CO2-rich phase corrosion prediction method based on random forest and improved ACE algorithm

Abstract

The invention provides a CO2-rich phase corrosion prediction method based on a random forest and an improved ACE algorithm, and relates to the technical field of corrosion prediction.The CO2-rich phase corrosion prediction method comprises the steps that corrosion influence data and corresponding corrosion rate data are collected, and a comprehensive database is established; based on the comprehensive database, analyzing the influence degree of each corrosion influence parameter on the corrosion rate to obtain influence degree data of different corrosion influence parameters on the corrosion rate; and establishing a carbon dioxide corrosion prediction model through the influence degree data and the comprehensive database. According to the invention, the applicable working conditions are wider: the CO2 conveying phase state is wider, and the system is suitable for various conveying phase state working conditions such as a gaseous state, a liquid state and a supercritical state; according to the invention, the impurity types are wider, and the method is suitable for the working conditions of impurities containing SO2, O2, H2O, H2S and NO2. Compared with a prediction model and other machine learning algorithm models in a traditional oil gas and transmission environment, the method is higher in prediction precision.

Description

Technical Field

[0001] The present invention relates to the technical field of corrosion prediction, and is applied to the prediction of internal corrosion in CO2 pipelines with different phases of impurities. Specifically, it relates to a method for predicting CO2-rich phase corrosion based on random forest and improved ACE algorithm. Background Art

[0002] CCUS (Carbon Capture, Utilization and Storage) is a key technology to address global climate change. Pipeline transportation is one of the most economical and efficient ways to transport CO2 from the capture site to the utilization site and the storage site. Statistical data from the Pipeline and Hazardous Materials Safety Administration (PHMSA) in the United States shows that 45% of CO2 pipeline failures are due to corrosion. Therefore, accurate prediction of the corrosion rate is very important for ensuring the safety of CO2 pipelines. However, there are many influencing factors for CO2 pipeline corrosion, which pose challenges to accurately predicting the corrosion rate. Restricted by the CO2 gas source, capture method and treatment cost, etc., impurities such as H2O, SO2, O2, NO2, H2S, etc. will inevitably exist in CO2. The effects of various impurities on CO2 corrosion are different, and the combined action of multiple impurities makes the corrosion of the pipeline very complex.

[0003] In view of the problems of the prior art, the present invention provides a method for predicting CO2-rich phase corrosion based on random forest and improved ACE algorithm. Summary of the Invention

[0004] In view of the problems of the current prior art, the present invention provides a method for predicting CO2-rich phase corrosion based on random forest and improved ACE algorithm, and the method includes:

[0005] Collect corrosion influence data and corresponding corrosion rate data, and establish a comprehensive database;

[0006] Based on the comprehensive database, analyze the influence degree of each corrosion influence parameter on the corrosion rate respectively, and obtain the influence degree data of different corrosion influence parameters on the corrosion rate;

[0007] Establish a carbon dioxide corrosion prediction model through the influence degree data and the comprehensive database.

[0008] According to an embodiment of the present invention, the corrosion influence data includes but is not limited to: environmental data, fluid data, flow data, pipeline data.

[0009] According to an embodiment of the present invention, the environmental data includes but is not limited to: transportation temperature, transportation pressure, and the phase state of carbon dioxide corresponding to different temperature and pressure conditions.

[0010] According to an embodiment of the present invention, the fluid data includes but is not limited to: types of impurities in carbon dioxide and proportions of each component, wherein the types of impurities include but are not limited to: H2O, H2S, NO2, O2, SO2, NO.

[0011] According to an embodiment of the present invention, the flow data includes but is not limited to: the flow rate of carbon dioxide.

[0012] According to an embodiment of the present invention, the pipeline data includes but is not limited to: the type of pipeline material for transportation, the surface roughness of the transportation pipeline, the inclination angle of the transportation pipeline.

[0013] According to an embodiment of the present invention, the corrosion influence parameters include but are not limited to: transportation temperature, transportation pressure, H2O, H2S, NO2, O2, SO2, NO, the flow rate of carbon dioxide, transportation time, the surface roughness of the transportation pipeline, the inclination angle of the transportation pipeline, the type of pipeline material for transportation.

[0014] According to an embodiment of the present invention, the comprehensive database is obtained through the following steps:

[0015] For the case of incomplete data, fill it with data under similar experimental conditions;

[0016] For the case of duplicate data, perform a deletion operation;

[0017] For outliers, delete them after drawing a box plot and performing cluster analysis.

[0018] According to an embodiment of the present invention, the influence degree data of different corrosion influence parameters on the corrosion rate is obtained through the following steps:

[0019] Based on the comprehensive database, use the random forest algorithm for feature selection to obtain the ranking of the influence degree of different corrosion influence parameters on the corrosion rate, as the influence degree data.

[0020] According to an embodiment of the present invention, the carbon dioxide corrosion prediction model is established through the following steps:

[0021] Using the improved ACE algorithm, calculate the optimal conversion coefficients of each corrosion influence parameter and the optimal conversion coefficient of the corrosion rate, and perform polynomial fitting on each corrosion influence parameter and its optimal conversion coefficient respectively to obtain the fitting formula of the optimal conversion coefficient of each corrosion influence parameter;

[0022] In the improved ACE algorithm, use the stepwise regression method to perform multiple linear fitting on the optimal conversion coefficient of the corrosion rate and the optimal conversion coefficients of each corrosion influence parameter to obtain the fitting formula of the optimal conversion coefficient of the corrosion rate;

[0023] Plot a scatter diagram of the corrosion rate and the optimal conversion coefficient of the corrosion rate, and perform polynomial fitting on the scatter diagram to obtain a corrosion rate fitting formula;

[0024] Based on the fitting formulas of the optimal conversion coefficients of the various corrosion influence parameters, the fitting formula of the optimal conversion coefficient of the corrosion rate, and the corrosion rate fitting formula, the carbon dioxide corrosion prediction model for calculating the corrosion rate through the various corrosion influence parameters is obtained.

[0025] According to an embodiment of the present invention, the improved ACE algorithm is an improved alternating conditional expectation algorithm.

