A multi-source data fusion accounting method for the entire life cycle carbon footprint

Abstract

This invention discloses a multi-source data fusion accounting method for the entire life cycle carbon footprint, belonging to the field of carbon footprint accounting technology. Specifically, it includes: inputting a set of purchase order numbers for the target product; automatically associating logistics fuel consumption records, production process energy consumption logs, and supplier indirect emission declaration forms based on the purchase order numbers; constructing a dynamic traceability tree with the purchase order as the root node; generating a main path by matching the order material code with the workpiece batch number in the process energy consumption log; when the difference between the real-time logistics energy consumption data and the supplier declaration form exceeds a set threshold, activating a disputed data branch reconstruction algorithm to maintain the continuity of the main path while creating parallel branches to temporarily replace the disputed data segments; finally, outputting a full life cycle carbon footprint data map with dispute markers, retaining the original association paths of all data sources. This invention provides reliable support for accurately tracing the carbon emissions throughout the entire product life cycle and improving the quality of carbon footprint accounting.

Description

Technical Field

[0001] This invention relates to the field of carbon footprint accounting technology, specifically to a multi-source data fusion accounting method for the entire life cycle carbon footprint. Background Technology

[0002] Currently, various carbon footprint accounting methods have emerged in the industry, quantifying emissions by integrating internal production data, supplier-reported information, and third-party monitoring data. For example, some methods obtain process-level energy consumption data by linking energy consumption logs from the production execution system and calculate carbon emissions in the production process by combining them with preset emission factors; simultaneously, they use fuel consumption records from logistics and transportation databases to calculate emissions in the transportation process and incorporate indirect emission data reported by suppliers to refine the accounting boundaries. These methods provide enterprises with basic carbon footprint accounting tools, meeting compliance and management requirements to a certain extent.

[0003] However, existing data from multiple sources, such as process energy consumption in production systems, fuel consumption in logistics databases, and indirect emissions reported by suppliers, suffer from inconsistent formats and standards and a lack of a unified traceability framework. This makes it difficult to organically connect data across the entire chain, resulting in fragmented accounting results that cannot form a complete carbon footprint map. Furthermore, when significant discrepancies occur in multi-source data, existing methods cannot dynamically correct disputed data while maintaining accounting continuity. They are also prone to accounting interruptions or result distortions due to data conflicts, severely impacting the completeness and accuracy of life-cycle carbon footprint accounting. Summary of the Invention

[0004] The purpose of this invention is to provide a multi-source data fusion and accounting method for the entire life cycle carbon footprint, and to solve the following technical problems:

[0005] Existing multi-source data from production, logistics, and suppliers suffer from inconsistent formats and traceability frameworks, leading to difficulties in data integration and fragmented results. Furthermore, existing methods cannot dynamically correct data deviations, which can easily cause accounting interruptions or distortions, affecting the completeness and accuracy of the full life-cycle carbon footprint accounting.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] A multi-source data fusion and accounting method for the entire life cycle carbon footprint includes the following steps:

[0008] S1. Input the set of purchase order numbers for the target product, wherein the set of purchase order numbers contains the unique identifiers of suppliers throughout the entire chain from raw material mining to finished product assembly;

[0009] S2. Automatically link fuel consumption records from the logistics and transportation database, process energy consumption logs from the production execution system, and indirect emission declaration forms submitted by suppliers based on the purchase order number;

[0010] S3. Construct a dynamic traceability tree with the purchase order as the root node. The dynamic traceability tree generates the main path by matching the order material code with the workpiece batch number in the process energy consumption log.

[0011] S4. When the difference between the real-time energy consumption data in the logistics and transportation database and the supplier declaration form exceeds the set energy consumption threshold, activate the disputed data branch reconstruction algorithm; while maintaining the continuity of the main path, create a parallel calculation branch to temporarily replace the disputed data segment;

[0012] S5. Output a full life cycle carbon footprint data map with disputed markers, wherein the full life cycle carbon footprint data map retains the original association paths of all data sources.

