Intelligent control method and system for food-grade carbon dioxide purification process

Abstract

This invention discloses an intelligent control method and system for food-grade carbon dioxide purification processes, relating to the field of purification control technology. The method includes: acquiring M process nodes of the purification production line, extracting historical anomaly characteristics to identify node anomalies, configuring a reliable window for each process node based on the anomaly level, continuously monitoring each node, analyzing operational risks, generating operational monitoring data and risk coefficient sequences for each node, and combining the risk coefficient of each node with the corresponding monitoring data set and a reliable time period identifier to perform risk control on the M process nodes. This invention solves the technical problem of existing carbon dioxide purification processes lacking real-time monitoring and adjustment mechanisms, leading to the inability to identify and handle risks in a timely manner, thus affecting the stability and safety of the purification process. It achieves the technical effect of timely identification and handling of risks through real-time monitoring and intelligent control mechanisms, thereby improving the stability and safety of the purification process.

Description

Technical Field

[0001] This invention relates to the field of purification control technology, specifically to an intelligent control method and system for food-grade carbon dioxide purification processes. Background Technology

[0002] In carbon dioxide purification, the operational status of each process node directly affects the purification effect and product quality. Traditional purification processes typically rely on manual experience and simple monitoring systems, making it difficult to achieve precise control and real-time monitoring of each process node. Existing monitoring systems often focus on monitoring parameters of a single node, lacking a comprehensive analysis of the interrelationships between multiple process nodes and failing to effectively identify and intelligently control real-time risks. This results in the failure to promptly detect potential anomalies during the process, thereby impacting purification efficiency and production safety. Summary of the Invention

[0003] This application provides an intelligent control method and system for food-grade carbon dioxide purification processes, which addresses the technical problem that the lack of real-time monitoring and adjustment mechanisms in existing carbon dioxide purification processes leads to the inability to identify and handle risks in a timely manner, thus affecting the stability and safety of the purification process.

[0004] The first aspect of this application provides an intelligent control method for a food-grade carbon dioxide purification process. The method includes: acquiring M process nodes of the purification production line; traversing the M process nodes to extract historical anomaly characteristics, obtaining a set of historical anomaly characteristics for the M process nodes; identifying the node anomaly degree of the set of historical anomaly characteristics for the M process nodes, obtaining M process node anomaly degrees; configuring M reliable windows for the M process nodes based on the M process node anomaly degrees; continuously monitoring the operation of the M process nodes and performing operational risk analysis using the M reliable windows as monitoring periods, obtaining a sequence of M process node window operation monitoring data sets and a sequence of M process node window risk coefficients, wherein each process node window risk coefficient and the corresponding process node window operation monitoring data set have a reliable time period identifier; and combining the reliable time period identifier, performing risk control on the M process nodes based on the sequence of M process node window risk coefficients and the sequence of M process node window operation monitoring data sets.

[0005] A second aspect of this application provides an intelligent control system for a food-grade carbon dioxide purification process. The system includes: an anomaly characterization extraction module, which acquires M process nodes of the purification line, traverses the M process nodes to extract historical anomaly characterizations, and obtains a set of historical anomaly characterizations for the M process nodes; a trusted window configuration module, which identifies the node anomaly degree of the historical anomaly characterizations for the M process nodes, obtains the anomaly degree of the M process nodes, and configures M trusted windows for the M process nodes based on the anomaly degrees of the M process nodes; an operational risk analysis module, which continuously monitors the operation of the M process nodes and performs operational risk analysis using the M trusted windows as monitoring periods, obtaining a sequence of operational monitoring data sets and a sequence of risk coefficients for the M process node windows, wherein each process node window risk coefficient and its corresponding operational monitoring data set has a trusted time period identifier; and a risk control module, which combines the trusted time period identifier with the sequence of risk coefficients and operational monitoring data sets for the M process nodes to perform risk control on the M process nodes.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0007] This application provides an intelligent control method and system for food-grade carbon dioxide purification processes, relating to the field of purification control technology. By acquiring M process nodes of the purification production line, extracting historical abnormal data and identifying the degree of abnormality, and configuring a reliable window for each node, the system continuously monitors the node's operating status based on the reliable window, analyzes risks, generates monitoring data and risk coefficient sequences, and combines these with reliable time period identifiers for control, ensuring the stability and safety of the purification process. This solves the technical problem of existing carbon dioxide purification processes lacking real-time monitoring and adjustment mechanisms, leading to the inability to identify and handle risks in a timely manner, thus affecting the stability and safety of the purification process. It achieves the technical effect of timely identification and handling of risks through real-time monitoring and intelligent control mechanisms, thereby improving the stability and safety of the purification process. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 A schematic diagram of the intelligent control method for food-grade carbon dioxide purification provided in the embodiments of this application;

[0010] Figure 2 This is a schematic diagram of the intelligent control system for food-grade carbon dioxide purification process provided in an embodiment of this application.

[0011] Figure labeling: 11 Anomaly characterization extraction module, 12 Trusted window configuration module, 13 Operational risk analysis module, 14 Risk control module. Detailed Implementation

[0012] This application provides an intelligent control method and system for food-grade carbon dioxide purification processes, which addresses the technical problem that the lack of real-time monitoring and adjustment mechanisms in existing carbon dioxide purification processes leads to the inability to identify and handle risks in a timely manner, thus affecting the stability and safety of the purification process.