[0026] According to an embodiment of the present invention, in the improved ACE algorithm, the fitting formula of the optimal conversion coefficient of the corrosion rate is obtained through the following steps:

[0027] Based on the influence degree data, sort the different corrosion influence parameters in descending order of the influence degree on the corrosion rate to obtain a sorted sequence;

[0028] Establish a fitting formula containing n corrosion influence parameters, and record the current goodness of fit;

[0029] n = n - 1, remove the corrosion influence parameter with the lowest influence degree in the sorted sequence to obtain a remaining sorted sequence, and establish a fitting formula containing n corrosion influence parameters, and record the current goodness of fit;

[0030] And so on, successively remove the corrosion influence parameter with the lowest influence degree in the remaining sorted sequence, construct a new fitting formula, and record the current goodness of fit;

[0031] Taking into account the number of corrosion influence parameters and all the goodness of fit values, select a fitting formula as the fitting formula of the optimal conversion coefficient of the corrosion rate.

[0032] According to an embodiment of the present invention, the fitting formula containing n corrosion influence parameters is:

[0033]

[0034] In the formula, is the optimal conversion coefficient of the corrosion rate; v corr is the corrosion rate; A0, A i are coefficients; is the optimal conversion coefficient of the i-th corrosion influence parameter; X i is the i-th corrosion influence parameter.

[0035] According to an embodiment of the present invention, the method further includes:

[0036] The carbon dioxide corrosion prediction model is respectively established in the gaseous, liquid, and supercritical states.

[0037] According to an embodiment of the present invention, the method further includes:

[0038] Compare the accuracy of the carbon dioxide corrosion prediction model with the corrosion prediction models established by SVM and BP neural network to obtain the error results of each model.

[0039] According to another aspect of the present invention, there is also provided a storage medium, which includes a series of instructions for executing the method steps described in any one of the above.

[0040] According to another aspect of the present invention, there is also provided a device for predicting the corrosion of a CO₂-rich phase based on a random forest and an improved ACE algorithm, which executes the method described in any one of the above. The device includes:

[0041] A comprehensive database module, which is used to collect corrosion influence data and corresponding corrosion rate data and establish a comprehensive database;

[0042] An influence degree module, which analyzes the influence degree of each corrosion influence parameter on the corrosion rate based on the comprehensive database to obtain the influence degree data of different corrosion influence parameters on the corrosion rate;

[0043] A corrosion prediction module, which establishes a carbon dioxide corrosion prediction model through the influence degree data and the comprehensive database.

[0044] The present invention provides a method for predicting the corrosion of a CO₂-rich phase based on a random forest and an improved ACE algorithm. Compared with the prior art, it has the following advantages:

[0045] (1) The working conditions applicable to the present invention are more extensive: The CO₂ transportation phase state of the present invention is more extensive, applicable to various transportation phase state working conditions such as gaseous, liquid, and supercritical states; the types of impurities in the present invention are more extensive, applicable to working conditions containing impurities such as SO₂, O₂, H₂O, H₂S, and NO₂.

[0046] (2) The present invention uses a random forest for feature selection to obtain the ranking of the influence degree of different corrosion influence parameters on the corrosion rate, has good tolerance for outliers and noise, and has better prediction and classification performance compared with decision trees.

[0047] (3) The present invention improves the Alternating Conditional Expectation (ACE) algorithm in the prior art, removes the corrosion influence parameters with the lowest influence degree in the remaining sorting sequence one by one, obtains multiple fitting formulas, and finally comprehensively considers the number of corrosion influence parameters and all goodness-of-fit values, and selects a fitting formula as the optimal conversion coefficient fitting formula for the corrosion rate, so as to select the best optimal conversion coefficient fitting formula for the corrosion rate.

[0048] (4) The CO2 prediction accuracy of the present invention is more accurate. Compared with the prediction models and other machine learning algorithm models in the traditional oil and gas and transportation environments, the prediction accuracy is higher.

[0049] Other features and advantages of the present invention will be described in the subsequent description, and in part, will become apparent from the description, or will be understood by implementing the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the structures specifically pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The drawings are used to provide a further understanding of the present invention, and constitute a part of the description, and are used together with the embodiments of the present invention to explain the present invention, and do not constitute a limitation to the present invention. In the drawings:

[0051] Figure 1 The flowchart showing the steps of a CO2-rich phase corrosion prediction method based on a random forest and an improved ACE algorithm according to an embodiment of the present invention;

[0052] Figure 2 The flowchart showing the steps of a CO2-rich phase corrosion prediction method based on a random forest and an improved ACE algorithm according to another embodiment of the present invention;

[0053] Figure 3 The bar chart showing the importance degree of different corrosion influence parameters according to an embodiment of the present invention;

[0054] Figure 4 The function relationship diagram showing the relationship between each corrosion influence parameter and its corresponding optimal conversion coefficient in the case of gaseous CO2 according to an embodiment of the present invention;

[0055] Figure 5 The relationship curve showing the relationship between the optimal conversion coefficient of the corrosion rate and the sum of the optimal conversion coefficients of each corrosion influence parameter in the case of gaseous CO2 according to an embodiment of the present invention;

[0056] Figs. 6(a)-6(i) show the function relationship diagrams showing the relationship between each corrosion influence parameter and its corresponding optimal conversion coefficient in the case of liquid CO2 according to an embodiment of the present invention;

[0057] Figure 7 Shows the relationship curve between the optimal conversion coefficient of the corrosion rate and the sum of the optimal conversion coefficients of each corrosion influencing parameter in the case of liquid CO2 according to an embodiment of the present invention;

[0058] Figure 8 Shows the functional relationship diagram between each corrosion influencing parameter and its corresponding optimal conversion coefficient in the case of supercritical CO2 according to an embodiment of the present invention;

[0059] Figure 9 Shows the relationship curve between the optimal conversion coefficient of the corrosion rate and the sum of the optimal conversion coefficients of each corrosion influencing parameter in the case of supercritical CO2 according to an embodiment of the present invention.

[0060] In the drawings, the same components are denoted by the same reference numerals. Additionally, the drawings are not drawn to actual scale. Detailed implementation manners

[0061] To make the objectives, technical solutions, and advantages of the present invention clearer, the following further describes the embodiments of the present invention in detail with reference to the accompanying drawings.

