Prediction method and device for hydrocarbon source rock in deep concave area and storage medium

Abstract

The invention discloses a prediction method and device for hydrocarbon source rocks in a deep concave area and a storage medium, and belongs to the technical field of oil and gas resource exploration, and the method comprises the steps: building a hydrocarbon source rock prediction model in the deep concave area according to a plurality of seismic attribute data and hydrocarbon source rock distribution attribute data of a to-be-researched block, wherein the plurality of seismic attribute data and the hydrocarbon source rock distribution attribute data are respectively input and output of the deep-concave region hydrocarbon source rock prediction model; acquiring a plurality of seismic attribute data of the deep concave area in the to-be-researched block; and substituting the plurality of seismic attribute data of the deep concave region in the to-be-researched block into the deep concave region hydrocarbon source rock prediction model to obtain deep concave region hydrocarbon source rock distribution attribute data. According to the method, the hydrocarbon source rocks in the deep concave area can be accurately predicted, and the exploration and development efficiency of the hydrocarbon source rocks in the deep concave area is improved.

Description

Technical Field

[0001] This application relates to the field of oil and gas resource exploration technology, and in particular to a method, apparatus and storage medium for predicting hydrocarbon source rocks in deep depressions. Background Technology

[0002] Source rocks are rocks rich in organic matter that can generate and provide industrial quantities of oil and gas, and they have significant research value in petroleum exploration and development. Due to the low exploration level and small number of wells in deep depressions, there is insufficient analytical and testing data, making it difficult to predict source rocks in many wellless areas.

[0003] In related technologies, seismic attribute data of deep depressions are mainly substituted into linear regression models to determine the distribution characteristics of source rocks in these areas. However, since the linear regression model only considers a single seismic attribute data point, the predicted distribution characteristics of source rocks differ significantly from the actual distribution characteristics, affecting the efficiency of exploration and development of source rocks in deep depressions. Summary of the Invention

[0004] In view of this, this application provides a method, device and storage medium for predicting source rocks in deep depressions, which can predict source rocks in deep depressions more accurately and improve the exploration and development efficiency of source rocks in deep depressions.

[0005] Specifically, the following technical solutions are included:

[0006] In a first aspect, embodiments of this application provide a method for predicting hydrocarbon source rocks in deep depressions, the method comprising:

[0007] Based on multiple seismic attribute data and source rock distribution attribute data of the block under study, a source rock prediction model for the deep depression area is established, wherein the multiple seismic attribute data and the source rock distribution attribute data are the input and output of the source rock prediction model for the deep depression area, respectively.

[0008] Acquire multiple seismic attribute data of the deep depression within the block to be studied;

[0009] Substitute multiple seismic attribute data of the deep depression area within the block under study into the hydrocarbon source rock prediction model of the deep depression area to obtain the hydrocarbon source rock distribution attribute data of the deep depression area.

[0010] In some embodiments, establishing a deep depression hydrocarbon source rock prediction model based on multiple seismic attribute data and source rock distribution attribute data of the block under study includes:

[0011] Acquire multiple seismic attribute data and source rock distribution attribute data of the block to be studied;

[0012] The multiple seismic attribute data were filtered to identify several strongly correlated seismic attribute data.

[0013] Based on the multiple strongly correlated seismic attribute data and the hydrocarbon source rock distribution attribute data, a preset learning model is trained to obtain the hydrocarbon source rock prediction model for the deep depression area.

[0014] In some embodiments, acquiring multiple seismic attribute data and source rock distribution attribute data of the block under study includes:

[0015] Obtain logging data, well logging data, and geochemical analysis data from all wells in the block under study to determine the source rock distribution attributes of the block under study.

[0016] Seismic data from each well within the study block are acquired, and the seismic data from each well within the study block are filtered to determine the seismic attribute data related to the distribution attributes of the source rock as multiple seismic attribute data of the study block.

[0017] In some embodiments, before filtering the plurality of seismic attribute data to determine a plurality of strongly correlated seismic attribute data, the method further includes:

[0018] Preprocessing is performed on multiple seismic attribute data and source rock distribution attribute data of the block to be studied.

