Enterprise carbon emission real-time monitoring system and method

Abstract

The invention relates to the technical field of environment monitoring and data processing, and discloses an enterprise carbon emission real-time monitoring system and method, and the system comprises a data collection module, a data transmission module, a cloud processing center, and a user terminal. A real-time accounting unit in the cloud processing center respectively calculates a first carbon emission amount and a second carbon emission amount by adopting an accounting method and a direct measurement method in parallel aiming at the same emission source which can be subjected to accounting and measurement at the same time; the emission factor dynamic correction module carries out feedback correction on the emission factors which are used by the accounting method and are related to the emission source according to the deviation between the emission factor dynamic correction module and the emission source dynamic correction module; a carbon anomaly source diagnostic module identifies one or more related active data sources that cause an anomaly when monitoring the data anomaly. According to the invention, through combination of dual-path accounting and feedback correction, the accuracy and real-time performance of carbon emission data are improved, rapid and automatic source diagnosis of abnormal events is realized, and reliable data support is provided for fine carbon management of enterprises.

Description

Technical Field

[0001] This invention relates to the field of environmental monitoring and data processing technology, specifically to a real-time monitoring system and method for enterprise carbon emissions. Background Technology

[0002] With increasing global attention to climate change and the proposal of carbon peaking and carbon neutrality goals, accurate and real-time monitoring and management of greenhouse gas emissions from industrial enterprises has become an inherent requirement for their sustainable development.

[0003] Currently, enterprise-level carbon emission accounting primarily relies on the accounting method, which involves collecting enterprise activity data (such as the consumption of various fossil fuels, electricity, and materials) and multiplying it by the corresponding emission factors to calculate carbon emissions. However, this method typically uses general emission factors published by the state or industry. These general factors are averages derived from extensive statistics and cannot reflect the dynamic changes in equipment efficiency, raw material quality, and process parameters in the actual production and operation of a specific enterprise. Therefore, this leads to a systematic deviation between the accounting results and the enterprise's actual carbon emissions, resulting in insufficient accuracy and real-time data.

[0004] To improve data accuracy, some large stationary emission sources install Continuous Flue Gas Monitoring Systems (CEMS) for direct measurement. While this method provides high-precision data, its applicability is typically limited to specific emission outlets and cannot cover all emission sources of an enterprise, such as emissions from purchased electricity, process emissions, or fugitive emissions. More importantly, whether based on periodic reports using accounting methods or monitoring data based on direct measurement, when data shows abnormal fluctuations, current technologies can usually only display the result of changes in total emissions, but cannot automatically and quickly trace and locate the root cause of the anomaly. Management and technical personnel often need to manually check the operating logs and process parameters of multiple production lines and numerous equipment. This process is not only inefficient but also heavily reliant on personal experience, making it difficult to meet the requirements of modern production management for rapid response and accurate decision-making.

[0005] Therefore, there is an urgent need in this field for a new technical solution that can not only improve the accuracy and real-time performance of carbon emission accounting, but also enable automated and rapid location of the source of the anomaly when emission data is abnormal. Summary of the Invention

[0006] The purpose of this invention is to provide a real-time monitoring system and method for enterprise carbon emissions, which solves the problems of insufficient accuracy in the calculation of existing carbon emission monitoring technologies due to their reliance on static emission factors, and the inability to automatically and quickly trace the root cause when carbon emission data is abnormal.

[0007] To achieve the above objectives, the first aspect of the present invention provides a real-time monitoring system for enterprise carbon emissions. The system includes a data acquisition module, a data transmission module, a cloud processing center, and a user terminal.

[0008] The data acquisition module is used to collect multi-source heterogeneous data related to carbon emissions. The cloud processing center is used to receive and process the multi-source heterogeneous data, and includes a real-time calculation unit, an emission factor dynamic correction module, and a carbon anomaly source diagnosis module.

[0009] In one technical solution of the present invention, after receiving multi-source heterogeneous data, the real-time accounting unit first divides the enterprise's carbon emission sources into a first emission source set and a second emission source set according to the system's preset mapping configuration.

[0010] The first set of emission sources refers to those emission sources that are monitored using both accounting methods (based on activity data) and direct measurement methods (such as CEMS systems) in an overlapping manner. For example, the emissions of a critical boiler flue equipped with CEMS can be calculated both by actual measurement by CEMS and by accounting for its fuel consumption (activity data) and emission factors.

[0011] The second set of emission sources refers to those emission sources that can only be accounted for using an accounting method. Examples include purchased electricity consumed by enterprises, process emissions generated during production, and fugitive emissions.

[0012] Based on this division, the real-time accounting unit performs the following calculations in parallel:

[0013] Calculate the first carbon emissions: It is defined as the carbon emissions calculated only for the first set of emission sources, based on their corresponding activity data (e.g., fuel consumption) and using a kernel algorithm (using the current emission factor).

[0014] Calculate the second carbon emissions: This is defined as the carbon emissions calculated only for the first set of emission sources, based on direct measurement methods (e.g., CEMS data).

[0015] Calculate other emissions: Independently, this portion of carbon emissions is calculated based on activity data (e.g., purchased electricity) corresponding to the second set of emission sources and using a kernel algorithm (using their respective emission factors, such as the grid factor).

