Carbon footprint tracing system and method fusing knowledge graph and spatiotemporal attention

Abstract

The application discloses a carbon footprint tracing system fusing a knowledge graph and space-time attention, which comprises a multi-source data fusion module, a power grid, product and auxiliary data are used to obtain space-time characteristics and a knowledge graph original data set; an electric carbon factor dynamic prediction module, preliminary prediction is carried out through a fusion model, and a dynamic error correction model is used to calibrate to obtain a final electric carbon factor; a knowledge graph construction module, based on preset entities and relations and attributes, extracts and fuses knowledge elements from original data, and constructs a special knowledge graph; a carbon footprint accurate accounting module, based on the correlation of the special knowledge graph, automatically matches the final electric carbon factor with the power consumption of each production process of the product, combines non-power carbon emission data, and obtains the total amount of product carbon footprint through an accounting model; and a carbon footprint tracing module, based on the special knowledge graph, traces the production link of the total amount of product carbon footprint. The application improves the product carbon footprint accounting accuracy and full-link tracing capability.

Description

Technical Field

[0001] This invention relates to the fields of energy and environmental engineering, data processing and artificial intelligence, specifically to a carbon footprint tracing system and method that integrates knowledge graphs and spatiotemporal attention. Background Technology

[0002] Current product carbon footprint accounting suffers from weak linkage between electricity carbon factor and carbon footprint accounting. Traditional accounting methods often use static, uniform electricity carbon factors, failing to implement time-sharing and region-based dynamic accounting as required by regulations, resulting in significant accounting errors. Carbon footprint traceability is also insufficient, with existing methods often producing black-box results that fail to clearly present the correlation between electricity carbon factors and carbon emission links, making it difficult to pinpoint the root causes of high carbon emissions. Furthermore, the requirements for accounting boundaries and data sources are not followed. In addition, multi-source data integration is difficult, with heterogeneous data from grid topology, power supply operation, and product manufacturing processes forming data silos. This leads to low accuracy in matching electricity carbon factors with electricity consumption at each stage, and the data quality does not meet relevant requirements. While spatiotemporal attention mechanisms and graph neural network technology can accurately capture spatiotemporal features and process graph structure data, they have not been jointly applied to carbon footprint accounting scenarios that comply with the standards for calculating the carbon factor, nor have they been integrated with the standard accounting formulas and regional time period division standards. Although knowledge graph technology can achieve structured data association and visual traceability, it lacks the ability to dynamically predict the carbon factor when applied alone, and cannot support real-time updates and accurate accounting of the carbon footprint. It has also not been integrated into the definitions and accounting logic of the carbon factor and carbon emissions in the relevant requirements. Summary of the Invention

[0003] The purpose of this invention is to provide a carbon footprint tracing system and method that integrates knowledge graphs and spatiotemporal attention. This invention improves the accuracy of product carbon footprint accounting and the end-to-end traceability capability.

[0004] To achieve this objective, the carbon footprint tracing system integrating knowledge graphs and spatiotemporal attention designed in this invention includes: The multi-source data fusion module is used to preprocess power grid-side control cloud and carbon emission data, product-side energy consumption data in production processes and carbon footprint traceability auxiliary data to obtain spatiotemporal feature datasets and knowledge graph raw datasets. The dynamic prediction module for the electric carbon factor is used to predict the preliminary electric carbon factor using the spatiotemporal feature dataset through a pre-trained fusion model, and to calibrate the preliminary electric carbon factor using a dynamic error correction model to obtain the final electric carbon factor. The knowledge graph construction module is used to extract knowledge elements from the original dataset of the knowledge graph based on the entities, relationships and attributes of the preset carbon factors and carbon footprints, perform knowledge fusion on the knowledge elements, and construct a knowledge graph dedicated to carbon factors and carbon footprints. The carbon footprint accurate accounting module is used to automatically match the final carbon factor with the electricity consumption of each production process in the product's entire life cycle based on the correlation between the electric carbon factor and the carbon footprint in the dedicated knowledge graph of carbon footprint. Combined with non-electric carbon emission data, the total carbon footprint of the product is calculated through the full life cycle carbon footprint accounting model. The carbon footprint traceability module is used to trace the generation chain of the total carbon footprint of the product based on the electric carbon factor and the carbon footprint-specific knowledge graph.

[0005] Preferably, the grid-side control cloud and carbon emission data include control cloud ledger data, control cloud operation data, and measured carbon emission data from thermal power plants; the energy consumption data in the product-side production process includes implicit carbon, electricity consumption, transportation energy consumption, and recycling efficiency data at each stage of the product's entire life cycle; and the carbon footprint traceability auxiliary data includes meteorological data, policy and standard data, and historical carbon factor data.

