Automatic diagenesis phase sample labeling method and system fusing affinity propagation clustering and graph convolutional neural network

Abstract

The invention provides a diagenesis phase sample automatic labeling method and system fusing affinity propagation clustering and a graph convolutional neural network, and belongs to the field of diagenesis phase intelligent identification. The method aims to solve the problems of small sample size and low accuracy of diagenetic phase labeling work. The method comprises the following steps: preprocessing logging curve data, labeling a small amount of labels, and constructing an automatic labeling data set; establishing an incidence relation between logging curve depth nodes by utilizing an affinity propagation clustering structure map structure; node features are aggregated through a graph convolution layer, and rapid and accurate marking of lithogenous phases is achieved. Compared with a traditional method, the method has the advantages that complicated correlation and spatial distribution characteristics between lithogenous phases can be better processed, and the graph structure and the logging curve characteristics are combined, so that the marking accuracy and stability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of diagenetic facies intelligent identification, and in particular to a diagenetic facies sample automatic labeling method and system integrating affinity propagation clustering and graph convolutional neural network. Background Art

[0002] Diagenetic facies sample labeling is the key link in the intelligent identification of diagenetic facies. Traditional diagenetic facies labeling methods mostly rely on the experience of geological experts. [1-3] , it is necessary to analyze the logging curve and make manual judgments, which has the problems of large data volume, time-consuming and labor-intensive; in addition, due to the quality problems of the logging curve itself and the errors of manual judgment, the accuracy of the final diagenetic phase identification will also be affected. Therefore, the automatic annotation of diagenetic phase samples has important practical significance for improving the speed and efficiency of reservoir evaluation. [4,5] .

[0003] Traditional automatic labeling techniques can be summarized into three types: semi-supervised learning, self-supervised learning, and deep learning. Semi-supervised learning trains the model by partially training the labeled data, and then preliminarily labels the unlabeled data, and improves the model accuracy through alternating training and correction. [6] Although this method reduces the reliance on fully annotated data, its effectiveness is limited by the quality of the model and data. Since the quality of well logging curve data is relatively weak or has certain deviations, it is difficult to play its advantages in the automatic annotation of diagenetic phase samples. Self-supervised learning learns through the internal structure of the data, which can greatly reduce the reliance on manual annotation. [7] , but the training process is complicated and requires a large amount of data sets. Due to the complex characteristics of diagenetic facies sample data and insufficient data volume, it is difficult to be effectively applied in the automatic annotation of diagenetic facies samples.

[0004] With the rapid development of deep learning technology, automatic labeling technology has also been significantly improved. Convolutional neural networks, recurrent neural networks and other models are usually used to identify and classify time series data or images, thereby automatically generating labels. [8,9] This type of method can effectively identify complex features at a fast speed, but it still faces some challenges in the field of automatic identification of diagenetic phase samples: first, the training process of deep learning usually requires a large number of samples, but due to cost factors, the number of diagenetic phase samples is difficult to support; second, due to the complexity and diversity of the geological environment, it is difficult for the model to fully capture all features, resulting in unsatisfactory accuracy.

[0005] In recent years, graph convolutional neural networks

[10] It is used to solve the problem of sample labeling of graph structure data. [11-13], especially in the few-label learning scenario, it can also maintain high learning efficiency and annotation accuracy. The diagenetic phase formation process is affected by sedimentation and various factors, and there are certain distribution characteristics in the spatial stratigraphic structure. Therefore, the use of graph convolutional neural networks can construct a graph structure to capture the correlation between different depths, thereby improving the accuracy of annotation.

[0006] At the same time, in order to further overcome the problem of few samples and optimize the graph structure, the Affinity Propagation (AP) algorithm is used. The AP algorithm works through "message passing" and dynamically adjusts the connection based on the similarity between nodes to determine the most suitable clustering center, rather than relying on fixed rules. [14-15] Ordinary graph structures are usually constructed based on spatial positions or depth sequences, and the connections between nodes are based on fixed rules. The AP algorithm can more effectively cluster nodes with similar logging response characteristics, and the generated graph structure better reflects the relationship between different diagenetic phases and the similar sedimentary characteristics of the strata.

[16] , from effectively dealing with the problem of small sample size. Summary of the invention

[0007] The technical problems to be solved by the present invention are: In order to solve the problem of small sample size and low accuracy in diagenetic phase annotation work.

[0008] The present invention provides a method for automatically labeling diagenetic facies samples by integrating affinity propagation clustering and graph convolutional neural network, comprising the following steps:

[0009] S100, data set establishment, including classification of diagenetic facies types and preprocessing of well logging curve data, wherein the well logging curve data includes well logging curve selection, missing value and outlier processing of the selected well logging curves, and sample labeling, and then labeling the well logging curve data according to the diagenetic facies type; S200, constructing an APM graph structure using affinity propagation clustering algorithm to connect nodes with similar characteristics but not adjacent to each other, in order to achieve diagenetic facies label sharing; S300, using a multi-layer graph convolutional network, combined with the affinity propagation clustering result obtained in step S200, a convolution operation is performed on the logging curve data processed in step S100, and after multiple iterations, an optimized model structure based on the graph convolution diagenetic facies annotation method is obtained.

