A method and system for monitoring corrosion of carbon dioxide capture equipment

Abstract

The present application provides a method and system for monitoring the corrosion of carbon dioxide capture equipment, relating to the technical field of corrosion monitoring. The method comprises: obtaining the real-time operating data, operating time, and pre-operation corrosion level of the current carbon dioxide capture equipment; obtaining the final corrosion level of similar equipment and training data containing historical operating data over the entire life cycle and corresponding corrosion rates; establishing a neural network model using a machine learning algorithm and training the model with the training data; inputting the real-time operating data into the trained model and outputting a predicted corrosion rate; and calculating the remaining life of the equipment based on the final corrosion level, pre-operation corrosion level, operating time, and predicted corrosion rate. The present application can monitor the corrosion status without shutting down the machine, solving the problem of delayed response of traditional methods to the complex corrosion mechanism of chemical absorption systems, and significantly improving the monitoring accuracy and reliability of long-cycle operating equipment.

Description

Technical Field

[0001] The present application relates to the technical field of corrosion monitoring, and in particular to a method and system for monitoring corrosion of carbon dioxide capture equipment. Background Art

[0002] In response to the urgent need for global climate change and environmental protection, carbon dioxide capture and storage (CCS) technology has become one of the key means to reduce greenhouse gas emissions. Chemical absorption is one of the most widely used and mature carbon capture technologies in industrial applications. In chemical absorption, organic amines have the advantages of fast absorption rate and large absorption capacity, making them the preferred absorbent. However, when carbon dioxide is dissolved in water or organic amine solution, it will react to form carbamate (pH drops from 10-12 to 6-8), causing the Fe3O4 passivation film of carbon steel equipment to dissolve (Fe3O4 + 8H + →3Fe 2+ + 4H2O); the heat-stable salts (HSS) produced by the degradation of amine solution at high temperature (110-130℃) in the regeneration tower cause the local pitting rate to reach 2-5mm / year, while the temperature alternation (40℃-120℃) and CO2 load rate fluctuation (0.2-0.8) caused by the circulation of lean and rich solutions will further induce stress corrosion cracking.

[0003] Traditional corrosion monitoring methods mainly use ultrasonic monitoring to regularly sample and analyze the degree of corrosion, and then calculate the real-time corrosion rate based on the sampling time difference (or use electrochemical monitoring to monitor the corrosion rate in real time); then calculate the remaining life of the equipment based on the real-time corrosion rate and the remaining corrosion degree of the equipment.

[0004] When using electrochemical monitoring to monitor corrosion rates in real time, if the corrosion rate is low or the rate of change is small, the sensor cannot sense the change in corrosion rate. Over time, the monitoring error of the corrosion level will drift. To overcome this drift, ultrasonic monitoring can be used to monitor the corrosion level. However, ultrasonic monitoring requires downtime inspection and is not suitable for inspections of CO2 capture equipment with long operating cycles. Therefore, for CO2 capture equipment with long operating cycles, it is difficult to ensure monitoring accuracy by using only electrochemical monitoring to monitor corrosion rate and remaining life. Summary of the Invention

[0005] In order to improve the accuracy of corrosion status monitoring results of carbon dioxide capture equipment with a long working cycle, the present application provides a carbon dioxide capture equipment corrosion monitoring method and system.

[0006] In a first aspect, the present application provides a method for monitoring corrosion of carbon dioxide capture equipment, which adopts the following technical solution:

[0007] A method for monitoring corrosion of carbon dioxide capture equipment, comprising:

[0008] Real-time data collection: Obtain real-time operating data of the current CO2 capture equipment, operating time, and corrosion level before operation;

[0009] Historical data collection: Obtain the final corrosion degree and training data of the same type of CO2 capture equipment. The training data includes: historical operating data of the same type of CO2 capture equipment throughout its life cycle, and the corrosion rate of the same type of CO2 capture equipment under each historical operating data;

[0010] Modeling and training: Use machine learning algorithms to build a neural network model, use training data to train the neural network model, and obtain the trained neural network model;

[0011] Predicting corrosion rate: Inputting real-time operating data into the trained neural network model, the output is the predicted corrosion rate;

[0012] Calculate remaining life: Calculate remaining life based on final corrosion level, corrosion level before operation, operation time and predicted corrosion rate.

[0013] The present application first obtains the real-time operating data, operating time and corrosion level before operation of the current carbon dioxide capture equipment, wherein the real-time operating data reflects the working status of the current carbon dioxide capture equipment, and the corrosion level before operation is combined with the operating time to calculate the current corrosion level of the equipment. Subsequently, the present application obtains the final corrosion level and training data of the same type of carbon dioxide capture equipment, and the final corrosion level of the same type of carbon dioxide capture equipment is used as the end point of life to calculate the remaining life in combination with the current corrosion level. Subsequently, the present application uses a machine learning algorithm to establish a neural network model, and uses the training data to train the neural network model. During the training process, the neural network model can automatically learn the potential laws and features in the data, thereby improving the prediction performance. Subsequently, the present application inputs the real-time operating data into the trained neural network model, and the trained neural network model will output the predicted corrosion rate. The present application can calculate the current corrosion state through the operating time and the corrosion level before operation, and is suitable for long-term carbon dioxide capture equipment that requires accurate calculation of the remaining life.

