Method for identifying cementation section and corrosion section of laumonite in volcaniclastic rock

Abstract

The invention belongs to the field of petroleum and natural gas exploration and development, and particularly discloses a volcaniclastic rock laumonite cementation section and corrosion section recognition method which comprises the following steps: performing lithofacies feature recognition on a rock sample, and dividing the types of laumonite cementation sections and corrosion sections; the method comprises the following steps: selecting a sensitive logging curve in response to laumontite by combining lithofacies characteristics, then extracting sensitive logging data of different cementation types of laumontite by combining the lithofacies characteristics, and establishing a database of a random forest algorithm; and carrying out model training on a pre-established random forest algorithm by utilizing the database, and obtaining a laumonite cementation section and corrosion section identification model by taking the sensitive logging curve data as input characteristics and the corresponding laumonite cementation section and corrosion section types as output characteristics during training. According to the method, continuous identification and accurate prediction of the laumonite cementation section and the corrosion section of the ultra-deep volcaniclastic rock reservoir can be realized.

Description

Technical Field

[0001] This application belongs to the field of oil and gas exploration and development, and more specifically, relates to a method for identifying cemented and dissolved sections of volcanic clastic rocks. Background Technology

[0002] With the strategic shift in oil and gas exploration from shallow to deep and ultra-deep reservoirs, ultra-deep volcanic clastic rocks have become important exploration targets. However, volcanic clastic reservoirs are characterized by strong zeolite mineral cementation and uneven spatial distribution of dissolution during burial, resulting in strong heterogeneity that severely restricts the exploration and development of ultra-deep (burial depth ≥ 6000m) volcanic clastic reservoirs.

[0003] Therefore, identifying the types of zeolite cemented sections and dissolution sections in ultra-deep volcanic clastic reservoirs, and predicting their distribution patterns, is crucial for predicting the quality of ultra-deep volcanic clastic reservoirs. Summary of the Invention

[0004] In view of the shortcomings of the existing technology, the purpose of this application is to provide a method for identifying cemented and dissolved sections of zeolite in volcanic clastic rocks, which can realize the continuous identification and accurate prediction of cemented and dissolved sections of zeolite in ultra-deep volcanic clastic rock reservoirs.

[0005] To achieve the above objectives, in a first aspect, this application provides a method for constructing a model for identifying cemented and dissolved sections of volcanic clastic rocks, comprising the following steps: S10, Obtain rock samples from the study area, identify the petrographic characteristics of the rock samples, and classify the types of zeolite cemented sections and dissolution sections; S20, combined with petrographic characteristics, well logging data of well locations with typical cementation and dissolution phenomena of zeolite in the study area were extracted. The well logging data was preprocessed, and sensitive well logging curves that respond to zeolite were selected. Then, sensitive well logging data of different cementation types of zeolite were extracted in combination with petrographic characteristics, and a database of random forest algorithm was established. S30, the pre-established random forest algorithm is trained using the database. During training, sensitive logging curve data is used as input features, and the corresponding zeolite cemented section and dissolution section types are used as output features to obtain the zeolite cemented section and dissolution section identification model.

[0006] The method for constructing a model for identifying cemented and dissolved sections of volcanic clastic rocks provided in this application has the following advantages: This application integrates drilling core analysis, image identification, conventional logging curve analysis, and random forest algorithm, and uses a combination of macroscopic and microscopic techniques to achieve continuous identification and accurate prediction of cemented and dissolved sections of volcanic clastic rocks in ultra-deep volcanic clastic reservoirs. It fundamentally solves the difficult problem of quantitative identification and prediction of cemented and dissolved sections of volcanic clastic rocks in existing technologies, and lays the foundation for accurate prediction of the quality of ultra-deep volcanic clastic reservoirs.

[0007] As a further preferred option, step S10 specifically includes: Rock samples were obtained from the strata in the study area. Core observation and thin section identification were used to identify and quantitatively analyze the petrographic features of the study area, as well as the petrographic features of cementation and dissolution of zeolite. Then, human-computer interactive identification methods were used to classify the types of cemented and dissoluted zeolite sections.

[0008] As a further preferred option, in step S10, the types of cemented and eroded sections of zeolite specifically include: Type I weakly cemented zeolite section with zeolite cement content of 0% to 5% and zeolite erosion amount of 0% to 5%; Type II moderately cemented to strongly eroded zeolite section with zeolite cement content of 5% to 15% and zeolite erosion amount of greater than 5%; and Type III strongly cemented to weakly eroded zeolite section with zeolite cement content of greater than 15% and zeolite erosion amount of 0% to 5%.