[0062] The prior arts (202010028987.3 A method for establishing a CO2 corrosion prediction chart considering multiple factors), the prior arts (202010692956.8 A method for predicting CO2 corrosion rate considering the influence of corrosion product film), the prior arts (Prediction model of CO2 corrosion in oil and gas fields, Chinese Journal of Corrosion and Protection, April 2005), the prior arts (Research on CO2 corrosion prediction model in offshore oil and gas fields, Contemporary Chemical Industry, October 2016) have established CO2 corrosion prediction models for their respective oil and gas environments. However, the CO2 corrosion prediction models established by these prior arts all belong to the traditional oil and gas extraction and transportation pipeline environments. The corrosion environment has an essential difference from the CO2 transportation environment applicable to the present invention. In the traditional oil and gas extraction and transportation pipeline environment, the solvent is H2O, and the solutes are CO2 and inorganic salts, etc. The solvent in the CO2 transportation environment applicable to the present invention is CO2, and the solutes are impurities such as H2O, SO2, and O2.

[0063] The prior art (A Mechanistic Model for Pipeline Steel Corrosion in Supercritical CO2-SO2-O2-H2O Environments, The Journal of Supercritical Fluids, June 2013) described the influence of the corrosive thin liquid film on the corrosion system based on the traditional CO2 corrosion prediction theory and established a corrosion prediction model for supercritical-phase CO2 containing SO2, O2, and H2O impurities. This model has a high degree of agreement with some experimental results, but its prediction accuracy is poor under liquid and gaseous CO2 transportation conditions and is not applicable to working conditions containing H2S and NO2.

[0064] Aiming at the above defects of the prior art, the present invention aims to establish a set of corrosion prediction models applicable to CO2 transportation conditions with different phases (liquid, gaseous, supercritical phase) containing impurities (SO2, O2, H2O, H2S, NO2).

[0065] Compared with the prior art, the present invention respectively constructs corrosion prediction models for gaseous, liquid, and supercritical-phase CO2 containing impurities based on the random forest and improved alternating conditional expectation (RF-MACE) algorithms. The types of impurities considered are: SO2, O2, H2O, H2S, NO2, NO, and the types of impurities are more comprehensive.

[0066] Figure 1 The flowchart of the steps of a CO2-rich phase corrosion prediction method based on the random forest and improved ACE algorithm according to an embodiment of the present invention is shown.

[0067] As Figure 1 shown, in step S101, corrosion influence data and corresponding corrosion rate data are collected to establish a comprehensive database.

[0068] In one embodiment, the corrosion influence data includes but is not limited to: environmental data, fluid data, flow data, and pipeline data.

[0069] Specifically, the environmental data includes but is not limited to: transportation temperature T, transportation pressure p, and the phase state of carbon dioxide corresponding to different temperature and pressure conditions.

[0070] Specifically, the fluid data includes but is not limited to: the types of impurities contained in carbon dioxide and the proportions of each component. Among them, the types of impurities include but are not limited to: H2O, H2S, NO2, O2, SO2, NO.

[0071] Specifically, the flow data includes but is not limited to: the flow rate v of carbon dioxide.

[0072] Specifically, the pipeline data includes but is not limited to: the material type of the conveying pipeline, the surface roughness Ra of the conveying pipeline, and the inclination angle θ of the conveying pipeline.

[0073] In one embodiment, the comprehensive database is obtained through the following steps: for the case of incomplete data, fill it with data under similar experimental conditions; for the case of duplicate data, perform a deletion operation; for outliers, delete them after drawing a box plot and performing cluster analysis.

[0074] Specifically, for the case of incomplete part of the experimental data in the database, fill it with data under similar experimental conditions; for the case of duplicate data, perform a deletion operation. For outliers, delete them after analyzing by methods such as drawing a box plot and performing cluster analysis.

[0075] As Figure 1 shown, in step S102, based on the comprehensive database, analyze the influence degree of each corrosion influence parameter on the corrosion rate respectively, and obtain the influence degree data of different corrosion influence parameters on the corrosion rate.

[0076] In one embodiment, the corrosion influence parameters include but are not limited to: the conveying temperature T, the conveying pressure p, H2O, H2S, NO2, O2, SO2, NO, the carbon dioxide flow rate v, the conveying time (or experimental time) t, the surface roughness Ra of the conveying pipeline, the inclination angle θ of the conveying pipeline, and the material type of the conveying pipeline.

[0077] In one embodiment, the influence degree data of different corrosion influence parameters on the corrosion rate is obtained through the following steps: based on the comprehensive database, use the random forest algorithm for feature selection to obtain the sorting of the influence degree of different corrosion influence parameters on the corrosion rate, which is used as the influence degree data.

[0078] Specifically, the Random Forest Algorithm is a classifier that uses multiple trees to train and predict samples, and the output category is determined by the mode of the categories output by individual trees. Random Forest is a classifier containing many decision trees, which can be used to handle classification and regression problems, and is also suitable for dimensionality reduction problems. It also has good tolerance for outliers and noise, and has better prediction and classification performance compared with decision trees.

[0079] As Figure 1 shown, in step S103, establish a carbon dioxide corrosion prediction model through the influence degree data and the comprehensive database.

[0080] In one embodiment, the carbon dioxide corrosion prediction model is established through steps S1031 - S1034.

[0081] In step S1031, using the improved ACE algorithm, the optimal conversion coefficients of each corrosion influence parameter and the optimal conversion coefficient of the corrosion rate are calculated, and polynomial fitting is performed on each corrosion influence parameter and its optimal conversion coefficient respectively to obtain the fitting formula of the optimal conversion coefficient of each corrosion influence parameter.

[0082] Specifically, the improved ACE algorithm (MACE algorithm) is used to calculate the optimal conversion coefficients of each corrosion influence parameter and the corrosion rate. Polynomial fitting is performed on the corrosion influence parameter and its optimal conversion coefficient respectively to obtain the fitting formula. Among them, the improved ACE algorithm is the improved alternating conditional expectation algorithm.

[0083] In step S1032, in the improved ACE algorithm, the stepwise regression method is used to perform multiple linear fitting on the optimal conversion coefficient of the corrosion rate and the optimal conversion coefficients of each corrosion influence parameter to obtain the fitting formula of the optimal conversion coefficient of the corrosion rate.

