tasks faster when processing large-scale sedimentary microfacies image data.

[0070] In step 5, the trained model is evaluated using an independent test set, which should contain data that the model has not seen before, to evaluate the generalization ability of the model and calculate the model's accuracy, precision, recall, F1 score and other indicators; the model's prediction results are compared with expert analysis to determine the model's error mode and error type, and the model's strengths and improvements are determined based on the model's evaluation indicators and expert analysis. The performance of the model under different categories and conditions is recorded, such as the accuracy and error analysis for various sedimentary microfacies; based on the evaluation results and expert feedback, the model's hyperparameters are optimized, such as learning rate, batch size, convolution kernel size, etc.; data enhancement techniques such as rotation, translation, scaling, and inversion are used to increase data diversity and improve the model's generalization ability; model structure optimization is carried out to change the model's architecture, such as adding or reducing convolutional layers, adjusting pooling strategies or fully connected layer structures, to further improve model performance; repeated verification and adjustment are performed until the model reaches an expert feedback satisfaction level of ≥90%, and the model is verified and re-evaluated regularly to ensure the stability and reliability of the model in long-term operation to cope with different sedimentary microfacies image automatic recognition and analysis tasks.

[0071] In step 6, the constraint data such as single well microfacies profile, seismic attribute plane map, seismic attribute profile map are collected and prepared, and the data are preprocessed, including noise removal, missing value filling, standardization, etc., to ensure the quality and consistency of the data; the constraint data are input into the sedimentary microfacies map recognition model, and matched in the microfacies pattern library using a relevant matching algorithm. The algorithm includes traditional pattern matching methods such as correlation analysis and template matching, or includes feature extraction and matching algorithms in deep learning; the model performs intelligent optimization and application of correction rules based on the matching results to ensure that the matching results conform to the laws and constraints of the sedimentary model. The correction rules are based on expert knowledge and experience, or based on data-driven methods; based on the optimized matching results, sedimentary microfacies plane maps and profiles are generated; the results are compared and discussed by geological experts to verify the accuracy and reliability of the model, and the model is optimized and adjusted based on the evaluation and verification results to improve its performance and applicability.

[0072] The technical solution of the present invention is further described below through specific embodiments.

[0073] The specific implementation methods of the present invention are further described by taking the actual drawing scene of the small layer sand thickness map of the W oil group in the L area of ​​an offshore oil field as an example and combining with the accompanying drawings.

[0074] The W oil group in the L area belongs to the braided river delta front deposits, mainly developing underwater distributary channels (including the main body and lateral edges), estuary bars, remote sand bars, sheet sands, and interchannel microfacies. The number of wells in the work area is relatively small.

[0075] First, based on the existing typical braided river delta sedimentary model, the sedimentary microfacies distribution pattern is clarified, and the empirical database data is constructed according to the sedimentary microfacies pattern diagram; the existing typical braided river delta sedimentary model is as follows Figure 2 shown.

[0076] Secondly, the sedimentary microfacies pattern images are preprocessed and a variety of key characteristic parameters are extracted from the preprocessed images;

[0077] Next, a convolutional neural network (CNN) is used to obtain sedimentation pattern image information and segment it. After the convolution layer classifies the pixels, it is trained based on the image features to build a training model. The convolutional neural network (CNN) image segmentation processing flow chart is shown in Figure 3 As shown by Figure 3 It can be seen that by inputting a pre-processed plane image of a sedimentary microfacies pattern, the convolution layer performs a convolution operation by sliding the convolution kernel on the image to extract the features of the image. The output of the convolution layer is then processed by an activation function (such as a nonlinear function ReLU). The output of the activation function enters the pooling layer, and the dimension of the data is reduced by downsampling operations (such as maximum pooling or average pooling) while retaining important feature information. The output of the pooling layer is used for feature map extraction, and the extracted feature map is sent to feature map analysis and processing to ensure the validity and accuracy of the feature map. After the feature map analysis, the data enters the segmentation algorithm stage, and a specific algorithm is used to segment the image to distinguish different objects or regions in the image. Finally, the image segmentation results are output, including images with different regions labeled or classification results for different objects in the image.

[0078] Then, the trained model is evaluated and verified by comparing expert analysis with existing data to evaluate the accuracy of the image recognition results, and the model is adjusted and optimized based on the feedback to ensure the accuracy and reliability of the model;

[0079] Finally, the small layer seismic attribute plane map (such as Figure 4 (a) shows the RMS amplitude seismic attribute map, Figure 4 (b) shows the relative impedance seismic attribute map, Figure 4 The Vp / Vs inversion seismic attribute map (shown in (c)) and the small-layer single-well sedimentary microfacies label data are input into the model. The model generates a pattern trend based on the trend of the input plane map and the small-layer single-well microfacies label data, and matches and optimizes it with the empirical microfacies pattern map obtained by model training, and then calculates and draws the final sedimentary microfacies plane map, as shown in Figure 5 The sedimentary microfacies diagram of the 7th layer of the W oil group shown in the figure is composed of Figure 5 It can be seen that the distribution and combination pattern of sedimentary microfacies conform to the geological model, and its distribution trend is strictly constrained by the distribution of seismic attribute values, meeting the actual data requirements of the work area and achieving the accuracy required for geological research.

