Abnormal formation pore pressure cause mechanism discrimination method, device and equipment

Abstract

The invention provides an abnormal formation pore pressure formation mechanism judgment method and device, computer equipment and a storage medium. The method comprises the steps that logging data of a target well are obtained, and an abnormal pressure section of formation pore pressure is determined based on the logging data; clustering and grouping the logging data corresponding to the abnormal pressure section according to the similarity to obtain a plurality of initial sample data groups; establishing a wave velocity density cross-plot based on the logging data, and analyzing the abnormal pressure cause mechanism corresponding to each initial sample data group by using the wave velocity density cross-plot so as to construct a sample set of the abnormal pressure cause mechanism; performing optimization training on an initial discrimination model of the abnormal pressure cause mechanism by using the sample set to obtain an abnormal pressure cause mechanism discrimination model; and inputting the logging data of the target layer into the abnormal pressure cause mechanism discrimination model to discriminate the abnormal pressure cause mechanism. According to the invention, the formation mechanism of the abnormal formation pore pressure can be automatically discriminated.

Description

Technical Field

[0001] The present invention relates to the technical field of well logging engineering, and in particular to a method, device, computer equipment and storage medium for distinguishing the formation mechanism of abnormal formation pore pressure. Background Art

[0002] Formation pore pressure refers to the pressure of fluids (water, oil, gas) in the pores or fractures of a formation. Accurately identifying and calculating abnormally high pressure (overpressure) is of great significance for studying oil and gas migration and accumulation and optimizing drilling engineering design. Many scholars at home and abroad have summarized more than ten causes of abnormally high pressure and proposed a method for comprehensively judging the causes of abnormally high pressure using logging curve comparison and wave velocity density crossplot. These methods can determine the depth of the formation and distinguish the main causes of abnormally high pressure in sedimentary rock formations, such as undercompaction, fluid expansion, and tectonic compression. However, these methods rely on expert experience in actual application and the identification methods are complex, making it impossible to quickly and effectively identify the mechanism of abnormal pressure formation. Summary of the Invention

[0003] Based on this, it is necessary to provide a method, device, computer equipment and storage medium for determining the cause of abnormal formation pore pressure in response to the above technical problems.

[0004] A method for determining the formation mechanism of abnormal formation pore pressure includes:

[0005] Acquire logging data of the entire target well section, and determine the abnormal pressure section of the target well formation pore pressure based on the logging data; wherein the logging data includes well depth, compressional wave velocity, and density;

[0006] Clustering and grouping the well logging data corresponding to the abnormal pressure section according to similarity to obtain multiple initial sample data groups;

[0007] establishing a wave velocity density cross plot based on the well logging data, and using the wave velocity density cross plot to analyze the abnormal pressure formation mechanism corresponding to each initial sample data group, thereby constructing a sample set of the abnormal pressure formation mechanism;

[0008] Utilizing the sample set of the abnormal pressure formation mechanism, optimizing and training the initial discrimination model of the abnormal pressure formation mechanism to obtain the discrimination model of the abnormal pressure formation mechanism;

[0009] The well logging data of the target layer is obtained, and the well logging data of the target layer is input into the abnormal pressure formation mechanism discrimination model to discriminate the abnormal pressure formation mechanism.

[0010] In one embodiment, the step of establishing a velocity-density crossplot based on the well logging data comprises:

[0011] Determine the normal compaction section and abnormal pressure section of the target well formation pore pressure based on the well logging data;

[0012] Based on the coordinate system of the longitudinal wave velocity and density, the logging data of the normal compaction section is fitted to generate a wave velocity density curve of the normal compaction section; based on the logging data of the abnormal pressure section, an initial wave velocity density cross plot of the abnormal pressure section is generated; and the wave velocity density curve of the normal compaction section is synthesized with the initial wave velocity density cross plot of the abnormal pressure section to obtain the wave velocity density cross plot.

[0013] In one embodiment, the step of analyzing the abnormal pressure formation mechanisms corresponding to each initial sample data group using the wave velocity density crossplot to construct a sample set of abnormal pressure formation mechanisms includes:

[0014] Based on the position of the initial sample data group relative to the wave velocity density curve in the wave velocity density cross-plot, the abnormal pressure formation mechanism corresponding to each initial sample data group is determined, thereby constructing a sample set of the abnormal pressure formation mechanism.

[0015] In one embodiment, the step of fitting the logging data of the normal compaction section based on the coordinate system of the longitudinal wave velocity and density to generate the velocity density curve of the normal compaction section includes:

[0016] Based on the coordinate system of the longitudinal wave velocity and density, the well logging data of the normal compaction section is fitted using the Gardner equation to generate the velocity density curve of the normal compaction section.

[0017] In one embodiment, the step of optimizing and training the initial discrimination model of the abnormal pressure formation mechanism using the sample set of the abnormal pressure formation mechanism to obtain the discrimination model of the abnormal pressure formation mechanism includes:

[0018] inputting the logging data of the sample set into an initial discrimination model for abnormal pressure causal mechanism according to the type of abnormal pressure causal mechanism to identify the abnormal pressure causal mechanism, obtaining a predicted abnormal pressure causal mechanism, and comparing the predicted abnormal pressure causal mechanism with the abnormal pressure causal mechanism to obtain a model identification accuracy rate;

[0019] The initial discrimination model of the abnormal pressure causal mechanism is optimized and trained by a Bayesian optimization algorithm based on a Gaussian process until the accuracy rate no longer increases, and the parameter combination corresponding to the highest accuracy rate is determined as the final parameter combination. The discrimination model of the abnormal pressure causal mechanism is obtained based on the final parameter combination.

[0020] In one embodiment, the step of establishing the initial discrimination model of the abnormal pressure formation mechanism includes:

[0021] The decision tree algorithm is selected, gradient boosting and histogram technology are combined, and an initial abnormal pressure mechanism determination model is determined based on a LightGBM algorithm.

[0022] In one of the embodiments, the step of determining the abnormal pressure section of the formation pore pressure of the target well based on the logging data comprises:

[0023] A P-wave velocity curve is established based on the well depth and the P-wave velocity.

[0024] The normal compaction section and the abnormal pressure section of the formation pore pressure of the target well are determined based on the change trend of the P-wave velocity curve with the well depth.

[0025] A device for determining the abnormal formation pore pressure mechanism comprises:

[0026] The acquisition module is configured to acquire logging data of a full well section of a target well, and determine an abnormal pressure section of formation pore pressure of the target well based on the logging data, wherein the logging data comprises well depth, P-wave velocity, and density.

[0027] The clustering module is configured to cluster and group the logging data corresponding to the abnormal pressure section according to similarity, and obtain a plurality of initial sample data groups.

[0028] The division module is configured to establish a wave velocity-density crossplot based on the logging data, analyze the abnormal pressure mechanism corresponding to each initial sample data group by using the wave velocity-density crossplot, and thereby construct a sample set of abnormal pressure mechanisms.

[0029] The training module is configured to optimize and train an initial abnormal pressure mechanism determination model by using the sample set of abnormal pressure mechanisms, and obtain an abnormal pressure mechanism determination model.

[0030] The determination module is configured to acquire logging data of a target layer, input the logging data of the target layer into the abnormal pressure mechanism determination model, and determine the abnormal pressure mechanism.

