Method and system for predicting micro-cracks of tight sandstone reservoir and storage medium

Through drilling cores, imaging logging and cast sheet image analysis, combined with Bayesian optimization and random forest algorithm, a discriminant model of micro-fractures in dense sandstone reservoirs was established, which solved the problem of micro-fracture identification of dense sandstone reservoirs and improved the accuracy of prediction Sex and credibility.

CN118938348AActive Publication Date: 2024-11-12CHINA UNIV OF GEOSCIENCES (WUHAN)
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202410927941.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-11
Publication Date
2024-11-12
Estimated Expiration
2044-07-11

AI Technical Summary

Technical Problem

The prior art is difficult to accurately identify and predict the causes, development density and dissolution products of micro-cracks in dense sandstone reservoirs, resulting in difficulty in accurately predicting physical dessert areas in dense sandstone reservoirs.

Method used

Through drilling cores, imaging logging images and cast sheet image analysis, combined with Bayesian optimization algorithm and random forest classification algorithm, a discriminant model of the combination type of micro-fractures, development density and erosion products is established to improve the recognition accuracy.

Benefits of technology

The scientific, integrated and systematic prediction of micro-cracks in tight sandstone reservoirs has been achieved, the operation process has been simplified, and the dependence on professional and technical personnel has been reduced. It is suitable for dense sandstone reservoir prediction in different regions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118938348A_ABST
    Figure CN118938348A_ABST
Patent Text Reader

Abstract

The invention provides a tight sandstone reservoir microfracture identification method and system and a storage medium, and belongs to the technical field of petroleum and natural gas exploration and development. According to the identification method, combination types of'microfracture causes, development density and corrosion products' of the compact sandstone are divided by comprehensively applying drilling cores, imaging logging, casting body slices and image analysis software, and discrimination models of different combination types are established by applying conventional logging data through a Bayesian optimized random forest classification algorithm, so that the identification accuracy is improved; the problem of difficulty in exploration and development of the compact sandstone reservoir at present is essentially solved, so that a compact sandstone reservoir microfracture prediction technology is more scientific, comprehensive and systematic, and the method can be suitable for prediction of compact sandstone reservoirs in different petroliferous basins.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of oil and gas exploration and development, and in particular relates to a prediction method, system and storage medium for micro-cracks in a dense sandstone reservoir. Background Art

[0002] Oil and gas exploration practice has proved that the development degree of microcracks in tight sandstone reservoirs has become one of the key factors in the development of physical sweet spots. The presence of microcracks can not only serve as migration channels for fluids in tight sandstone reservoirs, but also promote the dissolution of unstable minerals, thereby forming different combinations of types and distributions of dissolution products with different abundances. Tight sandstone reservoirs have undergone complex diagenetic transformations, resulting in complex identification of the causes of microcracks and distribution characteristics of dissolution products, which seriously restricts the accurate prediction of physical sweet spots in tight sandstone reservoirs. At present, a quantitative prediction method for the relationship between the genesis type, development density and dissolution products of microcracks in tight sandstones has yet to be established.

[0003] At present, the identification methods of microfractures in tight sandstone reservoirs mainly include: ① geological identification: identifying microfractures through field outcrop profiles and rock thin sections, and statistically analyzing microfracture-related parameters (such as aperture and filling degree, etc.); ② logging identification: curve fitting based on the conventional logging curve response characteristics of microfractures, and establishing empirical formulas for identification; ③ experimental identification: using core diversion test method, CT scanning method and acoustic emission test method to identify microfractures; ④ seismic identification: the presence of cracks will enhance the anisotropy of the formation, and identification is based on the abnormal seismic reflection characteristics generated in the seismic waves.

[0004] The above methods are currently important technical means for microfracture identification, but they all have certain shortcomings: ① The limitation of geological identification is that the characteristic parameters of microfractures are obtained through actual core samples, while the collection of drilling and field outcrop core samples is limited, which is difficult to meet the actual exploration and development needs; ② Well logging identification can be divided into imaging logging and conventional logging identification. The former method is intuitive and has high resolution, and can clearly determine the characteristic parameters of microfractures, but it is expensive and difficult to obtain in a comprehensive coverage of the work area; the latter needs to consider the accuracy and matching of conventional logging curves to identify microfractures, and requires a large amount of data to train the model; ③ The experimental method can identify the characteristic parameters of microfracture development and determine the main controlling factors of the development of dominant microfractures, but it is difficult to evaluate the impact of microfractures on the heterogeneity of tight sandstone reservoirs, and its guiding significance for oil and gas exploration and development is relatively limited; ④ The resolution of seismic identification is low, and it can often only carry out identification research on macroscopic and large-scale fractures, resulting in low accuracy and strong subjectivity in the identification results. Therefore, establishing a comprehensive prediction method for the causes of microfractures in tight sandstone reservoirs, development density and dissolution product combination, and effectively improving the credibility of microfracture identification in tight sandstone reservoirs has become a technical issue that is of general concern in the industry and urgently needs to be solved. Summary of the invention

[0005] The purpose of the present invention is to provide a prediction method, system and storage medium for microcracks in tight sandstone reservoirs in response to the above-mentioned deficiencies in the prior art, which can essentially solve the current difficult problems in the exploration and development of tight sandstone oil and gas reservoirs, and can be applicable to the identification and prediction of microcracks in tight sandstones in different regions.