[0019] In some embodiments, filtering the plurality of seismic attribute data to determine a plurality of strongly correlated seismic attribute data includes:

[0020] Obtain the correlation coefficient, first threshold, and second threshold for each seismic attribute and source rock distribution attribute;

[0021] In response to the seismic attribute being greater than the first threshold, multiple transitional seismic attributes are determined;

[0022] In response to the transitional seismic attribute being less than the second threshold, multiple strongly correlated seismic attributes are determined;

[0023] Based on the aforementioned multiple strongly correlated earthquake attributes, multiple strongly correlated earthquake attribute data were obtained.

[0024] In some embodiments, the number of preset learning models is multiple, and the step of training the preset learning models based on the multiple strongly correlated seismic attribute data and the hydrocarbon source rock distribution attribute data to obtain the deep depression hydrocarbon source rock prediction model includes:

[0025] Based on the multiple strongly correlated seismic attribute data and the hydrocarbon source rock distribution attribute data, each preset learning model is trained and tested to obtain the determination coefficient of each learning model, wherein the multiple strongly correlated seismic attribute data and the hydrocarbon source rock distribution attribute data are the input and output of each preset learning model, respectively.

[0026] The preset learning model corresponding to the maximum value of the coefficient of determination is determined as the prediction model for hydrocarbon source rocks in the deep depression area.

[0027] In some embodiments, the step of training and testing each preset learning model based on the plurality of strongly correlated seismic attribute data and the hydrocarbon source rock distribution attribute data to obtain the determination coefficient of each learning model includes:

[0028] The multiple strongly correlated seismic attribute data and the hydrocarbon source rock distribution attribute data are divided into a training data set and a test data set;

[0029] Based on the training data set, each preset learning model is trained to obtain the trained learning model;

[0030] Based on the test data set, the trained learning models are tested to obtain the determination coefficient of each learning model.

[0031] In some embodiments, the source rock distribution properties include lithology, thickness, and total organic carbon content.

[0032] Secondly, embodiments of this application also provide a device for predicting hydrocarbon source rocks in deep depressions, the device comprising:

[0033] The modeling module is used to establish a deep depression hydrocarbon source rock prediction model based on multiple seismic attribute data and hydrocarbon source rock distribution attribute data of the block to be studied, wherein the multiple seismic attribute data and the hydrocarbon source rock distribution attribute data are the input and output of the deep depression hydrocarbon source rock prediction model, respectively.

[0034] The acquisition module is used to acquire multiple seismic attribute data of the deep depression area within the block under study;

[0035] The module is used to substitute multiple seismic attribute data of the deep depression area within the block under study into the hydrocarbon source rock prediction model of the deep depression area, thereby obtaining the hydrocarbon source rock distribution attribute data of the deep depression area.

[0036] Thirdly, embodiments of this application also provide a non-volatile readable storage medium storing at least one program, which is loaded and executed by a processor to implement the prediction method for source rocks in deep depressions as described in any of the first aspects.

[0037] The beneficial effects of the technical solutions provided in this application include at least the following:

[0038] The method for predicting source rocks in deep depressions provided in this application establishes a prediction model for source rocks in deep depressions based on multiple seismic attribute data and source rock distribution attribute data of the block under study. Since the input of this prediction model is multiple seismic attribute data, in other words, the prediction model considers multiple seismic attribute data, so that multiple seismic attribute data are associated with source rock distribution attribute data. Therefore, when predicting the distribution attribute data of source rocks in deep depressions within the block under study, multiple seismic attribute data of deep depressions within the block under study can be substituted into the prediction model for source rocks in deep depressions for calculation. This can more accurately predict source rocks in deep depressions and improve the exploration and development efficiency of source rocks in deep depressions. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 A flowchart illustrating a method for predicting hydrocarbon source rocks in deep depressions, provided as an embodiment of this application;

[0041] Figure 2 The flowchart illustrates a method for predicting source rocks in deep depressions, provided in this application embodiment, which involves establishing a prediction model for source rocks in deep depressions based on multiple seismic attribute data and source rock distribution attribute data of the block under study.

