Carbon emission accounting method and system based on artificial intelligence technology

Abstract

The invention provides a carbon emission accounting method and system based on an artificial intelligence technology, and belongs to the field of carbon emission. According to the technical scheme, the method comprises the following steps: constructing a sensor network of an equipment operation state, obtaining real-time operation data, and preprocessing the real-time operation data through an edge computing node; establishing a dynamic prediction model of emission factor boundary parameters based on the real-time operation data; according to the dynamic prediction model, constructing a carbon emission propagation map, analyzing the spatial and temporal distribution trend of carbon emission intensity, and identifying the position of an abnormal emission source; according to the abnormal emission source position identification result, carbon emission trend prediction is executed, and emission early warning response is triggered in combination with a preset threshold value. The method has the beneficial effects that dynamic prediction of carbon emission factors, identification of emission abnormal sources and trend early warning response are realized, the real-time modeling and space-time propagation analysis capabilities of the system on carbon emission behaviors are enhanced, and the intelligent level and response efficiency of emission management are effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of carbon emissions, and in particular to a carbon emissions accounting method and system based on artificial intelligence technology. Background Art

[0002] As industrial enterprises' carbon emission regulations continue to intensify, traditional carbon emission accounting methods based on fixed emission factors or static models are no longer able to accurately reflect the dynamic changes in emission intensity in actual production. Existing carbon emission accounting systems usually rely on emission factors that are regularly calibrated by equipment or boundary parameters established by empirical rules. They are unable to effectively capture the real-time impact of factors such as equipment status fluctuations, changes in raw material ratios, and process adjustments on emission behavior, resulting in large errors in carbon emission estimation, delayed response, and limited anomaly detection capabilities. On the other hand, complex logistics, energy flows, and exhaust flows exist between multiple devices in industrial systems, and emission anomalies often manifest as a chain reaction from upstream devices to downstream devices. Existing emission monitoring solutions often focus on single-point measurements and local statistics, lacking the ability to comprehensively model the propagation paths and spatial aggregation characteristics of emission behavior within the system structure, making it difficult to accurately identify the location of emission sources and the extent of their impact on other devices. In addition, current carbon emission early warning systems are mostly based on static threshold triggering and lack the ability to predict future emission trends. Alarms are often issued only after actual exceeding of the standard occurs, making it impossible to achieve active intervention and process optimization control. Therefore, there is an urgent need for a system that can achieve dynamic modeling of equipment-level carbon emission factors, accurate identification of abnormal sources, and early warning of future trends, so as to improve the scientific nature, responsiveness and decision-making support capabilities of carbon emission management. Summary of the Invention

[0003] The purpose of the present invention is to provide a carbon emission accounting method and system based on artificial intelligence technology that realizes dynamic prediction of carbon emission factors, identification of abnormal emission sources and trend warning response, enhances the system's real-time modeling and spatiotemporal propagation analysis capabilities of carbon emission behavior, and effectively improves the intelligence level and response efficiency of emission management.

[0004] The present invention is achieved by the following measures: A carbon emissions accounting method based on artificial intelligence technology, characterized by including: Build a sensor network for equipment operating status, obtain real-time operating data, and pre-process the real-time operating data through edge computing nodes; establish a dynamic prediction model for emission factor boundary parameters based on real-time operating data; based on the dynamic prediction model, build a carbon emission propagation map, analyze the temporal and spatial distribution trend of carbon emission intensity and identify the location of abnormal emission sources; based on the abnormal emission source location identification results, perform carbon emission trend prediction, and trigger emission warning response in combination with preset thresholds.

[0005] Specific features of the present invention also include: The real-time operation data includes equipment operation condition parameter data and carbon emission pollutant concentration monitoring data; The equipment operation condition parameter data includes reaction temperature data, reaction pressure data, raw material inlet and outlet flow rate data, raw material ratio data, electric energy consumption data, heat energy input data, equipment start-stop state data, and equipment operation cycle data; The carbon emission pollutant concentration monitoring data includes carbon dioxide concentration data, methane concentration data, carbon monoxide concentration data, nitrogen oxide concentration data, volatile organic compound concentration data, and particulate matter concentration data.