[0006] To achieve the above object, the present invention adopts the following technical solution:

[0007] The first object of the present invention is to provide a method for identifying microcracks in a tight sandstone reservoir, comprising the following specific steps:

[0008] S1. Get the original data

[0009] S11. Obtain the genesis type of microcracks and the characteristic parameters of microcrack development based on the analysis of drilling cores, imaging logging images and casting thin section images;

[0010] S12, obtaining the characteristics of the dissolution products based on the casting thin section image analysis;

[0011] S2. Data Processing

[0012] Based on the three evaluation parameters of the genesis type, development characteristics and abundance of dissolved products of microcracks in tight sandstone reservoirs obtained in step S1, a quantitative classification standard for each parameter is established, and the original data with the combination type of labels "microcrack genesis, development density and dissolved products" are uniformly classified.

[0013] S3, build data set

[0014] Conventional logging curves are preprocessed by means of outlier removal, curve splicing, depth relocation and logging curve standardization, and logging curves that are sensitive to the combination of "microfracture genesis type, development density and dissolution products" are selected to obtain sensitive logging curve data corresponding to different combinations of "microfracture genesis type, development density and dissolution products", and a logging database of different combinations of "microfracture genesis type, development density and dissolution products" is established;

[0015] S4, performing missing value processing and outlier detection processing on the well logging database obtained in step S3 to construct a data set, and dividing the data set into a training data set, a verification data set and a test data set;

[0016] S5. Select the optimal hyperparameters of the random forest classification algorithm through the Bayesian optimization algorithm;

[0017] The Bayesian optimization algorithm optimizes the hyperparameters of the random forest algorithm, including the number of random forest trees and the minimum number of leaf nodes. The Bayesian optimization includes the following steps:

[0018] S51, define a hyperparameter space, determine the hyperparameters to be optimized and their value ranges;

[0019] S52, define the proxy model, select an appropriate proxy model, usually using Gaussian process;

[0020] S53, initializing the sample set, randomly selecting a set of initial samples in the hyperparameter space, and training the proxy model based on the performance evaluation of these samples;

[0021] S54, iterative optimization, based on the prediction of the proxy model, select the next hyperparameter combination that is most likely to improve performance; evaluate the selected hyperparameter combination on the real model and record its performance; update the proxy model with new samples and update the estimate of the objective function in the hyperparameter space; repeat the above steps until the preset number of iterations is reached or the stopping criterion is met;

[0022] S55, output optimal hyperparameters;

[0023] S6. Substitute the selected optimal hyperparameters into the random forest algorithm training model to establish a discriminant model for the combination type of “microcracks genesis, development density and dissolution products”;

[0024] S7. Perform cross-validation on the training set and evaluate the model prediction performance on the test set.

[0025] Furthermore, in step S11, the genesis types of the microcracks include tectonic genesis and diagenetic genesis, and the development characteristic parameters of the microcracks include opening, length and surface density. According to the surface density index, they are divided into low-density microcrack combinations, with surface density <5 cm / cm 2 ; Medium density micro crack combination, 5cm / cm 2 <Surface density<10cm / cm 2 ; and high-density microcrack combination, surface density> 10cm / cm 2 .

[0026] Furthermore, in step S12, the characteristics of the dissolution products include product type, product content and development location. The abundance of the dissolution products is quantitatively counted by Image J image analysis software, and divided into a low dissolution product combination, with a product abundance of <3%; a medium dissolution product combination, with a product abundance of 3% < 8%; and a high dissolution product combination, with a product abundance of >8%.

[0027] Furthermore, in step S2, the combination types are divided into 6 categories, specifically: Class I "structural-diagenetic microfractures, high development density, low dissolution products" combination, Class II "structural-diagenetic microfractures, medium development density, medium dissolution products" combination, Class III "structural-diagenetic microfractures, low development density, high dissolution products" combination, Class IV "diagenetic microfractures, high development density + low dissolution products" combination, Class V "diagenetic microfractures, medium development density, medium dissolution products" combination and Class VI "diagenetic microfractures, low development density, high dissolution products" combination.

[0028] Furthermore, in step S3, the logging curves include a natural potential curve, a natural gamma curve, an acoustic time difference curve, a supplementary neutron curve, a supplementary density curve, a deep lateral resistivity curve and a shallow lateral resistivity curve.

[0029] Furthermore, in step S3, the process of removing outliers is to delete the measured values ​​in a group of measured values ​​at a certain sample point whose deviation from the average value exceeds three times the standard deviation;

[0030] The process of curve splicing is to select the logging value of the formation at the same depth measured by two logging curves as the comparison standard, and make the two logging curves overlap by comparing and moving the logging curves;

[0031] The process of depth homing is to select the GR curve with high vertical resolution and obvious characteristic marks as the standard curve, and determine the movement of other logging curves relative to the standard curve by comparison to complete the depth homing of the logging curve;

[0032] The process of logging curve standardization is to calibrate the logging curves of different wells, select the standard layer of key wells with homogeneous formation and stable distribution, determine the correction amount required for each logging curve of other wells through the peak value of the logging curve frequency distribution histogram of the key drilling standard layer, and complete the standardization process of logging curves.

