Ai-driven industrial decarbonization process control method and system based on feedback regulation

Abstract

The application provides an AI-driven industrial decarbonization process control method and system based on feedback regulation, relating to the technical field of industrial control. The method comprises: performing industrial decarbonization process monitoring to obtain a process operation data stream sequence; performing process condition time sequence iterative analysis; performing matching in a pre-constructed causal knowledge base; traversing an associated causal clue set to perform historical data backtracking and abnormality investigation to obtain an abnormal associated causal clue set and an abnormality degree set; performing feedback control scheme identification to obtain a target feedback control scheme and generate a feedback regulation instruction, which is issued to a control execution module to execute the target feedback control scheme. The application solves the technical problem that, in the prior art, due to the fact that static model prediction or experience rule-based determination is mainly used, it is difficult to realize real-time tracing and dynamic intervention of process-level carbon emission abnormalities, resulting in an unsatisfactory carbon emission control effect in a complex environment, and improves the control effect of the industrial decarbonization process.

Description

Technical Field

[0001] This application relates to the field of industrial control technology, and in particular to an AI-driven industrial decarbonization process control method and system based on feedback regulation. Background Technology

[0002] With the continuous development of industrial automation and environmental protection technologies, the requirements for energy consumption and carbon emission control in industrial production processes are becoming increasingly stringent. However, existing industrial decarbonization process control methods mainly rely on static model predictions or judgments based on empirical rules, which are significantly insufficient in dealing with carbon emission changes under complex operating conditions. Static model prediction methods cannot fully reflect the dynamic changes in the production process, resulting in low prediction accuracy. In actual production processes, factors such as equipment parameters, energy consumption data, and environmental disturbance factors often change continuously, causing significant deviations between the prediction results of static models and the actual situation. Judgment methods based on empirical rules depend on the experience of operators and are difficult to standardize and scale up. When faced with complex operating conditions, the lack of systematic modeling and analysis tools makes it difficult to find the precise root cause when carbon emissions are abnormal, leading to unsatisfactory carbon emission control results.

[0003] In summary, existing technologies suffer from several problems. They rely primarily on static model predictions or empirical rule-based judgments, lacking the ability to systematically model the causes and manifestations of carbon emission changes under complex operating conditions. This makes it difficult to achieve real-time tracking and dynamic intervention of process-level carbon emission anomalies, resulting in unsatisfactory carbon emission control effects in complex environments. Summary of the Invention

[0004] The purpose of this application is to provide an AI-driven industrial decarbonization process control method and system based on feedback regulation, in order to solve the technical problems in the prior art, which are mainly based on static model prediction or judgment based on empirical rules, lack the ability to systematically model the causes and manifestations of carbon emission changes under complex operating conditions, make it difficult to achieve real-time tracing and dynamic intervention of process-level carbon emission anomalies, and thus lead to unsatisfactory carbon emission control effects in complex environments.

[0005] In view of the above problems, this application provides an AI-driven industrial decarbonization process control method and system based on feedback regulation.

[0006] Firstly, this application provides an AI-driven industrial decarbonization process control method based on feedback regulation. This method is implemented through an AI-driven industrial decarbonization process control system based on feedback regulation. The method includes: performing industrial decarbonization process monitoring to obtain a process operation data stream sequence, where each data stream includes equipment parameters, energy consumption data, carbon emission values, and environmental disturbance factors; traversing the process operation data stream sequence to perform iterative analysis of process operating conditions to obtain iterative abnormal operating condition state characteristics; matching the iterative abnormal operating condition state characteristics against a pre-constructed causal knowledge base to obtain a set of associated causal clues; traversing the set of associated causal clues to perform historical data backtracking and anomaly investigation to obtain an abnormal associated causal clue set and an anomaly degree set; identifying a feedback control scheme based on the abnormal associated causal clue set and the anomaly degree set to obtain a target feedback control scheme, generating a feedback regulation instruction, and issuing the feedback regulation instruction to the control execution module to execute the target feedback control scheme.

[0007] Optionally, a multimodal sensing array is deployed at key nodes of the target industrial decarbonization process. The multimodal sensing array includes an equipment parameter acquisition unit, an energy consumption metering unit, a carbon emission monitoring device, and an environmental disturbance sensing module. The multimodal sensing array is traversed to perform simultaneous sequential sensing of key nodes, thereby constructing a time-synchronized sequence of process operation data streams.

[0008] Optionally, a first process running data stream is extracted from the process running data stream sequence; the first process running data stream is subjected to condition status identification based on a preset process condition standard value to obtain a first process running condition status vector; the preset process condition standard value and the first process running condition status vector are added to a first iterative memory unit; a second process running data stream extracted from the process running data stream sequence is subjected to process condition time-series iterative analysis based on the first iterative memory unit to obtain a second process running condition status vector; the first iterative memory unit is updated based on the second process running condition status vector to obtain a second iterative memory unit, and so on, time-series iterative analysis is performed on the remaining process running condition status vectors in the process running data stream sequence to obtain a target iterative memory unit; anomaly feature identification is performed on the target iterative memory unit to obtain the iterative abnormal running condition status features.

[0009] Optionally, a preset process condition standard value is extracted from the first iterative memory unit, and the process condition status is identified on the second process running data stream to obtain a second initial process running condition status vector; a first process running condition status vector is extracted from the first iterative memory unit, and the second initial process running condition status vector is subjected to time-series iterative enhancement to obtain a second process running condition status vector.

[0010] Optionally, the element similarity between the first process running condition state vector and the second initial process running condition state vector is calculated to obtain an element similarity set; the element similarity set is matrix-processed to construct a temporal iterative enhancement matrix; the temporal iterative enhancement matrix is ​​used to perform convolution enhancement on the second initial process running condition state vector to obtain the second process running condition state vector.

[0011] Optionally, a set of historical abnormal related events in the industrial decarbonization process is obtained; causal clues are extracted by traversing the set of historical abnormal related events to obtain a set of historical abnormal causal clues and a corresponding set of historical abnormal operating condition features; the set of historical abnormal causal clues and the corresponding set of historical abnormal operating condition features are integrated to obtain an integrated set of historical abnormal causal clues and a corresponding integrated set of historical abnormal operating condition features; the integrated set of historical abnormal causal clues and the corresponding integrated set of historical abnormal operating condition features are associated and stored to obtain the causal knowledge base.

[0012] Optionally, the set of historical abnormal operating condition status features is integrated into a similar category, and the integration results are averaged to obtain an integrated set of historical abnormal operating condition status features; the set of historical abnormal causal clues is mapped and integrated based on the integrated set of historical abnormal operating condition status features, and the union of the mapping and integration results is obtained to obtain the integrated set of historical abnormal causal clues.

[0013] Optionally, in the historical data backtracking window, the set of related causal clues is traversed to perform historical data backtracking, obtaining a set of historical data sequences of related causal clues; the set of historical data sequences of related causal clues is traversed to identify standard value deviations, obtaining a set of deviation value sequences of related causal clues; based on the set of deviation value sequences of related causal clues, anomaly screening is performed on the set of related causal clues to obtain an abnormal set of related causal clues; the abnormal related causal clue deviation value sequences corresponding to the abnormal set of related causal clues are extracted from the set of deviation value sequences of related causal clues and the deviation value mean is processed to obtain the set of anomalies.