[0084] Specifically, in the improved ACE algorithm, the stepwise regression method is used for multiple linear fitting of the optimal conversion coefficient of the corrosion rate and the optimal conversion coefficients of each corrosion influence parameter.

[0085] In step S1033, a scatter plot of the corrosion rate and the optimal conversion coefficient of the corrosion rate is plotted, and polynomial fitting is performed on the scatter plot to obtain the corrosion rate fitting formula.

[0086] Specifically, a scatter plot of the corrosion rate v corr and the optimal conversion coefficient of the corrosion rate is plotted, and polynomial fitting is performed on the scatter plot to establish the mathematical relationship between v and, so that v corr can be calculated through . corr .

[0087] In step S1034, by synthesizing the fitting formulas of the optimal conversion coefficients of each corrosion influence parameter, the fitting formula of the optimal conversion coefficient of the corrosion rate, and the corrosion rate fitting formula, a carbon dioxide corrosion prediction model for calculating the corrosion rate through each corrosion influence parameter is obtained.

[0088] In one embodiment, in step S1032, in the improved ACE algorithm, the fitting formula of the optimal conversion coefficient of the corrosion rate is obtained through the following steps:

[0089] Step a: Based on the influence degree data, different corrosion influence parameters are sorted in descending order of their influence degree on the corrosion rate to obtain a sorting sequence. Specifically, the influence degree data predicted by the random forest is used to arrange the corrosion influence parameters from large to small according to the influence degree. ​

[0090] Step b: Establish a fitting formula with n corrosion influence parameters and record the current goodness of fit. Specifically, use Equation (1) to establish a fitting formula with n corrosion influence parameters and record the current goodness of fit R 2 .

[0091] In one embodiment, the fitting formula with n corrosion influence parameters is as follows:

[0092]

[0093] In the formula: is the optimal conversion coefficient of the corrosion rate; v corr is the corrosion rate; A0, A i are coefficients; is the optimal conversion coefficient of the i-th corrosion influence parameter; X i is the i-th corrosion influence parameter.

[0094] Step c: n = n - 1, remove the corrosion influence parameter with the lowest influence degree in the sorting sequence to obtain the remaining sorting sequence, and establish a fitting formula with n corrosion influence parameters and record the current goodness of fit. Specifically, n = n - 1, remove the corrosion influence parameter with the lowest influence degree, establish a fitting formula with n corrosion influence parameters, and record the current goodness of fit R 2 .

[0095] Step d: By analogy, gradually remove the corrosion influence parameter with the lowest influence degree in the remaining sorting sequence, construct a new fitting formula, and record the current goodness of fit. Specifically, by analogy, gradually remove the corrosion influence parameter with the lowest influence degree in the remaining corrosion influence parameter set, construct a new fitting formula, and record the current goodness of fit R 2 .

[0096] Step e: Considering comprehensively the number of corrosion influence parameters and all the goodness-of-fit values, select a fitting formula as the fitting formula for the optimal conversion coefficient of the corrosion rate. Specifically, considering comprehensively the number of corrosion influence parameters and all the goodness of fit R 2 , select the best fitting formula as the fitting formula for the optimal conversion coefficient of the corrosion rate.

[0097] Further, consider comprehensively the number of corrosion influence parameters and all the goodness of fit R through the following steps 2, select the best fitting formula as the corrosion rate optimal conversion coefficient fitting formula: sort the obtained multiple fitting formulas in descending order of goodness of fit, and sequentially select the currently highest-ranked fitting formula. Determine whether the number of corrosion influence parameters included in the current fitting formula meets the preset condition. If the judgment result is yes, select the current fitting formula as the corrosion rate optimal conversion coefficient fitting formula; if the judgment result is no, remove the current fitting formula and update the fitting formula ranking list. Among them, the preset condition is that the number of corrosion influence parameters included in the current fitting formula is greater than or equal to 5.

[0098] It should be noted that the goodness of fit refers to the degree of fitting of the regression line to the observed values. The statistic for measuring the goodness of fit is the coefficient of determination (also known as the determination coefficient) R 2 . R 2 The maximum value is 1. The closer the value of R 2 is to 1, the better the fitting degree of the regression line to the observed values; conversely, the smaller the value of R 2 , the worse the fitting degree of the regression line to the observed values. R2 measures the overall goodness of fit of the regression equation and expresses the overall relationship between the dependent variable and all independent variables.

[0099] The alternating conditional expectation (ACE) algorithm is a non-parametric regression and multivariate regression method for estimating the best transformation. Its transformation result is to maximize the correlation coefficient between a dependent variable and multiple independent variables. The basic idea is to repeatedly alternate conditional expectations to obtain the best transformation equation between the explanatory variable and the explained variable. Based on the alternating conditional expectation algorithm, the present invention proposes an improved alternating conditional expectation (MACE) algorithm, which sequentially removes the corrosion influence parameters with the lowest influence degree in the remaining sorting sequence, constructs multiple fitting formulas, and finally comprehensively considers the number of corrosion influence parameters and all goodness of fit values to select a fitting formula as the corrosion rate optimal conversion coefficient fitting formula, which can balance the goodness of fit and the number of corrosion influence parameters and obtain the best corrosion rate optimal conversion coefficient fitting formula.

[0100] In one embodiment, a CO2-rich phase corrosion prediction method based on random forest and improved ACE algorithm further includes: establishing carbon dioxide corrosion prediction models under gaseous, liquid, and supercritical states respectively.

[0101] In one embodiment, a CO2-rich phase corrosion prediction method based on random forest and improved ACE algorithm further includes: comparing the model accuracies of the carbon dioxide corrosion prediction model with the corrosion prediction models established by SVM and BP neural network to obtain the error results of each model. Specifically, compare the established corrosion prediction model with other models to verify the prediction accuracy of the model.

[0102] The present invention is applicable to the field of corrosion prediction in the CO2 transportation link during the CCUS process, especially for the corrosion prediction of CO2 transportation pipelines containing impurities. The present invention establishes a CO2 corrosion prediction model applicable to different phases (liquid, gas, supercritical phase) of impurities (SO2, O2, H2O, H2S, NO2), filling the gap in the corrosion prediction of the current multi-phase CO2 transportation working conditions with complex impurities, and having good application prospects in the CO2 transportation link during the CCUS process, especially in the field of CO2 transportation pipelines containing impurities.