[0013] As a further aspect of the present invention: In step S3, the process of constructing a dynamic traceability tree with the purchase order as the root node is as follows:

[0014] Parse the industry classification prefix of the material code in the purchase order and map the industry classification prefix to the preset carbon emission factor priority list; search the process energy consumption log of the production execution system in order according to the carbon emission factor priority list, and lock the workpiece processing time period containing the same material prefix in the process energy consumption log; extract the spindle power sampling sequence of all machine tools in the workpiece processing time period, and perform time dimension integration on the spindle power sampling sequence to convert it into process-level net energy consumption value.

[0015] The supplier's unique identifier is attached as a leaf node to the corresponding purchase order branch, forming a parent-child hierarchical data topology. When the same material code is detected to exist in multiple supplier links, the main path is automatically copied and a supply chain level identifier is attached. The matching process between the material code prefix and the workpiece batch number adopts a fuzzy mapping algorithm based on edit distance.

[0016] As a further aspect of the present invention: in step S4, setting the energy consumption threshold specifically involves:

[0017] Obtain the historical energy consumption record set and the corresponding supplier declaration value set for the time period from the logistics and transportation database; calculate the difference sequence between the historical energy consumption record set and the supplier declaration value set, and store the difference sequence in an independent cache area according to the transportation route.

[0018] For each type of transportation route, an independent moving window standard deviation calculation is performed on the difference sequence. The window width of the moving window standard deviation calculation adaptively expands as the number of data records increases.

[0019] The vehicle management system retrieves the vehicle's enabled year field to calculate the usage period parameter. The usage period parameter increases linearly with the number of calendar years. The result of the moving window standard deviation calculation is multiplied by the linear growth coefficient of the usage period parameter to generate the set energy consumption threshold for the current transportation route.

[0020] As a further aspect of the present invention: in step S4, the disputed data branch reconstruction algorithm specifically comprises:

[0021] Traverse the leaf node set of the dynamic source tree to locate the target node corresponding to the disputed data, copy all the computation states of the upstream path of the target node and keep the running thread active.

[0022] Based on the data type of the dispute, a preset alternative model is selected from the industry benchmark database. The preset alternative models in the industry benchmark database are indexed by transportation vehicle type and process classification.

[0023] The copied upstream path output value is input into the alternative model to generate a parallel branch data sequence. The parallel branch data sequence is then connected to the original downstream node path to form a complete computation chain. A data source change marker containing the model version number and timestamp is inserted at the target node position.

[0024] As a further aspect of the present invention: the data processing process of the alternative model is as follows:

[0025] Obtain the start and end timestamps of the disputed data segment, extract the output values ​​of adjacent nodes in the three sampling periods before the start timestamp as the reference data for the front transition zone; extract the input values ​​of downstream nodes in the three sampling periods after the end timestamp as the reference data for the back transition zone.

[0026] Linear interpolation is used to fuse the baseline data of the pre-transition zone with the head of the output sequence of the alternative model, and quadratic curve is used to smoothly connect the tail of the output sequence of the alternative model with the baseline data of the post-transition zone. The fusion connection result covers the time interval corresponding to the disputed data segment and is marked as the reconstructed data segment.

[0027] As a further solution of the present invention: when the same supplier's unique identifier appears in multiple purchase order branches, a dispute data association index is established, the declaration data sequence of the supplier in all purchase order branches is extracted, and the dynamic time-normalized distance matrix between each sequence is calculated.

[0028] When the minimum eigenvalue of the dynamic time-normalized distance matrix is ​​lower than the benchmark of similar suppliers, the cross-chain substitution model generation program is activated. The cross-chain substitution model generation program aggregates the upstream path output values ​​of multiple purchase order branches, trains and generates a joint substitution model, and uses the output of the joint substitution model to replace the original parallel calculation branch data sequence and update the data source change flag.

[0029] As a further aspect of the present invention: in step S5, the process of generating the full life cycle carbon footprint data map is as follows:

[0030] The parent-child node relationship of the dynamic traceability tree is converted into directed graph vertices and directed graph edges. A data source type attribute field is added to each directed graph vertex. The data source type attribute field contains three types of codes: enterprise measured data identifier, supplier declaration data identifier, and reconstructed branch data identifier.