[0014] Optionally, a feedback control scheme identifier is pre-built, and the feedback control scheme identifier is used to identify the set of abnormal correlation causal clues and the set of abnormality degree to obtain the target feedback control scheme.

[0015] Secondly, this application also provides an AI-driven industrial decarbonization process control system based on feedback regulation, used to execute the AI-driven industrial decarbonization process control method based on feedback regulation as described in the first aspect. The AI-driven industrial decarbonization process control system based on feedback regulation includes: a process monitoring module for performing industrial decarbonization process monitoring and obtaining a process operation data stream sequence, wherein each process operation data stream includes equipment parameters, energy consumption data, carbon emission values, and environmental disturbance factors; a time-series iterative analysis module for traversing the process operation data stream sequence to perform time-series iterative analysis of process operating conditions and obtain iterative abnormal operating condition state characteristics; a knowledge base matching module for matching based on the iterative abnormal operating condition state characteristics in a pre-built causal knowledge base to obtain a set of associated causal clues; an anomaly investigation module for traversing the set of associated causal clues to perform historical data backtracking anomaly investigation and obtain an abnormal associated causal clue set and an anomaly degree set; and a control scheme execution module for identifying feedback control schemes based on the abnormal associated causal clue set and anomaly degree set, obtaining a target feedback control scheme, generating a feedback regulation instruction, and sending the feedback regulation instruction to the control execution module to execute the target feedback control scheme.

[0016] One or more technical solutions provided in this application have at least the following beneficial effects:

[0017] By performing industrial decarbonization process monitoring, a process operation data stream sequence is obtained. Each process operation data stream includes equipment parameters, energy consumption data, carbon emission values, and environmental disturbance factors. The process operation data stream sequence is traversed to perform iterative analysis of process operating conditions to obtain iterative abnormal operating condition state characteristics. Based on the iterative abnormal operating condition state characteristics, a pre-built causal knowledge base is matched to obtain a set of related causal clues. The set of related causal clues is traversed to perform historical data backtracking and anomaly investigation to obtain an abnormal related causal clue set and an anomaly degree set. Based on the abnormal related causal clue set and the anomaly degree set, a feedback control scheme is identified to obtain a target feedback control scheme, and a feedback adjustment command is generated and sent to the control execution module to execute the target feedback control scheme. In other words, by collecting the process operation data flow sequence, performing time-series iterative analysis of the process operating conditions, extracting the state characteristics of abnormal operation, using a pre-built causal knowledge base to trace the causes of carbon emission anomalies, and by backtracking historical data, associating the causes of anomalies with actual operating conditions, determining the target feedback control scheme to dynamically intervene in the decarbonization process, the precise tracing and efficient control of carbon emission anomalies are achieved, greatly improving the control effect of industrial decarbonization processes.

[0018] The above description is merely an overview of the technical solution of this application. To better understand the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the AI-driven industrial decarbonization process control method based on feedback regulation proposed in this application.

[0021] Figure 2 This is a schematic diagram of the AI-driven industrial decarbonization process control system based on feedback regulation in this application.

[0022] Figure labeling: Process monitoring module 11, Time series iterative analysis module 12, Knowledge base matching module 13, Anomaly investigation module 14, Control scheme execution module 15. Detailed Implementation

[0023] This application provides an AI-driven industrial decarbonization process control method and system based on feedback regulation. It addresses the technical problem in existing technologies that rely primarily on static model predictions or empirical rule-based judgments, lacking the ability to systematically model the causes and manifestations of carbon emission changes under complex operating conditions. This makes it difficult to achieve real-time tracing and dynamic intervention of process-level carbon emission anomalies, resulting in unsatisfactory carbon emission control performance in complex environments. By collecting process operation data stream sequences and performing iterative analysis of process operating conditions, the application extracts the state characteristics of abnormal operation. Using a pre-built causal knowledge base, it traces the causes of carbon emission anomalies. Through historical data backtracking, it correlates the causes of anomalies with actual operating conditions, determines a target feedback control scheme, and dynamically intervenes in the decarbonization process. This achieves accurate tracing and efficient control of carbon emission anomalies, significantly improving the control effect of industrial decarbonization processes.

[0024] The technical solutions of this application will now be clearly and completely described 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. It should be understood that this application is not limited to the exemplary embodiments described herein. 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. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0025] Example 1, please refer to the appendix. Figure 1 This application provides an AI-driven industrial decarbonization process control method based on feedback regulation. The method is executed by an AI-driven industrial decarbonization process control system based on feedback regulation, and specifically includes the following steps:

[0026] S100: Perform industrial decarbonization process monitoring to obtain a sequence of process operation data streams, where each process operation data stream includes equipment parameters, energy consumption data, carbon emission values, and environmental disturbance factors.

[0027] Furthermore, this application S100 includes:

[0028] A multimodal sensing array is deployed at key nodes in the target industrial decarbonization process. The multimodal sensing array includes an equipment parameter acquisition unit, an energy consumption metering unit, a carbon emission monitoring device, and an environmental disturbance sensing module. The multimodal sensing array is traversed to perform simultaneous sequential sensing of key nodes, thereby constructing a time-synchronized sequence of process operation data streams.

[0029] Specifically, at key nodes in the industrial decarbonization process, multimodal sensing arrays are deployed to comprehensively monitor various factors that may affect carbon emissions during production. These arrays include equipment parameter acquisition units, energy consumption metering units, carbon emission monitoring equipment, and environmental disturbance sensing modules. Each sensing device is responsible for collecting different types of data in real time. For example, equipment parameter acquisition units collect data on temperature, pressure, flow rate, rotational speed, and valve opening; energy consumption metering units connect to current transformers, calorimeters, and gas flow meters to obtain real-time energy consumption data; carbon emission monitoring equipment uses CO2 concentration detectors, flue gas analyzers, and infrared sensor modules to collect carbon emission values; and environmental disturbance sensing modules include temperature and humidity sensors, raw material batch recorders, and external power supply stability detectors.

[0030] A multimodal sensing array is used to simultaneously monitor each key node of the target industrial decarbonization process from multiple dimensions. This simultaneous sequential sensing means that all sensing devices at each key point operate concurrently, collecting data and ensuring that this data is synchronized in time. The multimodal sensing array generates a time-synchronized process operation data stream sequence. This sequence integrates collected equipment parameters, energy consumption data, carbon emission values, and environmental disturbance factors into a unified process operation data stream, including time-synchronized data collected from various sensing devices. Time synchronization means that data from different sources are arranged in the same time unit, ensuring data order and temporal consistency and avoiding analytical errors caused by time deviations.

[0031] In the process of industrial decarbonization, the energy sector must be addressed first, as energy production and consumption are the core sources of carbon emissions. By combining sustainable energy solutions with clean energy sources such as photovoltaics, hydrogen, and wind power to replace traditional fossil fuels, carbon emissions from energy consumption can be effectively reduced, potentially decreasing direct carbon emissions by 30%-60%. For example, a steel company reduced its annual coal consumption by 120,000 tons and its carbon emission intensity by 42% by deploying a wind-solar hybrid power generation system. Multimodal sensing arrays monitor energy consumption and carbon emission levels in real time. Energy metering units monitor the use of clean energy, while carbon emission monitoring equipment records carbon emissions after conversion from traditional to clean energy sources. Simultaneously, data from waste heat recovery systems, combined with microgrid energy allocation, ensures efficient energy utilization and minimized carbon emissions. For example, a steel company reduced energy consumption by 30% through waste heat recovery technology, thereby reducing carbon emissions by 42%. Deploying multimodal sensing arrays comprehensively and in real-time monitors all factors affecting carbon emissions, ensuring the completeness and accuracy of the monitored data.