[0006] Preferably, the specific method for preprocessing power grid-side control cloud and carbon emission data, product-side energy consumption data in production processes, and carbon footprint traceability auxiliary data to obtain spatiotemporal feature datasets and original knowledge graph datasets is as follows: Data cleaning was performed on the grid-side control cloud and carbon emission data, energy consumption data in the product-side production process and carbon footprint traceability auxiliary data. Outliers were removed using the 3σ criterion, duplicate data was removed by deduplication, and missing values ​​were filled in by the mean filling method to obtain the cleaned data. Data standardization is performed on the cleaned data, heterogeneous data in the cleaned data is formatted and encoded, power grid topology data in the cleaned data is converted into graph structure data, and product production process data in the cleaned data is aligned by timestamps to obtain standardized data. The standardized data is integrated to correlate the power consumption of power grid nodes with the power consumption of each production process in the entire product life cycle, generating a spatiotemporal feature dataset and a knowledge graph raw dataset.

[0007] Preferably, the specific process of obtaining the preliminary electrocarbon factor by using the spatiotemporal feature dataset through a pre-trained fusion model is as follows: The spatiotemporal feature dataset is input into a pre-trained spatiotemporal attention and graph neural network fusion model; In the pre-trained spatiotemporal attention and graph neural network fusion model, within the spatiotemporal attention layer, the time attention sublayer dynamically assigns weights to data at different time points based on the temporal characteristics of the spatiotemporal feature dataset, according to peak and valley periods and seasonal variations, to obtain a time dimension feature vector; the spatial attention sublayer dynamically assigns weights to data at each grid node based on the correlation strength between grid nodes and the proportion of transmission power between grid nodes, to obtain a spatial dimension feature vector; combining the time dimension feature vector and the spatial dimension feature vector yields the spatiotemporal feature vector; In the pre-trained spatiotemporal attention and graph neural network fusion model, the graph neural network layer transforms the power grid topology graph structure data into an undirected graph. The graph convolutional network learns the carbon flow transport relationship between power grid nodes and obtains a node embedding vector that represents the spatial association attribute of each power grid node in the topology network. Initial weights are assigned to the spatiotemporal feature vector and the node embedding vector respectively, and the basic fusion features are obtained by weighted summation according to the initial weight ratio. Based on the basic fusion features, key dimensions are extracted from the spatiotemporal feature vector and the node embedding vector and concatenated to form a fusion feature vector containing spatiotemporal and spatial correlation characteristics. The fusion feature vector is input into the fully connected layer in the pre-trained spatiotemporal attention and graph neural network fusion model for calculation and mapping, and finally outputs the preliminary electric carbon factor corresponding to each power grid node and each prediction period.

[0008] Preferably, the specific process of using a dynamic error correction model to calibrate the initial electrocarbon factor to obtain the final electrocarbon factor is as follows: A dynamic error correction model is constructed using the material balance method. The actual carbon emissions of thermal power plants are calculated using this model to obtain the baseline carbon factor. The calculation formula is as follows: in, The amount of carbon dioxide emissions produced by fuel combustion. Fuel consumption For fuel carbon content, Fuel oxidation rate; The deviation rate between the preliminary carbon factor and the benchmark carbon factor is calculated. Based on the correlation between the deviation rate and the power generation type, time period and meteorological conditions, a piecewise linear correction function is established to perform differentiated calibration of the preliminary carbon factor under different scenarios, and the final carbon factor is obtained.

[0009] Preferably, the specific process for constructing the dedicated knowledge graph of the electrocarbon factor and carbon footprint is as follows: The entities, relationships, and attributes of the carbon factor and carbon footprint are predefined. Based on the policy text and the original dataset of the knowledge graph, the BERT model is used to extract entity attributes and accounting rules from the textual policy data. An entity recognition algorithm is used to extract knowledge elements from structured or semi-structured power grid data and product data. The extracted knowledge elements are aligned to eliminate entity conflicts, and a conflict detection algorithm is used to detect and resolve conflicts, forming consistent knowledge elements. The knowledge elements after knowledge fusion are then stored in a graph database to form a dedicated knowledge graph for the carbon factor and carbon footprint.

[0010] Preferably, the specific process for obtaining the total carbon footprint of the product is as follows: Based on the correlation between the electric carbon factor and the carbon footprint-specific knowledge graph of production process time, corresponding power grid area, corresponding power grid node, corresponding time period and electric carbon factor, the final electric carbon factor is automatically matched for the power consumption of each production process in the entire product life cycle according to the execution time and geographical location of the production process. Combining non-electricity carbon emission data, the carbon footprint is calculated using a life-cycle carbon footprint accounting model, with the following formula: Total product carbon footprint = Σ (electricity consumption of each production process × corresponding final carbon factor) + carbon emissions hidden in raw materials + total carbon emissions from transportation + total carbon emissions from other non-electricity processes - carbon offset from recycling and reuse; where Σ represents the summation of carbon footprint for each production process. The final electrical carbon factor corresponding to the power consumption of each production process is calculated using nodal carbon potential. The formula for calculating nodal carbon potential is as follows: in, Let be the carbon potential of grid node n at time t. Let n be the set of all branches connected to node n in the power grid. For carbon footprint flow through branch roads l carbon flow, For carbon footprint flow through branch roads l The active power; The carbon footprint flows through the branch road l carbon flow The calculation formula is: , carbon footprint branch l Carbon density at time t.

