Inematic rock fluid identification method based on CXDF algorithm

Abstract

The invention provides an igneous rock fluid identification method based on a CXDF algorithm, and the method specifically comprises the following modules: S1, data preprocessing: obtaining core and logging data of a research area, and sorting the data into a table file in an xlsx format; performing logarithm processing on the specified logging data; performing normalization processing on the logging data; 78% of the preprocessed data is a training data set, and 22% of the preprocessed data is a verification data set; dividing a training data set label into two parts, namely multi-hydrocarbon and low-hydrocarbon; s2, model establishment: constructing a fluid identification model based on a CXDF algorithm and other artificial intelligence algorithms; constructing a simulation well based on the verification data set; identifying the simulated well according to the fluid identification model; s3, model optimization: performing model evaluation and model comparison based on a table, a radar map and a histogram; and determining an optimal fluid identification model. According to the method, rebalance processing is carried out on the data set based on an adaptive multi-target group cross optimization algorithm, and recognition of the complex lithologic igneous rock reservoir fluid is realized in combination with an integrated learning model.

Description

Technical Field

[0001] The present invention relates to the field of geophysics, and in particular to an igneous fluid identification method based on a CXDF algorithm. Background Art

[0002] Igneous reservoirs are widely distributed in major oil-bearing basins on Earth. With the advancement of oil and gas exploration and development technology, the industry has gradually expanded to igneous reservoirs with complex lithology, which has opened up new avenues for the growth of oil and gas production and reserves. In order to effectively develop such reservoirs, accurate identification of reservoir fluids is crucial. Traditional reservoir fluid identification methods include standard water layer comparison method, reservoir minimum resistivity method, radial resistivity method and adjacent well curve comparison method. However, these methods are often limited by reservoir conditions, such as inconsistent lithology and similar formation water salinity, which makes fluid identification difficult. Although a variety of machine learning methods (such as support vector machines, extreme gradient boosting and random forests) have been used for reservoir evaluation, there is currently no research on the application of machine learning methods to fluid identification in complex lithology igneous reservoirs.

[0003] Ensemble learning methods combine multiple weak learners into a strong learner, which not only retains the characteristics of weak learners but also improves the overall effect of the model. Typical ensemble learning methods include random forest, AdaBoost, GBRT and XGBoost. These methods have been widely used in lithology identification, thin layer identification and reservoir physical property inversion. Ensemble learning methods provide potential advantages for the development of complex igneous reservoirs. However, the imbalance of conventional logging data sets poses a challenge to fluid identification. Classification models based on imbalanced data may be biased towards majority class samples, so the data set needs to be rebalanced. Summary of the invention

[0004] In view of the defects of the above-mentioned prior art, the present invention provides an igneous fluid identification method based on the CXDF algorithm. The present invention solves the problem of imbalance of the initial data set based on the adaptive multi-objective group cross optimization algorithm, and adopts an integrated learning method combining the extreme gradient boosting tree algorithm and the deep forest algorithm, which can realize the effective identification of complex lithology igneous reservoir fluids.

[0005] The present invention provides a method for identifying igneous fluids based on the CXDF algorithm, which specifically includes the following modules:

[0006] S1 Data preprocessing: Obtain core and logging data in the study area, organize the data into a table file in xlsx format; perform logarithmic processing on the specified logging data; normalize the logging data; 78% of the preprocessed data is the training data set, and 22% is the validation data set; divide the training data set labels into two parts, namely high hydrocarbon and low hydrocarbon.

[0007] S2 model establishment: Based on the CXDF algorithm and other artificial intelligence algorithms, a fluid identification model is constructed; a simulated well is constructed based on the validation data set; and a simulated well is identified according to the fluid identification model.

[0008] S3 model optimization: evaluate and compare models based on tables, radar charts, and bar charts; determine the best fluid identification model.

[0009] Preferably, the fluid types in the study area described in S1 include dry layers, low oil-yielding layers, water layers, oil layers, low water-yielding layers, oil-bearing water layers and gas layers.