[0032] S200: Traverse the process operation data stream sequence to perform process condition timing iterative analysis and obtain the characteristics of iterative abnormal operation conditions.

[0033] Furthermore, this application S200 includes:

[0034] Extract a first process operation data stream from the process operation data stream sequence; identify the operating status of the first process operation data stream based on a preset process operating condition standard value to obtain a first process operation operating condition state vector; add the preset process operating condition standard value and the first process operation operating condition state vector into a first iterative memory unit; perform a process operating condition time-series iterative analysis on the second process operation data stream extracted from the process operation data stream sequence based on the first iterative memory unit to obtain a second process operation operating condition state vector; update the first iterative memory unit based on the second process operation operating condition state vector to obtain a second iterative memory unit, and so on, perform time-series iterative analysis on the remaining process operation operating condition state vectors in the process operation data stream sequence to obtain a target iterative memory unit; identify abnormal features in the target iterative memory unit to obtain the iterative abnormal operation operating condition state features.

[0035] Specifically, the process traverses the sequence of process operation data streams, randomly extracting the first process operation data stream, which contains information such as equipment parameters, energy consumption data, and carbon emission values ​​within a certain time period. Based on the industrial equipment, process flow, and actual operating conditions, a preset process operating condition standard value is set, representing the parameter range that the equipment and process should achieve under normal operating conditions. Using the preset process operating condition standard value, the operating condition status of each parameter in the first process operation data stream is identified. By comparing the actual values ​​in the first process operation data stream with the preset process operating condition standard value, it is determined whether the current equipment and process are normal. For example, if the standard value is that the temperature of a certain piece of equipment in a chemical plant is between 700℃ and 950℃, while the actual temperature is 850℃, it indicates that the equipment is in a normal state; otherwise, it may be in an abnormal state. The first process operation condition state vector is a mathematical vector describing the first process operation data stream, containing the values ​​of various parameters of equipment operation (such as temperature, pressure, flow rate, etc.). Each state vector represents the operating condition state at a specific point in time.

[0036] The first process operation status vector and the preset process status standard value are stored together in the first iteration memory unit. By continuously analyzing new data and comparing the status vector with the standard value, the data in the memory unit is gradually optimized. The current state may be caused by other equipment or other processes, so it is also necessary to investigate the impact of other clues. If problems are found, they need to be addressed and adjusted accordingly.

[0037] The second process execution data stream is extracted from the process execution data stream sequence. This is the process execution data stream at the next moment after the first process execution data stream. Based on the first iterative memory unit, a process condition time-series iterative analysis is performed on the second process execution data stream. That is, the process condition state vector of the first process execution data stream is combined with the second process execution data stream to determine the condition state of the second process and optimize it.

[0038] Updating the first iterative memory unit by using the second process's operating condition state vector combines the operating condition information of the second process with that of the first process, resulting in a more comprehensive understanding of the entire process. The second process's operating condition state vector and preset process operating condition standard values ​​are added to the first iterative memory unit to form the second iterative memory unit. For example, if the first iterative memory unit stores the first process's operating condition state vector and preset process operating condition standard values, then during the update process, the second process's operating condition state vector and preset process operating condition standard values ​​will also be stored in the second iterative memory unit. The second iterative memory unit is a new memory unit obtained based on the update of the first iterative memory unit. Each time new process operation data (such as the second process's operating condition state vector) is added, the first iterative memory unit is updated, thus forming a new iterative memory unit.

[0039] This process continues, traversing the data flow sequence to extract subsequent data and progressively performing iterative time-series analysis of the process conditions to enrich the data. Each iteration generates a new iterative memory unit based on the previous iterative memory unit and the current process condition state vector. For example, if there are 100 data flows in the process data flow sequence, 99 iterations will be performed to finally obtain the target iterative memory unit. During each iteration, information from the previous iterative memory unit is used to enhance the current process condition state vector, resulting in more accurate condition information. Iterative time-series analysis is a method of analyzing current data based on historical data, using information from the previous point in time to predict and optimize future operating conditions. As the data flows are continuously iterated and updated, all process data is analyzed cyclically until the final target iterative memory unit is obtained.

[0040] The target iterative memory unit contains all data after time-series iterative analysis, representing a comprehensive view of the entire process. It includes the temporal evolution and trends of all key operating conditions (such as temperature, pressure, and flow rate) of the process system. In-depth analysis of historical operating condition data within the iterative memory unit identifies features that deviate from normal operating patterns or potential abnormal trends. Abnormal features typically manifest as drastic changes in certain operating condition values ​​or exceeding set standard ranges, indicating problems such as equipment failure, process instability, or low energy efficiency. Deviations in standard deviation and mean are usually used to determine whether data exceeds normal tolerance ranges. Abnormal features typically include operating conditions that deviate from expectations, experience failures, or exhibit abnormalities.

[0041] When the abnormal feature identification is completed, the abnormal operating condition features are extracted based on the identified abnormal state to obtain the iterative abnormal operating condition state features. These features describe the deviation of certain operating conditions from the normal state during the iteration process, including certain equipment failures, system abnormalities, or abnormal operating behaviors, revealing abnormal or atypical operating conditions in the industrial process.

[0042] Anomaly identification, as a dynamic feedback mechanism, is a key component of process optimization. By detecting anomalies in real-time data during production, potential problems in the process flow can be identified promptly, especially those not detected by traditional process parameter adjustments. Process optimization is like reprogramming the heart of industry; process reengineering unlocks hidden carbon reduction potential. Traditional industrial processes contain significant energy waste and redundant steps. Through process parameter optimization, lean production management, and supply chain collaboration, end-to-end energy reduction can be achieved. For example, an automobile manufacturer identified 20% energy waste in the casting process using process diagnostic tools from the SPARK theory; after modification, energy consumption per unit product decreased by 18%.

[0043] SPARK theory is a theoretical framework specifically designed for carbon reduction in the industrial sector, addressing the challenges faced by industry (process industries, which account for 72% of industrial carbon emissions) in achieving decarbonization goals. SPARK theory deeply integrates five key elements: sustainable energy, process optimization, AI analysis, real-time control, and knowledge systems, forming a co-evolutionary framework that provides a systematic solution for industrial carbon reduction. SPARK's five core elements include S, P, A, R, and K. S stands for Sustainable Energy, which involves replacing traditional fossil fuels with clean energy sources such as photovoltaics and hydrogen, combined with technologies like waste heat recovery and microgrid regulation, to reduce direct carbon emissions by 30%-60%. P stands for Process Optimization, achieving end-to-end energy conservation through process parameter optimization, lean production management, and supply chain collaboration. A stands for AI Analytics, utilizing machine vision, deep learning, and other technologies for real-time monitoring, prediction, and scheduling optimization. For example, a machine vision-based real-time energy consumption monitoring system can dynamically adjust equipment operating status and, combined with deep learning, predict carbon emission peaks to optimize production scheduling. R stands for Real-time Control, achieving closed-loop response and dynamically adjusting equipment parameters (such as temperature and pressure) to quickly control carbon emissions. K stands for Knowledge Embedding, used to standardize and encapsulate modules such as energy management, process parameters, and AI models to form a cross-industry knowledge graph.