[0011] Preferably, the specific process for tracing the generation chain of the total carbon footprint of the product based on the electrocarbon factor and the carbon footprint-specific knowledge graph is as follows: By using the electrical carbon factor and the carbon footprint-specific knowledge graph, a complete carbon footprint generation chain corresponding to the total carbon footprint of the product is automatically generated. The nodes of the carbon footprint generation chain include electrical carbon factor, power grid area, power grid node, power consumption, production process, carbon emission link and carbon footprint result. Each node in the carbon footprint generation chain is an entity in the electrical carbon factor and the carbon footprint-specific knowledge graph.

[0012] A carbon footprint tracing method integrating knowledge graphs and spatiotemporal attention includes the following steps: Preprocessing of power grid-side control cloud and carbon emission data, energy consumption data in product-side production processes and carbon footprint traceability auxiliary data yields spatiotemporal feature datasets and original knowledge graph datasets. The preliminary electric carbon factor is obtained by using the spatiotemporal feature dataset to make predictions through a pre-trained fusion model. The preliminary electric carbon factor is then calibrated for deviation using a dynamic error correction model to obtain the final electric carbon factor. Based on the entities, relationships, and attributes of the pre-defined carbon factors and carbon footprint, knowledge elements are extracted from the original dataset of the knowledge graph, and knowledge elements are fused to construct a unique knowledge graph for carbon factors and carbon footprint. Based on the correlation between the electric carbon factor and carbon footprint in the knowledge graph dedicated to the electric carbon factor and carbon footprint, the final electric carbon factor is automatically matched with the power consumption of each production process in the product's entire life cycle. Combined with non-electric carbon emission data, the total carbon footprint of the product is calculated through the full life cycle carbon footprint accounting model. The generation chain of the total carbon footprint of the product is traced based on the aforementioned electrocarbon factor and carbon footprint-specific knowledge graph.

[0013] A computer program product includes a computer program that, when executed by a processor, implements the steps of the above-described method.

[0014] The beneficial effects of this invention are: This invention reduces the prediction error rate of the carbon footprint factor and improves its accuracy by using a spatiotemporal attention and graph neural network fusion model combined with a standardized accounting formula for dynamic error correction. Based on a dedicated knowledge graph of the carbon footprint factor and its relation to carbon factors, it achieves automatic and accurate matching of the carbon footprint factor with the electricity consumption of each production process throughout the product's lifecycle, further reducing the error rate of the product's carbon footprint calculation. By tracing the product's production chain based on the dedicated knowledge graph of the carbon footprint factor and its relation to carbon factors, it achieves transparency in the accounting process, enhances interpretability, and improves the efficiency of integrating and processing multi-source data. This invention saves manpower and time costs through automated calculation of the product's carbon footprint. Attached Figure Description

[0015] Figure 1This is a schematic diagram of the structure of the present invention; Figure 2 This is a flowchart of the present invention. Detailed Implementation

[0016] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments: Example 1 A carbon footprint tracing system that integrates knowledge graphs and spatiotemporal attention, such as Figure 1 As shown, it includes: The multi-source data fusion module is used to preprocess the grid-side control cloud and carbon emission data, the energy consumption data in the product-side production process, and the carbon footprint traceability auxiliary data to obtain the spatiotemporal feature dataset and the knowledge graph original dataset. This design realizes the efficient integration of grid-side control cloud and carbon emission data, the energy consumption data in the product-side production process, and the carbon footprint traceability auxiliary data, providing a high-quality, standardized dataset for subsequent electric carbon factor prediction and carbon footprint accounting. The dynamic prediction module for carbon emission factor is used to predict the preliminary carbon emission factor using the spatiotemporal feature dataset through a pre-trained fusion model. The preliminary carbon emission factor is then calibrated using a dynamic error correction model to obtain the final carbon emission factor. This design can output high-precision, time-division and zone-division-level final carbon emission factors at the station level and line level, providing a dynamic and reliable data source for accurate carbon footprint accounting. The knowledge graph construction module is used to extract knowledge elements from the original dataset of the knowledge graph based on the entities, relationships and attributes of the preset carbon factors and carbon footprints, perform knowledge fusion on the knowledge elements, and construct a dedicated knowledge graph for carbon factors and carbon footprints. This design realizes the structured association between carbon factors and carbon footprint-related data, and provides a foundation for the automatic matching of carbon factors and electricity consumption and the full-link traceability of carbon footprints. The carbon footprint precision accounting module is used to automatically match the final carbon factor with the power consumption of each production process in the product's entire life cycle based on the correlation between the carbon factor and the carbon footprint in the dedicated knowledge graph of carbon footprint. Combined with non-electric carbon emission data, the total carbon footprint of the product is calculated through the full life cycle carbon footprint accounting model. This design realizes the dynamic and precise accounting of the carbon footprint of the product's entire life cycle by automatically matching the carbon factor with the power consumption of each production process of the product based on the correlation between the carbon factor and the carbon footprint in the dedicated knowledge graph of carbon footprint. The carbon footprint traceability module is used to trace the generation chain of the total carbon footprint of the product based on the electric carbon factor and the carbon footprint-specific knowledge graph. This design presents the complete carbon emission generation chain through the electric carbon factor and the carbon footprint-specific knowledge graph, and supports querying detailed accounting data of each link layer by layer, which can realize the visualized full-chain traceability of carbon footprint.