[0016] The emission factor dynamic correction module receives the first carbon emission and the second carbon emission from the real-time accounting unit.

[0017] Crucially, since both the first and second carbon emissions correspond to the same set of emission sources (i.e., the first emission source set), the deviation between them is a physically meaningful and comparable quantity. This deviation accurately reflects the degree of deviation of the emission factors used in the accounting algorithm (e.g., fuel emission factors) from actual operating conditions, without being affected by the emissions from the second emission source set (e.g., purchased electricity).

[0018] Based on this deviation, the emission factor dynamic correction module uses gradient descent or other iterative algorithms to iteratively correct the emission factors (e.g., fuel or feedstock factors related to the flue) in the kernel algorithm that are only related to the first emission source set.

[0019] In a preferred embodiment of the present invention, in order to prevent the calculation model from being incorrectly corrected due to the failure of the direct measurement method (such as CEMS), the emission factor dynamic correction module further includes a benchmark confidence verification unit.

[0020] The verification unit presets a reasonable threshold (for example, based on historical operating conditions or physicochemical limits, it is determined that the deviation between the verification algorithm and the measured method should not exceed 20%).

[0021] Before executing the correction algorithm, the verification unit first determines whether the absolute value of the current deviation exceeds the reasonableness threshold.

[0022] If the values ​​do not exceed the limit, the CEMS data is deemed reliable, and the system performs emission factor correction according to the aforementioned logic.

[0023] If the deviation exceeds 20% for 10 consecutive data points, it is determined that at least one of the CEMS data or accounting activity data (such as fuel batches) has a significant anomaly, and the gold standard status of CEMS is no longer reliable. The emission factor dynamic correction module will immediately stop emission factor correction (i.e., lock the current factor) and trigger a systemic deviation alarm, prompting maintenance personnel to calibrate the CEMS sensors or verify the activity data source (such as fuel calorific value).

[0024] This two-way verification mechanism avoids model contamination caused by blindly trusting benchmark data.

[0025] The emission factor dynamic correction module does not use this bias to correct emission factors associated with the second emission source set (e.g., grid factors for purchased electricity).

[0026] In one embodiment of the present invention, the total carbon emissions of an enterprise at any given time shall be summed by superimposing the second carbon emissions (as a baseline value verified by the measured method for the first set of emission sources) with the other emissions (calculated independently by the accounting algorithm).

[0027] The carbon anomaly source diagnosis module is connected to the emission factor dynamic correction module. Its function is to achieve real-time monitoring of carbon emission data and root cause analysis of abnormal events. The carbon anomaly source diagnosis module first monitors the carbon emission data. When an anomaly is detected, it analyzes the coupling relationship between the carbon emission data and multi-source heterogeneous data to determine one or more related activity data sources causing the anomaly.

[0028] In a preferred embodiment of the present invention, the coupling relationship analysis includes not only calculating the cross-correlation coefficient of a single variable and screening out all relevant sources that exceed a preset threshold, but also constructing a multiple regression model or using machine learning algorithms (such as random forest, gradient boosting tree, etc.) to identify combinations of factors that are statistically significant for carbon emission anomalies by analyzing the feature importance of the model.

[0029] A second aspect of this invention provides a method for real-time monitoring of enterprise carbon emissions. The method includes the following steps:

[0030] First, collect multi-source heterogeneous data related to carbon emissions.

[0031] Next, within a processing cycle, two calculations are performed in parallel: a first carbon emission is calculated based on a first portion of the multi-source heterogeneous data using a kernel algorithm; and a second carbon emission is calculated based on a second portion of the multi-source heterogeneous data using a direct measurement method.

[0032] Subsequently, based on the deviation between the first carbon emission and the second carbon emission, the emission factor used in the accounting algorithm is corrected so that the subsequently calculated first carbon emission is closer to the second carbon emission.

[0033] While performing the above steps, carbon emission data is continuously monitored. When an anomaly in carbon emission data is detected, the specific source causing the anomaly is determined by analyzing the cross-correlation between the carbon emission data and the multi-source heterogeneous data, and the source information is output.

[0034] In summary, the present invention has at least one of the following beneficial technical effects:

[0035] 1. This invention sets up a real-time calculation unit to execute the accounting algorithm and the direct measurement method in parallel, ensuring that both methods target the same set of overlapping emission sources (such as a specific boiler). Then, a dynamic emission factor correction module continuously compares the results from these two methods for the same source. The emission factor used in the accounting algorithm is corrected based on the discrepancy between the two methods. This design allows the emission factor to adaptively evolve from a static, industry-standard value to a localized dynamic value that highly matches the actual production conditions of the enterprise, thus ensuring that the final output carbon emission data accurately and in real-time reflects the true state of the enterprise's operations.

[0036] 2. The carbon anomaly source diagnosis module of this invention can automatically trigger a diagnostic process when abnormal fluctuations in carbon emission data are detected through statistical methods. This process quantitatively identifies the active data source most strongly correlated with the abnormal fluctuations by performing cross-correlation analysis on the carbon emission sequence within the time window preceding the anomaly. This approach transforms the complex, experience-dependent troubleshooting process into an automated, data-driven attribution analysis process, significantly shortening the anomaly response time.