[0103] Figure 2 The flowchart shows the steps of a CO2-rich phase corrosion prediction method based on the random forest and improved ACE algorithm according to another embodiment of the present invention.

[0104] In one embodiment, as Figure 2 shown, a gaseous CO2 corrosion prediction model is constructed:

[0105] As Figure 2 shown, in step S201, a comprehensive database is established: environmental data, fluid data, flow data, pipeline data, and corrosion rate data are collected.

[0106] Specifically, temperature, pressure, H2O, H2S, NO2, O2, SO2, NO, flow rate, surface roughness data, and corrosion rate data under the gaseous CO2 transportation working conditions of the X65 steel transportation pipeline are collected. Among them, in the dataset, the data of H2S, flow rate, surface roughness, and pipeline inclination angle are all constant and unchanged, so they are removed.

[0107] As Figure 2 shown, in step S202, data cleaning: incomplete data, duplicate data, and outlier data are processed.

[0108] Specifically, for the situation where some experimental data in the database is incomplete, data under similar experimental conditions are used for filling; for duplicate data, a deletion operation is performed. For outliers, they are deleted after analysis by methods such as drawing a boxplot and cluster analysis.

[0109] As Figure 2 shown, in step S203, a feature importance ranking sequence is established: the random forest algorithm is used to analyze the influence degree of each corrosion influence parameter, and they are ranked from large to small according to importance.

[0110] Specifically, the random forest algorithm is used for feature selection, and the influence degree of different corrosion influence parameters on the corrosion rate is obtained as shown in Appendix Figure 3 shown, from large to small: H2O, SO2, NO2, pressure, experimental time, temperature, O2, NO.

[0111] AsFigure 2 As shown in the figure, in step S204, a corrosion prediction model is established: a multiple linear regression formula for the optimal conversion coefficient of the corrosion rate and the optimal conversion coefficients of each corrosion influence parameter is constructed, and a corrosion prediction model based on the improved ACE algorithm is established.

[0112] 1. Calculate the optimal conversion coefficients of each corrosion influence parameter and the corrosion rate using the improved ACE algorithm Perform polynomial fitting on the corrosion influence parameter and its optimal conversion coefficient respectively, and the fitting results are as attached Figure 4 shown, and the polynomial coefficients are shown in Table 1.

[0113] Table 1 Fitting coefficients of the gaseous CO2 corrosion prediction model

[0114]

[0115] 2. Use the stepwise regression method to perform multiple linear fitting on and the optimal conversion coefficients of each corrosion influence parameter, and the specific steps are as follows:

[0116] 1) Arrange the parameters in the order of decreasing influence degree according to the results predicted by the random forest.

[0117] 2) Construct the polynomial linear fitting formula between and the optimal conversion coefficients corresponding to n corrosion influence parameters as formula (2) below, and R 2 is 0.89.

[0118]

[0119] 3) n = n - 1, remove the NO with the lowest influence degree in the sequence, and establish a fitting formula containing n parameters as formula (3) below, and R 2 is 0.93.

[0120]

[0121] 4) And so on, remove the parameter with the lowest influence degree in the remaining parameter set one by one, and construct a new fitting formula, and R 2 are 0.91, 0.92, 0.97, 0.98, 0.99, 1 respectively.

[0122] 5) Considering the total influence degree of the parameter set and the R 2 value comprehensively, the selected parameter set is H2O, SO2, NO2, pressure, experimental time, temperature, O2, and the finally selected fitting formula is formula (3).

[0123] 3. Plot the scatter diagram of the corrosion rate v corr and the optimal conversion coefficient of the corrosion rate as Figure 5As shown, the polynomial fitting formulas of the two are as shown in (4):

[0124]

[0125] As Figure 2 shown, in step S205, model verification: compare the established corrosion prediction model with other models to verify the prediction accuracy of the model.

[0126] Specifically, compare the established corrosion prediction model with SVM and BP neural network. The prediction results are shown in Table 2. The results show that the MAE error (Mean Absolute Error) of the gaseous CO2 corrosion prediction model established by the present invention is 0.0215, which is better than that of SVM (0.0377) and BP neural network (0.0433).

[0127] Table 2 Prediction error table of RF-MACE, SVM, and BP neural network algorithms

[0128] Phase state RF-MACE SVM BP neural network Gaseous state 0.0215 0.0377 0.0433

[0129] In one embodiment, as Figure 2 shown, construct a liquid CO2 corrosion prediction model:

[0130] As Figure 2 shown, in step S201, establish a comprehensive database: collect environmental data, fluid data, flow data, pipeline data, and corrosion rate data.

[0131] Specifically, collect temperature, pressure, H2O, H2S, NO2, O2, SO2, NO, flow rate, surface roughness data and corrosion rate data when the pipe material is X65 steel under the conditions of liquid CO2 transportation. Among them, the surface roughness and pipeline inclination data are both constant and unchanged, so they are removed.

[0132] As Figure 2 shown, in step S202, data cleaning: process incomplete data, duplicate data, and outlier data.

[0133] Specifically, for the situation where some experimental data in the database is incomplete, use data under similar experimental conditions to fill it; for duplicate data, perform a deletion operation. For outliers, analyze and delete them by methods such as drawing a boxplot and cluster analysis.

[0134] As Figure 2 shown, in step S203, establish a feature importance ranking sequence: use the random forest algorithm to analyze the influence degree of each corrosion influence parameter, and sort them from large to small according to importance.

[0135] Specifically, the random forest algorithm is used for feature selection, and the influence degrees of different corrosion influence parameters on the corrosion rate are obtained as shown in the appendix Figure 3 as follows, from large to small: H2O, SO2, flow rate, NO2, pressure, experimental time, H2S, temperature, O2, NO.

[0136] As Figure 2 shown, in step S204, a corrosion prediction model is established: a multiple linear regression formula for the optimal conversion coefficient of the corrosion rate and the optimal conversion coefficients of each corrosion influence parameter is constructed, and a corrosion prediction model based on the improved ACE algorithm is established.

[0137] 1. Calculate the optimal conversion coefficients of each corrosion influence parameter and the corrosion rate by using the improved ACE algorithm Polynomial fitting is performed on the corrosion influence parameters and their optimal conversion coefficients respectively, and the fitting results are shown in Figures 6(a)-6(i) of the appendix. The polynomial coefficients are shown in Table 3.