[0017] In the above technical solution, the grid-side control cloud and carbon emission data include control cloud ledger data, control cloud operation data, and measured carbon emission data from thermal power plants. The energy consumption data in the product-side production process includes implicit carbon, electricity consumption, transportation energy consumption, and recycling efficiency data at each stage of the product's entire life cycle. The carbon footprint traceability auxiliary data includes meteorological data, policy and standard data (including accounting specifications), and historical electricity carbon factor data. The above design refines the specific composition of each type of data, ensuring that data collection fully conforms to industry standards and actual accounting needs, and avoiding the impact of missing data or incompatible data types on the effectiveness of subsequent stages.

[0018] The specific method for preprocessing the power grid-side control cloud and carbon emission data, the energy consumption data in the product-side production process, and the carbon footprint traceability auxiliary data to obtain the spatiotemporal feature dataset and the original knowledge graph dataset in the above technical solution is as follows: Data cleaning was performed on the grid-side control cloud and carbon emission data, energy consumption data in the product-side production process and carbon footprint traceability auxiliary data. Outliers were removed using the 3σ criterion, duplicate data was removed by deduplication, and missing values ​​were filled in by the mean filling method to obtain the cleaned data. Data standardization is performed on the cleaned data, heterogeneous data in the cleaned data is formatted and encoded, power grid topology data in the cleaned data is converted into graph structure data, and product production process data in the cleaned data is aligned by timestamps to obtain standardized data. The standardized data is integrated to correlate the power consumption of power grid nodes with the power consumption of each production process in the product's entire life cycle, generating a spatiotemporal feature dataset and a knowledge graph raw dataset. The above design clarifies the specific methods for data cleaning, standardization, and integration, ensuring that the final spatiotemporal feature dataset and knowledge graph raw dataset better meet the requirements.

[0019] For grid nodes, some optimized technical solutions include: In physical entities, grid nodes refer to the basic connection points and power conversion points in the physical network of the power grid, including substations, converter stations, and grid connection points of power plants. These nodes are the basic elements constituting the grid topology structure. In the dedicated knowledge graph of electric carbon factor and carbon footprint, a grid node refers to a core entity of the electric carbon factor class, indicating that each physical grid node has a corresponding structured digital representation in the dedicated knowledge graph of electric carbon factor and carbon footprint, including the attributes of the grid node (such as node ID, voltage level, and region) and its association with other entities (such as power plants, electric carbon factor values, and power consumption points).

[0020] In the above technical solution, the specific process of obtaining the preliminary electrocarbon factor by using the spatiotemporal feature dataset through a pre-trained fusion model is as follows: The spatiotemporal feature dataset (which includes power grid operation data, meteorological data, and historical carbon factor data) is input into a pre-trained spatiotemporal attention and graph neural network fusion model; In the pre-trained spatiotemporal attention and graph neural network fusion model, within the spatiotemporal attention layer, the time attention sublayer dynamically assigns weights to data at different time points based on the temporal characteristics of the spatiotemporal feature dataset, according to peak and valley periods and seasonal variations (e.g., assigning higher weights to data during peak electricity consumption periods to strengthen the impact of peak electricity consumption periods on the predicted carbon factor), thus obtaining a time dimension feature vector. The spatial attention sublayer dynamically assigns weights to data of each grid node (different grid nodes represent different spatial locations) based on the correlation strength between grid nodes and the proportion of transmission power between grid nodes (e.g., assigning higher weights to data of associated grid nodes that frequently exchange power with the target grid node), thus obtaining a spatial dimension feature vector. Combining the time dimension feature vector and the spatial dimension feature vector yields a spatiotemporal feature vector that integrates key spatiotemporal information. In the pre-trained spatiotemporal attention and graph neural network fusion model, the graph neural network layer transforms the power grid topology graph structure data into an undirected graph (the power grid topology graph structure data is obtained through a spatiotemporal feature dataset, where node attributes include equipment type and installed capacity, and edge attributes include line parameters and transmission power). The graph convolutional network learns the carbon flow transmission relationship between power grid nodes, and obtains a node embedding vector that represents the spatial association attribute of each power grid node in the topology network. Initial weights are assigned to the spatiotemporal feature vector and the node embedding vector respectively, and the basic fusion features are obtained by weighted summation according to the initial weight ratio. Based on the basic fusion features, key dimensions (such as "time period identifier" in the spatiotemporal feature vector and "access power type" in the node embedding vector) are extracted and concatenated to form a fusion feature vector with richer dimensions and containing spatiotemporal and spatial correlation characteristics. The fusion feature vector is input into the fully connected layer of the pre-trained spatiotemporal attention and graph neural network fusion model for calculation and mapping, and finally outputs the preliminary carbon factor corresponding to each grid node and each prediction time period. The above design defines in detail the two-layer architecture of the spatiotemporal attention and graph neural network fusion model and the specific working method of each layer, and clarifies the label data and feature vector fusion method in the training process, which can reduce the prediction error of the spatiotemporal attention and graph neural network fusion model and improve the prediction accuracy of the preliminary carbon factor.