[0031] A computer device comprises a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:

[0032] The logging data of a full well section of a target well is acquired, and an abnormal pressure section of formation pore pressure of the target well is determined based on the logging data, wherein the logging data comprises well depth, P-wave velocity, and density.

[0033] The logging data corresponding to the abnormal pressure section is clustered and grouped according to similarity, and a plurality of initial sample data groups are obtained.

[0034] establish a wave velocity-density crossplot based on the logging data, analyze the abnormal pressure formation mechanism corresponding to each initial sample data group respectively by using the wave velocity-density crossplot, and thus construct a sample set of abnormal pressure formation mechanisms;

[0035] train the initial discrimination model of the abnormal pressure formation mechanism by using the sample set of abnormal pressure formation mechanisms, and obtain a discrimination model of abnormal pressure formation mechanisms;

[0036] obtain the logging data of the target layer, and input the logging data of the target layer into the discrimination model of abnormal pressure formation mechanisms for discrimination of the abnormal pressure formation mechanism.

[0037] A computer-readable storage medium has a computer program stored thereon, and the computer program is executed by a processor to implement the following steps:

[0038] obtain the logging data of the target well, determine an abnormal pressure section of the formation pore pressure of the target well based on the logging data, and wherein the logging data comprises well depth, P-wave velocity and density;

[0039] group the logging data corresponding to the abnormal pressure section according to similarity, and obtain a plurality of initial sample data groups;

[0040] establish a wave velocity-density crossplot based on the logging data, analyze the abnormal pressure formation mechanism corresponding to each initial sample data group respectively by using the wave velocity-density crossplot, and thus construct a sample set of abnormal pressure formation mechanisms;

[0041] train the initial discrimination model of the abnormal pressure formation mechanism by using the sample set of abnormal pressure formation mechanisms, and obtain a discrimination model of abnormal pressure formation mechanisms;

[0042] obtain the logging data of the target layer, and input the logging data of the target layer into the discrimination model of abnormal pressure formation mechanisms for discrimination of the abnormal pressure formation mechanism.

[0043] The discrimination method, device, computer equipment and storage medium of the abnormal formation pore pressure formation mechanism described above obtain a sample set of abnormal pressure formation mechanisms through clustering and grouping of logging data and a wave velocity-density crossplot, automatically learn the patterns and rules of the sample set by using a machine learning algorithm, process complex logging features and nonlinear relationships of abnormal formation pore pressure formation mechanisms, and thus accurately discriminate the abnormal formation pore pressure formation mechanism. BRIEF DESCRIPTION OF DRAWINGS

[0044] Figure 1 It is a flowchart of the discrimination method of the abnormal formation pore pressure formation mechanism in one embodiment.

[0045] Figure 2 is a specific flow chart of a method for determining the formation mechanism of abnormal formation pore pressure in another embodiment;

[0046] Figure 3 Schematic diagram of the steps of a method for determining the formation mechanism of abnormal formation pore pressure in one embodiment;

[0047] Figure 4 is a schematic diagram of a wave velocity density crossplot in one embodiment;

[0048] Figure 5 This is a sample distribution diagram of the training set after balancing by the SMOTE algorithm in one embodiment;

[0049] Figure 6 A diagram showing a process of optimizing a LightGBM model using a Bayesian algorithm in one embodiment;

[0050] Figure 7 FIG. 1 is a confusion matrix result diagram of a discriminant model of the abnormal formation pore pressure formation mechanism on a test set in one embodiment;

[0051] Figure 8 is a structural block diagram of a device for determining the formation mechanism of abnormal formation pore pressure in one embodiment;

[0052] Figure 9 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0054] Example 1

[0055] In this embodiment, Figure 1 As shown, a method for determining the mechanism of abnormal formation pore pressure is provided, which includes:

[0056] Step 110 , obtaining logging data of the entire target well section, and determining the abnormal pressure section of the target well formation pore pressure based on the logging data; wherein the logging data includes well depth, compressional wave velocity, and density.

[0057] In this embodiment, the logging data of the entire well section of the target well is the logging data associated with the abnormal formation pore pressure formation mechanism, including well depth data, compressional wave velocity data, density data, porosity data, and mud content data, etc., which are not limited in this embodiment.

[0058] In this embodiment, the normal compaction section and the abnormal pressure section are preliminarily divided using the logging data to obtain the logging data corresponding to the abnormal pressure section. Subsequently, only the logging data corresponding to the abnormal pressure section needs to be used to identify the abnormal pressure cause mechanism to ensure the accuracy of the input sample data.

[0059] Step 120 : Clustering and grouping the logging data corresponding to the abnormal pressure section according to similarity to obtain a plurality of initial sample data groups.

[0060] In this embodiment, an unsupervised learning algorithm is used to cluster the well logging data, and the well logging data is divided into a plurality of initial sample data groups according to similarity.

[0061] In one embodiment, the unsupervised learning method employed is a hierarchical clustering algorithm, which hierarchically divides the well logging data into multiple distinct categories. Hierarchical clustering is used to hierarchically divide data samples into distinct categories, thereby forming a cluster tree structure (also known as a dendrogram or tree map). This method progressively merges or splits data to construct clusters at different levels.

[0062] The main steps in hierarchical clustering involve defining a similarity or distance metric between data points. Common metrics include Euclidean distance, Manhattan distance, and correlation systems. Starting with each data point, clusters are gradually merged or split based on similarity or distance, creating a cluster tree.

[0063] Step 130 : establishing a wave velocity density cross plot based on the well logging data, and using the wave velocity density cross plot to analyze the abnormal pressure formation mechanism corresponding to each initial sample data group, thereby constructing a sample set of the abnormal pressure formation mechanism.

[0064] In this embodiment, the velocity-density crossplot uses density as the horizontal coordinate and longitudinal wave velocity as the vertical coordinate, and plots each logging point in order of well depth; the position of each logging point in the velocity-density crossplot facilitates identification of the abnormal pressure formation mechanism.

[0065] In this embodiment, the logging data corresponding to the abnormal pressure section are first clustered and grouped, and then the abnormal pressure formation mechanism of the multiple initial sample data groups obtained by clustering is judged based on the wave velocity density intersection diagram, which facilitates the division of the initial sample data group, ensures the sample accuracy of the data subsequently input into the machine learning model, and improves the model's recognition accuracy of the abnormal pressure formation mechanism.

[0066] Step 140 : Optimize and train the initial discrimination model of the abnormal pressure formation mechanism using the sample set of the abnormal pressure formation mechanism to obtain the discrimination model of the abnormal pressure formation mechanism.

[0067] In this embodiment, a sample set of abnormal pressure causal mechanisms, namely, a sample set corresponding to different abnormal pressure causal mechanisms, is input into a machine learning model, which is then optimized and trained to obtain a model for distinguishing abnormal pressure causal mechanisms. This embodiment utilizes a machine learning model to distinguish abnormal pressure causal mechanisms, without relying on expert experience, enabling efficient and accurate identification.

[0068] Step 150: Acquire the well logging data of the target layer, and input the well logging data of the target layer into the abnormal pressure formation mechanism identification model to identify the abnormal pressure formation mechanism.

[0069] In this embodiment, the well logging data of the target layer is input into the abnormal pressure formation mechanism identification model established above to identify the abnormal pressure formation mechanism.