[0010] Furthermore, in step S100, it specifically includes: S110. Summarize the diagenetic facies types, classify the diagenetic facies types according to compaction, lithology, cementation and dissolution, classify the reservoirs according to the reservoir properties, and classify the diagenetic facies types according to the reservoir types; S120, pre-processing well logging curve data, including well logging curve selection, well logging curve pre-processing and sample labeling: S121, well logging curve selection, analyze the well logging response characteristics of various diagenetic phases, use the decision tree scoring method to score the importance of each well logging curve, and select the well logging curve type with the greatest impact on diagenetic phase analysis; S122, preprocessing the logging curve data obtained in step S121, including processing missing values ​​and abnormal values; The missing values ​​of the logging curve data are predicted using the regression interpolation method. The missing value processing uses formula (1) to build a linear regression model. When interpolating, the regression coefficient is calculated using formula (2). Then, the intercept is obtained using formula (3). The missing values ​​are predicted using the linear regression model. The linear regression model is used to detect outliers. Outlier detection uses formula (4) to analyze the residuals to identify outliers. If the residual value of a data point is too large, these points are judged to be outliers. The formula is as follows: (1) (2) (3) (4) (5) in, is the response variable, is the input variable, The logging characteristic value of the i-th logging depth node, The response variable value of the i-th logging depth node, is the number of logging depth nodes, is the error term, is the regression coefficient, is the intercept, is the residual, It is the depth node of the well logging curve. The output obtained after prediction; S123, sample labeling, normalizing the well logging curve data pre-processed in step S122 to make different well logging curve data within the same measurement range; analyzing and labeling the casting thin section images and scanning electron microscope images at corresponding depths to form a data set of automatically labeled diagenetic phase samples.

[0011] Furthermore, the diagenetic facies types are divided according to compaction, lithology, cementation and dissolution, including weakly compacted and weakly cemented dissolved siltstone facies, weakly compacted and weakly cemented dissolved fine sandstone facies, medium-strong compacted and weakly cemented dissolved fine sandstone facies, medium-strong compacted and weakly cemented dissolved siltstone facies, medium-strong compacted and weakly dissolved cemented fine sandstone facies and mudstone facies; the diagenetic facies types are divided according to reservoir properties, including Class I, Class II and Class III reservoirs, Class I reservoirs correspond to weakly compacted and weakly cemented dissolved siltstone facies, weakly compacted and weakly cemented dissolved fine sandstone facies and medium-strong compacted and weakly cemented dissolved fine sandstone facies; Class II reservoirs correspond to medium-strong compacted and weakly cemented dissolved siltstone facies and medium-strong compacted and weakly dissolved cemented fine sandstone facies; Class III reservoirs correspond to mudstone facies.

[0012] Furthermore, logging curves with great influence on diagenetic phase analysis are selected, including natural gamma, sonic time difference, deep lateral resistivity, shallow lateral resistivity, natural potential and well diameter.

[0013] Furthermore, in step S200, it specifically includes: S210, read the logging curve data, select the logging curve features according to step S121, form the feature vector of each node, and generate the feature matrix, as shown in the following formula (6): (6) in, , is the depth node of the logging curve, is the number of features; S220, the depth nodes of the logging curve are sorted according to the depth, adjacent depth nodes are connected by edges, an empty undirected graph structure based on the depth order is constructed, and the corresponding feature vector is extracted; S230, clustering nodes with non-adjacent depths using affinity propagation clustering algorithm, selecting logging curve features according to step S121, classifying each logging curve depth node into a corresponding cluster, establishing connections between nodes with similar logging curve features, and then calculating the Euclidean distance between nodes and performing normalization processing to make each feature value within the same range: (7) Similarity Matrix The negative distance matrix is: (8) (9) S240. Generate an adjacency matrix based on the diagenetic phase labels of the clusters. If two nodes belong to the same cluster, the corresponding position in the adjacency matrix is ​​1, otherwise it is 0: (10) in, Indicates Diagenetic facies labels for depth nodes of well log curves; Generate the adjacency matrix according to the above steps Feature Matrix .

[0014] Further, in step S300, the first convolution layer uses the GCNConv operation, the input dimension is num_features, and the output dimension is 64; the second convolution layer uses the GCNConv operation, the input dimension is 64, and the output dimension is 32; the third convolution layer uses the GCNConv operation, the input dimension is 32, and the output dimension is 6; S310, the feature matrix and the adjacency matrix Input, the specific convolution operation is as follows: (11) (12) (13) in, is the identity matrix, is the degree of node i, It is The feature matrix of the layer is initially the input feature matrix , It is The weight matrix of the layer, is a nonlinear activation function; S320, the output layer uses the softmax function to classify the diagenetic phases: (14) (15) Among them, Z outputs the category probability distribution; S330. Update model parameters using the Adam optimizer: (16) in, are model parameters, is the learning rate, is a parameter The gradient operation, is the loss function.

[0015] Furthermore, the number of iterations is set to 2000.

[0016] The present invention discloses a system for automatically labeling diagenetic facies samples by integrating affinity propagation clustering and graph convolutional neural network. The system has a program module corresponding to the above steps, and executes the steps in the method for automatically labeling diagenetic facies samples by integrating affinity propagation clustering and graph convolutional neural network during operation.

[0017] The present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program is configured to implement the steps of a method for automatically labeling diagenetic phase samples by integrating affinity propagation clustering and graph convolutional neural network when called by a processor.

[0018] Compared with the prior art, the present invention has the following beneficial effects: (1) The graph neural network is used to annotate diagenetic facies samples and the graph structure constructed by affinity propagation clustering can capture the complex associations and spatial distribution characteristics between diagenetic facies, and combine the graph structure with the characteristics of well logging curves to improve the annotation accuracy.

[0019] (2) The graph structure constructed based on depth order and affinity propagation clustering method performs best in diagenetic facies sample annotation. Compared with the graph without constructed edges and the graph structure based only on depth order, affinity propagation clustering can more accurately capture the similarities between diagenetic facies and better express the relationship between nodes, thereby generating a more reasonable graph structure to improve the accuracy and stability of annotation.

[0020] (3) The precision of the APM-GCN method of the present invention is above 86%, the recall is above 90%, and the F1 score is above 89%. Compared with the traditional GCN, GraphSAGE, SVM and RF methods, the diagenetic facies annotation effect is better. APM-GCN first classifies the data through clustering technology, and then uses graph convolution to effectively transfer and learn features, which improves the annotation accuracy. However, the annotation accuracy is easily affected by the similarity of logging curve features.