[0021] For the pre-trained spatiotemporal attention and graph neural network fusion model, some optimization techniques include: using historical carbon factor data calculated in accordance with the relevant requirements of the "Power Grid Regional Time-Division and Zoning Carbon Factor Calculation Specification" as the real label, inputting the historical spatiotemporal feature dataset into the spatiotemporal attention and graph neural network fusion model, performing forward propagation to obtain the predicted preliminary carbon factor, calculating the error between the predicted preliminary carbon factor and the real label, backpropagating the loss error through the gradient descent optimization algorithm, dynamically adjusting the weight allocation parameters of the spatiotemporal attention layer, the graph convolution parameters of the graph neural network layer, and the weight parameters of the feature fusion layer in the spatiotemporal attention and graph neural network fusion model, and iteratively optimizing the parameters of the spatiotemporal attention and graph neural network fusion model until the prediction error rate of the spatiotemporal attention and graph neural network fusion model on the validation set is lower than a preset threshold (the preset threshold is 5%), and the training is completed, resulting in the pre-trained spatiotemporal attention and graph neural network fusion model.

[0022] In the above technical solution, the specific process of using a dynamic error correction model to calibrate the initial electrocarbon factor and obtain the final electrocarbon factor is as follows: A dynamic error correction model is constructed using the material balance method. The actual carbon emissions of thermal power plants are calculated using this model to obtain the baseline carbon factor. The calculation formula is as follows: in, The amount of carbon dioxide emissions produced by fuel combustion. Fuel consumption For fuel carbon content, This represents the fuel oxidation rate; 44 / 12 is the molecular weight conversion factor between carbon dioxide (CO2) and carbon (C). For constructing a dynamic error correction model, some optimized technical solutions include: A dynamic error correction model can also be constructed using experimental methods. The actual carbon emissions of thermal power plants can then be calculated using this model to obtain the baseline carbon factor. The calculation formula is as follows: in, The amount of carbon dioxide emissions produced by fuel combustion. Dry flue gas flow rate, This represents the volume fraction of CO2 in the flue gas. The molecular weight of CO2 is... The average molecular weight of air; This is a conversion factor used to convert flue gas analysis measurements into carbon dioxide mass emissions under standard conditions. The deviation rate between the preliminary carbon emission factor and the benchmark carbon emission factor is calculated. Based on the correlation between the deviation rate and power generation type, time period and meteorological conditions, a piecewise linear correction function is established to perform differentiated calibration of the preliminary carbon emission factor under different scenarios, thus obtaining the final carbon emission factor. The above design realizes differentiated calibration of the prediction results of the preliminary carbon emission factor, further improving the accuracy of the final carbon emission factor, and making the corrected final carbon emission factor more consistent with the actual carbon emission situation of the power grid.

[0023] In the above technical solution, the specific process of constructing the dedicated knowledge graph of the electrocarbon factor and carbon footprint is as follows: This design predefines the entities, relationships, and attributes of the carbon factor and carbon footprint. Based on policy texts (including the "Specification for Time-of-Use and Zonal Carbon Factor Calculation in Power Grid Areas") and the original dataset of the knowledge graph, it uses the BERT model to extract entity attributes and calculation rules from the textual policy data. An entity recognition algorithm is then used to extract knowledge elements from structured or semi-structured power grid and product data. The extracted knowledge elements are aligned to eliminate entity conflicts, and a conflict detection algorithm is used to detect and resolve conflicts, resulting in consistent knowledge elements. These fused knowledge elements are then stored in a graph database to form a dedicated knowledge graph for the carbon factor and carbon footprint. This design refines the entities, relationships, and attributes of the carbon factor and carbon footprint, clarifies the specific content of various entities, and outlines the core relationships between them. This ensures that the dedicated knowledge graph for the carbon factor and carbon footprint can match the prediction results of the carbon factor and changes in product production data in real time, improving data accuracy.

[0024] Regarding the entities, relationships, and attributes of electrical carbon factors and carbon footprints, some optimized technical solutions include: determining the core entity types, attributes, and relationships between entities for electrical carbon factors and carbon footprints. Entity types include electrical carbon factor entities, product manufacturing entities, carbon emission entities, and policy / standard entities. The core entities included in each entity type are: Entities related to the carbon factor include power plants, grid nodes, grid regions, time periods, carbon factor values, fuel types, and carbon emission sources. Product manufacturing entities include products, production processes, raw materials, electricity consumption points, transportation methods, and recycling processes; Carbon emission entities include emission sources, emission stages, carbon emission amounts, branch carbon flows, branch carbon density, and nodal carbon potential; Policy and standard entities include accounting standards, measurement standards, and emission reduction targets; Attributes are features or values ​​used to describe a specific entity; Preset the relationships between entities, such as: "Power plant - Corresponding - Power grid node", "Power grid node - Belonging - Power grid area", "Production process - Consumption - Electricity", "Electricity consumption - Association - Power grid node" and "Accounting specification - Constraint - Accounting process".