[0009] As a further preferred embodiment, in step S20, the sensitive logging curves include natural gamma, sonic transit time, neutron and density logging.

[0010] As a further preferred embodiment, in step S20, the preprocessing includes eliminating outliers in the logging data, stitching logging curves, depth correction, and standardizing logging curve data.

[0011] As a further preferred option, step S20 also includes: The sensitive logging curves are standardized to eliminate the influence of dimensions between them, while improving the convergence speed and stability of the random forest algorithm optimization process, thus completing the standardization of the logging curves.

[0012] As a further preferred method, the sensitive logging curves are standardized using the Z-score standardization method. The calculation formula for the Z-score standardization is as follows:

[0013] In the formula, z is the normalized value of a sensitive logging curve, and x is the original value of that sensitive logging curve. This is the average value of the data from this sensitive logging curve. This represents the standard deviation of the data for this sensitive logging curve.

[0014] As a further preferred embodiment, in step S30, the process of establishing the random forest model is specifically as follows: A sample database is established, wherein the feature vector of the database has n+1 dimensions, where n is the number of sensitive logging curve types and 1 is a type label for a set of turbid zeolite cemented and dissolved sections. For each decision tree, samples are randomly drawn with replacement from the original database. The undrawn samples constitute out-of-bag data for model testing. When splitting at each non-leaf node of each tree, m features are randomly selected, where m is no greater than the square root of the total number of features. The Gini index is used as the evaluation criterion for decision tree splitting. If the reduction in Gini impurity of all candidate splitting schemes does not exceed the minimum gain threshold, the splitting is stopped. At the same time, the maximum depth of the decision tree is set to optimize the model generation speed and prevent overfitting. The random forest model is generated using the majority voting method; Output the classification results. As a further preferred option, the following steps are also included: S40 uses accuracy, precision, recall, and F1 score to validate and evaluate the trained model. Model training is completed when the accuracy reaches 85% or higher.

[0015] Secondly, this application provides a method for identifying cemented and dissolved sections of volcanic clastic rocks, comprising the following steps: The cemented and dissolved sections of the target strata are identified and predicted using the cemented and dissolved sections identification model of zeolite. The cemented and dissolved sections identification model of zeolite is constructed using the construction method of the cemented and dissolved sections identification model of volcanic clastic rocks described in any one of the above descriptions.

[0016] It is understandable that the beneficial effects of the second aspect mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here. Attached Figure Description

[0017] Figure 1 This is a flowchart of the method for constructing the identification model of cemented and dissolved sections of volcanic clastic rocks provided in this application; Figure 2 This is a flowchart of the method for identifying the cemented and dissolved sections of turbidite in ultra-deep volcanic clastic reservoirs provided in the embodiments of this application; Figure 3These are characteristic images of cemented and dissolved cast thin sections of zeolite provided in the embodiments of this application; wherein, A is a weakly cemented section of type I zeolite, well A is 7598.3m; B is a moderately cemented to strongly dissolved section of type II zeolite, well A is 6958m; C is a strongly cemented to weakly dissolved section of type III zeolite, well A is 7269m. Figure 4 These are the sensitive logging curve response characteristics of different types of cemented and dissolved sections of zeolite provided in the embodiments of this application; wherein, A is a weakly cemented section of type I zeolite, with a depth of 7598.3m in well A; B is a moderately cemented to strongly dissolved section of type II zeolite, with a depth of 6958m in well A; and C is a strongly cemented to weakly dissolved section of type III zeolite, with a depth of 7269m in well A. Figure 5 This is a schematic diagram of the random forest model discrimination provided in the embodiments of this application; Figure 6 This is a diagram showing the confusion matrix identification result of the random forest method provided in the embodiments of this application; Figure 7 This application provides an analysis of the cemented and dissolved sections of turbidite in the untrained identification section provided in the embodiments of this application; wherein, A is a type I weakly cemented section, with a depth of 7627.2m in well A; B is a type II mediumly cemented to strongly dissolved section, with a depth of 7056.5m in well A; and C is a type III strongly cemented to weakly dissolved section, with a depth of 7480m in well A. Figure 8 This is a graph showing the prediction results of an untrained well segment based on a random forest model provided in this application embodiment. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0019] This application provides a method for identifying cemented and dissolved sections of zeolite in volcanic clastic rocks, relating to the current field of deep-to-ultra-deep volcanic clastic rock oil and gas exploration and development technology. The method includes: selecting ultra-deep volcanic clastic rock formations in the study area as target formations, obtaining rock samples from the target formations, and comprehensively utilizing drilling cores, microscopic image identification, and conventional logging curves. First, the zeolite cement content and zeolite dissolution intensity of the rock samples are determined, establishing a standard for dividing zeolite cemented and dissolved sections, and identifying the vertical distribution characteristics of these sections. Second, sensitive logging curves of the target formations in the study area are selected, and the sensitive logging curves are standardized to analyze the logging curve response characteristics of zeolite cemented and dissolved sections. Third, the sensitive logging curve data are processed, and a random forest model is used to predict the development intensity and distribution pattern of zeolite cemented and dissolved sections in the target formations.