[0014] Optionally, the historical operation data includes m types of operation indicators. After executing the step of collecting historical data and before executing the step of modeling and training, the following steps are further included:

[0015] Clustering: K-means clustering algorithm is used to cluster the historical operation data under the i-th operation indicator to obtain clusters, where the value of K is 2;

[0016] Determine the threshold: Maximize the inter-class variance of the two clusters through the great law, and record the historical operation data corresponding to the maximum inter-class variance as the alarm threshold of the i-th operation indicator;

[0017] Threshold judgment: Determine in turn whether each real-time running data is greater than the corresponding alarm threshold. If so, a data unqualified signal is issued; if not, the modeling and training steps are executed.

[0018] This application uses the K-means clustering algorithm to cluster historical operating data for each operating indicator (K is 2). Clustering aims to separate normal operating data and abnormal operating data into different clusters. This approach simplifies complex data sets into two representative clusters, helping to discover underlying patterns and distribution characteristics in the data. Subsequently, this application uses the Otsu's method to maximize the inter-cluster variance between the two clusters to determine the alarm threshold. The Otsu method automatically finds an optimal split point in the data that maximizes the difference between the two clusters, thereby determining a reasonable alarm threshold. This threshold determination method is more scientific and accurate than subjective threshold setting and better reflects the actual data. This application also sets alarm thresholds for each operating indicator, fully considering the characteristics and differences of different indicators. Different operating indicators may have different data distributions and variation ranges. Determining personalized alarm thresholds for each indicator can improve the relevance and effectiveness of the thresholds. Subsequently, this application sequentially determines whether the real-time operating data exceeds the corresponding alarm threshold. If data failure is detected, a data failure signal is immediately issued, thereby promptly identifying abnormalities in equipment operation.

[0019] Optionally, after performing the step of determining the threshold and before performing the step of judging the threshold, the method further includes:

[0020] Calculate the average value: calculate the average value of historical operation data under each operation indicator;

[0021] Data analysis: Using a PDP-like method, the historical operating data under the i-th operating indicator and the average values ​​corresponding to the remaining operating indicators are input into the neural network model in turn to obtain the change in the predicted result of the i-th operating indicator. The calculation model is as follows:

[0022] ;

[0023] in, is the output probability when the jth historical operation data in the i-th operation indicator is input; is the output probability of the kth remaining operating indicator when the jth historical operating data in the i-th operating indicator is input; n is the number of historical operating data in the i-th operating indicator; m is the number of operating indicators;

[0024] Obtain change trends: Obtain all changes and set the priority of the operating indicators in descending order according to the changes;

[0025] In the threshold judgment step, the real-time operation data is compared with the corresponding alarm threshold according to the priority.

[0026] This application calculates the average value of historical operating data for each operating indicator. Then, using a PDP-like (partial dependence plot) approach, the historical operating data for the i-th operating indicator and the average values ​​corresponding to the remaining operating indicators are sequentially input into a neural network model. This method obtains the change in the predicted result for the i-th operating indicator, quantifying the independent impact of each operating indicator on the prediction result and helping to identify which operating indicators are most critical to the corrosion rate or remaining life prediction of the equipment. By calculating the change, it is possible to more intuitively understand how changes in each operating indicator lead to changes in the prediction result. Subsequently, this application obtains all the changes and prioritizes the operating indicators based on their change from largest to smallest, enabling more targeted threshold determination, prioritizing those with the greatest impact on the prediction result. Indicators with large changes and significant impact on the prediction result are given higher priority for threshold determination, enabling more effective identification of potential issues and improving the efficiency and accuracy of equipment monitoring. During the threshold determination step, real-time operating data is compared against the corresponding alarm thresholds based on priority. This prioritized determination approach allows for faster detection of abnormalities in key operating indicators. Once it is discovered that the data of high-priority indicators exceeds the alarm threshold, timely measures can be taken to improve the response speed of equipment maintenance.

[0027] Optionally, the method further includes:

[0028] Indicator screening: Perform Pareto analysis on the change in each operating indicator to obtain key operating indicators, and update the key operating indicators to operating indicators.

[0029] This application uses Pareto analysis to screen the changes in each operating indicator, thereby identifying the key operating indicators that have the most significant impact on the prediction results. Among the many operating indicators, there are often only a few indicators that play a decisive role in the corrosion rate or remaining life prediction of the equipment. Pareto analysis can quickly find these key indicators, thereby improving the efficiency of data analysis. This application updates the screened key operating indicators as operating indicators for subsequent analysis, reducing the data dimensions that need to be processed. In subsequent steps such as obtaining change trends and making threshold judgments, you only need to focus on key indicators, which reduces the complexity of the calculation and makes the entire process more concise and efficient.

[0030] Optionally, after executing the step of predicting the corrosion rate and before executing the step of calculating the remaining life, the method further includes:

[0031] Environmental acquisition: obtain solvent environment data and corrosion current data with the same timestamp as historical operation data; obtain solvent environment data at the current moment;

[0032] Modeling: Establish a random forest model and train it using solvent environment data and corrosion current data to obtain a trained random forest model;

[0033] Predicting corrosion current: Input the current solvent environment data into the trained random forest model and output the predicted corrosion current data;

[0034] Calculate corrosion rate: Calculate the theoretical corrosion rate using Faraday's law, and calculate the difference between the theoretical corrosion rate and the predicted corrosion rate;

[0035] Corrosion rate judgment: judge whether the predicted corrosion rate difference is greater than the preset corrosion rate threshold. If so, an alarm signal is issued; if not, the step of calculating the remaining life is executed.