[0044] The process of performing time-series iterative analysis of the process operation data flow sequence is equivalent to the P-factor in SPARK theory. Through real-time monitoring and analysis of operating conditions, potential problems can be identified in the process, allowing for timely adjustments and ensuring production stability, thus preventing production interruptions or safety accidents caused by sudden failures. By continuously optimizing and updating the target iterative memory unit, high-energy-consumption or high-carbon-emission operating conditions are identified, and adjustments are made based on this abnormal data, thereby achieving overall energy efficiency improvement and emission reduction effects throughout the process.

[0045] Furthermore, this application also includes the following steps:

[0046] The preset process operating condition standard value is extracted from the first iterative memory unit, and the operating condition status is identified for the second process running data stream to obtain the second initial process running operating condition status vector; the first process running operating condition status vector is extracted from the first iterative memory unit, and the second initial process running operating condition status vector is subjected to time-series iterative enhancement to obtain the second process running operating condition status vector.

[0047] Calculate the element similarity between the first process running condition state vector and the second initial process running condition state vector to obtain an element similarity set; perform matrix processing based on the element similarity set to construct a temporal iterative enhancement matrix; use the temporal iterative enhancement matrix to perform convolution enhancement on the second initial process running condition state vector to obtain the second process running condition state vector.

[0048] Specifically, preset process condition standard values ​​are extracted from the first iteration memory unit, representing the normal range of various key parameters in the production process. Based on these preset process condition standard values, the operating status of the second process's data stream is identified by comparing the actual parameters in the second process's data stream with the preset operating condition standard values. For example, if the temperature value is higher than the preset standard value, the operating status may be identified as an overload state; if the flow rate exceeds the set range, it may be an abnormal state. The core of operating status identification is finding the deviation between each process parameter (such as temperature, pressure, etc.) and the standard value. After operating status identification, a preliminary state vector containing the second process's operation is obtained, which includes the preliminary state vector of the second process's data stream. For example, the temperature is 850℃, the pressure is 28kPa, the flow rate is 55L / min, and the carbon dioxide concentration is 13.1%.

[0049] The first process operation condition state vector is extracted from the first iterative memory unit. Then, a temporal iterative enhancement is performed on the second initial process operation condition state vector. This involves comparing, optimizing, and enhancing the states of the first and second process operation data streams based on their temporal relationship, identifying potential deviations or anomalies in the second process operation data stream, and improving the accuracy and robustness of the state vector. Temporal iterative enhancement is a dynamic adjustment process that considers the relationship between preceding and following data streams, improving the accuracy of operation condition identification through iterative updates. After temporal iterative enhancement, the resulting second process operation condition state vector is more accurate than the second initial process operation condition state vector, reflecting a more realistic operation condition.

[0050] Specifically, the element similarity is calculated between the first process operating condition state vector and the second initial process operating condition state vector. Each operating condition state vector contains multiple elements (such as temperature, pressure, etc.), therefore, it is necessary to calculate the similarity between each pair of corresponding elements. Element similarity is used to quantify the similarity between corresponding elements in the first process operating condition state vector and the second initial process operating condition state vector, helping to identify the similarities and differences between the two operating condition state vectors. For example, the similarity of temperature values, the similarity of pressure values, etc. Euclidean distance is used to measure the distance between two vectors in space; a smaller Euclidean distance indicates a higher similarity between the two vectors. The most suitable similarity calculation method is selected based on the different elements in the operating condition state vectors. For each pair of parameters (such as temperature, pressure, flow rate, etc.), the calculated similarity value is stored in the element similarity set. For example, the operating state vector of the first process is: temperature 880℃, pressure 40kPa, flow rate 60L / min, and CO2 concentration 12%; the operating state vector of the second initial process is: temperature 890℃, pressure 40.1kPa, flow rate 61L / min, and CO2 concentration 12.2%. Element similarity is calculated: temperature similarity is 10, pressure similarity is 0.1, flow rate similarity is 1, and CO2 concentration similarity is 0.2.

[0051] The element similarity set is normalized to convert the similarity values ​​to a standard range (e.g., [0,1]). The processed data is then filled into an initially empty matrix to obtain the time-series iterative enhancement matrix. Each row and column of the matrix represents the process state at different time periods, and each element in the matrix corresponds to the similarity between two working states. The time-series iterative enhancement matrix is ​​a matrix obtained through matrix transformation, forming a matrix that can describe the changing patterns of time-series data. It is used to enhance the characteristics of time-series data and further improve the accuracy and reliability of the data.

[0052] The second initial process operation state vector is enhanced by convolution using a temporal iterative enhancement matrix. Convolution operations are used to extract local features and temporal patterns from data. By performing convolution operations between the convolution kernel and the temporal matrix, the temporal variation patterns in the data are amplified. Convolution enhancement is a signal processing technique used to enhance local features of data. In the processing of the operation state vector, convolution enhancement enhances the temporal features in the data by performing convolution operations on the temporal iterative enhancement matrix, thereby improving the accuracy and robustness of predictions. Convolution operations typically use a sliding window approach, multiplying the convolution kernel (a small matrix) with the input matrix and moving the window to capture temporal features. Through convolution enhancement, the operation states at adjacent time points can be correlated, enhancing their temporal information. After convolution enhancement, an optimized second process operation state vector is obtained, which contains the results of improving and optimizing the operation states of the second time period. For example, the second initial process operating condition state vector (temperature 890℃, pressure 40.1kPa, flow rate 61L / min, CO2 concentration 12.2%) is convolved and enhanced to finally obtain a temperature of 889℃, a pressure of 40.5kPa, a flow rate of 61.2L / min, and a CO2 concentration of 12.5%.

[0053] By extracting preset operating condition standard values ​​from the first iteration memory unit and combining them with a time-series iterative enhancement method, the operating condition state vector of the current process is accurately identified and adjusted, making the judgment of the current operating condition more accurate and providing strong data support for subsequent optimization, thereby improving the efficiency and stability of the entire industrial decarbonization and optimization process.

[0054] S300: Based on the characteristics of iterative abnormal operation conditions, match them with a pre-built causal knowledge base to obtain a set of related causal clues.

[0055] Furthermore, this application S300 includes:

[0056] Obtain a set of historical abnormal related events for the industrial decarbonization process; traverse the set of historical abnormal related events to extract causal clues, obtaining a set of historical abnormal causal clues and a corresponding set of historical abnormal operating condition status features; integrate the set of historical abnormal causal clues and the corresponding set of historical abnormal operating condition status features to obtain an integrated set of historical abnormal causal clues and a corresponding integrated set of historical abnormal operating condition status features; associate and store the integrated set of historical abnormal causal clues and the corresponding integrated set of historical abnormal operating condition status features to obtain the causal knowledge base.