[0020] like Figure 1 As shown, the method for constructing the identification model of cemented and dissolved sections of volcanic clastic rocks provided in this application includes steps S10 to S30, which are detailed below: Step S10: Obtain rock samples from the study area, identify the petrographic characteristics of the rock samples, and classify the types of cemented and dissolved sections of zeolite.

[0021] In this application, step S10 may be: obtaining rock samples from the strata of the study area, comprehensively using core observation and casting thin section identification, identifying and quantitatively statistically analyzing the petrographic characteristics of the study area, the petrographic characteristics of cementation and dissolution of zeolite, and then combining human-computer interactive identification methods to classify the types of cemented and dissoluted zeolite sections.

[0022] Specifically, the types of cemented and eroded sections of zeolite are divided as follows: Type I weakly cemented zeolite section with zeolite cement content of 0% to 5% and zeolite erosion of 0% to 5%; Type II moderately cemented to strongly eroded zeolite section with zeolite cement content of 5% to 15% and zeolite erosion of greater than 5%; and Type III strongly cemented to weakly eroded zeolite section with zeolite cement content of greater than 15% and zeolite erosion of 0% to 5%.

[0023] Step S20: Combining petrographic characteristics, extract logging data from well locations in the study area that exhibit typical cementation and dissolution phenomena of zeolite. Preprocess the logging data, select sensitive logging curves that respond to zeolite, and then extract sensitive logging data for different cementation types of zeolite based on petrographic characteristics to establish a database for the random forest algorithm.

[0024] In this application, preprocessing includes removing outliers from the logging data, stitching logging curves, depth correction, and standardization of the logging curve data. Sensitive logging curves include natural gamma, sonic transit time, neutron, and density logging.

[0025] Step S30: Use the database to train the pre-established random forest algorithm model. During training, use sensitive logging curve data as input features and the corresponding turbid zeolite cemented section and dissolution section types as output features to obtain the turbid zeolite cemented section and dissolution section identification model.

[0026] The method for constructing a model for identifying cemented and dissolved sections of volcanic clastic rocks provided in this application has the following advantages: This application integrates drilling core analysis, image identification, conventional logging curve analysis, and random forest algorithm, and uses a combination of macroscopic and microscopic techniques to achieve continuous identification and accurate prediction of cemented and dissolved sections of volcanic clastic rocks in ultra-deep volcanic clastic reservoirs. It fundamentally solves the difficult problem of quantitative identification and prediction of cemented and dissolved sections of volcanic clastic rocks in existing technologies, and lays the foundation for accurate prediction of the quality of ultra-deep volcanic clastic reservoirs.

[0027] Based on the same inventive concept, this application also provides a method for identifying the cemented and dissolved sections of volcanic clastic rocks, comprising the following steps: The cemented and dissolved sections of the zeolite were identified and predicted using a model for identifying cemented and dissolved sections of the zeolite in the target formation.

[0028] The identification model for cemented and dissolved sections of turbidite provided in this application is constructed using the construction method of the above-mentioned identification model for cemented and dissolved sections of turbidite in volcanic clastic rocks.