[0036] This application obtains solvent environment data and corrosion current data with the same timestamp as historical operating data and uses this data in the modeling process, enabling a more comprehensive consideration of factors influencing the corrosion process. By incorporating solvent environment data, the random forest model can learn the variation patterns of corrosion current under different environmental conditions, thereby better adapting to the complex and changing environment of actual operation. Subsequently, this application obtains the solvent environment data at the current moment and inputs it into the trained random forest model to predict the corrosion current data in real time. The theoretical corrosion rate is then calculated using Faraday's law, and the corrosion rate difference between the theoretical and predicted corrosion rates is calculated. This method combines theoretical calculations with actual predictions to more accurately reflect the actual corrosion situation of the equipment. The theoretical corrosion rate provides a benchmark for prediction, while the corrosion rate difference reflects the deviation between the actual corrosion situation and the theoretical situation. The predicted corrosion rate difference is then determined to be greater than a preset corrosion rate threshold. When the corrosion rate difference exceeds the preset corrosion rate threshold, an alarm signal is promptly issued to alert management personnel that the equipment may be abnormal and require further inspection and maintenance. Only when the corrosion rate difference does not exceed the threshold value, the present application will perform the step of calculating the remaining life to accurately assess the remaining service life of the carbon dioxide capture equipment under normal corrosion conditions.

[0037] Optionally, the random forest model includes p decision trees, and after executing the step of acquiring the environment and before executing the step of modeling, further includes:

[0038] Constructing a data set: Integrate all solvent environment data into a solvent environment data set, construct p-1 sub-data sets, and randomly take a solvent environment data into any sub-data set;

[0039] Iteration: put the selected solvent environment data back into the solvent environment data set and re-execute the steps of constructing the data set until the number of samples in all sub-data sets is greater than the number of samples in the solvent environment data set, and update all sub-data sets to the solvent environment data set;

[0040] In the modeling step, the p-th solvent environment dataset is used to train the p-th decision tree in the random forest model.

[0041] When constructing a data set, this application integrates all solvent environment data into one solvent environment data set, then constructs p-1 sub-data sets and randomly places the solvent environment data into the sub-data sets. This application increases the diversity of the data through random assignment, so that each sub-data set has a different sample combination. By increasing data diversity, each decision tree in the random forest model can be exposed to a wider range of data samples during training, reducing the risk of a single decision tree overfitting to specific data. Different decision trees are trained based on different sub-data sets and learn different data characteristics and patterns, thereby improving the generalization ability of the entire random forest model, enabling it to make more accurate predictions when faced with new solvent environment data. During the iterative process, this application puts the selected solvent environment data back into the solvent environment data set and re-executes the steps of constructing the data set until the number of samples in all sub-data sets is greater than the number of samples in the solvent environment data set. This method gives each solvent environment data the opportunity to be assigned to different sub-data sets multiple times, thereby being used by multiple decision trees. This application fully mines the information in the data, improves data utilization, and helps the model learn more comprehensive data features. Since each decision tree is trained based on a different sub-data set, the random forest model is less dependent on the data. Even if a sub-dataset contains some outliers or noise, it will not have a significant impact on the entire random forest model. This stability allows the random forest model to maintain good performance under different data distributions and working conditions.

[0042] Optionally, the step of constructing the data set further includes: dividing the solvent environment data set into q layers, randomly generating a random number x from 1 to q, and randomly picking a solvent environment data in the x layer and putting it into any sub-data set.

[0043] This application divides the solvent environment dataset into q layers, then extracts data from each of these q layers and places it into sub-datasets. This ensures that data from different layers have the opportunity to be extracted into the sub-datasets, allowing the sub-datasets to more comprehensively cover the various features of the solvent environment dataset. Subsequently, this application randomly selects a solvent environment data point from the x layer and places it into the sub-dataset. This increases the randomness of the data extraction, helps break down any correlations that may exist in the original dataset, and makes the data in the sub-datasets more independent. When training on these independent sub-datasets, decision trees can learn a wider range of data features and patterns, reducing their reliance on specific data patterns. This improves the model's generalization ability and enables it to make more accurate predictions when faced with new, unseen solvent environment data. Randomly extracting data prevents the model from overfitting to specific data samples during training. Since each sub-dataset is randomly extracted, the model needs to adapt to different data combinations rather than just memorizing specific samples, which helps improve the model's robustness. Different decision trees can learn different data features based on different sub-datasets, thus providing richer and more accurate predictions when the model is integrated.

[0044] Optionally, the method further includes:

[0045] Life span judgment: judging whether the remaining life span is less than a preset life span threshold, if so, executing the early warning step; if not, re-executing the real-time data collection step;

[0046] Early warning: Send out early warning signals.

[0047] By determining whether the remaining life is less than a preset life threshold, this application can promptly issue an early warning signal when the remaining life of the equipment is insufficient, allowing equipment managers to understand the potential risks of the equipment in advance and take appropriate measures. The life judgment and early warning steps, together with real-time data collection and remaining life assessment steps, form a closed-loop management system. By continuously collecting data, assessing remaining life, issuing early warnings, and addressing problems, the equipment management process can be continuously optimized. Based on early warning information and equipment operation status, managers can summarize experiences and lessons learned, adjust management strategies and methods, and improve the scientific nature and effectiveness of equipment management.