[0042] Figure 3 A flowchart illustrating a method for obtaining multiple seismic attribute data and source rock distribution attribute data of a block under study in a method for predicting source rocks in a deep depression area, as provided in an embodiment of this application;

[0043] Figure 4 The flowchart illustrates a method for filtering multiple seismic attribute data and determining multiple strongly correlated seismic attribute data in a method for predicting source rocks in deep depressions provided in this application embodiment.

[0044] Figure 5 The flowchart illustrates a method for predicting source rocks in deep depressions, provided in this application embodiment. This method trains a preset learning model based on multiple strongly correlated seismic attribute data and source rock distribution attribute data to obtain a prediction model for source rocks in deep depressions.

[0045] Figure 6 This is a schematic diagram of a device for predicting source rocks in deep depressions, provided in an embodiment of this application.

[0046] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0047] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0048] Unless otherwise defined, all technical terms used in the embodiments of this application have the same meaning as commonly understood by one of ordinary skill in the art.

[0049] To make the technical solutions and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0050] Source rocks are rocks rich in organic matter that can generate and provide industrial quantities of oil and gas, and they have significant research value in petroleum exploration and development. Traditional source rock evaluation mainly relies on geochemical testing of core samples and cuttings or well logging data analysis. These methods can only determine the distribution characteristics of source rocks near oil wells and often fail to consider the planar heterogeneity of source rocks, resulting in extremely limited spatial scale for determining source rock properties. Due to the low exploration level and small number of wells in deep depressions, the available analytical data is insufficient, making it difficult to predict source rocks in many wellless areas.

[0051] In related technologies, seismic attribute data of deep depressions are mainly substituted into linear regression models to determine the distribution characteristics of source rocks in these areas. However, since the linear regression model only considers a single seismic attribute data point, the predicted distribution characteristics of source rocks differ significantly from the actual distribution characteristics, affecting the efficiency of exploration and development of source rocks in deep depressions.

[0052] To address the problems existing in related technologies, this application provides a method and apparatus for predicting source rocks in deep depressions, which can accurately predict source rocks in deep depressions and improve the exploration and development efficiency of source rocks in deep depressions.

[0053] Figure 1 A flowchart illustrating a method for predicting hydrocarbon source rocks in deep depressions, as provided in this application embodiment, is shown below. Figure 1 The method includes the following steps:

[0054] Step 101: Based on multiple seismic attribute data and source rock distribution attribute data of the block to be studied, establish a source rock prediction model for the deep depression area, where multiple seismic attribute data and source rock distribution attribute data are the input and output of the source rock prediction model for the deep depression area, respectively.

[0055] Since well logging is scarce or nonexistent in the deep depression area of ​​the block, to predict the distribution attributes of source rocks in the deep depression area by training a model, it is necessary to first train the model using source rock distribution attribute data obtained from existing well logging in the block and multiple seismic attribute data. The model is trained by using multiple seismic attribute data from well logging in the block as input and source rock distribution attribute data from well logging in the block as output, thus establishing a prediction model for source rocks in the deep depression area.

[0056] In some embodiments, the distribution properties of source rocks include lithology, thickness, and total organic carbon content.

[0057] In some embodiments, see Figure 2 Step 101 includes the following sub-steps:

[0058] Step 1011: Obtain multiple seismic attribute data and source rock distribution attribute data for the block to be studied.

[0059] In some embodiments, see Figure 3 Step 1011 includes the following sub-steps:

[0060] Step 10111: Obtain logging data, well logging data, and geochemical analysis data from all wells in the block to be studied, and determine the source rock distribution attributes of the block to be studied.

[0061] Step 10112: Obtain seismic data from each well within the block to be studied, and filter the seismic data from each well within the block to be studied to determine the seismic attribute data related to the distribution attributes of source rocks as multiple seismic attribute data of the block to be studied.

[0062] After obtaining seismic data from all wells in the block under study, the seismic data is first preliminarily screened to obtain seismic attribute data that are correlated with the distribution attributes of source rocks.