[0138] Table 3 Fitting coefficients of the liquid CO2 corrosion prediction model

[0139]

[0140]

[0141] 2. Use the stepwise regression method to perform multiple linear fitting on and the optimal conversion coefficients of each corrosion influence parameter. The specific steps are as follows:

[0142] 1) Arrange the parameters in the order of influence degree from large to small according to the results predicted by the random forest.

[0143] 2) Construct a polynomial linear fitting formula between and the optimal conversion coefficients corresponding to n corrosion influence parameters as shown in the following formula (5), and R 2 is 0.94.

[0144]

[0145] 3) n = n - 1, remove the NO with the lowest influence degree in the sequence, and establish a fitting formula containing n parameters as shown in the following formula (3), and R 2 is 0.97.

[0146]

[0147] 4) And so on, remove the parameter with the lowest influence degree in the remaining parameter set one by one, construct a new fitting formula, and R 2 are 0.95, 0.96, 0.98, 0.99, 0.99, 1, 1, 1 respectively.

[0148] 5) Considering the total influence degree of the parameter set and the R 2 value, the selected parameter set is H2O, SO2, NO2, pressure, experimental time, temperature, O2, and the finally selected fitting formula is formula (6).

[0149] 3. Plot the corrosion rate v corr and the optimal conversion coefficient of the corrosion rate The scatter plot is as Figure 7 shown, and the polynomial fitting formula of the two is as shown in (7):

[0150]

[0151] As Figure 2 shown, in step S205, model verification: compare the established corrosion prediction model with other models to verify the prediction accuracy of the model.

[0152] Specifically, compare the established corrosion prediction model with SVM and BP neural network, and the prediction results are shown in Table 4. The results show that the MAE error of the gaseous CO2 corrosion prediction model established by the present invention is 0.0342, which is better than SVM (0.0417) and BP neural network (0.0492).

[0153] Table 4 Prediction error table of RF-MACE, SVM, and BP neural network algorithms

[0154] Phase state RF-MACE SVM BP neural network Liquid state 0.0342 0.0417 0.0492

[0155] In one embodiment, as Figure 2 shown, construct a supercritical CO2 corrosion prediction model:

[0156] As Figure 2 shown, in step S201, establish a comprehensive database: collect environmental data, fluid data, flow data, pipeline data, and corrosion rate data.

[0157] Specifically, collect the temperature, pressure, H2O, H2S, NO2, SO2, NO, flow rate, surface roughness data and corrosion rate data of the pipe material X65 steel under the working conditions of supercritical CO2 transportation. Among them, the O2, surface roughness, and pipeline inclination data are all constant and unchanged, so they are removed.

[0158] As Figure 2 shown, in step S202, data cleaning: process incomplete data, duplicate data, and outlier data.

[0159] Specifically, for the case where some experimental data in the database is incomplete, data under similar experimental conditions are used for filling; for the case of duplicate data, a deletion operation is performed. For outliers, they are deleted after analysis through methods such as drawing a boxplot and cluster analysis.

[0160] As Figure 2 shown, in step S203, a sequence of feature importance rankings is established: the random forest algorithm is used to analyze the influence degrees of various corrosion influence parameters, and they are ranked from the most important to the least important.

[0161] Specifically, the random forest algorithm is used for feature selection, and the influence degrees of different corrosion influence parameters on the corrosion rate are obtained as shown in the appendix Figure 3 shown, from the most important to the least important: H2O, SO2, flow rate, NO2, pressure, experimental time, H2S, temperature, NO.

[0162] As Figure 2 shown, in step S204, a corrosion prediction model is established: a multiple linear regression formula for the optimal conversion coefficient of the corrosion rate and the optimal conversion coefficients of various corrosion influence parameters is constructed, and a corrosion prediction model based on the improved ACE algorithm is established.

[0163] 1. The optimal conversion coefficients of various corrosion influence parameters and the corrosion rate are calculated using the improved ACE algorithm Polynomial fittings are respectively performed on the corrosion influence parameters and their optimal conversion coefficients, and the fitting results are as shown in the appendix Figure 8 shown, and the polynomial coefficients are shown in Table 5.

[0164] Table 5 Fitting coefficients of the supercritical phase CO2 corrosion prediction model

[0165]

[0166] 2. The stepwise regression method is used to perform multiple linear fitting on and the optimal conversion coefficients of various corrosion influence parameters. The specific steps are as follows:

[0167] 1) Arrange the parameters according to the influence degree from the most important to the least important based on the results predicted by the random forest.

[0168] 2) Construct a polynomial linear fitting formula between and the optimal conversion coefficients corresponding to n corrosion influence parameters as shown in the following formula (8), and R 2 is 0.95.

[0169]

[0170] 3) n = n - 1, remove NO with the lowest influence degree in the sequence, and establish a fitting formula containing n parameters as shown in the following formula (9), and R2 is 0.97.

[0171]

[0172] 4) And so on, remove one by one the parameter with the lowest influence degree in the remaining parameter set, construct a new fitting formula, R 2 are 0.96, 0.96, 0.97, 0.99, 0.99, 0.99, 1 respectively.

[0173] 5) Considering comprehensively the total influence degree of the parameter set and R 2 value, the selected parameter set is H2O, SO2, flow rate, NO2, pressure, experimental time, H2S, temperature, and the finally selected fitting formula is formula (9).

[0174] 3. Plot the scatter diagram of the corrosion rate v corr and the optimal conversion coefficient of the corrosion rate as shown in Figure 9 shown, and the polynomial fitting formula of the two is as shown in (10):

[0175]

[0176] As Figure 2 shown, in step S205, model verification: compare the established corrosion prediction model with other models to verify the prediction accuracy of the model.

[0177] Specifically, compare the established corrosion prediction model with SVM and BP neural network, and the prediction results are shown in Table 4. The results show that the MAE error of the gaseous CO2 corrosion prediction model established by the present invention is 0.0426, which is better than that of SVM (0.0535) and BP neural network (0.0645).