[0057] Furthermore, this application also includes the following steps:

[0058] The historical abnormal operating condition status feature set is integrated into a similar category, and the integration results are averaged to obtain an integrated historical abnormal operating condition status feature set; based on the integrated historical abnormal operating condition status feature set, the historical abnormal causal clue set is mapped and integrated, and the union of the mapping and integration results is obtained to obtain the integrated historical abnormal causal clue set.

[0059] Specifically, through equipment self-inspection systems, operator reports, and automated monitoring systems, all past abnormal events related to the industrial decarbonization process are acquired. Each event includes the operating conditions at the time of the abnormality and possible causal relationships, resulting in a set of historical abnormal event associations. This set of historical abnormal event associations is then traversed to extract causal clues and identify the causal relationships between abnormal events. For example, if excessive pressure in a piece of equipment leads to a temperature increase, this is a typical causal relationship.

[0060] By analyzing the relationships between anomalous events, the root causes behind the anomalies can be identified. Algorithms such as Apriori or FP-growth can be used to discover which anomalous events frequently occur together. For example, it might be found that excessively high reactor temperatures are always accompanied by cooling system failures or heating system overloads. In the process of extracting causal clues, in addition to discovering the causal relationships between anomalous events, it is also necessary to extract the operating parameters (temperature, pressure, flow rate, etc.) at the time of each event. These parameters provide context for the occurrence of the anomalous events. For example, some anomalous events might mention excessively high temperatures, equipment overload, or equipment damage; the corresponding operating characteristics might be a reactor temperature of 750°C, a pressure of 15 kPa, and a flow rate of 50 L / min.

[0061] All causal relationships extracted from historical anomalies are summarized to form a historical anomaly causal clue set. This not only indicates the chronological order of certain anomalies but may also reveal underlying causes. For each anomaly, relevant operating condition data is collected, including parameters such as temperature, pressure, flow rate, and energy consumption. For example, for an anomaly of excessive temperature, operating condition data such as temperature, pressure, fuel flow rate, and oxygen content might be recorded. This data can help analyze the cause of the excessive temperature. The historical anomaly causal clue set is the collection of causal relationships between all identified anomalies. Each causal clue represents one or more pairs of events, revealing the cause of the anomaly. The historical anomaly operating condition status feature set refers to the set of operating condition features associated with each anomaly. These features include parameters such as temperature, pressure, flow rate, and energy consumption. They are used to analyze the operating environment and equipment status at the time of the anomaly, thereby revealing possible causes of the anomaly.

[0062] The historical abnormal operating condition feature set is integrated by merging or classifying operating condition data with similar properties or types. For example, all temperature-related operating condition data is merged into one major category, and pressure data into another. This integration reduces redundant information in the operating condition data and extracts the main features. For each category of integrated operating condition data, a mean is calculated, i.e., averaging the operating condition data within the same category. Mean averaging reduces the impact of individual extreme values ​​on the overall analysis, making the data more stable. For example, for all recorded temperature values, the average of these temperature values ​​is calculated to obtain the mean representing the operating condition data of that category. Through mean averaging, an integrated historical abnormal operating condition feature set is obtained, containing the average operating condition data for each category, which can represent the typical operating condition states under different categories.

[0063] Based on the integrated set of historical abnormal operating condition features, the historical abnormal causal clue set is mapped and integrated. Elements in the two datasets are associated or merged through a mapping relationship (such as numerical range mapping, functional relationship, etc.). The integrated set of historical abnormal operating condition features serves as the foundation for mapping the historical abnormal causal clue set, thus connecting operating condition features with causal clues. The features in the integrated set of historical abnormal operating condition features are connected and mapped with the causal relationships in the historical abnormal causal clue set, forming a more comprehensive information set by combining operating condition features and abnormal causal relationships. The purpose of mapping and integration is to further improve and optimize the knowledge base by combining features and causal relationships. For example, if certain abnormal states (such as excessively high temperature) are always accompanied by a certain causal clue (such as equipment failure), these two data points are associated. The purpose of mapping is to unify operating condition features and causal clues into a single set, forming a new dataset.

[0064] The union of the mapping and integration results is performed, merging all obtained causal clues and removing duplicates to form the final integrated historical anomaly causal clue set. This set represents the possible causal relationships in all historical anomaly events and their corresponding operating condition characteristics. The integrated historical anomaly causal clue set is a new set of causal clues obtained based on mapping and integration and the union, containing the causal relationships of historical anomaly events, and these causal relationships correspond to the processed operating condition characteristics. The final integrated historical anomaly causal clue set and its corresponding integrated historical anomaly operating condition status characteristic set are stored in the causal knowledge base. The causal knowledge base is a database containing all anomaly events and their causal relationships, including information such as anomalies occurring under different operating conditions, their causes, and corresponding operating condition characteristics.

[0065] For example, equipment failure occurs when the temperature is 950℃, the pressure is 45kPa, and the flow rate is 70L / min, the corresponding cause being high temperature; low efficiency occurs when the temperature is 920℃, the pressure is 20kPa, and the flow rate is 80L / min, the corresponding cause being high temperature and abnormal pressure; high energy consumption occurs when the temperature is 900℃, the pressure is 30kPa, and the flow rate is 25L / min, the corresponding cause being high temperature and low flow rate. Extracting causal relationships from historical abnormal event sets reveals that excessively high temperature corresponds to equipment failure, excessively high temperature corresponds to low efficiency, excessively high temperature corresponds to high energy consumption, and excessively high temperature combined with pressure fluctuations corresponds to equipment failure. By constructing a causal knowledge base, a large amount of historical experience and lessons can be accumulated, improving the ability to predict future abnormal events. When similar operating conditions occur in the future, potential anomalies can be quickly identified and corresponding preventative measures can be taken.

[0066] Based on the characteristics of iterative abnormal operating conditions, semantic matching mechanisms (such as vector distance and label matching) are used to search the causal knowledge base to determine the set of associated causal clues. Vector distance refers to converting the abnormal operating condition characteristics into vector form and evaluating their similarity to known operating conditions in the causal knowledge base by calculating the distance between vectors (such as cosine similarity, Euclidean distance, etc.). If the operating condition characteristics have clear label classifications, the label matching mechanism is used to compare them with the labels in the causal knowledge base to find relevant causal relationships. Based on the extracted iterative abnormal operating condition characteristics, the selected matching mechanism is used to search the causal knowledge base for relevant causal clues. For example, if the current state characteristic is excessively high temperature, all causal clues related to excessively high temperature can be found through vector distance matching or label matching mechanisms.

[0067] Based on the matching results, causal clues associated with the current abnormal operating condition characteristics are extracted from the causal knowledge base, resulting in a set of associated causal clues. This helps to understand the possible causes of the anomaly and further assists in developing targeted handling strategies. By introducing vector distance matching and label matching mechanisms, the most relevant historical causal clues to the current abnormal state are accurately identified, thereby improving the accuracy of anomaly detection.

[0068] S400: Traverse the set of related causal clues to perform historical data backtracking and anomaly investigation, and obtain the set of abnormal related causal clues and the set of anomalies.

