Carbon footprint accounting method and system based on full life cycle multi-dimensional analysis

Abstract

This invention provides a carbon footprint accounting method and system based on multi-dimensional analysis of the entire life cycle, relating to the field of carbon emission accounting technology. The method includes: establishing a material transformation chain and extracting process nodes by acquiring material flow data; matching an emission factor library to obtain preliminary emissions; performing transitive tracing based on node dependencies to obtain dependency transmission paths and calculating path weights for weighted accumulation to obtain extended emissions; generating a multi-dimensional emission feature matrix based on associated attributes and identifying key factors to construct a factor correlation graph; performing topological analysis to extract core impact paths and tracing back to source nodes to obtain a set of source attributes; calculating dynamic adjustment coefficients and iteratively correcting extended emissions; and calculating corrected weights based on core impact paths and factor correlation graphs and redistributing them to obtain the carbon footprint accounting result.

Description

Technical Field

[0001] This invention relates to the field of carbon emission accounting technology, and in particular to a carbon footprint accounting method and system based on multi-dimensional analysis of the entire life cycle. Background Technology

[0002] Carbon footprint accounting is an important tool for assessing greenhouse gas emissions throughout a product's entire life cycle and is widely used in environmental impact assessments and sustainable development decision-making. Current conventional practices are primarily based on the life cycle assessment framework, collecting material and energy input and output data from stages such as raw material acquisition, production, transportation, use, and disposal, and then conducting inventory analysis using a standard emission factor database. During the accounting process, direct emissions from each stage are typically summed up item by process category to obtain the final total carbon emissions.

[0003] However, existing accounting methods fail to fully reflect the complex dependencies between process nodes in a production network. In a real production chain, the output of one process may serve as the input for multiple downstream processes, resulting in indirect emissions contributions across stages, such as the implicit impact of upstream energy consumption on midstream manufacturing. Existing linear summation methods assume that the emissions of each process are independent contributions, ignoring the cumulative effect of such indirect transmission paths. This leads to a systematic underestimation of actual carbon emissions in the accounting results, especially for systems with long supply chains or recycling characteristics, where the error is more pronounced.

[0004] Furthermore, existing accounting processes rely on pre-set fixed emission factors and lack dynamic response mechanisms to changes in key influencing factors over time or under different operating conditions. In actual production, factors such as process parameters, raw material sources, and energy structure fluctuate continuously. Small deviations in initial material flow data can be amplified through transmission paths, and existing methods cannot automatically identify the sensitivity of source attributes to the overall accounting results. The accounting results cannot self-correct or adjust, making it difficult to adapt to the accuracy requirements of different scenarios, thus limiting the reliability of carbon emission reduction strategy optimization and priority decision-making. Summary of the Invention

[0005] This invention provides a carbon footprint accounting method and system based on multi-dimensional analysis of the entire life cycle, which can at least solve some of the problems existing in the prior art.

[0006] A first aspect of this invention provides a carbon footprint accounting method based on multi-dimensional analysis throughout the entire life cycle, comprising:

[0007] Acquire material flow data of the target object during its complete life cycle, establish a material conversion chain based on the material flow data and extract process nodes, and match and query the process nodes with a preset emission factor library to obtain preliminary emission amounts;

[0008] Based on the dependencies between process nodes, the initial emissions are traced transitively to obtain the dependency transmission path. Cross-stage indirect contribution paths in the dependency transmission path are identified and path weights are calculated. The initial emissions are weighted and accumulated based on the path weights to obtain the extended emissions.

[0009] A multidimensional emission feature matrix is ​​generated by associating and labeling the extended emissions with the attributes of the process nodes. Key factors are identified and a factor correlation map is constructed based on the multidimensional emission feature matrix. Topological analysis is performed on the factor correlation map to extract the core influence path. The source node of the material conversion chain is traced back along the core influence path to obtain the source attribute set. The dynamic adjustment coefficient is calculated based on the response relationship between the source attribute set and the extended emissions.

[0010] The extended emissions and the dynamic adjustment coefficient are iteratively corrected to obtain the corrected emissions. The corrected weights are calculated based on the core impact path and the factor correlation map, and the corrected emissions are redistributed to obtain the carbon footprint accounting result and output.

[0011] In one alternative implementation,

[0012] Acquire material flow data of the target object throughout its complete life cycle, establish a material transformation chain based on the material flow data and extract process nodes, and match and query the process nodes with a preset emission factor database to obtain preliminary emission amounts, including:

[0013] The material input and output records of the target object throughout its entire life cycle are collected and the material flow data is extracted. A directed graph is constructed on the material flow data according to the material flow direction. The circular flow path in the directed graph is identified and the loop is broken to obtain a directed acyclic graph. The predecessor and successor relationship is established based on the directed acyclic graph to obtain the material conversion chain. The material conversion chain is topologically sorted and the node attributes after sorting are extracted to obtain the process node set.

[0014] The node feature combination associated with each process node is extracted from the set of process nodes. The node feature combination is semantically embedded and mapped to obtain a feature vector. The feature vector is then compared with the factor feature vector in the preset emission factor library to obtain a similarity distribution. Based on the similarity distribution, a set of candidate factors with similarity exceeding a preset similarity threshold is selected. The emission factors in the candidate factor set are then subjected to secondary verification according to applicable boundary conditions to obtain the optimal emission factor. The optimal emission factor is then jointly calculated with the quantity of substances associated with the process node and the conversion efficiency coefficient to obtain the node emission amount. The node emission amounts are then accumulated along the topological order of the substance conversion chain to obtain the preliminary emission amount.

[0015] In one alternative implementation,

[0016] Based on the dependencies between process nodes, the initial emissions are traced transitively to obtain dependency propagation paths. Cross-stage indirect contribution paths within these paths are identified and their weights are calculated. The initial emissions are then weighted and accumulated based on these weights to obtain extended emissions, including:

[0017] The predecessor and successor connections between the process nodes are extracted from the material conversion chain. A dependency graph is constructed based on the predecessor and successor connections, and the stage identifier corresponding to each process node is marked. The dependency graph is traversed in reverse depth-first order. The complete path from each process node back to all predecessor nodes is recorded, and the corresponding material flow and energy conversion loss are extracted to determine the dependency transmission path set. For each path in the dependency transmission path set, the stage identifier of the adjacent nodes is detected along the path direction. When a change in the stage identifier is detected, the current node pair is marked as a cross-stage connection point. The path containing the cross-stage connection point is extracted to obtain the cross-stage indirect contribution path.