[0006] The preprocessing includes data integrity verification, standardization processing, time series alignment, and formatting conversion operations on the real-time operation data.

[0007] The establishment of the dynamic prediction model for emission factor boundary parameters includes: constructing a continuous time series input based on the equipment operation condition parameter data and carbon emission pollutant concentration monitoring data, dividing the continuous time series input into sliding time windows of a fixed length, and inputting them into a dynamic prediction model composed of a time series modeling unit and a boundary inference unit; The time series modeling unit is used to extract the dynamic correlation features between the operation state and the emission factor, encode the input sequence using a long short-term memory neural network, and generate the emission factor prediction value corresponding to the current window; The boundary inference unit calculates the credible upper and lower boundaries of the current prediction value based on the encoding of the time series modeling unit and the historical fluctuation statistical information, and outputs the emission factor boundary parameters; The emission factor boundary parameters include the emission factor prediction value, the upper bound parameter, and the lower bound parameter, which are used to define the reasonable range of the emission factor under the current operating conditions.

[0008] The construction of the carbon emission propagation map includes: using the emission factor boundary parameters generated by the dynamic prediction model and the real-time operation data as inputs; Taking equipment units as graph nodes within a set time window, with edges representing transmission paths, configuring an attribute vector containing the emission factor prediction value, operation state vector, and historical emission records for each node, and configuring attribute parameters representing emission correlation for the edges; Modeling a directed graph structure through a graph neural network, calculating the propagation path of carbon emission intensity in the directed graph structure using the connection relationship between nodes and the attribute vector of historical emission records, and judging whether the emission level of each node exceeds the emission factor boundary parameters based on the threshold range obtained from training with historical data, and marking the over-limit nodes as candidate positions for abnormal emission sources within a given time window.

[0009] The analysis of the spatiotemporal distribution trend of carbon emission intensity includes: within a sliding time window, based on the temporal emission trajectory of abnormal nodes and the spatial connection relationship in the graph, constructing a spatiotemporal mapping structure of carbon emission intensity to express the distribution trend and propagation evolution path of emission anomalies.

[0010] The results of the abnormal emission source location identification include: extracting the carbon emission factor prediction value of each node in the carbon emission propagation map and comparing it with the corresponding boundary parameters. If the carbon emission factor prediction value exceeds the upper and lower limits of the corresponding boundary parameters, the node is marked as an abnormal node; Based on the edge connectivity, emission propagation intensity, material transmission rate, and energy correlation parameters in the carbon emission propagation graph, the degree of emission correlation between abnormal nodes and other nodes and the location of abnormal emission sources in the directed graph structure are analyzed.

[0011] The method of performing carbon emission trend prediction and triggering an emission warning response based on a preset threshold includes: extracting the corresponding carbon emission factor prediction value, operating status data, and spatiotemporal distribution trend within the corresponding time window based on the identified abnormal emission source location, and constructing an emission intensity change trajectory of the device node in chronological order; Based on the connection relationship of the device nodes in the carbon emission propagation map, the emission sequence within the time window is updated; Trend extension processing is performed on the basis of the updated time series to generate the carbon emission factor change trajectory of the node in subsequent multiple time steps, and the predicted results are compared with the preset emission threshold. When the predicted value exceeds the threshold, an emission warning response is triggered.

[0012] A carbon emissions accounting system based on artificial intelligence technology, characterized by: Data processing module: builds a sensor network for equipment operation status, obtains real-time operation data, and pre-processes the real-time operation data through edge computing nodes; dynamic prediction model module: establishes a dynamic prediction model for emission factor boundary parameters based on real-time operation data; analysis module: based on the dynamic prediction model, builds a carbon emission propagation map, analyzes the temporal and spatial distribution trend of carbon emission intensity and identifies the location of abnormal emission sources; early warning module: performs carbon emission trend prediction based on the abnormal emission source location identification results, and triggers emission early warning response in combination with preset thresholds.