[0021] (4) After adding different Gaussian noises, the mAA and mAP of APM-GCN have lower standard deviations and smaller fluctuations, indicating that the model has higher stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a flow chart of a method for automatically labeling diagenetic facies samples by integrating affinity propagation clustering and graph convolutional neural network in an embodiment of the present invention; Figure 2 It is a scoring diagram of a well logging curve decision tree in an embodiment of the present invention; Figure 3The thin-section images and scanning electron microscope images of the casting of the Fuyu oil layer in the Sanzhao Depression in the embodiment of the present invention are, among which (a), (b), and (c) are all thin-section images of the casting of the Fuyu oil layer in the Sanzhao Depression, and (d), (e), and (f) are all scanning electron microscope images of the casting of the Fuyu oil layer in the Sanzhao Depression; Figure 4 This is a structural diagram of the APM network model in an embodiment of the present invention; Figure 5 This is a structural diagram of a graph convolutional network model in an embodiment of the present invention; Figure 6 This is a graph of loss iteration results during the model training process in an embodiment of the present invention; Figure 7 A comparison chart of the Matthews coefficient, blind well accuracy and test accuracy of the map construction method in an embodiment of the present invention; Figure 8 It is a line graph of the accuracy of the k-means clustering method, the Gaussian mixture clustering method, the aggregation clustering method and the affinity propagation clustering method in the embodiment of the present invention; Fig. 9 : The confusion matrix result diagrams in the embodiments of the present invention, wherein (a) is the confusion matrix result diagram of the APM-GCN method of the present invention, (b) is the confusion matrix result diagram of the GCN method, (c) is the confusion matrix result diagram of the GraphSAGE method, (d) is the confusion matrix result diagram of the RF method, and (e) is the confusion matrix result diagram of the SVM method; Fig.10 The figures are the calculation results of Precision, Recall and F1 scores of the APM-GCN method, GCN method, GraphSAGE method, RF method and SVM method of the present invention in the embodiments of the present invention, wherein (a) is a line graph of the Precision calculation results of the above five methods, (b) is a line graph of the Recall calculation results of the above five methods, and (c) is a line graph of the F1 score calculation results of the above five methods; Fig.11 Graph showing the relationship between different data amounts and mAA and mAP in an embodiment of the present invention; Fig.12 The comparison diagram of mAA and mAP calculation results of various diagenetic facies annotation effects with different data amounts in the embodiment of the present invention, wherein (a) is a diagram of the calculation results of Accuracy value and mAA value in various diagenetic facies annotation effects, and (b) is a diagram of the calculation results of AP' value and mAP value in various diagenetic facies annotation effects; Fig.13 This is a graph of mAA and mAP calculation results after adding Gaussian noise in an embodiment of the present invention. DETAILED DESCRIPTION

[0023] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0024] Combination Figure 1 As shown in the figure, the present invention proposes an automatic labeling method for petrofacies samples using affinity propagation clustering and graph convolutional neural network (APM-GCN). First, the diagenetic facies types are divided and key well logging curves are selected for data preprocessing, and the well logging curve data are labeled as different diagenetic facies types; secondly, the AP algorithm is used to construct a graph structure to connect nodes with similar features but non-adjacent to achieve diagenetic facies label sharing; then, a multi-layer graph convolutional network is used, combined with the AP clustering results, to perform convolution operations on the processed well logging curve data to improve the classification accuracy and efficiency of automatic diagenetic facies labeling; finally, the effectiveness of the method proposed in the present invention is verified through comparative experiments.

[0025] Specific implementation plan 1: Combine Figures 1 to 6 As shown, the present invention provides a method for automatically labeling diagenetic facies samples by integrating affinity propagation clustering and graph convolutional neural network, comprising the following steps: S100, data set establishment, data set establishment includes two steps: summarizing diagenetic facies types and preprocessing well logging curve data. S110, summarize the diagenetic facies type, take the Fuyu oil layer in the Sanzhao Sag of the Songliao Basin as the target area, the reservoir has the characteristics of fine grain size and high mud content

[17] The diagenetic facies types are classified according to compaction, lithology, cementation and dissolution. [18,19], including: Weakly compacted weakly cemented dissolved siltstone phase (Wip), Weakly compacted weakly cemented dissolved fine sandstone phase (Wap), Medium to strong compaction of weakly cemented dissolved fine sandstone phase (Map), Medium to strong compaction of weakly cemented dissolved siltstone phase (Mip), Medium to strong compaction of weakly dissolved colluvial fine sandstone phase (Msap) and Mudstone phase (Mp). According to the reservoir properties, the above diagenetic facies types are divided into Class I, Class II and Class III reservoirs. Class I reservoirs correspond to Wip, Wap and Map; Class II reservoirs correspond to Mip and Msap; Class III reservoirs correspond to Mp.