[0178] Table 6 Prediction error table of RF - MACE, SVM, and BP neural network algorithms

[0179] Phase state RF-MACE SVM BP neural network Supercritical state 0.0426 0.0535 0.0645

[0180] A method for predicting CO2 - rich phase corrosion based on random forest and improved ACE algorithm provided by the present invention can also cooperate with a computer - readable storage medium. A computer program is stored on the storage medium, and the computer program is executed to run a method for predicting CO2 - rich phase corrosion based on random forest and improved ACE algorithm. The computer program can run computer instructions, and the computer instructions include computer program codes. The computer program codes can be in the form of source code, object code, executable file, or some intermediate form, etc.

[0181] A computer-readable storage medium may include: any entity or device capable of carrying computer program code, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0182] It should be noted that the content included in the computer-readable storage medium can be appropriately increased or decreased according to the requirements of legislation and patent practice within the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable storage medium does not include electrical carrier signals and telecommunication signals.

[0183] According to another aspect of the present invention, a rich CO2 phase corrosion prediction device based on a random forest and an improved ACE algorithm is also provided, which executes a rich CO2 phase corrosion prediction method based on a random forest and an improved ACE algorithm. The device includes: a comprehensive database module, an influence degree module, and a corrosion prediction module.

[0184] The comprehensive database module is used to collect corrosion influence data and corresponding corrosion rate data, and establish a comprehensive database; the influence degree module, based on the comprehensive database, respectively analyzes the influence degree of each corrosion influence parameter on the corrosion rate, and obtains the influence degree data of different corrosion influence parameters on the corrosion rate; the corrosion prediction module, through the influence degree data and the comprehensive database, establishes a carbon dioxide corrosion prediction model.

[0185] In summary, the present invention provides a rich CO2 phase corrosion prediction method based on a random forest and an improved ACE algorithm. Compared with the prior art, it has the following advantages:

[0186] (1) The working conditions applicable to the present invention are more extensive: the CO2 transportation phase state of the present invention is more extensive, applicable to various transportation phase state working conditions such as gaseous, liquid, and supercritical states; the impurity types of the present invention are more extensive, applicable to working conditions containing impurities such as SO2, O2, H2O, H2S, and NO2.

[0187] (2) The present invention uses a random forest for feature selection, obtains the ranking of the influence degree of different corrosion influence parameters on the corrosion rate, has good tolerance for outliers and noise, and has better prediction and classification performance compared with decision trees.

[0188] (3) The present invention improves the Alternating Conditional Expectation (ACE) algorithm in the prior art, removes the corrosion influence parameters with the lowest influence degree in the remaining sorting sequence one by one to obtain multiple fitting formulas, and finally comprehensively considers the number of corrosion influence parameters and all goodness-of-fit values to select a fitting formula as the optimal conversion coefficient fitting formula for the corrosion rate, so as to be able to select the best optimal conversion coefficient fitting formula for the corrosion rate.

[0189] (4) The CO2 prediction accuracy of the present invention is more accurate. Compared with the prediction models and other machine learning algorithm models in the traditional oil and gas and transportation environments, the prediction accuracy is higher.

[0190] It should be understood that the embodiments disclosed in the present invention are not limited to the specific structures, processing steps or materials disclosed herein, but should extend to equivalent alternatives of these features understood by those of ordinary skill in the relevant art. It should also be understood that the terms used herein are only for the purpose of describing specific embodiments and do not mean to limit.

[0191] In the description of the present invention, unless otherwise specified, "a plurality of" means two or more; the orientation or positional relationship indicated by terms such as "upper", "lower", "left", "right", "inner", "outer", "front end", "rear end", "head", "tail", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of the present invention. In addition, terms such as "first", "second", "third", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0192] In the description of the present invention, it should be noted that unless otherwise clearly specified and limited, the terms "connected" and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0193] Certain terms are used throughout this application to refer to particular system components. As those skilled in the art will recognize, the same components can generally be referred to by different names, and thus this application does not intend to distinguish between components that differ only in name and not in function. In this application, the terms "comprise," "include," and "have" are used in an open-ended fashion and should therefore be interpreted to mean "including but not limited to...". In addition, the terms "substantially," "essentially," or "approximately" as may be used herein refer to the industry-accepted tolerances for the corresponding terms. The term "coupled" as may be employed herein includes direct coupling and indirect coupling via additional components, elements, circuits, or modules, where for indirect coupling, the intervening components, elements, circuits, or modules do not change the information of the signal but may adjust its current level, voltage level, and / or power level. Inferred coupling (e.g., where one element is inferred to be coupled to another element) includes direct and indirect coupling between the two elements in the same manner as "coupled".

[0194] The phrase "an embodiment" or "embodiments" mentioned in the specification means that the particular features, structures, or characteristics described in connection with the embodiments are included in at least one embodiment of the present invention. Thus, the phrases "an embodiment" or "embodiments" that appear throughout the specification do not necessarily all refer to the same embodiment.

[0195] Embodiments of the present invention are given for purposes of illustration and description and are not exhaustive or limit the present invention to the disclosed form. Many modifications and variations are obvious to those of ordinary skill in the art. The embodiments are chosen and described in order to better illustrate the principles of the present invention and its practical application and to enable those of ordinary skill in the art to understand the present invention so as to design various embodiments with various modifications suitable for a particular purpose.

[0196] Although the embodiments disclosed in the present invention are as above, the content described is only the embodiments adopted for the convenience of understanding the present invention and is not used to limit the present invention. Any person skilled in the art within the technical field to which the present invention pertains may make any modifications and variations in the form of implementation and details without departing from the spirit and scope disclosed by the present invention. However, the scope of patent protection of the present invention shall still be subject to the scope defined by the appended claims.

Claims

1. A method for predicting CO2-rich phase corrosion based on random forest and improved ACE algorithm, characterized in that, The method includes: Collecting corrosion impact data and corresponding corrosion rate data, and establishing a comprehensive database; Based on the comprehensive database, analyzing the influence degree of each corrosion impact parameter on the corrosion rate respectively, and obtaining the influence degree data of different corrosion impact parameters on the corrosion rate; Establishing a carbon dioxide corrosion prediction model through the influence degree data and the comprehensive database.

2. The method for predicting rich CO2 phase corrosion based on random forest and improved ACE algorithm according to claim 1, wherein, The corrosion impact data includes but is not limited to: environmental data, fluid data, flow data, pipeline data.