[0018] The loss ratio of mass flow at each node in the cross-stage indirect contribution path is calculated and the cumulative mass transfer efficiency is solved. The number of cross-stage connection points traversed by the cross-stage indirect contribution path is counted and the path weight is calculated in combination with the cumulative mass transfer efficiency. The initial emission amount corresponding to each process node on the cross-stage indirect contribution path is correlated with the path weight to obtain the path indirect emission contribution amount. The path indirect emission contribution amounts of all cross-stage indirect contribution paths are summarized and combined with the initial emission amount to obtain the extended emission amount.

[0019] In one alternative implementation,

[0020] The process involves associating and labeling the extended emissions with the attributes of the process nodes to generate a multidimensional emission feature matrix. Based on this multidimensional emission feature matrix, key factors are identified and a factor correlation map is constructed, including:

[0021] Extract the material type and process parameters corresponding to the process node as node attributes, map the node attributes to the corresponding extended emissions and construct an attribute emission association table, expand each node in the attribute emission association table according to the dimension of the node attribute and generate a node feature vector, arrange the node feature vectors into row vectors according to the topological order of the material conversion chain and combine them into a matrix to obtain a multidimensional emission feature matrix.

[0022] Singular value decomposition is performed on the multidimensional emission feature matrix to obtain singular values ​​in the direction of the principal components in the feature space. Based on the distribution of the singular values, principal components whose contribution exceeds a preset contribution threshold are identified and their corresponding feature vector components are extracted. Loading coefficients of each dimension in the feature vector components are calculated, and a set of key factors is determined based on the loading coefficients. Attribute columns corresponding to the set of key factors are extracted from the multidimensional emission feature matrix, and Pearson correlation coefficients between different attribute columns are calculated. Edge weights between key factors are constructed based on the Pearson correlation coefficients. The set of key factors is used as nodes, and connection relationships are established by combining the edge weights to obtain a factor correlation graph.

[0023] In one alternative implementation,

[0024] Topological analysis is performed on the factor correlation map to extract the core influence path. The core influence path is then traced back to the source node of the material transformation chain to obtain the source attribute set. Based on the response relationship between the source attribute set and the extended emissions, a dynamic adjustment coefficient is calculated, including:

[0025] The edge weights in the factor association graph are normalized and a weighted adjacency matrix is ​​constructed. The betweenness centrality of each node is calculated based on the weighted adjacency matrix. Nodes whose betweenness centrality exceeds a preset centrality threshold are identified as hub nodes. Random walk sampling is performed on the weighted adjacency matrix starting from the hub nodes and the access frequency is recorded. Power iteration is performed based on the access frequency until convergence is obtained to obtain a steady-state distribution vector. The set of core influence paths is determined based on the steady-state distribution vector.

[0026] The path-end nodes are extracted from the core impact path set and mapped to the material conversion chain. The preceding connections of the material conversion chain are traced back to the source nodes, and the corresponding material types and process parameters are extracted as source attributes to construct a source attribute set. For the source nodes, the loss ratio of material flow in each conversion stage is extracted along the forward path, and the cumulative attenuation coefficient is calculated. Based on the cumulative attenuation coefficient and the extended emissions of each process node on the corresponding path, the transmission contribution is obtained. A linear fit is performed on the source attributes and the transmission contribution to obtain regression coefficients. A perturbation is applied to the regression coefficients, and the rate of change of the transmission contribution is calculated to obtain the marginal response intensity. Based on the marginal response intensity and the transmission contribution, the dynamic adjustment coefficient is obtained.

[0027] In one alternative implementation,

[0028] The loss ratio of mass flow in each stage of conversion is extracted from the source node along the forward path, and the cumulative attenuation coefficient is calculated. Based on the cumulative attenuation coefficient and the extended emissions of each process node on the corresponding path, the transmission contribution is obtained, including:

[0029] Starting from the source node, extract the forward material flow path along the subsequent connection relationship of the material conversion chain, calculate the node entropy increase loss corresponding to each process node in the forward material flow path, calculate the effective energy loss ratio of material flow in single-stage conversion based on the material input-output relationship between adjacent nodes and the entropy increase loss to obtain the single-stage loss coefficient, and perform geometric mean aggregation on the single-stage loss coefficient and combine it with a preset path length correction factor to obtain the cumulative attenuation coefficient.

[0030] The cumulative attenuation coefficient is quantified using the Markov chain Monte Carlo sampling method to obtain the attenuation coefficient distribution range. The expected value of the attenuation coefficient distribution range is correlated with the extended emissions element by element to obtain the attenuation contribution component. The attenuation contribution component is accumulated and aggregated along the positive material flow path and solved by combining a preset time discount factor to obtain the path transmission contribution. All positive material flow paths originating from the source node are counted and the corresponding path transmission contributions are extracted to construct a contribution set. The path transmission contributions in the contribution set are nonlinearly mapped and aggregated according to the path topology complexity and the number of material cycles to obtain the transmission contribution of the source node.

[0031] In one alternative implementation,

[0032] The extended emissions and the dynamic adjustment coefficient are iteratively corrected to obtain the corrected emissions. Based on the core impact path and the factor correlation map, the correction weights are calculated, and the corrected emissions are redistributed to obtain the carbon footprint accounting results, which are then output as follows:

[0033] The extended emissions are arranged into emission vectors according to the process nodes. The dynamic adjustment coefficients are propagated along the core influence path to obtain the node adjustment coefficients. The initial corrected emission vector is obtained by combining the emission vectors. The deviation between the initial corrected emission vector and the emission vector is calculated and iteratively corrected based on a preset convergence threshold to obtain the corrected emissions.

[0034] Edge weights are extracted from the factor association graph and a weighted adjacency matrix is ​​constructed. The weighted adjacency matrix is ​​normalized based on the access frequency of nodes in the core influence path to obtain a transition probability matrix. The steady-state eigenvector of the transition probability matrix is ​​calculated and nonlinearly mapped by combining it with the betweenness centrality of nodes in the core influence path to obtain a corrected weight vector.