[0033] Furthermore, in step S4, the process of missing value processing is to use the code in MATLAB software to traverse the original logging data column data to see if there is more than 20% missing, and if missing, directly delete it; the process of outlier detection processing is to select the sigma detection method to perform outlier detection processing on the data.

[0034] Furthermore, in step S4, the ratio of the training data set, the validation data set, and the test data set is 8:1:1.

[0035] The second object of the present invention is to provide a system for identifying microcracks in a tight sandstone reservoir, wherein the system is used to implement the above-mentioned method for identifying microcracks in a tight sandstone reservoir, and at least comprises:

[0036] The data acquisition module is configured to obtain the original data of the combination type of microcracks with the label "microcrack genesis, development density and dissolution product"; specifically, by using the three evaluation parameters of the genesis type, development characteristics and abundance of dissolution products of microcracks in tight sandstone reservoirs, a quantitative classification standard for each parameter is established, and the original data of the combination type with the label "microcrack genesis, development density and dissolution product" is uniformly classified;

[0037] The microcracks have the following types of genesis: tectonic genesis and diagenetic genesis; the microcracks have the following development characteristic parameters: opening, length and surface density; and the microcracks are divided into low-density microcrack combination, medium-density microcrack combination and high-density microcrack combination according to the surface density index;

[0038] The characteristics of the dissolution products include product type, product content and development location. The abundance of the dissolution products is quantitatively counted by Image J image analysis software, and the dissolution products are divided into low dissolution product combination, medium dissolution product combination and high dissolution product combination according to the abundance index of the dissolution products.

[0039] The abnormality elimination module is configured to process the logging curves by means of abnormal value elimination, curve splicing, depth relocation and logging curve standardization, select the logging curves that are sensitive to the combination type of "microfracture genesis type, development density and dissolution product", obtain the sensitive logging curve data corresponding to different combinations of "microfracture genesis type, development density and dissolution product", and establish a logging database of different combinations of "microfracture genesis type, development density and dissolution product";

[0040] A data set construction module is configured to perform missing value processing and outlier detection processing on the well logging database to construct a data set, and divide the data set into a training data set, a validation data set and a test data set;

[0041] The classification and discrimination module is configured to use the random forest classification algorithm selected by the Bayesian optimization algorithm as the discrimination model of the combination type of "microfracture genesis, development density and dissolution products", input the sample labels and logging data of different combination types of "microfracture genesis type, development density and dissolution products" into the model, and randomly divide the logging data into a training data set, a validation data set and a test data set at a ratio of 8:1:1. After the model training output, the model classification confusion matrix is ​​obtained, and the classification result is obtained through the classification accuracy displayed by the confusion matrix;

[0042] The recognition module is configured to use the trained discrimination model to recognize the combination type of "microcrack genesis, development density and dissolution product" of the microcracks to be tested.

[0043] The third object of the present invention is to provide a computer-readable storage medium, wherein the computer-readable storage medium stores a program, and the program can be executed by one or more processors to implement the above-mentioned method for identifying microcracks in tight sandstone reservoirs.

[0044] Compared with the prior art, the technical solution provided by the present invention has the following beneficial effects:

[0045] (1) The present invention provides a method for identifying microfractures in tight sandstone reservoirs. By comprehensively using drilling cores, imaging logging, cast thin sections and image analysis software, the combination types of "microfracture genesis, development density and dissolution products" of tight sandstone are divided. Conventional logging data are used to establish a discrimination model of different combination types through a Bayesian optimized random forest classification algorithm, thereby improving the recognition accuracy and essentially solving the current difficult problems in the exploration and development of tight sandstone reservoirs. This makes the microfracture prediction technology of tight sandstone reservoirs more scientific, comprehensive and systematic, and can be applied to the prediction of tight sandstone reservoirs in different oil and gas basins.

[0046] (2) Compared with traditional methods, this method simplifies the operation process, reduces the dependence on professional technicians, and helps to quickly identify microfractures in tight sandstone reservoirs in a wider range of application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 This is a technical flow chart of a method for identifying micro-cracks in a tight sandstone reservoir provided by the present invention;

[0048] Figure 2 The present invention provides a diagram of the genesis and development characteristics of microcracks in the tight sandstone reservoir of the Shihezi Formation in the Ordos Basin. In the diagram, (AB) imaging logging, structural microcracks; (CD) core image, structural microcracks; (E) casting thin section image, structural microcracks; (EF) casting thin section image, diagenetic microcracks;

[0049] Figure 3 This is a combined type diagram of "micro-crack genesis, development density and dissolution products" of tight sandstone of Shihezi Formation in Ordos Basin provided by the present invention;

[0050] Figure 4 It is a schematic diagram of establishing a discriminant model based on the random forest algorithm provided by the present invention;

[0051] Figure 5 It is a basic principle diagram of the Bayesian optimization hyperparameter discrimination model provided by the present invention;

[0052] Figure 6a to Figure 6c It is a confusion matrix diagram of the data prediction results of the random forest classification model optimized by Bayesian method provided by the present invention;

[0053] Figure 7 This is a recognition result diagram of the combination type of "micro-crack genesis, development density and dissolution products" of the tight sandstone of the Shihezi Formation in the Ordos Basin provided by the present invention;

[0054] Figure 8 The present invention is a schematic structural diagram of a system for identifying micro-cracks in a dense sandstone reservoir. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical scheme and advantages of the present invention clearer, the specific implementation of the present invention is further described in detail below in conjunction with specific embodiments and drawings. If no specific technology or conditions are specified in the embodiments, the technology or conditions described in the literature in the art or the product instructions are used.

