Carbon emission data intelligent monitoring method and system based on digital twinning

Abstract

The application provides a carbon emission data intelligent monitoring method and system based on digital twinning. The method comprises the following steps: acquiring multi-source monitoring data of an enterprise production system, and constructing an energy-carbon flow integrated digital twinning model; predicting carbon emission and intensity of each production line, and positioning high-carbon production lines in combination with real-time data; determining high-carbon units in the high-carbon production lines; performing root cause analysis on the high-carbon units and screening candidate modification schemes; simulating the candidate schemes by using the digital twinning model, determining recommended modification schemes, and issuing the recommended modification schemes for execution; in addition, the carbon market signal is used to dynamically optimize the allocation of carbon quotas and production scheduling in the enterprise, so as to realize closed-loop management and control from passive monitoring to active optimization. The application can realize intelligent monitoring of the whole process of carbon emission, provide a scientific basis for enterprise carbon emission management and control, effectively improve the carbon emission monitoring efficiency and management and control precision, and help enterprises to realize low-carbon production.

Description

Technical Field

[0001] This application relates to the field of carbon emission technology, and more specifically, to a method and system for intelligent monitoring of carbon emission data based on digital twins. Background Technology

[0002] Enterprises have an increasingly urgent need for precise monitoring and efficient management of carbon emissions during the production process. Currently, enterprise carbon emission monitoring largely relies on traditional manual statistics or single-dimensional sensor monitoring, which has several shortcomings: First, monitoring data sources are scattered, and there is a lack of effective linkage between energy consumption, equipment operating conditions, and carbon emission data, making it difficult to form a comprehensive, integrated perspective on carbon flow management. Second, carbon emission prediction largely depends on empirical models, resulting in low accuracy and an inability to predict high-carbon emission risks in advance. Furthermore, the location of high-carbon processes is vague, making it difficult to accurately pinpoint carbon emission weaknesses at the production line and unit level. In addition, there is a lack of scientific simulation and verification methods for modification schemes targeting high-carbon processes. Blindly implementing modifications can easily lead to cost waste and production interruptions. Simultaneously, the lack of coordinated optimization between carbon quota allocation and production scheduling makes it impossible to achieve a balance between carbon emission management and production efficiency. Digital twin technology, as a core technology for achieving precise mapping and real-time linkage between physical and virtual systems, has been widely used in industrial production. However, a comprehensive, intelligent carbon emission monitoring and optimization solution combining digital twins has not yet been formed, making it difficult to meet the needs of enterprises for refined and efficient carbon emission management. Therefore, there is an urgent need for an intelligent carbon emission data monitoring method based on digital twins to solve problems such as inaccurate monitoring, ambiguous high-carbon positioning, lack of verification of transformation schemes, and low management efficiency in existing technologies. Summary of the Invention

[0003] The purpose of this application is to provide a digital twin-based intelligent monitoring method and system for carbon emission data, which overcomes the shortcomings of existing technologies such as low accuracy of enterprise carbon emission monitoring, vague positioning of high-carbon links, lack of scientific verification of transformation schemes, and insufficient control efficiency. It aims to achieve intelligent monitoring of carbon emission data throughout the entire process, accurate positioning of high-carbon links, and scientific optimization of transformation schemes, thereby improving the intelligence and refinement of enterprise carbon emission control and helping enterprises achieve the dual goals of low-carbon production and improved efficiency.

[0004] This application also provides a method for intelligent monitoring of carbon emission data based on digital twins, including the following steps:

[0005] Acquire multi-source monitoring data from enterprise production systems and construct an integrated digital twin model of energy and carbon flow;

[0006] The carbon emissions and carbon intensity of each production line are predicted, and high-carbon production lines are located by combining real-time monitoring of carbon emissions and carbon intensity.

[0007] High-carbon units are determined based on the carbon emissions and carbon emission intensity of each production unit within the high-carbon production line.

[0008] Root cause analysis was conducted on high-carbon units, and candidate modification schemes were screened.

[0009] The candidate modification schemes are simulated using a digital twin model, and the recommended modification scheme is then sent to the production line control system for execution.

[0010] Optionally, in the intelligent monitoring method for carbon emission data based on digital twins described in this application, the step of acquiring multi-source monitoring data from the enterprise's production system and constructing an integrated energy-carbon flow digital twin model includes:

[0011] Acquire multi-source monitoring data from the enterprise's production system, including energy consumption data, equipment operating data, and carbon emission data;

[0012] Based on the physical topology and operational logic of the production system, and driven in real time by multi-source monitoring data, an integrated digital twin model of energy and carbon flow is constructed.

[0013] Optionally, in the intelligent monitoring method for carbon emission data based on digital twins described in this application, the step of predicting the carbon emissions and carbon emission intensity of each production line, and locating high-carbon production lines by combining the real-time monitored carbon emissions and carbon emission intensity, includes:

[0014] Acquire historical multi-source monitoring data for each production line, including energy consumption, production output, production operating conditions data, and corresponding historical carbon emissions;

[0015] The XGBoost model is trained based on historical multi-source monitoring data to obtain the carbon emission prediction model for each production line.

[0016] Acquire production plan settings data for each production line within a preset time period in the future and extract expected output, expected energy consumption and expected production conditions data;

[0017] The expected output, expected energy consumption and expected production conditions are input into the corresponding carbon emission prediction model for processing to obtain the predicted carbon emission values ​​of each production line in the future preset time period.

[0018] The ratio of predicted carbon emissions to projected production is used as the predicted carbon intensity.

[0019] Obtain the real-time carbon emissions and real-time output of each production line, and use the ratio of real-time carbon emissions to real-time output as the current carbon intensity.

[0020] Determine whether the current carbon emissions of the production line exceed the preset carbon emission threshold or whether the current carbon emission intensity exceeds the preset carbon emission intensity threshold;

[0021] Determine whether the predicted carbon emissions of the production line are greater than the preset carbon emission threshold or whether the predicted carbon emission intensity is greater than the preset carbon emission intensity threshold.

[0022] If any of the above judgment conditions is true, an early warning will be triggered based on the integrated energy-carbon flow digital twin model, and the production line will be marked as a high-carbon production line in its visualization interface.

[0023] Optionally, in the intelligent monitoring method for carbon emission data based on digital twins described in this application, the determination of high-carbon units based on the carbon emission amount and carbon emission intensity of each production unit within the high-carbon production line includes:

[0024] Obtain the carbon emissions and carbon emission intensity of each production unit in the high-carbon production line, and plot the carbon emission change curve and carbon emission intensity change curve in the integrated energy-carbon flow digital twin model;

[0025] Obtain the pre-set expected range of carbon emissions and the expected range of carbon emission intensity for each production unit within the high-carbon production line;

[0026] Determine whether the proportion of data points in the carbon emission change curve within a preset time window whose carbon emissions exceed the expected range is greater than a preset threshold.

[0027] Determine whether the proportion of data points in the carbon emission intensity change curve that exceed the expected range of carbon emission intensity within a preset time window is greater than a preset threshold.

[0028] If a production unit meets any of the above conditions, it will be marked as a high-carbon unit and highlighted in the interface of the integrated energy-carbon flow digital twin model.

[0029] Optionally, in the intelligent monitoring method for carbon emission data based on digital twins described in this application, the step of performing root cause analysis on high-carbon units and screening candidate modification schemes includes:

[0030] Acquire multi-dimensional operational data of the high-carbon unit during abnormal periods, including energy consumption, process parameters, equipment operating load, product output, and workshop temperature and humidity;

[0031] By comparing multi-dimensional operational data with multi-dimensional operational data under historical healthy operating conditions, the temporal deviation of each dimension of data is obtained, and the deviation pattern feature data is obtained through processing.

[0032] The deviation pattern feature data is input into a pre-set anomaly diagnosis rule base for matching and identification to obtain one or more potential root causes;

[0033] Potential root causes are input into a pre-set knowledge base of modification schemes for matching and identification, resulting in at least one candidate modification scheme.

[0034] Optionally, in the intelligent monitoring method for carbon emission data based on digital twins described in this application, the step of simulating candidate modification schemes using a digital twin model to determine the recommended modification scheme and then issuing it to the production line control system for execution includes:

[0035] The energy-carbon flow integrated digital twin model was used to simulate the operation of each candidate retrofit scheme to obtain the simulated energy consumption, simulated carbon emissions and simulated carbon emission intensity of the high-carbon unit.