[0035] Based on the corrected emissions and the corresponding elements in the corrected weight vector, the allocation benchmark is calculated. The cumulative distribution of the allocation benchmark in each stage of the material conversion chain is calculated and emission concentration areas are identified. The corrected emissions in the emission concentration areas are proportionally redistributed according to the proportion of material flow to obtain a centralized allocation result. Non-concentrated areas are directly allocated according to the corrected weight vector to obtain a discrete allocation result. The discrete allocation result and the centralized allocation result are merged to obtain the carbon footprint accounting result and output.

[0036] A second aspect of this invention provides a carbon footprint accounting system based on multi-dimensional analysis throughout the entire life cycle, comprising:

[0037] The preliminary calculation unit is used to acquire material flow data of the target object during its complete life cycle, establish a material conversion chain based on the material flow data and extract process nodes, and match and query the process nodes with a preset emission factor library to obtain the preliminary emission amount.

[0038] The path weighting unit is used to trace the initial emissions based on the dependencies between the process nodes to obtain the dependency transmission path, identify the cross-stage indirect contribution paths in the dependency transmission path and calculate the path weights, and perform weighted summation of the initial emissions based on the path weights to obtain the extended emissions.

[0039] The graph source tracing unit is used to associate and label the extended emissions with the attributes of the process nodes to generate a multidimensional emission feature matrix, identify key factors based on the multidimensional emission feature matrix and construct a factor association graph, perform topological analysis on the factor association graph to extract the core influence path, trace back along the core influence path to the source node of the material conversion chain to obtain the source attribute set, and calculate the dynamic adjustment coefficient based on the response relationship between the source attribute set and the extended emissions.

[0040] The iterative correction unit is used to iteratively correct the extended emissions and the dynamic adjustment coefficient to obtain the corrected emissions, calculate the correction weight based on the core influence path and the factor correlation map, and redistribute the corrected emissions to obtain the carbon footprint accounting result and output it.

[0041] A third aspect of the present invention provides an electronic device, comprising:

[0042] A processor and a memory for storing processor-executable instructions, wherein the processor is configured to invoke instructions stored in the memory to perform the aforementioned method.

[0043] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0044] This invention achieves precise quantification of material flow throughout the entire lifecycle by extracting material transformation chains and process nodes. Combined with the weighted accumulation of dependency transmission paths and cross-stage indirect contributions, the extended emissions accurately reflect the cumulative environmental impact of upstream and downstream processes, significantly improving the system integrity and cross-stage attribution accuracy of the accounting results. This avoids underestimation or misalignment of emissions due to ignoring indirect paths. By extracting core impact paths through topological analysis and tracing back to source attributes, the invention accurately identifies key factors and their source links that dominate the overall carbon footprint, providing clear guidance for developing targeted emission reduction measures. The introduction of dynamic adjustment coefficients enables the accounting model to respond in real-time to changes in source attributes, enhancing the method's flexibility and adaptability. The iterative correction mechanism, combined with the redistribution of correction weights, effectively suppresses convergence fluctuations caused by initial parameter errors or path dependency biases, ensuring that corrected emissions gradually approach the true value. The final carbon footprint accounting results possess high reliability and interpretability, meeting the accuracy requirements of carbon management throughout the entire lifecycle of complex products and providing a solid data foundation for multi-dimensional source tracing analysis and emission reduction potential mining. Attached Figure Description

[0045] Figure 1 This is a flowchart illustrating the carbon footprint accounting method based on multi-dimensional analysis throughout the entire life cycle, as described in an embodiment of the present invention.

[0046] Figure 2 This is a flowchart illustrating the dynamic adjustment coefficient calculation process of the carbon footprint accounting method based on multi-dimensional analysis throughout the entire life cycle, as described in this embodiment of the invention. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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.

[0048] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0049] Figure 1This is a flowchart illustrating the carbon footprint accounting method based on multi-dimensional analysis throughout the entire life cycle, as described in an embodiment of the present invention. Figure 1 As shown, the method includes:

[0050] Acquire material flow data of the target object during its complete life cycle, establish a material conversion chain based on the material flow data and extract process nodes, and match and query the process nodes with a preset emission factor library to obtain preliminary emission amounts;

[0051] Based on the dependencies between process nodes, the initial emissions are traced transitively to obtain the dependency transmission path. Cross-stage indirect contribution paths in the dependency transmission path are identified and path weights are calculated. The initial emissions are weighted and accumulated based on the path weights to obtain the extended emissions.

[0052] A multidimensional emission feature matrix is ​​generated by associating and labeling the extended emissions with the attributes of the process nodes. Key factors are identified and a factor correlation map is constructed based on the multidimensional emission feature matrix. Topological analysis is performed on the factor correlation map to extract the core influence path. The source node of the material conversion chain is traced back along the core influence path to obtain the source attribute set. The dynamic adjustment coefficient is calculated based on the response relationship between the source attribute set and the extended emissions.

[0053] The extended emissions and the dynamic adjustment coefficient are iteratively corrected to obtain the corrected emissions. The corrected weights are calculated based on the core impact path and the factor correlation map, and the corrected emissions are redistributed to obtain the carbon footprint accounting result and output.

[0054] In one alternative implementation,

[0055] Acquire material flow data of the target object throughout its complete life cycle, establish a material transformation chain based on the material flow data and extract process nodes, and match and query the process nodes with a preset emission factor database to obtain preliminary emission amounts, including:

[0056] The material input and output records of the target object throughout its entire life cycle are collected and the material flow data is extracted. A directed graph is constructed on the material flow data according to the material flow direction. The circular flow path in the directed graph is identified and the loop is broken to obtain a directed acyclic graph. The predecessor and successor relationship is established based on the directed acyclic graph to obtain the material conversion chain. The material conversion chain is topologically sorted and the node attributes after sorting are extracted to obtain the process node set.

[0057] The node feature combination associated with each process node is extracted from the set of process nodes. The node feature combination is semantically embedded and mapped to obtain a feature vector. The feature vector is then compared with the factor feature vector in the preset emission factor library to obtain a similarity distribution. Based on the similarity distribution, a set of candidate factors with similarity exceeding a preset similarity threshold is selected. The emission factors in the candidate factor set are then subjected to secondary verification according to applicable boundary conditions to obtain the optimal emission factor. The optimal emission factor is then jointly calculated with the quantity of substances associated with the process node and the conversion efficiency coefficient to obtain the node emission amount. The node emission amounts are then accumulated along the topological order of the substance conversion chain to obtain the preliminary emission amount.