[0048] In a second aspect, the present application provides a carbon dioxide capture equipment corrosion monitoring system, which adopts the following technical solutions:

[0049] A carbon dioxide capture equipment corrosion monitoring system comprising:

[0050] Monitoring module, including real-time data acquisition unit and historical data acquisition unit;

[0051] A real-time data acquisition unit, used to obtain the real-time operating data, operating time and corrosion degree of the current carbon dioxide capture equipment;

[0052] a historical data acquisition unit for obtaining the final corrosion degree and training data of the same type of carbon dioxide capture equipment, wherein the training data includes: historical operating data of the same type of carbon dioxide capture equipment over its entire life cycle, and the corrosion rate of the same type of carbon dioxide capture equipment under each historical operating data;

[0053] Edge computing module, including modeling and training unit, corrosion rate prediction unit and remaining life calculation unit;

[0054] A modeling and training unit is used to establish a neural network model using a machine learning algorithm, train the neural network model using training data, and obtain a trained neural network model;

[0055] The corrosion rate prediction unit is used to input real-time operating data into the trained neural network model and output the predicted corrosion rate;

[0056] The remaining life calculation unit is used to calculate the remaining life based on the final corrosion level, the corrosion level before operation, the operation time and the predicted corrosion rate.

[0057] Optionally, the system further comprises:

[0058] a memory storing a computer-readable storage medium and training data;

[0059] A processor is configured to execute the steps of the method when the processor calls the program code in the memory.

[0060] In summary, this application includes at least one of the following beneficial technical effects:

[0061] 1. The present application first obtains the real-time operating data, operating time and corrosion level before operation of the current carbon dioxide capture equipment, wherein the real-time operating data reflects the working status of the current carbon dioxide capture equipment, and the corrosion level before operation is combined with the operating time to calculate the current corrosion level of the equipment. Subsequently, the present application obtains the final corrosion level and training data of the same type of carbon dioxide capture equipment, and the final corrosion level of the same type of carbon dioxide capture equipment is used as the end point of life, which is used to calculate the remaining life in combination with the current corrosion level. Subsequently, the present application uses a machine learning algorithm to establish a neural network model, and uses the training data to train the neural network model. During the training process, the neural network model can automatically learn the potential laws and features in the data, thereby improving the prediction performance. Subsequently, the present application inputs the real-time operating data into the trained neural network model, and the trained neural network model will output the predicted corrosion rate. The present application can calculate the current corrosion state through the operating time and the corrosion level before operation, and is suitable for long-term carbon dioxide capture equipment that requires accurate calculation of the remaining life.

[0062] 2. This application uses the K-means clustering algorithm to cluster historical operating data for each operating indicator (K is 2). Clustering aims to separate normal operating data and abnormal operating data into separate clusters. This approach simplifies complex data sets into two representative clusters, helping to discover underlying patterns and distribution characteristics within the data. Subsequently, this application uses the Otsu's method to maximize the inter-cluster variance between the two clusters to determine the alarm threshold. The Otsu method automatically finds an optimal split point in the data that maximizes the difference between the two clusters, thereby determining a reasonable alarm threshold. This threshold determination method is more scientific and accurate than subjective threshold setting and better reflects the actual data. This application also sets alarm thresholds for each operating indicator, fully considering the characteristics and differences of different indicators. Different operating indicators may have different data distributions and ranges of variation. Determining personalized alarm thresholds for each indicator can improve the relevance and effectiveness of the thresholds. Subsequently, this application sequentially determines whether the real-time operating data exceeds the corresponding alarm threshold. If data failure is detected, a data failure signal is immediately issued, thereby promptly identifying abnormalities in equipment operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 is a flow chart of Example 1 of the present application;

[0064] Figure 2 This is a flow chart of Example 2 of the present application;

[0065] Figure 3 This is a flowchart of Example 3 of the present application. DETAILED DESCRIPTION

[0066] The following combination Figures 1 to 3 This application is described in further detail.

[0067] The carbon dioxide capture equipment mentioned in this application includes an absorption tower, a desorption tower, a heat exchanger, a pump, a compressor, a separator and a control system.

[0068] Example 1: This example discloses a method for monitoring corrosion of carbon dioxide capture equipment. Figure 1 The method includes: S11 real-time data acquisition, S12 historical data acquisition, S13 modeling and training, S14 corrosion rate prediction, and S15 remaining life calculation. In this embodiment, the real-time operation data, operation time, and pre-operation corrosion level of the current carbon dioxide capture equipment are first obtained; the final corrosion level of the same type of equipment and training data containing historical operation data of the entire life cycle and the corresponding corrosion rate are obtained; then a neural network model is established using a machine learning algorithm, and the model is trained using the training data; then the real-time operation data is input into the trained model, and the predicted corrosion rate is output; finally, the remaining life of the equipment is calculated based on the final corrosion level, pre-operation corrosion level, operation time, and predicted corrosion rate. The process of this embodiment is as follows:

[0069] S11 real-time data acquisition, obtains the real-time operation data, operation time and corrosion degree before operation of the current carbon dioxide capture equipment.

[0070] Operational data reflects the current operating status of the CO2 capture equipment, covering the following aspects:

[0071] Process parameters: including temperature, pressure, flow rate, temperature gradient ( ), O2 content;

[0072] Equipment parameters: including equipment material parameters.

[0073] The duration from the start of the current operation of the equipment to the current moment (i.e., the operating time) is recorded. The ultrasonic detection method is used to measure the corrosion level of the current carbon dioxide capture equipment before the current operation. The corrosion level refers to the corrosion wall thickness, i.e., the thickness of the carbon steel passivation film corrosion.