[0036] The emission reduction rate is calculated based on the carbon emissions of high-carbon units and simulated carbon emissions, and the emission intensity reduction rate is calculated based on the carbon emission intensity of high-carbon units and simulated carbon emission intensity.

[0037] The energy consumption reduction rate is obtained by calculating the energy consumption and simulated energy consumption of the high-carbon unit. The carbon efficiency improvement rate corresponding to each candidate retrofit scheme is obtained by weighted averaging the emission reduction rate, emission intensity reduction rate and energy consumption reduction rate.

[0038] The simulated carbon emissions and simulated carbon emission intensity are compared with the preset expected ranges for carbon emissions and carbon emission intensity, respectively. The candidate retrofit scheme with the highest carbon efficiency improvement rate that passes both the comparison of carbon emissions and carbon emission intensity is selected as the recommended retrofit scheme.

[0039] The recommended renovation plan will be pushed to the visualization interface of the integrated energy-carbon flow digital twin model.

[0040] Optionally, the intelligent monitoring method for carbon emission data based on digital twins described in this application further includes:

[0041] Obtain the enterprise's carbon emission allowances and carbon emission trading market prices, as well as the production plans of each production unit in the production system;

[0042] Based on the production plan of each production unit, simulation is carried out using an integrated energy-carbon flow digital twin model to obtain simulated production profits and simulated carbon emissions.

[0043] A reward function is constructed using simulated production profit, carbon purchase cost, and carbon sales revenue as core indicators. A carbon quota allocation optimization model based on reinforcement learning algorithm is then built. The carbon purchase cost and carbon sales revenue are determined by multiplying the difference between the carbon emission quota and the simulated carbon emission amount by the carbon quota price.

[0044] A carbon quota allocation optimization model is trained with the reward function as the optimization objective, and the output is the carbon quota allocation scheme for each production unit that maximizes the reward function value.

[0045] Optionally, the intelligent monitoring method for carbon emission data based on digital twins described in this application further includes:

[0046] Based on the production plan data and carbon quota allocation scheme of each production unit, multiple candidate production scheduling schemes are generated in the integrated energy-carbon flow digital twin model.

[0047] The digital twin model was used to simulate each candidate scheduling scheme to obtain the simulated carbon emissions of each production unit under each scheme;

[0048] The carbon quota compliance rate of a production unit is obtained by calculating the ratio of the number of production units whose simulated carbon emissions do not exceed their allocated carbon quotas to the total number of production units.

[0049] Calculate the ratio of the number of production units that have completed the production plan to the total number of production units to obtain the production plan completion rate;

[0050] The carbon-constrained production completion rate for each candidate scheduling scheme is obtained by weighting the carbon quota compliance rate of the production unit and the production plan completion rate.

[0051] The candidate scheduling scheme with the highest overall completion rate of carbon-constrained production will be output as the recommended production scheduling scheme.

[0052] Secondly, this application provides a digital twin-based intelligent monitoring system for carbon emission data. The system includes a memory and a processor. The memory stores a program for a digital twin-based intelligent monitoring method for carbon emission data. When the program is executed by the processor, it performs the following steps:

[0053] Acquire multi-source monitoring data from enterprise production systems and construct an integrated digital twin model of energy and carbon flow;

[0054] The carbon emissions and carbon intensity of each production line are predicted, and high-carbon production lines are located by combining real-time monitoring of carbon emissions and carbon intensity.

[0055] High-carbon units are determined based on the carbon emissions and carbon emission intensity of each production unit within the high-carbon production line.

[0056] Root cause analysis was conducted on high-carbon units, and candidate modification schemes were screened.

[0057] The candidate modification schemes are simulated using a digital twin model, and the recommended modification scheme is then sent to the production line control system for execution.

[0058] Optionally, in the intelligent monitoring system for carbon emission data based on digital twins described in this application, the step of acquiring multi-source monitoring data from the enterprise's production system and constructing an integrated energy-carbon flow digital twin model includes:

[0059] Acquire multi-source monitoring data from the enterprise's production system, including energy consumption data, equipment operating data, and carbon emission data;

[0060] Based on the physical topology and operational logic of the production system, and driven in real time by multi-source monitoring data, an integrated digital twin model of energy and carbon flow is constructed.

[0061] As can be seen from the above, the intelligent carbon emission data monitoring method and system based on digital twins provided in this application overcomes the shortcomings of existing technologies, such as low accuracy of enterprise carbon emission monitoring, vague positioning of high-carbon links, lack of scientific verification of transformation schemes, and insufficient control efficiency. It realizes intelligent monitoring of carbon emission data throughout the entire process, accurate positioning of high-carbon links, and scientific optimization of transformation schemes, thereby improving the intelligence and refinement of enterprise carbon emission control and helping enterprises achieve the dual goals of low-carbon production and improved efficiency.

[0062] Other features and advantages of this application will be set forth in the following description and will be apparent in part from the description or may be learned by practicing embodiments of this application. The objectives and other advantages of this application may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings. Attached Figure Description

[0063] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0064] Figure 1 A flowchart illustrating the intelligent monitoring method for carbon emission data based on digital twins provided in this application embodiment;

[0065] Figure 2 A flowchart illustrating the location of high-carbon production lines using a digital twin-based intelligent carbon emission data monitoring method provided in this application embodiment;

[0066] Figure 3 A flowchart illustrating the determination of high-carbon units using a digital twin-based intelligent carbon emission data monitoring method provided in this application embodiment. Detailed Implementation

[0067] 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 the embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0068] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0069] Please refer to Figure 1 , Figure 1 This is a flowchart of a digital twin-based intelligent monitoring method for carbon emission data according to some embodiments of this application. This digital twin-based intelligent monitoring method for carbon emission data is used in terminal devices, such as computers and mobile phones. The digital twin-based intelligent monitoring method for carbon emission data includes the following steps:

[0070] S11. Obtain multi-source monitoring data from the enterprise's production system and construct an integrated digital twin model of energy and carbon flow;

[0071] S12. Predict the carbon emissions and carbon emission intensity of each production line, and locate high-carbon production lines by combining the real-time monitored carbon emissions and carbon emission intensity.

[0072] S13. Determine the high-carbon unit based on the carbon emissions and carbon emission intensity of each production unit within the high-carbon production line;

[0073] S14. Conduct root cause analysis on high-carbon units and screen candidate modification schemes;

[0074] S15. Use a digital twin model to simulate the candidate modification schemes, determine the recommended modification scheme, and send it to the production line control system for execution.

[0075] It should be noted that this application, through a full-process design of "model building - high-carbon positioning - root cause analysis - solution simulation - execution closed loop," deeply integrates digital twin technology with carbon emission monitoring, realizing intelligent management and control of enterprise carbon emissions across the entire chain from data collection to solution implementation. By accurately locating high-carbon production lines and providing feasible optimization solutions, it effectively improves the efficiency and accuracy of enterprise carbon emission monitoring, helps enterprises quickly identify carbon emission shortcomings, reduces the cost of low-carbon transformation, and provides systematic methodological support for enterprises to achieve their carbon emission reduction goals.

[0076] According to an embodiment of the present invention, the step of acquiring multi-source monitoring data of an enterprise production system and constructing an integrated energy-carbon flow digital twin model includes:

[0077] Acquire multi-source monitoring data from the enterprise's production system, including energy consumption data, equipment operating data, and carbon emission data;

[0078] Based on the physical topology and operational logic of the production system, and driven in real time by multi-source monitoring data, an integrated digital twin model of energy and carbon flow is constructed.

[0079] It should be noted that acquiring multi-source monitoring data from an enterprise's production system and constructing an integrated digital twin model of energy and carbon flow involves, specifically, first collecting multi-source monitoring data such as energy consumption data, equipment operating condition data, and carbon emission data from the enterprise's production system. Then, based on the physical topology and operational logic of the production system, the aforementioned multi-source monitoring data is used for real-time driving to construct an integrated digital twin model that can accurately map the carbon flow and energy flow of the production system, providing a foundation for subsequent monitoring, prediction, and simulation. Specifically, energy consumption data includes energy type data, real-time and cumulative consumption of various energy sources, and the associated production processes and energy supply parameters corresponding to the time points of energy consumption. Equipment operating condition data specifically includes core operating parameters such as equipment operating load, speed, temperature, pressure, start / stop status, and fault alarm signals. Carbon emission data includes emission quantification data (real-time and cumulative carbon emissions of each production unit and each production process), emission correlation data (carbon emission time points, corresponding emission sources (such as energy combustion emissions, direct emissions from production processes), and associated energy consumption types (such as carbon emissions corresponding to coal combustion, and indirect carbon emissions corresponding to electricity consumption)).