[0056] like Figure 1 FIG. 1 is a flow chart of a method for identifying microcracks in a tight sandstone reservoir according to the present invention, which comprises the following steps:

[0057] (1) Obtaining raw data

[0058] The genesis type of microcracks and the characteristic parameters of microcrack development are obtained based on the analysis of drilling cores, imaging logging images and casting thin section images;

[0059] The characteristics of the dissolution products were obtained based on the analysis of the casting thin section images;

[0060] (2) Data processing

[0061] Based on the three evaluation parameters of the genesis type, development characteristics and abundance of dissolved products of microcracks in tight sandstone reservoirs obtained in step S1, a quantitative classification standard for each parameter is established, and the original data with the combination type of labels "microcrack genesis, development density and dissolved products" are uniformly classified.

[0062] (3) Constructing a dataset

[0063] Conventional logging curves are preprocessed by means of outlier removal, curve splicing, depth relocation and logging curve standardization, and logging curves that are sensitive to the combination of "microfracture genesis type, development density and dissolution products" are selected to obtain sensitive logging curve data corresponding to different combinations of "microfracture genesis type, development density and dissolution products", and a logging database of different combinations of "microfracture genesis type, development density and dissolution products" is established;

[0064] (4) performing missing value processing and outlier detection processing on the well logging database obtained in step (3) to construct a data set, and dividing the data set into a training data set, a validation data set, and a test data set;

[0065] (5) Select the optimal hyperparameters of the random forest classification algorithm through the Bayesian optimization algorithm;

[0066] The Bayesian optimization algorithm optimizes the hyperparameters of the random forest algorithm, including the number of random forest trees and the minimum number of leaf nodes. The Bayesian optimization includes the following steps:

[0067] (51) Define the hyperparameter space and determine the hyperparameters to be optimized and their value ranges;

[0068] (52) Define the surrogate model and select an appropriate surrogate model, usually using a Gaussian process;

[0069] (53) Initialize the sample set, randomly select a set of initial samples in the hyperparameter space, and train the proxy model based on the performance evaluation of these samples;

[0070] (54) Iterative optimization: based on the prediction of the proxy model, select the next hyperparameter combination that is most likely to improve performance; evaluate the selected hyperparameter combination on the real model and record its performance; update the proxy model using new samples and update the estimate of the objective function in the hyperparameter space; repeat the above steps until the preset number of iterations is reached or the stopping criterion is met;

[0071] (55) Output optimal hyperparameters;

[0072] (6) Substitute the selected optimal hyperparameters into the random forest algorithm training model to establish a discriminant model for the combination type of “microcrack genesis, development density and dissolution products”;

[0073] (7) Perform cross-validation on the training set and evaluate the model prediction performance on the test set.

[0074] The technical terms in the present invention are explained as follows:

[0075] Drilling core means using coring tools to get underground rocks to the surface in blocks during the drilling process. This kind of rock block is called drilling core;

[0076] Development density, which means the width of cracks per unit area;

[0077] The abundance of dissolution products refers to the content of secondary products formed by mineral dissolution in the core sample;

[0078] Well logging curve refers to the curve formed during the logging process, which can reflect the characteristics of different lithologies and stratigraphic positions;

[0079] Sensitive logging curve means in this patent a logging curve that has a sensitive response to the changing characteristics of the combination of microfractures and their dissolution products;

[0080] Well logging data refers to the data values ​​obtained from different well logging curves.

[0081] Example 1

[0082] The specific technical scheme of the present invention is explained by taking the tight sandstone reservoir of the Shihezi Formation of the Permian System in the Fuxian area of ​​the Ordos Basin as an example.

[0083] Step 1: Development characteristics of micro-fractures in tight sandstone

[0084] Based on the comprehensive analysis of drilling cores, imaging logging images, and casting thin section images, it is believed that the tight sandstone reservoirs of the Shihezi Formation have two types of microcracks: structural microcracks and diagenetic microcracks (such as Figure 2 As shown in Table 1, the opening of the structural microcracks is greater than 100 μm, the length is greater than 10 mm, and the surface density is generally greater than 5 cm / cm 2 ; The opening of diagenetic microcracks is less than 100μm, the length is less than 10mm, and the surface density is generally less than 5cm / cm 2 According to the microcrack surface density index, it is divided into low density (<5cm / cm 2 ), medium density (5~10cm / cm 2 ) and high density (>10cm / cm 2 ) Microcrack combination.