[0074] S12 historical data collection, obtaining the final corrosion degree and training data of the same type of carbon dioxide capture equipment, wherein the training data includes: historical operating data of the same type of carbon dioxide capture equipment throughout its life cycle, and the corrosion rate of the same type of carbon dioxide capture equipment under each historical operating data.

[0075] Collect data on the corrosion levels achieved by similar CO2 capture equipment at the end of its lifecycle. This data can come from decommissioned or scrapped CO2 capture equipment. Ultrasonic testing will be used to measure the corrosion levels at various locations. The average value will be calculated as the final corrosion level benchmark.

[0076] Training data is key to building accurate models and consists of two main parts:

[0077] Historical operating data for the entire life cycle of the same type of CO2 capture equipment. This data records the operating parameters of the same type of CO2 capture equipment at each stage of its life cycle, from commissioning to decommissioning. Similar to real-time operating data, historical operating data also includes the following parameters:

[0078] Process parameters: including temperature, pressure, flow rate, temperature gradient ( ), O2 content;

[0079] Equipment parameters: including equipment material parameters.

[0080] But the difference is that it covers the changes throughout the entire life cycle of the equipment and can reflect the characteristics and laws of carbon dioxide capture equipment at different operating stages.

[0081] The corrosion rate of the same type of CO2 capture equipment for each historical operating data point. The corrosion rate is the rate of change of the equipment's corrosion level over time. By combining historical operating data with the corresponding corrosion rate, a correlation between the equipment's operating status and corrosion status can be established.

[0082] S13 modeling and training, using a machine learning algorithm to establish a neural network model, the initial parameters of the neural network model are shown in Table 1.

[0083] Table 1 Initial parameters of the neural network model

[0084] Network Depth Convolution kernel Activation Function Number of iterations Max Pooling 2 64，5×5 ReLU 128 2×2

[0085] The collected training data is input into the neural network model, and the internal parameters of the model are continuously adjusted to make the output results of the model as close as possible to the actual corrosion rate.

[0086] The training process uses a stochastic gradient descent algorithm to minimize the model's prediction error. The training data is divided into multiple batches, with a portion of the data used for training at a time. The model parameters are updated through multiple iterations until satisfactory performance is achieved. Upon completion, a trained neural network model is obtained, which has learned the underlying relationship between the equipment's operating status and corrosion rate.

[0087] S14 predicts corrosion rates by feeding real-time operating data into a trained neural network model. The trained neural network calculates and analyzes the input real-time data based on the previously learned relationship between the CO2 capture equipment's operating status and corrosion rate, ultimately outputting a predicted corrosion rate.

[0088] S15 calculates the remaining lifespan. Based on the pre-operation and final corrosion levels, the maximum allowable corrosion level for the CO2 capture equipment over its entire lifespan can be determined. Combining the operating time with the predicted corrosion rate, the remaining lifespan is calculated based on the current corrosion rate of the CO2 capture equipment, indicating how long it will take to reach the final corrosion level. This time is the remaining lifespan of the equipment.

[0089] This embodiment first performs real-time data acquisition to obtain the real-time operating data, operating time, and pre-operation corrosion level of the current carbon dioxide capture equipment. Then, historical data acquisition is carried out to collect the final corrosion levels of similar equipment and training data containing historical operating data and corresponding corrosion rates throughout the entire life cycle. Subsequently, a neural network model is established using a machine learning algorithm, and the model is trained using the training data. The real-time operating data is then input into the trained model, and a predicted corrosion rate is output. Finally, based on the final corrosion level, pre-operation corrosion level, operating time, and predicted corrosion rate, the remaining life of the equipment is calculated. This allows for prediction of the corrosion condition of the carbon dioxide capture equipment and assessment of its remaining life, providing decision support for equipment operation management.

[0090] In other embodiments, the method further comprises:

[0091] Life span judgment, judging whether the remaining life span is less than a preset life span threshold, if so, executing the early warning step; if not, re-executing S11 real-time data collection.

[0092] Early warning, issue an early warning signal.

[0093] Example 2: Reference Figure 2 The difference between this embodiment and embodiment 1 is that after executing S12 historical data collection and before executing S13 modeling and training, it also includes:

[0094] S20 indicator screening: Perform a Pareto analysis on the change in each operating indicator. Pareto analysis is an analytical method based on the "80 / 20 principle," which states that a small number of key factors often have a major impact on the results. In this step, the change in all operating indicators is ranked to identify the key operating indicators that have the greatest impact on the forecast results.

[0095] Through Pareto analysis, we can determine which operating indicators play a key role in the model's prediction results, and use these key operating indicators as operating indicators for subsequent analysis.

[0096] S21 Clustering: Use the K-means clustering algorithm to cluster the historical operating data for the i-th operating indicator. K-means clustering is a commonly used unsupervised learning algorithm that aims to partition a dataset into K distinct clusters, ensuring high similarity between data points within a cluster and low similarity between data points in different clusters. In this step, the algorithm is applied to the historical operating data for the i-th operating indicator, with K set to 2, effectively dividing the historical operating data for the i-th operating indicator into two clusters.

[0097] S22 determines a threshold by maximizing the inter-class variance of the two clusters using the Grand Law. The inter-class variance reflects the degree of separation between the two clusters; a larger inter-class variance indicates a more pronounced difference between the two clusters. When the Grand Law finds the historical operating data that maximizes the inter-class variance, that historical operating data value is determined as the alarm threshold for the i-th operating indicator.