[0080] Specifically, the method for constructing the integrated energy-carbon flow digital twin model is as follows:

[0081] (1) Physical Topology Mapping: The equipment, processes, and pipelines in the enterprise's production system are abstracted into a directed graph structure, where nodes represent equipment or process units, and edges represent energy flow paths and material flow paths. Each node is associated with three types of attributes: energy consumption parameters (including the consumption of various types of energy such as electricity, natural gas, steam, and coal), carbon emission factors (based on the carbon emission factors of the corresponding energy type in the National Greenhouse Gas Inventory Guidelines published by the Intergovernmental Panel on Climate Change), and equipment operating parameters (load rate, operating time, etc.). The attributes of the edges include energy transmission efficiency and transmission loss coefficient. Through the above mapping, the physical topology of the production system is transformed into a computable graph model structure.

[0082] (2) Run logic mapping: Construct input-output energy balance equations and carbon emission calculation equations for each node. For the i-th node, its energy balance equation is: ,in Let be the total input energy of node i at time t. Let be the total output energy of node i at time t. Let be the loss during the transmission and conversion process of node i at time t. ,in Let be the overall energy conversion efficiency of node i. The equation for calculating the carbon emissions of node i at time t is: ,in Let be the consumption of the k-th type of energy at node i at time t. Let be the carbon emission factor of the k-th energy type, and n be the number of energy types. By simultaneously solving the energy balance equations and carbon emission calculation equations for all nodes, a set of model equations covering the entire production system is constructed.

[0083] (3) Unified Mapping of Multi-Source Heterogeneous Data: To address the issue of inconsistent time scales in multi-source monitoring data, a unified timestamp alignment strategy is adopted. The minimum sampling period of the data acquisition system (1 minute in this embodiment) is used as the benchmark time granularity. For data with a sampling frequency higher than the benchmark, arithmetic mean is used for downsampling, and for data with a sampling frequency lower than the benchmark, linear interpolation is used for upsampling. To address the issue of inconsistent spatial scales, the device ID is used as the association primary key. Energy consumption data, equipment operating condition data, and carbon emission data are associated and mapped through the device ID to ensure that each data record can be located to a specific device node.

[0084] (4) Real-time data-driven: The model is driven by a combination of sliding window mean filtering and incremental parameter updates. Specifically, the multi-source monitoring data collected in real time is first subjected to sliding window mean filtering for noise reduction. The sliding window length is set to 5 sampling periods (i.e., 5 minutes) to eliminate the influence of instantaneous fluctuations in the sensors. Then, the incremental parameter update method is used to update the energy conversion efficiency of the nodes in the model. Perform online correction and update the formula as follows: ,in To update the step size, the value is set to 0.1. Let be the efficiency value of node i at time t. Let be the efficiency value of node i at time t-1. Using the above method, the model can continuously adjust its internal parameters based on real-time monitoring data, maintaining synchronization with the physical system state.

[0085] Please refer to Figure 2 , Figure 2 This is a flowchart illustrating the location of high-carbon production lines using a digital twin-based intelligent carbon emission data monitoring method, as described in some embodiments of this application. According to embodiments of the present invention, the step of predicting the carbon emissions and carbon emission intensity of each production line, and combining this with real-time monitored carbon emissions and carbon emission intensity to locate high-carbon production lines, includes:

[0086] S21. Obtain historical multi-source monitoring data for each production line, including energy consumption, production output, production condition data, and corresponding historical carbon emissions.

[0087] S22. Train the preset XGBoost model based on historical multi-source monitoring data to obtain the carbon emission prediction model corresponding to each production line.

[0088] S23. Obtain production plan setting data for each production line within a future preset time period and extract expected output, expected energy consumption and expected production conditions data.

[0089] S24. Input the expected output, expected energy consumption and expected production conditions into the corresponding carbon emission prediction model for processing, and obtain the predicted carbon emission values ​​of each production line in the future preset time period.

[0090] S25. Use the ratio of predicted carbon emissions to projected production as the predicted carbon emission intensity.

[0091] S26. Obtain the real-time carbon emissions and real-time output of each production line, and use the ratio of real-time carbon emissions to real-time output as the current carbon emission intensity.

[0092] S27. Determine whether the current carbon emissions of the production line are greater than the preset carbon emission threshold or whether the current carbon emission intensity is greater than the preset carbon emission intensity threshold.

[0093] S28. Determine whether the predicted carbon emissions of the production line are greater than the preset carbon emission threshold or whether the predicted carbon emission intensity is greater than the preset carbon emission intensity threshold.

[0094] S29. If any of the above judgment conditions is true, then an early warning will be triggered based on the integrated energy-carbon flow digital twin model and the production line will be marked as a high-carbon production line in its visualization interface.

[0095] It should be noted that energy consumption data includes the consumption and power data of various energy sources, and production condition data includes core process parameters and workshop temperature and humidity data. Core process parameters include reaction temperature, reaction pressure, medium flow rate, equipment speed, and valve opening. Workshop temperature and humidity data includes workshop temperature and relative humidity. A pre-set XGBoost (eXtreme Gradient Boosting) model is trained based on historical multi-source monitoring data to obtain carbon emission prediction models for each production line. The input features of the carbon emission prediction model are constructed by feature engineering of historically collected energy consumption, production output, and production condition data. Specifically, these include: raw instantaneous features: directly taken from historically collected raw data, including power, production output, core process parameter values ​​(reaction temperature, reaction pressure, medium flow rate, equipment speed, valve opening), and workshop temperature and humidity (temperature, relative humidity) for each energy-consuming branch; and statistical and trend features: calculated based on a sliding time window for key time-series data such as energy consumption and process parameters, extracting their mean, variance, and slope of change to characterize the dynamic trend of the operating status. By constructing the above multi-dimensional features, multi-source monitoring data is transformed into feature vectors that the model can learn, thereby accurately depicting the mapping relationship between production status and carbon emissions.

[0096] Specifically, the XGBoost model is trained based on historical multi-source monitoring data to obtain a carbon emission prediction model for each production line, including the following steps:

[0097] (1) Dataset partitioning: Historical multi-source monitoring data were partitioned according to time series order. The first 70% of the data was used as the training set, the middle 15% as the validation set, and the last 15% as the test set. Time series partitioning was used instead of random partitioning to avoid future data leakage into the training process and to ensure the authenticity of model evaluation.

[0098] (2) Input Feature Construction: The input features of the carbon emission prediction model are constructed by performing feature engineering on historically collected energy consumption data, production output data, and production condition data. Specifically, they include two types of features:

[0099] (a) Raw instantaneous characteristics: directly taken from historically collected raw data, including power of each energy-consuming branch, production output, core process parameter values ​​(reaction temperature, reaction pressure, medium flow rate, equipment speed, valve opening) and workshop temperature and humidity (temperature, relative humidity).

[0100] (b) Statistical and trend characteristics: Calculations are performed on key time-series data such as energy consumption, production output, and process parameters based on a sliding time window. The sliding window size is set to 15 minutes (corresponding to 15 sampling periods), and the following statistics are extracted: the mean, variance, and slope of change of the data within the window (i.e., the slope of linear regression on the data points within the window) to characterize the dynamic trend of the operating status.

[0101] (3) XGBoost Model Hyperparameter Setting and Optimization: The initial settings of the model hyperparameters are as follows: learning rate is set to 0.1, maximum tree depth is set to 6, number of trees is set to 200, subsampling ratio is set to 0.8, column sampling ratio is set to 0.8, and minimum number of leaf node samples is set to 3. Hyperparameter optimization is performed using a combination of grid search and early stopping strategy: the mean squared error (MSE) on the validation set is used as the evaluation index, and a search is conducted among candidate hyperparameter combinations. When the MSE on the validation set no longer decreases in 20 consecutive iterations, early stopping is triggered, and the hyperparameter combination with the smallest MSE on the validation set is selected as the final model parameters. The objective function is set as squared loss, with the squared error between the predicted and actual carbon emissions as the optimization objective.