[0070] In the above embodiment, a sample set of abnormal pressure genesis mechanisms is obtained by clustering and grouping the logging data and performing wave velocity density cross-plots. A machine learning algorithm is used to automatically learn the patterns and regularities of the logging data, and to process the nonlinear relationship between complex logging characteristics and abnormal formation pore pressure genesis mechanisms, thereby accurately identifying the abnormal pressure genesis mechanisms.

[0071] In one embodiment, the step of establishing a wave velocity density intersection diagram based on the logging data includes: determining the normal compaction section and the abnormal pressure section of the target well formation pore pressure based on the logging data; fitting the logging data of the normal compaction section based on the coordinate system of the longitudinal wave velocity and density to generate a wave velocity density curve of the normal compaction section; generating an initial wave velocity density intersection diagram of the abnormal pressure section based on the logging data of the abnormal pressure section; and synthesizing the wave velocity density curve of the normal compaction section with the initial wave velocity density intersection diagram of the abnormal pressure section to obtain the wave velocity density intersection diagram.

[0072] A velocity-density crossplot is a graphical representation method used to demonstrate the relationship between velocity and density in a formation. In one embodiment, based on the well logging data, an initial velocity-density crossplot is generated in a coordinate system with density as the abscissa and P-wave velocity as the ordinate. The initial sample data group is distributed on the initial velocity-density crossplot. The well logging data of the normal compaction section is fitted in the P-wave velocity and density coordinate system to generate a velocity-density curve for the normal compaction section. This velocity-density curve is then synthesized as a loading curve onto the initial velocity-density crossplot to generate a velocity-density crossplot. This velocity-density crossplot facilitates the classification of abnormal pressure generation mechanisms within the initial sample data group.

[0073] In this embodiment, an initial wave velocity density cross plot is established based on the logging data. In order to facilitate the division of the abnormal pressure causal mechanism of the logging data corresponding to the abnormal pressure section, a normal trend line is added to the initial wave velocity density cross plot to fit the wave velocity density curve of the normal compaction section. According to the relative position of the initial sample data group and the normal trend line, the abnormal pressure causal mechanism of the initial sample data group is determined.

[0074] In one embodiment, the step of using the wave velocity density intersection diagram to analyze the abnormal pressure causal mechanism corresponding to each initial sample data group, thereby constructing a sample set of abnormal pressure causal mechanism includes: based on the position of the initial sample data group relative to the wave velocity density curve in the wave velocity density intersection diagram, determining the abnormal pressure causal mechanism corresponding to each initial sample data group, thereby constructing a sample set of abnormal pressure causal mechanism.

[0075] In the prior art, there are more than a dozen identified causes of abnormal formation pore pressure, but the most common ones include undercompaction, fluid expansion, and mudstone diapir. In this embodiment, the initial sample data group is divided into the three common abnormal pressure causes of undercompaction, fluid expansion, and mudstone diapir through the wave velocity density intersection diagram, which can not only realize the identification of most abnormal bottom pore pressure causes, but also ensure the identification efficiency and accuracy.

[0076] In this embodiment, the initial sample data group is classified into abnormal pressure genesis mechanisms based on their position relative to the velocity-density curve of the normally compacted section in the velocity-density crossplot, that is, their relative position relative to the normal trend line. Specifically, the undercompaction mechanism is assigned label 0, the fluid expansion mechanism is assigned label 1, and the mudstone diapir mechanism is assigned label 2. A label of 0 indicates that the abnormal pressure genesis mechanism corresponds to undercompaction, which deviates from the normal trend line toward lower velocity and density values, but remains distributed around or follows a similar trend line. Therefore, samples with label 0 are located below and to the left and right of the normal trend line. A label of 1 indicates that the abnormal pressure genesis mechanism corresponds to fluid expansion, which exhibits a rapid decrease in velocity and a slight decrease in density. Consequently, samples are located directly below or slightly to the lower left of the normal trend line, deviating toward lower velocity values ​​and are relatively concentrated near the unloading curve within this depth range. Therefore, samples with label 1 are located below and to the center of the normal trend line. The sample labeled 2, that is, the sample whose abnormal pressure mechanism corresponds to mudstone diapir, has a certain degree of increase in wave velocity relative to the normal wave velocity, and the density increases slightly or remains unchanged. Therefore, the sample labeled 2 is relatively concentrated at the high wave velocity value just above or slightly to the upper right of the normal trend line in this depth range.

[0077] In this embodiment, the logging data of the abnormal pressure section is first clustered and grouped to obtain multiple initial sample data groups; then, through the wave velocity density cross plot, the abnormal pressure causal mechanism corresponding to each initial sample data group is analyzed, thereby constructing a sample set of the abnormal pressure causal mechanism, so that the abnormal pressure causal mechanism can be divided for the initial sample data group more accurately and quickly. Compared with the prior art method of directly establishing a wave velocity density cross plot and relying on expert experience to identify the abnormal pressure causal mechanism in the cross plot, this embodiment does not rely on expert experience, but uses the wave velocity density cross plot to divide the samples, and combines it with a machine learning model to identify the abnormal pressure causal mechanism, ensuring the accuracy of the samples when input into the model for training, facilitating the machine learning model to automatically learn the laws of the logging data corresponding to the abnormal pressure section, and processing the nonlinear relationship between complex logging characteristics and the abnormal formation pore pressure causal mechanism, thereby accurately classifying the abnormal formation pore pressure mechanism formation.

[0078] In one embodiment, the step of fitting the logging data of the normal compaction section based on the coordinate system of the longitudinal wave velocity and density to generate the wave velocity density curve of the normal compaction section includes: fitting the logging data of the normal compaction section using the Gardner equation based on the coordinate system of the longitudinal wave velocity and density to generate the wave velocity density curve of the normal compaction section.

[0079] In this embodiment, the velocity-density curve for the normal compaction period is fitted based on the Gardner equation, establishing a normal trend line for the velocity-density crossplot. Furthermore, the coefficients of the Gardner equation are modified using the least squares method to optimize the fitted curve. Compared to the empirical Gardner model used in the prior art, this embodiment improves the fitting accuracy of the velocity-density curve for the normal compaction period, ensuring accurate classification of the abnormal pressure generation mechanism within the initial sample data set.

[0080] In one embodiment, the step of optimizing and training the initial discrimination model of the abnormal pressure causal mechanism using the sample set of the abnormal pressure causal mechanism to obtain the abnormal pressure causal mechanism discrimination model includes: inputting the logging data of the sample set into the initial discrimination model of the abnormal pressure causal mechanism according to the type of the abnormal pressure causal mechanism to identify the abnormal pressure causal mechanism, obtaining a predicted abnormal pressure causal mechanism, comparing the predicted abnormal pressure causal mechanism with the abnormal pressure causal mechanism, and obtaining the model recognition accuracy; optimizing and training the initial discrimination model of the abnormal pressure causal mechanism by a Bayesian optimization algorithm based on a Gaussian process until the accuracy no longer improves, determining the parameter combination corresponding to the highest accuracy as the final parameter combination, and obtaining the abnormal pressure causal mechanism discrimination model based on the final parameter combination.