[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0014] It should be noted that the terms "first," "second," etc., in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to such processes, methods, products, or devices.

[0015] Example 1, as Figure 1 As shown, this application provides an intelligent control method for a food-grade carbon dioxide purification process, the method comprising:

[0016] P10: Obtain M process nodes of the purification production line, traverse the M process nodes to extract historical anomaly representations, and obtain a set of historical anomaly representations for the M process nodes.

[0017] Specifically, the first step is to comprehensively review and analyze the entire purification production line, identifying all the process nodes within it. These process nodes are crucial links in the carbon dioxide purification process, encompassing multiple steps from raw gas pretreatment, carbon dioxide separation, impurity removal, purification, to final product testing. Each process node has independent functions and monitorable parameters, and its operating status and parameter settings have a vital impact on the quality of the final product and production efficiency. For example, the pretreatment stage may include nodes such as gas compression, cooling, and drying; the separation stage may involve nodes such as membrane separation or adsorption separation; and the purification stage may include nodes such as condensation and evaporation. The operating parameters and status data of these nodes are the foundation for intelligent control.

[0018] After acquiring all process nodes, it is necessary to extract historical anomaly representations for each node. Historical anomaly representations refer to the characteristic behaviors of each process node when abnormal situations occurred during past production processes, which may include equipment failures, parameter fluctuations exceeding normal ranges, and excessive impurity content. These anomaly features can be extracted by analyzing historical data to form a historical anomaly representation set. For example, firstly, historical operating data for each process node is collected from the production control system. This data includes equipment parameters, sensor data, and operation logs. Then, the collected data is cleaned and normalized to remove noisy and invalid data, ensuring data accuracy and consistency. Next, anomaly detection algorithms, such as statistical anomaly detection and machine learning-based anomaly detection, are used to identify anomalies in the data. Features are extracted from the identified anomalies to form a historical anomaly representation set for each process node. These features may include the time of the anomaly, the duration of the anomaly, and changes in key parameters during the anomaly.

[0019] By traversing M process nodes, historical anomaly representations are extracted from each node, forming a set of M historical anomaly representations for each process node. This set is a dataset containing detailed records of various anomalies that occurred at each process node in past production. This historical dataset forms the basis for subsequent analysis and risk prediction, providing a basis for node anomaly assessment. Through data mining and pattern recognition methods, representative anomaly features can be extracted, thus providing accurate data support for subsequent anomaly identification and risk control.

[0020] By following the steps above, a set of historical anomaly characteristics for each process node can be obtained. This set provides crucial data support for subsequent node anomaly identification and reliable window configuration, and is a key foundation for achieving intelligent control.

[0021] P20: Perform node anomaly degree identification on the historical anomaly representation set of the M process nodes to obtain the anomaly degree of the M process nodes, and configure the M reliable windows of the M process nodes according to the anomaly degree of the M process nodes.

[0022] Furthermore, step P20 in this embodiment of the application also includes:

[0023] P21: Cluster the historical anomaly representations of the M process nodes with the same type of anomaly representation within each process node to determine the M clusters of historical anomaly representations of the process nodes; P22: Traverse the M clusters of historical anomaly representations of the process nodes and perform weighted analysis to determine the anomaly degree of the M process nodes.

[0024] It should be understood that after extracting the historical anomaly representation set of the M process nodes of the purification production line, the next step is to conduct in-depth analysis of these anomaly representations to identify the degree of anomaly for each process node, i.e., the node anomaly degree. This process is crucial for subsequent confidence window configuration because the node anomaly degree reflects the frequency and severity of anomalies that occurred at each process node during historical operation.

[0025] Specifically, the first step is to perform cluster analysis on the historical anomaly representations for each process node. Clustering of similar anomaly representations within a process node means that for a specific process node, all anomalies (such as excessively high temperature or excessively low pressure) are categorized according to their severity. For example, if the anomaly representation for a process node is "excessively high pressure," it can be clustered according to different degrees of pressure exceeding the range (such as slightly exceeding the limit or severely exceeding the limit), thus forming multiple subsets. Each subset represents a specific type of anomaly. This process can utilize clustering algorithms, such as the K-means algorithm and DBSCAN, common data clustering methods. The goal of clustering is to group anomalies with similar properties into one category, ensuring that the identification of anomalies is targeted and accurate.

[0026] Each process node, after undergoing such clustering analysis, generates M clusters representing historical anomalies. These clusters contain representational data similar to the anomalies of that node, helping us to more accurately understand the node's anomaly patterns, their frequency, and severity. Different clusters can represent different types of anomalies exhibited by the process node throughout its history, such as the distinction between minor and severe anomalies.

[0027] Next, a weighted analysis is performed on the historical anomaly clusters of the M clustered process nodes to determine the anomaly degree of each process node. Specifically, different types of anomalies have varying degrees of impact on process nodes; for example, some anomalies may have a significant impact on product quality or equipment safety, while others have a smaller impact. Therefore, in the weighted analysis, a weight coefficient needs to be assigned to each cluster based on the severity and frequency of each type of anomaly. The weight allocation can be determined based on factors such as the severity of the anomaly, its frequency, and its impact on product quality and production efficiency. For example, clusters of anomalies that occur frequently and have a significant impact on product quality can be assigned higher weights, while clusters of anomalies that occur occasionally and have a smaller impact can be assigned lower weights. Through the weighted analysis of each cluster, the impact of different anomalies can be comprehensively considered, thereby obtaining the anomaly degree of each process node. This anomaly degree is a quantitative indicator that can intuitively reflect the anomalies of each process node in its historical operation. The higher the anomaly degree value, the more frequent or severe the anomalies of that process node, and the higher the risk. The anomaly level of each process node is dynamically adjusted, and is updated as the process node operates and the monitoring data changes.

