Geothermal water layer quality rapid identification method and system based on algorithm model

Abstract

The invention discloses a geothermal water layer quality rapid identification method and system based on an algorithm model, belongs to the technical field of geological exploration, and aims to correct logging curve data of a thermal reservoir research well area, effectively avoid possible system errors in a logging data acquisition process, including improper operation modes or instrument self-measurement errors and the like. Standardized logging data can more remarkably reflect regional geologic features compared with original data, principal component analysis is conducted on sample data through a PCA method, an original sample space is reconstructed according to the contribution rate of principal components, the dimensionality of an input space is effectively reduced, meanwhile, the correlation between variables of the input space is eliminated, and the accuracy of the input space is improved. The identification precision is improved; according to the method, the recognition accuracy is greatly improved, a more scientific and reliable method is provided for well logging data thermal reservoir recognition, and the method has practical significance in exploring geothermal resource development.

Description

Technical Field

[0001] The invention belongs to the technical field of geological exploration, and in particular relates to a method and system for quickly identifying the quality of geothermal water layers based on an algorithm model. Background Art

[0002] Global "carbon neutrality" is an important indicator of the coordinated evolution of human beings and the earth. It is related to the livable environment of the earth. It has risen to the national strategic level and reached a global consensus. Earth science can play a vital role in achieving "carbon neutrality". The use of renewable energy instead of fossil energy and carbon capture and storage are geological solutions provided by earth science. my country has a good endowment of basin-type geothermal resources. Vigorously developing and utilizing geothermal resources has important practical significance for optimizing my country's energy structure and reducing carbon emissions. Especially the hydrothermal resources associated with oil fields in young hot basins, but the aquifer identification standards established based on oil and gas reservoirs in the past are difficult to apply to geothermal water layers. At present, there is a lack of fast and effective geothermal water layer identification technology, especially the identification of geothermal water layer quality, which restricts the in-depth development of geothermal resources in oil fields. Therefore, it is of great significance to design an efficient and intelligent geothermal water layer quality identification method. Summary of the invention

[0003] In order to solve the above problems, the present invention proposes: a method for quickly identifying the quality of geothermal water layers based on an algorithm model, comprising the following steps:

[0004] S1. Based on the oil test data, the oil-water layer, dry layer and water layer of the oil reservoir are divided into different categories, and the identifiability of the logging curves for different categories is clarified;

[0005] S2. Classify the quality of geothermal water layers based on geothermal reservoir evaluation methods;

[0006] S3, according to the oil-water layer, dry layer and water layer divided by the secondary logging interpretation result, the logging curve data is processed to establish the geothermal water layer label based on the secondary logging interpretation;

[0007] S4. Use PCA principal component analysis method to study the correlation of logging curve feature vectors, perform principal component analysis on various logging curve data, reduce data dimension while retaining more data features, and use the obtained k principal components as data sets;

[0008] S5, dividing the data set;

[0009] S6. Establish an SVM classifier model, design kernel function, kernel parameters, and penalty factor parameter combination strategy, use the training set as input for training, use the grid search algorithm to traverse the parameter combination during the training process, find the optimal parameter combination based on accuracy, and finally obtain the optimal model; then send the test set to the SVM model with the optimal parameter combination to obtain the identified water layer result;

[0010] S7. On the basis of rapid identification of geothermal water layers, a well logging curve data set of three classifications of water layer quality is further established according to the results of water layer quality classification in the geothermal reservoir evaluation method;

[0011] S8. Repeat step S6 to obtain a specific quality identification result of the water layer.

[0012] Furthermore, in step S1, the histogram method is used to correct the logging curve to reduce the error caused by the difference in the test system; combined with the results of the oil-water layer, dry layer and water layer in the oil test data, the scatter plot is used to perform the correlation of the logging curve to clarify the identifiability of different logging curves to the water layer.

[0013] Furthermore, in step S2, based on the geothermal reservoir evaluation method, the water layer with a porosity greater than 25% is defined as Class I water layer, the water layer with a porosity between 15% and 25% is defined as Class II water layer, and the water layer with a porosity less than 15% is defined as Class III water layer.