[0013] The beneficial effects of the present invention are as follows: the present invention overcomes the problems of static emission factors, limited anomaly identification, and delayed early warning responses in existing carbon emission accounting methods, and has significant technological advancement and practical application value. By establishing data collection on equipment operating status and implementing data standardization and time series alignment processing at edge computing nodes, the present invention ensures the integrity of input data in terms of structure, time, and physical properties, laying a high-quality foundation for model training and prediction. Furthermore, by introducing a dynamic prediction model that combines temporal modeling and boundary inference mechanisms, the estimation of emission factors can respond in real time to changes in the operating state of equipment, solving the problem that traditional emission factors are difficult to reflect dynamic operating condition differences, and improving the accuracy and adaptive ability of emission accounting; At the same time, the present invention also first introduces emission behavior into the graph structure modeling paradigm, maps the connection relationships between equipment nodes and their material flow and energy flow in the industrial process into a carbon emission propagation map, realizes the structured expression of emission information in the spatio-temporal dimension, and conducts propagation feature learning based on graph neural networks. It can not only identify local emission anomaly nodes, but also combine the connection path, propagation intensity and time continuity of the nodes in the graph to realize the system-level identification and positioning of abnormal emission sources, significantly enhancing the causal interpretation ability and spatial identification accuracy of the model.

[0014] Based on the anomaly identification, the present invention further introduces a trend prediction mechanism. Based on historical emission trajectories and equipment states, an emission intensity prediction sequence for a future time period is constructed and compared with a set threshold, so as to issue an early warning response before the emission level exceeds the regulatory red line, with significant forward-looking, responsiveness and operability. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is a carbon emission accounting method based on artificial intelligence technology provided in Embodiment 1 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0016] To clearly illustrate the technical features of the present solution, the present solution will be described below through specific embodiments.

[0017] Embodiment 1 See Figure 1 , a carbon emission accounting method based on artificial intelligence technology, characterized by comprising: S1. Construct a sensing network for the operating state of equipment, obtain real-time operating data, and preprocess the real-time operating data through edge computing nodes; Collect real-time operating data through sensors arranged at key nodes of the equipment; The real-time operating data includes equipment operating condition parameter data and carbon emission pollutant concentration monitoring data; The equipment operating condition parameter data includes reaction temperature data, reaction pressure data, raw material inlet and outlet flow rate data, raw material ratio data, electric energy consumption data, heat energy input data, equipment start-stop state data and equipment operating cycle data; The carbon emission pollutant concentration monitoring data includes carbon dioxide concentration data, methane concentration data, carbon monoxide concentration data, nitrogen oxide concentration data, volatile organic compound concentration data and particulate matter concentration data.