[0098] S23 calculates an average value. For each operating indicator, such as flow rate and vibration frequency of the equipment, all historical operating data are added up and then divided by the amount of data to obtain the average value of the operating indicator.

[0099] S24 data analysis uses a PDP-like method. For the i-th operating indicator, each historical operating data contained in the i-th operating indicator is combined with the average value of the remaining operating indicators to form multiple input samples. For example, if there are three operating indicators: temperature, pressure, and flow, when analyzing the temperature indicator, each temperature historical data is combined with the average value of pressure and flow, and then input into the neural network model to obtain the change in the predicted result of the i-th operating indicator. The change in the predicted result of the i-th operating indicator is The calculation model is as follows:

[0100] ;

[0101] in, is the output probability when the jth historical operation data in the i-th operation indicator is input; is the output probability of the kth remaining operating indicator when the jth historical operating data in the i-th operating indicator is input; n is the number of historical operating data in the i-th operating indicator; m is the number of operating indicators.

[0102] The PDP-like method (Partial Dependence Plot) is a method used to explain machine learning models. It can show the impact of a single feature on the model's prediction results. In this step, the PDP-like method is applied to the neural network model.

[0103] S25 obtains the change trend and the change amount of all key operating indicators, and sets the priority of the operating indicators in descending order of change amount. The larger the change amount, the more significant the impact of the operating indicator on the model prediction result, and thus it is given a higher priority.

[0104] For example, if the changes in the two operating indicators of temperature and pressure are 0.3 and 0.1 respectively, then the priority of temperature is higher than that of pressure, and the temperature indicator will be given priority in subsequent analysis and judgment.

[0105] S26 threshold judgment, according to the priority of the operating indicators set by the change trend obtained in S25, check in turn whether the data of each key operating indicator at the current moment exceeds its corresponding alarm threshold. For example, if the temperature has the highest priority, first determine whether the temperature data at the current moment is greater than the alarm threshold of the temperature indicator. If the data of a key operating indicator at the current moment exceeds its alarm threshold, it means that the operating indicator of the equipment has an abnormality. At this time, a data unqualified signal is issued to remind relevant personnel to pay attention to the operating status of the equipment. If the data of all key operating indicators at the current moment do not exceed the corresponding alarm threshold, it means that the operating status of the equipment is normal, and S13 modeling and training can be continued.

[0106] This embodiment first uses a K-means clustering algorithm (K is 2) to cluster the historical operating data under the i-th operating indicator to obtain clusters. Then, the inter-class variance of the two clusters is maximized through the large law. The historical operating data corresponding to the maximum inter-class variance is determined as the alarm threshold of the i-th operating indicator. At the same time, the average value of the historical operating data under each operating indicator is calculated respectively. Then, a quasi-PDP method is used to input the historical operating data under the i-th operating indicator and the average values ​​corresponding to the remaining operating indicators into the neural network model in sequence to obtain the change in the prediction result of the i-th operating indicator. Then, a Pareto analysis is performed on the change in each operating indicator to screen out key operating indicators and update the operating indicator set. Then, the priority of the operating indicators is set from large to small according to the change. Finally, each real-time operating data is determined to be greater than the corresponding alarm threshold according to the priority. If it is greater than the corresponding alarm threshold, a data failure signal is issued. If not, the modeling and training steps are performed.

[0107] Example 3: Reference Figure 3 The difference between this embodiment and embodiment 1 is that after executing S14 to predict the corrosion rate and before executing S15 to calculate the remaining life, the following steps are further included:

[0108] S31 environment acquisition, obtains solvent environment data and corrosion current data with the same timestamp as historical operation data. The historical operation data records the operating status of the equipment at different times in the past, and the solvent environment data and corrosion current data with the same timestamp can reflect the solvent environment conditions and corresponding corrosion current conditions in the equipment under this operating state.

[0109] Obtain the current solvent environment data, which reflects the actual environmental conditions of the solvent. Real-time monitoring of the CO2 capture equipment's solvent environment data, such as total amine concentration, carbamate concentration, carbonate concentration, amine solution pH, heat-stable salt concentration (HSS concentration), and CO2 loading rate (α value), will be used as input into the trained model to predict the current corrosion current.

[0110] S32 constructs a dataset by integrating all solvent environment data into a single dataset. From this dataset, randomly extracting solvent environment data to construct p - 1 sub-datasets. During the extraction process, each solvent environment data point has a chance of being included in any sub-dataset. This extraction method improves the diversity and randomness of the sub-datasets.

[0111] To further increase the randomness and representativeness of the sub-datasets, the solvent environment dataset is divided into q strata. A random number x is then generated between 1 and q, and a solvent environment data point is randomly selected from the corresponding x-stratum. This data point is then placed into any sub-dataset. The selected solvent environment data point is then placed back into the solvent environment dataset. This stratified random sampling approach ensures that the sub-datasets better capture the diverse characteristics of the solvent environment dataset.

[0112] For example, the solvent environment data set is divided into 5 layers according to a certain feature (such as total amine concentration), and the randomly generated number x is 3. Then a solvent environment data is randomly selected from the 3rd layer and put into the sub-data set.

[0113] In S33, after each sub-dataset is constructed, the previously selected solvent environment data is placed back into the solvent environment dataset, restoring the solvent environment dataset to its original state. S32 is then executed again to construct the dataset, performing random sampling and allocation again to generate a new sub-dataset. This iterative process is repeated until the number of samples in each sub-dataset exceeds the number of samples in the solvent environment dataset. Finally, all generated sub-datasets are sequentially updated to become the solvent environment dataset.