[0028] Finally, based on the calculated anomalies of the M process nodes, a confidence window is configured for each process node. The confidence window refers to the allowable range of parameter fluctuations for that process node during normal operation. Under normal circumstances, the operating parameters of the process node should remain within this window range; behavior exceeding this range will be considered abnormal and trigger the risk control mechanism. The configuration of the confidence window not only depends on the anomaly rate but also considers the characteristics and historical performance of the process node, thereby ensuring its safety and efficiency in actual production. In this way, suitable monitoring and control strategies can be tailored for each process node, thereby improving the stability of the entire purification process and product quality.

[0029] Furthermore, step P20 in this embodiment of the application also includes:

[0030] P23: Divide the anomaly degree of each of the M process nodes by the sum of the anomalies of the M process nodes to obtain the confidence coefficients of the M process nodes; P24: Multiply the confidence coefficients of the M process nodes by the standard confidence window to obtain the M confidence windows.

[0031] Optionally, in further implementation, the configuration of the trusted window can be further optimized and adjusted based on the calculated process node anomalies.

[0032] First, the reliability coefficient of each process node is calculated by ratioing the anomaly degree of each process node to the sum of the anomalies of all process nodes. Specifically, this process aims to normalize the anomaly degree of each node into a relative value, facilitating the measurement of the relative risk level of different nodes throughout the purification process. The reliability coefficient calculated in this way reflects the proportion of abnormal risk for each process node in the overall purification process. A lower reliability coefficient indicates that the process node has experienced relatively few anomalies in its historical operation, suggesting a relatively stable operating state; conversely, a higher reliability coefficient indicates a relatively higher anomaly risk for the process node, requiring more stringent monitoring and control.

[0033] Next, the confidence coefficient of each process node is multiplied by a standard confidence window to obtain M confidence windows. The standard confidence window is typically a fixed reference value that can be set according to actual production needs and process requirements, representing the range of operating parameters that each process node should maintain under normal operating conditions. In this way, the confidence window of each process node can be dynamically adjusted according to its anomaly risk. For process nodes with lower anomaly risk, their confidence windows are relatively wide, meaning that the node's operating status is relatively stable within that time period, and the monitoring frequency can be appropriately relaxed. Conversely, for process nodes with higher anomaly risk, their confidence windows are relatively narrow, requiring more frequent monitoring and control to ensure that their operating status remains within a controllable range.

[0034] Through the above steps, the embodiments of this application can not only identify the anomaly degree of each process node, but also calculate the confidence coefficient based on the anomaly degree, and configure a confidence window suitable for each process node accordingly, thereby realizing refined and intelligent control of the purification process, effectively improving the stability of the entire purification system and product quality, while reducing the risks in the production process.

[0035] P30: Using the M trusted windows as monitoring periods, continuously monitor the operation of the M process nodes and perform operational risk analysis to obtain the M process node window operation monitoring data set sequence and the M process node window risk coefficient sequence. Each process node window risk coefficient and the corresponding process node window operation monitoring data set have a trusted time period identifier.

[0036] Furthermore, step P30 in this embodiment of the application also includes:

[0037] P31: When the purification production line starts up, obtain the first monitoring time point; P32: At the first monitoring time point, monitor the operation of the M process nodes to obtain M sets of first process node window operation monitoring data, and perform operation risk analysis on the M sets of first process node window operation monitoring data to obtain M first process node window risk coefficients; P33: Using the M reliable windows as the monitoring period, determine M second monitoring time points in combination with the first monitoring time point; P34: Based on the M second monitoring time points, monitor the operation of the M process nodes to obtain M sets of second process node window operation monitoring data, and perform operation risk analysis on the M sets of second process node window operation monitoring data to obtain M second process node window risk coefficients; P35: And so on, using the M reliable windows as the monitoring period, continuously monitor and perform operation risk analysis on the M process nodes to obtain the sequence of the M process node window operation monitoring data sets and the sequence of M process node window risk coefficients.

[0038] It should be understood that after configuring the trusted windows for the M process nodes of the purification production line, the next step is to continuously monitor the operation of these process nodes and conduct operational risk analysis. Using the M trusted windows as the monitoring period, each node is monitored in real time to ensure it operates within its normal operating range.

[0039] When the purification production line starts up, the system automatically records the start-up time as the first monitoring time point, i.e., the initial monitoring point. It then comprehensively monitors the operation of M process nodes, collecting operating parameters and status information for each node, such as key indicators like temperature, pressure, flow rate, and gas composition, forming M sets of first-process-node window operating monitoring data. Subsequently, a pre-defined risk assessment model is used to analyze the operational risks of these monitoring data sets, calculating the risk coefficient for each process node at that time point, thus obtaining the risk coefficients for the M first-process-node windows. The calculation of the risk coefficients is based on multiple factors, such as the degree to which parameters deviate from normal ranges and the stability of equipment operation, quantitatively reflecting the operational risk level of each process node.