[0085] Table 1. Statistical table of microcracks genesis types and characteristic parameters

[0086] Microcracks Cause Type Opening(μm) Length(mm) <![CDATA[Areal density (cm / cm 2 )]]> Tectonic causes >100 >10 >5 Diagenesis <100 <10 <5

[0087] Step 2: Characteristics of Dissolution Products

[0088] Using cast thin section image analysis, the dissolution mineral types of the tight sandstone reservoir of the Shihezi Formation mainly include feldspar and rock fragments particles and a small amount of carbonate cement, and the dissolution products mainly include quartz, clay and carbonate minerals. The abundance of dissolution products was quantitatively counted by ImageJ image analysis software, and the dissolution product combinations were divided into small (<3%), medium (3%-8%), and large (>8%) dissolution product combinations according to the abundance index of dissolution products.

[0089] Step 3: Combination type of “microcrack genesis, development density and dissolution products”

[0090] Taking into account the three indicators of microcrack genesis, microcrack development density and dissolution product abundance in the tight sandstone reservoir of the Shihezi Formation, the combination types of "microcrack genesis, development density and dissolution product" of the tight sandstone of the Shihezi Formation are divided into 6 categories (such as Figure 3 ① Type I combination of “structural-diagenetic microfractures, high development density, and low dissolution products” ( Figure 3 ② Type II combination of "structural-diagenetic microfractures, medium development density, medium dissolution products" ( Figure 3Middle B); ③ Type III “structural-diagenetic microfractures, low development density, high dissolution products” combination ( Figure 3 Middle C); ④ Type IV "diagenetic microcracks, high development density + low dissolution products" combination ( Figure 3 ⑤ Type V “diagenetic microcracks, medium development density, medium dissolution products” combination ( Figure 3 Middle E); ⑥ Type VI “diagenetic microcracks, low development density, high dissolution products” combination ( Figure 3 Middle F).

[0091] Step 4: Well logging curve processing and well logging database

[0092] The logging curve data of the tight sandstone reservoir of the Permian Shihezi Formation in the Ordos Basin are processed, mainly including outlier removal, curve splicing, depth relocation correction and logging curve standardization, in order to eliminate the depth error and offset error between each logging series and ensure the consistency between the core depth and the logging depth.

[0093] The well logging curve analysis found that: (1) spontaneous potential (SP) and natural gamma (GR) curves have good response characteristics to the brittle minerals, plastic minerals and mud content of tight sandstone reservoirs; (2) acoustic transit time (AC), supplementary neutron (CNL) and supplementary density (DEN) curves can reflect the reservoir porosity and permeability conditions and the degree of microfracture development; (3) the response of deep lateral resistivity (LLD) and shallow lateral resistivity (LLS) curves is closely related to the properties of formation water, oil and gas content and the degree of dissolution development. Therefore, seven types of sensitive well logging curves, GR, SP, DEN, CNC, AC, LLS and LLD, were selected, and missing value processing and outlier detection processing were performed.

[0094] Based on the core observation, thin section analysis, fracture genesis type and development density, and quantitative statistics of dissolution product abundance of the Shihezi Formation in the Ordos Basin, the logging curve response characteristics of six types of "microfracture genesis, development density and dissolution product" combinations are summarized: ① Type I "structural-diagenetic microfractures, high development density, low dissolution products" combination has the characteristics of "low SP, high GR, low LLD and LLS, high AC, low DEN, and high CNL"; ② Type II "structural-diagenetic microfractures, medium development density, and medium dissolution products" combination has the characteristics of "high SP, low GR, high LLD and LLS, medium AC, medium DEN, and low CNL"; ③ Type III "structural-diagenetic microfractures, low development density, and medium dissolution products" combination has the characteristics of "high SP, low GR, high LLD and LLS, medium AC, medium DEN, and low CNL"; ④The combination of type IV "diagenetic microfractures, high development density + low dissolution products" has the characteristics of "medium SP, high GR, medium LLD and LLS, low AC, low DEN, and high CNL"; ⑤The combination of type V "diagenetic microfractures, medium development density, and medium dissolution products" has the characteristics of "high SP, low GR, high LLD and LLS, medium AC, high DEN, and low CNL"; ⑥The combination of type VI "diagenetic microfractures, low development density, and high dissolution products" has the characteristics of "medium SP, medium GR, low LLD and LLS, low AC, high DEN, and medium CNL". Through the above logging response characteristic analysis and data extraction, a logging database with different combinations of "microfracture genesis type, development density, and dissolution products" was established.

[0095] Step 5: Discriminant Model

[0096] (1) Use the code in MATLAB software to traverse the original well logging data column data to see if there are more than 20% missing. If missing, delete it directly, and then perform initial data detection on the row data. Traverse the data to automatically identify the character data in the original data and convert it into floating point data.

[0097] (2) Select the sigma detection method to detect outliers in the data. The specific steps are as follows: ① Read the data set X, which contains n data points; ② Calculate the mean (μ) and standard deviation (σ) of the data set; ③ According to the properties of the normal distribution, about 68% of the data points fall within the range of ±1σ of the mean, about 95% of the data points fall within the range of ±2σ of the mean, and about 99.7% of the data points fall within the range of ±3σ of the mean; ④ Using the 3σ principle, data points in the data set that differ from the mean by more than 3 times the standard deviation are defined as outliers; ⑤ Check whether there are outliers in the data set, and if so, mark them as outliers.