[0014] Furthermore, in step S3, the logging curve data is cleaned to remove abnormal values ​​in the logging curve; the logging data is standardized to eliminate the adverse effects of dimensional differences so that the data have the same scale;

[0015] The standard for judging abnormal values ​​in the box plot is based on the upper and lower limits. The logging curve data that is greater than the upper limit of the box plot or less than the lower limit of the box plot is defined as an abnormal value and deleted. The definition formula of the lower and upper limits is:

[0016] Lower limit=Q1-1.5(Q3-Q1); upper limit=Q3+1.5(Q3-Q1)

[0017] The lower quartile Q1 refers to the value corresponding to the 25% quantile of the data; the upper quartile Q3 refers to the value corresponding to the 75% quantile of the data;

[0018] The logging data is standardized, and each logging curve is processed separately, using the following standardization formula:

[0019] Among them, z i is the variable after standardization,

[0020] Furthermore, in step S4, PCA is used to calculate multiple logging curves such as natural potential, natural gamma, acoustic time difference, density, deep lateral resistivity and shallow lateral resistivity to obtain new feature data information, and the required amount of data is determined from the multiple new features according to the cumulative contribution rate greater than 85%, and the newly formed data is used as the input data of the model.

[0021] Furthermore, in step S5, the data set is divided into a training set and a test set in proportion, the training set is used for model training and parameter adjustment, and the test set is used for final verification of the model to obtain the model test results, wherein the training set accounts for 60% and the test set accounts for 40%.

[0022] Furthermore, in step S6, the grid search method is used to optimize the parameters of the SVM to obtain the best parameter combination to achieve the best model effect; the grid search method arranges and combines all the parameters to form a parameter grid, and then trains each group of parameters in turn, and evaluates the model performance through a cross-validation method, and selects the parameter combination with the best cross-validation result as the final parameter of the model.

[0023] Furthermore, in step S7, according to the results of water layers I, II, and III divided in step S2, the data set is further divided into a training set and a test set in proportion, which is the same as step S5.

[0024] Furthermore, in step S8, a classification model is re-established based on the water layer, and the grid search method is used to optimize the parameters of the SVM to obtain the best classification model, and then the test data is input into the model for classification to obtain the water layer quality classification result of the model.

[0025] A geothermal water layer quality rapid identification system based on an algorithm model, comprising:

[0026] The well logging curve preprocessing unit preprocesses the well logging curve data of the target well in the study area;

[0027] The secondary logging interpretation unit divides the oil-water layer, dry layer and water layer of the oil zone reservoir into categories based on the oil test data, and establishes the corresponding relationship between the logging curve and different categories;

[0028] Water layer division unit, which divides the quality of geothermal water layers based on geothermal reservoir evaluation methods;

[0029] The principal component analysis unit is used to study the correlation of the characteristic vectors of the well logging curve using the PCA principal component analysis method, perform principal component analysis operations on various well logging curve data, reduce the data dimension while retaining more data features, and use the obtained k principal components as a data set;

[0030] The data set partitioning unit is used to partition the training set and the test set, and divide the data of different categories in proportion through stratified sampling;

[0031] The model training and testing unit is used to establish a support vector machine model. Based on the training set, the optimal hyperparameter combination is selected through the grid search method to obtain the optimal training model, and the test set is tested to obtain the optimal selection model.

[0032] A water layer identification unit, which uses the selected optimal model to identify the water layer;

[0033] The system management unit is used to manage system user information, organizational information, permission information, metadata information, and operation log information to support the above units.

[0034] The beneficial effects of the present invention are:

[0035] 1. The present invention corrects the logging curve data of the thermal reservoir research well area, effectively avoiding the systematic errors that may exist in the logging data acquisition process, including improper operation or instrument measurement errors. By comparison, it can be found that the standardized logging data can more significantly reflect the regional geological characteristics than the original data.

[0036] 2. The present invention uses the PCA method to perform principal component analysis on the sample data and reconstructs the original sample space according to the contribution rate of the principal component, which effectively reduces the dimension of the input space and eliminates the correlation between the variables in the input space, thereby improving the recognition accuracy.

[0037] 3. The present invention sets up multiple groups of hyperparameter selection combinations for SVM, and performs screening and verification through grid search and cross-validation algorithms to ensure that the optimal hyperparameter combination of SVM can be screened out, providing accuracy guarantee for subsequent experiments.