[0063] Step 1012: Preprocess the multiple seismic attribute data and source rock distribution attribute data of the block to be studied.

[0064] Before using multiple seismic attribute data and source rock distribution attribute data of the block under study, these data need to be preprocessed. Preprocessing includes checking the data and handling outliers and missing data to ensure the integrity and reasonableness of the data.

[0065] For example, if a certain item is missing in seismic attribute data arranged at fixed intervals, the missing item is filled in using adjacent data.

[0066] Step 1013: Filter multiple seismic attribute data to identify multiple strongly correlated seismic attribute data.

[0067] Understandably, since there are many types of seismic attribute data and their correlation with the distribution attributes of source rocks are different, it is necessary to filter multiple seismic attribute data and select the seismic attribute data that has a high correlation with the distribution attributes of source rocks as the model input when training the model.

[0068] In some embodiments, see Figure 4 Step 1013 includes the following sub-steps:

[0069] Step 10131: Obtain the correlation coefficient, first threshold, and second threshold for each seismic attribute and source rock distribution attribute.

[0070] In some embodiments, the correlation coefficient between each seismic attribute and the source rock distribution attribute is calculated using the Pearson correlation coefficient test. The Pearson correlation coefficient measures the linear correlation between two variables X and Y. The correlation coefficient ranges from -1 to 1, where 1 indicates a perfect positive correlation and -1 indicates a perfect negative correlation. The formula for calculating the correlation coefficient is as follows:

[0071]

[0072] Where, ρ X,Y σ is the correlation coefficient between X and Y; cov(X,Y) is the covariance between X and Y; σ X ,σ Y Let X and Y be the standard deviations; E(X) and E(Y) be the expected values ​​of X and Y, respectively.

[0073] For example, when calculating the correlation coefficient between each seismic attribute and the source rock distribution attribute, the seismic attribute is substituted into the position of X in the formula, and the source rock distribution attribute is substituted into the position of Y in the formula.

[0074] Step 10132: In response to the seismic attribute being greater than a first threshold, determine multiple transitional seismic attributes.

[0075] By setting a first threshold, seismic attribute data with high correlation to source rock distribution attributes are selected from multiple seismic attribute data. For example, taking lithology as an example of source rock distribution attributes, the correlation coefficients between multiple seismic attribute data and source rock lithology are shown in Table 1:

[0076] Table 1. Correlation coefficients between multiple seismic attribute data and source rock lithology.

[0077]

[0078] As shown in Table 1 above, when the first threshold is set to 0.15, several transitional seismic attributes can be identified. These attributes include average instantaneous frequency, average reflection intensity, average instantaneous phase, positive and negative sampling rate of change, total energy, and total amplitude. In other words, these six seismic attributes are strongly correlated with lithology, and therefore can be used as transitional seismic attributes for secondary screening in subsequent steps.

[0079] Step 10133: In response to the transitional earthquake attribute being less than the second threshold, determine multiple strongly correlated earthquake attributes.

[0080] After obtaining the transitional seismic attributes, a second screening is required based on a second threshold. This process necessitates using the formula from the Pearson correlation test to calculate the correlation coefficients between the transitional seismic attributes. This ensures that the resulting strongly correlated seismic attributes have a weak correlation and low mutual influence, thereby guaranteeing that the seismic attribute data used in training the model are relatively independent. For example, taking lithology in the source rock distribution attributes as an example, the correlation coefficients between transitional seismic attributes are shown in Table 2 below:

[0081] Table 2. Correlation coefficients between transitional earthquake attributes

[0082]

[0083] As shown in Table 2 above, when the second threshold is set to 0.5, the final obtained source rock lithology has several strongly correlated seismic attributes, including average instantaneous frequency, average reflection intensity, positive and negative sampling change rate, and total amplitude.

[0084] For example, the strongly correlated seismic attributes corresponding to the source rock lithology obtained by the final screening include the average instantaneous frequency, average reflection intensity, positive and negative sampling change rate, and total amplitude; the strongly correlated seismic attributes corresponding to the source rock thickness include the average instantaneous frequency, average energy, and average amplitude; and the strongly correlated seismic attributes corresponding to the total organic carbon content of the source rock include the positive and negative change rate, average peak amplitude, energy half-life, and average reflection intensity.