[0037] 3. This invention integrates high-precision real-time calculation, adaptive model correction, and automated anomaly diagnosis into a closed-loop technical solution, forming a complete management tool from data monitoring to root cause insight. It not only presents carbon emission results but also reveals the key driving factors leading to changes in those results. This provides enterprise managers with direct and reliable data support for optimizing production processes, evaluating the effectiveness of energy-saving measures, and formulating scientific carbon reduction strategies, thereby enhancing the precision of enterprises in the field of carbon emission management. Attached Figure Description

[0038] Figure 1 This is a block diagram of the enterprise carbon emission real-time monitoring system of the present invention;

[0039] Figure 2 This is a schematic diagram of the internal structure of the data acquisition and transmission module of the present invention;

[0040] Figure 3 This is a flowchart of the initialization phase of the method of the present invention;

[0041] Figure 4 This is a flowchart of the stable operation phase of the method of the present invention;

[0042] Figure 5 This is a flowchart of the early warning and response phases of the method of the present invention;

[0043] Figure 6 This is a schematic diagram of the carbon anomaly detection process of the present invention. Detailed Implementation

[0044] The following is in conjunction with the appendix Figure 1 -Appendix Figure 6 The present invention will be further described in detail below.

[0045] This invention provides a real-time monitoring system and method for enterprise carbon emissions.

[0046] See attached document Figure 1 The present invention provides a real-time monitoring system for enterprise carbon emissions, the system comprising: a data acquisition module, a data transmission module, a cloud processing center, and a user terminal.

[0047] The data acquisition module is deployed at the enterprise's production site to collect multi-source heterogeneous data related to carbon emissions in real time and synchronously. This multi-source heterogeneous data includes: energy consumption data acquired from energy consumption sensors, material consumption data acquired from material balance sensors, and flue gas emission data acquired from flue gas emission sensors.

[0048] The data transmission module is connected to the data acquisition module and the cloud processing center, and is used to send the multi-source heterogeneous data acquired by the data acquisition module to the cloud processing center through a specified network transmission protocol.

[0049] The cloud processing center, connected to the data transmission module, is used to receive the multi-source heterogeneous data and perform parallel carbon emission accounting, self-calibration of the accounting model, and carbon anomaly source diagnosis. Specifically, the cloud processing center includes: a real-time accounting unit, a dynamic emission factor correction module, and a carbon anomaly source diagnosis module.

[0050] Before officially executing monitoring tasks at the cloud processing center, the system first performs an initialization configuration procedure to establish an emission source mapping matrix. This matrix logically divides all of the enterprise's carbon emission nodes into two non-overlapping sets:

[0051] Dual-monitoring emission source set: This set includes emission points that have both direct measurement equipment (such as a CEMS system) installed and corresponding activity data (such as fuel consumption) recorded. In the cement clinker production line scenario of this embodiment, this set specifically refers to the rotary kiln flue gas emission outlet. The system maps the CEMS sensor data of this emission outlet as a baseline data source and maps the coal consumption data of the rotary kiln as a source to be calibrated and calculated.

[0052] Single-source emission accounting set: This set contains emission points that can only be accounted for through activity data. In this embodiment, the set includes purchased electricity consumption (corresponding to smart meter data for the entire plant), raw material carbonate decomposition (corresponding to raw material feed amount data), and emissions from in-plant transport vehicles.

[0053] Through this logical division, the cloud processing center can clearly identify which data streams are used for model calibration and which data streams are only used for total accumulation in subsequent processing, thereby achieving decoupling at the algorithm level.

[0054] The real-time calculation unit is used to execute two independent carbon emission calculation paths in parallel. Path one is an emission factor-based calculation method, which calculates the first carbon emission amount based on the received energy consumption data and material consumption data. Path two is a direct measurement method based on sensor measurements, which calculates the second carbon emission amount based on the received flue gas emission data.

[0055] The emission factor dynamic correction module, connected to the real-time calculation unit, receives the first carbon emission and the second carbon emission, calculates the deviation between them, and iteratively updates the emission factors used in the calculation algorithm based on the deviation to generate a set of dynamically localized emission factors. During the monitoring period, the real-time calculation unit processes the data stream in parallel according to a preset mapping matrix.

[0056] For the set of dual-monitored emission sources (i.e., rotary kiln flue gas emission outlets), the real-time accounting unit performs dual-path calculations:

[0057] Path A (Calculated Value): Obtain the real-time coal feed rate of the rotary kiln (activity data), call the current coal emission factor, and calculate the first carbon emission component.

[0058] Path B (Measured Value): Obtain the flue gas velocity and carbon dioxide concentration uploaded by the CEMS system, and calculate the second carbon emission component by integration.

[0059] For a single set of emission sources (i.e., purchased electricity, etc.), the real-time accounting unit independently calculates the third carbon emission component based solely on the readings of smart meters and the average emission factor of the power grid.

[0060] The emission factor dynamic correction module first checks whether the deviation value has exceeded the preset reasonableness threshold (e.g., 20%) for consecutive periods (e.g., N consecutive periods).

[0061] Scenario 1: If the threshold is not exceeded, the system determines that the direct measurement method (CEMS) data is reliable. The emission factor dynamic correction module inputs the deviation value into the PID control algorithm or gradient descent algorithm, and only iteratively updates the coal-fired emission factor to make it closer to the true value.