[0040] Next, the completion point of the first round of monitoring, starting from the first monitoring time point, will be used as the starting point for the second round of monitoring. M second monitoring time points will be calculated and determined using M confidence windows as the monitoring cycle. The monitoring cycle is set based on the confidence window of each process node, ensuring that each node is monitored at an appropriate frequency during operation to promptly capture any potential risks. The determination of the second monitoring time points not only depends on the setting of the confidence window but also takes into account the actual operating conditions of the process nodes, ensuring that the monitoring of each node is flexible and meets actual needs.

[0041] Next, a second round of node operation monitoring is conducted on the M process nodes at M second monitoring time points. This monitoring process is similar to step P32. At the second monitoring time point, the system again monitors the node operation of the M process nodes, collects new operating parameters and status information, forms a set of operating monitoring data for the M second process node windows, and performs operational risk analysis on these data sets to obtain the risk coefficients for the M second process node windows. This process is similar to the monitoring and analysis at the first monitoring time point, but the time points are determined according to the reliable window period, ensuring the periodicity and continuity of monitoring.

[0042] Similarly, steps P33 and P34 will continue to be repeated. In each subsequent monitoring cycle, the system will continue to monitor and analyze the risks of each process node based on the completion time of the previous monitoring cycle. Each new monitoring cycle will generate a new dataset and risk coefficient sequence, ensuring that the system can track the operating status of each process node in real time, assess its potential risks, and adjust monitoring and risk control strategies according to changes.

[0043] Furthermore, each process node's window risk coefficient and corresponding window operation monitoring data set are marked with a reliable time period identifier. This identifier ensures that the monitoring data and risk analysis results correspond to the relevant reliable window, thus providing an accurate time reference for subsequent risk control. In this way, the system can monitor the operating status of each process node in real time, promptly identify potential risk points, and provide crucial data support for subsequent risk control, ensuring the stability of the entire purification process and product quality.

[0044] Furthermore, step P30 in this embodiment of the application also includes:

[0045] P30a: Obtain M first credible time periods, and identify the risk coefficients of M first process node windows and the corresponding M first process node window operation monitoring data sets. The starting time of the M first credible time periods is the first monitoring time point, and the duration of the M first credible time periods is the M credible windows.

[0046] In one possible embodiment of this application, the system will acquire M first credible time periods and identify the risk coefficients of the M first process node windows and the corresponding sets of M first process node window operation monitoring data.

[0047] First, M first-degree-of-sense time periods need to be obtained. The starting point of these degree-of-sense time periods is determined by the first monitoring time point, i.e., from the moment the purification production line starts. The duration of these time periods is the same as the time range of the M degree-of-sense windows, i.e., the monitoring cycle of each process node. Specifically, the degree-of-sense window of each process node has been configured with a suitable range of operating parameters based on its anomaly degree and confidence coefficient, and these ranges determine the monitoring cycle of each process node.

[0048] Next, the risk coefficients and monitoring data sets for the M first process node windows of each process node are identified. This identification serves to associate the monitoring data and risk coefficients of each process node within a specified first credible time period with that time period for subsequent analysis and processing. These data and risk coefficients represent the operational status and potential risks of each process node within the first credible time period.

[0049] Specifically, the system binds the monitoring data set of each process node and its corresponding risk coefficient to the corresponding first reliable time period. The data and risk coefficient of each process node will be organized according to time period, so that the subsequent monitoring system can clearly identify which process nodes have abnormal operating status and which nodes have high risk coefficients within a specific time period, thereby providing a basis for subsequent process control and risk warning.

[0050] This identification process enables the system to clearly distinguish the operational status of different process nodes at different time periods and to perform dynamic monitoring and comparison based on historical data. As subsequent monitoring cycles continue, the system can use this historical data as a reference to determine the changing trends of process nodes in different monitoring cycles, thereby more accurately assessing the risks and operational status of process nodes.

[0051] P40: Combining reliable time period identifiers, risk control is performed on M process nodes based on the risk coefficient sequence of M process node windows and the set of operational monitoring data of M process node windows.

[0052] Furthermore, step P40 in this embodiment of the application also includes:

[0053] P41: Extract the maximum value of the process node window risk coefficient from the sequence of M process node windows, and take the corresponding process node as the risk process node; P42: Based on the reliable time period identifier corresponding to the maximum value of the process node window risk coefficient, extract the M-1 cooperating process node window operation monitoring data sets of the same reliable time period in the sequence of M process node window operation monitoring data sets, and the risk process node window operation monitoring data set of the risk process node; P43: Perform risk control based on the risk process node window operation monitoring data set of the risk process node and the M-1 cooperating process node window operation monitoring data set of the M-1 cooperating process nodes.

[0054] It should be understood that, based on the previously determined reliable time period identifier, the risk coefficient sequence of M process node windows, and the set of operational monitoring data of M process node windows, precise risk control is carried out on the M process nodes to ensure that all process nodes can maintain a safe working state during operation. Furthermore, through comprehensive analysis of the interrelationships between nodes, operating parameters are adjusted in a timely manner to avoid the impact of potential risks on the purification process.