[0026] S120, pre-processing the logging curve data, including three steps: logging curve selection, logging curve pre-processing and sample labeling. S121. Well logging curve selection: Analyze the well logging response characteristics of various diagenetic phases

[20] , the decision tree scoring method is used to score the importance of each logging curve, such as Figure 2 As shown in the figure, the well logging curves with the largest scores on the diagenetic phase analysis features (FeatureImportances) were selected, including natural gamma (GR), acoustic time difference (AC), deep lateral resistivity (LLD), shallow lateral resistivity (LLS), natural potential (SP) and wellbore (CAL) [21,22] ; S122, preprocessing the logging curve data obtained in step S121, including processing missing values ​​and abnormal values; The missing values ​​of the logging curve data are predicted using the regression interpolation method, and then the linear regression model is used to detect outliers, and the missing values ​​and outliers are processed. The missing value processing uses formula (1) to build a linear regression model, and when interpolating, the regression coefficient is calculated using formula (2). Then, the intercept is obtained using formula (3), and the missing values ​​are predicted using the linear regression model. The outlier detection uses formula (4) to analyze the residuals to identify outliers. If the residual values ​​of some data points are too large, that is, the residual values ​​are greater than 3.5, these points are judged to be outliers. The formula is as follows: (1) (2) (3) (4) (5) in, is the response variable, is the input variable, The logging characteristic value of the i-th logging depth node, The response variable value of the i-th logging depth node, is the number of logging depth nodes, is the error term, is the regression coefficient, is the intercept, is the residual, It is the depth node of the well logging curve. The output obtained after prediction; S123, sample labeling, normalizing the well logging curve data pre-processed in step S122 to ensure that different well logging curve data are in the same measurement range, and the node depth interval is set to 0.05 meters. Finally, according to the casting thin section image and scanning electron microscope image (such as Figure 3 Detailed analysis and annotation of the samples (as shown) are performed to form a dataset of automatic annotation of diagenetic facies samples; S200, using AP algorithm to build APM graph structure, connecting nodes with similar characteristics but non-adjacent nodes, to achieve diagenetic facies label sharing; the design of APM graph structure can not only form connections between logging curve depth nodes with similar logging curve characteristics but non-adjacent depths, but also ensure that these logging curve depth nodes share the same diagenetic facies label. APM network model such as Figure 4 As shown in the figure, the construction process of the APM graph structure specifically includes the following four steps: S210, reading the logging curve data, forming a feature vector for each node according to the six features of 'GR', 'AC', 'LLD', 'LLS', 'SP' and 'CAL' obtained in step S121, and generating a feature matrix, as shown in the following formula 6: (6) in, , is the depth node of the logging curve, is the number of features; S220, the depth nodes of the well logging curve are sorted according to the depth (DEPT). Due to the continuity of the sedimentary process, each depth node at a certain depth of the formation has diagenetic phase similarity with other depth nodes adjacent in space. The certain depth here refers to the depth range of a specific formation thickness or layer segment that has the performance of the sedimentary process. Therefore, adjacent depth nodes are connected by edges to construct an empty undirected graph structure based on the depth order, and the corresponding feature vector is extracted; S230, clustering nodes with non-adjacent depths using the AP clustering algorithm, classifying each well logging curve depth node into a corresponding cluster according to the six features of 'GR', 'AC', 'LLD', 'LLS', 'SP' and 'CAL', and establishing connections between nodes with similar well logging curve features. Calculate the Euclidean distance between nodes and perform normalization to ensure that each feature value is in the same range.

[0027] (7) The AP algorithm performs clustering through the similarity matrix and outputs the clustering label of each logging curve depth node. Therefore, the similarity matrix Take the negative distance matrix: (8) (9) S240. Generate an adjacency matrix based on the diagenetic phase labels of the clusters. If two nodes belong to the same cluster, the corresponding position in the adjacency matrix is ​​1, otherwise it is 0: (10) in, Indicates Diagenetic facies labels for depth nodes of well log curves; Generate the adjacency matrix according to the above steps Feature Matrix , laying the foundation for the subsequent automatic labeling task of diagenetic phase samples.

[0028] S300, using a multi-layer graph convolutional network, combined with AP clustering results, convolution operation is performed on the processed logging curve data; the model structure based on the graph convolution diagenetic facies annotation method is as follows Figure 5 As shown, the first convolution layer uses GCNConv operation, input dimension: num_features, output dimension: 64; the second convolution layer uses GCNConv operation, input dimension: 64, output dimension: 32; the third convolution layer uses GCNConv operation, input dimension: 32, output dimension: 16. Specifically include: S310, the feature matrix and the adjacency matrix Input, the specific convolution operation is as follows: (11) (12) (13) in, is the identity matrix, is the degree of node i, It is The feature matrix of the layer is initially the input feature matrix , It is The weight matrix of the layer, is a non-linear activation function (ReLU).

[0029] S320, the output layer uses the softmax function to classify the diagenetic phases, and Z outputs the category probability distribution: (14) (15) S330. Update model parameters using the Adam optimizer: (16) in, are model parameters, is the learning rate, is a parameter The gradient operation, is the loss function.

[0030] The loss iteration during model training is as follows Figure 6 As shown in the figure, the number of iterations is set to 2000. It can be observed that after about 500 iterations, the error of the model gradually approaches zero and reaches the preset training accuracy, which indicates that the model can fit the training data well after sufficient training.

[0031] Specific implementation scheme two: The present invention provides a system for automatic labeling of diagenetic phase samples that integrates affinity propagation clustering and graph convolutional neural network. The system has a program module corresponding to the above steps, and executes the steps in the above-mentioned method for automatic labeling of diagenetic phase samples that integrates affinity propagation clustering and graph convolutional neural network during operation.

[0032] The other combinations and connection relationships of this embodiment are the same as those of the first embodiment.

[0033] Specific implementation scheme three: The present invention provides a computer-readable storage medium, which stores a computer program, and the computer program is configured to implement the steps of a method for automatic labeling of diagenetic phase samples by integrating affinity propagation clustering and graph convolutional neural network when called by a processor.

[0034] The other combinations and connection relationships of this embodiment are the same as those of the first embodiment.

[0035] Simulation experiment To verify the effectiveness of the proposed method, the present invention conducts graph structure comparison evaluation experiments, clustering method comparison evaluation experiments, model performance evaluation experiments, and model stability evaluation experiments. The specific parameters of the experimental equipment are as follows: CPU is Intel XeonSilver 4210R, memory is 64G, GPU is RTX 6000; operating system is Ubuntu 20.04.3; experimental framework is PyTorch 1.8.0.