[0058] The process involves collecting material input and output records for the target object throughout its entire lifecycle, covering the complete stages from raw material acquisition, production and processing, transportation and distribution, use and maintenance, to waste disposal. The sources of these records include various types of structured and semi-structured data, such as production ledgers, energy consumption statistics reports, transportation documents, and waste disposal records. After cleaning and integrating these records, the inflow and outflow volumes, material type identifiers, and corresponding timestamps for each material in each stage are extracted to form a material flow dataset covering the entire lifecycle. This dataset uses material flow events as the basic unit, and each record includes fields such as material identifier, flow direction, quantity, and the lifecycle stage it belongs to, providing structured input for subsequent directed graph construction.

[0059] Based on material flow data, directed edge modeling is performed on each flow event according to the material flow direction. A directed graph is constructed with the starting process or stage of the material flow as the source node and the ending process or stage as the target node. In actual industrial scenarios, material flow often involves recycling and reuse, leading to loop structures in the directed graph. The existence of loop flow paths can invalidate subsequent topological sorting, therefore, loop breaking is necessary. When identifying loop flow paths, a depth-first search strategy is used to traverse the directed graph, recording back edges and marking node pairs forming loops. The specific loop breaking method is as follows: for each identified back edge, the corresponding circulating material flow quantity is extracted separately and retained as an additional attribute on the node, then the back edge is deleted, transforming the directed graph into a directed acyclic graph. The circulating material quantity information retained during the loop breaking process will participate in the joint calculation as a correction term in subsequent node emission calculations, ensuring that the contribution of circulating flow to carbon emissions is not overlooked.

[0060] After obtaining the directed acyclic graph (DAG), predecessor-successor relationships are established based on the edge relationships between nodes. Specifically, for each node, the sets of all its direct predecessor nodes and direct successor nodes are recorded, forming a material conversion chain. This chain, with its predecessor-successor relationships as the core data structure, fully describes the entire process path of a substance from its initial input through a series of conversion processes, ultimately transforming it into a product or waste. The material conversion chain is topologically sorted, and the Kahn algorithm is used to progressively peel away nodes based on their in-degree, numbering them according to topological order. After topological sorting, the attribute information of each node is extracted, including its life cycle stage, substance type identifier, process type description, substance quantity, and conversion efficiency coefficient, forming a process node set. Each process node in this set carries complete attribute information, providing a sufficient semantic foundation for subsequent emission factor matching.

[0061] When extracting the node feature combination associated with each process node from the process node set, the node feature combination is composed of multi-dimensional attribute fields such as material type identifier, process type description, life cycle stage label, and geographical region label. Semantic embedding mapping is performed on the node feature combination, transforming the text-based feature combination into a fixed-dimensional feature vector through a pre-trained semantic encoding model, ensuring that semantically similar node features have a small distance in the vector space. Each emission factor record in the pre-set emission factor library undergoes the same semantic embedding process, pre-generating corresponding factor feature vectors and establishing an index.

[0062] The cosine similarity is calculated between the feature vector of a process node and the feature vectors of all factors in the emission factor database to obtain the similarity distribution corresponding to that process node. The cosine similarity is calculated as follows: Let the feature vector of the process node be... The eigenvector of a certain emission factor is The cosine similarity between the two is... for: ,in, Represents the dot product of two vectors. and Let represent the Euclidean norms of the two vectors, respectively. Based on the similarity distribution, similarity is filtered. Exceeding the preset similarity threshold The emission factors are used to form a candidate factor set. A preset similarity threshold is set. The value is configured based on the coverage density of the emission factor library and the characteristic granularity of the process nodes, typically ranging from 0.75 to 0.95, in order to achieve a balance between recall and precision.

[0063] The emission factors in the candidate factor set undergo secondary validation based on applicable boundary conditions. These boundary conditions include constraint fields such as the applicable geographical scope, applicable time interval, applicable process type, and applicable substance purity interval. During secondary validation, the geographical region label of the process node, the time range corresponding to its life cycle stage, process type, and substance purity are matched against the boundary condition fields of each candidate factor, eliminating those that do not meet the boundary conditions. If multiple candidate factors still meet the boundary conditions after secondary validation, similarity is selected. The highest value is selected as the optimal emission factor. If no candidate factor passes the second verification, the process reverts to the second-best candidate in the similarity distribution and re-executes the boundary condition verification until the optimal emission factor is determined.

[0064] After determining the optimal emission factor, the quantity of substances associated with it and the conversion efficiency coefficient are jointly calculated to obtain the node emissions. Let the quantity of substances associated with a certain process node be... The optimal emission factor value is The conversion efficiency coefficient is Then the node emissions of that node for: Among them, the conversion efficiency coefficient This reflects the proportion of the input material that is converted into the target product in this process, and the value range is [value range missing]. The mass of recycled material retained during the loop-breaking treatment stage is equivalent to the amount of additional material. Similarly, the optimal emission factor and conversion efficiency coefficient are jointly calculated and then superimposed onto the node emission of the corresponding node to ensure that the contribution of the circulating flow is fully accounted for.

[0065] Following the topological order of the material transformation chain, the node emissions of each process node are sequentially analyzed. By summing up, a preliminary emission figure can be obtained. The accumulation process strictly follows the topological order to ensure that emissions from upstream nodes are fully included in the statistics before downstream nodes complete their calculations, avoiding duplicate calculations or omissions due to incorrect order. Preliminary emissions. It covers the direct emission contributions of all process nodes throughout the target object's entire life cycle, providing foundational data for subsequent dependency propagation path analysis and extended emission calculations. After completing the initial emission calculations, all node emissions and their corresponding optimal emission factors, substance quantities, conversion efficiency coefficients, and other intermediate calculation results are retained to support the construction of subsequent multidimensional emission feature matrices and the tracing analysis of core impact paths.

[0066] In one alternative implementation,

[0067] Based on the dependencies between process nodes, the initial emissions are traced transitively to obtain dependency propagation paths. Cross-stage indirect contribution paths within these paths are identified and their weights are calculated. The initial emissions are then weighted and accumulated based on these weights to obtain extended emissions, including:

[0068] The predecessor and successor connections between the process nodes are extracted from the material conversion chain. A dependency graph is constructed based on the predecessor and successor connections, and the stage identifier corresponding to each process node is marked. The dependency graph is traversed in reverse depth-first order. The complete path from each process node back to all predecessor nodes is recorded, and the corresponding material flow and energy conversion loss are extracted to determine the dependency transmission path set. For each path in the dependency transmission path set, the stage identifier of the adjacent nodes is detected along the path direction. When a change in the stage identifier is detected, the current node pair is marked as a cross-stage connection point. The path containing the cross-stage connection point is extracted to obtain the cross-stage indirect contribution path.