[0038] 4. The thermal reservoir identification method based on principal component analysis and support vector machine proposed in the present invention greatly improves the accuracy of identification compared with the manual method, provides a more scientific and reliable method for thermal reservoir identification based on logging data, and has practical significance for exploring geothermal resource development. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a technical flow diagram of the method and system for quickly identifying the quality of geothermal water layers based on an algorithm model of the present invention;

[0040] Figure 2 Well logging interpretation histograms for reservoir types and water layer qualities of the present invention;

[0041] Figure 3 It is a scatter plot of the correlation between reservoir type and logging curve of the present invention;

[0042] Figure 4 A confusion matrix for thermal reservoir identification of the present invention;

[0043] Figure 5 The confusion matrix for water layer identification of the present invention;

[0044] Figure 6 Verify the bar graph for quick identification of water layer quality;

[0045] Figure 7 is a system structure block diagram of the present invention;

[0046] Figure 8 It is the interface of the method and system for rapid identification of geothermal water layer quality based on algorithm model. DETAILED DESCRIPTION

[0047] Embodiment 1

[0048] In order to make the technical means and objectives of the present invention easy to understand, the present invention is further described below in combination with specific implementation methods. A method for quickly identifying the quality of geothermal water layers based on an algorithm model is as follows: Figure 1 As shown in the figure, the thermal reservoir is identified based on principal component analysis and support vector machine, including:

[0049] (1) Based on the oil test data, the oil-water layer, dry layer and water layer of the oil-bearing area reservoir are divided into different categories, and the identifiability of the logging curves for different categories is clarified;

[0050] (2) Based on the geothermal reservoir evaluation method (NB / T 10263-2019), the geothermal water layer is classified into good (Class I), medium (Class II) and poor (Class III) quality;

[0051] (3) According to the oil-water layer, dry layer and water layer divided by the secondary interpretation of well logging, the well logging curve data is processed to establish the geothermal water layer label based on the secondary well logging interpretation;

[0052] (4) The PCA principal component analysis method is used to study the correlation of the characteristic vectors of the well logging curves. The principal component analysis operation is performed on various well logging curve data to reduce the data dimension while retaining more data features. The obtained k principal components are used as the data set;

[0053] (5) Divide the data set into a training set and a test set in proportion. The training set is used for model training and parameter adjustment, and the test set is used for final model verification to obtain model test results. The training set accounts for 60% and the test set accounts for 40%;

[0054] (6) Establish an SVM classifier model, design the kernel function, kernel parameters, and penalty factor parameter combination strategy, use the training set as input for training, use the grid search algorithm to traverse the parameter combination during the training process, find the optimal parameter combination based on accuracy, and finally obtain the optimal model. Then send the test set to the SVM model with the optimal parameter combination to obtain the identified water layer results.

[0055] (7) Based on the rapid identification of geothermal water layers, a well logging curve dataset with three types of water layer quality was further established according to the results of the water layer quality classification (Class I, Class II, and Class III) in the geothermal reservoir evaluation method (NB / T 10263-2019).

[0056] (8) Repeat step (6) to obtain the specific quality identification result of the water layer.

[0057] The present invention performs thermal reservoir identification based on principal component analysis and support vector machine, which greatly improves the accuracy of identification compared with manual methods, provides a more scientific and reliable method for thermal reservoir identification based on well logging data, and has practical significance for exploring geothermal resource development.

[0058] Embodiment 2

[0059] like Figure 1-8 As shown, Figure 1 As shown, Figure 1 The schematic diagram of the thermal reservoir identification method based on principal component analysis and support vector machine of the present invention specifically includes the following steps:

[0060] Step 1), correct the logging curve, eliminate the system error, and form a new logging curve histogram, such as Figure 2 .

[0061] Step 2) Identify the oil-water layer, dry layer and water layer through the scatter plot of the logging curve, and clarify the identifiability of the logging curve to the water layer, such as Figure 3 .

[0062] Step 3) Clean and standardize the logging curve data of the target wells in the study area: clean the logging curve data and remove abnormal values ​​in the logging curve; standardize the logging data to eliminate the adverse effects of dimensional differences and make the data have the same scale.

[0063] To clean the logging curve data, it is necessary to remove the outliers in the logging curve. Outliers refer to sample points whose values ​​in the sample obviously deviate from the rest of the values, so they are also called outliers. Their existence will interfere with the identification experiment. A large amount of data proves that the results of the box plot in identifying outliers are more objective and have certain advantages in outlier processing. Therefore, the box plot is used here to detect and remove outliers.