[0081] In this embodiment, the sample set corresponding to each abnormal pressure formation mechanism is divided into a training set and a test set according to a preset ratio; the logging data of the training set is used as the input layer data set of the initial discriminant model, and the predicted abnormal pressure formation mechanism is output; the predicted abnormal pressure formation mechanism is compared with the abnormal pressure formation mechanism to obtain the model recognition accuracy.

[0082] The initial discriminant model is optimized and trained using a Bayesian optimization algorithm based on a Gaussian process, specifically comprising: according to the prediction results of the Gaussian process model, using the Bayesian optimization algorithm to determine hyperparameters for training the initial discriminant model to obtain an optimized hyperparameter combination; retraining the initial discriminant model according to the optimized hyperparameter combination to obtain a discriminant model for identifying the causal mechanism of abnormal pressure; and testing the prediction accuracy of the discriminant model using the test set.

[0083] In this embodiment, a Bayesian optimization algorithm based on Gaussian process is introduced to perform hyperparameter optimization training on the initial discriminant model to ensure that the optimal solution is found within a limited number of iterations, so as to obtain the abnormal pressure cause mechanism discrimination model more efficiently.

[0084] Specifically, the Bayesian Optimization (BO) algorithm based on Gaussian processes is a method for optimizing black-box functions that combines Gaussian processes as a probabilistic model of the objective function with Bayesian inference for parameter optimization. At each step, this algorithm combines historical observations with predictions from the Gaussian process model to infer the most promising next parameter value. Specifically, the algorithm first models the objective function using a Gaussian process to obtain a prior distribution over the parameter space; it then maximizes the expected improvement in the objective function by continuously selecting the next parameter value, i.e., the best parameter value is expected to be obtained under the current best estimate. After each observation, the Gaussian process model is updated, and the next most promising parameter value is recalculated based on the new information.

[0085] Probabilistic surrogate models and acquisition functions are the core components of Bayesian optimization. Probabilistic surrogate models are generally parametric or non-parametric, with non-parametric models using Gaussian processes as surrogate models being the most widely used. By using Gaussian process models, Bayesian optimization algorithms can select the locations in the search space most likely to reach the global optimal solution, rather than just the local optimal solution. This reduces unnecessary sampling and allows for finding the optimal solution within a limited number of sampling attempts.

[0086] A Gaussian process is a combination of random variables. A Gaussian process function is composed of:

[0087] f(x)~GP(m(x),k(x,x′))

[0088] Where, m(x) is the mean function, which is usually set to 0. k(x,x′) is the covariance function, and x′ is a random variable.

[0089] Consider the prior distribution p(f|X, β) with mean 0 as:

[0090] p(f|X,β)=N(0,∑)

[0091] Where X is the training set, f is the set of function values ​​of the unknown function, and ∑ is the matrix composed of k(x,x′).

[0092] When there is observation noise ε, p(ε)=N(0,σ 2 )), σ 2 Is the variance, the resulting likelihood distribution is:

[0093] p(y|f)=N(f,σ 2 I)

[0094] Where y is the set of observations and I is the identity matrix.

[0095] Then the boundary likelihood distribution is obtained as:

[0096] p(y|X,β)=N(0,∑+σ 2 I)

[0097] According to the properties of Gaussian process, we can get:

[0098]

[0099] Where, f * is the predicted value, K * is the covariance matrix, K * The transposed matrix, K ** is the covariance between the predicted input values. Then the distribution of the prediction is:

[0100]

[0101]

[0102] In the formula, cov(f * ) represents the covariance of the predicted values.

[0103] The acquisition function is an important basis for determining the next evaluation point. When the confidence interval-based strategy is adopted, the next evaluation point is:

[0104]

[0105] In the formula, the parameter α t is a constant that balances exploration and exploitation, μ t (x) is the mean, σ t (x) is the standard deviation.

[0106] In one embodiment, the step of establishing the initial discrimination model of the abnormal pressure formation mechanism includes:

[0107] The decision tree algorithm was selected, combined with gradient boosting and histogram technology, and the initial discriminant model of the abnormal pressure causal mechanism was determined based on the LightGBM algorithm.

[0108] In this embodiment, LightGBM is used as the initial discriminant model for the abnormal pressure formation mechanism. LightGBM (LightGradient Boosting Machine) is an efficient machine learning algorithm based on the gradient boosting framework. LightGBM improves the performance of algorithm learning through strategies such as vertical parallel training, histogram optimization, and feature parallelization. In this embodiment, the LightGBM model is used as the discriminant model for the abnormal formation pore pressure formation mechanism to achieve effective and efficient discrimination of the abnormal formation pore pressure formation mechanism. It improves the training speed and memory efficiency of the model by optimizing the algorithm and data structure. The following is the principle of the lightweight gradient boosting algorithm:

[0109] Vertically parallelized training algorithm: LightGBM uses a vertically parallelized training algorithm called GOSS. In each iteration, the GOSS algorithm first sorts the samples in descending order based on their gradient values, then selects a subset of samples with larger gradient values ​​for training, while sampling samples with smaller gradient values ​​to reduce the size of the dataset. This significantly reduces training time and memory consumption while maintaining model accuracy.

[0110] Histogram Optimization Algorithm: LightGBM also uses a histogram optimization algorithm called GOH. When building a decision tree, the traditional gradient boosting tree algorithm needs to sort the feature values ​​in order to select the best partitioning point. The GOH algorithm buckets the feature values ​​into histograms and then partitions based on the histograms. This can greatly reduce the time complexity of sorting and improve training speed.

[0111] Feature parallelization and histogram compression: LightGBM also uses feature parallelization and histogram compression techniques to further improve training speed and memory efficiency. Feature parallelization allows parallel processing of histogram construction of different features, while histogram compression reduces memory consumption by using discrete gradient histograms.

[0112] Leaf-wise growth strategy: LightGBM uses a leaf-wise growth strategy, which is different from the traditional level-wise growth strategy. In the leaf-wise strategy, the leaf node with the largest gradient is selected for splitting each time, thereby focusing more on samples with larger gradients and accelerating the learning speed of the model.

[0113] Decision tree with depth limit: To further reduce memory consumption, LightGBM uses a decision tree with depth limit. Traditional gradient boosting tree algorithms usually use fully grown decision trees, while LightGBM sets a maximum depth threshold to limit the growth of the decision tree, thereby reducing the complexity of the model and memory usage.

[0114] Overall, LightGBM effectively improves the training speed and memory efficiency of the gradient boosting tree algorithm by introducing technologies such as GOSS and GOH algorithms, Leaf-wise growth strategy, feature parallelization, and histogram compression. It performs well in processing large-scale data sets and high-dimensional features, and has achieved good performance in various machine learning tasks.

[0115] In one embodiment, the Bayesian optimization algorithm of the Gaussian process is introduced to perform combinatorial optimization on the hyperparameters of the LightGBM model, where the hyperparameters include the number of base models, the depth of the decision tree, and the learning rate. Specifically, the search range of the base model data is [1, 100], the search range of the depth of the decision tree is [1, 50], and the search range of the learning rate is [0, 1]. Furthermore, the number of iterative search runs of the Bayesian optimization algorithm is set to 100 times, and the initial search number is 10 to ensure that the optimal hyperparameter combination can be searched, thereby obtaining the combined parameters of the LightGBM model with the highest recognition accuracy.