[0102] Please refer to Figure 3 , Figure 3 This is a flowchart illustrating the identification of high-carbon units using a digital twin-based intelligent carbon emission data monitoring method, as described in some embodiments of this application. According to embodiments of the present invention, the identification of high-carbon units based on the carbon emission amount and carbon emission intensity of each production unit within a high-carbon production line includes:

[0103] S31. Obtain the carbon emissions and carbon emission intensity of each production unit in the high-carbon production line, and plot the carbon emission change curve and carbon emission intensity change curve in the integrated energy-carbon flow digital twin model.

[0104] S32. Obtain the expected range of carbon emissions and the expected range of carbon emission intensity for each production unit in the high-carbon production line.

[0105] S33. Determine whether the proportion of data points in the carbon emission change curve within the preset time window whose carbon emissions exceed the expected range is greater than a preset threshold.

[0106] S34. Determine whether the proportion of data points in the carbon emission intensity change curve within the preset time window that exceed the expected range of carbon emission intensity is greater than the preset threshold.

[0107] S35. If a production unit meets any of the above conditions, then the production unit will be marked as a high-carbon unit and highlighted in the interface of the integrated energy-carbon flow digital twin model.

[0108] It should be noted that this application achieves precise targeting of high-carbon processes from the production line level to the production unit level. By plotting the changes in carbon emissions and carbon emission intensity, and combining this with the proportion of abnormal data within a preset time window to identify high-carbon units, it can accurately capture dynamic anomalies in carbon emissions of production units, avoiding misjudgments of normal fluctuations. At the same time, by highlighting high-carbon units through a digital twin model, it is easy for staff to quickly locate the specific location of carbon emission shortcomings, solving the problem of vague high-carbon positioning and difficulty in precise policy implementation in traditional methods. This provides a precise target for subsequent root cause analysis and the formulation of transformation plans, further improving the level of precision in carbon emission management.

[0109] According to an embodiment of the present invention, the step of performing root cause analysis and screening candidate modification schemes for high-carbon units includes:

[0110] Acquire multi-dimensional operational data of the high-carbon unit during abnormal periods, including energy consumption, process parameters, equipment operating load, product output, and workshop temperature and humidity;

[0111] By comparing multi-dimensional operational data with multi-dimensional operational data under historical healthy operating conditions, the temporal deviation of each dimension of data is obtained, and the deviation pattern feature data is obtained through processing.

[0112] The deviation pattern feature data is input into a pre-set anomaly diagnosis rule base for matching and identification to obtain one or more potential root causes;

[0113] Potential root causes are input into a pre-set knowledge base of modification schemes for matching and identification, resulting in at least one candidate modification scheme.

[0114] It should be noted that this application achieves accurate diagnosis of the root causes of anomalies in high-carbon units through multi-dimensional operational data comparison and deviation pattern feature extraction (including deviation amplitude, duration, and trend of change), avoiding the problems of strong blindness and low efficiency in traditional root cause analysis. Through a pre-built diagnostic rule base and modification scheme knowledge base, potential root causes and candidate modification schemes can be quickly matched, significantly shortening the time for root cause analysis and scheme selection, and improving work efficiency. The process parameter data includes reaction temperature, reaction pressure, medium flow rate, equipment speed, and valve opening; the workshop temperature and humidity include workshop temperature and relative humidity.

[0115] Specifically, the anomaly diagnosis rule base is constructed as follows: based on the enterprise's historical operational data and the experience of domain experts, common anomaly patterns in high-carbon units are summarized into rules. Each rule includes triggering conditions (composed of deviation thresholds and durations of data from various dimensions) and corresponding root cause labels. The anomaly diagnosis rule base stores several diagnostic rules; the following are exemplary rules:

[0116] Rule 1: If the energy consumption deviation exceeds 15% of the healthy operating condition baseline and lasts for more than 30 minutes, and the temperature deviation exceeds 10% of the baseline, the potential root cause is determined to be "decreased heat exchange efficiency".

[0117] Rule 2: If the deviation of the equipment operating load exceeds 20% of the benchmark value and lasts for more than 20 minutes, and the deviation of the product output exceeds 10% of the benchmark value (i.e., the output drops significantly), the potential root cause is determined to be "inefficient equipment operation or mismatch between load and output".

[0118] Rule 3: If the deviation of a process parameter (including any one of reaction temperature, pressure, or flow rate) exceeds twice the set tolerance of that parameter and lasts for more than 15 minutes, the potential root cause is determined to be "process parameter out of control".

[0119] Rule 4: If multiple deviations occur simultaneously (at least 3 dimensions of data deviation exceed 10% of their respective baseline values) and the duration exceeds 30 minutes, the potential root cause is determined to be "comprehensive equipment failure".

[0120] Rule 5: If the workshop temperature deviation exceeds 20% of the baseline value and the energy consumption deviation exceeds 10%, but the process parameter deviation is within the tolerance range, the potential root cause is determined to be "deterioration of environmental conditions leading to reduced energy efficiency".

[0121] The anomaly diagnosis rule base is stored using a "condition-root cause" mapping structure. When deviation pattern feature data is input into the rule base, the system matches them sequentially according to the rule number. If all the triggering conditions of a rule are met, the corresponding root cause label is output. If multiple rules are matched successfully at the same time, multiple potential root causes are output for subsequent scheme matching.

[0122] The modification solution knowledge base is constructed as follows: For each potential root cause, domain experts, combining equipment technical manuals and historical modification cases, formulate corresponding candidate modification solutions. The modification solution knowledge base stores the mapping relationship between "root causes" and "solutions," as exemplified below:

[0123] Option 1: Targeting the root cause of "decreased heat exchange efficiency", candidate modification solutions include "cleaning the surface of the heat exchanger" and "replacing the heat exchange medium".

[0124] Option 2: For the root cause of "inefficient equipment operation or mismatch between load and output", the candidate modification solutions include "adjusting the production schedule to increase equipment load rate" and "optimizing equipment selection or implementing frequency conversion modification".

[0125] Option 3: For the root cause of "process parameter malfunction", the candidate modification solutions include "recalibrating the process parameter setpoints" and "upgrading the PID control parameters of the control system".

[0126] Option 4: For the root cause "comprehensive equipment failure", the candidate modification solutions include "arranging planned equipment maintenance" and "replacing faulty parts".

[0127] Option 5: In response to the root cause of "deteriorating environmental conditions leading to reduced energy efficiency", candidate improvement solutions include "activating the workshop environmental control system" and "adjusting the production schedule to avoid periods of extreme environmental conditions".

[0128] Once potential root causes are input into the remediation scheme knowledge base, the system retrieves corresponding candidate remediation schemes based on the "root cause-scheme" mapping relationship. If a root cause corresponds to multiple schemes, all of them are output as candidate schemes for subsequent digital twin simulation evaluation.

[0129] According to an embodiment of the present invention, the step of simulating the candidate modification schemes using a digital twin model and determining the recommended modification scheme for execution by issuing it to the production line control system includes:

[0130] The energy-carbon flow integrated digital twin model was used to simulate the operation of each candidate retrofit scheme to obtain the simulated energy consumption, simulated carbon emissions and simulated carbon emission intensity of the high-carbon unit.

[0131] The emission reduction rate is calculated based on the carbon emissions of high-carbon units and simulated carbon emissions, and the emission intensity reduction rate is calculated based on the carbon emission intensity of high-carbon units and simulated carbon emission intensity.

[0132] The energy consumption reduction rate is obtained by calculating the energy consumption and simulated energy consumption of the high-carbon unit. The carbon efficiency improvement rate corresponding to each candidate retrofit scheme is obtained by weighted averaging the emission reduction rate, emission intensity reduction rate and energy consumption reduction rate.

[0133] The simulated carbon emissions and simulated carbon emission intensity are compared with the preset expected ranges for carbon emissions and carbon emission intensity, respectively. The candidate retrofit scheme with the highest carbon efficiency improvement rate that passes both the comparison of carbon emissions and carbon emission intensity is selected as the recommended retrofit scheme.

[0134] The recommended renovation plan will be pushed to the visualization interface of the integrated energy-carbon flow digital twin model.