[0064] The standard for judging abnormal values ​​in the box plot is based on the upper and lower limits. The logging curve data that is greater than the upper limit of the box plot or less than the lower limit of the box plot is defined as an abnormal value and deleted. The definition formula of the lower and upper limits is:

[0065] Lower limit=Q1-1.5(Q3-Q1); upper limit=Q3+1.5(Q3-Q1)

[0066] Among them, Q1 (lower quartile) refers to the value corresponding to the 25% quantile of the data; Q3 (upper quartile) refers to the value corresponding to the 75% quantile of the data.

[0067] The logging data is standardized, and each logging curve is processed separately, using the following standardization formula:

[0068]

[0069] Among them, z i is the variable after standardization,

[0070] Step 4), PCA is used to calculate various logging curves such as natural potential, natural gamma, acoustic time difference, density, deep lateral resistivity and shallow lateral resistivity to obtain new feature data information, and the required data volume is delineated from multiple new features according to the specified cumulative contribution rate, and the newly formed data is used as the input data of the model. The steps of PCA principal component analysis are as follows:

[0071] Assume that the original sample data is X=(x1,x2,...,x m ), each sample is n-dimensional.

[0072] Compute the covariance matrix of the sample:

[0073]

[0074] Use Jacobi method to find the eigenvalues ​​and corresponding eigenvectors of the covariance matrix C;

[0075] Arrange the eigenvectors from large to small according to the corresponding eigenvalues, and select the first k principal components according to the cumulative contribution rate of the principal components. The contribution rate refers to the proportion of the variance of a principal component to the total variance. That is:

[0076]

[0077] Among them, b j is the contribution rate of each component. The larger the contribution rate, the more information the principal component contains about the original variable. kIt is the cumulative contribution rate of the first k principal components. Generally, the cumulative contribution rate is required to reach more than 85% so that most of the information of the original variables can be integrated.

[0078] Here, the first four principal components are selected through PCA experiment, with a contribution rate of 97%, forming the matrix P = (p1, p2, p3, p4) to obtain the principal components.

[0079] Step 5) Divide the data set into a training set and a test set in proportion. The training set is used for model training and parameter adjustment, and the test set is used for final model verification to obtain model test results. The training set accounts for 60% and the test set accounts for 40%.

[0080] Step 6), establish an SVM model, use the grid search method to optimize the parameters of the SVM, obtain the best parameter combination, and achieve the best model effect. The grid search method is a commonly used hyperparameter optimization method. Its main idea is to arrange and combine all parameters to form a parameter grid, then train each group of parameters in turn, and evaluate the model performance through the cross-validation method, and select the parameter combination with the best cross-validation result as the final parameter of the model. The main parameters of the SVM algorithm are the penalty factor C, the kernel function, and the parameters (kernel parameters) that appear in the kernel function used. Different parameters have different value ranges. The radial basis kernel function (RBF) has a better application effect. Whether it is a small sample or a large sample, high dimension or low dimension, RBF is applicable. Therefore, RBF is selected as the kernel function, and then the penalty factor C and the kernel parameter gamma in RBF are optimized. The value range of C is set to [1, 10, 100, 1000], and the value range of gamma is set to [0.1, 0.01, 0.001, 0.0001]. The optimal parameter combination finally selected is C=1000 and gamma=0.0001.

[0081] The test confusion matrix is ​​established to show the accuracy of the support vector machine model training effect. The rows, columns and main diagonal of the confusion matrix represent the reservoir type predicted by the trained support vector machine model, the actual reservoir type of the test sample and the number of correctly identified reservoirs, respectively. Figure 4 As shown, there were 360 ​​dry layer cases in the test set, 325 cases were correctly identified, and the recognition rate was 87.1%; 277 water layer cases were correctly identified, 260 cases were correctly identified, and the recognition rate was 90.3%; 51 oil layer cases were correctly identified, 20 cases were correctly identified, and the recognition rate was 37.0%; the overall recognition rate was 85%.

[0082] Step 7), based on the results of water layer I, II, and III divided in step (2), the data set is further divided into a training set and a test set in proportion, the same as step (5).