[0069] Furthermore, this application S400 includes:

[0070] In the historical data backtracking window, the set of related causal clues is traversed to perform historical data backtracking, obtaining a set of historical data sequences of related causal clues; the set of historical data sequences of related causal clues is traversed to identify standard value deviations, obtaining a set of deviation value sequences of related causal clues; based on the set of deviation value sequences of related causal clues, anomaly screening is performed on the set of related causal clues to obtain an abnormal set of related causal clues; from the set of deviation value sequences of related causal clues, the abnormal related causal clue deviation value sequences corresponding to the abnormal set of related causal clues are extracted and the deviation value mean is processed to obtain the set of anomalies.

[0071] Specifically, a historical data backtracking window is defined, which involves selecting a historical time period within the time series and performing backtracking analysis on the data within that period. Based on the defined historical data backtracking window, historical data is analyzed backtracked. The purpose of backtracking analysis is to compare current anomaly causal clues with historical data to gain a deeper understanding of the possible root causes of the anomalies. Based on previously identified causal relationships, the corresponding historical data sequences are reviewed and analyzed to identify past trends and potential problems. Each causal clue may correspond to a specific production process or operation, and the associated historical data sequences may include variable data such as temperature, pressure, and flow rate for that process. Through the backtracking window and associated causal clues, historical data is extracted and a set of associated causal clue historical data sequences is formed. This set contains multiple data items related to each causal clue, forming a time series. The associated causal clue historical data sequence set is extracted from the associated causal clue set and consists of historical data sequences associated with each causal clue. Each data sequence contains historical data that occurred within the backtracking window, which helps in analyzing and identifying the causes of anomalies.

[0072] For the historical data sequence set of causal clues, standard deviation identification is performed. The deviation between the actual operating conditions and the standard operating conditions is calculated, i.e., compared with the preset process operating condition standard value, resulting in a set of deviation value sequences for causal clues. Each deviation value in this set is compared with a preset threshold to determine the number of deviation values ​​greater than or equal to the threshold. If this number exceeds the allowable number, it is considered abnormal and identified as an abnormal causal clue, thus obtaining a set of abnormal causal clues. The preset threshold is a predefined value used to determine whether the deviation value deviates from the normal range. The preset threshold is used to filter out obviously abnormal data. The allowable number is an acceptable threshold used to control how many deviation values ​​are allowed to exceed the normal range. For example, in 10 consecutive measurements, if only one exceeds the threshold, it may be acceptable; however, if the number of exceedances exceeds the allowable number, it is considered an anomaly. For example, if the identified set of deviation values ​​for causal clues includes: 80, 70, 85, 95, 90, with a preset threshold of 80 and an allowed number of deviations of 3, then the set of deviation values ​​greater than 80 exceeds the preset threshold of 3, and is therefore judged as an abnormal causal clue set. The deviation values ​​of the abnormal causal clues are averaged: (80+85+95+90) / 4 = 87.5.

[0073] This process extracts abnormal causal correlation clue deviation value sequences corresponding to abnormal causal correlation clue sets from the set of correlation causal clue deviation value sequences. The deviation values ​​of these abnormal causal correlation clue deviation value sequences are then averaged to calculate the anomaly degree set of these abnormal states, reflecting the severity of the anomaly. Anomaly degree quantifies the severity of abnormal behavior and is typically reflected by averaging the anomaly deviation values; a larger average value indicates a more severe anomaly. Through historical data backtracking, standard value deviation identification, and anomaly screening, potential anomalies are effectively identified, avoiding missed detections. The averaging of deviation values ​​quantifies the severity of anomalies, determining whether emergency measures or long-term optimization measures are needed.

[0074] S500: Based on the set of abnormal correlation causal clues and the set of abnormality degree, identify the feedback control scheme, obtain the target feedback control scheme, generate a feedback adjustment instruction, and send the feedback adjustment instruction to the control execution module to execute the target feedback control scheme.

[0075] Furthermore, this application S500 includes:

[0076] A pre-built feedback control scheme identifier is used to identify the set of abnormal correlation causal clues and the set of abnormality degree to obtain the target feedback control scheme.

[0077] Specifically, the feedback control scheme identifier is a module used to automatically identify and evaluate feedback control strategies. Based on input data (such as sets of anomalous causal clues, sets of anomalousness, etc.), it determines the control scheme to be applied using pre-built rules or algorithms. Simply put, the feedback control scheme identifier can generate suitable control schemes based on anomalous states.

[0078] Collect historical anomaly data, including causal clues, anomaly severity, and system status data. Process the collected anomaly data to ensure clarity and consistency. This typically involves noise removal, missing value imputation, and numerical range standardization. Extract the causal relationships leading to the anomalies from the historical anomaly data. Calculate the severity of each causal clue by analyzing deviations in the historical data, forming an anomaly severity set. Based on historical data, expert experience, or data analysis models, construct decision rules for feedback control schemes. Validate the established model to ensure it can accurately generate appropriate control schemes under different anomaly scenarios.

[0079] The set of causal clues and anomaly degree sets related to anomalies are input into the feedback control scheme identifier to identify the most suitable control scheme. For example, if overload operation is detected accompanied by high temperature deviation, the problem needs to be solved by reducing the load or activating the backup cooling system. The goal of the control scheme is to restore the system to a normal or optimal state by optimizing and adjusting various operations of the system, avoiding excessive energy waste or equipment damage. Once the target feedback control scheme is determined, specific operation instructions are generated based on the target feedback control scheme, such as adjusting parameters, activating equipment, and adjusting processes. The set of causal clues related to anomalies, the set of anomaly degree sets, and the target feedback scheme are stored in the control memory for subsequent decarbonization process control. After receiving the feedback adjustment instructions, the control execution module adjusts the operating state of the equipment according to the instructions to ensure that the feedback control scheme can be accurately executed, thereby effectively repairing the abnormal state. SPARK emphasizes real-time data-driven feedback control, using methods such as causal relationship analysis and anomaly degree identification to trigger appropriate control operations. Through the pre-built feedback control scheme identifier, it automatically adapts to anomalies, adjusts the process flow, and optimizes operating parameters, which is the embodiment of adaptive process adjustment in SPARK. By providing feedback and adjustment instructions, closed-loop optimization can be achieved, deviations in the process flow can be adjusted, and the efficiency and stability of the production process can be ensured.

[0080] The expected effects of implementing the target feedback control scheme through the control execution module are as follows: The typical investment cost of the S module is 5-20 million yuan, with an emission reduction benefit of 12,000-30,000 tons of CO2 per year and an ROI period of 5-8 years; the typical investment cost of the P module is 500,000-3 million yuan, with an emission reduction benefit of 3,000-8,000 tons of CO2 per year and an ROI period of 1-3 years; the typical investment cost of the A+R module is 800,000-5 million yuan, with an emission reduction benefit of 5,000-15,000 tons of CO2 per year and an ROI period of 2-4 years; the typical investment cost of the K module is 200,000-1 million yuan, with an emission reduction benefit of 30% industry-wide replication and efficiency improvement, and an ROI period of 1 year. By constructing a feedback control scheme identifier, reasonable control schemes are automatically generated for abnormal states, reducing manual intervention. Automatic optimization of the control scheme helps reduce energy waste; for example, automatically adjusting equipment load or activating backup systems can reduce energy consumption without sacrificing production efficiency.