[0069] The loss ratio of mass flow at each node in the cross-stage indirect contribution path is calculated and the cumulative mass transfer efficiency is solved. The number of cross-stage connection points traversed by the cross-stage indirect contribution path is counted and the path weight is calculated in combination with the cumulative mass transfer efficiency. The initial emission amount corresponding to each process node on the cross-stage indirect contribution path is correlated with the path weight to obtain the path indirect emission contribution amount. The path indirect emission contribution amounts of all cross-stage indirect contribution paths are summarized and combined with the initial emission amount to obtain the extended emission amount.

[0070] When extracting predecessor-successor connections between process nodes in a material transformation chain, it is necessary to determine the directed connections between nodes based on the input-output relationships recorded in the material flow data. Specifically, if the output material of one process node directly serves as the input material of another process node, a directed edge is established between them, indicating that the predecessor node has a direct dependency on the successor node. A dependency graph is constructed based on all directed edges. This graph uses process nodes as vertices and the direction of material flow as the edge direction, and each vertex is appended with a stage identifier of its life cycle stage, such as raw material acquisition stage, manufacturing stage, usage stage, and disposal stage. The introduction of stage identifiers is the core basis for subsequent identification of cross-stage connection points, ensuring that the boundaries between different life cycle stages can be distinguished in the topology.

[0071] When performing a reverse depth-first traversal on the dependency graph, starting from each process node, the process backtracks layer by layer along the reverse direction of the edges to all predecessor nodes until it reaches the source node without predecessors. During the traversal, the sequence of nodes traversed by each backtracking path is fully recorded, and the mass flow values ​​between nodes and the losses generated during energy conversion are extracted synchronously on each path. Mass flow reflects the actual amount of matter transferred between adjacent nodes, while energy conversion loss characterizes the attenuation effect of the conversion process within a node on the effective transfer of matter. All backtracking paths and their corresponding mass flow and energy conversion loss data are summarized to form a dependency transfer path set, which covers all node connection paths with dependencies throughout the complete lifecycle of the target object.

[0072] For each path in the dependency transfer path set, the stage identifiers carried by adjacent node pairs are sequentially detected along the path direction. When the stage identifiers of a pair of adjacent nodes are inconsistent, it indicates that the material flow crosses the life cycle stage boundary at this point, and the node pair is marked as a cross-stage connection point. All paths in the dependency transfer path set containing at least one cross-stage connection point are extracted to form the cross-stage indirect contribution path set. These paths reflect the indirect emission transfer relationships between different life cycle stages and are an indispensable component in extended emission calculations. Traditional direct emission accounting methods often overlook such cross-stage contributions, leading to underestimation of the calculated results.

[0073] The calculation of path weights involves two related steps. For a cross-stage indirect contribution path, let the path pass through... 1 node, adjacent nodes With nodes The ratio of material flow loss between them is denoted as The mass transfer efficiency of this path is... Defined as the product of the effective transmission ratios of all adjacent nodes on the path, expressed as: ,in The range of values ​​is Reflecting the first The degree of material loss during the transmission process. The closer the value is to 1, the less loss occurs during material transport along the path, and the stronger the actual contribution of the path. Based on this, the total number of inter-stage connection points traversed by the path is counted. Combined with the efficiency of material accumulation and transfer Calculate path weights : The physical meaning of this formula is that the effective contribution of a path increases with the improvement of mass transfer efficiency, while decreasing with the increase of the number of cross-stage connection points. This is because the uncertainty and attenuation effect of indirect contribution accumulate with each stage boundary. Therefore, the number of cross-stage connection points is used as the denominator to normalize and correct the path weight, so that paths with a large number of cross-stages will not excessively amplify their indirect emission contribution.

[0074] After obtaining the path weights, the preliminary emissions corresponding to each process node on the cross-stage indirect contribution path are correlated and calculated. Let the first process node on a certain cross-stage indirect contribution path be... The initial emissions of each node are The path weight is The indirect emission contribution of this node under the current path is... for: The path indirect emission contribution of this cross-stage indirect contribution path is obtained by summing the indirect emission contributions of all nodes along the path. : When a process node appears on multiple cross-stage indirect contribution paths, the indirect emission contributions of that node in all paths must be calculated separately and then summed to avoid double counting or omissions. Different paths correspond to different material accumulation and transfer efficiencies and the number of cross-stage connection points. Therefore, the path weights of the same node under different paths are different, and its indirect emission contributions are also different. This reflects the differentiated impact of multi-path indirect transfer on carbon emission accounting.

[0075] The total indirect emissions from all cross-stage indirect contribution pathways are summed to obtain the global total indirect emissions from all cross-stage indirect contribution pathways. : ,in For the index of cross-stage indirect contribution paths, iterate through all identified cross-stage indirect contribution paths. Compared with initial emissions By merging, we obtain the expanded emissions. : Expanding emissions This approach incorporates additional emissions from all indirect contribution pathways across different stages, building upon the initial emissions figures. This allows for a more comprehensive reflection of the emission transfer effects across all stages of the target product's complete lifecycle. This method is particularly important for products with complex supply chain structures or multi-stage material recycling, as it effectively identifies and quantifies hidden carbon emission sources that are difficult to detect from a single-stage perspective. This provides a more accurate data foundation for the subsequent construction of a multi-dimensional emission characteristic matrix and the calculation of dynamic adjustment coefficients.

[0076] In one alternative implementation,

[0077] The process involves associating and labeling the extended emissions with the attributes of the process nodes to generate a multidimensional emission feature matrix. Based on this multidimensional emission feature matrix, key factors are identified and a factor correlation map is constructed, including:

[0078] Extract the material type and process parameters corresponding to the process node as node attributes, map the node attributes to the corresponding extended emissions and construct an attribute emission association table, expand each node in the attribute emission association table according to the dimension of the node attribute and generate a node feature vector, arrange the node feature vectors into row vectors according to the topological order of the material conversion chain and combine them into a matrix to obtain a multidimensional emission feature matrix.