[0135] It should be noted that this application uses a digital twin model to simulate the candidate retrofit schemes, verifying their feasibility without interrupting actual production. This solves the problems of production interruption and cost waste caused by the blind implementation of traditional retrofit schemes. By calculating indicators such as emission reduction rate, emission intensity reduction rate, and carbon efficiency improvement rate, and comparing simulated emission data with preset thresholds, the optimization effect of each candidate scheme can be quantitatively evaluated, avoiding the problem of unreasonable scheme selection due to subjective judgment. Finally, the scheme with the best carbon efficiency and emission compliance is selected and implemented, realizing the scientific optimization of the retrofit scheme, ensuring the carbon emission reduction effect, and maximizing the economic and social benefits of the enterprise's low-carbon transformation.

[0136] Specifically, the carbon efficiency improvement rate is calculated using the following weighted average formula:

[0137] Carbon efficiency improvement rate = α × emission reduction rate + β × emission intensity reduction rate + γ × energy consumption reduction rate;

[0138] Among them, the emission reduction rate = (carbon emissions before retrofit - simulated carbon emissions) / carbon emissions before retrofit × 100%, the emission intensity reduction rate = (carbon emission intensity before retrofit - simulated carbon emission intensity) / carbon emission intensity before retrofit × 100%, and the energy consumption reduction rate = (energy consumption before retrofit - simulated energy consumption) / energy consumption before retrofit × 100%.

[0139] The weights are as follows: emission reduction rate weight α = 0.5, emission intensity reduction rate weight β = 0.3, and energy consumption reduction rate weight γ = 0.2, with α + β + γ = 1. The weights are determined as follows: the emission reduction rate directly reflects the absolute effect of carbon emission reduction and is the most crucial indicator for evaluating retrofit plans, hence it is assigned the highest weight of 0.5; the emission intensity reduction rate reflects the efficiency of carbon emission improvement per unit of output, reflecting the synergistic relationship between emission reduction and production, and is assigned the second highest weight of 0.3; the energy consumption reduction rate reflects the degree of improvement in energy utilization efficiency and is an indirect driving factor for carbon emission reduction, assigned a weight of 0.2. These weights are default values ​​and can be adjusted according to individual emission reduction targets and energy structure characteristics. The sum of the three weights is 1.

[0140] According to an embodiment of the present invention, it further includes:

[0141] Obtain the enterprise's carbon emission allowances and carbon emission trading market prices, as well as the production plans of each production unit in the production system;

[0142] Based on the production plan of each production unit, simulation is carried out using an integrated energy-carbon flow digital twin model to obtain simulated production profits and simulated carbon emissions.

[0143] A reward function is constructed using simulated production profit, carbon purchase cost, and carbon sales revenue as core indicators. A carbon quota allocation optimization model based on reinforcement learning algorithm is then built. The carbon purchase cost and carbon sales revenue are determined by multiplying the difference between the carbon emission quota and the simulated carbon emission amount by the carbon quota price.

[0144] A carbon quota allocation optimization model is trained with the reward function as the optimization objective, and the output is the carbon quota allocation scheme for each production unit that maximizes the reward function value.

[0145] It should be noted that this application constructs a reward function using simulated production profit, carbon purchase cost, and carbon sales revenue as core indicators, and combines this with a reinforcement learning algorithm (in this embodiment, the Multi-Agent Deep Deterministic Policy Gradient (MADDPG) algorithm) to construct a carbon quota allocation optimization model. Specifically, if the simulated carbon emissions exceed the enterprise's carbon emission quota, a carbon purchase cost is incurred; if the simulated carbon emissions are less than the carbon emission quota, carbon sales revenue is incurred; and if the simulated carbon emissions equal the carbon emission quota, both the carbon purchase cost and carbon sales revenue are zero.

[0146] Specifically, the carbon quota allocation optimization model based on reinforcement learning algorithms employs the MADDPG (Multi-Agent Deep Deterministic Policy Gradient) algorithm, treating each production unit as an agent, with each agent collaboratively optimizing the overall carbon quota allocation scheme for the enterprise. The specific definition of this model is as follows:

[0147] (1) State space S: The state vector of each agent (production unit i) It consists of the following elements: the carbon emissions of this production unit at the current moment. Current output carbon quota balance =Allocated quotas - Consumed quotas, carbon quota price Production plan completion progress = Completed output / Planned output, i.e. The global state S is formed by concatenating the state vectors of all agents.

[0148] (2) Action space A: Actions of each agent The carbon allowance allocation ratio for this production unit, with a value ranging from [0,1], represents the share of the company's total carbon allowance obtained by this unit. The sum of the actions of all agents is normalized to 1 to ensure that the total allowance is fully allocated among the company's units. Each agent uses a deterministic policy network (Actor network) to output continuous action values.

[0149] (3) Reward function The mathematical expression for the reward function is: ;

[0150] in, Simulated production profits obtained from digital twin model simulation; To determine the cost of purchasing carbon credits, when a company's total simulated carbon emissions exceed its total carbon allowance, Where m is the number of production units. Let i be the carbon emissions of production unit i. For the company's total carbon allowance, otherwise =0; To generate revenue from selling carbon credits, when a company's total simulated carbon emissions are less than its total carbon allowances, ,otherwise =0. The default value for the weighting coefficient is: =1.0, =1.0, =0.8. The above weighting means that production profit and carbon purchase cost are treated equally. This reflects the equal impact of profit growth and increased carbon costs on corporate profitability; the weighting of carbon sales revenue is slightly lower. =0.8), reflecting that the realizable value of carbon allowance surpluses is slightly lower than production and operating income. The above weights are default values, and enterprises can adjust them according to their carbon market trading strategies. The discount factor is set to 0.95 to balance current and long-term returns.

[0151] (4) Model Training: The training data comes from the simulation data generated by the integrated energy-carbon flow digital twin model. In each training round, each agent outputs a quota allocation action according to the current policy, and the digital twin model simulates the allocation scheme and calculates the reward function value. An experience replay mechanism is used to store the four-tuple of state, action, reward, and next state of each round of interaction in the replay buffer. The buffer size is set to 10,000 records, and 256 records are randomly selected for batch learning in each training. Both the Actor network and the Critic network use three-layer fully connected neural networks with 128 and 64 hidden layer nodes, respectively, and the ReLU function is used as the activation function. The Critic network is updated by minimizing the temporal difference (TD) error, and the Actor network is updated by policy gradient ascent. The target network uses a soft update method with an update coefficient of 0.01. The exploration noise uses the Ornstein-Uhlenbeck (OU) process with a noise scale of 0.1.

[0152] (5) Convergence criterion: When the rate of change of the reward function value is less than 1% in 100 consecutive training rounds, the model is determined to be converged, training is stopped, and the carbon quota allocation scheme of each production unit that maximizes the reward function value under the current strategy is output.

[0153] (6) Multi-agent cooperation and competition: In the MADDPG framework of this application, there is a game relationship of cooperation and competition among the agents (production units). Cooperation is reflected in the fact that each agent shares the total carbon quota pool of the enterprise and must work together to ensure that the total carbon emissions do not exceed the total quota of the enterprise in order to avoid carbon purchase costs; competition is reflected in the fact that each agent pursues the maximization of its own production profits, forming competition under limited quota resources. The MADDPG algorithm handles the above relationships through a centralized training and distributed execution architecture—in the training phase, each Critic network can observe the state and action information of all agents, thereby learning a strategy that considers the global impact; in the execution phase, each Actor network only allocates actions according to its own state output to achieve distributed decision-making.

[0154] According to an embodiment of the present invention, it further includes:

[0155] Based on the production plan data and carbon quota allocation scheme of each production unit, multiple candidate production scheduling schemes are generated in the integrated energy-carbon flow digital twin model.

[0156] The digital twin model was used to simulate each candidate scheduling scheme to obtain the simulated carbon emissions of each production unit under each scheme;

[0157] The carbon quota compliance rate of a production unit is obtained by calculating the ratio of the number of production units whose simulated carbon emissions do not exceed their allocated carbon quotas to the total number of production units.

[0158] Calculate the ratio of the number of production units that have completed the production plan to the total number of production units to obtain the production plan completion rate;

[0159] The carbon-constrained production completion rate for each candidate scheduling scheme is obtained by weighting the carbon quota compliance rate of the production unit and the production plan completion rate.

[0160] The candidate scheduling scheme with the highest overall completion rate of carbon-constrained production will be output as the recommended production scheduling scheme.