[0116] In one embodiment, the step of determining the abnormal pressure section of the target well formation pore pressure based on the logging data includes: establishing a P-wave velocity curve based on the well depth and P-wave velocity; and determining the normal compaction section and abnormal pressure section of the target well formation pore pressure based on the changing trend of the P-wave velocity curve with well depth.

[0117] In this embodiment, based on the well depth and P-wave velocity in the logging data, a P-wave velocity curve is generated to determine the normal compaction section and abnormal pressure section of the formation pore pressure by analyzing the changing trend of the P-wave velocity curve with the well depth. Under normal circumstances, as the well depth increases, the P-wave velocity may show a gradual increase or a relatively stable changing trend, while the abnormal pressure section may be manifested as a sudden decrease in the P-wave velocity or a discontinuous change. Therefore, in this embodiment, the normal compaction section and the abnormal pressure section are first determined by the changing trend of the P-wave velocity curve with the well depth, and the logging data is divided. Subsequently, only the logging data of the abnormal pressure section needs to be used to identify the abnormal pressure formation mechanism to ensure the accuracy of the sample data input into the model.

[0118] In some other embodiments, the normal compaction section and abnormal pressure section of the pore pressure may be determined by porosity, permeability, etc., which is not limited in this embodiment.

[0119] In one embodiment, the sample set of the abnormal pressure causation mechanism is used to optimize and train the initial discrimination model of the abnormal pressure causation mechanism, and the step of obtaining the discrimination model of the abnormal pressure causation mechanism also includes: dividing the sample set corresponding to the abnormal pressure causation mechanism into a training set and a test set according to a preset ratio; and balancing the number of samples corresponding to each abnormal pressure causation mechanism in the training set by using the SMOTE algorithm.

[0120] In this embodiment, the obtained sample set is divided into a training set and a test set according to a preset ratio. Specifically, it can be randomly divided into a training set and a test set in a ratio of 7:3. Because the number of samples corresponding to each abnormal pressure causal mechanism obtained is different, the number of samples corresponding to a certain abnormal pressure causal mechanism may be too small, resulting in the inability of the computational learning model to train it. Therefore, in this embodiment, after the sample set is divided into a training set and a test set, the number of samples corresponding to each abnormal pressure causal mechanism in the training set is also balanced by the SMOTE algorithm, that is, the samples corresponding to one or two types of abnormal pressure causal mechanisms with a smaller number of samples in the training set are expanded by the SMOTE algorithm to effectively balance the sample ratios corresponding to each abnormal pressure causal mechanism in the training set, so that the model can fully learn the characteristics of each abnormal formation pore pressure causal mechanism.

[0121] The SMOTE algorithm uses original samples to generate new samples that are distinct from the original. This algorithm effectively addresses sample imbalance, thus preventing overfitting, a common problem in prediction algorithms. The advantage of this algorithm lies in its rationality in preserving the characteristics of the original samples while interpolating them. The key idea is to select a sample x from a set of similar samples, find K neighboring samples near x, and then randomly select a sample x' from these samples. Interpolation is then performed on the line connecting x and x' according to specific strategies and rules, thus preserving the similar characteristics of similar samples.

[0122] SMOTE sets the algorithm parameters: sample set S∈{A, B}, minority class samples are positive samples A and multi-class samples are negative samples B. Empty set X new 、S K 、S new ; Neighborhood parameter K; Sample imbalance rate IR = A / B; Sampling rate S R Output: The sample set S′ after sampling by the SMOTE algorithm.

[0123] SMOTE algorithm implementation steps:

[0124] The sample set S is divided into positive and negative sample categories, with the minority class samples being positive samples A and the multi-class samples being negative samples B.

[0125] Randomly select a sample x from the positive sample A;

[0126] Find K neighboring samples near the x sample, set as S K gather;

[0127] Select all S K A sample in is x ′ , use the following mathematical formula to calculate the new sample x new :

[0128] x new =x+rand(0,1)×(x ′ -x)

[0129] The generated x new Add to X new middle:

[0130] S new =S new ∪X new .

[0131] The obtained sample S new Perform a union operation on sample A and sample B to achieve oversampling of the sample, which is set as S′.

[0132] In one embodiment, after the training set samples are balanced and expanded, the maximum and minimum normalization algorithm is used to normalize the well logging data to eliminate the dimensional effects between different data.

[0133] It should be understood that although Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. In addition, Figure 1 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0134] Example 2

[0135] In this embodiment, Figures 2-7 As shown in the figure, a method for distinguishing the causal mechanism of abnormal formation pore pressure in a basin is provided to solve the problem that the existing distinction method is too dependent on expert experience and cannot quickly and effectively distinguish the causal mechanism of abnormal pressure.

[0136] like Figure 2 and Figure 3 As shown, the method includes:

[0137] Step 161, collect the full-well logging data of the target well, determine the normal compaction section and abnormal pressure section of the target well formation pore pressure based on the trend of the longitudinal wave velocity curve with well depth, cluster based on the logging parameters using an unsupervised learning algorithm, divide the abnormal pressure cause mechanism into different groups, and mine sample data groups with high similarity.

[0138] In this example, logging data from the entire target well was collected. Specifically, the logging data originated from the Rio de la Ré Basin in Cameroon and included data such as well depth, acoustic velocity, density, porosity, and shale content. Based on the logging data, the trend of the P-wave velocity curve with well depth was determined. Analysis of this trend with acoustic velocity revealed that the normally compacted section is located below 2730 m, and the abnormally pressurized section is located above 2730 m.

[0139] Clustering is performed based on logging parameters using an unsupervised learning algorithm to divide sample data groups with higher similarity. Specifically, the unsupervised learning method used is a hierarchical clustering algorithm. When clustering the logging data, it is determined that when the number of categories is 9, the classification effect of the sample data group is more obvious. Therefore, the sample data is hierarchically divided into 9 different categories.

[0140] Step 162: Fit the velocity-density curve for the normally compacted section using the Gardner equation to establish a normal trend line for the velocity-density crossplot. This normal trend line is applied as the loading curve to the crossplot for the abnormally pressured section. Using conventional discriminant methods, the data group is divided into three categories: undercompaction, fluid expansion, and mudstone diapir.

[0141] In this embodiment, the wave velocity-density intersection Figure 1 A graphical representation method is used to demonstrate the relationship between velocity and density in a formation. An initial velocity-density crossplot is established based on the well logging data corresponding to the abnormal pressure section. To facilitate the classification of the abnormal pressure-causing mechanism in the well logging data corresponding to the abnormal pressure section, a normal trend line is added to the initial velocity-density crossplot to fit the velocity-density curve of the normal compaction section. Specifically, the velocity-density curve of the normal compaction section is fitted based on the Gardner equation to establish a normal trend line for the velocity-density crossplot. Furthermore, the coefficients of the Gardner equation are corrected using the least squares method based on the normal trend line established based on the well logging data corresponding to the normal compaction section, and the fitted curve is optimized, thereby improving the fitting accuracy of the normal trend line compared to the empirical Gardner model.