[0025] In the above technical solution, the specific process for obtaining the total carbon footprint of the product is as follows: Based on the correlation between the electric carbon factor and the carbon footprint-specific knowledge graph of production process time, corresponding power grid area, corresponding power grid node, corresponding time period and electric carbon factor, the final electric carbon factor is automatically matched for the power consumption of each production process in the entire product life cycle according to the execution time and geographical location of the production process. Combining non-electricity carbon emission data, the carbon footprint is calculated using a life-cycle carbon footprint accounting model, with the following formula: Total product carbon footprint = Σ (electricity consumption of each production process × corresponding final carbon factor) + carbon emissions hidden in raw materials + total carbon emissions from transportation + total carbon emissions from other non-electricity processes - carbon offset from recycling and reuse; where Σ represents the summation of carbon footprint for each production process. The final electrical carbon factor corresponding to the power consumption of each production process is calculated using nodal carbon potential. The formula for calculating nodal carbon potential is as follows: in, Let be the carbon potential of grid node n at time t. Let n be the set of all branches connected to node n in the power grid. For carbon footprint flow through branch roads l carbon flow, For carbon footprint flow through branch roads l The active power; The carbon footprint flows through the branch road l carbon flow The calculation formula is: , carbon footprint branch l The carbon density at time t; the above design, through the full life cycle carbon footprint accounting model, clarifies the automatic matching rules of the electrical carbon factor, refines the calculation formula of the node carbon potential and the meaning of each parameter, and ensures that the accounting process can respond in a timely manner to changes in the electrical carbon factor and product production data, so that the accounting results always fit the actual situation and improve the standardization and accuracy of carbon footprint accounting.

[0026] In the above technical solution, the specific process of visually tracing and displaying the generation chain of the total carbon footprint of the product based on the electrocarbon factor and the carbon footprint-specific knowledge graph, and outputting emission reduction optimization suggestions based on the analysis results, is as follows: Through the electric carbon factor and carbon footprint-specific knowledge graph, a complete carbon footprint generation chain corresponding to the total carbon footprint of the product is automatically generated. The nodes of the carbon footprint generation chain include electric carbon factor, power grid area, power grid node, power consumption, production process, carbon emission link and carbon footprint result. Each node in the carbon footprint generation chain is an entity in the electric carbon factor and carbon footprint-specific knowledge graph. The generation chain of the total carbon footprint of the product is visualized and displayed through the dedicated knowledge graph of the electric carbon factor and carbon footprint. It presents the complete chain from the electric carbon factor to the grid node to the power consumption point to the production process to the carbon emission link and then to the carbon footprint result. It also supports drilling down layer by layer to view the data details of each chain. Based on graph neural network analysis, the data in the complete traceability chain is traced to uncover the correlation between high-carbon emission links, electrical carbon factors, and various production processes. If the carbon emissions of a certain link account for more than a preset threshold of the total carbon footprint of the product, then that link is a high-carbon emission link. Based on the analysis results, emission reduction optimization suggestions are output, and the carbon emission reduction effect after implementation is predicted by associating electrical carbon factors with a carbon footprint-specific knowledge graph. The above design refines the visualization link drilling dimensions of the traceability optimization link, the criteria for judging high-carbon emission links, the types of emission reduction suggestions, and the requirements for effect prediction, thereby improving the completeness of carbon footprint traceability, the accuracy of high-carbon emission link positioning, and the practicality of emission reduction suggestions.

[0027] Example 2 A carbon footprint tracing method that integrates knowledge graphs and spatiotemporal attention, such as Figure 2 As shown, the original dataset of spatiotemporal features and knowledge graph is obtained using power grid, product, and auxiliary data; preliminary prediction is performed through a fusion model, and the final carbon factor is obtained by calibration using a dynamic error correction model; based on preset entities, relationships, and attributes, knowledge elements are extracted and fused from the original data to construct a dedicated knowledge graph; based on the association relationships of the dedicated knowledge graph, the final carbon factor is automatically matched with the power consumption of each production process of the product, and combined with non-power carbon emission data, the total carbon footprint of the product is obtained through an accounting model; the generation chain of the total carbon footprint of the product is traced based on the dedicated knowledge graph.

[0028] The specific methods for carbon footprint tracing include the following steps: Preprocessing of power grid-side control cloud and carbon emission data, energy consumption data in product-side production processes and carbon footprint traceability auxiliary data yields spatiotemporal feature datasets and original knowledge graph datasets. The preliminary electric carbon factor is obtained by using the spatiotemporal feature dataset to make predictions through a pre-trained fusion model. The preliminary electric carbon factor is then calibrated for deviation using a dynamic error correction model to obtain the final electric carbon factor. Based on the entities, relationships, and attributes of the pre-defined carbon factors and carbon footprint, knowledge elements are extracted from the original dataset of the knowledge graph, and knowledge elements are fused to construct a unique knowledge graph for carbon factors and carbon footprint. Based on the correlation between the electric carbon factor and carbon footprint in the knowledge graph dedicated to the electric carbon factor and carbon footprint, the final electric carbon factor is automatically matched with the power consumption of each production process in the product's entire life cycle. Combined with non-electric carbon emission data, the total carbon footprint of the product is calculated through the full life cycle carbon footprint accounting model. Based on the aforementioned electrocarbon factor and carbon footprint-specific knowledge graph, the generation chain of the total carbon footprint of the product is traced. Example 3 A computer program product includes a computer program that, when executed by a processor, implements the steps of the method described in Embodiment 2.