[0010] Preferably, the lithology of the study area described in S1 includes basalt, trachyte, diabase, andesite and tuff.

[0011] Preferably, the logging data described in S1 include neutron (CNL), density (DEN), natural gamma (GR), deep lateral resistivity (RLLD) and micro lateral resistivity (RMLL). Among them, the designated logging data described in S1 are deep lateral resistivity (RLLD) and micro lateral resistivity (RMLL).

[0012] Preferably, the multi-hydrocarbons described in S1 include oil layers and gas layers; the low hydrocarbons described in S1 include dry layers, low oil-producing layers, water layers, low water-producing layers and oil-containing water layers.

[0013] Preferably, the CXDF algorithm described in S2 is based on the optimization data set of the adaptive multi-objective group cross optimization (AMSCO) algorithm, and is combined with an integrated learning model of the XGBoost algorithm and the deep forest (DF) algorithm.

[0014] Preferably, the AMSCO algorithm proposes an adaptively optimized data sampling model to solve the data imbalance problem in classification scenarios. According to the adaptive characteristics of the model, a composite evaluation standard of reliable accuracy (RA) is proposed.

[0015] Preferably, the rebalancing process of the AMSCO algorithm specifically includes: reading the preprocessed training data set as the current training data set; evaluating the reliable accuracy based on the current training data set; dividing the current training data set that does not meet the standard into two subsets, the minority class and the majority class; optimizing the minority class subset using the oversampling technology (OSMOTE), and optimizing the majority class subset using the group instance selection technology (SIS(o_major)); coupling the two optimizations together as a unified iterator, that is, by iteratively crossing the results of the two optimizations, gradually enhancing the data mixing effect of the two optimizations until a high-quality data set is generated.

[0016] Preferably, the radar chart described in module S3 is based on F betaThe three evaluation indicators of F, precision and recall rate show the recognition effect of the fluid recognition model on the seven fluids in the simulated well. beta It is a statistical measure used to evaluate the accuracy of binary classification models. Different beta values ​​correspond to different weights of recall and precision.

[0017] Furthermore, based on the comprehensive recognition effects of tables, radar charts and bar charts, the best igneous fluid recognition model is determined to achieve accurate recognition of igneous fluids with complex lithology. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A flow chart of an igneous fluid identification method based on the CXDF algorithm provided by the present invention.

[0019] Figure 2 A flowchart of a rebalancing training data set for the AMSCO algorithm provided in an embodiment of the present invention.

[0020] Figure 3 A cross-plot of SVM classification boundaries for a preprocessed and rebalanced training data set provided by an embodiment of the present invention.

[0021] Figure 4 A structural diagram of the XDF algorithm provided in an embodiment of the present invention.

[0022] Figure 5 A radar chart showing the identification of the fluid in the model well by the fluid identification model provided in an embodiment of the present invention.

[0023] Figure 6 A histogram showing the identification of the model well fluid by the fluid identification model provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0024] In order to enable those skilled in the art to more fully understand the technical solution of this specification, the embodiments of this specification will be described in detail and completely according to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments in this specification, not all of them. Based on one or more embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the embodiments of this specification.

[0025] The present invention provides a method for identifying igneous fluids based on the CXDF algorithm, which specifically includes the following modules:

[0026] S1: data preprocessing;

[0027] S2: Model building;

[0028] S3: Model optimization.

[0029] Specifically, the S1 module includes the following steps:

[0030] (1) Obtain core and logging data of the study area and organize the data into a table file in xlsx format;

[0031] (2) Logarithmic processing of the specified logging data;

[0032] (3) Normalize the logging data;

[0033] (4) 78% of the preprocessed data is the training data set and 22% is the validation data set;

[0034] (5) The labels of the training dataset are divided into two parts, namely, high hydrocarbon and low hydrocarbon.