[0029] In one embodiment, the technical solution to achieve the above objective can be as follows: This embodiment develops a quantitative evaluation method for the cemented and dissolved sections of turbidite in ultra-deep volcanic clastic reservoirs, combining petrological characteristics, sensitive logging curve response characteristics, and random forest algorithms. Figure 2 As shown, the specific technical solution is as follows: Step 1: Petrographic identification of cemented and dissolved sections of zeolite Core and cast thin section samples were collected from the study area. Core observation and cast thin section observation were used to identify and quantitatively analyze the petrographic features of the study area, as well as the petrographic features of cementation and dissolution of zeolite. Human-computer interaction identification methods were used to classify the types of cemented and dissoluted zeolite sections.

[0030] Step 2: Database of logging curves for cemented and dissolved sections of zeolite. Well logging data from wells exhibiting typical cementation and dissolution phenomena of zeolite in the study area were extracted based on petrographic characteristics. The well logging data underwent preprocessing, selecting sensitive logging curves responsive to zeolite. The main steps included: eliminating outliers, stitching logging curves, depth correction, and standardizing the logging curve data. Sensitive logging data for different cementation types of zeolite were extracted based on petrographic characteristics, and a database using a random forest algorithm was established.

[0031] Step 3: Random Forest Algorithm to Identify the Model The process of building a random forest model is as follows: (1) Establish a sample database. The feature vector dimension of the database in this method is n+1, where n is the number of sensitive logging curve types and 1 is a set of data labels (types of cemented and dissolved sections of turbid zeolite).

[0032] (2) For each decision tree, samples are randomly drawn with replacement from the original database. The undrawn samples constitute out-of-bag data for model testing.

[0033] (3) When splitting at each non-leaf node of each tree, randomly select m features (m is not greater than the square root of the total number of features) and use the Gini index as the evaluation criterion for decision tree splitting. If the reduction in Gini impurity of all candidate splitting schemes does not exceed the minimum gain threshold, then stop splitting. At the same time, set the maximum depth of the decision tree to optimize the model generation speed and prevent overfitting.

[0034] (4) The random forest model is generated by majority voting.

[0035] (5) Output the classification results.

[0036] Step 4: Model Result Output and Validation The random forest model is validated using a validation database, and the validation results are output. The prediction performance of the random forest model is validated and evaluated using four parameters: accuracy, precision, recall, and F1 score, combined with the confusion matrix. An accuracy higher than 85% indicates that the random model has been successfully trained.

[0037] The method provided in this embodiment has the following advantages: it comprehensively utilizes drilling core analysis, image identification, conventional logging curve analysis, and random forest algorithms to accurately predict the cemented and dissolved sections of zeolite in ultra-deep volcanic clastic reservoirs. Zeolite is an important mineral for the formation of volcanic clastic reservoirs, and accurate prediction of zeolite is beneficial for the exploration and development of ultra-deep volcanic clastic reservoirs. This method combines big data models with geological knowledge, leveraging the advantage of computer methods in processing massive amounts of data at low cost. It solves the problems of complex response characteristics, long time consumption, and high cost associated with traditional geological methods such as rock sample identification and logging curve prediction in predicting the cemented and dissolved sections of zeolite in ultra-deep volcanic clastic reservoirs. This method can also be applied to accurately predict other minerals with special responses in logging curves in the vertical direction.

[0038] The following is a specific implementation example of this application: The specific technical solution of this application is illustrated using the Shawan Depression in the Junggar Basin as an example: Step 1: Petrographic identification of cemented and dissolved sections of zeolite This embodiment collects ultra-deep volcanic clastic rock samples from Well A in the study area of ​​the Shawan Depression in the Junggar Basin. Combining core observation and thin-section analysis, microscopic characterization of zeolite cementation and dissolution phenomena is performed, such as... Figure 3 As shown.

[0039] Based on extensive sample observation and quantitative statistics, the cemented and dissolved sections of zeolite in the study area were divided into three categories: Category I weakly cemented zeolite, Category II moderately cemented-strongly dissolved zeolite, and Category III strongly cemented-weakly dissolved zeolite, as shown in Table 1.

[0040] Table 1. Classification Criteria for Cemented and Dissolved Sections of Turbidite

[0041] Step 2: Database of logging curves for cemented and dissolved sections of zeolite. This example collects ultra-deep logging data from Well A in the study area of ​​the Shawan Depression in the Junggar Basin, and preprocesses the logging data, including eliminating outliers, splicing logging curves, depth correction, and curve standardization.