[0029] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0030] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A system that specifies functions in one or more boxes.

[0031] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including an instruction set implemented in a process. Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0032] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0033] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit its scope of protection. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that after reading the present invention, they can still make various changes, modifications or equivalent substitutions to the specific implementation of the invention, but these changes, modifications or equivalent substitutions are all within the scope of protection of the pending claims of the invention.

[0034] The contents not described in detail in this specification are existing technologies known to those skilled in the art.

Claims

1. A carbon footprint tracing system integrating knowledge graphs and spatiotemporal attention, characterized in that, It includes: The multi-source data fusion module is used to preprocess power grid-side control cloud and carbon emission data, product-side energy consumption data in production processes and carbon footprint traceability auxiliary data to obtain spatiotemporal feature datasets and knowledge graph raw datasets. The dynamic prediction module for the electric carbon factor is used to predict the preliminary electric carbon factor using the spatiotemporal feature dataset through a pre-trained fusion model, and to calibrate the preliminary electric carbon factor using a dynamic error correction model to obtain the final electric carbon factor. The knowledge graph construction module is used to extract knowledge elements from the original dataset of the knowledge graph based on the entities, relationships and attributes of the preset carbon factors and carbon footprints, perform knowledge fusion on the knowledge elements, and construct a knowledge graph dedicated to carbon factors and carbon footprints. The carbon footprint accurate accounting module is used to automatically match the final carbon factor with the electricity consumption of each production process in the product's entire life cycle based on the correlation between the electric carbon factor and the carbon footprint in the dedicated knowledge graph of carbon footprint. Combined with non-electric carbon emission data, the total carbon footprint of the product is calculated through the full life cycle carbon footprint accounting model. The carbon footprint traceability module is used to trace the generation chain of the total carbon footprint of the product based on the electric carbon factor and the carbon footprint-specific knowledge graph.

2. The carbon footprint tracing system integrating knowledge graphs and spatiotemporal attention as described in claim 1, characterized in that: The grid-side control cloud and carbon emission data include control cloud ledger data, control cloud operation data, and measured carbon emission data from thermal power plants. The energy consumption data in the product-side production process includes implicit carbon, electricity consumption, transportation energy consumption, and recycling efficiency data at each stage of the product's entire life cycle. The carbon footprint traceability auxiliary data includes meteorological data, policy and standard data, and historical carbon factor data.

3. The carbon footprint tracing system integrating knowledge graphs and spatiotemporal attention as described in claim 1, characterized in that: The specific method for preprocessing power grid-side control cloud and carbon emission data, product-side energy consumption data in production processes, and carbon footprint traceability auxiliary data to obtain spatiotemporal feature datasets and original knowledge graph datasets is as follows: Data cleaning was performed on the grid-side control cloud and carbon emission data, energy consumption data in the product-side production process and carbon footprint traceability auxiliary data. Outliers were removed using the 3σ criterion, duplicate data was removed by deduplication, and missing values ​​were filled in by the mean filling method to obtain the cleaned data. Data standardization is performed on the cleaned data, heterogeneous data in the cleaned data is formatted and encoded, power grid topology data in the cleaned data is converted into graph structure data, and product production process data in the cleaned data is aligned by timestamps to obtain standardized data. The standardized data is integrated to correlate the power consumption of power grid nodes with the power consumption of each production process in the entire product life cycle, generating a spatiotemporal feature dataset and a knowledge graph raw dataset.

4. The carbon footprint tracing system integrating knowledge graphs and spatiotemporal attention as described in claim 1, characterized in that: The specific process of obtaining the preliminary electrocarbon factor by using the spatiotemporal feature dataset through the pre-trained fusion model is as follows: The spatiotemporal feature dataset is input into a pre-trained spatiotemporal attention and graph neural network fusion model; In the pre-trained spatiotemporal attention and graph neural network fusion model, within the spatiotemporal attention layer, the time attention sublayer dynamically assigns weights to data at different time points based on the temporal characteristics of the spatiotemporal feature dataset, according to peak and valley periods and seasonal variations, to obtain a time dimension feature vector; the spatial attention sublayer dynamically assigns weights to data at each grid node based on the correlation strength between grid nodes and the proportion of transmission power between grid nodes, to obtain a spatial dimension feature vector; combining the time dimension feature vector and the spatial dimension feature vector yields the spatiotemporal feature vector; In the pre-trained spatiotemporal attention and graph neural network fusion model, the graph neural network layer transforms the power grid topology graph structure data into an undirected graph. The graph convolutional network learns the carbon flow transport relationship between power grid nodes and obtains a node embedding vector that represents the spatial association attribute of each power grid node in the topology network. The spatiotemporal feature vector and the node embedding vector are assigned initial weights respectively, and the basic fusion features are obtained by weighted summation according to the initial weight ratio; Based on the basic fusion features, key dimensions from the spatiotemporal feature vector and node embedding vector are extracted and concatenated to form a fusion feature vector containing spatiotemporal and spatial correlation characteristics. The fusion feature vector is then input into the fully connected layer of the pre-trained spatiotemporal attention and graph neural network fusion model for calculation and mapping, and finally outputs the preliminary electric carbon factor corresponding to each power grid node and each prediction period.