[0062] Scenario 2: If the threshold is exceeded continuously, the system determines that one of the CEMS sensors or accounting activity data sources (such as sudden changes in coal calorific value) has experienced a major malfunction or change, rendering the CEMS baseline invalid. In this case, the emission factor dynamic correction module immediately stops executing the correction algorithm, locks the coal emission factor at the current value, and pushes a systematic deviation alarm between the CEMS baseline and the accounting model to the user terminal, requiring manual intervention for inspection.

[0063] Ultimately, the company's real-time total carbon emissions are generated by adding the second carbon emission component (measured value, as a baseline) and the third carbon emission component (calculated value, as a supplement), ensuring the integrity and accuracy of the data.

[0064] The carbon anomaly source diagnosis module is connected to the emission factor dynamic correction module and is used to monitor the corrected carbon emission data in real time. When an anomaly in carbon emission data is detected, a diagnostic program is triggered. This program analyzes the cross-correlation between the carbon emission data and the activity data of each source to determine the specific source causing the carbon emission anomaly.

[0065] When the carbon anomaly source diagnosis module detects anomalies in total emissions data, it first performs preliminary location based on the aforementioned source set division:

[0066] If the abnormal fluctuations primarily originate from the second carbon emission component (measured channel), the carbon anomaly source diagnosis module will limit the diagnostic scope to the set of emission sources monitored by both systems. At this point, the carbon anomaly source diagnosis module triggers a multivariate coupling analysis program. This program calculates the correlation coefficient between the emission component sequence and specific process parameter sequences such as coal feeding rate, kiln head temperature, and air supply volume. It then filters out all parameters whose absolute correlation coefficients exceed a preset significance threshold, presenting them as a list of suspected related sources, rather than unilaterally attributing the cause to a single factor.

[0067] If the abnormal fluctuations primarily originate from the third carbon emission component (accounting channel), the carbon anomaly source diagnosis module automatically skips the analysis of the combustion system and instead focuses on a single set of accounting emission sources. The carbon anomaly source diagnosis module analyzes the correlation between the electricity load curves of each workshop and emission fluctuations. If the electricity loads of multiple workshops show a strong correlation (all exceeding the threshold), the system will simultaneously list these workshops and indicate that there may be plant-wide electricity fluctuations or a surge in emissions caused by the coordinated operation of multiple production lines, rather than misjudging it as a single workshop malfunction.

[0068] By employing this hierarchical and multi-factor-supporting diagnostic strategy, the system avoids erroneously investigating rotary kiln combustion faults during sudden changes in power load, and overcomes the limitations of single-cause attribution, greatly improving the efficiency and accuracy of fault location.

[0069] The user terminal is connected to the cloud processing center and is used to receive and display real-time carbon emission data, dynamic localized emission factors, and diagnostic reports on carbon anomaly sources.

[0070] In this embodiment of the invention, the system operates as follows: the data acquisition module continuously collects heterogeneous data from multiple sources and sends it to the cloud processing center via the data transmission module. The real-time processing unit within the cloud processing center receives the data and processes it within a preset time window. Internal parallel computing:

[0071] First carbon emissions The calculation formula is as follows:

[0072] ;

[0073] in, For electricity consumption activity data, The carbon emission factor corresponding to electricity consumption; For fossil fuel consumption activity data, The carbon emission factor corresponding to fossil fuel consumption; For the first Activity data on the consumption of various materials For the first The carbon emission factor corresponding to the consumption of a certain material.

[0074] Second carbon emissions The calculation formula is as follows:

[0075] ;

[0076] in, for The volume concentration of carbon dioxide at time t. for The velocity of the smoke stream at any given moment. The cross-sectional area of ​​the flue is... This is the comprehensive conversion factor.

[0077] The emission factor dynamic correction module receives and First, calculate the deviation between the two. Subsequently, based on the aforementioned deviation The emission factor vector EF is updated using the following formula:

[0078] ;

[0079] in, and The first and the The emission factor vector for each time window, For learning rate, For the first Deviation within a time window, For the first A vector of active data for each time window.

[0080] The carbon anomaly source diagnosis module analyzes the corrected carbon emission data sequence. Monitoring is performed. When an anomaly is detected using control chart algorithms such as Exponentially Weighted Moving Average (EWMA), this module extracts all activity data sequences before and after the anomaly. And calculate the Pearson correlation coefficient between the carbon emission series and each activity data series.

[0081] ;

[0082] in, and These are the means of their respective sequences within the analysis window. The active data source with the largest absolute correlation coefficient. It was identified as the source of the anomaly.

[0083] Ultimately, the cloud processing center will process real-time carbon emission data and updated emission factor vectors. and including the source of the anomaly The diagnostic report is sent to the user's terminal for display.

[0084] See attached document Figure 2 The data acquisition module and the data transmission module work together to acquire and report multi-source heterogeneous data.

[0085] The data acquisition module is specifically implemented by multiple sets of sensors deployed on-site within the enterprise. To ensure the comprehensiveness and accuracy of the data, the selection and deployment of sensors follow this scheme: For energy consumption data, electricity consumption is collected using smart meters that support the Modbus / TCP communication protocol. These meters are deployed at the enterprise's main power inlet and on the independent power supply circuits of key production lines or high-energy-consuming equipment to obtain the total power consumption and individual power consumption. Fossil fuel consumption is collected using thermal gas mass flow meters, which are installed on the main gas pipelines and branch pipelines leading to the main combustion units such as boilers and heating furnaces. These flow meters output a 4-mA standard current signal that is proportional to the mass flow rate.