[0018] The preprocessing includes data integrity verification, standardization processing, time series alignment, and formatting conversion operations on real-time operation data; The edge computing node performs time series synchronization processing, field normalization conversion, and data missing value filling on the data, and uploads the processed data to the central processing platform in a structured format for model training. The normalization methods include Z-score standardization and maximum-minimum normalization methods. Data synchronization uses a unified reference clock to perform sliding window alignment processing on data sampled at different time series to ensure that all input data fields have time consistency and structural compatibility; Among them, data integrity verification is used to identify missing items, invalid values, and mutation outliers, and mark, eliminate, or repair them according to preset data quality rules; Standardization processing normalizes or adjusts the distribution of the original field values based on the data type to unify the input dimension range and numerical distribution; Time series alignment includes mapping the data from asynchronous sampling devices through a unified time reference to construct a consistent data time axis; Formatting conversion is used to encapsulate the preprocessed data into a model input format with consistent structure and continuous time series to meet the requirements of the subsequent carbon emission accounting model for the data input structure; The further preprocessing specifically includes: First, use the edge computing node to perform integrity checks on the received original sensing data, identify missing fields, invalid data, and abnormal deviation values, and judge whether each data field is complete according to the preset data quality rules. If the proportion of missing fields exceeds the set threshold (such as 10%), it is marked as an unavailable data segment and excluded from the training set. If the missing is within the tolerable range, it is filled by combining linear interpolation and sliding window mean. The size of the sliding window is dynamically set between 5 and 15 time periods according to the actual data volatility, and sliding weighted interpolation is used for filling; Subsequently, standardization processing is performed on the data of each field, and the normalization strategy is selected according to the physical type to which the field belongs; Fields such as temperature, pressure, and energy consumption are processed using the Z-score standardization method, that is, each record is converted by dividing the mean difference by the standard deviation. Emission concentration fields use the maximum-minimum normalization method to compress the values into the [0,1] interval to unify the numerical range and distribution structure of the input data; After standardization, time series alignment operations are performed on the multi-channel asynchronous sampling data. A unified time reference (such as a central time server) is used to synchronize and map the data of each channel, and a standard time axis is constructed for unifying the window structure. At the same time, the data at different sampling points are mapped to a unified time node through a sliding interpolation mechanism to ensure that all fields have valid values in each time segment; Time series alignment, uniformly mapping asynchronous sampling data to a standard time axis, and finally organizing the processed data into a tensor input format with a unified structure and continuous time series. The system encapsulates the preprocessed data of consecutive time points within a sliding window into a unified tensor format; S2. Establish a dynamic prediction model for emission factor boundary parameters based on real-time operation data; The establishment of the dynamic prediction model for emission factor boundary parameters includes: constructing a continuous time series input based on device operating condition parameter data and carbon emission pollutant concentration monitoring data, dividing the continuous time series input into sliding time windows of a fixed length, and inputting them into a dynamic prediction model composed of a time series modeling unit and a boundary inference unit; The time series modeling unit is used to extract the dynamic correlation features between the operating state and the emission factor, encode the input sequence using a long short-term memory neural network, and generate the predicted emission factor value corresponding to the current window:

[0019] is the current time step, is the sliding window length, is the historical time index within the window, represents time the predicted emission factor value, is the reaction temperature of the device at time moment, is the pressure, is the raw material flow rate, is the raw material ratio, is the power consumption, is the CO2 concentration, 、 are empirical coefficients related to thermodynamics; The boundary inference unit calculates the credible upper and lower boundaries of the current predicted value based on the encoding of the time series modeling unit and historical fluctuation statistical information, and outputs the emission factor boundary parameters:

[0020] represents the dynamic standard deviation boundary estimated based on historical errors, is the prediction upper boundary, is the prediction lower boundary; The emission factor boundary parameters include the predicted emission factor value, the upper bound parameter, and the lower bound parameter, which are used to define the reasonable range of the emission factor under the current operating conditions; After receiving the real-time operation status data and carbon emission monitoring data of each new time step, the dynamic prediction model of emission factor boundary parameters automatically updates the content of the sliding time window, and re-executes the emission factor prediction and boundary parameter calculation based on the updated sequence to achieve continuous dynamic update of the carbon emission factor boundary parameters; S3. According to the dynamic prediction model, construct a carbon emission propagation map, analyze the spatio-temporal distribution trend of carbon emission intensity, and identify the locations of abnormal emission sources; The construction of the carbon emission propagation map includes: using the emission factor boundary parameters and real-time operation data generated by the dynamic prediction model as inputs; Taking equipment units as graph nodes within a set time window, with edges representing transmission paths, configuring an attribute vector containing the predicted emission factor value, operation status vector, and historical emission records for each node, and configuring attribute parameters representing emission correlation for the edges; Model a directed graph structure through a graph neural network, calculate the propagation path of carbon emission intensity in the directed graph structure using the connection relationship between nodes and the attribute vector of historical emission records, and judge whether the emission level of each node exceeds the emission factor boundary parameters based on the threshold range obtained from training with historical data, and mark the over-limit nodes as candidate locations for abnormal emission sources within a given time window; Among them, each node is configured with an attribute vector containing the predicted emission factor value, operation status vector (such as temperature, pressure, energy consumption, electric power, etc.), and historical emission records within the sliding time window (such as continuous emission factor sequences or pollutant concentration sequences); Each edge is configured with attribute parameters representing emission correlation, including the material transfer rate , energy coupling coefficient , process path distance , and the emission correlation factor calculated from the emission factor sequence within the time window , which is used to reflect the historical cooperation degree of emission behaviors between nodes; Among them, each node is configured with an attribute vector containing the predicted emission factor value, operation status vector (such as temperature, pressure, energy consumption, electric power, etc.), and historical emission records within the sliding time window (such as continuous emission factor sequences or pollutant concentration sequences); Each edge is configured with attribute parameters representing emission correlation, including the material transfer rate , energy coupling coefficient , process path distance , and the emission correlation factor calculated from the emission factor sequence within the time window , which is used to reflect the historical cooperation degree of emission behaviors between nodes; The calculation formula for the graph edge propagation intensity is:

[0021] Among them represents the node to the emission propagation intensity, which is used as the edge weight input in the graph neural network. The graph neural network is modeled based on the above graph structure, fusing the connection relationship between nodes and the attribute vector information, and calculating the propagation path and propagation trend of the carbon emission factor in the network structure; The analysis of the spatio-temporal distribution trend of the carbon emission intensity includes: within a sliding time window, based on the temporal emission trajectory of abnormal nodes and the spatial connection relationship in the graph spectrum, constructing a spatio-temporal mapping structure of the carbon emission intensity to express the distribution trend of emission anomalies and the propagation and evolution path, supporting the trend prediction and control strategy formulation of regional abnormal diffusion; The identification result according to the abnormal emission source location includes: extracting the predicted values of the carbon emission factors of each node in the carbon emission propagation graph spectrum and comparing them with their corresponding boundary parameters. If the predicted value of the carbon emission factor exceeds the upper and lower limit ranges of its corresponding boundary parameters, then mark this node as an abnormal node; Based on the connection relationship of the edges, emission propagation intensity, material transfer rate, and energy correlation parameters in the carbon emission propagation graph spectrum, analyze the emission correlation degree between abnormal nodes and other nodes and their abnormal emission source locations in the directed graph structure; Specifically: The system determines whether there is an emission anomaly at the current node by combining the emission boundary parameters output by the dynamic prediction model according to the following anomaly determination conditions:

[0022] If the predicted value of the node emission factor exceeds its credible boundary range , then mark it as a candidate abnormal emission source node; After marking multiple abnormal nodes, the system further identifies the final abnormal emission source location based on the propagation graph spectrum. This process comprehensively judges the propagation influence and time continuity of each candidate abnormal node through an anomaly scoring function. The scoring function is defined as follows:

[0023] Among them represents the set of adjacent nodes of node , represents whether the adjacent node is marked as an abnormal node (1 if yes), represents the number of time steps that node is continuously determined to be abnormal, represents the window length, is a hyperparameter for adjusting the propagation intensity and time weight contribution.

[0024] The final abnormal emission source node is the node with the largest score:

[0025] in is the set of all abnormal candidate nodes, That is the final abnormal emission source location, and its corresponding equipment number and time information can be further output and used for early warning response.

[0026] S4. Based on the results of identifying the location of abnormal emission sources, perform carbon emission trend forecasting and trigger emission warning responses based on preset thresholds.