[0114] S34 modeling, build a random forest model. Random forest is an ensemble learning algorithm that consists of multiple decision trees. Each decision tree performs independent classification or regression prediction on the data. Ultimately, the accuracy and stability of the model are improved by combining the results of multiple decision trees.

[0115] The pth decision tree in the random forest model is trained using the constructed pth solvent environment dataset. During training, the decision tree learns the mapping relationship between data based on the features and labels (if any) in the solvent environment dataset, enabling predictions for new solvent environment data. For example, the features in the solvent environment dataset include total amine concentration, carbamate concentration, carbonate concentration, amine solution pH, heat-stable salt concentration (HSS concentration), CO2 loading rate (α value), etc., and the labels are the corresponding corrosion current data. The decision tree learns the relationship between these features and labels to construct the corresponding decision rules.

[0116] After training all decision trees, this embodiment can obtain a trained random forest model. The trained random forest model has learned the complex relationship between solvent environment data and corrosion current, and can be used to predict corrosion current for new solvent environment data.

[0117] S35 predicts the corrosion current by inputting the current solvent environment data into the trained random forest model. The random forest model analyzes and processes the current solvent environment data based on the previously learned relationship between solvent environment data and corrosion current. Based on the input solvent environment data, the random forest model uses a comprehensive judgment of multiple decision trees to output predicted corrosion current data.

[0118] S36 calculates the corrosion rate and calculates the theoretical corrosion rate using Faraday's law. The calculation model is as follows:

[0119] ;

[0120] in, is the theoretical corrosion rate; I is the predicted corrosion current data; E is the chemical equivalent of the metal; N is the valence of the metal type used in the CO2 capture device; F is the Faraday constant, which is 96485; R is the surface area of ​​the metal, which is obtained through the design parameters of the CO2 capture device or actual measurement.

[0121] The corrosion rate difference is calculated based on the theoretical corrosion rate and the predicted corrosion rate.

[0122] S37 corrosion rate judgment, judge whether the predicted corrosion rate difference is greater than the preset corrosion rate threshold, if so, issue an alarm signal; if not, execute S15 to calculate the remaining life.

[0123] This embodiment first obtains solvent environment data and corrosion current data with the same timestamp as historical operation data, as well as solvent environment data at the current moment. Then, all solvent environment data are integrated into a data set and p-1 sub-data sets are constructed. Stratified random sampling is continuously iterated until the number of samples in the sub-data set is greater than the original solvent environment data set, and then the sub-data set is updated to the solvent environment data set. Then, a random forest model containing p decision trees is established, and the p-th decision tree is trained with the p-th solvent environment data set to obtain a trained model. The current solvent environment data is then input into the model to output predicted corrosion current data. The theoretical corrosion rate is calculated using Faraday's law and the difference with the predicted value is obtained. Finally, it is determined whether the corrosion rate difference is greater than a preset threshold. If it is greater, an alarm signal is issued; if it is less than or equal to, the remaining life calculation step is executed.

[0124] Example 4: This example discloses a carbon dioxide capture equipment corrosion monitoring system, the system comprising:

[0125] The monitoring module includes a real-time data acquisition unit and a historical data acquisition unit.

[0126] The real-time data acquisition unit is used to collect data in real time through a sensor network (such as temperature sensors, pressure sensors, corrosion sensors, etc.) to obtain the real-time operating data, operating time and corrosion degree before operation of the current carbon dioxide capture equipment.

[0127] The historical data acquisition unit is used to obtain the final corrosion degree and training data of the same type of carbon dioxide capture equipment. The training data includes: historical operation data of the same type of carbon dioxide capture equipment throughout its life cycle, and the corrosion rate of the same type of carbon dioxide capture equipment under each historical operation data.

[0128] Edge computing module, including modeling and training unit, corrosion rate prediction unit and remaining life calculation unit;

[0129] The modeling and training unit is used to establish a corrosion prediction model using machine learning algorithms (such as neural networks, random forests, support vector machines, etc.), and use training data to train the neural network model to obtain the trained neural network model for subsequent corrosion rate prediction.

[0130] The corrosion rate prediction unit is used to input real-time operation data into the trained neural network model and output the predicted corrosion rate.

[0131] The remaining life calculation unit is used to calculate the remaining life based on the final corrosion level, the corrosion level before operation, the operation time and the predicted corrosion rate.

[0132] The system further includes a memory storing a computer-readable storage medium and training data.

[0133] A processor is configured to execute the steps of the method when the processor calls the program code in the memory.

[0134] The above are all preferred embodiments of the present application, and are not intended to limit the scope of protection of the present application. Therefore, any equivalent changes made based on the structure, shape, and principle of the present application should be included in the scope of protection of the present application.