[0079] Singular value decomposition is performed on the multidimensional emission feature matrix to obtain singular values ​​in the direction of the principal components in the feature space. Based on the distribution of the singular values, principal components whose contribution exceeds a preset contribution threshold are identified and their corresponding feature vector components are extracted. Loading coefficients of each dimension in the feature vector components are calculated, and a set of key factors is determined based on the loading coefficients. Attribute columns corresponding to the set of key factors are extracted from the multidimensional emission feature matrix, and Pearson correlation coefficients between different attribute columns are calculated. Edge weights between key factors are constructed based on the Pearson correlation coefficients. The set of key factors is used as nodes, and connection relationships are established by combining the edge weights to obtain a factor correlation graph.

[0080] The material types and process parameters corresponding to each process node are extracted as node attributes to construct a multidimensional emission feature matrix. Material type attributes include information such as the types of raw materials processed by the node, the categories of intermediate products, and the form of waste. Process parameter attributes cover quantifiable process control variables such as reaction temperature, pressure, energy intensity, and equipment efficiency. For each process node, these two types of attributes are combined to form a complete node attribute description, which is then compared with the corresponding extended emission levels. A one-to-one mapping is performed to establish an attribute emission association table. This association table uses nodes as row indexes and each attribute dimension and its corresponding emission amount as columns, ensuring that the emission characteristics of each node can be traced and compared through the attribute dimensions.

[0081] For each node in the attribute emission association table, the node is expanded according to its attribute dimensions. The substance type is encoded as a discrete category feature, process parameters are retained as continuous numerical features, and extended emissions are included as an additional dimension, thus generating a node feature vector for each node. Let the dimension of the node feature vector be... That is, the attributes and emission information of each node together constitute a length of The vector representation of the emission features is obtained by arranging the feature vectors of all nodes into row vectors according to the topological order of the material transformation chain, and then combining them into a matrix. ,in The number of rows equals the total number of process nodes, and the number of columns equals the feature dimension. The topological order ensures the consistency of material flow logic between matrix rows, enabling subsequent decomposition operations to capture emission evolution patterns along the life cycle direction.

[0082] Multidimensional emission characteristic matrix Perform singular value decomposition to decompose it into left singular matrices. Diagonal singular value matrix And right singular matrix Three parts, namely .in, diagonal elements These are the singular values ​​corresponding to the directions of each principal component. For matrix The rank of the singular values. The magnitude of the singular values ​​reflects the proportion of variance explained by the corresponding principal component direction in the feature space; the larger the singular value, the more significant the contribution of that direction to the overall emission characteristic variation. Based on the distribution of singular values, the contribution of each principal component is calculated. , defined as the first The ratio of the square of one singular value to the sum of the squares of all singular values, i.e. Let the preset contribution threshold be... Identify and satisfy The principal components are identified, and the corresponding right singular matrices are extracted. The feature vector components in the image are denoted as the set of extracted feature vector components. The number of columns corresponds to the number of selected principal components, and the number of rows corresponds to the feature dimensions. .

[0083] Calculate eigenvector components The loading coefficients for each dimension reflect the projection intensity of each attribute dimension onto the selected principal component direction. For each row (i.e., each attribute dimension), calculate the sum of squared loadings on all selected principal components to obtain the comprehensive loading coefficient for that dimension. ,in For indexing attribute dimensions, The comprehensive load factor Sort in descending order and select those with load factors exceeding the set load threshold. The dimensions were determined as the set of key factors. Key factor set Each element in the matrix corresponds to a multidimensional emission feature matrix. One of the attribute columns represents the type of material or process parameter dimension that has significant explanatory power for carbon emission characteristics. Through the loading coefficient screening mechanism, redundant attributes that contribute little to emission characteristics can be effectively filtered out, focusing on the core factors that truly drive emission differences.

[0084] From the multidimensional emission characteristic matrix Extracting key factor sets The corresponding attribute columns constitute the key factor submatrix. .right Pearson correlation coefficients are calculated between different attribute columns to quantify the degree of linear correlation between any two key factors. Let two key factors be... and The corresponding attribute column vectors are respectively and Pearson correlation coefficient Defined as the product of the covariance of two column vectors divided by their respective standard deviations, i.e. ,in and They are respectively and The mean, For node row indexes. The range of values ​​is The larger the absolute value, the stronger the linear correlation between the two key factors; a positive value indicates a positive correlation, and a negative value indicates a negative correlation. This is based on the Pearson correlation coefficient. Construct edge weights between key factors, and As a factor With factors The weight value of the edge between them, when Below the set minimum correlation threshold In this case, it is assumed that there is no effective correlation between two key factors, and no edge connection is established to avoid introducing noisy edges into the graph structure.

[0085] Set of key factors Each key factor in the graph is used as a node, and the connection relationships between nodes are established using the calculated edge weights, thus obtaining the factor association graph. Atlas The set of nodes corresponds to the set of key factors. The edge set consists of factor pairs that satisfy the relevant threshold condition, with each edge carrying the absolute value of the corresponding Pearson correlation coefficient as its weight. Factor correlation graph. The graph structure visually presents the interrelationships between key factors, providing a structured analytical foundation for subsequent topological analysis and extraction of core influence paths. Through the combined application of singular value decomposition and Pearson correlation coefficient, dimensionality reduction and structured representation from a high-dimensional emission feature matrix to a low-dimensional key factor map are achieved, effectively supporting the in-depth exploration of multi-dimensional factor interactions in carbon footprint accounting.

[0086] Figure 2 This is a flowchart illustrating the dynamic adjustment coefficient calculation process of the carbon footprint accounting method based on multi-dimensional analysis throughout the entire life cycle, as described in this embodiment of the invention.

[0087] In one alternative implementation,

[0088] Topological analysis is performed on the factor correlation map to extract the core influence path. The core influence path is then traced back to the source node of the material transformation chain to obtain the source attribute set. Based on the response relationship between the source attribute set and the extended emissions, a dynamic adjustment coefficient is calculated, including:

[0089] The edge weights in the factor association graph are normalized and a weighted adjacency matrix is ​​constructed. The betweenness centrality of each node is calculated based on the weighted adjacency matrix. Nodes whose betweenness centrality exceeds a preset centrality threshold are identified as hub nodes. Random walk sampling is performed on the weighted adjacency matrix starting from the hub nodes and the access frequency is recorded. Power iteration is performed based on the access frequency until convergence is obtained to obtain a steady-state distribution vector. The set of core influence paths is determined based on the steady-state distribution vector.