[0142] In this embodiment, the abnormal pressure genesis mechanism is divided for the initial sample data group based on the position of the initial sample data group relative to the normal trend line in the wave velocity density cross plot. The undercompaction genesis mechanism is defined as label 0, the fluid expansion genesis mechanism is defined as label 1, and the mudstone diapir genesis mechanism is defined as label 2. Figure 4 As shown, the sample labeled 2 is located to the upper right of the normal trend line, and its abnormal pressure mechanism corresponds to mudstone diapir. The sample labeled 1 is located slightly below the center of the normal trend line, and its abnormal pressure mechanism corresponds to fluid expansion. The sample labeled 0 is located to the left and right of the normal trend line, and its abnormal pressure mechanism corresponds to undercompaction. In this example, the number of sample sets corresponding to the three abnormal pressure mechanisms of undercompaction, fluid expansion, and mudstone diapir is 613 in total.

[0143] Step 163, the obtained abnormal pressure sample is divided into a training set and a test set, and the SMOTE algorithm is used to balance the training set sample. The maximum and minimum normalization algorithm is used to normalize the logging features, so as to eliminate the influence of the dimension of different data. The decision tree algorithm is selected to regard the numerical value as a category symbol, and the LightGBM algorithm is established by combining the gradient boosting and histogram technology as an initial discrimination model of the abnormal pressure mechanism.

[0144] In the embodiment, the sample set corresponding to the abnormal pressure mechanism is divided into a training set and a test set according to a preset proportion. Specifically, the training set and the test set can be randomly divided according to a proportion of 7:3. In the embodiment, after the sample set is divided into the training set and the test set, the SMOTE algorithm is used to balance the number of samples corresponding to each abnormal pressure mechanism in the training set, so as to effectively balance the sample proportion corresponding to the abnormal pressure mechanism, and enable the model to sufficiently learn the characteristics of each abnormal formation pore pressure mechanism. Specifically, as shown in FIG. 2, after the SMOTE algorithm is expanded, the number of samples corresponding to three abnormal formation pore pressure mechanisms in the training set is 216 groups. Figure 5

[0145] In the embodiment, the initial discrimination model of the abnormal pressure mechanism selects a LightGBM model. The LightGBM (Light Gradient Boosting Machine) is a high-efficiency machine learning algorithm based on a gradient boosting framework. In the embodiment, the LightGBM model is used as the discrimination model of the abnormal formation pore pressure mechanism, so as to effectively and efficiently discriminate the abnormal formation pore pressure mechanism.

[0146] In the embodiment, the hyperparameters required to be optimized by the LightGBM model mainly include the number of base models, the depth of the decision tree, and the learning rate.

[0147] Step 164, the abnormal pressure mechanism is taken as an output result, and the initial discrimination model of the abnormal pressure mechanism is optimized and trained by using the Gaussian process-based Bayesian optimization algorithm to obtain the discrimination model of the abnormal pressure mechanism.

[0148] ​In this embodiment, the logging data of the sample set is input into the abnormal pressure mechanism initial discrimination model according to the types of abnormal pressure mechanism formation mechanism to identify the abnormal pressure mechanism formation mechanism, and the predicted abnormal pressure mechanism formation mechanism is obtained. By comparing the predicted abnormal pressure mechanism formation mechanism with the abnormal pressure mechanism formation mechanism, the model recognition accuracy is obtained; the abnormal pressure mechanism initial discrimination model is optimized and trained based on the Gaussian process-based Bayesian optimization algorithm until the accuracy no longer improves, and the parameter combination corresponding to the highest accuracy is determined as the final parameter combination. The abnormal pressure mechanism discrimination model is obtained based on the final parameter combination; and the prediction accuracy of the discrimination model is tested by the test set.

[0149] Referring to Figure 6 As shown in the figure, in this embodiment, the Gaussian process-based Bayesian optimization algorithm is introduced to combine and optimize the base model number, the depth of the decision tree and the learning rate of the LightGBM model (i.e. the abnormal pressure mechanism initial discrimination model). The search range of the base model data is [1, 100], the search range of the depth of the decision tree is [1, 50], and the search range of the learning rate is [0, 1]. Further, the number of iterative search runs of the Bayesian optimization algorithm is set to 100 times, and the initial search number is 10, so as to ensure that the best hyperparameter combination can be searched, so as to obtain the combined parameters of the LightGBM model with the highest recognition accuracy.

[0150] In this embodiment, the Gaussian process-based Bayesian optimization algorithm is used to optimize and train the hyperparameters of the LightGBM model, and the final combination of the hyperparameters is obtained. The optimal result of the base model number is 28, the optimal structure of the depth of the decision tree is 26, and the optimal result of the learning rate is 0.688. The recognition accuracy of the abnormal pressure mechanism discrimination model reaches 0.967.

[0151] Step 165, acquiring the logging data of the target layer, inputting the logging data of the target layer into the abnormal pressure mechanism discrimination model to discriminate the abnormal pressure mechanism.

[0152] In this embodiment, by inputting the logging data of the target layer into the abnormal pressure mechanism discrimination model, including acoustic velocity, formation density, shale content and porosity, the abnormal formation pore pressure mechanism at this depth can be discriminated.

[0153] Referring to Figure 7 As shown in the figure, Figure 7This is the confusion matrix result diagram of the abnormal formation pore pressure mechanism discrimination model on the test samples. The confusion matrix can intuitively observe the number of correct and misidentified abnormal pressure mechanisms by the abnormal formation pore pressure mechanism discrimination model on the test samples. It can be found that only 6 undercompacted samples were misidentified, that is, the abnormal formation pore pressure mechanism discrimination method based on machine learning can accurately distinguish different abnormal formation pore pressure mechanisms on most test samples.

[0154] Example 3

[0155] In this embodiment, Figure 8 As shown, a device for determining the formation mechanism of abnormal formation pore pressure is provided, comprising:

[0156] An acquisition module 210 is configured to acquire logging data of the entire target well section and determine the abnormal pressure section of the target well formation pore pressure based on the logging data; wherein the logging data includes well depth, compressional wave velocity, and density;

[0157] A clustering module 220 is configured to cluster and group the well logging data corresponding to the abnormal pressure section according to similarity to obtain a plurality of initial sample data groups;

[0158] a partitioning module 230 for establishing a wave velocity density cross plot based on the well logging data, and using the wave velocity density cross plot to analyze the abnormal pressure formation mechanism corresponding to each initial sample data group, thereby constructing a sample set of the abnormal pressure formation mechanism;

[0159] A training module 240 is configured to optimize and train the initial discrimination model of the abnormal pressure formation mechanism using the sample set of the abnormal pressure formation mechanism to obtain a discrimination model of the abnormal pressure formation mechanism;

[0160] The identification module 250 is used to obtain the logging data of the target layer, and input the logging data of the target layer into the abnormal pressure formation mechanism identification model to identify the abnormal pressure formation mechanism.

[0161] In one embodiment, the division module is also used to determine the normal compaction section and abnormal pressure section of the target well formation pore pressure based on the logging data; based on the coordinate system of the longitudinal wave velocity and density, the logging data of the normal compaction section is fitted to generate a wave velocity density curve of the normal compaction section; based on the logging data of the abnormal pressure section, an initial wave velocity density intersection diagram of the abnormal pressure section is generated; the wave velocity density curve of the normal compaction section is synthesized with the initial wave velocity density intersection diagram of the abnormal pressure section to obtain the wave velocity density intersection diagram.