5. The carbon footprint tracing system integrating knowledge graphs and spatiotemporal attention as described in claim 1, characterized in that: The specific process of using a dynamic error correction model to calibrate the initial electrocarbon factor and obtain the final electrocarbon factor is as follows: A dynamic error correction model is constructed using the material balance method. The actual carbon emissions of thermal power plants are calculated using this model to obtain the baseline carbon factor. The calculation formula is as follows: in, The amount of carbon dioxide emissions produced by fuel combustion. Fuel consumption For fuel carbon content, Fuel oxidation rate; The deviation rate between the preliminary carbon factor and the benchmark carbon factor is calculated. Based on the correlation between the deviation rate and the power generation type, time period and meteorological conditions, a piecewise linear correction function is established to perform differentiated calibration of the preliminary carbon factor under different scenarios, and the final carbon factor is obtained.

6. The carbon footprint tracing system integrating knowledge graphs and spatiotemporal attention as described in claim 1, characterized in that: The specific process for constructing the dedicated knowledge graph of electrocarbon factor and carbon footprint is as follows: The entities, relationships, and attributes of the carbon factor and carbon footprint are predefined. Based on the policy text and the original dataset of the knowledge graph, the BERT model is used to extract entity attributes and accounting rules from the textual policy data. An entity recognition algorithm is used to extract knowledge elements from structured or semi-structured power grid data and product data. The extracted knowledge elements are aligned to eliminate entity conflicts, and a conflict detection algorithm is used to detect and resolve conflicts, forming consistent knowledge elements. The knowledge elements after knowledge fusion are then stored in a graph database to form a dedicated knowledge graph for the carbon factor and carbon footprint.

7. The carbon footprint tracing system integrating knowledge graphs and spatiotemporal attention as described in claim 1, characterized in that: The specific process for obtaining the total carbon footprint of the product is as follows: Based on the correlation between the electric carbon factor and the carbon footprint-specific knowledge graph of production process time, corresponding power grid area, corresponding power grid node, corresponding time period and electric carbon factor, the final electric carbon factor is automatically matched for the power consumption of each production process in the entire product life cycle according to the execution time and geographical location of the production process. Combining non-electricity carbon emission data, the carbon footprint is calculated using a life-cycle carbon footprint accounting model, with the following formula: Total product carbon footprint = Σ (electricity consumption of each production process × corresponding final carbon factor) + carbon emissions hidden in raw materials + total carbon emissions from transportation + total carbon emissions from other non-electricity processes - carbon offset from recycling and reuse; where Σ represents the summation of carbon footprint for each production process. The final electrical carbon factor corresponding to the power consumption of each production process is calculated using nodal carbon potential. The formula for calculating nodal carbon potential is as follows: in, Let be the carbon potential of grid node n at time t. Let n be the set of all branches connected to node n in the power grid. For carbon footprint flow through branch roads l carbon flow, For carbon footprint flow through branch roads l The active power; The carbon footprint flows through the branch road l carbon flow The calculation formula is: , carbon footprint branch l Carbon density at time t.

8. The carbon footprint tracing system integrating knowledge graphs and spatiotemporal attention as described in claim 1, characterized in that: The specific process of tracing the generation chain of the total carbon footprint of the product based on the aforementioned electrocarbon factor and carbon footprint-specific knowledge graph is as follows: By using the electrical carbon factor and the carbon footprint-specific knowledge graph, a complete carbon footprint generation chain corresponding to the total carbon footprint of the product is automatically generated. The nodes of the carbon footprint generation chain include electrical carbon factor, power grid area, power grid node, power consumption, production process, carbon emission link and carbon footprint result. Each node in the carbon footprint generation chain is an entity in the electrical carbon factor and the carbon footprint-specific knowledge graph.

9. A carbon footprint tracing method integrating knowledge graphs and spatiotemporal attention, characterized in that, It includes the following steps: Preprocessing of power grid-side control cloud and carbon emission data, energy consumption data in product-side production processes and carbon footprint traceability auxiliary data yields spatiotemporal feature datasets and original knowledge graph datasets. The preliminary electric carbon factor is obtained by using the spatiotemporal feature dataset to make predictions through a pre-trained fusion model. The preliminary electric carbon factor is then calibrated for deviation using a dynamic error correction model to obtain the final electric carbon factor. Based on the entities, relationships, and attributes of the pre-defined carbon factors and carbon footprint, knowledge elements are extracted from the original dataset of the knowledge graph, and knowledge elements are fused to construct a unique knowledge graph for carbon factors and carbon footprint. Based on the correlation between the electric carbon factor and carbon footprint in the knowledge graph dedicated to the electric carbon factor and carbon footprint, the final electric carbon factor is automatically matched with the power consumption of each production process in the product's entire life cycle. Combined with non-electric carbon emission data, the total carbon footprint of the product is calculated through the full life cycle carbon footprint accounting model. The generation chain of the total carbon footprint of the product is traced based on the aforementioned electrocarbon factor and carbon footprint-specific knowledge graph.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method of claim 9.

Images