[0083] Step 8), a classification model is established again based on the water layer, and the grid search method is used to optimize the parameters of the SVM to obtain the best classification model. The specific process is consistent with the reservoir classification model. Then the test data is input into the model for classification to obtain the classification results of the model.

[0084] By building a test confusion matrix, the accuracy of the support vector machine model training effect is shown. Figure 5 As shown in the figure, there are 38 cases of Class I water layers in the test set, 31 of which are correctly identified, with a recognition rate of 81.6%; 168 cases of Class II water layers, 143 of which are correctly identified, with a recognition rate of 86.1%; 137 cases of Class III, 121 of which are correctly identified, with a recognition rate of 81.2%; the overall recognition rate is 86%. To verify the effectiveness of the model, a single well was further tested to obtain the predicted water layer category results, as shown in the figure below. Figure 6 shown.

[0085] The present invention conducts secondary interpretation based on the previous logging curves, establishes reservoir labels, and solves the problem of inaccurate interpretation of water layers in previous logging. The present invention uses artificial intelligence methods to complete the effective division of thermal reservoirs with low cost, high accuracy and low time consumption.

[0086] The embodiment of the present application provides a geothermal water layer quality rapid identification system based on an algorithm model, which performs thermal reservoir identification based on principal component analysis and support vector machine, such as Figure 7 shown, including:

[0087] A well logging curve preprocessing unit is used to preprocess the well logging curve data of the target well in the study area;

[0088] The secondary logging interpretation unit is used to conduct secondary interpretation of the logging curves of the target wells in the study area that were previously interpreted for oil and gas layers, and to classify reservoir types based on the secondary logging interpretation results, and then correspond them with the pre-processed logging curve data, and finally establish thermal reservoir labels based on the secondary logging interpretation;

[0089] Aquifer division unit, which further divides the identified water layers to obtain three different levels of aquifers;

[0090] The principal component analysis unit is used to study the correlation of the characteristic vectors of the well logging curve using the PCA principal component analysis method, perform principal component analysis operations on various well logging curve data, reduce the data dimension while retaining more data features, and use the results as a data set;

[0091] The data set partitioning unit is used to partition the training set and the test set, and divides the data of different categories in proportion through stratified sampling; the model training and testing unit is used to establish a support vector machine model. Based on the training set, the grid search method is used to select the optimal hyperparameter combination to obtain the optimal training model, and the test set is tested to obtain the optimal selection model.

[0092] The water layer identification unit is used to identify the water layer using the selected optimal model; the system management unit is used to manage system user information, organizational information, authority information, metadata information, operation log information, etc. to support the above units.

[0093] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical solutions and concepts of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for quickly identifying geothermal water layer quality based on an algorithm model, characterized in that: The steps include: S1. Based on the oil test data, the oil-water layer, dry layer and water layer of the oil reservoir are divided into different categories, and the identifiability of the logging curves for different categories is clarified; S2. Classify the quality of geothermal water layers based on geothermal reservoir evaluation methods; S3, according to the oil-water layer, dry layer and water layer divided by the secondary logging interpretation result, the logging curve data is processed to establish the geothermal water layer label based on the secondary logging interpretation; S4. Use PCA principal component analysis method to study the correlation of logging curve feature vectors, perform principal component analysis on various logging curve data, reduce data dimension while retaining more data features, and use the obtained k principal components as data sets; S5, dividing the data set; S6. Establish an SVM classifier model, design kernel function, kernel parameters, and penalty factor parameter combination strategy, use the training set as input for training, use the grid search algorithm to traverse the parameter combination during the training process, find the optimal parameter combination based on accuracy, and finally obtain the optimal model; then send the test set to the SVM model with the optimal parameter combination to obtain the identified water layer result; S7. On the basis of rapid identification of geothermal water layers, a well logging curve data set of three classifications of water layer quality is further established according to the results of water layer quality classification in the geothermal reservoir evaluation method; S8. Repeat step S6 to obtain a specific quality identification result of the water layer.

2. The method for rapid identification of geothermal water layer quality based on an algorithm model according to claim 1, characterized in that: In step S1, the histogram method is used to correct the logging curve to reduce the error caused by the difference in the test system; combined with the results of the oil-water layer, dry layer and water layer in the oil test data, the scatter plot is used to perform the correlation of the logging curve to clarify the identifiability of different logging curves to the water layer.