[0055] First, the risks of each process node are compared, and the process node with the highest risk in the current monitoring period is selected. To do this, the maximum value of the risk coefficient is extracted from the sequence of risk coefficients for M process nodes. This value represents which process node has the highest risk level in the current monitoring period. The risk coefficient of each process node reflects the degree of deviation between its operating state and the normal parameter range; the larger the value, the higher the risk. When the system identifies a risky process node, this node becomes the priority node to monitor and control throughout the purification process. The node with the highest risk coefficient may have significant deviations or potential faults, directly affecting product quality or system stability, and therefore requires focused adjustment.

[0056] Next, based on the reliable time period identifier corresponding to the maximum risk coefficient, the system extracts the operational monitoring data sets of the other M-1 process nodes within the same reliable time period from the sequence of operational monitoring data sets of the M process node windows, as well as the operational monitoring data set of the risky process node itself. The purpose of this filtering process is to obtain the operational status of other process nodes when the highest-risk process node experiences an anomaly, thereby providing comprehensive data support for subsequent risk control. In this way, the system can understand whether other process nodes also have potential anomalies or parameters that need adjustment when a problem occurs at the risky process node, enabling coordinated control.

[0057] Finally, taking into account the abnormal situations of high-risk process nodes and the operating status of other process nodes, the system reduces the risk level of high-risk process nodes and ensures the stable operation of the entire purification process by adjusting relevant process parameters, optimizing equipment operating conditions, or taking other necessary measures. Specifically, the system will assess the interrelationships between process nodes based on this monitoring data and risk analysis. For example, if the risk factor of a high-risk process node increases due to excessive pressure, the system may adjust the operating parameters of relevant equipment, such as reducing the compressor speed or increasing the power of the cooling system, while checking whether other cooperating process nodes need corresponding adjustments to ensure the coordinated operation of the entire system.

[0058] This approach not only identifies current risk points but also allows for coordinated control by considering their correlation with other nodes, ensuring the safety and stability of the purification process. This precise control method effectively avoids the negative impact of a single node anomaly on the entire system, thereby improving the efficiency of the purification process and the quality of the product.

[0059] Furthermore, step P43 in this embodiment of the application also includes:

[0060] P43-1: Adjust the risk parameters of the risk process nodes using the risk process node window operation monitoring data set to obtain risk process node adjustment parameters; P43-2: Based on the connection relationship between M process nodes, identify the adjustment parameters of the M-1 cooperating process nodes according to the risk process node adjustment parameters and the M-1 cooperating process node window operation monitoring data set to obtain M-1 cooperating process node adjustment parameters; P43-3: Perform risk control based on the risk process node adjustment parameters and the M-1 cooperating process node adjustment parameters. Specifically, adjusting the risk parameters of the risk process nodes using the risk process node window operation monitoring data set includes: pre-constructing an adjustment parameter identifier, and using the adjustment parameter identifier to identify parameters in the risk process node window operation monitoring data set to obtain risk process node adjustment parameters.

[0061] Optionally, to achieve precise risk control, the stability and safety of the entire purification process can be ensured by adjusting the parameters of the risk process nodes and the corresponding process nodes.

[0062] First, the operational monitoring data set of the risky process nodes is analyzed to determine the control parameters for these nodes. By analyzing various parameters (such as pressure, temperature, and flow rate) in the monitoring data set, the main factors affecting the risk are identified, and corresponding control measures are proposed. For example, a pre-built control parameter identifier is used to perform in-depth analysis of the monitoring data. This identifier, which can be built based on machine learning or data mining techniques, can automatically identify key parameters associated with the risky nodes and predict the parameter values ​​that need to be adjusted based on historical data. For instance, if the risky process node is a gas pretreatment node, and the risk manifests as poor water removal (e.g., excessive moisture), the control parameter identifier might identify parameter adjustment suggestions such as increasing cooling efficiency or enhancing the adsorption capacity of the molecular sieve.

[0063] After identifying the control parameters for the high-risk process nodes, the next step is to identify the control parameters for the M-1 cooperating process nodes based on the connections between the M process nodes. The core of this operation is to identify which parameters of the cooperating process nodes need adjustment by analyzing the relationship between the control parameters of the high-risk process nodes and the monitoring data sets of the M-1 cooperating process nodes. Since there are usually close interactions between the process nodes, when one process node experiences a significant risk, the operating status of other nodes may also be affected, or they may work together to mitigate the risk. For example, if the temperature of the high-risk process node is too high, the cooling, filtration, and adsorption operations of other nodes may need to be increased or adjusted. Therefore, based on the control parameters of the high-risk nodes and the operating data of other cooperating nodes, the control parameters of the M-1 cooperating process nodes can be calculated. These control parameters of the cooperating nodes will provide a basis for subsequent coordinated control, ensuring the stability of the entire process chain.

[0064] In this process, comprehensive control is performed based on the risk process node control parameters obtained from the first two steps and the control parameters of M-1 cooperating process nodes. This ensures that when an anomaly occurs at a risk process node, the impact of the risk can be minimized by adjusting the parameters of the cooperating process nodes. Specific control strategies will consider the mutual influence and dependencies between nodes to ensure that all process nodes can cooperate to achieve optimal operating results. For example, when an anomaly occurs at a certain process node, the operating conditions (such as temperature, pressure, flow rate, etc.) of its downstream or related nodes may need to be adjusted synchronously according to the adjusted control parameters to prevent an anomaly in one location from causing failures in other nodes.