[0086] For material balance data, the real-time weight of bulk materials stored in silos or bins is obtained by installing resistance strain gauge load cells on their supporting structures; for materials transported by conveyor belts, electronic belt scales are installed below the conveyor belts to obtain the mass of material passing through per unit time. These sensors all output electrical signals directly related to mass.

[0087] To monitor flue gas emissions, a continuous flue gas monitoring subsystem was installed at the location specified in national environmental monitoring standards on the company's main emission duct. This subsystem includes a carbon dioxide analyzer using non-dispersive infrared (NDIR) technology to measure the volumetric concentration of carbon dioxide. An oxygen analyzer based on the zirconia principle is used to measure oxygen concentration for combustion efficiency analysis and emissions calculation; a flue gas velocity meter using either the Pitot tube differential pressure method or the ultrasonic method is used to measure flue gas velocity. ; and temperature and pressure sensors, used to convert operating condition data into standard condition data.

[0088] The data transmission module is specifically implemented as one or more edge computing gateways deployed on the production site. Each edge computing gateway has multiple physical interfaces, including an Ethernet port for connecting smart meters and flue gas monitoring subsystems, an RS-485 serial port for connecting flow meters and weighing sensors, and an analog input (AI) channel.

[0089] The program running on the edge computing gateway performs the following operations: polling or receiving raw data signals from various sensors through its physical interface; parsing and converting data from different protocols (such as Modbus protocol, 4-mA current signal); attaching a high-precision timestamp in a uniform format to all data, and integrating the data into structured data frames, such as JSON format data packets.

[0090] During the data reporting phase, the edge computing gateway uses the Message Queuing Telemetry Transport (MQTT) protocol to publish encapsulated structured data frames to the server address specified by the cloud processing center. To ensure the integrity and confidentiality of data transmission, the communication link between the edge computing gateway and the cloud processing center is established on top of the Transport Layer Security (TLS) protocol, encrypting the entire communication session. Simultaneously, the Quality of Service (QoS) level of the MQTT protocol is set to 2 (ExactlyOnce) to ensure that each data frame is received by the cloud processing center exactly once.

[0091] See attached document Figure 3 The cloud processing center is a computing entity that executes the core method of this invention. It receives data reported by the data transmission module, performs a series of processes, and finally outputs accurate carbon emission data and diagnostic information.

[0092] The implementation of the cloud processing center relies on a set of software programs deployed on a cloud server, and its specific functions are carried out by the following units and modules:

[0093] The real-time accounting unit is used to perform dual-path parallel processing on the received multi-source heterogeneous data. In path one, i.e., the implementation of the accounting algorithm, this unit maintains an emission factor database. This database is populated during system initialization according to publicly available authoritative standards, such as the IPCC National Greenhouse Gas Inventory Guidelines or industry-specific emission accounting standards, pre-setting an initial emission factor for each type of activity data. In path two, i.e., the implementation of the direct measurement method, to ensure the accuracy of the calculation, the comprehensive conversion coefficient... It consists of multiple sub-coefficients. Specifically, The calculation formula is:

[0094] ;

[0095] in, Here is the molar mass of carbon dioxide. This represents the molar volume of the gas under standard conditions. The first part of this formula... Used to convert gas volume into mass. The second and third parts are used to convert actual operating conditions ( The flue gas flow rate under the specified conditions is corrected to the standard condition. The flow rate under ( ), where the actual temperature and actual pressure The data is provided in real time by temperature and pressure sensors in the data acquisition module.

[0096] The core function of the emission factor dynamic correction module is to adaptively calibrate the emission factors in the kernel algorithm using the results of the direct measurement method. This module first calculates the deviation... Used as an error signal to drive the emission factor vector The update algorithm employs gradient descent with an adaptive learning rate to improve convergence efficiency and stability. Specifically, the learning rate... The value of can be determined using adaptive adjustment strategies such as AdaGrad, that is, for each emission factor An independent learning rate is assigned, which dynamically changes as the historical gradient of the factor accumulates, thereby assigning a larger update step size to sparse activity data (infrequent emission activities) and a smaller update step size to frequent activity data.

[0097] To ensure the stable operation of the emission factor dynamic correction module, the module also includes a convergence judgment mechanism and a boundary constraint mechanism. The convergence judgment mechanism calculates the moving average of the absolute value of the deviation. When this average value is less than a preset convergence threshold for N consecutive time windows (N is a preset integer), the convergence is determined. When the system determines that the model has converged, it can pause or reduce the frequency of factor updates. The boundary constraint mechanism applies to each emission factor. Preset a reasonable range of values This range is set based on physicochemical principles or industry experience. After each iteration, if the calculated new factor... If the value exceeds this range, it will be forced to be set to the closest boundary value to prevent the model from overfitting or generating factor values ​​that do not conform to physical reality.