[0036] 1. Graph structure comparison and evaluation test In order to compare the performance of graph structures in the diagenetic facies sample labeling task, the present invention uses the Matthews correlation coefficient (MCC) and the precision of the diagenetic facies labeling results as the main evaluation indicators.

[0037] MCC: A commonly used indicator for evaluating the performance of classification models. The calculation formula is: (17) Precision: The proportion of correctly labeled diagenetic facies types identified, calculated as: (18) Among them, TP represents the number of samples correctly labeled as positive, TN represents the number of samples correctly labeled as negative, FP represents the number of samples incorrectly labeled as positive, and FN represents the number of samples incorrectly labeled as negative.

[0038] The experiment used 8000 well logging curve depth nodes, and designed three graph structure construction methods to analyze their effects on diagenetic facies annotation: the first is no edge construction (NE method), directly using the original well logging curve as input data; the second is a graph structure based on deep depth sequence (DS method); the third is a graph structure based on depth sequence and clustering (SC method). The same training data and test data were used in the experiment to calculate the MCC and Precision of different graph construction methods. To further verify the generalization ability of the three graph structures on the model, a blind well test was conducted using 2101 well logging curve depth nodes of the F361 well. The well logging curve data of F361 was used for graph construction, but did not participate in the training of the model. Finally, the precision of the diagenetic facies type annotation results was calculated.

[0039] 2. Clustering method comparison and evaluation experiment The clustering method comparison experiment uses the precision of diagenetic phase annotation as the evaluation index, where the meaning and calculation method of Precision are the same as those in Experiment 1. The experimental process is based on the selection of the graph construction method with the best annotation effect in Experiment 1. This experiment uses the same training data and test data to compare the selection of four different clustering methods, namely k-means clustering, Gaussian mixture clustering, aggregation clustering and affinity propagation clustering, and compares the precision of diagenetic phase annotation to evaluate the impact of clustering methods on annotation effect.

[0040] 3. Model performance evaluation test The present invention establishes a confusion matrix to calculate the precision, recall, and F1 score of the annotation results of each diagenetic phase type.

[23] As an evaluation indicator, we conduct model performance evaluation experiments.

[0041] Recall represents the proportion of the correct number in the actual component, which can be expressed by formula (19): (19) (20) The experimental process is based on the construction method of the graph in Experiment 2 to further verify the effects of the APM-GCN and other deep learning algorithms in diagenetic facies annotation. First, the standardized (i.e., pre-processed) well logging curve data is randomly split in a ratio of 6:4. Then, the graph structure is constructed based on the deep sequence and AP clustering method, and the standardized well logging curve data is used as the feature matrix. Finally, the validation set is used to verify the trained graph convolutional neural network GCN, GraphSAGE (graph sampling and aggregation), SVM (support vector machine), and RF (random forest) for diagenetic facies sample annotation to evaluate the effect of the APM-GCN algorithm.

[0042] 4. Model stability evaluation test This experiment uses the mean Average-Accuracy (mA)

[24] and mAP (mean Average-Precision, mAP)

[25] As an evaluation indicator.

[0043] mAA represents the average accuracy of multiple diagenetic facies type identification. Accuracy represents the proportion of correctly identified samples in the total number. The formula is as follows: TP represents the number of correctly identified diagenetic facies, and N represents the total number of diagenetic facies samples.

[0044] (twenty one) mAP represents the average of the average precision (AP') of multiple diagenetic facies types. AP' represents the area under the curve of Precision and Recall. The definitions of Precision and Recall are the same as those in Experiment 3 above.

[0045] The experimental process is divided into two parts: (1) The effect of different data volumes on the annotation effect. Different numbers of data sets were selected from different wells in the same area, namely 2000, 4000, 6000, 8000 and 10000 depth nodes. The Accuracy and AP' of the APM-GCN algorithm in the annotation of each diagenetic phase were calculated, and the mA value and mAP value were calculated.

[0046] (2) To further verify the stability of the model, Gaussian noise was added to the dataset with the best annotation effect. The model was trained and evaluated on the well logging curve dataset with added noise. The same mAA and mAP indicators were used for performance evaluation. The changes in model performance before and after adding noise were compared to analyze the stability of the model.

[0047] Experimental Results 1. Graph structure comparison experimental results The Matthews coefficient and accuracy of the graph construction method are as follows Figure 7 As shown in the figure, when the APM-GCN model of the present invention only inputs the original data, the test data accuracy can reach 0.8. Although the correct annotation rate of the diagenetic phase sample is 10% lower than the input constructed graph structure, it is still higher than the SVM and RF annotation accuracy, proving that the use of graph structure data can significantly improve the accuracy of diagenetic phase annotation. The DS method is based on the graph structure of depth order, considering the geological factors in the formation process of diagenetic phase, and the test accuracy is 0.91. The SC method combines the depth order and clustering method, and the test accuracy is 0.9, which is slightly lower than DS. The reason for this result is mainly because the training and test data are randomly selected, so there is a high similarity between the training and test data.

[0048] Depend on Figure 7 It can be seen that in the F361 well test, the three graph structure annotation accuracies were 0.62, 0.7, and 0.75, respectively. The results show that SC and DS have similar accuracy in the diagenetic facies labeling task, but the generalization ability of the SC training model is significantly better than that of DS. Therefore, the SC method shows good accuracy and generalization ability in the diagenetic facies labeling.

[0049] 2. Clustering method comparison experimental results The accuracy of the four cluster selection methods is as follows Figure 8 As shown in the figure, affinity propagation clustering performs best in diagenetic phase labeling accuracy, reaching 0.90; in contrast, the labeling accuracy of Gaussian mixture clustering and clustering clustering are 0.87 and 0.86 respectively, and the accuracy of k-means clustering is the lowest, only 0.85. This shows that although Gaussian mixture clustering and clustering clustering can better capture the potential structure in the data, they are still slightly inferior to affinity propagation clustering in processing well logging curve data. Although the k-means clustering method has great advantages in computing speed and implementation, its effect is not as good as other methods. The calculation results show that in the test data, the model based on affinity propagation clustering algorithm has a higher accuracy than other clustering algorithms.