[0098] (3) Based on the missing value processing and outlier detection processing, MATLAB software is used to build a discriminant model based on the random forest algorithm, and Bayesian optimization processing is performed at the same time. The specific steps are as follows:

[0099] ① Data preparation: The data required for this invention are logging data values ​​of six different combinations of "microfracture causes, development density and dissolution products" of the Shihezi Formation tight sandstone (as shown in Table 2). In order to detect the accuracy of the training model, all the data are randomly divided according to the ratio of "training set: validation set: test set = 8:1:1". Each sample has a set of features and corresponding category labels. These samples will be used to train the classification model.

[0100] Table 2. Well logging database of six combinations of “microfracture genesis, development density and dissolution products” in tight sandstone of Shihezi Formation

[0101]

[0102]

[0103] ② Random forest algorithm model establishment

[0104] The random forest classification model is implemented by calling the TreeBagger function that comes with MATLAB. The random forest algorithm is a collection of multiple unpruned decision trees. The training set of each decision tree is sampled from the original training set using the Bagging method. During the construction of each decision tree, each node selects the best split feature based on information gain. Finally, each tree votes with equal weight to decide the prediction type of each instance as the final classification result (such as Figure 4 shown).

[0105] ③ Bayesian optimization random forest classification algorithm

[0106] There are some shortcomings when using the random forest algorithm for classification: when the correlation between trees is greater, the error rate is greater; when the training data is noisy, it is easy to produce overfitting. To solve the above shortcomings, the Bayesian optimization random forest classification algorithm (such as Figure 5 shown):

[0107] 1) Define the hyperparameter space: determine the hyperparameters to be optimized and their value ranges;

[0108] 2) Define the surrogate model: Select an appropriate surrogate model, usually using a Gaussian process;

[0109] 3) Initialization sample set: randomly select a set of initial samples in the hyperparameter space and train the proxy model based on the performance evaluation of these samples;

[0110] 4) Iterative optimization: Based on the prediction of the proxy model, select the next hyperparameter combination that is most likely to improve performance; evaluate the selected hyperparameter combination on the real model and record its performance; update the proxy model with new samples and update the estimate of the objective function in the hyperparameter space; repeat the above steps until the preset number of iterations is reached or the stopping criterion is met;

[0111] 5) Output the best hyperparameters: Based on the optimization results, output the best hyperparameter combination for training the final model;

[0112] 6) The optimization results show that when the number of decision trees is 167 and the minimum number of leaf nodes is 3, the classification result is the best; the results show that all data sets are divided into training set: validation set: test set = 8:1:1 ratio and then the classification and discrimination model is trained, validated and tested (such as Figure 6a-6c As shown in the figure); the training set confusion matrix, each row represents the true category of the data, and each column represents the predicted category given by the model. Taking the first column of the confusion matrix as an example, it can be determined that 16 of the 19 data generated by the classification model are accurate and 3 are wrong. Finally, the accuracy of the model training set is 0.95 (as shown in the figure). Figure 6a ); The principles of the verification set confusion matrix and the test set confusion matrix are as above, where the verification set accuracy is 0.85 (such as Figure 6b ), the test set accuracy is 0.90 (such as Figure 6c ).

[0113] Step 6: Output and verification of judgment results

[0114] The established model was used to identify the tight sandstone reservoirs of the typical wells of the Shihezi Formation in the Ordos Basin, and the identification results of different combinations of "microfracture genesis type, development density and dissolution products" were obtained. After comparison with the actual drilling core and thin section results, the accuracy of the identification results was 87.3% (e.g. Figure 7 Therefore, it is believed that this method can be promoted and used.

[0115] Example 2

[0116] This embodiment is based on the design of embodiment 1 and discloses a system for identifying micro-cracks in a dense sandstone reservoir. Figure 8 As shown, the system is used to implement the above-mentioned method for identifying micro-fractures in tight sandstone reservoirs, and at least includes:

[0117] The data acquisition module is configured to obtain the original data of the combination type of microcracks with the label "microcrack genesis, development density and dissolution product"; specifically, by using the three evaluation parameters of the genesis type, development characteristics and abundance of dissolution products of microcracks in tight sandstone reservoirs, a quantitative classification standard for each parameter is established, and the original data of the combination type with the label "microcrack genesis, development density and dissolution product" is uniformly classified;

[0118] The microcracks have the following types of genesis: tectonic genesis and diagenetic genesis; the microcracks have the following development characteristic parameters: opening, length and surface density; and the microcracks are divided into low-density microcrack combination, medium-density microcrack combination and high-density microcrack combination according to the surface density index;

[0119] The characteristics of the dissolution products include product type, product content and development location. The abundance of the dissolution products is quantitatively counted by Image J image analysis software, and the dissolution products are divided into low dissolution product combination, medium dissolution product combination and high dissolution product combination according to the abundance index of the dissolution products.