[0035] Specifically, the S2 module includes the following steps:

[0036] (6) Construct a fluid recognition model based on the CXDF algorithm and other artificial intelligence algorithms;

[0037] (7) constructing simulated wells based on the validation data set;

[0038] (8) Identify the simulated wells based on the fluid identification model.

[0039] Specifically, the S3 module includes the following steps:

[0040] (9) Model evaluation and comparison based on tables, radar charts, and bar charts;

[0041] (10) Determine the optimal fluid identification model.

[0042] The fluid types in the study area include dry layers, low oil-producing layers, water layers, oil layers, low water-producing layers, oil-bearing water layers, and gas layers.

[0043] The lithologies in the study area include basalt, trachyte, diabase, andesite and tuff.

[0044] Well logging data include neutron (CNL), density (DEN), natural gamma (GR), deep lateral resistivity (RLLD) and micro lateral resistivity (RMLL).

[0045] As shown in Table 1, the core and logging data of the study area are organized into a table file in xlsx format based on Excel to facilitate subsequent processing.

[0046] Table 1: Part of the initial dataset

[0047]

[0048] The designated logging data are deep lateral resistivity (RLLD) and micro lateral resistivity (RMLL). Among them, RLLD and RMLL are processed by logarithm according to the "LOG" function in the Excel table, such as the initial value of RLLD is 6.720, which is 0.827 after processing; the initial value of RMLL is 5.310, which is 0.725 after processing.

[0049] Specifically, based on formula (1), the logging data is normalized and the expression is as follows:

[0050]

[0051] In the formula, X represents the initial data set, represents the mean of the data set, and s represents the standard deviation of the data set.

[0052] As shown in Table 2, in the preprocessed data set, the number of oil layer samples is significantly higher than that of other fluid samples, while gas layer samples only account for 1.98% of the total sample volume. This data set shows obvious data imbalance characteristics.

[0053] Table 2: Distribution of reservoir types after preprocessing

[0054]

[0055] High hydrocarbons include oil layers and gas layers; low hydrocarbons include dry layers, low oil-producing layers, water layers, low water-producing layers and oil-bearing water layers.

[0056] Based on the CXDF algorithm and other artificial intelligence algorithms, the training data sets were trained separately to establish corresponding fluid recognition models. Among them, the fluid recognition models constructed by other artificial intelligence algorithms were used as the control group, and the fluid recognition models constructed by the CXDF algorithm were compared with the models to select the best model.

[0057] The CXDF algorithm is an ensemble learning model based on the optimization dataset of the Adaptive Multi-Objective Group Cross Optimization (AMSCO) algorithm and combined with the Extreme Gradient Boosting Tree (XGBoost) algorithm and the Deep Forest (DF) algorithm.

[0058] The core optimizer of the AMSCO algorithm is particle swarm optimization (PSO). PSO is a typical optimization algorithm for simulating the foraging process of birds, with the advantages of simple implementation, fast convergence speed and few parameters.

[0059] The optimization process of PSO is as follows:

[0060] The initial dataset is read as the current dataset; the quality of the dataset is verified by the candidate classifier; the dataset is optimized twice in parallel to change the sample distribution until the performance of the candidate classifier reaches the threshold.

[0061] Specifically, two parallel optimizations produced two candidate solutions, which were different in effect. By selecting the best dataset instance and performing cross-output merging, a comprehensive dataset with a size close to the initial dataset was finally formed. During the optimization process, an iterative cycle optimization strategy was adopted. When the performance of the candidate classifier was lower than the preset threshold, the dataset was re-partitioned and two parallel group optimization processes were performed.

[0062] During the iterative loop optimization process, each particle will iterate according to its own speed and position. The mathematical process of particle iteration is as follows:

[0063]

[0064]

[0065] In the formula, ω is the inertia weight, d is the spatial dimension, i is the number of particles, j is the number of iterations, and c is 1 and c 2 is the speed weight, r 1 and c 2 is the random weight, V is the velocity of the current particle. At the beginning of the iteration, the parameters of the particle are random, P is the fitness index, and X is the set of all particles.