[0027] The method of performing carbon emission trend prediction and triggering an emission warning response based on a preset threshold includes: extracting the corresponding carbon emission factor prediction value, operating status data, and spatiotemporal distribution trend within the corresponding time window based on the identified abnormal emission source location, and constructing an emission intensity change trajectory of the device node in chronological order; Based on the connection relationship of the device nodes in the carbon emission propagation map, the emission sequence within the time window is updated; Perform trend extension processing on the updated time series to generate a carbon emission factor change trajectory for the node in subsequent time steps, and compare the predicted result with a preset emission threshold. When the predicted value exceeds the threshold, an emission warning response is triggered; Specifically, after the system identifies the final abnormal emission source location through the propagation map, it automatically extracts the carbon emission factor prediction value, operating status data (such as temperature, pressure, energy consumption) of the node in the current time window, as well as the spatial topological structure in the propagation map and the connection relationship with surrounding nodes to form a time series sample set and a spatial propagation relationship set.

[0028] Based on the time series data of the emission source node, the system constructs a chronological trajectory of its emission intensity changes at multiple time steps, both current and historical, forming a complete evolutionary path. This path is used to capture the trend direction, amplitude, and periodic characteristics of the device's emission fluctuations in the recent period.

[0029] Subsequently, combined with the upstream and downstream connection relationship of the node in the carbon emission propagation map (i.e., the material / energy path and emission correlation between it and other equipment), the emission sequence input within the time window is updated to correct its emission change trend and enhance the contextual consistency and systematicness of the trend prediction.

[0030] Based on the updated time series, the system performs trend extension processing, that is, predicts the emission factors for multiple subsequent time steps through time series modeling methods (such as LSTM or graph neural time series prediction structures) to generate continuous trend trajectories and capture the potential upward or sudden change risks of future carbon emissions.

[0031] Embodiment 2 A carbon emission accounting system based on artificial intelligence technology, characterized in that: Data processing module: construct a sensing network for the operating status of the device, obtain real-time operating data, and preprocess the real-time operating data through edge computing nodes; Dynamic prediction model module: establish a dynamic prediction model for the boundary parameters of emission factors based on real-time operating data; Analysis module: construct a carbon emission propagation map according to the dynamic prediction model, analyze the spatio-temporal distribution trend of carbon emission intensity and identify the location of abnormal emission sources; Early warning module: according to the identification result of the location of abnormal emission sources, perform carbon emission trend prediction and trigger an emission early warning response in combination with a preset threshold.

[0032] The technical features not described in the present invention can be achieved by or adopted from the prior art, and will not be elaborated here. Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by those of ordinary skill in the art within the scope of the essence of the present invention should also fall within the protection scope of the present invention.

Claims

1. A carbon emission accounting method based on artificial intelligence technology, characterized in that, Including: Construct a sensing network for the operating status of the device, obtain real-time operating data, and preprocess the real-time operating data through edge computing nodes; Based on the real-time operating data, establish a dynamic prediction model for emission factor boundary parameters; according to the dynamic prediction model, construct a carbon emission propagation map, analyze the temporal and spatial distribution trends of carbon emission intensity, and identify the locations of abnormal emission sources; according to the identification results of the abnormal emission source locations, perform carbon emission trend prediction, and trigger an emission warning response in combination with a preset threshold.

2. The carbon emission accounting method based on artificial intelligence technology according to claim 1, wherein, The real-time operating data includes device operating condition parameter data and carbon emission pollutant concentration monitoring data; The device operating condition parameter data includes reaction temperature data, reaction pressure data, raw material inlet and outlet flow rate data, raw material ratio data, power consumption data, heat input data, device start-stop status data, and device operating cycle data; The carbon emission pollutant concentration monitoring data includes carbon dioxide concentration data, methane concentration data, carbon monoxide concentration data, nitrogen oxide concentration data, volatile organic compound concentration data, and particulate matter concentration data.

3. The carbon emission accounting method based on artificial intelligence technology according to claim 2, wherein The preprocessing includes data integrity verification, standardization processing, time series alignment, and formatting conversion operations on the real-time operating data.