[0162] In an embodiment, the dividing module is further configured to determine abnormal pressure mechanism corresponding to each initial sample data group based on the position of the initial sample data group in the wave velocity-density crossplot, so as to construct a sample set of abnormal pressure mechanism.

[0163] In an embodiment, the dividing module is further configured to fit the logging data of the normal compaction section by using Gardner equation based on the coordinate system of the P-wave velocity and density, so as to generate the wave velocity-density curve of the normal compaction section.

[0164] In an embodiment, the training module is further configured to input the logging data of the sample set into an initial discriminant model of abnormal pressure mechanism according to the type of abnormal pressure mechanism, so as to identify the abnormal pressure mechanism and obtain a predicted abnormal pressure mechanism, compare the predicted abnormal pressure mechanism with the abnormal pressure mechanism, and obtain a model identification accuracy; optimize and train the initial discriminant model of abnormal pressure mechanism by using a Gaussian process-based Bayesian optimization algorithm until the accuracy no longer improves, determine a parameter combination corresponding to the highest accuracy as a final parameter combination, and obtain the discriminant model of abnormal pressure mechanism based on the final parameter combination.

[0165] In an embodiment, the training module is further configured to select a decision tree algorithm, combine gradient boosting and histogram technology, and determine the initial discriminant model of abnormal pressure mechanism based on a LightGBM algorithm.

[0166] In an embodiment, the obtaining module is further configured to establish a P-wave velocity curve based on the well depth and P-wave velocity; and determine a normal compaction section and an abnormal pressure section of the formation pore pressure of the target well based on the variation trend of the P-wave velocity curve with the well depth.

[0167] The specific limitations of the abnormal formation pore pressure mechanism discrimination device can refer to the limitations of the abnormal formation pore pressure mechanism discrimination method described above, which will not be repeated here. Each unit in the above abnormal formation pore pressure mechanism discrimination device can be realized by software, hardware, and combinations thereof, in whole or in part. The above units can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory in the computer device in software form, so as to be called and executed by the processor to perform the operations corresponding to each unit.

[0168] Embodiment Four

[0169] In this embodiment, a computer device is provided. Its internal structure diagram can be as follows: Figure 9As shown in the figure. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected by a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium, an internal memory. The non-volatile storage medium stores an operating system and a computer program, and the non-volatile storage medium is deployed with a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The network interface of the computer device is used to communicate with other computer devices deployed with application software. The computer program is executed by the processor to implement a cargo warehouse data processing method. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the computer device shell, or an external keyboard, touchpad or mouse, etc.

[0170] Those skilled in the art can understand that, Figure 9 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0171] In one embodiment, a computer device is provided, comprising a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the following steps:

[0172] Obtain the logging data of the target well full well section, and determine the abnormal pressure section of the target well formation pore pressure based on the logging data; wherein the logging data includes well depth, P-wave velocity and density;

[0173] According to the similarity, the logging data corresponding to the abnormal pressure section is clustered and grouped to obtain a plurality of initial sample data groups;

[0174] Based on the logging data, a wave velocity-density crossplot is established, and the abnormal pressure formation mechanism corresponding to each initial sample data group is analyzed by using the wave velocity-density crossplot, so as to construct a sample set of abnormal pressure formation mechanism;

[0175] Using the sample set of abnormal pressure formation mechanism, an initial discriminant model of abnormal pressure formation mechanism is optimized and trained to obtain an abnormal pressure formation mechanism discriminant model;

[0176] Obtain the logging data of the target layer, and input the logging data of the target layer into the abnormal pressure formation mechanism discriminant model to discriminate the abnormal pressure formation mechanism.

[0177] In one embodiment, the processor, when executing the computer program, also implements the following steps:

[0178] determining a normal compaction section and an abnormal pressure section of a target well formation pore pressure based on the logging data;

[0179] fitting the logging data of the normal compaction section based on the coordinate system of the P-wave velocity and density, to generate a wave velocity-density curve of the normal compaction section; generating an initial wave velocity-density crossplot of the abnormal pressure section according to the logging data of the abnormal pressure section; and synthesizing the wave velocity-density curve of the normal compaction section and the initial wave velocity-density crossplot of the abnormal pressure section to obtain the wave velocity-density crossplot.

[0180] In one embodiment, the processor, when executing the computer program, also implements the following steps:

[0181] determining the abnormal pressure formation mechanism corresponding to each initial sample data group based on the position of the initial sample data group in the wave velocity-density crossplot relative to the wave velocity-density curve, to thereby construct a sample set of abnormal pressure formation mechanisms.

[0182] In one embodiment, the processor, when executing the computer program, also implements the following steps:

[0183] fitting the logging data of the normal compaction section based on the coordinate system of the P-wave velocity and density using the Gardner equation, to generate the wave velocity-density curve of the normal compaction section.

[0184] In one embodiment, the processor, when executing the computer program, also implements the following steps:

[0185] inputting the logging data of the sample set into an initial discriminant model of abnormal pressure formation mechanisms according to the type of abnormal pressure formation mechanism, to identify the abnormal pressure formation mechanism, to obtain a predicted abnormal pressure formation mechanism, and comparing the predicted abnormal pressure formation mechanism with the abnormal pressure formation mechanism to obtain a model identification accuracy rate;

[0186] optimizing and training the initial discriminant model of abnormal pressure formation mechanisms based on a Gaussian process-based Bayesian optimization algorithm until the accuracy rate no longer improves, determining a parameter combination corresponding to the highest accuracy rate as a final parameter combination, and obtaining the discriminant model of abnormal pressure formation mechanisms based on the final parameter combination.

[0187] In one embodiment, the processor, when executing the computer program, also implements the following steps:

[0188] selecting a decision tree algorithm, combining gradient boosting and histogram technology, and determining the initial discriminant model of abnormal pressure formation mechanisms based on a LightGBM algorithm.

[0189] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0190] establishing a P-wave velocity curve based on the well depth and the P-wave velocity;

[0191] The normal compaction section and abnormal pressure section of the target well formation pore pressure are determined based on the variation trend of the P-wave velocity curve with well depth.

[0192] Example 5

[0193] In this embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0194] Acquire logging data of the entire target well section, and determine the abnormal pressure section of the target well formation pore pressure based on the logging data; wherein the logging data includes well depth, compressional wave velocity, and density;

[0195] Clustering and grouping the well logging data corresponding to the abnormal pressure section according to similarity to obtain multiple initial sample data groups;

[0196] establishing a wave velocity density cross plot based on the well logging data, and using the wave velocity density cross plot to analyze the abnormal pressure formation mechanism corresponding to each initial sample data group, thereby constructing a sample set of the abnormal pressure formation mechanism;

[0197] Utilizing the sample set of the abnormal pressure formation mechanism, optimizing and training the initial discrimination model of the abnormal pressure formation mechanism to obtain the discrimination model of the abnormal pressure formation mechanism;

[0198] The well logging data of the target layer is obtained, and the well logging data of the target layer is input into the abnormal pressure formation mechanism discrimination model to discriminate the abnormal pressure formation mechanism.

[0199] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0200] Determine the normal compaction section and abnormal pressure section of the target well formation pore pressure based on the well logging data;

[0201] Based on the coordinate system of the longitudinal wave velocity and density, the logging data of the normal compaction section is fitted to generate a wave velocity density curve of the normal compaction section; based on the logging data of the abnormal pressure section, an initial wave velocity density cross plot of the abnormal pressure section is generated; and the wave velocity density curve of the normal compaction section is synthesized with the initial wave velocity density cross plot of the abnormal pressure section to obtain the wave velocity density cross plot.