[0085] Step 10134: Based on multiple strongly correlated earthquake attributes, obtain multiple strongly correlated earthquake attribute data.

[0086] Based on the selected strongly correlated earthquake attribute types, the data of these strongly correlated earthquake attributes are obtained for use in the next step of model training.

[0087] Step 1014: Based on multiple strongly correlated seismic attribute data and source rock distribution attribute data, train the preset learning model to obtain a source rock prediction model for deep depression areas.

[0088] Multiple strongly correlated seismic attribute data with high correlation to the distribution attributes of source rocks are used as input to a pre-set learning model, and the distribution attribute data of source rocks are used as output to train the pre-set learning model, thereby obtaining a source rock prediction model for deep depression areas.

[0089] In some embodiments, the number of preset learning models is multiple.

[0090] In the embodiments of this application, the preset learning model includes gradient boosting decision tree, k-nearest neighbor algorithm, artificial neural network and support vector machine.

[0091] Gradient boosting decision trees are additive models based on the boosting ensemble concept. Compared to basic decision tree algorithms, they improve generalization ability through accumulation. During training, a forward distribution algorithm is used for greedy learning, where each iteration learns a regression tree to fit the residuals between the predictions of the previous t-1 trees and the true values ​​of the training samples. The calculation formula is:

[0092]

[0093] In the formula: —The final predicted values ​​of the target variables (source rock lithology, thickness, total organic carbon content) using machine learning algorithms;

[0094] f t (x i — For the i-th sample x i The t-th regression tree model.

[0095] Its iterative process is represented as:

[0096]

[0097] Artificial neural network models simulate signal transmission between neurons through mathematical expressions, thereby establishing a nonlinear equation with input-output relationships that can be visualized through a network. This model includes an input layer, multiple hidden layers, and an output layer. Here, the input layer refers to the seismic attribute parameters used to predict the properties of source rocks; its number of nodes equals the number of input seismic attribute parameters. The hidden layers abstract the input parameters, ultimately decomposing the input into multiple classes of valuable feature information. A neuron is the basic unit in a neural network; it receives multiple input signals, calculates and generates an output signal, which is then transmitted to other neurons. The formula for calculating a neuron is:

[0098]

[0099] Where: y is the output value of the neuron; g(...) is the activation function; w i For x i The weights; x i is the i-th input value; b is the bias.

[0100] The activation function of a neural network is a non-linear transformation of neurons, used to enhance the expressiveness of the network. Commonly used activation functions include the Tanh function, Maxout function, and sigmoid function. The formula for the Tanh function is:

[0101]

[0102] In some embodiments, see Figure 5 Step 1014 includes the following sub-steps:

[0103] Step 10141: Based on multiple strongly correlated seismic attribute data and source rock distribution attribute data, train and test each preset learning model to obtain the determination coefficient of each learning model, wherein the multiple strongly correlated seismic attribute data and source rock distribution attribute data are the input and output of each preset learning model, respectively.

[0104] After training the preset learning model, the trained models are tested using a test dataset. The coefficient of determination is calculated based on the predicted values ​​of source rock distribution attributes output by the model and the actual values ​​of source rock distribution attributes corresponding to multiple seismic attributes in the test dataset. The model with the best prediction effect on source rock lithology, thickness and total organic carbon content is determined based on the coefficient of determination.

[0105] In some embodiments, step 10141 specifically includes: dividing multiple strongly correlated seismic attribute data and source rock distribution attribute data into a training data set and a test data set; training each preset learning model based on the training data set to obtain a trained learning model; and testing the trained learning model based on the test data set to obtain the determination coefficient of each learning model.

[0106] In some embodiments, the ratio of the training dataset to the test dataset can be 8:2.

[0107] In some embodiments, the coefficient of determination R 2 The value range of the coefficient of determination is [0,1]. A larger coefficient of determination indicates a better prediction effect. The formula for its calculation is:

[0108]

[0109] In the formula: y i The true values ​​of the target variables (source rock lithology, thickness, TOC); The predicted values ​​of the target variables (source rock lithology, thickness, TOC) are obtained using machine learning algorithms; This is the average of the true values ​​of all target variables (source rock lithology, thickness, TOC).