[0066] In an embodiment of the present invention, the AMSCO algorithm proposes an adaptively optimized data sampling model to solve the data imbalance problem in classification scenarios. According to the adaptive characteristics of the model, a composite evaluation standard of reliable accuracy (RA) is proposed, and its expression is as follows:

[0067]

[0068]

[0069]

[0070]

[0071]

[0072] In the formula, Accuracy is the accuracy, Kappa is the credibility index, P is positive, N is negative, TP is true positive, TN is true negative, FP is false positive, FN is false negative ...TP is true positive, TN is true negative, TP is true positive, TP is true positive, 0 is the percentage of consistency, P c It is a consistent trend indicator.

[0073] The rebalancing process of the AMSCO algorithm specifically includes: reading the preprocessed training data set as the current training data set; evaluating the reliable accuracy based on the current training data set; dividing the current training data set that does not meet the standard into two subsets: the minority class and the majority class; optimizing the minority class subset using the oversampling technique (OSMOTE), and optimizing the majority class subset using the group instance selection technique (SIS(o_major)); coupling the two optimizations together as a unified iterator, that is, by iteratively crossing the results of the two optimizations, gradually enhancing the data mixing effect of the two optimizations until a high-quality data set is generated, such as Figure 2 As shown. In this process, there are three types of evaluation rules:

[0074]

[0075]

[0076]

[0077] In the above equations (4)-(6), each subgroup divides and searches the space in parallel and shares the best information to find the optimal solution in the shortest time.

[0078] The OSMOTE technique inserts the generated samples into any or several minority class nearest neighbor segments within the data range and optimizes them through PSO; the SIS (o_major) technique uses the classification evaluation metric of the learner to select the optimal length between instances.

[0079] The AMSCO algorithm has the following beneficial effects:

[0080] (1) Rebalance the instances through group optimization and gradually optimize the dataset until it reaches the optimal state.

[0081] (2) Optimize the imbalanced dataset through evaluation indicators to make it a balanced dataset.

[0082] (3) Balance the dataset without changing its size.

[0083] Furthermore, the classification boundary comparison cross-plot of the preprocessed logging data and the rebalanced logging data was drawn based on the support vector machine (SVM) algorithm to intuitively judge the optimization effect of the rebalanced data set.

[0084] Specifically, the support vector machine (SVM) algorithm is based on the RBF kernel function, and the accuracy evaluation index is used to show the comparison effect of the two-dimensional classification boundary. The above-mentioned well logging data uses natural gamma (GR) and neutron (CNL). Among them, Accuracy represents the overall prediction ability of the model, as shown in formula (5).

[0085] Figure 3 In the figure, blue scatter points represent the majority class of dry layer data, and the blue background represents the classification range of dry layer data; red scatter points represent the minority class of oil layer data, and the red background represents the classification range of oil layer data. After rebalancing, the classification range of oil layer has been expanded, and the classification boundary has moved toward the dry layer. At the same time, the classification accuracy has increased from 76.596% to 82.879%. This shows that after the AMSCO algorithm rebalances the data set, the classification range of the minority class has been expanded, the classification boundary has been compressed to the majority class area, and the classification accuracy has been improved.

[0086] Other artificial intelligence algorithms in the embodiments of the present invention include SVM algorithm, XGBoost algorithm and XDF algorithm. The specific learning process of each algorithm is as follows:

[0087] The support vector machine (SVM) algorithm is a machine learning algorithm that aims to achieve the classification of data points of different categories by finding the optimal hyperplane. The SVM algorithm uses the support vectors in the training data set to establish the optimal decision boundary and can handle linear and nonlinear classification problems and regression modeling. When dealing with nonlinear problems, the SVM algorithm maps the input data to a high-dimensional feature space by introducing a kernel function, thereby performing inner product operations in the high-dimensional space to improve the classification and regression capabilities.