[0031] When a directed graph vertex contains a data source change marker, add a red topological break symbol at the connection of the corresponding directed graph edge; arrange the positions of the directed graph vertices in the order of raw material mining stage vertices, component production stage vertices, logistics and transportation stage vertices, and product use stage vertices.

[0032] The trajectory line width is divided into five levels based on the range of carbon emission values, with the trajectory line width level increasing stepwise as the carbon emission value increases.

[0033] As a further aspect of the present invention: the visualization processing of the full life cycle carbon footprint data map is as follows:

[0034] Parse the model version number and timestamp in the data source change mark, match the model type description document in the industry benchmark database according to the model version number, and convert the timestamp into a readable date format with local time and display it.

[0035] An interactive hotspot that responds to mouse click events is embedded at the red topological fracture symbol. When a mouse click event is triggered, a two-layer data card is loaded. The first layer of the two-layer data card displays a numerical comparison curve between the parallel calculation branch data sequence and the original disputed data segment. The second layer of the two-layer data card displays a summary of the model type description document and data generation time information.

[0036] The beneficial effects of this invention are:

[0037] This invention constructs a dynamic traceability tree with purchase orders as the root node, utilizes fuzzy mapping between material codes and process energy consumption logs, and identifies supply chain levels to achieve organic connection of multi-source data across the entire chain from raw materials to finished products. This solves the problems of weak data correlation and fragmented accounting results in existing technologies. By setting an adaptive energy consumption threshold based on historical data and the service life of transportation vehicles, a disputed data branch reconstruction algorithm is activated when data differences exceed the limit. This algorithm replicates the upstream calculation state and adopts an industry benchmark substitution model, combining pre- and post-processing data for smooth fusion. This maintains accounting continuity while dynamically correcting disputed data, overcoming the defects of accounting interruption or distortion due to data deviation. By generating a full life-cycle carbon footprint data map with disputed markers, the data source association path is converted into a directed graph containing type attributes. Interactive hotspots are embedded to display data comparison and model information, fully preserving the original calculation logic and improving the transparency, credibility, and auditability of the results. Furthermore, the cross-chain substitution model generation program optimizes data consistency for multiple orders from the same supplier, further strengthening the completeness and accuracy of full life-cycle carbon footprint accounting. Attached Figure Description

[0038] The invention will now be further described with reference to the accompanying drawings.

[0039] Figure 1 This is a flowchart illustrating the present invention. Detailed Implementation

[0040] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] Please see Figure 1 As shown, this invention provides a multi-source data fusion and accounting method for the entire life cycle carbon footprint, comprising the following steps:

[0042] S1. Input a set of purchase order numbers for the target product. This set covers the entire product chain from raw material mining to finished product assembly, including unique supplier identifiers for each stage. These identifiers are used to accurately locate each supplier involved in the production process, ensuring that subsequent data traceability covers all nodes of the entire lifecycle.

[0043] S2. Based on the input purchase order number, the system automatically links to multiple data sources. These include fuel consumption information recorded in the logistics and transportation database, energy consumption logs for each process stored in the production execution system, and indirect emissions declaration forms submitted by suppliers. Through the purchase order number as a key identifier, the system automatically matches data from different stages, integrating scattered information such as transportation, production, and indirect emissions into a unified accounting framework.

[0044] S3. Construct a dynamic traceability tree with the purchase order as the root node. The main path of this traceability tree is generated by matching the order material code with the workpiece batch number in the process energy consumption log of the production execution system. The system will parse the industry classification information in the material code, associate it with the corresponding production process, and extract the energy consumption records of related equipment within that process, forming a data mainline that runs through the entire production process. When the same material code appears in multiple supplier stages, the system will automatically copy the main path and add a supply chain level identifier to ensure clear association of data across multiple levels.