[0110] Step 10142: The preset learning model corresponding to the maximum value of the coefficient of determination is determined as the prediction model for hydrocarbon source rocks in the deep depression area.

[0111] Coefficient of determination R 2 The larger the value, the higher the accuracy of the model's predictions. For example, the coefficient of determination R of multiple trained models... 2 As shown in Table 3 below:

[0112] Table 3. Model Determination Coefficients

[0113]

[0114] As shown in Table 3, the most accurate prediction model for the lithology of source rocks is the one built using gradient boosting decision trees; the most accurate prediction model for the thickness of source rocks is the one built using artificial neural networks; and the most accurate prediction model for the total organic carbon content of source rocks is the one built using gradient boosting decision trees.

[0115] Step 102: Obtain multiple seismic attribute data for the deep depression within the block to be studied.

[0116] After establishing a source rock prediction model for the deep depression area using existing source rock distribution attribute data from well logging in this block and multiple seismic attribute data, multiple seismic attribute data of the deep depression area in the block are used as inputs to the source rock prediction model for the deep depression area in order to complete the prediction of the source rock distribution attributes in the deep depression area.

[0117] Step 103: Substitute multiple seismic attribute data of the deep depression area within the block to be studied into the deep depression area source rock prediction model to obtain the distribution attribute data of the deep depression area source rock.

[0118] Since the training process involves using multiple seismic attribute data as input to the preset model and source rock distribution attribute data as output, after training is completed and a source rock prediction model for the deep depression area is obtained, substituting the multiple seismic attribute data of the deep depression area into the prediction model yields the source rock distribution attribute data for the deep depression area. Based on this data, the distribution of source rocks in the deep depression area can be determined, providing data support for the exploration and development of source rocks in the deep depression area.

[0119] Therefore, the prediction method for source rocks in deep depressions provided in this application establishes a prediction model for source rocks in deep depressions based on multiple seismic attribute data and source rock distribution attribute data of the block under study. Since the input of this prediction model is multiple seismic attribute data, in other words, the prediction model considers multiple seismic attribute data, so that multiple seismic attribute data are associated with source rock distribution attribute data. Therefore, when predicting the distribution attribute data of source rocks in deep depressions within the block under study, multiple seismic attribute data of deep depressions within the block under study can be substituted into the prediction model for source rocks in deep depressions for calculation. This can more accurately predict source rocks in deep depressions and improve the exploration and development efficiency of source rocks in deep depressions.

[0120] Figure 6 A schematic diagram of a prediction device for hydrocarbon source rocks in a deep depression area provided in this application embodiment is shown below. Figure 6 The device 600 includes:

[0121] Modeling module 601 is used to establish a deep depression hydrocarbon source rock prediction model based on multiple seismic attribute data and hydrocarbon source rock distribution attribute data of the block to be studied. The multiple seismic attribute data and hydrocarbon source rock distribution attribute data are the input and output of the deep depression hydrocarbon source rock prediction model, respectively.

[0122] Module 602 is used to acquire multiple seismic attribute data of the deep depression area within the block under study;

[0123] Module 603 is obtained, which is used to substitute multiple seismic attribute data of the deep depression area within the block to be studied into the deep depression area source rock prediction model to obtain the distribution attribute data of the deep depression area source rock.

[0124] In some embodiments, the modeling module includes:

[0125] The first acquisition submodule is used to acquire multiple seismic attribute data and source rock distribution attribute data of the block to be studied;

[0126] The first determination submodule is used to filter multiple seismic attribute data and determine multiple strongly correlated seismic attribute data.

[0127] The first submodule is used to train a preset learning model based on multiple strongly correlated seismic attribute data and source rock distribution attribute data to obtain a source rock prediction model for deep depression areas.