[0050] 3. Experimental results of model performance evaluation The confusion matrix results are as follows Fig. 9 The calculation results of Precision, Recall and F1 scores are shown in Fig.10 As shown. Fig. 9It can be seen that the APM-GCN method has the best effect in diagenetic facies labeling, followed by the GCN and GraphSAGE methods, and the SVM and RF methods perform the worst. This is because APM-GCN, GCN and GraphSAGE use graph structure data containing the relationship between samples, while SVM and RF only use tabular data of independent samples, which proves that the use of graph structure data can significantly improve the accuracy of diagenetic facies labeling. Compared with the GCN method and GraphSAGE method, the APM-GCN model performs better in the accuracy of diagenetic facies sample labeling and can better realize the diagenetic facies sample labeling of tight reservoirs.

[0051] In addition, the experimental results show that Wip has the highest number of correct classifications, but a small number of them are misclassified into the Wap category. There are more misclassifications in the Wap category, which are misclassified into Wip and Mip. Mip category is misclassified into Wap and Map, and the recognition performance of Map category is good, but there are some misclassifications into Msap category. Msap category recognition performance is good, but there are cases of misclassification into Map category. Mp category recognition performance is poor, with a large number of misclassifications, especially misclassification into Msap category. The reason for the misjudgment may be that the response characteristics of the logging curve are not only affected by the diagenetic phase, but also by the geology and drilling operations of clay minerals, pore fluids, and mud pressure.

[0052] Depend on Fig.10 It can be seen that the Precision, Recall and F1 score values ​​of the APM-GCN method are higher than those of the GraphSAGE method, SVM method and RF method, while the GCN method and GraphSAGE method are higher than the SVM method and RF method. The four methods used in the Mp category all have good results, which indicates that the logging response characteristics of this diagenetic phase are significantly different from those of other diagenetic phases. Fig. 9 (a) It can be seen that there are many misclassifications in the Wap category and the annotation accuracy is low. The main reason is that Wap and Msap have similar logging response characteristics, so there is a phenomenon that Msap is misjudged as Wap. The same phenomenon also exists in Wip and Mip. The APM-GCN method has an annotation accuracy of more than 86% for various diagenetic phases, indicating that it has a good annotation effect. Fig. 9 In (b), the recall of the APM-GCN method for various diagenetic phases is above 90%, indicating that it has a good recall. Fig. 9 In (c), the APM-GCN method has the highest F1 score, which is above 89%, indicating that this method has a good effect on the labeling of various diagenetic facies and basically meets the accuracy requirements of diagenetic facies labeling.

[0053] 4. Experimental results of model stability evaluation (1) The mAA and mAP calculation results of different data volume annotation effects are as follows Fig.11 The relationship between different data amounts and mAA and mAP is shown in Fig.12 shown.

[0054] Depend on Fig.11 It can be seen that as the data volume increases from 2000 to 10000, the accuracy of diagenetic facies annotation gradually increases, but the growth rate slows down at 6000 data volumes, indicating that increasing the data volume can effectively improve the accuracy of diagenetic facies annotation, but after reaching a certain data volume, the effect will gradually decrease. Fig.12 It can be seen that at 8000 depth nodes, the annotation accuracy of various diagenetic facies is the best, about 90%. With the increase of data volume, the Accuracy value and mAA value in the annotation effect of various diagenetic facies show an overall upward trend, and the accuracy of most diagenetic facies annotations has improved. The mAP value increased from 0.72 (2000 data volume) to 0.81 (10000 data volume), but the increase tends to decrease when the data volume is large. However, the increase gradually decreases when the data volume is large. Since 8000 depth nodes is the best annotation result, this data volume is used in the accuracy experiment.

[0055] (2) The mAA and mAP calculation results after adding Gaussian noise are as follows Fig.13 As shown. After adding Gaussian noise, mAA and mAP values ​​showed small fluctuations, with average values ​​of 89.99 and 81.935, and standard deviations of 0.76 and 0.837, respectively, indicating that the model performance is stable. The maximum and minimum mAA values ​​are 90.85 and 88.48, respectively, with a fluctuation range of 2.37; the maximum and minimum mAP values ​​are 83.15 and 80.84, with a fluctuation range of 2.31. Overall, although the addition of Gaussian noise has a certain impact on the mAA and mAP values ​​of the model, the overall fluctuation is small, proving that the model exhibits strong stability and robustness.

[0056] Although the present invention is disclosed as above, the protection scope of the present invention is not limited thereto. Those skilled in the art may make various changes and modifications without departing from the spirit and scope of the present invention, and these changes and modifications will fall within the protection scope of the present invention.