Claims

1. A method for monitoring corrosion of carbon dioxide capture equipment, characterized in that: include: Real-time data collection: Obtain real-time operating data of the current CO2 capture equipment, operating time, and corrosion level before operation; Historical data collection: Obtain the final corrosion degree and training data of the same type of CO2 capture equipment. The training data includes: historical operating data of the same type of CO2 capture equipment throughout its life cycle, and the corrosion rate of the same type of CO2 capture equipment under each historical operating data; Modeling and training: Use machine learning algorithms to build a neural network model, use training data to train the neural network model, and obtain the trained neural network model; Predicting corrosion rate: Inputting real-time operating data into the trained neural network model, the output is the predicted corrosion rate; Calculate the remaining life: Calculate the remaining life based on the final corrosion level, corrosion level before operation, operation time and predicted corrosion rate; The historical operation data includes m types of operation indicators. After executing the step of collecting historical data and before executing the step of modeling and training, the following steps are also included: Clustering: K-means clustering algorithm is used to cluster the historical operation data under the i-th operation indicator to obtain clusters, where the value of K is 2; Determine the threshold: Maximize the inter-class variance of the two clusters through the great law, and record the historical operation data corresponding to the maximum inter-class variance as the alarm threshold of the i-th operation indicator; Threshold judgment: judge whether each real-time running data is greater than the corresponding alarm threshold in turn. If so, a data unqualified signal is issued; if not, the modeling and training steps are executed; After executing the step of determining the threshold and before executing the step of judging the threshold, the method further includes: Calculate the average value: calculate the average value of historical operation data under each operation indicator; Data analysis: Using a PDP-like method, the historical operating data under the i-th operating indicator and the average values ​​corresponding to the remaining operating indicators are input into the neural network model in turn to obtain the change in the predicted result of the i-th operating indicator. The calculation model is as follows: ； in, is the output probability when the jth historical operation data in the i-th operation indicator is input; is the output probability of the kth remaining operating indicator when the jth historical operating data in the i-th operating indicator is input; n is the number of historical operating data in the i-th operating indicator; m is the number of operating indicators; Obtain change trends: Obtain all changes and set the priority of the operating indicators in descending order according to the changes; In the threshold judgment step, the real-time operation data is compared with the corresponding alarm threshold according to the priority.

2. The method for monitoring corrosion of carbon dioxide capture equipment according to claim 1, characterized in that: The method further comprises: Indicator screening: Perform Pareto analysis on the change in each operating indicator to obtain key operating indicators, and update the key operating indicators to operating indicators.

3. The method for monitoring corrosion of carbon dioxide capture equipment according to claim 1 or 2, characterized in that: After executing the step of predicting the corrosion rate and before executing the step of calculating the remaining life, the method further includes: Environmental acquisition: obtain solvent environment data and corrosion current data with the same timestamp as historical operation data; obtain solvent environment data at the current moment; Modeling: Establish a random forest model and train it using solvent environment data and corrosion current data to obtain a trained random forest model; Predicting corrosion current: Input the current solvent environment data into the trained random forest model and output the predicted corrosion current data; Calculate corrosion rate: Calculate the theoretical corrosion rate using Faraday's law, and calculate the difference between the theoretical corrosion rate and the predicted corrosion rate; Corrosion rate judgment: judge whether the predicted corrosion rate difference is greater than the preset corrosion rate threshold. If so, issue a warning signal; if not, execute the step of calculating the remaining life.

4. The method for monitoring corrosion of carbon dioxide capture equipment according to claim 3, wherein: The random forest model includes p decision trees, and after executing the step of acquiring the environment and before executing the step of modeling, further includes: Constructing a data set: Integrate all solvent environment data into a solvent environment data set, construct p-1 sub-data sets, and randomly take a solvent environment data into any sub-data set; Iteration: put the selected solvent environment data back into the solvent environment data set and re-execute the steps of constructing the data set until the number of samples in all sub-data sets is greater than the number of samples in the solvent environment data set, and update all sub-data sets to the solvent environment data set; In the modeling step, the p-th solvent environment dataset is used to train the p-th decision tree in the random forest model.

5. The method for monitoring corrosion of carbon dioxide capture equipment according to claim 4, characterized in that: The step of constructing the data set also includes: dividing the solvent environment data set into q layers, randomly generating a random number x from 1 to q, and randomly picking a solvent environment data in the x layer and putting it into any sub-data set.

6. The method for monitoring corrosion of carbon dioxide capture equipment according to claim 1 or 2, characterized in that: The method further comprises: Life span judgment: judging whether the remaining life span is less than a preset life span threshold, if so, executing the early warning step; if not, re-executing the real-time data collection step; Early warning: Send out early warning signals.

7. A carbon dioxide capture equipment corrosion monitoring system, characterized in that: The system is used to perform the method according to any one of claims 1 to 6, and the system includes: Monitoring module, including real-time data acquisition unit and historical data acquisition unit; A real-time data acquisition unit, used to obtain the real-time operating data, operating time and corrosion degree of the current carbon dioxide capture equipment; a historical data acquisition unit for obtaining the final corrosion degree and training data of the same type of carbon dioxide capture equipment, wherein the training data includes: historical operating data of the same type of carbon dioxide capture equipment over its entire life cycle, and the corrosion rate of the same type of carbon dioxide capture equipment under each historical operating data; Edge computing module, including modeling and training unit, corrosion rate prediction unit and remaining life calculation unit; A modeling and training unit is used to establish a neural network model using a machine learning algorithm, train the neural network model using training data, and obtain a trained neural network model; A corrosion rate prediction unit is used to input real-time operating data into the trained neural network model and output the predicted corrosion rate; The remaining life calculation unit is used to calculate the remaining life based on the final corrosion level, the corrosion level before operation, the operation time and the predicted corrosion rate.

8. A corrosion monitoring device for carbon dioxide capture equipment, characterized in that: include: a memory, wherein program code is stored in the memory; A processor, wherein the processor executes the steps of the method according to any one of claims 1 to 6 when calling the program code in the memory.

Images