[0128] In some embodiments, the first acquisition submodule includes:

[0129] The second acquisition submodule is used to acquire well logging data, well logging data and geochemical analysis data of all wells in the block to be studied, and to determine the distribution attribute data of source rocks in the block to be studied.

[0130] The third acquisition submodule is used to acquire seismic data from each well in the block to be studied, and to filter the seismic data from each well in the block to be studied to determine the seismic attribute data related to the distribution attributes of source rocks as multiple seismic attribute data of the block to be studied.

[0131] In some embodiments, the modeling module further includes:

[0132] The preprocessing submodule is used to preprocess multiple seismic attribute data and source rock distribution attribute data of the block to be studied.

[0133] In some embodiments, the first determining submodule includes:

[0134] The fourth acquisition submodule is used to acquire the correlation coefficient, first threshold and second threshold of each seismic attribute and source rock distribution attribute;

[0135] The second determination submodule is used to determine multiple transitional seismic attributes in response to a seismic attribute being greater than a first threshold.

[0136] The third determination submodule is used to determine multiple strongly correlated earthquake attributes in response to the transitional earthquake attribute being less than the second threshold.

[0137] The second submodule is used to obtain multiple strongly correlated earthquake attribute data based on multiple strongly correlated earthquake attributes.

[0138] In some embodiments, the number of preset learning models is multiple, and the first sub-module includes:

[0139] The third submodule is used to train and test each preset learning model based on multiple strongly correlated seismic attribute data and source rock distribution attribute data, and to obtain the determination coefficient of each learning model. The multiple strongly correlated seismic attribute data and source rock distribution attribute data are the input and output of each preset learning model, respectively.

[0140] The fourth determination submodule is used to determine the preset learning model corresponding to the maximum value of the determination coefficient as the prediction model for hydrocarbon source rocks in the deep depression area.

[0141] In some embodiments, the third submodule includes:

[0142] The sub-module is used to divide multiple strongly correlated seismic attribute data and source rock distribution attribute data into training data sets and test data sets.

[0143] The fourth submodule is used to train each preset learning model based on the training data set to obtain the trained learning model.

[0144] The fifth submodule is used to test the trained learning models based on the test data set and obtain the determination coefficient of each learning model.

[0145] In some embodiments, the distribution properties of source rocks include lithology, thickness, and total organic carbon content.

[0146] Therefore, the prediction device for source rocks in deep depressions provided in this application establishes a prediction model for source rocks in deep depressions based on multiple seismic attribute data and source rock distribution attribute data of the block under study. Since the input of the prediction model is multiple seismic attribute data, in other words, the prediction model considers multiple seismic attribute data, so that multiple seismic attribute data are associated with source rock distribution attribute data. Therefore, when predicting the distribution attribute data of source rocks in deep depressions within the block under study, multiple seismic attribute data of deep depressions within the block under study can be substituted into the prediction model for source rocks in deep depressions for calculation. This can more accurately predict source rocks in deep depressions and improve the exploration and development efficiency of source rocks in deep depressions.

[0147] This application also provides a non-volatile readable storage medium storing at least one program that is loaded and executed by a processor to implement the method for predicting source rocks in deep depressions in any embodiment.

[0148] In this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance. The term "multiple" refers to two or more unless otherwise expressly defined.

[0149] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the application disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only.

[0150] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A method for predicting hydrocarbon source rocks in deep depressions, characterized in that, The method includes: Based on multiple seismic attribute data and source rock distribution attribute data of the block under study, a source rock prediction model for the deep depression area is established, wherein the multiple seismic attribute data and the source rock distribution attribute data are the input and output of the source rock prediction model for the deep depression area, respectively. Acquire multiple seismic attribute data of the deep depression within the block to be studied; Substitute multiple seismic attribute data of the deep depression area within the block under study into the hydrocarbon source rock prediction model of the deep depression area to obtain the hydrocarbon source rock distribution attribute data of the deep depression area.