[0090] The path-end nodes are extracted from the core impact path set and mapped to the material conversion chain. The preceding connections of the material conversion chain are traced back to the source nodes, and the corresponding material types and process parameters are extracted as source attributes to construct a source attribute set. For the source nodes, the loss ratio of material flow in each conversion stage is extracted along the forward path, and the cumulative attenuation coefficient is calculated. Based on the cumulative attenuation coefficient and the extended emissions of each process node on the corresponding path, the transmission contribution is obtained. A linear fit is performed on the source attributes and the transmission contribution to obtain regression coefficients. A perturbation is applied to the regression coefficients, and the rate of change of the transmission contribution is calculated to obtain the marginal response intensity. Based on the marginal response intensity and the transmission contribution, the dynamic adjustment coefficient is obtained.

[0091] In obtaining factor correlation maps Next, the weights of each edge in the graph are normalized, and the original correlation coefficients are mapped to... This eliminates the interference of dimensional differences on subsequent topological analysis by defining intervals. Let the total number of nodes in the graph be... Construct a weighted adjacency matrix from the normalized edge weights. , where matrix elements Represents a node With nodes The normalized association strength between nodes is calculated, with the corresponding element having a value of zero if no edge exists between the two nodes. The construction of the weighted adjacency matrix allows for the quantitative expression of the graph's structural information, providing a numerical basis for subsequent calculations of betweenness centrality.

[0092] Based on the weighted adjacency matrix Calculate the betweenness centrality of each node. The centrality of the middle Defined as the path between all node pairs in the graph. The proportion of the number of shortest paths to the total number of shortest paths is calculated using a weighted shortest path algorithm, with the reciprocal of the edge weight as the path distance metric, ensuring that edges with higher association strength have higher priority in path search. Let node pairs... The total number of shortest paths between them is , which passes through nodes The quantity is ,but Identify cases where the centrality of betweenness exceeds a preset centrality threshold. The nodes are defined as the hub node set. Hub nodes play a core transit role in information transmission within the factor correlation graph, and their corresponding factors have a crucial bridging role in the carbon emission impact chain.

[0093] From the hub node set Starting with the weighted adjacency matrix Perform a random walk sampling process. In each step of the walk, use the normalized weights of the edges adjacent to the current node as the transition probability, randomly select the next hop node, and record the cumulative number of times each node is visited. Divide the number of visits by the total number of steps to obtain the visit frequency vector for each node. Nodes with high access frequency indicate that they are frequently traversed during random walks, reflecting their influence on the overall graph structure. This is represented by the access frequency vector. As the initial distribution, the normalized transition probability matrix is ​​subjected to exponential iteration. Let the... The distribution vector after the nth iteration is The iterative update rule is as follows: ,in For the weighted adjacency matrix The random transition matrix is ​​obtained by row normalization. The L2 norm of the difference between two consecutive iterations is less than the convergence threshold. The iteration terminates at this point, at which point the steady-state distribution vector is obtained. Nodes with larger component values ​​in the steady-state distribution vector are the core nodes in the graph. Connection paths are extracted along these core nodes, and paths whose cumulative weight exceeds a preset path strength threshold are selected. The path forms a core set of influential paths. .

[0094] From the core impact path set Extract the terminal nodes of each path and map them to the corresponding process nodes in the material conversion chain. Following the predecessor connections of the material conversion chain, perform reverse backtracking from each terminal node, tracing upstream nodes level by level until reaching the source node in the material conversion chain where no predecessor node exists. For each source node, extract its corresponding material type label and process parameter information. The material type label describes the type of raw material processed by the node, and the process parameters cover operating conditions affecting material conversion efficiency, such as temperature, pressure, and reaction time. Summarize the material types and process parameters of all source nodes to construct a source attribute set. .

[0095] The loss ratio of material flow in each conversion stage is extracted step by step along the forward path from the source node, assuming the process from the source node to a certain downstream process node takes... Level conversion, the first The loss ratio of the stage conversion is Then the cumulative attenuation coefficient along this path Defined as This represents the proportion of matter that actually reaches the target node after undergoing multiple stages of transformation from the source. It is based on the cumulative attenuation coefficient. And the extended emissions of each process node along the corresponding path, to calculate the transmission contribution. Let the path be the first... The extended emissions of the level node are The contribution amount is then expressed as This quantitative result reflects the actual transmission intensity of source attributes to downstream emissions, while taking into account the loss effect of substances in multi-stage transformation.

[0096] For the source attribute set The attribute variables and their corresponding transitive contributions Perform linear fitting and construct a regression model ,in For the source attribute vector, Let be the vector of regression coefficients to be estimated. This represents the intercept term. The regression coefficient vector is solved using the least squares method. This yields the linear influence weights of each source attribute on the transmission contribution. Regression coefficients. The components of the coefficient reflect the direction and intensity of the contribution of different source attributes to carbon emission transmission. Attribute dimensions with larger absolute values ​​of coefficients indicate that their impact on the transmission contribution is more significant.

[0097] Obtaining the regression coefficient vector Then, apply a magnitude of to each of its components. For small perturbations, calculate the rate of change of the transmitted contribution before and after the perturbation, and thus obtain the marginal response intensity vector. Let the first... Each regression coefficient component Apply perturbation Afterwards, the change in the amount of contribution transferred is Then the first Marginal response strength of each attribute dimension This value measures the local sensitivity of the source attribute to the transmission contribution. The greater the marginal response strength, the more significant the fluctuation in the transmission contribution will be caused by a small change in the corresponding attribute, and the higher the priority of this attribute in carbon emission regulation should be.

[0098] Comprehensive marginal response strength vector With contribution amount Calculate the dynamic adjustment coefficient Specifically, the transmission contribution is normalized and corrected using the weighted average of the marginal response intensities, calculated as follows: ,in To prevent extremely small positive numbers with a denominator of zero, a dynamic adjustment coefficient is used. This coefficient comprehensively reflects the sensitivity of source attributes to emission transmission and their actual contribution. Subsequent calculations will iteratively correct the extended emissions based on this coefficient, enabling the carbon footprint accounting results to dynamically respond to changes in source attributes and improve the accuracy and adaptability of the calculations.