[0065] Specific examples of adjustment strategies include: When the risk at the gas pretreatment node manifests as poor water removal or ineffective removal of particulate matter during coarse filtration, the adjustment strategies include: improving the cooling efficiency of the compression and cooling nodes to lower the gas temperature and ensure that condensed moisture is effectively removed, preventing it from entering downstream equipment; increasing the adsorption capacity or extending the adsorption cycle of the molecular sieve adsorption nodes to ensure further removal of moisture and impurities from the gas; and improving the filtration precision of the precision filtration and adsorption nodes to ensure that smaller particles and dissolved substances are effectively removed. When the risk at the separation node manifests as low separation efficiency leading to insufficient carbon dioxide purity, the adjustment strategies include: optimizing the pretreatment process to ensure that the gas quality entering the separation node meets requirements; and adjusting the parameters of the purification process, such as increasing the condensation temperature or optimizing the evaporation process, to further improve the purity of carbon dioxide. When the risk at the purification node manifests as residual impurities or purity fluctuations during the purification process, the adjustment strategies include: optimizing the separation process to ensure that the purity of the gas entering the purification node meets requirements; and increasing the detection frequency to promptly identify and adjust parameters in the purification process to ensure the quality of the final product.

[0066] This comprehensive control strategy enables precise adjustments to each process node, ensuring that even under high-risk conditions, the coordinated adjustments of other process nodes can effectively guarantee the safe and stable operation of the entire purification process.

[0067] Furthermore, the embodiments of this application also include step P50:

[0068] A risk control feedback window is determined, and anomalies are identified for M process nodes within the risk control feedback window. Early warning instructions are generated based on the identification results.

[0069] Specifically, after completing risk control of M process nodes in the purification production line, the risk control feedback window can be used to continuously identify anomalies in all M process nodes, ensuring that the system can capture potential anomalies in real time during operation and issue alarms in a timely manner when risks occur, so that corresponding measures can be taken to intervene and adjust.

[0070] Specifically, the first step is to determine the risk control feedback window. This window refers to the time period used to assess the effectiveness of risk control measures and monitor new anomalies after their implementation. The length of the feedback window can be set according to actual production needs and process characteristics, and should generally cover the critical operating cycle after the implementation of control measures to ensure a comprehensive assessment of the control effects.

[0071] Within the risk control feedback window, the system will perform comprehensive anomaly identification for M process nodes. For example, a pre-defined anomaly identification model can be used, combined with operational monitoring data from each process node, to analyze the operational status of each node in real time. The anomaly identification model can be built based on historical data, statistical analysis, or machine learning algorithms, and can identify abnormal situations that deviate from normal operating conditions. For instance, if the temperature or pressure parameters of a process node still exceed the normal range after adjustment, or if new abnormal fluctuations occur, these will be identified as abnormal situations.

[0072] Based on the anomaly identification results, corresponding early warning instructions are generated. These instructions are responses to identified anomalies, designed to alert operators or the automated control system to take timely further checks or adjustments. Early warning instructions may include, but are not limited to: issuing alarm signals, prompting operators to conduct checks, automatically adjusting relevant process parameters, or activating backup equipment. These instructions ensure that after risk control measures are implemented, the system can promptly detect and handle new anomalies, thereby maintaining the stability and safety of the entire purification process.

[0073] Furthermore, in certain situations, if multiple process nodes experience simultaneous or interconnected anomalies, the system may generate more complex early warning commands to coordinate the control actions of multiple nodes. This coordinated control ensures a comprehensive response from the entire system to risks, thereby minimizing the impact of risks on the purification process and improving system stability and safety. Through this real-time monitoring and response mechanism, the system can detect and handle anomalies in the shortest possible time, ensuring the smooth operation of the purification process.

[0074] In summary, the embodiments of this application have at least the following technical effects:

[0075] This application acquires M process nodes from the purification production line, extracts historical anomaly characteristics, identifies node anomalies, and configures a reliable window for each process node based on the anomaly level. Based on the reliable window, each node is continuously monitored, operational risks are analyzed, and operational monitoring data and risk coefficient sequences are generated. The risk coefficient of each node and its corresponding monitoring data set are combined with a reliable time period identifier to perform precise risk control on the M process nodes, ensuring stable and safe operation of the purification process.

[0076] The technology achieves the effect of timely identification and handling of risks through real-time monitoring and intelligent control mechanisms, thereby improving the stability and safety of the purification process.

[0077] Example 2, based on the same inventive concept as the intelligent control method for the food-grade carbon dioxide purification process in the foregoing examples, such as... Figure 2As shown, this application provides an intelligent control system for a food-grade carbon dioxide purification process. The system and method embodiments in this application are based on the same inventive concept. The system includes:

[0078] Anomaly characterization extraction module 11 is used to obtain M process nodes of the purification production line, traverse the M process nodes to extract historical anomaly characterizations, and obtain a set of historical anomaly characterizations of the M process nodes.

[0079] The trusted window configuration module 12 is used to identify the node anomaly degree of the M process nodes based on the historical anomaly representation set of the M process nodes, obtain the anomaly degree of the M process nodes, and configure the M trusted windows of the M process nodes according to the anomaly degree of the M process nodes.