[0161] It should be noted that this application relies on an integrated energy-carbon flow digital twin model, combining production plan data of each production unit with carbon quota allocation schemes to generate multiple candidate production scheduling schemes. By simulating each candidate scheme, the simulated carbon emissions of each production unit are accurately obtained. By calculating the carbon quota compliance rate of each production unit, the degree of fit between the carbon emissions of each unit and the allocated quota is quantified, effectively avoiding the risks associated with exceeding the carbon quota. At the same time, the production plan completion rate ensures the implementation of the company's established production goals, avoiding extreme situations of "emphasizing emission reduction while neglecting production" or "emphasizing production while neglecting emission reduction". Based on the two core indicators, a weighted calculation is performed to obtain the comprehensive completion rate of carbon-constrained production, which can scientifically balance carbon quota compliance requirements and production execution efficiency. By selecting the scheme with the highest comprehensive completion rate as the recommended production scheduling scheme, the rationality and feasibility of the scheduling scheme are ensured, promoting the deep integration of carbon management and production operations, and helping enterprises achieve efficient, low-carbon, and sustainable production operations while ensuring the smooth completion of production plans and achieving carbon quota compliance goals.

[0162] This invention also discloses a digital twin-based intelligent carbon emission data monitoring system, including a memory and a processor. The memory stores a digital twin-based intelligent carbon emission data monitoring method program. When the processor executes the digital twin-based intelligent carbon emission data monitoring method program, it performs the following steps:

[0163] Acquire multi-source monitoring data from enterprise production systems and construct an integrated digital twin model of energy and carbon flow;

[0164] The carbon emissions and carbon intensity of each production line are predicted, and high-carbon production lines are located by combining real-time monitoring of carbon emissions and carbon intensity.

[0165] High-carbon units are determined based on the carbon emissions and carbon emission intensity of each production unit within the high-carbon production line.

[0166] Root cause analysis was conducted on high-carbon units, and candidate modification schemes were screened.

[0167] The candidate modification schemes are simulated using a digital twin model, and the recommended modification scheme is then sent to the production line control system for execution.

[0168] It should be noted that this application, through a full-process design of "model building - high-carbon positioning - root cause analysis - solution simulation - execution closed loop," deeply integrates digital twin technology with carbon emission monitoring, realizing intelligent management and control of enterprise carbon emissions across the entire chain from data collection to solution implementation. By accurately locating high-carbon production lines and providing feasible optimization solutions, it effectively improves the efficiency and accuracy of enterprise carbon emission monitoring, helps enterprises quickly identify carbon emission shortcomings, reduces the cost of low-carbon transformation, and provides systematic methodological support for enterprises to achieve their carbon emission reduction goals.

[0169] According to an embodiment of the present invention, the step of acquiring multi-source monitoring data of an enterprise production system and constructing an integrated energy-carbon flow digital twin model includes:

[0170] Acquire multi-source monitoring data from the enterprise's production system, including energy consumption data, equipment operating data, and carbon emission data;

[0171] Based on the physical topology and operational logic of the production system, and driven in real time by multi-source monitoring data, an integrated digital twin model of energy and carbon flow is constructed.

[0172] It should be noted that acquiring multi-source monitoring data from an enterprise's production system and constructing an integrated digital twin model of energy and carbon flow involves, specifically, first collecting multi-source monitoring data such as energy consumption data, equipment operating condition data, and carbon emission data from the enterprise's production system. Then, based on the physical topology and operational logic of the production system, the aforementioned multi-source monitoring data is used for real-time driving to construct an integrated digital twin model that can accurately map the carbon flow and energy flow of the production system, providing a foundation for subsequent monitoring, prediction, and simulation. Specifically, energy consumption data includes energy type data, real-time and cumulative consumption of various energy sources, and the associated production processes and energy supply parameters corresponding to the time points of energy consumption. Equipment operating condition data specifically includes core operating parameters such as equipment operating load, speed, temperature, pressure, start / stop status, and fault alarm signals. Carbon emission data includes emission quantification data (real-time and cumulative carbon emissions of each production unit and each production process), emission correlation data (carbon emission time points, corresponding emission sources (such as energy combustion emissions, direct emissions from production processes), and associated energy consumption types (such as carbon emissions corresponding to coal combustion, and indirect carbon emissions corresponding to electricity consumption)).

[0173] According to an embodiment of the present invention, the prediction of carbon emissions and carbon emission intensity of each production line, and the location of high-carbon production lines by combining real-time monitoring of carbon emissions and carbon emission intensity, includes:

[0174] Acquire historical multi-source monitoring data for each production line, including energy consumption, production output, production operating conditions data, and corresponding historical carbon emissions;

[0175] The XGBoost model is trained based on historical multi-source monitoring data to obtain the carbon emission prediction model for each production line.

[0176] Acquire production plan settings data for each production line within a preset time period in the future and extract expected output, expected energy consumption and expected production conditions data;

[0177] The expected output, expected energy consumption and expected production conditions are input into the corresponding carbon emission prediction model for processing to obtain the predicted carbon emission values ​​of each production line in the future preset time period.

[0178] The ratio of predicted carbon emissions to projected production is used as the predicted carbon intensity.

[0179] Obtain the real-time carbon emissions and real-time output of each production line, and use the ratio of real-time carbon emissions to real-time output as the current carbon intensity.

[0180] Determine whether the current carbon emissions of the production line exceed the preset carbon emission threshold or whether the current carbon emission intensity exceeds the preset carbon emission intensity threshold;

[0181] Determine whether the predicted carbon emissions of the production line are greater than the preset carbon emission threshold or whether the predicted carbon emission intensity is greater than the preset carbon emission intensity threshold.

[0182] If any of the above judgment conditions is true, an early warning will be triggered based on the integrated energy-carbon flow digital twin model, and the production line will be marked as a high-carbon production line in its visualization interface.

[0183] It should be noted that energy consumption data includes the consumption and power data of various energy sources, and production condition data includes core process parameters and workshop temperature and humidity data. Core process parameters include reaction temperature, reaction pressure, medium flow rate, equipment speed, and valve opening. Workshop temperature and humidity data includes workshop temperature and relative humidity. A pre-set XGBoost (eXtreme Gradient Boosting) model is trained based on historical multi-source monitoring data to obtain carbon emission prediction models for each production line. The input features of the carbon emission prediction model are constructed by feature engineering of historically collected energy consumption, production output, and production condition data. Specifically, these include: raw instantaneous features: directly taken from historically collected raw data, including power, production output, core process parameter values ​​(reaction temperature, reaction pressure, medium flow rate, equipment speed, valve opening), and workshop temperature and humidity (temperature, relative humidity) for each energy-consuming branch; and statistical and trend features: calculated based on a sliding time window for key time-series data such as energy consumption and process parameters, extracting their mean, variance, and slope of change to characterize the dynamic trend of the operating status. By constructing the above multi-dimensional features, multi-source monitoring data is transformed into feature vectors that the model can learn, thereby accurately depicting the mapping relationship between production status and carbon emissions.

[0184] According to an embodiment of the present invention, determining the high-carbon unit based on the carbon emissions and carbon emission intensity of each production unit within the high-carbon production line includes:

[0185] Obtain the carbon emissions and carbon emission intensity of each production unit in the high-carbon production line, and plot the carbon emission change curve and carbon emission intensity change curve in the integrated energy-carbon flow digital twin model;

[0186] Obtain the pre-set expected range of carbon emissions and the expected range of carbon emission intensity for each production unit within the high-carbon production line;

[0187] Determine whether the proportion of data points in the carbon emission change curve within a preset time window whose carbon emissions exceed the expected range is greater than a preset threshold.

[0188] Determine whether the proportion of data points in the carbon emission intensity change curve that exceed the expected range of carbon emission intensity within a preset time window is greater than a preset threshold.

[0189] If a production unit meets any of the above conditions, it will be marked as a high-carbon unit and highlighted in the interface of the integrated energy-carbon flow digital twin model.