[0081] In summary, the AI-driven industrial decarbonization process control method based on feedback regulation provided in this application has the following beneficial effects:

[0082] By performing industrial decarbonization process monitoring, a process operation data stream sequence is obtained. Each process operation data stream includes equipment parameters, energy consumption data, carbon emission values, and environmental disturbance factors. The process operation data stream sequence is traversed to perform iterative analysis of process operating conditions to obtain iterative abnormal operating condition state characteristics. Based on the iterative abnormal operating condition state characteristics, a pre-built causal knowledge base is matched to obtain a set of related causal clues. The set of related causal clues is traversed to perform historical data backtracking and anomaly investigation to obtain an abnormal related causal clue set and an anomaly degree set. Based on the abnormal related causal clue set and the anomaly degree set, a feedback control scheme is identified to obtain a target feedback control scheme, and a feedback adjustment command is generated and sent to the control execution module to execute the target feedback control scheme. In other words, by collecting the process operation data flow sequence, performing time-series iterative analysis of the process operating conditions, extracting the state characteristics of abnormal operation, using a pre-built causal knowledge base to trace the causes of carbon emission anomalies, and by backtracking historical data, associating the causes of anomalies with actual operating conditions, determining the target feedback control scheme to dynamically intervene in the decarbonization process, the precise tracing and efficient control of carbon emission anomalies are achieved, greatly improving the control effect of industrial decarbonization processes.

[0083] Example 2: Based on the same inventive concept as the AI-driven industrial decarbonization process control method based on feedback regulation in Example 1, this application also provides an AI-driven industrial decarbonization process control system based on feedback regulation. Please refer to the appendix. Figure 2 The AI-driven industrial decarbonization process control system based on feedback regulation includes:

[0084] The process monitoring module 11 is used to perform industrial decarbonization process monitoring and obtain a process operation data stream sequence, wherein each process operation data stream includes equipment parameters, energy consumption data, carbon emission values, and environmental disturbance factors; the time series iterative analysis module 12 is used to traverse the process operation data stream sequence to perform time series iterative analysis of process operating conditions and obtain iterative abnormal operating condition state characteristics; the knowledge base matching module 13 is used to match based on the iterative abnormal operating condition state characteristics in a pre-built causal knowledge base to obtain a set of related causal clues; the anomaly investigation module 14 is used to traverse the set of related causal clues to perform historical data backtracking anomaly investigation and obtain an abnormal related causal clue set and an anomaly degree set; the control scheme execution module 15 is used to identify feedback control schemes based on the abnormal related causal clue set and anomaly degree set, obtain a target feedback control scheme, generate feedback adjustment instructions, and send the feedback adjustment instructions to the control execution module to execute the target feedback control scheme.

[0085] Furthermore, the process monitoring module 11 in the AI-driven industrial decarbonization process control system based on feedback regulation is also used for:

[0086] A multimodal sensing array is deployed at key nodes in the target industrial decarbonization process. The multimodal sensing array includes an equipment parameter acquisition unit, an energy consumption metering unit, a carbon emission monitoring device, and an environmental disturbance sensing module. The multimodal sensing array is traversed to perform simultaneous sequential sensing of key nodes, thereby constructing a time-synchronized sequence of process operation data streams.

[0087] Furthermore, the time-series iterative analysis module 12 in the AI-driven industrial decarbonization process control system based on feedback regulation is also used for:

[0088] Extract a first process operation data stream from the process operation data stream sequence; identify the operating status of the first process operation data stream based on a preset process operating condition standard value to obtain a first process operation operating condition state vector; add the preset process operating condition standard value and the first process operation operating condition state vector into a first iterative memory unit; perform a process operating condition time-series iterative analysis on the second process operation data stream extracted from the process operation data stream sequence based on the first iterative memory unit to obtain a second process operation operating condition state vector; update the first iterative memory unit based on the second process operation operating condition state vector to obtain a second iterative memory unit, and so on, perform time-series iterative analysis on the remaining process operation operating condition state vectors in the process operation data stream sequence to obtain a target iterative memory unit; identify abnormal features in the target iterative memory unit to obtain the iterative abnormal operation operating condition state features.

[0089] Furthermore, the time-series iterative analysis module 12 in the AI-driven industrial decarbonization process control system based on feedback regulation is also used for:

[0090] The preset process operating condition standard value is extracted from the first iterative memory unit, and the operating condition status is identified for the second process running data stream to obtain the second initial process running operating condition status vector; the first process running operating condition status vector is extracted from the first iterative memory unit, and the second initial process running operating condition status vector is subjected to time-series iterative enhancement to obtain the second process running operating condition status vector.

[0091] Furthermore, the time-series iterative analysis module 12 in the AI-driven industrial decarbonization process control system based on feedback regulation is also used for:

[0092] Calculate the element similarity between the first process running condition state vector and the second initial process running condition state vector to obtain an element similarity set; perform matrix processing based on the element similarity set to construct a temporal iterative enhancement matrix; use the temporal iterative enhancement matrix to perform convolution enhancement on the second initial process running condition state vector to obtain the second process running condition state vector.

[0093] Furthermore, the knowledge base matching module 13 in the AI-driven industrial decarbonization process control system based on feedback adjustment is also used for:

[0094] Obtain a set of historical abnormal related events for the industrial decarbonization process; traverse the set of historical abnormal related events to extract causal clues, obtaining a set of historical abnormal causal clues and a corresponding set of historical abnormal operating condition status features; integrate the set of historical abnormal causal clues and the corresponding set of historical abnormal operating condition status features to obtain an integrated set of historical abnormal causal clues and a corresponding integrated set of historical abnormal operating condition status features; associate and store the integrated set of historical abnormal causal clues and the corresponding integrated set of historical abnormal operating condition status features to obtain the causal knowledge base.

[0095] Furthermore, the knowledge base matching module 13 in the AI-driven industrial decarbonization process control system based on feedback adjustment is also used for:

[0096] The historical abnormal operating condition status feature set is integrated into a similar category, and the integration results are averaged to obtain an integrated historical abnormal operating condition status feature set; based on the integrated historical abnormal operating condition status feature set, the historical abnormal causal clue set is mapped and integrated, and the union of the mapping and integration results is obtained to obtain the integrated historical abnormal causal clue set.

[0097] Furthermore, the anomaly detection module 14 in the AI-driven industrial decarbonization process control system based on feedback adjustment is also used for:

[0098] In the historical data backtracking window, the set of related causal clues is traversed to perform historical data backtracking, obtaining a set of historical data sequences of related causal clues; the set of historical data sequences of related causal clues is traversed to identify standard value deviations, obtaining a set of deviation value sequences of related causal clues; based on the set of deviation value sequences of related causal clues, anomaly screening is performed on the set of related causal clues to obtain an abnormal set of related causal clues; from the set of deviation value sequences of related causal clues, the abnormal related causal clue deviation value sequences corresponding to the abnormal set of related causal clues are extracted and the deviation value mean is processed to obtain the set of anomalies.

[0099] Furthermore, the control scheme execution module 15 in the AI-driven industrial decarbonization process control system based on feedback adjustment is also used for:

[0100] A pre-built feedback control scheme identifier is used to identify the set of abnormal correlation causal clues and the set of abnormality degree to obtain the target feedback control scheme.