[0088] The SVM algorithm in the embodiment of the present invention is a machine learning algorithm based on radial basis function (RBF kernel function), and the specific formula is as follows:

[0089]

[0090] In the formula, K(x i ,x j ) is the inner product kernel, x i and x j are two different samples; σ is a free parameter.

[0091] The extreme gradient boosting (XGBoost) algorithm is an ensemble learning method and belongs to the additive model in the boosting tree (Boosting) algorithm. XGBoost effectively suppresses the overfitting of the model by combining the gradient boosting mechanism with L1 and L2 regularization, thereby enhancing its generalization ability. Its calculation formula is as follows:

[0092]

[0093] In the formula, is the prediction result of sample i after the tth iteration of the model, is the prediction result of the first t-1 weak classifiers, f t (X i) is the model of the tth weak classifier, and F is the set of all possible weak classifiers.

[0094] Specifically, the objective function of the XGBoost algorithm consists of a loss function and a regular term. The objective function is minimized through iteration. The objective function expression after iteration is as follows:

[0095]

[0096]

[0097] In the formula, L is the loss function, n is the number of samples, Obj (t) is the objective function of the tth iteration, and Ω is a regular term that suppresses the complexity of the model to prevent overfitting.

[0098] Specifically, the Boosting algorithm is characterized by integrating multiple decision trees (weak classifiers) together to build a strong classifier. The decision tree uses the CART regression tree model, and its implementation process includes the following steps:

[0099] The first step is to determine the best segmentation variable j and segmentation point s to obtain the (j, s) with the minimum error.

[0100]

[0101] In the second step, the sample features are divided into two nodes using the best (j, s), and the above steps are repeated until the termination condition is met. Finally, the input space is divided into M regions, and a decision tree is generated.

[0102] c m =average(y i |x i ∈R m (j,s)) (17)

[0103]

[0104] The XDF algorithm is an integrated learning model that combines the Extreme Gradient Boosting Tree (XGBoost) algorithm and the Deep Forest (DF) algorithm. The essential difference between the XDF algorithm and the CXDF algorithm is that the XDF algorithm does not use the AMSCO algorithm to rebalance the initial data set.

[0105] Specifically, Deep Forest (DF) is a multi-layer ensemble learning method based on decision trees, where each layer consists of multiple forests. In the data sampling stage, Deep Forest uses a multi-granularity scanning method to enrich the sample data set. In the basic Deep Forest model, the output feature vector of each layer of the forest is synthesized by averaging the output vector of the layer. The number of layers of Deep Forest is not a fixed hyperparameter, but can be automatically adjusted according to the complexity of the data set. In addition, Deep Forest has strong robustness, and good performance can usually be obtained using the default parameters, which indicates that no special parameter optimization is required when applying Deep Forest.

[0106] In the embodiment of the present invention, for continuous data, there is usually a strong spatial correlation between adjacent data. In order to improve the performance of the classifier, a multi-granularity scanning method can be used to fill the data samples. Specifically, taking a data sequence with a length of 600 as an example, the scanning window length is set to 100, the step size is set to 1, and the final generated data length is 3006. Figure 4 shown.

[0107] Figure 4 In the deep forest, the data is input into a deep forest structure, which consists of two or more different types of random forests at each layer, namely random forest and full random forest. The full random forest contains 1000 decision trees, and each decision tree randomly selects a feature when selecting a branch node. The random forest also contains 1000 decision trees, but the number of features selected by each decision tree is the square root of the total number of features in the dataset, and the Gini coefficient is used to determine the branching node. In each layer of the forest, each decision tree generates a feature vector, which is aggregated by extreme gradient boosting technology to form the output feature vector of the layer. After each layer is trained, the sample data is verified using the K-fold cross validation method. The entire deep forest structure is output layer by layer, and the final result is aggregated by the XGBoost algorithm as a weak learner.