[0190] It should be noted that this application achieves precise targeting of high-carbon processes from the production line level to the production unit level. By plotting the changes in carbon emissions and carbon emission intensity, and combining this with the proportion of abnormal data within a preset time window to identify high-carbon units, it can accurately capture dynamic anomalies in carbon emissions of production units, avoiding misjudgments of normal fluctuations. At the same time, by highlighting high-carbon units through a digital twin model, it is easy for staff to quickly locate the specific location of carbon emission shortcomings, solving the problem of vague high-carbon positioning and difficulty in precise policy implementation in traditional methods. This provides a precise target for subsequent root cause analysis and the formulation of transformation plans, further improving the level of precision in carbon emission management.

[0191] According to an embodiment of the present invention, the step of performing root cause analysis and screening candidate modification schemes for high-carbon units includes:

[0192] Acquire multi-dimensional operational data of the high-carbon unit during abnormal periods, including energy consumption, process parameters, equipment operating load, product output, and workshop temperature and humidity;

[0193] By comparing multi-dimensional operational data with multi-dimensional operational data under historical healthy operating conditions, the temporal deviation of each dimension of data is obtained, and the deviation pattern feature data is obtained through processing.

[0194] The deviation pattern feature data is input into a pre-set anomaly diagnosis rule base for matching and identification to obtain one or more potential root causes;

[0195] Potential root causes are input into a pre-set knowledge base of modification schemes for matching and identification, resulting in at least one candidate modification scheme.

[0196] It should be noted that this application achieves accurate diagnosis of the root causes of anomalies in high-carbon units through multi-dimensional operational data comparison and deviation pattern feature extraction. These deviation pattern features include deviation amplitude, duration, and trend, avoiding the problems of blindness and low efficiency in traditional root cause analysis. Through a pre-built diagnostic rule base and modification scheme knowledge base, potential root causes and candidate modification schemes can be quickly matched, significantly shortening the time for root cause analysis and scheme selection, and improving work efficiency.

[0197] According to an embodiment of the present invention, the step of simulating the candidate modification schemes using a digital twin model and determining the recommended modification scheme for execution by issuing it to the production line control system includes:

[0198] The energy-carbon flow integrated digital twin model was used to simulate the operation of each candidate retrofit scheme to obtain the simulated energy consumption, simulated carbon emissions and simulated carbon emission intensity of the high-carbon unit.

[0199] The emission reduction rate is calculated based on the carbon emissions of high-carbon units and simulated carbon emissions, and the emission intensity reduction rate is calculated based on the carbon emission intensity of high-carbon units and simulated carbon emission intensity.

[0200] The energy consumption reduction rate is obtained by calculating the energy consumption and simulated energy consumption of the high-carbon unit. The carbon efficiency improvement rate corresponding to each candidate retrofit scheme is obtained by weighted averaging the emission reduction rate, emission intensity reduction rate and energy consumption reduction rate.

[0201] The simulated carbon emissions and simulated carbon emission intensity are compared with the preset expected ranges for carbon emissions and carbon emission intensity, respectively. The candidate retrofit scheme with the highest carbon efficiency improvement rate that passes both the comparison of carbon emissions and carbon emission intensity is selected as the recommended retrofit scheme.

[0202] The recommended renovation plan will be pushed to the visualization interface of the integrated energy-carbon flow digital twin model.

[0203] It should be noted that this application uses a digital twin model to simulate the candidate retrofit schemes, verifying their feasibility without interrupting actual production. This solves the problems of production interruption and cost waste caused by the blind implementation of traditional retrofit schemes. By calculating indicators such as emission reduction rate, emission intensity reduction rate, and carbon efficiency improvement rate, and comparing simulated emission data with preset thresholds, the optimization effect of each candidate scheme can be quantitatively evaluated, avoiding the problem of unreasonable scheme selection due to subjective judgment. Finally, the scheme with the best carbon efficiency and emission compliance is selected and implemented, realizing the scientific optimization of the retrofit scheme, ensuring the carbon emission reduction effect, and maximizing the economic and social benefits of the enterprise's low-carbon transformation.

[0204] According to an embodiment of the present invention, it further includes:

[0205] Obtain the enterprise's carbon emission allowances and carbon emission trading market prices, as well as the production plans of each production unit in the production system;

[0206] Based on the production plan of each production unit, simulation is carried out using an integrated energy-carbon flow digital twin model to obtain simulated production profits and simulated carbon emissions.

[0207] A reward function is constructed using simulated production profit, carbon purchase cost, and carbon sales revenue as core indicators. A carbon quota allocation optimization model based on reinforcement learning algorithm is then built. The carbon purchase cost and carbon sales revenue are determined by multiplying the difference between the carbon emission quota and the simulated carbon emission amount by the carbon quota price.

[0208] A carbon quota allocation optimization model is trained with the reward function as the optimization objective, and the output is the carbon quota allocation scheme for each production unit that maximizes the reward function value.

[0209] It should be noted that this application constructs a reward function using simulated production profit, carbon purchase cost, and carbon sales revenue as core indicators, and combines this with a reinforcement learning algorithm (in this embodiment, the Multi-Agent Deep Deterministic Policy Gradient (MADDPG) algorithm) to construct a carbon quota allocation optimization model. Specifically, if the simulated carbon emissions exceed the enterprise's carbon emission quota, a carbon purchase cost is incurred; if the simulated carbon emissions are less than the carbon emission quota, carbon sales revenue is incurred; and if the simulated carbon emissions equal the carbon emission quota, both the carbon purchase cost and carbon sales revenue are zero.

[0210] According to an embodiment of the present invention, it further includes:

[0211] Based on the production plan data and carbon quota allocation scheme of each production unit, multiple candidate production scheduling schemes are generated in the integrated energy-carbon flow digital twin model.

[0212] The digital twin model was used to simulate each candidate scheduling scheme to obtain the simulated carbon emissions of each production unit under each scheme;

[0213] The carbon quota compliance rate of a production unit is obtained by calculating the ratio of the number of production units whose simulated carbon emissions do not exceed their allocated carbon quotas to the total number of production units.

[0214] Calculate the ratio of the number of production units that have completed the production plan to the total number of production units to obtain the production plan completion rate;

[0215] The carbon-constrained production completion rate for each candidate scheduling scheme is obtained by weighting the carbon quota compliance rate of the production unit and the production plan completion rate.

[0216] The candidate scheduling scheme with the highest overall completion rate of carbon-constrained production will be output as the recommended production scheduling scheme.

[0217] It should be noted that this application relies on an integrated energy-carbon flow digital twin model, combining production plan data of each production unit with carbon quota allocation schemes to generate multiple candidate production scheduling schemes. By simulating each candidate scheme, the simulated carbon emissions of each production unit are accurately obtained. By calculating the carbon quota compliance rate of each production unit, the degree of fit between the carbon emissions of each unit and the allocated quota is quantified, effectively avoiding the risks associated with exceeding the carbon quota. At the same time, the production plan completion rate ensures the implementation of the company's established production goals, avoiding extreme situations of "emphasizing emission reduction while neglecting production" or "emphasizing production while neglecting emission reduction". Based on the two core indicators, a weighted calculation is performed to obtain the comprehensive completion rate of carbon-constrained production, which can scientifically balance carbon quota compliance requirements and production execution efficiency. By selecting the scheme with the highest comprehensive completion rate as the recommended production scheduling scheme, the rationality and feasibility of the scheduling scheme are ensured, promoting the deep integration of carbon management and production operations, and helping enterprises achieve efficient, low-carbon, and sustainable production operations while ensuring the smooth completion of production plans and achieving carbon quota compliance goals.

[0218] The present invention discloses a digital twin-based intelligent carbon emission data monitoring method and system, which overcomes the shortcomings of existing technologies such as low accuracy of enterprise carbon emission monitoring, vague positioning of high-carbon links, lack of scientific verification of transformation schemes, and insufficient control efficiency. It realizes intelligent monitoring of carbon emission data throughout the entire process, accurate positioning of high-carbon links, and scientific optimization of transformation schemes, thereby improving the intelligence and refinement of enterprise carbon emission control and helping enterprises achieve the dual goals of low-carbon production and improved efficiency.

[0219] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0220] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0221] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

Claims

1. A method for intelligent monitoring of carbon emission data based on digital twins, characterized in that, Includes the following steps: Acquire multi-source monitoring data from enterprise production systems and construct an integrated digital twin model of energy and carbon flow; The carbon emissions and carbon intensity of each production line are predicted, and high-carbon production lines are located by combining real-time monitoring of carbon emissions and carbon intensity. High-carbon units are determined based on the carbon emissions and carbon emission intensity of each production unit within the high-carbon production line. Root cause analysis was conducted on high-carbon units, and candidate modification schemes were screened. The candidate modification schemes are simulated using a digital twin model, and the recommended modification scheme is then sent to the production line control system for execution.