[0098] The carbon anomaly source diagnosis module is implemented in two sequential steps: anomaly detection and source diagnosis. In the anomaly detection step, the module uses an exponentially weighted moving average (EWMA) control chart to analyze the corrected carbon emission sequence. Monitoring is performed. The formulas for calculating the control limits (UCL and LCL) of the EWMA control chart are as follows:

[0099] ;

[0100] in, and , respectively, represent the historical mean and standard deviation of the carbon emission series during the stable operation phase, and L is the coefficient for the control limit width, typically taken as 3. When the EWMA statistic... An abnormal signal is triggered when the value exceeds the UCL or LCL.

[0101] The cloud processing center also includes a data storage and visualization module, which is attached... Figure 3 This module is not shown separately. It provides data persistence services for other functional modules of the cloud processing center (i.e., the real-time accounting unit, the emission factor dynamic correction module, and the carbon anomaly source diagnosis module), and provides data query and display interfaces for the user terminal.

[0102] The data storage and visualization module, in its specific implementation, includes a time-series database and a relational database. The time-series database stores all sequence data with high time resolution, specifically including all raw sensor data uploaded by the data acquisition module, such as instantaneous power. Instantaneous flow rate Material consumption rate Gas concentration etc.; and computational sequence data generated by the real-time accounting unit, including the first carbon emissions. Second carbon emissions .

[0103] The relational database is used to store system configuration information, static parameters, and low-frequency event data. The configuration information includes: unique identifiers, physical locations, measurement ranges, and communication protocol addresses for each sensor in the data acquisition module; and the initial values ​​of each emission factor in the emission factor dynamic correction module. Learning rate and preset factor value boundary constraints The control limit parameter L used for the EWMA algorithm and the correlation threshold used for source determination in the carbon anomaly source diagnosis module.

[0104] The relational database is also used to store the event-type data, specifically including: emission factor vectors generated by the emission factor dynamic correction module after each iteration update. Historical records; and a carbon anomaly diagnosis report generated by the carbon anomaly source diagnosis module after each successful diagnosis, which includes at least the time of the anomaly occurrence. The type of exception and the most relevant source of the exception to be identified. and its correlation coefficient value .

[0105] The data storage and visualization module also includes a data service interface, specifically implemented as a RESTful API based on the HTTP protocol. This interface provides a set of data query endpoints to the user terminal (e.g., a web application running on a PC browser or mobile device), which retrieves the data required for its display by calling these endpoints.

[0106] The user terminal interface is specifically implemented as a visual dashboard, which is configured with multiple data display components to provide a comprehensive view of the system's operating status. These components include a real-time carbon emission curve component, which queries a time-series database in real time through a data service interface to obtain the latest data. and The sequence is used to plot two carbon emission curves in the same coordinate system, and the real-time deviation between the two is calculated and displayed. .

[0107] Factor convergence monitoring component, which queries a relational database. Historical records, and charts showing one or more key emission factors (such as...) The convergence process in which the value of ) changes over time iteration.

[0108] The anomaly diagnosis list component queries the carbon anomaly diagnosis reports stored in a relational database and displays a list of all occurring anomaly events in reverse chronological order. When the user selects any event in the list, the system will retrieve and display the complete diagnosis report details for that event.

[0109] The initialization phase specifically includes the following steps:

[0110] See attached document Figure 4 The implementation process of the method of the present invention first performs an initialization phase before entering the stable operation phase. This phase is used to configure the necessary baseline parameters and initial models for each functional module of the cloud processing center.

[0111] First, the sensors and data channels are configured. System administrators, through the user terminal, input the unique identifier of each sensor in the data acquisition module, its physical deployment location, the specific physical quantity measured (e.g., total power of production line 2), data type, measurement unit, and the corresponding data address or communication point in the edge computing gateway of the data transmission module into the relational database within the cloud processing center. This step establishes a logical connection between physical sensors and specific activity data (e.g.,...) at the cloud processing center's logical level. The mapping relationship between ).

[0112] Next, the initial emission factor library is loaded. The real-time calculation unit accesses a built-in or externally connected standard factor database, reads the initial emission factors corresponding to the configured activity data, and constructs an initial emission factor vector. These initial values, for example , and The initial values ​​are derived from the IPCC National Greenhouse Gas Inventory Guidelines or industry-specific accounting standards. These values ​​are loaded into the configuration table of the relational database and used by the emission factor dynamic correction module. The starting baseline used at time (i.e., the first iteration).

[0113] Subsequently, the algorithm hyperparameters are set. System administrators, through the user terminal, set and save a set of operating parameters for the algorithm module in the cloud processing center:

[0114] For the aforementioned emission factor dynamic correction module, a learning rate is set for the gradient descent algorithm. The deviation threshold used to determine model convergence And for each emission factor Set reasonable boundary constraints .

[0115] For the aforementioned carbon anomaly source diagnosis module, a smoothing coefficient is set for the EWMA control chart. (in And the coefficient L used to calculate the control limit width. Simultaneously, a significance threshold for the Pearson correlation coefficient, used to determine the source of the anomaly, is set. .

[0116] Then, baseline data acquisition is performed. This is because the EWMA algorithm in the carbon anomaly source diagnosis module relies on the historical average of carbon emission data. and historical standard deviation However, the system lacks this historical data upon initial startup. Therefore, the system first enters a baseline data acquisition period after startup. In that During the cycle (e.g., a complete production cycle or 24 hours), the data acquisition module, data transmission module, and real-time accounting unit (especially the direct measurement method of path B) operate normally, continuously generating and storing real-time second carbon emissions. During this period, the emission factor dynamic correction module and the carbon anomaly source diagnosis module remain inactive, meaning that factor correction or anomaly diagnosis is not performed.