[0080] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A reservoir sedimentary microfacies characterization method based on artificial intelligence image recognition, characterized in that: The following steps are involved: Step 1: Based on the existing geological prior knowledge of sedimentary patterns, different types of sedimentary microfacies patterns are collected and sorted, including sedimentary microfacies plane maps and well-connected phase cross-section maps, to construct a sedimentary microfacies pattern library; Step 2, obtaining image data in the sedimentary microfacies pattern library and preprocessing the image, the preprocessing includes denoising, smoothing, transformation, enhancement and standardization; Step 3, extracting key feature parameters from the preprocessed image, including but not limited to morphology, scale, combination relationship, superposition relationship, distribution direction, and change law; Step 4: Use deep learning algorithms to train the extracted image features, obtain recognition rules and decision rules, and build a sedimentary microfacies recognition model; Step 5: Evaluate and verify the trained model, compare expert analysis with existing data, evaluate the accuracy of image recognition results, and adjust and optimize the model based on feedback to ensure the accuracy and reliability of the model; Step 6: Input the limited constraint data into the sedimentary microfacies identification model. Based on the relevant matching algorithm and correction rules, the model intelligently optimizes the matching microfacies pattern and calculates and draws sedimentary microfacies plan and cross-sectional diagrams that fully match the data and conform to the laws of the sedimentary pattern.

2. The reservoir sedimentary microfacies characterization method based on artificial intelligence image recognition according to claim 1 is characterized in that: Step 1 includes: Arrange and classify the collected sedimentary microfacies patterns, including sedimentary microfacies plane maps and well-connected facies cross-section maps; Classify sedimentary microfacies model maps according to different sedimentary environments, sedimentary systems and geological periods, and systematically establish a sedimentary microfacies model library by comprehensively applying knowledge and methods of multiple disciplines such as geology, geostatistics and geological information technology to ensure the quality and application effect of the sedimentary microfacies model library; The sedimentary microfacies pattern library is updated regularly to absorb new research results and practical experience.

3. The reservoir sedimentary microfacies characterization method based on artificial intelligence image recognition according to claim 1 is characterized in that: In step 2, when preprocessing the image, it is necessary to select appropriate methods and parameters according to the specific situation to perform denoising, smoothing, transformation, enhancement and standardization on the image, such as median filtering or Gaussian filtering denoising, mean filtering or Gaussian filtering smoothing, image geometric transformation, histogram equalization or contrast stretching enhancement, grayscale standardization, edge detection, and evaluate and adjust the preprocessing effect to ensure the accuracy of subsequent image recognition.

4. The reservoir sedimentary microfacies characterization method based on artificial intelligence image recognition according to claim 3 is characterized in that: In step 3, when extracting key characteristic parameters, it is necessary to select appropriate methods and indicators in combination with the actual characteristics of sedimentary microfacies and the purpose of analysis. At the same time, for complex images and features, it is necessary to combine multiple methods for comprehensive analysis and extraction, including using edge detection and morphological operations to extract morphological features such as sediment shape, size, and contour; The scale characteristics of sediment length, width and thickness are extracted by pixel counting and contour area calculation. The combination relationship characteristics between sedimentary microfacies are extracted by image segmentation and connectivity analysis. The superposition relationship characteristics between sedimentary microfacies are extracted through texture analysis, edge detection and morphological operation. Directional filtering and Hough transform are used to extract the distribution direction and trend characteristics of sedimentary microfacies. Statistical analysis, spatial analysis and image feature description are used to extract the variation patterns and spatial distribution characteristics of sedimentary microfacies.

5. The reservoir sedimentary microfacies characterization method based on artificial intelligence image recognition according to claim 4 is characterized in that: Step 4 includes: Design and build a model architecture suitable for sedimentary microfacies image recognition, design and adjust the model architecture according to the requirements of the task, and train it based on the sedimentary microfacies images that have been preprocessed and feature extracted; The model parameters are updated through the back-propagation algorithm. During the training process, common optimizers and loss functions are used to minimize the prediction error of the model. Recognition rules and decision rules are obtained based on the training, thus constructing a sedimentary microfacies recognition model.

6. The reservoir sedimentary microfacies characterization method based on artificial intelligence image recognition according to claim 5 is characterized in that: Step 5 includes: Evaluate the trained model using an independent test set that contains data that the model has not seen before. Compare the model's predictions with expert analysis to determine the model's error patterns and error types. Based on the model's evaluation metrics and expert analysis, identify the model's strengths and areas for improvement. Record the model's performance in different categories and under different conditions. Optimize the model’s hyperparameters based on evaluation results and expert feedback; Use data augmentation techniques to increase data diversity and improve the generalization ability of the model; Carry out model structure optimization and change the model architecture to further improve model performance; Repeat verification and adjustment until the model reaches a satisfactory performance level, and regularly verify and re-evaluate the model to ensure its stability and reliability in long-term operation to cope with different automatic identification and analysis tasks of sedimentary microfacies images.

7. The reservoir sedimentary microfacies characterization method based on artificial intelligence image recognition according to claim 1 is characterized in that: Step 6 includes: Collect and prepare constraint data such as single well microfacies profiles, seismic attribute planes, and seismic attribute profiles, and preprocess the data, including removing noise, filling missing values, and standardizing to ensure data quality and consistency; Input the constraint data into the sedimentary microfacies map recognition model, and use the relevant matching algorithm to match it in the microfacies pattern library. The algorithm includes a traditional pattern matching method or a feature extraction and matching algorithm in deep learning; The model performs intelligent optimization and application of correction rules based on the matching results to ensure that the matching results conform to the laws and constraints of the sedimentation pattern. The correction rules are based on expert knowledge and experience, or on a data-driven approach. Generate sedimentary microfacies plane and cross-sectional images based on the optimized matching results; The results were compared and discussed by geological experts to verify the accuracy and reliability of the model. Based on the evaluation and verification results, the model was optimized and adjusted to improve its performance and applicability.

Images