[0202] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0203] Based on the position of the initial sample data group relative to the wave velocity density curve in the wave velocity density cross-plot, the abnormal pressure formation mechanism corresponding to each initial sample data group is determined, thereby constructing a sample set of the abnormal pressure formation mechanism.

[0204] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0205] Based on the coordinate system of the longitudinal wave velocity and density, the well logging data of the normal compaction section is fitted using the Gardner equation to generate the velocity density curve of the normal compaction section.

[0206] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0207] inputting the logging data of the sample set into an initial discrimination model for abnormal pressure causal mechanism according to the type of abnormal pressure causal mechanism to identify the abnormal pressure causal mechanism, obtaining a predicted abnormal pressure causal mechanism, and comparing the predicted abnormal pressure causal mechanism with the abnormal pressure causal mechanism to obtain a model identification accuracy rate;

[0208] The initial discrimination model of the abnormal pressure causal mechanism is optimized and trained by a Bayesian optimization algorithm based on a Gaussian process until the accuracy rate no longer increases, and the parameter combination corresponding to the highest accuracy rate is determined as the final parameter combination. The discrimination model of the abnormal pressure causal mechanism is obtained based on the final parameter combination.

[0209] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0210] The decision tree algorithm was selected, combined with gradient boosting and histogram technology, and the initial discriminant model of the abnormal pressure causal mechanism was determined based on the LightGBM algorithm.

[0211] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0212] establishing a P-wave velocity curve based on the well depth and the P-wave velocity;

[0213] The normal compaction section and abnormal pressure section of the target well formation pore pressure are determined based on the variation trend of the P-wave velocity curve with well depth.

[0214] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0215] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0216] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. A method for determining the mechanism of formation of abnormal formation pore pressure, characterized in that: include: Acquire logging data of the entire target well section, and determine the abnormal pressure section of the target well formation pore pressure based on the logging data; wherein the logging data includes well depth, compressional wave velocity, and density; Clustering and grouping the well logging data corresponding to the abnormal pressure section according to similarity to obtain multiple initial sample data groups; establishing a wave velocity density cross plot based on the well logging data, and using the wave velocity density cross plot to analyze the abnormal pressure formation mechanism corresponding to each initial sample data group, thereby constructing a sample set of the abnormal pressure formation mechanism; Utilizing the sample set of the abnormal pressure formation mechanism, an initial discrimination model of the abnormal pressure formation mechanism is optimized and trained to obtain a discrimination model of the abnormal pressure formation mechanism; The well logging data of the target layer is obtained, and the well logging data of the target layer is input into the abnormal pressure formation mechanism discrimination model to discriminate the abnormal pressure formation mechanism.

2. The method for determining the formation mechanism of abnormal formation pore pressure according to claim 1, characterized in that: The step of establishing a wave velocity density crossplot based on the well logging data comprises: Determine the normal compaction section and abnormal pressure section of the target well formation pore pressure based on the well logging data; Based on the coordinate system of the longitudinal wave velocity and density, the logging data of the normal compaction section is fitted to generate a wave velocity density curve of the normal compaction section; based on the logging data of the abnormal pressure section, an initial wave velocity density cross plot of the abnormal pressure section is generated; and the wave velocity density curve of the normal compaction section is synthesized with the initial wave velocity density cross plot of the abnormal pressure section to obtain the wave velocity density cross plot.

3. The method for determining the formation mechanism of abnormal formation pore pressure according to claim 2, characterized in that: The step of analyzing the abnormal pressure formation mechanism corresponding to each initial sample data group using the wave velocity density cross plot to construct a sample set of the abnormal pressure formation mechanism includes: Based on the position of the initial sample data group relative to the wave velocity density curve in the wave velocity density cross-plot, the abnormal pressure formation mechanism corresponding to each initial sample data group is determined, thereby constructing a sample set of the abnormal pressure formation mechanism.

4. The method for determining the formation mechanism of abnormal formation pore pressure according to claim 2, characterized in that: The step of fitting the logging data of the normal compaction section based on the coordinate system of the longitudinal wave velocity and density to generate the velocity density curve of the normal compaction section includes: Based on the coordinate system of the longitudinal wave velocity and density, the well logging data of the normal compaction section is fitted using the Gardner equation to generate the velocity density curve of the normal compaction section.

5. The method for determining the formation mechanism of abnormal formation pore pressure according to claim 1, characterized in that: The step of optimizing and training the initial discrimination model of the abnormal pressure formation mechanism using the sample set of the abnormal pressure formation mechanism to obtain the discrimination model of the abnormal pressure formation mechanism comprises: inputting the logging data of the sample set into an initial discrimination model for abnormal pressure causal mechanism according to the type of abnormal pressure causal mechanism to identify the abnormal pressure causal mechanism, obtaining a predicted abnormal pressure causal mechanism, and comparing the predicted abnormal pressure causal mechanism with the abnormal pressure causal mechanism to obtain a model identification accuracy rate; The initial discrimination model of the abnormal pressure causal mechanism is optimized and trained by a Bayesian optimization algorithm based on a Gaussian process until the accuracy rate no longer increases, and the parameter combination corresponding to the highest accuracy rate is determined as the final parameter combination. The discrimination model of the abnormal pressure causal mechanism is obtained based on the final parameter combination.

6. The method for determining the formation mechanism of abnormal formation pore pressure according to claim 1, characterized in that: The steps of establishing the initial discrimination model of the abnormal pressure formation mechanism include: The decision tree algorithm was selected, combined with gradient boosting and histogram technology, and the initial discriminant model of the abnormal pressure causal mechanism was determined based on the LightGBM algorithm.

7. The method for determining the formation mechanism of abnormal formation pore pressure according to claim 1, characterized in that: The step of determining the abnormal pressure section of the target well formation pore pressure based on the well logging data comprises: establishing a P-wave velocity curve based on the well depth and the P-wave velocity; The normal compaction section and abnormal pressure section of the target well formation pore pressure are determined based on the variation trend of the P-wave velocity curve with well depth.

8. A device for determining the mechanism of formation of abnormal pore pressure, characterized in that: include: An acquisition module is used to acquire logging data of the entire target well section, and determine the abnormal pressure section of the target well formation pore pressure based on the logging data; wherein the logging data includes well depth, compressional wave velocity and density; A clustering module, configured to cluster and group the well logging data corresponding to the abnormal pressure section according to similarity to obtain a plurality of initial sample data groups; a partitioning module for establishing a wave velocity density cross plot based on the well logging data, and using the wave velocity density cross plot to analyze the abnormal pressure formation mechanism corresponding to each initial sample data group, thereby constructing a sample set of the abnormal pressure formation mechanism; a training module, configured to optimize and train the initial discrimination model of the abnormal pressure formation mechanism using the sample set of the abnormal pressure formation mechanism to obtain the discrimination model of the abnormal pressure formation mechanism; The discrimination module is used to obtain the logging data of the target layer, and input the logging data of the target layer into the abnormal pressure formation mechanism discrimination model to discriminate the abnormal pressure formation mechanism.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images