[0080] The operation risk analysis module 13 is used to continuously monitor the operation of M process nodes and perform operation risk analysis with the M reliable windows as monitoring periods, respectively, to obtain the M process node window operation monitoring data set sequence and the M process node window risk coefficient sequence, wherein each process node window risk coefficient and the corresponding process node window operation monitoring data set have a reliable time period identifier.

[0081] Risk control module 14 is used to combine a reliable time period identifier and perform risk control on M process nodes based on the risk coefficient sequence of M process node windows and the set of operation monitoring data of M process node windows.

[0082] Furthermore, the trusted window configuration module 12 is also used to perform the following steps:

[0083] For each of the M historical anomaly representation sets of process nodes, cluster the anomaly representations of the same type within the process node to determine the M clusters of historical anomaly representations of process nodes; traverse the M clusters of historical anomaly representations of process nodes and perform weighted analysis to determine the anomaly degree of the M process nodes.

[0084] Furthermore, the trusted window configuration module 12 is also used to perform the following steps:

[0085] The M process node anomalies are each divided by the sum of the M process node anomalies to obtain the M process node confidence coefficients; the M process node confidence coefficients are then multiplied by the standard confidence window to obtain the M confidence windows.

[0086] Furthermore, the operational risk analysis module 13 is also used to perform the following steps:

[0087] When the purification production line starts up, a first monitoring time point is obtained. At the first monitoring time point, node operation monitoring is performed on the M process nodes to obtain M sets of first process node window operation monitoring data. Operational risk analysis is then performed on these M sets of first process node window operation monitoring data to obtain M risk coefficients for the first process node windows. Using the M reliable windows as the monitoring cycle, M second monitoring time points are determined in conjunction with the first monitoring time point. Based on the M second monitoring time points, node operation monitoring is performed on the M process nodes to obtain M sets of second process node window operation monitoring data. Operational risk analysis is then performed on these M sets of second process node window operation monitoring data to obtain M risk coefficients for the second process node windows. This process is repeated, using the M reliable windows as the monitoring cycle, to continuously monitor and perform operational risk analysis on the M process nodes, obtaining a sequence of M process node window operation monitoring data sets and a sequence of M process node window risk coefficients.

[0088] Furthermore, the operational risk analysis module 13 is also used to perform the following steps:

[0089] Get M first credible time periods, and identify the risk coefficients of M first process node windows and the corresponding M first process node window operation monitoring data sets. The starting time of the M first credible time periods is the first monitoring time point, and the duration of the M first credible time periods is the M credible windows.

[0090] Furthermore, the risk control module 14 is also used to perform the following steps:

[0091] Extract the maximum value of the process node window risk coefficient from the sequence of M process node windows, and designate the corresponding process node as the risk process node. Based on the reliable time period identifier corresponding to the maximum value of the process node window risk coefficient, extract the M-1 cooperative process node window operation monitoring data sets of the same reliable time period from the sequence of M process node window operation monitoring data sets, as well as the risk process node window operation monitoring data set of the risk process node. Perform risk control based on the risk process node window operation monitoring data set of the risk process node and the M-1 cooperative process node window operation monitoring data sets of the M-1 cooperative process nodes.

[0092] Furthermore, the risk control module 14 is also used to perform the following steps:

[0093] Risk parameters of the risk process nodes are adjusted based on the risk process node window operation monitoring data set to obtain risk process node adjustment parameters; based on the connection relationship between M process nodes, adjustment parameters of the M-1 cooperating process nodes are identified according to the risk process node adjustment parameters and the M-1 cooperating process node window operation monitoring data set to obtain M-1 cooperating process node adjustment parameters; risk adjustment is performed according to the risk process node adjustment parameters and the M-1 cooperating process node adjustment parameters.

[0094] Furthermore, the risk control module 14 is also used to perform the following steps:

[0095] A pre-built control parameter identifier is used to identify parameters in the risk process node window operation monitoring data set to obtain the control parameters of the risk process node.

[0096] Furthermore, the system also includes a risk control feedback module, which is used to determine a risk control feedback window, identify anomalies in M ​​process nodes in the risk control feedback window, and generate early warning instructions based on the identification results.

[0097] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0098] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0099] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application intends to include such modifications and variations.

Claims

1. An intelligent control method for food-grade carbon dioxide purification processes, characterized in that, The method includes: Get M process nodes of the purification production line, traverse the M process nodes to extract historical anomaly characteristics, and obtain a set of historical anomaly characteristics of the M process nodes. The node anomaly degree is identified by performing node anomaly degree identification on the historical anomaly representation set of the M process nodes to obtain the anomaly degree of the M process nodes, and the M confidence windows of the M process nodes are configured according to the anomaly degree of the M process nodes. Using the M trusted windows as monitoring periods, continuous monitoring of node operation and operation risk analysis are performed on the M process nodes to obtain the M process node window operation monitoring data set sequence and the M process node window risk coefficient sequence. Each process node window risk coefficient and the corresponding process node window operation monitoring data set have a trusted time period identifier. By combining reliable time period identifiers, risk control is carried out on M process nodes based on the risk coefficient sequence of M process node windows and the set of operational monitoring data of M process node windows.