3. The method for rapid identification of geothermal water layer quality based on an algorithm model as claimed in claim 2, characterized in that: In step S2, based on the geothermal reservoir evaluation method, the water layer with a porosity greater than 25% is defined as Class I water layer, the water layer with a porosity between 15% and 25% is defined as Class II water layer, and the water layer with a porosity less than 15% is defined as Class III water layer.

4. The method for rapid identification of geothermal water layer quality based on an algorithm model as claimed in claim 3, characterized in that: In the step S3, the logging curve data is cleaned to remove abnormal values ​​in the logging curve; the logging data is standardized to eliminate the adverse effects of dimensional differences so that the data have the same scale; The standard for judging abnormal values ​​in the box plot is based on the upper and lower limits. The logging curve data that is greater than the upper limit of the box plot or less than the lower limit of the box plot is defined as an abnormal value and deleted. The definition formula of the lower and upper limits is: Lower limit=Q1-1.5(Q3-Q1); upper limit=Q3+1.5(Q3-Q1) The lower quartile Q1 refers to the value corresponding to the 25% quantile of the data; the upper quartile Q3 refers to the value corresponding to the 75% quantile of the data; The logging data is standardized, and each logging curve is processed separately, using the following standardization formula: Among them, z i is the variable after standardization, 5. The method for rapid identification of geothermal water layer quality based on an algorithm model according to claim 4, characterized in that: In step S4, PCA is used to calculate multiple logging curves such as natural potential, natural gamma, acoustic wave time difference, density, deep lateral resistivity and shallow lateral resistivity to obtain new feature data information, and the required data volume is delineated from the multiple new features according to the cumulative contribution rate greater than 85%, and the newly formed data is used as the input data of the model.

6. The method for rapid identification of geothermal water layer quality based on an algorithm model according to claim 5, characterized in that: In step S5, the data set is divided into a training set and a test set in proportion, the training set is used for model training and parameter adjustment, and the test set is used for final verification of the model to obtain the model test results, wherein the training set accounts for 60% and the test set accounts for 40%.

7. The method for rapid identification of geothermal water layer quality based on an algorithm model according to claim 6, characterized in that: In step S6, the grid search method is used to optimize the parameters of the SVM to obtain the best parameter combination to achieve the best model effect; the grid search method arranges and combines all the parameters to form a parameter grid, and then trains each group of parameters in turn, and evaluates the model performance through a cross-validation method, and selects the parameter combination with the best cross-validation result as the final parameter of the model.

8. The method for rapid identification of geothermal water layer quality based on an algorithm model according to claim 7, characterized in that: In step S7, according to the results of the water layers I, II, and III divided in step S2, the data set is further divided into a training set and a test set in proportion, which is the same as step S5.

9. The method for rapid identification of geothermal water layer quality based on an algorithm model according to claim 8, characterized in that: In step S8, a classification model is re-established based on the water layer, and the grid search method is used to optimize the parameters of the SVM to obtain the best classification model. Then, the test data is input into the model for classification to obtain the water layer quality classification result of the model.

10. A geothermal water layer quality rapid identification system based on an algorithm model, characterized in that: include: The well logging curve preprocessing unit preprocesses the well logging curve data of the target well in the study area; The secondary logging interpretation unit divides the oil-water layer, dry layer and water layer of the oil zone reservoir into categories based on the oil test data, and establishes the corresponding relationship between the logging curve and different categories; Water layer division unit, which divides the quality of geothermal water layers based on geothermal reservoir evaluation methods; The principal component analysis unit is used to study the correlation of the characteristic vectors of the well logging curve using the PCA principal component analysis method, perform principal component analysis operations on various well logging curve data, reduce the data dimension while retaining more data features, and use the obtained k principal components as a data set; The data set partitioning unit is used to partition the training set and the test set, and divide the data of different categories in proportion through stratified sampling; The model training and testing unit is used to establish a support vector machine model. Based on the training set, the optimal hyperparameter combination is selected through the grid search method to obtain the optimal training model, and the test set is tested to obtain the optimal selection model. A water layer identification unit, which uses the selected optimal model to identify the water layer; The system management unit is used to manage system user information, organizational information, permission information, metadata information, and operation log information to support the above units.

Images