[0117] During the reference data acquisition period After completion, the system automatically performs statistical calculations, including all data collected during this period. The data are calculated to have an arithmetic mean, and this value is set as the initial historical mean. Simultaneously calculate the standard deviation of these data and set it as the initial historical standard deviation. .

[0118] Finally, the diagnostic control limits are calculated and activated. The carbon anomaly source diagnostic module retrieves the values ​​calculated in the previous step. , And the preset coefficient L, the initial control upper limit is calculated according to the following formula. and control lower limit

[0119] ;

[0120] ;

[0121] exist and After the calculations are completed and the configuration is saved, the emission factor dynamic correction module and the carbon anomaly source diagnosis module are activated. At this point, the initialization phase is complete, and the method of this invention automatically switches to the stable operation phase.

[0122] See attached document Figure 5After the initialization phase 3.1 is completed, the method of the present invention automatically enters the stable operation phase. This phase is a continuously running, cyclical data processing and model optimization closed loop. In each processing loop of the stable operation phase, firstly, the data acquisition module continuously collects real-time data from all monitoring points according to the sensor configuration set in the initialization phase. The edge computing gateway in the data transmission module parses the collected raw signals, unifies the timestamps and encapsulates the data frames, and sends the structured data frames to the cloud processing center through the data transmission module according to a preset reporting period (e.g., every 1 second) or an event triggering mechanism.

[0123] The real-time computing unit of the cloud processing center continuously receives and buffers incoming data frames. This unit operates within a preset time window. Within (for example, 1 minute), all relevant activity data received by that window (such as...) (To perform cumulative or integral calculations.)

[0124] In each time window At the end of the process, the real-time accounting unit immediately executes dual-path accounting in parallel. Path 1: This unit retrieves the current (i.e., the first) result from the relational database. (a window) of effective emission factor vectors The first carbon emission was calculated based on the accounting formula. Path Two: This unit uses the direct measurement method formula and calls the aforementioned comprehensive conversion coefficient. Calculate the second carbon emissions .

[0125] exist and After all calculations are completed, the real-time calculation unit will combine these two result values ​​with the activity data vector corresponding to the time window. They are then transmitted together to the emission factor dynamic correction module.

[0126] The emission factor dynamic correction module is triggered to perform an iterative update. This module first calculates the deviation. The module then checks... Is the absolute value less than the convergence threshold set during the initialization phase? If not (i.e., not converged), the module then uses the adaptive gradient descent algorithm described above. and Calculate an update increment and generate a new emission factor vector. . After being processed by the aforementioned boundary constraint mechanism, it is written into the relational database, replacing... As the first A new benchmark for kernel algorithm calculations within a time window.

[0127] While adjusting the emission factor, the real-time accounting unit uses the baseline carbon emission data for this time window (e.g., using...) The baseline carbon emission data is sent to the carbon anomaly source diagnosis module. The anomaly detection function of the carbon anomaly source diagnosis module continues to operate. This module receives new baseline carbon emission data. Then, immediately apply the EWMA algorithm formula. Update the current EWMA statistics. .

[0128] The carbon anomaly source diagnosis module is being updated. Then, immediately compare its value with the control limit stored in the configuration. and control lower limit Compare them.

[0129] like The value is in and Between (i.e.) If the current system status is normal, the current cycle ends, and the system waits for the next time window. The data is then processed, and all steps of the stable operation phase described above are repeated.

[0130] like The value exceeded or (Right now or If a carbon emission anomaly is detected, the carbon anomaly source diagnosis module immediately interrupts the current stable operation process and triggers the early warning and response phase.

[0131] See attached document Figure 6 During the stable operation phase, the carbon anomaly source diagnosis module detects the EWMA statistic. The value exceeded the preset control limit. Or control lower limit When the system detects an abnormal carbon emission, it automatically triggers the early warning and response phase.

[0132] In this phase, firstly, the carbon anomaly source diagnosis module immediately identifies the moment the anomaly was triggered. The module then sends a data snapshot extraction command to the data storage and visualization module (specifically, the time-series database). This command specifies a preset time window W (e.g., an abnormal time period). (The previous 60 time steps), and request the carbon emission sequence within that time window W. and all related activity data sequences ( (Index to the data sources for each activity).

[0133] After receiving the data snapshot, the carbon anomaly source diagnosis module performs a diagnostic test on each active data sequence. With carbon emission sequence Within the time window W, cross-correlation analysis is performed. Specifically, this module calculates the Pearson correlation coefficient according to the following formula.

[0134] ;

[0135] in, and These are the arithmetic means of their respective sequences within a time window W. The correlation coefficients of all active data sources are then calculated. Then, the carbon anomaly source diagnosis module executes the source determination logic. This module first determines each... absolute value Compared with the salience threshold set in the initialization phase The module performs a comparison and filters out all that meet the requirements. Activity data source .

[0136] Subsequently, the module sorts these eligible activity data sources according to... Sort the values ​​in descending order and select the highest-ranking active data source. This has been identified as the primary source of the recent carbon emission anomaly.