[0120] The abnormality elimination module is configured to process the logging curves by means of abnormal value elimination, curve splicing, depth relocation and logging curve standardization, select the logging curves that are sensitive to the combination type of "microfracture genesis type, development density and dissolution product", obtain the sensitive logging curve data corresponding to different combinations of "microfracture genesis type, development density and dissolution product", and establish a logging database of different combinations of "microfracture genesis type, development density and dissolution product";

[0121] A data set construction module is configured to perform missing value processing and outlier detection processing on the well logging database to construct a data set, and divide the data set into a training data set, a validation data set and a test data set;

[0122] The classification and discrimination module is configured to use the random forest classification algorithm selected by the Bayesian optimization algorithm as the discrimination model of the combination type of "microfracture genesis, development density and dissolution products", input the sample labels and logging data of different combination types of "microfracture genesis type, development density and dissolution products" into the model, and randomly divide the logging data into a training data set, a validation data set and a test data set at a ratio of 8:1:1. After the model training output, the model classification confusion matrix is ​​obtained, and the classification result is obtained through the classification accuracy displayed by the confusion matrix;

[0123] The recognition module is configured to use the trained discrimination model to recognize the combination type of "microcrack genesis, development density and dissolution product" of the microcracks to be tested.

[0124] Example 3

[0125] Based on the design foundation of Embodiments 1 and 2, this embodiment discloses a computer-readable storage medium, which stores a program. The program can be executed by one or more processors to implement the above-mentioned method for identifying microcracks in tight sandstone reservoirs.

[0126] In the absence of conflict, the above embodiments and features in the embodiments may be combined with each other.

[0127] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for identifying microcracks in a tight sandstone reservoir, characterized in that: The specific steps include: S1. Get the original data S11. Obtain the genesis type of microcracks and the characteristic parameters of microcrack development based on the analysis of drilling cores, imaging logging images and casting thin section images; S12, obtaining the characteristics of the dissolution products based on the casting thin section image analysis; S2. Data Processing According to the three evaluation parameters of the genesis type, development characteristics and abundance of dissolved products of microcracks in the tight sandstone reservoir obtained in step S1, a quantitative classification standard for each parameter is established, and the raw data with the combination type labeled "microcrack genesis, development density and dissolved products" is uniformly classified; S3, build data set Conventional logging curves are preprocessed by means of outlier removal, curve splicing, depth relocation and logging curve standardization, and logging curves that are sensitive to the combination of "microfracture genesis type, development density and dissolution products" are selected to obtain sensitive logging curve data corresponding to different combinations of "microfracture genesis type, development density and dissolution products", and a logging database of different combinations of "microfracture genesis type, development density and dissolution products" is established; S4, performing missing value processing and outlier detection processing on the well logging database obtained in step S3 to construct a data set, and dividing the data set into a training data set, a verification data set and a test data set; S5. Select the optimal hyperparameters of the random forest classification algorithm through the Bayesian optimization algorithm; The Bayesian optimization algorithm optimizes the hyperparameters of the random forest algorithm, including the number of random forest trees and the minimum number of leaf nodes. The Bayesian optimization includes the following steps: S51, define a hyperparameter space, determine the hyperparameters to be optimized and their value ranges; S52, define the proxy model, select an appropriate proxy model, usually using Gaussian process; S53, initializing the sample set, randomly selecting a set of initial samples in the hyperparameter space, and training the proxy model based on the performance evaluation of these samples; S54, iterative optimization, based on the prediction of the proxy model, select the next hyperparameter combination that is most likely to improve performance; evaluate the selected hyperparameter combination on the real model and record its performance; update the proxy model with new samples and update the estimate of the objective function in the hyperparameter space; repeat the above steps until the preset number of iterations is reached or the stopping criterion is met; S55, output optimal hyperparameters; S6. Substitute the selected optimal hyperparameters into the random forest algorithm training model to establish a discrimination model for the combination type of "microcracks genesis, development density and dissolution products"; S7. Perform cross-validation on the training set and evaluate the model prediction performance on the test set.

2. The method for identifying microcracks in a tight sandstone reservoir according to claim 1, characterized in that: In step S11, the genesis types of the microcracks include tectonic genesis and diagenetic genesis, and the development characteristic parameters of the microcracks include opening, length and surface density. According to the surface density index, they are divided into low-density microcrack combinations, with surface density <5 cm / cm 2 ; Medium density micro crack combination, 5cm / cm 2 <Surface density<10cm / cm 2 ; and high-density microcrack combination, surface density> 10cm / cm 2 .

3. The method for identifying microcracks in a tight sandstone reservoir according to claim 2, characterized in that: In step S12, the characteristics of the dissolution products include product type, product content and development location. The abundance of the dissolution products is quantitatively counted by Image J image analysis software, and divided into a low dissolution product combination, product abundance <3%; a medium dissolution product combination, 3% < product abundance <8%; and a high dissolution product combination, product abundance >8%.