[0042] In this example, the mudstone at the top of the Urho Formation in the study area was selected as the standard layer, the GR curve was selected as the standard curve, and the positions of other sensitive logging curves were adjusted to complete the depth correction of the logging curves.

[0043] The characteristic values ​​of various sensitive logging curves in the standard layer of well A were statistically analyzed. Z-score normalization was then applied to standardize the sensitive logging curve data of the study area, eliminating the influence of dimensions between sensitive logging curves and improving the convergence speed and stability of the subsequent random forest model optimization process. This completed the standardization of the logging curves of well A. The Z-score normalization calculation formula is as follows:

[0044] In the formula, z is the normalized value of a sensitive logging curve, and x is the original value of that sensitive logging curve. This is the average value of the data from this sensitive logging curve. This represents the standard deviation of the data from the sensitive logging curve. Outliers with a Z-score greater than 3 or less than -3 are removed to complete the outlier removal process.

[0045] The study area selected sensitive logging curve types for the response of zeolite cementation and dissolution sites, mainly including four types: natural gamma (GR), sonic transit time (AC), neutron (CNL), and density logging (DEN), such as... Figure 4 As shown.

[0046] The weakly cemented section of Class I turbid zeolite is developed between 7595m and 7599.4m. Sensitive logging curves show the following characteristics: GR values ​​range from 43.10 to 54.76 API, with an average of 49.42 API and a standard deviation of 2.63; AC values ​​range from 62.06 to 72.04 μs / ft, with an average of 67.10 μs / ft and a standard deviation of 2.09; CNL values ​​range from 9.18% to 14.80%, with an average of 12.16% and a standard deviation of 1.39; and DEN values ​​range from 2.45 to 2.54 g / cm³. 3 Average value 2.50 g / cm³ 3 The standard deviation is 0.0219.

[0047] The moderately cemented to strongly dissolved section of the Type II turbidite is developed at depths of 6953m to 6959m. Sensitive logging curves show the following characteristics: GR values ​​range from 36.68 to 54.96 API, with an average of 43.40 API and a standard deviation of 4.09; AC values ​​range from 64.45 to 76.36 μs / ft, with an average of 68.98 μs / ft and a standard deviation of 3.10; CNL values ​​range from 10.99% to 15.63%, with an average of 14.02% and a standard deviation of 1.25; and DEN values ​​range from 2.40 to 2.55 g / cm³. 3 The average value is 2.47 g / cm³. 3 The standard deviation is 0.0402.

[0048] The strongly cemented to weakly dissolved section of Type III turbid zeolite is developed at 7266.1m to 7270.1m. Sensitive logging curves show the following characteristics: GR values ​​range from 32.72 to 44.91 API, with an average of 37.94 API and a standard deviation of 3.35; AC values ​​range from 59.24 to 64.70 μs / ft, with an average of 61.70 API and a standard deviation of 1.48; CNL values ​​range from 10.03% to 14.56%, with an average of 12.06% and a standard deviation of 1.14; and DEN values ​​range from 2.44 to 2.51 g / cm³. 3 Average value 2.48 g / cm³ 3 The standard deviation is 0.0190.

[0049] Four types of sensitive logging data—natural gamma (GR), density (DEN), neutron (CNL), and acoustic transit time (AC)—were selected. The cemented and dissolved sections of zeolite were labeled as one group, forming a raw database of sensitive logging curves for different types of cemented and dissolved sections of zeolite, as shown in Table 2. 180 groups (60 groups for each type of cemented and dissolved zeolite) were randomly selected and standardized as the standardized database for the random forest model. Of these, 144 groups were used as the training database, and 36 groups were used as the validation database, as shown in Table 3.

[0050] Table 2. Original database of sensitive logging curves for cemented and dissolved sections of different types of zeolite.

[0051] Table 3 Standardized database for the random forest model

[0052] Step 3: Random Forest Algorithm to Identify the Model Use PyCharm software to import the random forest model code from the sklearn database, reference the training database, and run the program.

[0053] The debugging parameters for the random forest code provided in this embodiment are as follows: Number of Trees: This represents the number of decision trees in the random forest model. A larger number of decision trees results in better model stability and fit, but also increases computational cost. The number of trees is related to the database size. In this example, the training database and the validation database contain a total of 120 samples, which is a small database; therefore, the number of trees in this example is 100.