2. The intelligent control method for food-grade carbon dioxide purification process as described in claim 1, characterized in that, The anomaly scores of the M process nodes are identified by performing node anomaly score identification on the historical anomaly characterization sets of the M process nodes, including: For each of the M historical anomaly representation sets of process nodes, cluster the anomaly representations of the same type within the process node to determine the M clusters of historical anomaly representations of process nodes. The anomaly degree of the M process nodes is determined by performing weighted analysis on the historical anomaly characterization clusters of the M clustered process nodes.

3. The intelligent control method for food-grade carbon dioxide purification process as described in claim 1, characterized in that, Configure M confidence windows for M process nodes based on the anomalies of the M process nodes, including: The confidence coefficients of the M process nodes are obtained by dividing the anomaly degree of each of the M process nodes by the sum of the anomaly degrees of the M process nodes. The M confidence coefficients of the process nodes are multiplied by the standard confidence window to obtain the M confidence windows.

4. The intelligent control method for food-grade carbon dioxide purification process as described in claim 1, characterized in that, Using the M reliable windows as monitoring periods, continuous monitoring and operational risk analysis are performed on the M process nodes to obtain a sequence of operational monitoring data sets for the M process node windows and a sequence of risk coefficients for the M process node windows, including: The first monitoring time point is obtained when the purification production line is started; At the first monitoring time point, node operation monitoring is performed on the M process nodes to obtain a set of M first process node window operation monitoring data, and operation risk analysis is performed on the set of M first process node window operation monitoring data to obtain the risk coefficient of M first process node window. Using the M trusted windows as the monitoring period, and combining the first monitoring time point, determine the M second monitoring time points; Based on the M second monitoring time points, node operation monitoring is performed on the M process nodes to obtain a set of M second process node window operation monitoring data, and operation risk analysis is performed on the set of M second process node window operation monitoring data to obtain the risk coefficient of M second process node window. Similarly, using the M reliable windows as the monitoring period, continuous monitoring and operational risk analysis are performed on the M process nodes to obtain the sequence of operational monitoring data sets and the sequence of risk coefficients for the M process node windows.

5. The intelligent control method for food-grade carbon dioxide purification process as described in claim 4, characterized in that, Get M first credible time periods, and identify the risk coefficients of M first process node windows and the corresponding M first process node window operation monitoring data sets. The starting time of the M first credible time periods is the first monitoring time point, and the duration of the M first credible time periods is the M credible windows.

6. The intelligent control method for food-grade carbon dioxide purification process as described in claim 1, characterized in that, Combining reliable time period identifiers, risk control is performed on M process nodes based on the risk coefficient sequence of M process node windows and the operational monitoring data set sequence of M process node windows, including: Extract the maximum value of the process node window risk coefficient from the sequence of M process node windows, and take the corresponding process node as the risk process node. Based on the reliable time period identifier corresponding to the maximum risk coefficient of the process node window, extract the M-1 cooperative process node window operation monitoring data sets of the same reliable time period in the sequence of M process node window operation monitoring data sets, as well as the risk process node window operation monitoring data set of the risk process node. Risk control is performed based on the risk process node window operation monitoring data set of the risk process node and the M-1 cooperating process node window operation monitoring data set of the M-1 cooperating process nodes.

7. The intelligent control method for food-grade carbon dioxide purification process as described in claim 6, characterized in that, Risk control is performed based on the risk process node window operation monitoring data set of the aforementioned risk process node and the M-1 cooperating process node window operation monitoring data set of the aforementioned M-1 cooperating process nodes, including: The risk parameters of the risk process nodes are adjusted by analyzing the risk process node window operation monitoring data set to obtain the risk process node adjustment parameters. Based on the connection relationship between M process nodes, the control parameters of the M-1 cooperating process nodes are identified according to the control parameters of the risk process node and the window operation monitoring data set of the M-1 cooperating process nodes, so as to obtain the control parameters of the M-1 cooperating process nodes. Risk control is performed based on the risk process node control parameters and the control parameters of the M-1 cooperating process nodes.

8. The intelligent control method for food-grade carbon dioxide purification process as described in claim 7, characterized in that, A pre-built control parameter identifier is used to identify parameters in the risk process node window operation monitoring data set to obtain the control parameters of the risk process node.

9. The intelligent control method for food-grade carbon dioxide purification process as described in claim 1, characterized in that, include: A risk control feedback window is determined, and anomalies are identified for M process nodes within the risk control feedback window. Early warning instructions are generated based on the identification results.

10. An intelligent control system for food-grade carbon dioxide purification processes, characterized in that, The system includes: An anomaly characterization extraction module is used to obtain M process nodes of the purification production line, traverse the M process nodes to extract historical anomaly characterizations, and obtain a set of historical anomaly characterizations of the M process nodes. A trusted window configuration module is used to identify the node anomaly degree of the M process nodes based on the historical anomaly representation set of the M process nodes, obtain the anomaly degree of the M process nodes, and configure the M trusted windows of the M process nodes according to the anomaly degree of the M process nodes. The operation risk analysis module is used to continuously monitor the operation of M process nodes and perform operation risk analysis with the M reliable windows as monitoring periods, respectively, to obtain the M process node window operation monitoring data set sequence and the M process node window risk coefficient sequence, wherein each process node window risk coefficient and the corresponding process node window operation monitoring data set have a reliable time period identifier. The risk control module is used to perform risk control on M process nodes by combining a reliable time period identifier and based on the risk coefficient sequence of M process node windows and the set sequence of M process node window operation monitoring data.

Images