[0101] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Figure 1 The AI-driven industrial decarbonization process control method and specific examples based on feedback regulation in Example 1 are also applicable to the AI-driven industrial decarbonization process control system based on feedback regulation in this example. Through the foregoing detailed description of the AI-driven industrial decarbonization process control method based on feedback regulation, those skilled in the art can clearly understand the AI-driven industrial decarbonization process control system based on feedback regulation in this example. Therefore, for the sake of brevity, it will not be described in detail here.

[0102] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0103] Obviously, those skilled in the art can make several improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. AI-driven industrial decarbonization process control method based on feedback regulation, characterized in that, The method comprises the steps of: performing industrial decarbonization process monitoring to obtain a process operation data stream sequence, wherein each process operation data stream comprises device parameters, energy consumption data, carbon emission values, and environmental disturbance factors; iterative analysis of process operating conditions is performed on the process operation data stream sequence to obtain iterative abnormal operating condition state features; matching is performed on the pre-constructed causal knowledge base based on the iterative abnormal operating condition state features to obtain a set of associated causal clues; abnormal association causal clue sets and abnormal degree sets are obtained by traversing the set of associated causal clues for historical data backtracking and abnormal investigation; a target feedback control scheme is identified based on the abnormal association causal clue sets and the abnormal degree sets, and a feedback adjustment instruction is generated to execute the target feedback control scheme in the control execution module; The method comprises the steps of: deploying a multi-modal perception array at key nodes of the target industrial decarbonization process, wherein the multi-modal perception array comprises a device parameter collector, an energy consumption metering unit, a carbon emission monitoring device, and an environmental disturbance perception module; time-synchronized process operation data stream sequences are constructed by simultaneously sensing the key nodes using the multi-modal perception array; iterative analysis of process operating conditions is performed on the process operation data stream sequence to obtain iterative abnormal operating condition state features, including: extracting a first process operation data stream from the process operation data stream sequence; performing operating condition state identification on the first process operation data stream based on a pre-set process operating condition standard value to obtain a first process operating condition state vector, and adding the pre-set process operating condition standard value and the first process operating condition state vector to a first iteration memory unit; performing iterative analysis of process operating conditions on a second process operation data stream extracted from the process operation data stream sequence based on the first iteration memory unit to obtain a second process operating condition state vector; updating the first iteration memory unit based on the second process operating condition state vector to obtain a second iteration memory unit, and iteratively analyzing the remaining process operating condition state vectors in the process operation data stream sequence in the same manner to obtain a target iteration memory unit; abnormal feature identification is performed on the target iteration memory unit to obtain the iterative abnormal operating condition state features; matching is performed on the pre-constructed causal knowledge base based on the iterative abnormal operating condition state features to obtain a set of associated causal clues, including: obtaining a set of historical abnormal association events of the industrial decarbonization process; extracting causal clues from the set of historical abnormal association events to obtain a set of historical abnormal causal clues and a corresponding set of historical abnormal operating condition state features; integrating the set of historical abnormal causal clues and the corresponding set of historical abnormal operating condition state features to obtain an integrated set of historical abnormal causal clues and an integrated set of corresponding historical abnormal operating condition state features; The integrated historical abnormal causal clue set and the corresponding integrated historical abnormal operating condition state feature set are stored in association to obtain the causal knowledge base.

2. The AI-driven industrial decarbonization process control method based on feedback adjustment of claim 1, wherein, The second process operation data stream extracted from the first iterative memory unit is subjected to process condition time sequence iterative analysis based on the first iterative memory unit to obtain a second process operation condition state vector, including: A preset process condition standard value is extracted from the first iterative memory unit, and the second process operation data stream is subjected to condition state recognition to obtain a second initial process operation condition state vector. A first process operation condition state vector is extracted from the first iterative memory unit, and the second initial process operation condition state vector is subjected to time sequence iterative enhancement to obtain a second process operation condition state vector.

3. The AI-driven industrial decarbonization process control method based on feedback adjustment of claim 2, wherein, A first process operation condition state vector is extracted from the first iterative memory unit, and the second initial process operation condition state vector is subjected to time sequence iterative enhancement to obtain a second process operation condition state vector, including: The element similarity of the first process operation condition state vector and the second initial process operation condition state vector is calculated to obtain an element similarity set. The element similarity set is subjected to matrix processing based on the element similarity set to construct a time sequence iterative enhancement matrix. The second initial process operation condition state vector is subjected to convolution enhancement using the time sequence iterative enhancement matrix to obtain the second process operation condition state vector.

4. The AI-driven industrial decarbonization process control method based on feedback adjustment of claim 1, wherein, The historical abnormal causal clue set and the corresponding historical abnormal operating condition state feature set are integrated to obtain an integrated historical abnormal causal clue set and a corresponding integrated historical abnormal operating condition state feature set, including: The historical abnormal operating condition state feature set is subjected to similar integration, and the integration results are subjected to mean value processing, respectively, to obtain an integrated historical abnormal operating condition state feature set. The historical abnormal causal clue set is subjected to mapping integration based on the integrated historical abnormal operating condition state feature set, and the mapping integration results are subjected to set union to obtain the integrated historical abnormal causal clue set.

5. The AI-driven industrial decarbonization process control method based on feedback adjustment of claim 1, wherein, The associated causal clue set is traversed to perform historical data backtracking abnormality investigation to obtain an abnormal associated causal clue set and an abnormality degree set, including: The associated causal clue set is traversed to perform historical data backtracking to obtain an associated causal clue historical data sequence set in a historical data backtracking window. The associated causal clue historical data sequence set is traversed to perform standard value deviation identification to obtain an associated causal clue deviation value sequence set. The associated causal clue set is obtained by performing abnormality investigation on the associated causal clue set based on the associated causal clue deviation value sequence set. The abnormality degree set is obtained by performing deviation value mean value processing on the abnormal associated causal clue deviation value sequence corresponding to the abnormal associated causal clue set from the associated causal clue deviation value sequence set.

6. The AI-driven industrial decarbonization process control method based on feedback adjustment of claim 1, wherein, A pre-constructed feedback control scheme identifier is used to identify the abnormal associated causal clue set and the abnormality degree set to obtain the target feedback control scheme.

7. AI-driven industrial decarbonization process control system based on feedback regulation, characterized in that, Steps for implementing the AI-driven industrial decarbonization process control method based on feedback regulation according to any one of claims 1-6, the AI-driven industrial decarbonization process control system based on feedback regulation comprising: a process monitoring module for performing industrial decarbonization process monitoring to obtain a sequence of process operation data streams, wherein each process operation data stream includes equipment parameters, energy consumption data, carbon emission values, and environmental disturbance factors; a time series iteration analysis module for performing process working condition time series iteration analysis on the sequence of process operation data streams to obtain iteration abnormal operation working condition state features; a knowledge base matching module for matching the iteration abnormal operation working condition state features in a pre-constructed causal knowledge base to obtain an associated causal clue set; an abnormality investigation module for performing historical data backtracking abnormality investigation on the associated causal clue set to obtain an abnormal associated causal clue set and an abnormality degree set; a control scheme execution module for identifying a target feedback control scheme based on the abnormal associated causal clue set and the abnormality degree set, generating a feedback regulation instruction, and issuing the feedback regulation instruction to a control execution module to execute the target feedback control scheme.

Images