[0057] [1] De Ros, Luiz Fernando, and Karin Goldberg. "Reservoir petrofacies: a tool for quality characterization and prediction." AAPG, annual convention and exhibition, Long Beach, abstracts volume . 2007. [2]Watney, WL, et al. "Memoir 71, Chapter 6: PetrofaciesAnalysis--A Petrophysical Tool for Geologic / Engineering ReservoirCharacterization." (1999): 73-90. [3] [1] Liu Jingjing. Prediction of high-quality tight sandstone reservoirs based on artificial intelligence[D]. Chang'an University, 2023. DOI: 10.26976 / d.cnki.gchau.2022.000152. [4] Zou Caineng, Tao Shizhen, Zhou Hui, et al. Formation, classification and quantitative evaluation of diagenetic facies[J]. Petroleum Exploration and Development, 2008(05):526-540. [5] Cui, Yufeng, et al. "Prediction of diagenetic facies using welllogs–A case study from the upper Triassic Yanchang Formation, Ordos Basin,China." Marine and Petroleum Geology 81 (2017): 50-65. [6] Chapelle, Olivier, Bernhard Scholkopf, and Alexander Zien. "Semi-supervised learning (chapelle, o. et al., eds.; 2006) [book reviews]." IEEETransactions on Neural Networks 20.3 (2009): 542-542. [7]Yann, L., and Misra Ishan. "Self-supervised learning: The darkmatter of intelligence." Meta AI Blog (2021). [8] Krizhevsky, Alex, Ilya Sutskever, and Geoffrey E. Hinton. "Imagenet classification with deep convolutional neural networks." Advances in neural information processing systems 25 (2012). [9] Pattnaik, Sonali, et al. "Automatic carbonate rock faciesidentification with deep learning." SPE Annual Technical Conference andExhibition?. SPE, 2020.

[10] Wu, Zonghan, et al. "A comprehensive survey on graph neuralnetworks." IEEE transactions on neural networks and learning systems 32.1(2020): 4-24.

[11] Kipf, Thomas N., and Max Welling. "Semi-supervised classificationwith graph convolutional networks." arxiv preprint arxiv:1609.02907 (2016).

[12] Lewinsohn, Daniel P., et al. "Consensus label propagation withgraph convolutional networks for single-cell RNA sequencing cell typeannotation." Bioinformatics 39.6 (2023): btad360.

[13] Lu, Guoqing, et al. "Lithology identification using graph neuralnetwork in continental shale oil reservoirs: A case study in Mahu Sag,Junggar Basin, Western China." Marine and Petroleum Geology 150 (2023):106168.

[14] Frey, Brendan J., and Delbert Dueck. "Clustering by passingmessages between data points." science 315.5814 (2007): 972-976.

[15] Dueck, Delbert. Affinity propagation: clustering data by passing messages . Toronto, ON, Canada: University of Toronto, 2009.

[16] Givoni, Inmar E., and Brendan J. Frey. "A binary variable model for affinity propagation." Neural computation 21.6 (2009): 1589-1600.

[17] Liu, YX, et al. "Diagenesis and pore evolution of reservoir of the Member4 of Lower Cretaceous Quantou Formation in Sanzhao Sag, northern Songliao Basin." J Palaeogeogr 12.4 (2010): 480-7.

[18] Ding Sheng, Zhong Siying, Gao Guoqiang, et al. Quantitative evaluation of diagenetic facies of low permeability reservoirs by combining well logging and geology[J]. Journal of Southwest Petroleum University (Natural Science Edition), 2012, 34(04): 83-87.

[19] Tao Liu, Zongbao Liu, Kejia Zhang, Chunsheng Li, Yan Zhang, Zihao Mu, Fang Liu, Xiaowen Liu, Mengning Mu. Research on the generation and annotation method of thin section images of tight oil reservoir based on deep learning. Scientific Reports, 2024.

[20] Yang Wei, et al. "Carbonate diagenetic facies research method and its application: A case study of the oolitic reservoir of the Feixianguan Formation in the northern margin of the Yangtze Block." Acta Petrologica Sinica 27.3 (2011): 749-756.

[21] Zhao, M., Cao, G., Huang, X., Yang, L. (2022). HybridTransformer-CNN for Real Image Denoising. IEEE Signal Processing Letters, 29,1252-1256.

[22] Zhang, **yu, William Ambrose, and Wei **e. "Applyingconvolutional neural networks to identify lithofacies of large-n cores fromthe Permian Basin and Gulf of Mexico: The importance of the quantity andquality of training data." Marine and Petroleum Geology 133 (2021): 105307.

[23] Maria Navin, J. R., Pankaja, R. (2016). Performance analysis oftext classification algorithms using confusion matrix. Int. J. Eng. Tech.Res. IJETR, 6, 75-78.

[24] Antariksa, G., Muammar, R., Lee, J. (2022). Performanceevaluation of machine learning-based classification with rock-physicsanalysis of geological lithofacies in Tarakan Basin, Indonesia. Journal ofPetroleum Science and Engineering, 208, 109250.

[25] N. Pereira. (2022). PereiraASLNet: ASL letter recognition withYOLOX taking Mean Average Precision and Inference Time considerations. In2022 2nd International Conference on Artificial Intelligence and SignalProcessing (AISP), Vijayawada, India, 2022, pp. 1-6, doi: 10.1109 / AISP53593.2022.9760665.

Claims

1. A method for automatic annotation of diagenetic facies samples by integrating affinity propagation clustering and graph convolutional neural network, characterized in that: The following steps are involved: S100, data set establishment, including classification of diagenetic facies types and preprocessing of well logging curve data, wherein the well logging curve data includes well logging curve selection, missing value and outlier processing of the selected well logging curves, and sample labeling, and then labeling the well logging curve data according to the diagenetic facies type; S200, constructing an APM graph structure using affinity propagation clustering algorithm to connect nodes with similar characteristics but not adjacent to each other, in order to achieve diagenetic facies label sharing; S300, using a multi-layer graph convolutional network, combined with the affinity propagation clustering result obtained in step S200, a convolution operation is performed on the logging curve data processed in step S100, and after multiple iterations, an optimized model structure based on the graph convolution diagenetic facies annotation method is obtained.