[0045] S4. During data processing, if the difference between the real-time energy consumption data in the logistics and transportation database and the data in the declaration form submitted by the supplier exceeds the set energy consumption threshold, the system will automatically activate the disputed data branch reconstruction algorithm. This algorithm creates parallel calculation branches to temporarily replace the disputed data segments while maintaining the continuity of the main data path. The data source for the parallel calculation branches comes from a preset model in the industry benchmark database, ensuring that the calculation of disputed data segments can continue without interrupting the overall process.

[0046] S5. The final output is a full life-cycle carbon footprint data map with disputed data markers. This map fully preserves the original correlation paths of all data sources, clearly showing the flow of carbon emission data from raw material mining to finished product assembly. For disputed data segments that have been reconstructed through branching, the map will distinguish them with specific markers, while recording the source and processing of data replacements to ensure the traceability and transparency of the entire accounting result.

[0047] In step S3, when constructing a dynamic traceability tree with the purchase order as the root node, the material codes contained in the purchase order are first parsed to extract the industry classification prefixes at the top of the codes. These prefixes are used to distinguish the industry sectors to which the materials belong, such as electronic components, metal materials, etc. The system maps the parsed industry classification prefixes to a preset carbon emission factor priority list, which pre-sets the order of data retrieval based on the carbon emission characteristics of different industries.

[0048] Based on the carbon emission factor priority list obtained through mapping, the system sequentially retrieves the process energy consumption logs stored in the production execution system, filtering out records containing the same material prefix to pinpoint the corresponding workpiece processing time period for that material during production. Within the determined processing time period, the system extracts the spindle power sampling sequences of all machine tools involved in production; these sequences represent real-time power data collected during machine tool operation. By performing time-dimensional cumulative calculations on the sampling sequences, the instantaneous power data is converted into the net energy consumption value of that process over the entire processing time period.

[0049] After integrating production data, the system uses the unique supplier identifiers for each stage as leaf nodes, attaching them to the branches derived from the purchase order. This forms a clear parent-child hierarchical data topology, intuitively presenting the relationship between purchase orders and suppliers at each level. When the same material code is detected appearing in multiple supplier stages, the system automatically copies the current main path and adds supply chain level identifiers, such as "Tier 1 Supplier" and "Tier 2 Supplier," to the copied path to distinguish between different levels of supply chain data.

[0050] When matching the material code prefix with the workpiece batch number in the process energy consumption log, the system uses a fuzzy mapping algorithm based on edit distance. This algorithm can handle subtle differences that may exist in the code, such as the addition or deletion of characters, or the adjustment of order. It determines the matching degree by calculating the degree of character difference between the two, ensuring accurate association even if there are non-critical errors in the code.

[0051] In step S4, setting the energy consumption threshold specifically involves:

[0052] First, a set of historical energy consumption records needs to be extracted from the logistics and transportation database, along with the set of supplier declaration values ​​for the corresponding time periods. During extraction, the timestamp of each historical energy consumption record must be verified one by one to ensure a perfect correspondence with the time interval of the supplier declaration values, avoiding distortion in subsequent difference calculations due to time misalignment. Then, the difference between the historical energy consumption records and the corresponding supplier declaration values ​​is calculated, forming a continuous difference sequence. When classifying these sequences by transportation route, the routes must first be standardized and coded to distinguish different types such as trunk transportation and branch line delivery, ensuring that difference sequences for the same type of route are centrally stored in an independent cache, laying the foundation for subsequent targeted analysis.

[0053] For each type of transportation route, an independent moving window standard deviation calculation is required. This calculation analyzes the fluctuation of the difference within the window by dynamically defining the data window. The width of the window is not a fixed value, but rather expands adaptively as the number of data records for that route increases. When more than 10 new historical records are added to a route, the window width automatically increases by one unit, allowing more historical data to be included in the calculation, thus more accurately reflecting the long-term fluctuation trend.

[0054] Next, the vehicle management system needs to be retrieved to determine the year the transportation vehicles used on the route were put into operation, and the service life parameter is calculated accordingly. The calculation is based on the calendar year, subtracting the year of activation from the current year and taking the integer. For example, if the year of activation was 2018 and the current year is 2025, the service life parameter is 7. This parameter increases linearly with the increase in the service time of the transportation vehicles, thus reflecting the potential impact of equipment aging on energy consumption fluctuations.