[0054] Random state parameter: This represents the value of the random seed (initial value). The same seed value generates the same random results, ensuring the reproducibility of experimental data. In computer science, the random state parameter is typically set to 42.

[0055] Maximum feature number: This represents the maximum number of features considered at each split. To balance the bias and diversity of each decision tree, the balance value is the square root of the total number of features. In this example, the features are 4 types of sensitivity curves, the total number of features is 4, and the maximum feature number is 2.

[0056] Maximum depth: This represents the maximum number of splits in the decision tree, used to prevent overfitting and loss of generality. In this example, the value is 5.

[0057] The process of building a random forest model is as follows: (1) Import the sample database. The feature vector of the database of this method has a dimension of 5, which are four types of sensitive curve data: natural gamma (GR), density (DEN), neutron (CNL) and acoustic transit time (AC), and one set of labels for the cemented and dissolved sections of turbid zeolite.

[0058] (2) For each decision tree, a sample is randomly drawn with replacement from the sample database and the test database is used for model testing.

[0059] (3) When splitting at each non-leaf node of each tree, randomly select m features (m is not greater than the square root of the total number of features) and use the Gini index as the evaluation criterion for decision tree splitting. If the reduction in Gini impurity of all candidate splitting schemes does not exceed the minimum gain threshold, then stop splitting. At the same time, set the maximum depth of the decision tree to optimize the model generation speed and prevent overfitting.

[0060] (4) The random forest model is generated by majority voting.

[0061] (5) Output the classification results, such as Figure 5 As shown.

[0062] Step 4: Model Result Output and Validation The prediction performance of the random forest model was validated and evaluated using four parameters: accuracy, precision, recall, and F1 score, combined with the confusion matrix of the validation set.

[0063] Accuracy represents the correctness of the model's predictions, and the calculation formula is as follows:

[0064] Precision rate represents the accuracy with which the model predicts positive examples, and is calculated using the following formula:

[0065] Recall rate represents the model's ability to capture true positive examples, and is calculated using the following formula:

[0066] The F1 score is the harmonic average of precision and recall, calculated using the following formula:

[0067] In the formula, TP (True Positive) represents a correctly predicted positive example, TN (True Negative) represents a correctly predicted negative example, FP (False Positive) represents a incorrectly predicted positive example, and FN (False Negative) represents a incorrectly predicted negative example, as shown in Table 4.

[0068] Table 4 Model Output Parameter Results

[0069] After cross-validation of the model, the random forest model's prediction results for 36 sets of test data are output, returning accuracy, precision, recall, F1 score, and confusion matrix. Data shows that the random forest model's F1 score for Class I prediction is 0.89, for Class II prediction is 0.80, and for Class III prediction is 0.91. The model demonstrates excellent ability to identify and capture positive examples for each class, with an accuracy of 92.31% for Class I, 81.82% for Class II, and 83.33% for Class III, for an average accuracy of 86.11%. A mixture matrix model is used to examine the random forest model's sensitivity to identifying three types of cemented and eroded segments in the training database. The mixture matrix shows that the model can accurately predict the cemented and eroded segment types of zeolite. Figure 6 As shown.

[0070] The generated random forest model was used to identify and predict the cemented and dissolved segments of zeolite that were not trained and identified (a total of 126 sets of standardized data). The prediction results of the cemented and dissolved segments of zeolite are shown in Table 5.

[0071] Table 5. Prediction results of the random forest model for untrained well sections.

[0072] Model predictions show that the 7626.88m–7627.38m section is a weakly cemented (Type I) zeolite section, the 7054m–7058m section is a moderately cemented–strongly dissolved (Type II) zeolite section, and the 7476.9m–7485.5m section is a strongly cemented–weakly dissolved (Type III) zeolite section. Based on thin section observations, the random forest model predictions are consistent with the actual reservoir development characteristics. Figure 7 As shown, the model performs well in predicting well sections that were not trained and identified. Therefore, importing the prediction results into the Resform software platform can achieve continuous and accurate prediction of the types of cemented and dissolved sections in deep volcanic clastic reservoirs, such as... Figure 8 As shown.