2. The automatic annotation method for diagenetic facies samples integrating affinity propagation clustering and graph convolutional neural network according to claim 1 is characterized in that: In step S100, it specifically includes: S110. Summarize the diagenetic facies types, classify the diagenetic facies types according to compaction, lithology, cementation and dissolution, classify the reservoirs according to the reservoir properties, and classify the diagenetic facies types according to the reservoir types; S120, pre-processing well logging curve data, including well logging curve selection, well logging curve pre-processing and sample labeling: S121, well logging curve selection, analyze the well logging response characteristics of various diagenetic phases, use the decision tree scoring method to score the importance of each well logging curve, and select the well logging curve type with the greatest impact on diagenetic phase analysis; S122, preprocessing the logging curve data obtained in step S121, including processing missing values ​​and abnormal values; The missing values ​​of the logging curve data are predicted using the regression interpolation method. The missing value processing uses formula (1) to build a linear regression model. When interpolating, the regression coefficient is calculated using formula (2). Then, the intercept is obtained using formula (3). The missing values ​​are predicted using the linear regression model. The linear regression model is used to detect outliers. Outlier detection uses formula (4) to analyze the residuals to identify outliers. If the residual value of a data point is too large, these points are judged to be outliers. The formula is as follows: （1） （2） （3） （4） （5） in, is the response variable, is the input variable, The logging characteristic value of the i-th logging depth node, The response variable value of the i-th logging depth node, is the number of logging depth nodes, is the error term, is the regression coefficient, is the intercept, is the residual, It is the depth node of the well logging curve. The output obtained after prediction; S123, sample labeling, normalizing the well logging curve data pre-processed in step S122 to make different well logging curve data within the same measurement range; analyzing and labeling the casting thin section images and scanning electron microscope images at corresponding depths to form a data set of automatically labeled diagenetic phase samples.

3. The automatic annotation method of diagenetic facies samples integrating affinity propagation clustering and graph convolutional neural network according to claim 2 is characterized by: The diagenetic facies types are divided according to compaction, lithology, cementation and dissolution, including weak compaction and weak cementation dissolved siltstone facies, weak compaction and weak cementation dissolved fine sandstone facies, medium-strong compaction and weak cementation dissolved fine sandstone facies, medium-strong compaction and weak cementation dissolved siltstone facies, medium-strong compaction and weak dissolution cementation fine sandstone facies and mudstone facies; According to reservoir properties, diagenetic facies types are divided into Type I, Type II and Type III reservoirs. Type I reservoirs correspond to weakly compacted and weakly cemented dissolved siltstone facies, weakly compacted and weakly cemented dissolved fine sandstone facies and medium-strong compacted and weakly cemented dissolved fine sandstone facies; Type II reservoirs correspond to medium-strong compacted and weakly cemented dissolved siltstone facies and medium-strong compacted and weakly dissolved and cemented fine sandstone facies; Type III reservoirs correspond to mudstone facies.

4. The automatic annotation method of diagenetic facies samples integrating affinity propagation clustering and graph convolutional neural network according to claim 3 is characterized by: Well logging curves with great influence on diagenetic phase analysis are selected, including natural gamma, sonic time difference, deep lateral resistivity, shallow lateral resistivity, natural potential and well diameter.

5. The method for automatic annotation of diagenetic facies samples by integrating affinity propagation clustering and graph convolutional neural network according to claim 2 is characterized in that: In step S200, it specifically includes: S210, read the logging curve data, select the logging curve features according to step S121, form the feature vector of each node, and generate the feature matrix, as shown in the following formula (6): （6） in, , is the depth node of the logging curve, is the number of features; S220, the depth nodes of the logging curve are sorted according to the depth, adjacent depth nodes are connected by edges, an empty undirected graph structure based on the depth order is constructed, and the corresponding feature vector is extracted; S230, clustering nodes with non-adjacent depths using affinity propagation clustering algorithm, selecting logging curve features according to step S121, classifying each logging curve depth node into a corresponding cluster, establishing connections between nodes with similar logging curve features, and then calculating the Euclidean distance between nodes and performing normalization processing to make each feature value within the same range: （7） Similarity Matrix The negative distance matrix is: （8） （9） S240. Generate an adjacency matrix based on the diagenetic phase labels of the clusters. If two nodes belong to the same cluster, the corresponding position in the adjacency matrix is ​​1, otherwise it is 0: （10） in, Indicates Diagenetic facies labels for depth nodes of well logs; Generate the adjacency matrix according to the above steps Feature Matrix .

6. The automatic annotation method of diagenetic facies samples integrating affinity propagation clustering and graph convolutional neural network according to claim 5 is characterized by: In step S300, the first convolution layer uses the GCNConv operation, with an input dimension of num_features and an output dimension of 64; the second convolution layer uses the GCNConv operation, with an input dimension of 64 and an output dimension of 32; the third convolution layer uses the GCNConv operation, with an input dimension of 32 and an output dimension of 6; S310, the feature matrix and the adjacency matrix Input, the specific convolution operation is as follows: （11） （12） （13） in, is the identity matrix, is the degree of node i, It is The feature matrix of the layer is initially the input feature matrix , It is The weight matrix of the layer, is a nonlinear activation function; S320, the output layer uses the softmax function to classify the diagenetic phases: （14） （15） Among them, Z outputs the category probability distribution; S330. Update model parameters using the Adam optimizer: （16） in, are model parameters, is the learning rate, is a parameter The gradient operation, is the loss function.

7. The method for automatic annotation of diagenetic facies samples by integrating affinity propagation clustering and graph convolutional neural network according to claim 6 is characterized by: Set the number of iterations to 2000.

8. An automatic annotation system for diagenetic facies samples integrating affinity propagation clustering and graph convolutional neural network, characterized by: The system has a program module corresponding to the steps of any one of claims 1 to 7 above, and executes the steps in the above-mentioned method for automatic labeling of diagenetic phase samples by integrating affinity propagation clustering and graph convolutional neural network when running.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program is configured to implement the steps of the automatic annotation method of diagenetic phase samples integrating affinity propagation clustering and graph convolutional neural network described in any one of claims 1 to 7 when called by a processor.

Images