[0055] Finally, the result of the moving window standard deviation calculation is multiplied by the linear growth coefficient corresponding to the service life parameter to obtain the set energy consumption threshold for the current transportation route. The linear growth coefficient gradually increases with the service life, reflecting the greater energy consumption fluctuations of older transportation vehicles. This allows the threshold setting to dynamically adapt to changes caused by equipment aging, better reflecting the energy consumption fluctuation patterns of actual transportation scenarios.

[0056] In step S4, the disputed data branch reconstruction algorithm is specifically as follows:

[0057] First, the leaf node set of the dynamic traceability tree needs to be traversed to locate the target node corresponding to the disputed data. The leaf node set contains the data source information at the very end of each stage of the entire lifecycle. The traversal process checks the data matching status of each node hierarchically from bottom to top. When the difference between the real-time energy consumption data of logistics transportation and the supplier's declared data of a certain node exceeds a set threshold, that node is locked as the target node. After the location is completed, all the calculation status of the upstream path of the target node needs to be copied, including the energy consumption conversion parameters of each process, historical data correlations, intermediate calculation results, etc. At the same time, the running threads of these upstream paths need to be kept active. This means that real-time updates of upstream data can still be continuously transmitted to the calculation link, ensuring that the subsequent reconstruction process will not be halted due to data interruption.

[0058] Next, based on the specific type of disputed data, a pre-defined alternative model is selected from the industry benchmark database. This database encompasses standardized data models for different scenarios, with clear indexes organized by transportation vehicle type and process category. For example, the transportation vehicle type index includes heavy-duty trucks, container ships, and air freight, while the process category index covers metal smelting, electronic component assembly, and plastic molding. When the disputed data pertains to road transportation, the "heavy-duty truck" category is matched from the transportation vehicle type index, and then, combined with specific parameters such as transportation distance and load capacity, the most suitable alternative model is selected.

[0059] Subsequently, the copied upstream path output values ​​are input into the selected alternative model to generate parallel computation branch data sequences. These output values ​​contain the cumulative energy consumption and emissions data of all upstream links of the target node. The alternative model generates sequence data with the same format as the original data by simulating energy consumption changes under similar scenarios. The generated parallel branch data sequences are then connected to the original downstream node path to continue the computational link with the downstream data, forming a complete carbon footprint accounting chain. Simultaneously, a data source change marker is inserted at the target node location. This marker includes the version number of the alternative model and a generation timestamp. The version number distinguishes the model's iterative update status, while the timestamp precisely records the exact moment the model was activated, providing a basis for subsequent data traceability.

[0060] The data processing of the replacement model requires a rigorous transition and connection process. First, the start and end timestamps of the disputed data segment are obtained to determine the data interval that needs to be replaced. Then, the output values ​​of adjacent nodes in the three sampling periods preceding the start timestamp are extracted as baseline data for the pre-transition zone. This data reflects the stable state before the start of the disputed data segment. For example, if a disputed data segment in a transportation process begins at 8:00, energy consumption data at 7:45, 7:50, and 7:55 are extracted as baselines. Simultaneously, the input values ​​of downstream nodes in the three sampling periods following the end timestamp are extracted as baseline data for the post-transition zone. For instance, if the disputed data segment ends at 9:00, input data at 9:05, 9:10, and 9:15 are extracted.

[0061] Next, the baseline data from the pre-transition zone is linearly interpolated and fused with the first part of the output sequence of the alternative model. By gradually adjusting the data slope, a smooth transition from the original data to the model output is achieved, avoiding subsequent calculation deviations caused by data abrupt changes. For the tail of the output sequence of the alternative model, a quadratic curve is used to smoothly connect it with the baseline data from the post-transition zone. Through curve fitting, the model output gradually conforms to the downstream actual data, ensuring the continuity of the entire data chain. After the fusion and connection are completed, the resulting data will cover the time interval corresponding to the disputed data segment and will be clearly marked as the reconstructed data segment, forming a clear distinction from the original data.