4. The carbon emission accounting method based on artificial intelligence technology according to claim 3, wherein The establishment of the dynamic prediction model for emission factor boundary parameters includes: constructing a continuous time series input based on the device operating condition parameter data and the carbon emission pollutant concentration monitoring data, dividing the continuous time series input into sliding time windows of a fixed length, and inputting them into a dynamic prediction model composed of a time series modeling unit and a boundary inference unit; The time series modeling unit is used to extract the dynamic correlation features between the operating status and the emission factors, encode the input sequence using a long short-term memory neural network, and generate the emission factor prediction value corresponding to the current window; The boundary inference unit calculates the credible upper and lower boundaries of the current prediction value based on the encoding of the time series modeling unit and the historical fluctuation statistical information, and outputs the emission factor boundary parameters; The emission factor boundary parameters include the emission factor prediction value, the upper bound parameter, and the lower bound parameter, which are used to define the reasonable range of the emission factor under the current operating conditions.

5. The carbon emission accounting method based on artificial intelligence technology according to claim 4, wherein The construction of the carbon emission propagation map includes: using the emission factor boundary parameters generated by the dynamic prediction model and the real-time operating data as inputs; Taking the device unit as the graph node within a set time window, with the edge representing the transmission path, configuring an attribute vector containing the emission factor prediction value, the operating status vector, and the historical emission record for each node, and configuring an attribute parameter representing the emission correlation for the edge; Model the directed graph structure through a graph neural network, calculate the propagation path of the carbon emission intensity in the directed graph structure using the connection relationship between nodes and the attribute vector of the historical emission record, and judge whether the emission level of each node exceeds the emission factor boundary parameters based on the threshold range obtained from training with historical data, and mark the over-limit nodes as candidate locations for abnormal emission sources within a given time window.

6. The carbon emission accounting method based on artificial intelligence technology according to claim 5, wherein, The analysis of the spatiotemporal distribution trend of carbon emission intensity includes: within a sliding time window, based on the temporal emission trajectory of abnormal nodes and the spatial connection relationship in the graph, constructing a spatiotemporal mapping structure of carbon emission intensity to express the distribution trend and propagation evolution path of emission anomalies.

7. The carbon emission accounting method based on artificial intelligence technology according to claim 6, wherein The results of the abnormal emission source location identification include: extracting the carbon emission factor prediction value of each node in the carbon emission propagation map and comparing it with the corresponding boundary parameters. If the carbon emission factor prediction value exceeds the upper and lower limits of the corresponding boundary parameters, the node is marked as an abnormal node; Based on the edge connectivity, emission propagation intensity, material transmission rate, and energy correlation parameters in the carbon emission propagation graph, the degree of emission correlation between abnormal nodes and other nodes and the location of abnormal emission sources in the directed graph structure are analyzed.

8. The carbon emission accounting method based on artificial intelligence technology according to claim 7, characterized in that, The method of performing carbon emission trend prediction and triggering an emission warning response based on a preset threshold includes: extracting the corresponding carbon emission factor prediction value, operating status data, and spatiotemporal distribution trend within the corresponding time window based on the identified abnormal emission source location, and constructing an emission intensity change trajectory of the device node in chronological order; Based on the connection relationship of the device nodes in the carbon emission propagation map, the emission sequence within the time window is updated; Trend extension processing is performed on the basis of the updated time series to generate the carbon emission factor change trajectory of the node in subsequent multiple time steps, and the predicted results are compared with the preset emission threshold. When the predicted value exceeds the threshold, an emission warning response is triggered.

9. A carbon emission accounting system based on artificial intelligence technology using the method according to any one of claims 1 to 8, characterized in that: Data processing module: builds a sensor network for equipment operation status, obtains real-time operation data, and pre-processes the real-time operation data through edge computing nodes; dynamic prediction model module: establishes a dynamic prediction model for emission factor boundary parameters based on real-time operation data; analysis module: based on the dynamic prediction model, builds a carbon emission propagation map, analyzes the temporal and spatial distribution trend of carbon emission intensity and identifies the location of abnormal emission sources; early warning module: performs carbon emission trend prediction based on the abnormal emission source location identification results, and triggers emission early warning response in combination with preset thresholds.

Images