2. The intelligent monitoring method for carbon emission data based on digital twins according to claim 1, characterized in that, The acquisition of multi-source monitoring data from the enterprise's production system and the construction of an integrated energy-carbon flow digital twin model include: Acquire multi-source monitoring data from the enterprise's production system, including energy consumption data, equipment operating data, and carbon emission data; Based on the physical topology and operational logic of the production system, and driven in real time by multi-source monitoring data, an integrated digital twin model of energy and carbon flow is constructed.

3. The intelligent monitoring method for carbon emission data based on digital twins according to claim 2, characterized in that, The prediction of carbon emissions and carbon intensity for each production line, combined with real-time monitoring of carbon emissions and carbon intensity to pinpoint high-carbon production lines, includes: Acquire historical multi-source monitoring data for each production line, including energy consumption, production output, production operating conditions data, and corresponding historical carbon emissions; The XGBoost model is trained based on historical multi-source monitoring data to obtain carbon emission prediction models for each production line. Acquire production plan settings data for each production line within a preset time period in the future and extract expected output, expected energy consumption and expected production conditions data; The expected output, expected energy consumption and expected production conditions are input into the corresponding carbon emission prediction model for processing to obtain the predicted carbon emission values ​​of each production line in the future preset time period. The ratio of predicted carbon emissions to projected production is used as the predicted carbon intensity. Obtain the real-time carbon emissions and real-time output of each production line, and use the ratio of real-time carbon emissions to real-time output as the current carbon intensity. Determine whether the current carbon emissions of the production line exceed the preset carbon emission threshold or whether the current carbon emission intensity exceeds the preset carbon emission intensity threshold; Determine whether the predicted carbon emissions of the production line are greater than the preset carbon emission threshold or whether the predicted carbon emission intensity is greater than the preset carbon emission intensity threshold. If any of the above judgment conditions is true, an early warning will be triggered based on the integrated energy-carbon flow digital twin model, and the production line will be marked as a high-carbon production line in its visualization interface.

4. The intelligent monitoring method for carbon emission data based on digital twins according to claim 3, characterized in that, The determination of high-carbon units based on the carbon emissions and carbon emission intensity of each production unit within the high-carbon production line includes: Obtain the carbon emissions and carbon emission intensity of each production unit in the high-carbon production line, and plot the carbon emission change curve and carbon emission intensity change curve in the integrated energy-carbon flow digital twin model; Obtain the pre-set expected range of carbon emissions and the expected range of carbon emission intensity for each production unit within the high-carbon production line; Determine whether the proportion of data points in the carbon emission change curve within a preset time window whose carbon emissions exceed the expected range is greater than a preset threshold. Determine whether the proportion of data points in the carbon emission intensity change curve that exceed the expected range of carbon emission intensity within a preset time window is greater than a preset threshold. If a production unit meets any of the above conditions, it will be marked as a high-carbon unit and highlighted in the interface of the integrated energy-carbon flow digital twin model.

5. The intelligent monitoring method for carbon emission data based on digital twins according to claim 4, characterized in that, The root cause analysis and screening of candidate modification schemes for high-carbon units includes: Acquire multi-dimensional operational data of the high-carbon unit during abnormal periods, including energy consumption, process parameters, equipment operating load, product output, and workshop temperature and humidity; By comparing multi-dimensional operational data with multi-dimensional operational data under historical healthy operating conditions, the temporal deviation of each dimension of data is obtained, and the deviation pattern feature data is obtained through processing. The deviation pattern feature data is input into a pre-set anomaly diagnosis rule base for matching and identification to obtain one or more potential root causes; Potential root causes are input into a pre-set knowledge base of modification schemes for matching and identification, resulting in at least one candidate modification scheme.

6. The intelligent monitoring method for carbon emission data based on digital twins according to claim 5, characterized in that, The process of simulating candidate modification schemes using a digital twin model to determine the recommended modification scheme and then sending it to the production line control system for execution includes: The energy-carbon flow integrated digital twin model was used to simulate the operation of each candidate retrofit scheme to obtain the simulated energy consumption, simulated carbon emissions and simulated carbon emission intensity of the high-carbon unit. The emission reduction rate is calculated based on the carbon emissions of high-carbon units and simulated carbon emissions, and the emission intensity reduction rate is calculated based on the carbon emission intensity of high-carbon units and simulated carbon emission intensity. The energy consumption reduction rate is obtained by calculating the energy consumption and simulated energy consumption of the high-carbon unit. The carbon efficiency improvement rate corresponding to each candidate retrofit scheme is obtained by weighted averaging the emission reduction rate, emission intensity reduction rate and energy consumption reduction rate. The simulated carbon emissions and simulated carbon emission intensity are compared with the preset expected ranges for carbon emissions and carbon emission intensity, respectively. The candidate retrofit scheme with the highest carbon efficiency improvement rate that passes both the comparison of carbon emissions and carbon emission intensity is selected as the recommended retrofit scheme. The recommended renovation plan will be pushed to the visualization interface of the integrated energy-carbon flow digital twin model.

7. The intelligent monitoring method for carbon emission data based on digital twins according to claim 6, characterized in that, Also includes: Obtain the enterprise's carbon emission allowances and carbon emission trading market prices, as well as the production plans of each production unit in the production system; Based on the production plan of each production unit, simulation is carried out using an integrated energy-carbon flow digital twin model to obtain simulated production profits and simulated carbon emissions. A reward function is constructed using simulated production profit, carbon purchase cost, and carbon sales revenue as core indicators. A carbon quota allocation optimization model based on reinforcement learning algorithm is then built. The carbon purchase cost and carbon sales revenue are determined by multiplying the difference between the carbon emission quota and the simulated carbon emission amount by the carbon quota price. A carbon quota allocation optimization model is trained with the reward function as the optimization objective, and the output is the carbon quota allocation scheme for each production unit that maximizes the reward function value.

8. The intelligent monitoring method for carbon emission data based on digital twins according to claim 7, characterized in that, Also includes: Based on the production plan data and carbon quota allocation scheme of each production unit, multiple candidate production scheduling schemes are generated in the integrated energy-carbon flow digital twin model. The digital twin model was used to simulate each candidate scheduling scheme to obtain the simulated carbon emissions of each production unit under each scheme; The carbon quota compliance rate of a production unit is obtained by calculating the ratio of the number of production units whose simulated carbon emissions do not exceed their allocated carbon quotas to the total number of production units. Calculate the ratio of the number of production units that have completed the production plan to the total number of production units to obtain the production plan completion rate; The carbon-constrained production completion rate for each candidate scheduling scheme is obtained by weighting the carbon quota compliance rate of the production unit and the production plan completion rate. The candidate scheduling scheme with the highest overall completion rate of carbon-constrained production will be output as the recommended production scheduling scheme.

9. A digital twin-based intelligent monitoring system for carbon emission data, characterized in that, The system includes a memory and a processor. The memory stores a program for an intelligent monitoring method for carbon emission data based on digital twins. When the program for the intelligent monitoring method for carbon emission data based on digital twins is executed by the processor, it performs the following steps: Acquire multi-source monitoring data from enterprise production systems and construct an integrated digital twin model of energy and carbon flow; The carbon emissions and carbon intensity of each production line are predicted, and high-carbon production lines are located by combining real-time monitoring of carbon emissions and carbon intensity. High-carbon units are determined based on the carbon emissions and carbon emission intensity of each production unit within the high-carbon production line. Root cause analysis was conducted on high-carbon units, and candidate modification schemes were screened. The candidate modification schemes are simulated using a digital twin model, and the recommended modification scheme is then sent to the production line control system for execution.

10. The intelligent carbon emission data monitoring system based on digital twins according to claim 9, characterized in that, The acquisition of multi-source monitoring data from the enterprise's production system and the construction of an integrated energy-carbon flow digital twin model include: Acquire multi-source monitoring data from the enterprise's production system, including energy consumption data, equipment operating data, and carbon emission data; Based on the physical topology and operational logic of the production system, and driven in real time by multi-source monitoring data, an integrated digital twin model of energy and carbon flow is constructed.

Images