4. The method for identifying microcracks in a tight sandstone reservoir according to claim 3, characterized in that: In step S2, the combination types are divided into 6 categories, specifically: Class I "structural-diagenetic microfractures, high development density, low dissolution products" combination, Class II "structural-diagenetic microfractures, medium development density, medium dissolution products" combination, Class III "structural-diagenetic microfractures, low development density, high dissolution products" combination, Class IV "diagenetic microfractures, high development density + low dissolution products" combination, Class V "diagenetic microfractures, medium development density, medium dissolution products" combination and Class VI "diagenetic microfractures, low development density, high dissolution products" combination.

5. The method for identifying micro-cracks in a tight sandstone reservoir according to claim 1, characterized in that: In step S3, the logging curves include a natural potential curve, a natural gamma curve, an acoustic time difference curve, a supplementary neutron curve, a supplementary density curve, a deep lateral resistivity curve and a shallow lateral resistivity curve.

6. The method for identifying micro-cracks in a tight sandstone reservoir according to claim 5, characterized in that: In step S3, the process of removing outliers is to delete the measured values ​​whose deviation from the average value exceeds three times the standard deviation in a group of measured values ​​at a certain sample point; The process of curve splicing is to select the logging value of the formation at the same depth measured by two logging curves as the comparison standard, and make the two logging curves overlap by comparing and moving the logging curves; The process of depth homing is to select the GR curve with high vertical resolution and obvious characteristic marks as the standard curve, and determine the movement of other logging curves relative to the standard curve by comparison to complete the depth homing of the logging curve; The process of logging curve standardization is to calibrate the logging curves of different wells, select the standard layer of key wells with homogeneous formation and stable distribution, determine the correction amount required for each logging curve of other wells through the peak value of the logging curve frequency distribution histogram of the key drilling standard layer, and complete the standardization process of logging curves.

7. The method for identifying micro-fractures in a tight sandstone reservoir according to claim 1, characterized in that: In step S4, the process of missing value processing is to use the code in MATLAB software to traverse the original logging data column data to see if there is more than 20% missing, and if missing, delete it directly; the process of outlier detection processing is to select the sigma detection method to perform outlier detection processing on the data.

8. The method for identifying micro-fractures in a tight sandstone reservoir according to claim 1, characterized in that: In step S4, the ratio of the training data set, the validation data set, and the test data set is 8:1:

1.

9. A system for identifying micro-fractures in tight sandstone reservoirs, characterized in that: The system is used to implement the method for identifying micro-fractures in a tight sandstone reservoir according to any one of claims 1 to 8, and at least comprises: The data acquisition module is configured to obtain the original data of the combination type of microcracks with the label "microcrack genesis, development density and dissolution product"; specifically, by using the three evaluation parameters of the genesis type, development characteristics and abundance of dissolution products of microcracks in tight sandstone reservoirs, a quantitative classification standard for each parameter is established, and the original data of the combination type of microcracks with the label "microcrack genesis, development density and dissolution product" is uniformly classified; The microcracks have the following types of genesis: tectonic genesis and diagenetic genesis; the microcracks have the following development characteristic parameters: opening, length and surface density; and the microcracks are divided into low-density microcrack combination, medium-density microcrack combination and high-density microcrack combination according to the surface density index; The characteristics of the dissolution products include product type, product content and development location. The abundance of the dissolution products is quantitatively counted by Image J image analysis software, and the dissolution products are divided into low dissolution product combination, medium dissolution product combination and high dissolution product combination according to the abundance index of the dissolution products. The abnormality elimination module is configured to process the logging curves by means of abnormal value elimination, curve splicing, depth relocation and logging curve standardization, select the logging curves that are sensitive to the combination type of "microfracture genesis type, development density and dissolution product", obtain the sensitive logging curve data corresponding to different combinations of "microfracture genesis type, development density and dissolution product", and establish a logging database of different combinations of "microfracture genesis type, development density and dissolution product"; A data set construction module is configured to perform missing value processing and outlier detection processing on the well logging database to construct a data set, and divide the data set into a training data set, a validation data set and a test data set; The classification and discrimination module is configured to use the random forest classification algorithm selected by the Bayesian optimization algorithm as the discrimination model of the combination type of "microfracture genesis, development density and dissolution products", input the sample labels and logging data of different combination types of "microfracture genesis type, development density and dissolution products" into the model, and randomly divide the logging data into a training data set, a validation data set and a test data set at a ratio of 8:1:

1. After the model training output, the model classification confusion matrix is ​​obtained, and the classification result is obtained through the classification accuracy displayed by the confusion matrix; The recognition module is configured to use the trained discrimination model to recognize the combination type of "microcrack genesis, development density and dissolution product" of the microcracks to be tested.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program, and the program can be executed by one or more processors to implement the method for identifying micro-fractures in a tight sandstone reservoir according to any one of claims 1 to 8.

Citation Information

Patent Citations

  • Deep compact sandstone reservoir evaluation method based on artificial intelligence reservoir diagenesis phase recognition

    CN113820754A

  • Sandstone corrosion type distinguishing method and device, electronic equipment and medium

    CN114065828A

  • Method and device for determining compact sandstone microfracture development degree and storage medium

    CN114135264A

  • Hydraulic fracturing in hydrocarbon-bearing reservoirs

    US20220112422A1