[0108] Furthermore, a reservoir model with a depth of 28.126 m was constructed based on 226 sets of data of 7 fluid types in the validation dataset for the simulated well.

[0109] The table shows the recognition effect of the fluid recognition model on the seven fluids in the simulated well based on the two evaluation indicators of precision and recall, as shown in Table 3, Table 4, Table 5 and Table 6.

[0110] Precision indicates the accuracy of the model's prediction of positive samples, and its expression is as follows:

[0111]

[0112] Recall rate indicates the model's ability to identify positive samples, and its expression is as follows:

[0113]

[0114] Table 3: Reservoir identification results of CXDF model

[0115]

[0116] Table 4: Reservoir identification results of SVM model

[0117]

[0118] Table 5: Reservoir identification results of the XGboost model

[0119]

[0120] Table 6: Reservoir identification results of the XDF model

[0121]

[0122] Tables 3 to 6 show that the fluid identification models constructed by the SVM algorithm, XGBoost algorithm, and XDF algorithm performed poorly in the identification of water layer data, and most water layers were misidentified as low-yield reservoirs or oil reservoirs. The recall rates of these three methods were 0.104, 0.125, and 0.042, respectively. The CXDF algorithm has improved in this regard, and all 48 water layer samples were correctly identified. At the same time, in the control group, there were also cases where gas layers were misidentified as oil layers. Specifically, the SVM algorithm failed to identify the gas layer; the XGBoost algorithm only accurately identified 13 of the 30 gas layer samples; and the XDF algorithm accurately identified 11 gas layer samples. In contrast, the CXDF algorithm can accurately identify gas layer data, and its precision and recall are both 1. In the identification results of the CXDF algorithm, the precision of low-yield water reservoirs is 0.769, mainly due to the lack of low-yield water reservoir data sets, which also affects the recall rate of oil reservoirs. In addition, there are errors in the identification between water layers and oil-bearing water layers, among which 4 oil-bearing water layer samples are misidentified as water layers, which is related to the similarity of the physical properties of the two fluids, but the error is within an acceptable range. Overall, the CXDF method improves the precision and recall rate indicators of the model by improving data imbalance.

[0123] Radar chart based on F beta The three evaluation indicators of accuracy, precision and recall rate show the recognition effect of the fluid identification model on the seven fluids in the simulated wells.

[0124] F betaIt is a statistical measure used to evaluate the accuracy of binary classification models. It is a comprehensive evaluation standard that aims to reflect the overall performance of the classifier. Different beta values ​​correspond to different weights of recall and accuracy, thereby adjusting the focus of model evaluation. Its expression is as follows:

[0125]

[0126] Figure 5 In the figure, each radar chart represents a fluid category, which is represented by the prediction rate, recall rate, F 0.5 、F 1 、F 2 Etc. evaluation indexes. beta The evaluation index can evaluate the fluid identification performance of the model from different angles. The red line in the figure represents the evaluation index of the CXDF algorithm. Although the recognition effect of low water-yielding layers is inferior to that of other methods, overall, the CXDF algorithm performs better than other methods in the control group in the radar chart.

[0127] Figure 6 It is shown that the CXDF algorithm made errors in identifying the oil layer and the boundary between the water-bearing oil layer and the water layer. According to the conventional curve, the incorrectly identified layer segments showed corresponding fluctuations on the curve. However, the other fluid identification results were quite accurate. At the same time, the identification results of the control group were inferior to the CXDF method in all cases.

[0128] In summary, the CXDF algorithm performs better than other artificial intelligence algorithms in tables, radar charts, and bar charts. A fluid identification model is constructed based on the CXDF algorithm to accurately identify igneous fluids with complex lithology.

[0129] Those skilled in the art should understand that the above one or more embodiments are only used to describe the technical solution of the present invention, rather than to limit it. One or more embodiments of this specification may have various modifications and changes without making the essence of the corresponding technical solution depart from the scope of the technical solution of each embodiment of the present invention.