2. The method for predicting source rocks in deep depressions according to claim 1, characterized in that, The step of establishing a source rock prediction model for deep depression areas based on multiple seismic attribute data and source rock distribution attribute data of the block under study includes: Acquire multiple seismic attribute data and source rock distribution attribute data of the block to be studied; The multiple seismic attribute data were filtered to identify several strongly correlated seismic attribute data. Based on the multiple strongly correlated seismic attribute data and the hydrocarbon source rock distribution attribute data, a preset learning model is trained to obtain the hydrocarbon source rock prediction model for the deep depression area.

3. The method for predicting source rocks in deep depressions according to claim 2, characterized in that, The acquisition of multiple seismic attribute data and source rock distribution attribute data of the block under study includes: Obtain logging data, well logging data, and geochemical analysis data from all wells in the block under study to determine the source rock distribution attributes of the block under study. Seismic data from each well within the study block are acquired, and the seismic data from each well within the study block are filtered to determine the seismic attribute data related to the distribution attributes of the source rock as multiple seismic attribute data of the study block.

4. The method for predicting source rocks in deep depressions according to claim 2, characterized in that, Before filtering the multiple seismic attribute data to determine multiple strongly correlated seismic attribute data, the method further includes: Preprocessing is performed on multiple seismic attribute data and source rock distribution attribute data of the block to be studied.

5. The method for predicting source rocks in deep depressions according to claim 2, characterized in that, The process of filtering the multiple seismic attribute data to determine multiple strongly correlated seismic attribute data includes: Obtain the correlation coefficient, first threshold, and second threshold for each seismic attribute and source rock distribution attribute; In response to the seismic attribute being greater than the first threshold, multiple transitional seismic attributes are determined; In response to the transitional seismic attribute being less than the second threshold, multiple strongly correlated seismic attributes are determined; Based on the aforementioned multiple strongly correlated earthquake attributes, multiple strongly correlated earthquake attribute data were obtained.

6. The method for predicting source rocks in deep depressions according to claim 2, characterized in that, The number of preset learning models is multiple. The process of training the preset learning models based on the multiple strongly correlated seismic attribute data and the hydrocarbon source rock distribution attribute data to obtain the hydrocarbon source rock prediction model for the deep depression area includes: Based on the multiple strongly correlated seismic attribute data and the hydrocarbon source rock distribution attribute data, each preset learning model is trained and tested to obtain the determination coefficient of each learning model, wherein the multiple strongly correlated seismic attribute data and the hydrocarbon source rock distribution attribute data are the input and output of each preset learning model, respectively. The preset learning model corresponding to the maximum value of the coefficient of determination is determined as the prediction model for hydrocarbon source rocks in the deep depression area.

7. The method for predicting source rocks in deep depressions according to claim 6, characterized in that, Based on the multiple strongly correlated seismic attribute data and the hydrocarbon source rock distribution attribute data, each preset learning model is trained and tested to obtain the determination coefficients of each learning model, including: The multiple strongly correlated seismic attribute data and the hydrocarbon source rock distribution attribute data are divided into a training data set and a test data set; Based on the training data set, each preset learning model is trained to obtain the trained learning model; Based on the test data set, the trained learning models are tested to obtain the determination coefficient of each learning model.

8. The method for predicting source rocks in deep depressions according to claim 1, characterized in that, The distribution attributes of the source rocks include lithology, thickness, and total organic carbon content.

9. A device for predicting hydrocarbon source rocks in deep depressions, characterized in that, The device includes: The modeling module is used to establish a deep depression hydrocarbon source rock prediction model based on multiple seismic attribute data and hydrocarbon source rock distribution attribute data of the block to be studied, wherein the multiple seismic attribute data and the hydrocarbon source rock distribution attribute data are the input and output of the deep depression hydrocarbon source rock prediction model, respectively. The acquisition module is used to acquire multiple seismic attribute data of the deep depression area within the block under study; The module is used to substitute multiple seismic attribute data of the deep depression area within the block under study into the hydrocarbon source rock prediction model of the deep depression area, thereby obtaining the hydrocarbon source rock distribution attribute data of the deep depression area.

10. A non-volatile readable storage medium, characterized in that, The non-volatile readable storage medium stores at least one program, which is loaded and executed by a processor to implement the prediction method for source rocks in deep depressions as described in any one of claims 1 to 8.

Images