3. The method for predicting rich CO2 phase corrosion based on random forest and improved ACE algorithm according to claim 2, characterized in that, The environmental data includes but is not limited to: transportation temperature, transportation pressure, and the phase state of carbon dioxide corresponding to different temperature and pressure conditions.

4. A method for predicting rich CO2 phase corrosion based on random forest and improved ACE algorithm according to claim 2, characterized in that, The fluid data includes but is not limited to: the types of impurities contained in carbon dioxide and the proportion of each component. Among them, the types of impurities include but are not limited to: H2O, H2S, NO2, O2, SO2, NO.

5. The method for predicting CO2-rich phase corrosion based on random forest and improved ACE algorithm according to claim 2, characterized in that, The flow data includes but is not limited to: the flow rate of carbon dioxide.

6. The method for predicting CO2-rich phase corrosion based on random forest and improved ACE algorithm according to claim 2, characterized in that, The pipeline data includes but is not limited to: the type of pipeline material for transportation, the surface roughness of the transportation pipeline, and the inclination angle of the transportation pipeline.

7. A method for predicting rich CO2 phase corrosion based on random forest and improved ACE algorithm according to any one of claims 1-6, characterized in that The corrosion impact parameters include but are not limited to: transportation temperature, transportation pressure, H2O, H2S, NO2, O2, SO2, NO, carbon dioxide flow rate, transportation time, surface roughness of the transportation pipeline, inclination angle of the transportation pipeline, type of pipeline material for transportation.

8. A method for predicting rich CO2 phase corrosion based on random forest and improved ACE algorithm according to any one of claims 1-7, characterized in that The comprehensive database is obtained through the following steps: For the case of incomplete data, filling with data under similar experimental conditions; For the case of duplicate data, performing a deletion operation; For outliers, deleting them after drawing a box plot and performing cluster analysis.

9. A method for predicting corrosion in a CO₂-rich phase based on a random forest and an improved ACE algorithm according to any one of claims 1-8, characterized in that, The influence degree data of different corrosion impact parameters on the corrosion rate is obtained through the following steps: Based on the comprehensive database, using the random forest algorithm for feature selection, obtaining the influence degree ranking of different corrosion impact parameters on the corrosion rate, which is used as the influence degree data.

10. A method for predicting CO2-rich phase corrosion based on random forest and improved ACE algorithm according to claim 9, characterized in that, The carbon dioxide corrosion prediction model is established through the following steps: Using the improved ACE algorithm, calculating the optimal conversion coefficients of each corrosion impact parameter and the optimal conversion coefficient of the corrosion rate, respectively performing polynomial fitting on each corrosion impact parameter and its optimal conversion coefficient, and obtaining the fitting formula of the optimal conversion coefficient of each corrosion impact parameter; In the improved ACE algorithm, using the stepwise regression method to perform multiple linear fitting on the optimal conversion coefficient of the corrosion rate and the optimal conversion coefficients of each corrosion impact parameter, and obtaining the fitting formula of the optimal conversion coefficient of the corrosion rate; Drawing a scatter plot of the corrosion rate and the optimal conversion coefficient of the corrosion rate, and performing polynomial fitting on the scatter plot to obtain the corrosion rate fitting formula; Integrating the fitting formulas of the optimal conversion coefficients of each corrosion impact parameter, the fitting formula of the optimal conversion coefficient of the corrosion rate, and the corrosion rate fitting formula, obtaining the carbon dioxide corrosion prediction model for calculating the corrosion rate through each corrosion impact parameter.

11. A method for predicting CO2-rich phase corrosion based on random forest and improved ACE algorithm according to claim 10, characterized in that, The improved ACE algorithm is the improved alternating conditional expectation algorithm.

12. A method for predicting CO2-rich phase corrosion based on random forest and improved ACE algorithm according to claim 10 or 11, characterized in that, In the improved ACE algorithm, the fitting formula of the optimal conversion coefficient of the corrosion rate is obtained through the following steps: Based on the influence degree data, sort different corrosion influence parameters in descending order according to the influence degree on the corrosion rate to obtain a sorting sequence; Establish a fitting formula containing n corrosion influence parameters and record the current goodness of fit; n = n - 1, remove the corrosion influence parameter with the lowest influence degree in the sorting sequence to obtain a remaining sorting sequence, establish a fitting formula containing n corrosion influence parameters, and record the current goodness of fit; And so on, gradually remove the corrosion influence parameter with the lowest influence degree in the remaining sorting sequence, construct a new fitting formula, and record the current goodness of fit; Considering comprehensively the number of corrosion influence parameters and all goodness of fit values, select a fitting formula as the optimal conversion coefficient fitting formula for the corrosion rate.

13. A method for predicting CO2-rich phase corrosion based on random forest and improved ACE algorithm according to claim 12, characterized in that, The fitting formula containing n corrosion influence parameters is: In the formula, is the optimal conversion coefficient of the corrosion rate; v corr is the corrosion rate; A0, A i are coefficients; is the optimal conversion coefficient of the i-th corrosion influence parameter; X i is the i-th corrosion influence parameter.

14. A method for predicting CO2-rich phase corrosion based on random forest and improved ACE algorithm according to any one of claims 1-13, characterized in that, The method further includes: Under gaseous, liquid, and supercritical states, establish the carbon dioxide corrosion prediction model respectively.

15. A method for predicting CO2-rich phase corrosion based on random forest and improved ACE algorithm according to any one of claims 1-14, characterized in that, The method further includes: Compare the model accuracy of the carbon dioxide corrosion prediction model with the corrosion prediction models established by SVM and BP neural network to obtain the error results of each model.

16. A storage medium, characterized in that, It includes a series of instructions for executing the method steps described in any one of claims 1-15.

17. A rich CO2 phase corrosion prediction device based on a random forest and an improved ACE algorithm, characterized in that, Execute the method described in any one of claims 1-15. The device includes: A comprehensive database module, which is used to collect corrosion influence data and corresponding corrosion rate data and establish a comprehensive database; An influence degree module, which analyzes the influence degree of each corrosion influence parameter on the corrosion rate based on the comprehensive database to obtain the influence degree data of different corrosion influence parameters on the corrosion rate; A corrosion prediction module, which establishes a carbon dioxide corrosion prediction model through the influence degree data and the comprehensive database.

Images