[0062] When the same supplier's unique identifier appears in multiple purchase order branches, a disputed data association index needs to be established. This association index links data sequences involving that supplier across different purchase order branches using the supplier's unique identifier, forming a unified data management unit. It is necessary to extract the supplier's declared data sequences across all purchase order branches and calculate the dynamic time-normalized distance matrix between these sequences. This matrix measures the similarity of different sequences over time. For example, the matching degree of energy consumption fluctuation patterns in the same supplier's raw material transportation data for two orders can be reflected by the matrix values.

[0063] When the minimum eigenvalue of the dynamic time-normalized distance matrix is ​​lower than that of similar suppliers, it indicates significant inconsistencies in the supplier's declared data across different order branches. In this case, the cross-chain substitution model generation program needs to be activated. This program aggregates upstream path output values ​​from multiple procurement order branches. These values ​​contain data on the supplier's upstream processes in different orders, such as raw material procurement volume and transportation origin. By integrating this data, the program trains and generates a joint substitution model, enabling the model to comprehensively reflect the supplier's data characteristics across multiple scenarios. Finally, the output of the joint substitution model replaces the original parallel calculation branch data sequence, updates the data source change markers, and supplements the joint model's training parameters and data coverage to ensure the reconstructed data better reflects the supplier's actual operational status.

[0064] In S5, the process of generating the full life cycle carbon footprint data map is as follows:

[0065] First, the hierarchical relationship between parent and child nodes in the dynamic source tree is converted into vertices and edges of a directed graph. In the dynamic source tree, the parent node represents the summary data of the previous stage, and the child node corresponds to the specific data of the subdivided stage. This relationship is mapped to the connecting lines between vertices in the directed graph—each vertex corresponds to a data node, and the directed graph edges indicate the direction of data flow from the upstream stage to the downstream stage. During the conversion process, a data source type attribute field needs to be added to each directed graph vertex. This field distinguishes the original source of the data through specific codes: the enterprise measured data identifier is used to mark data directly collected by internal equipment such as production execution systems and IoT sensors, such as machine tool energy consumption and factory gas consumption; the supplier declaration data identifier corresponds to the indirect emission declaration form information submitted by the supplier, such as carbon emissions in the raw material production process; and the reconstructed branch data identifier specifically marks the alternative data after being processed by the disputed data branch reconstruction algorithm, such as the model output value replaced due to data conflict in the transportation stage. The three types of codes are distinguished by different character combinations to ensure that users can quickly identify the data source of each vertex.

[0066] If a directed graph vertex contains a data source change marker, indicating that the data at that node has been reconstructed using a replacement model, a red topological break symbol should be added at the connection point of the corresponding directed graph edge. This symbol is jagged, contrasting sharply with normal smooth lines, visually indicating that the data flow has been interrupted and reconstructed, helping users quickly locate the adjusted stage. Simultaneously, the arrangement of directed graph vertices must follow the temporal logic of the entire lifecycle: vertices are arranged sequentially in the order of raw material mining stage, component production stage, logistics and transportation stage, and product use stage. Vertices from each stage are grouped together, with groups separated by blank areas. For example, vertices from the raw material mining stage will cluster on the left side of the graph, including sub-nodes such as ore mining and plant cultivation; vertices from the component production stage follow closely, covering process nodes such as forging and injection molding, thus forming a linear flow from left to right, closely matching the actual flow path of the product from source to finished product.

[0067] Furthermore, the width of the trajectory line needs to be divided into five levels based on the range of carbon emission values. The trajectory line is a visual representation of the directed graph edges, and the level division is based on the actual carbon emission values: the smallest value corresponds to the thinnest line, and as the value increases, the line width increases in a fixed gradient, reaching its widest point at the highest value range. This design allows users to quickly identify high-emission stages throughout the entire lifecycle without needing to read specific values, simply by observing the line thickness. For example, if the trajectory line is significantly thicker in the logistics and transportation stage than in other stages, it can be directly identified as a key target for carbon emission control.