Claims

1. The present invention provides a method for identifying igneous fluids based on the CXDF algorithm, characterized in that: Specifically includes the following modules: S1 Data preprocessing: Obtain the core and logging data of the study area, organize the data into a table file in xlsx format; perform logarithmic processing on the specified logging data; and normalize the logging data; 78% of the preprocessed data is used as training data set, and 22% is used as validation data set; the training data set labels are divided into two parts, namely, high hydrocarbon and low hydrocarbon; S2 model establishment: build a fluid identification model based on the CXDF algorithm and other artificial intelligence algorithms; build a simulated well based on the validation data set; identify the simulated well according to the fluid identification model; S3 model optimization: evaluate and compare models based on tables, radar charts, and bar charts; determine the best fluid identification model.

2. The igneous fluid identification method based on the CXDF algorithm according to claim 1, characterized in that: The fluid types in the study area described in S1 include dry layers, low oil-producing layers, water layers, oil layers, low water-producing layers, oil-bearing water layers, and gas layers.

3. The igneous fluid identification method based on the CXDF algorithm according to claim 1 is characterized in that: The lithologies of the study area described in S1 include basalt, trachyte, diabase, andesite, and tuff.

4. The igneous fluid identification method based on the CXDF algorithm according to claim 1 is characterized in that: The logging data described in S1 include neutron (CNL), density (DEN), natural gamma (GR), deep lateral resistivity (RLLD) and micro lateral resistivity (RMLL), and the designated logging data described in S1 are deep lateral resistivity (RLLD) and micro lateral resistivity (RMLL).

5. The igneous fluid identification method based on the CXDF algorithm according to claim 1 is characterized in that: The multi-hydrocarbons described in S1 include oil layers and gas layers; the low hydrocarbons described in S1 include dry layers, low oil-producing layers, water layers, low water-producing layers and oil-containing water layers.

6. The igneous fluid identification method based on the CXDF algorithm according to claim 1, characterized in that: The CXDF algorithm described in S2 is an optimization dataset based on the adaptive multi-objective group cross optimization (AMSCO) algorithm, combined with an integrated learning model of the XGBoost algorithm and the Deep Forest (DF) algorithm.

7. The igneous fluid identification method based on the CXDF algorithm according to claim 6 is characterized by: The AMSCO algorithm proposes an adaptively optimized data sampling model to solve the data imbalance problem in classification scenarios. According to the adaptive characteristics of the model, a composite evaluation criterion of reliable accuracy (RA) is proposed, which is expressed as follows: Among them, Accuracy is the accuracy, Kappa is the credibility index, P is positive, N is negative, TP is true positive, TN is true negative, FP is false positive, FN is false negative, P0 is the consistency percentage, P c It is a consistent trend indicator.

8. The igneous fluid identification method based on the CXDF algorithm according to claim 6 or 7, characterized in that: The rebalancing process of the AMSCO algorithm specifically includes: The preprocessed training dataset is read as the current training dataset; the reliable accuracy is evaluated based on the current training dataset; the current training dataset that does not meet the standard is divided into two subsets: the minority class and the majority class; the minority class subset is optimized using the oversampling technology (OSMOTE), and the majority class subset is optimized using the group instance selection technology (SIS(o_major)); the two optimizations are coupled together as a unified iterator, that is, by iteratively crossing the results of the two optimizations, the data mixing effect of the two optimizations is gradually enhanced until a high-quality dataset is generated.

9. The igneous fluid identification method based on the CXDF algorithm according to claim 1, characterized in that: The radar chart described in S3 is based on F beta The three evaluation indicators of precision and recall rate show the recognition effect of the fluid recognition model on the seven fluids in the simulated well. beta It is a statistical measure used to evaluate the accuracy of a binary classification model. Different beta values ​​correspond to different weights of recall and precision. Its expression is as follows: Among them, Precision is the precision rate and Recall is the recall rate.

Images