[0073] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for constructing a model for identifying cemented and dissolved sections of volcanic clastic rocks, characterized in that, Includes the following steps: S10, Obtain rock samples from the study area, identify the petrographic characteristics of the rock samples, and classify the types of zeolite cemented sections and dissolution sections; S20, combined with petrographic characteristics, well logging data of well locations with typical cementation and dissolution phenomena of zeolite in the study area were extracted. The well logging data was preprocessed, and sensitive well logging curves that respond to zeolite were selected. Then, sensitive well logging data of different cementation types of zeolite were extracted in combination with petrographic characteristics, and a database of random forest algorithm was established. S30, the pre-established random forest algorithm is trained using the database. During training, sensitive logging curve data is used as input features, and the corresponding zeolite cemented section and dissolution section types are used as output features to obtain the zeolite cemented section and dissolution section identification model.

2. The method for constructing the identification model of cemented and dissolved sections of volcanic clastic rocks as described in claim 1, characterized in that, Step S10 is as follows: Rock samples were obtained from the strata in the study area. Core observation and thin section identification were used to identify and quantitatively analyze the petrographic features of the study area, as well as the petrographic features of cementation and dissolution of zeolite. Then, human-computer interactive identification methods were used to classify the types of cemented and dissoluted zeolite sections.

3. The method for constructing the identification model of cemented and dissolved sections of volcanic clastic rock as described in claim 1, characterized in that, In step S10, the specific types of cemented and eroded sections of zeolite include: Type I weakly cemented zeolite section with zeolite cement content of 0% to 5% and zeolite erosion of 0% to 5%; Type II moderately cemented to strongly eroded zeolite section with zeolite cement content of 5% to 15% and zeolite erosion of greater than 5%; and Type III strongly cemented to weakly eroded zeolite section with zeolite cement content of greater than 15% and zeolite erosion of 0% to 5%.

4. The method for constructing the identification model of cemented and dissolved sections of volcanic clastic rock as described in claim 1, characterized in that, In step S20, the sensitive logging curves include natural gamma, sonic transit time, neutron and density logging.

5. The method for constructing the identification model of cemented and dissolved sections of volcanic clastic rock as described in claim 1, characterized in that, In step S20, the preprocessing includes eliminating outliers in the logging data, stitching logging curves, depth correction, and standardizing logging curve data.

6. The method for constructing the identification model of cemented and dissolved sections of volcanic clastic rock as described in claim 1, characterized in that, Step S20 also includes: The sensitive logging curves are standardized to eliminate the influence of dimensions between them, while improving the convergence speed and stability of the random forest algorithm optimization process, thus completing the standardization of the logging curves.

7. The method for constructing the identification model of cemented and dissolved sections of volcanic clastic rock as described in claim 6, characterized in that, The sensitive logging curves are standardized using the Z-score standardization method. The formula for calculating the Z-score standardization is as follows: In the formula, z is the normalized value of a sensitive logging curve, and x is the original value of that sensitive logging curve. This is the average value of the data from this sensitive logging curve. This represents the standard deviation of the data for this sensitive logging curve.

8. The method for constructing the identification model of cemented and dissolved sections of volcanic clastic rock as described in claim 1, characterized in that, In step S30, the process of establishing the random forest model is as follows: A sample database is established, wherein the feature vector of the database has n+1 dimensions, where n is the number of sensitive logging curve types and 1 is a type label for a set of turbid zeolite cemented and dissolved sections. For each decision tree, samples are randomly drawn with replacement from the original database. The undrawn samples constitute out-of-bag data for model testing. When splitting at each non-leaf node of each tree, m features are randomly selected, where m is no greater than the square root of the total number of features. The Gini index is used as the evaluation criterion for decision tree splitting. If the reduction in Gini impurity of all candidate splitting schemes does not exceed the minimum gain threshold, the splitting is stopped. At the same time, the maximum depth of the decision tree is set to optimize the model generation speed and prevent overfitting. The random forest model is generated using the majority voting method. Output the classification results.

9. The method for constructing the identification model of cemented and dissolved sections of volcanic clastic rock as described in claim 1, characterized in that, It also includes the following steps: S40 uses accuracy, precision, recall, and F1 score to validate and evaluate the trained model. Model training is completed when the accuracy reaches 85% or higher.

10. A method for identifying cemented and dissolved sections of volcanic clastic rocks, characterized in that, Includes the following steps: The cemented and dissolved sections of the target strata are identified and predicted using a cemented and dissolved section identification model of zeolite. The cemented and dissolved section identification model of zeolite is constructed using the construction method of the cemented and dissolved section identification model of volcanic clastic rocks as described in any one of claims 1 to 9.

Images