[0068] The visualization of the full life-cycle carbon footprint data map further enhances the readability and interactivity of the data, specifically including:

[0069] The system parses the model version number and timestamp from the data source change marker. The model version number is a string of letters and numbers used to identify the iterative version of the replacement model. For example, "T-2023v3" represents the third iteration of the transportation model in 2023. The system automatically searches the industry benchmark database based on this version number and matches the corresponding model type description document, which contains core information such as the model's applicable scenarios, parameter settings, and training data sources. The timestamp is a sequence of numbers accurate to the second. The system converts it into a readable format of "year-month-day-hour:minute:second" and simultaneously indicates the local time zone, such as "2025-06-12-09:45:30 Beijing Time," allowing users to clearly understand the specific time of data reconstruction.

[0070] An interactive hotspot is embedded at the red topological fracture symbol, responding to mouse click events. The hotspot is a rectangular area surrounding the symbol; when the mouse pointer moves over this area, the pointer style changes to a hand, indicating that the user can interact with it. Clicking the hotspot loads a two-layer data card: the first layer displays a numerical comparison curve between the parallel computation branch data sequence and the original disputed data segment. The curve is plotted with time on the horizontal axis and carbon emissions on the vertical axis. Blue lines represent the parallel branch data output by the alternative model, and gray lines represent the original disputed data. The two lines are presented side-by-side on the same coordinate system for easy and intuitive comparison of data differences. The second layer displays a summary of the model type description document and data generation time information. The summary extracts key information from the document, such as the model's input parameters and applicable transportation types. The data generation time is the converted, readable timestamp, helping users to deeply understand the generation logic of the alternative data.

[0071] Through such generation and visualization, the full life cycle carbon footprint data map can not only completely preserve the original correlation path of the data, but also make the complex carbon footprint data intuitive and interpretable through structured design and interactive functions, providing users with clear visual guidance for tracing emission sources and analyzing high-carbon links.

[0072] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A multi-source data fusion and accounting method for the entire life cycle carbon footprint, characterized in that, Includes the following steps: S1. Input the set of purchase order numbers for the target product, wherein the set of purchase order numbers contains the unique identifiers of suppliers throughout the entire chain from raw material mining to finished product assembly; S2. Automatically link fuel consumption records from the logistics and transportation database, process energy consumption logs from the production execution system, and indirect emission declaration forms submitted by suppliers based on the purchase order number; S3. Construct a dynamic traceability tree with the purchase order as the root node. The dynamic traceability tree generates the main path by matching the order material code with the workpiece batch number in the process energy consumption log. S4. When the difference between the real-time energy consumption data in the logistics and transportation database and the supplier declaration form exceeds the set energy consumption threshold, activate the disputed data branch reconstruction algorithm; while maintaining the continuity of the main path, create a parallel calculation branch to temporarily replace the disputed data segment; S5. Output a full life cycle carbon footprint data map with disputed markers, wherein the full life cycle carbon footprint data map retains the original association paths of all data sources; In S3, the process of constructing a dynamic traceability tree with the purchase order as the root node is as follows: Parse the industry classification prefix of the material code in the purchase order and map the industry classification prefix to a preset carbon emission factor priority list; search the process energy consumption log of the production execution system in order according to the carbon emission factor priority list, and lock the processing time of the workpiece containing the same material prefix in the process energy consumption log; Extract the spindle power sampling sequence of all machine tools during the workpiece processing period, and perform time-dimensional integration on the spindle power sampling sequence to convert it into process-level net energy consumption value. The supplier's unique identifier is attached as a leaf node to the corresponding purchase order branch, forming a parent-child hierarchical data topology. When the same material code is detected to exist in multiple supplier links, the main path is automatically copied and a supply chain level identifier is attached. The matching process between the material code prefix and the workpiece batch number adopts a fuzzy mapping algorithm based on edit distance; In step S4, the disputed data branch reconstruction algorithm is specifically as follows: Traverse the leaf node set of the dynamic source tree to locate the target node corresponding to the disputed data, copy all the computation states of the upstream path of the target node and keep the running thread active. Based on the data type of the dispute, a preset alternative model is selected from the industry benchmark database. The preset alternative models in the industry benchmark database are indexed by transportation vehicle type and process classification. The copied upstream path output value is input into the alternative model to generate a parallel branch data sequence. The parallel branch data sequence is then connected to the original downstream node path to form a complete computation chain. A data source change marker containing the model version number and timestamp is inserted at the target node position.