[0137] The carbon anomaly source diagnosis module then generates a structured carbon anomaly diagnosis report. This report includes at least: the time the anomaly was triggered. , Anomaly type (e.g., exceeding the upper limit UCL), and the primary source of the anomaly to be determined. The name of the source (e.g., power consumption of production line number two) and the corresponding correlation coefficient value. If multiple secondary sources meet the criteria, the report also lists them together in order of their relevance.

[0138] The generated carbon anomaly diagnostic report is sent to the data storage and visualization module and persistently stored in the relational database for historical traceability.

[0139] Simultaneously, the cloud processing center proactively pushes the latest diagnostic report to the connected user terminal via the data service interface. Upon receiving the report, the user terminal issues a warning to the user using a preset alert format (e.g., interface highlighting or pop-up window) and immediately displays the diagnostic source information contained in the report. This diagnostic information provides management personnel with a basis for decision-making regarding subsequent interventions (e.g., equipment maintenance or process parameter adjustments).

[0140] After the early warning and response phase is completed, the system returns to the stable operation phase, clears the abnormal state, and continues the cycle of monitoring, calculation and correction based on the data of the next time window.

Claims

1. A real-time monitoring system for enterprise carbon emissions, characterized in that, include: The data acquisition module is used to collect multi-source heterogeneous data related to carbon emissions; The data transmission module is used to transmit the multi-source heterogeneous data; A cloud processing center is used to receive and process the multi-source heterogeneous data, and the cloud processing center includes: A real-time calculation unit is used to perform calculations and direct measurements simultaneously on an emission source in parallel: The first carbon emission is calculated based on the activity data of the emission source using a kernel algorithm, and the second carbon emission is calculated based on the measured data of the emission source using a direct measurement method. The emission factor dynamic correction module is used to correct the emission factor used in the accounting algorithm that corresponds to the activity data based on the deviation between the first carbon emission and the second carbon emission. The carbon anomaly source diagnosis module is used to determine one or more related activity data sources that cause the anomaly by analyzing the coupling correlation between the carbon emission data and multi-source heterogeneous data when an anomaly in carbon emission data is detected. The user terminal is used to receive and display carbon emission data and sources of anomalies.

2. The enterprise carbon emission real-time monitoring system according to claim 1, characterized in that, The multi-source heterogeneous data includes: Energy consumption data acquired by energy consumption sensors, material consumption data acquired by material balance sensors, and flue gas emission data acquired by flue gas emission sensors.

3. The enterprise carbon emission real-time monitoring system according to claim 2, characterized in that, The real-time accounting unit is specifically used for: The energy consumption data and the material consumption data are used as the first part of the data to calculate the first carbon emission; The flue gas emission data is used as the second part of the data to calculate the second carbon emission.

4. The enterprise carbon emission real-time monitoring system according to claim 1, characterized in that, The emission factor dynamic correction module is specifically used for: Calculate the deviation between the first carbon emission and the second carbon emission; Based on the deviation, the emission factor corresponding to the first part of the data in the kernel algorithm is iteratively updated.

5. The enterprise carbon emission real-time monitoring system according to claim 1, characterized in that, When iteratively updating the emission factor, the emission factor dynamic correction module also uses the activity data in the first part of the data as the basis for updating.

6. The enterprise carbon emission real-time monitoring system according to claim 1, characterized in that, The carbon anomaly source diagnosis module is specifically used for: The carbon emission data, consisting of the second carbon emission amount or the first carbon emission amount corrected by the emission factor, is monitored to detect anomalies. When an anomaly is detected, the carbon emission data and the multi-source heterogeneous data within a preset time window before and after the time of the anomaly are extracted. Calculate the cross-correlation coefficient between the carbon emission data and the activity data in the multi-source heterogeneous data; The active data source with the largest absolute value of the cross-correlation coefficient is identified as the source of the anomaly.

7. The enterprise carbon emission real-time monitoring system according to claim 6, characterized in that, The carbon anomaly source diagnosis module detects anomalies by comparing the statistical values ​​of the carbon emission data with preset control upper and lower limits.

8. The enterprise carbon emission real-time monitoring system according to claim 1, characterized in that, The data acquisition module includes smart meters, gas mass flow meters, electronic belt scales, and a continuous flue gas monitoring subsystem deployed at the enterprise's production site.

9. The enterprise carbon emission real-time monitoring system according to claim 1, characterized in that, The data transmission module is an edge computing gateway, which is used to encrypt and send the multi-source heterogeneous data to the cloud processing center through a message queue telemetry transmission protocol.

10. A method for real-time monitoring of enterprise carbon emissions, characterized in that, Includes the following steps: Collect multi-source heterogeneous data related to carbon emissions; For an emission source that can be both accounted for and directly measured simultaneously, in parallel: The first carbon emission is calculated based on the activity data of the emission source using a kernel algorithm, and the second carbon emission is calculated based on the measured data of the emission source using a direct measurement method. Based on the deviation between the first carbon emission and the second carbon emission, the emission factor used in the accounting algorithm and corresponding to the activity data is corrected; When anomalies in carbon emission data are detected, the coupling correlation between the carbon emission data and multi-source heterogeneous data is analyzed to determine one or more related activity data sources that caused the anomalies, and the source is output.

Images