2. The multi-source data fusion and accounting method for the entire life cycle carbon footprint according to claim 1, characterized in that, In step S4, setting the energy consumption threshold specifically involves: Obtain the historical energy consumption record set and the corresponding supplier declaration value set for the time period from the logistics and transportation database; calculate the difference sequence between the historical energy consumption record set and the supplier declaration value set, and store the difference sequence in an independent cache area according to the transportation route. For each type of transportation route, an independent moving window standard deviation calculation is performed on the difference sequence. The window width of the moving window standard deviation calculation adaptively expands as the number of data records increases. The vehicle management system retrieves the vehicle's enabled year field to calculate the usage period parameter. The usage period parameter increases linearly with the number of calendar years. The result of the moving window standard deviation calculation is multiplied by the linear growth coefficient of the usage period parameter to generate the set energy consumption threshold for the current transportation route.

3. The multi-source data fusion and accounting method for the entire life cycle carbon footprint according to claim 1, characterized in that, The process of data processing in the alternative model is as follows: Obtain the start and end timestamps of the disputed data segment, extract the output values ​​of adjacent nodes in the three sampling periods before the start timestamp as the reference data for the front transition zone; extract the input values ​​of downstream nodes in the three sampling periods after the end timestamp as the reference data for the back transition zone. Linear interpolation is used to fuse the baseline data of the pre-transition zone with the head of the output sequence of the alternative model, and quadratic curve is used to smoothly connect the tail of the output sequence of the alternative model with the baseline data of the post-transition zone. The fusion connection result covers the time interval corresponding to the disputed data segment and is marked as the reconstructed data segment.

4. The multi-source data fusion and accounting method for the entire life cycle carbon footprint according to claim 1, characterized in that, When the same supplier's unique identifier appears in multiple purchase order branches, a dispute data association index is established, the supplier's declaration data sequence in all purchase order branches is extracted, and the dynamic time-normalized distance matrix between each sequence is calculated. When the minimum eigenvalue of the dynamic time-normalized distance matrix is ​​lower than the benchmark of similar suppliers, the cross-chain substitution model generation program is activated. The cross-chain substitution model generation program aggregates the upstream path output values ​​of multiple purchase order branches, trains and generates a joint substitution model, and uses the output of the joint substitution model to replace the original parallel calculation branch data sequence and update the data source change flag.

5. The multi-source data fusion and accounting method for the entire life cycle carbon footprint according to claim 1, characterized in that, In S5, the process of generating the full life cycle carbon footprint data map is as follows: The parent-child node relationship of the dynamic traceability tree is converted into directed graph vertices and directed graph edges. A data source type attribute field is added to each directed graph vertex. The data source type attribute field contains three types of codes: enterprise measured data identifier, supplier declaration data identifier, and reconstructed branch data identifier. When a directed graph vertex contains a data source change marker, add a red topological break symbol at the connection of the corresponding directed graph edge; arrange the positions of the directed graph vertices in the order of raw material mining stage vertices, component production stage vertices, logistics and transportation stage vertices, and product use stage vertices. The trajectory line width is divided into five levels based on the range of carbon emission values, with the trajectory line width level increasing stepwise as the carbon emission value increases.

6. The multi-source data fusion and accounting method for the entire life cycle carbon footprint according to claim 5, characterized in that, The visualization processing of the full life cycle carbon footprint data map is as follows: Parse the model version number and timestamp in the data source change mark, match the model type description document in the industry benchmark database according to the model version number, and convert the timestamp into a readable date format with local time and display it. An interactive hotspot that responds to mouse click events is embedded at the red topological fracture symbol. When a mouse click event is triggered, a two-layer data card is loaded. The first layer of the two-layer data card displays a numerical comparison curve between the parallel calculation branch data sequence and the original disputed data segment. The second layer of the two-layer data card displays a summary of the model type description document and data generation time information.

Images