[0099] In one alternative implementation,

[0100] The loss ratio of mass flow in each stage of conversion is extracted from the source node along the forward path, and the cumulative attenuation coefficient is calculated. Based on the cumulative attenuation coefficient and the extended emissions of each process node on the corresponding path, the transmission contribution is obtained, including:

[0101] Starting from the source node, extract the forward material flow path along the subsequent connection relationship of the material conversion chain, calculate the node entropy increase loss corresponding to each process node in the forward material flow path, calculate the effective energy loss ratio of material flow in single-stage conversion based on the material input-output relationship between adjacent nodes and the entropy increase loss to obtain the single-stage loss coefficient, and perform geometric mean aggregation on the single-stage loss coefficient and combine it with a preset path length correction factor to obtain the cumulative attenuation coefficient.

[0102] The cumulative attenuation coefficient is quantified using the Markov chain Monte Carlo sampling method to obtain the attenuation coefficient distribution range. The expected value of the attenuation coefficient distribution range is correlated with the extended emissions element by element to obtain the attenuation contribution component. The attenuation contribution component is accumulated and aggregated along the positive material flow path and solved by combining a preset time discount factor to obtain the path transmission contribution. All positive material flow paths originating from the source node are counted and the corresponding path transmission contributions are extracted to construct a contribution set. The path transmission contributions in the contribution set are nonlinearly mapped and aggregated according to the path topology complexity and the number of material cycles to obtain the transmission contribution of the source node.

[0103] Starting from the source node, when extracting the forward mass flow path along the successive connections of the mass conversion chain, it is necessary to traverse all reachable paths in the mass conversion chain that extend downstream from the source node along directed edges. For each process node on the forward mass flow path, the node entropy increase loss is calculated, specifically the reduction in effective energy at that node due to thermodynamic irreversibility, mass diffusion, or energy dissipation during the mass conversion process. Let the... The material input at the first-level process node is The amount of material output is The entropy increase loss coefficient of this node It can be represented by the ratio of the difference between the input and output quantities to the input quantity, i.e. Based on this, and considering the material input-output relationship between adjacent nodes, the entropy increase loss coefficient is... Corrected to single-stage loss coefficient This coefficient comprehensively considers the effective energy loss during the material transfer process between adjacent nodes, reflecting the degree of attenuation of the actual available material flow in a single-stage conversion.

[0104] Single-stage loss coefficients for all stages on all forward paths Geometric mean aggregation is performed to obtain the average single-stage loss coefficient of the path. The calculation formula is: ,in This represents the number of transformation stages from the source node to the target node. The geometric mean aggregation method avoids the oversensitivity of the arithmetic mean to extreme values, more accurately reflecting the cumulative effect of losses in multi-stage transformations. Based on this, a preset path length correction factor is introduced. This correction factor is used to compensate for the bias of path length in the cumulative attenuation estimate, and typically varies with the path order. The cumulative attenuation coefficient is appropriately increased to suppress excessive attenuation. The average single-stage loss coefficient, determined by geometric averaging, is jointly calculated using the path length correction factor, and is expressed as follows: ,in It reflects The overall degree of decrease in mass flow rate after continuous conversion. The results are then subjected to length-dependent adaptive corrections.

[0105] For the cumulative attenuation coefficient Uncertainty quantification is performed by constructing the posterior distribution of the attenuation coefficient using a Markov chain Monte Carlo sampling method. The loss coefficients of each stage are then... Treating the sample as a random variable, its prior distribution is set based on historical data or expert prior knowledge. The Metropolis-Hastings sampling algorithm is used to iteratively sample in the parameter space. Each sample accepts or rejects a new state based on the transition probability, and finally, after sufficient iterations, a convergent posterior sample set is obtained. Statistical analysis is performed on the cumulative decay coefficient samples in the posterior sample set to extract the mean and confidence interval of the samples, thus constructing the decay coefficient distribution interval. And the expected value of this distribution interval is denoted as ,Right now During MCMC sampling, a sufficiently long burn-in period needs to be set to eliminate the influence of the initial state, and the convergence of the chain should be monitored using the Gelman-Rubin statistic to ensure the reliability of the sampling results.

[0106] The expected value of the attenuation coefficient distribution range Extended emissions at each process node along the path Perform element-wise association operations to obtain the attenuation contribution components of each node. The calculation method is as follows This element-wise operation directly applies the quantified attenuation effect to the extended emissions at each node, ensuring that the calculation of the contribution components fully reflects the actual impact of mass flow attenuation on emission transmission. The attenuation contribution components for each node are as follows: Accumulation and aggregation are performed along the forward material flow path to obtain the preliminary cumulative contribution at the path level, expressed as follows: Based on this, a preset time discount factor is introduced. This factor is used to reflect the weighting of emission contributions at different life cycle stages over time, and is typically discounted according to the distance of the emission occurrence time relative to a reference time. Path transfer contribution. It is given by the product of the cumulative aggregation result and the time discount factor, i.e. This allows the calculation of the contribution to transmission to simultaneously encompass both the material decay effect and the time value.

[0107] Statistically analyze all positive material flow paths originating from the source node and extract the path transfer contribution for each path. Construct a set of contributions ,in This represents the total number of forward paths. The contribution of each path in the contribution set is transferred, and aggregated non-linearly based on the path topological complexity and the number of material cycles. Path topological complexity. The complexity of the path structure is reflected by comprehensively considering the number of branches, the number of converging nodes, and the path length; the number of material cycles. This indicates the number of times material recirculation or recycling occurs in the path; a higher number of recycling cycles indicates a stronger retention effect of the material within the system. The nonlinear mapping aggregation function uses a weighted exponential form, with the first... The propagation contribution of each path is weighted by a non-linear weight determined by both topological complexity and the number of iterations. The calculation method is as follows This form degenerates into a weight determined solely by topological complexity when the number of iterations is zero, but it increases exponentially with the number of iterations, reasonably reflecting the amplification effect of material cycling on the contribution.

[0108] The contribution of the source node The sum is obtained by normalizing and aggregating the contributions from each path in the contribution set through nonlinear mapping, and the expression is: This normalization process ensures that the propagation contributions between source nodes with different numbers of paths are comparable, avoiding systematic bias in the final result caused by differences in the number of paths. In actual calculations, when the number of paths... For larger datasets, a path pruning strategy can be employed, prioritizing paths with higher topological complexity and material cycle counts for aggregation. This reduces computational complexity while ensuring complete coverage of core contributing paths. Through this complete process, an accurate solution for the contribution of material decay, uncertainty quantification, time discounting, and path topology transmission is achieved, starting from the source node and comprehensively considering these factors. This provides a reliable numerical foundation for subsequent calculations of dynamically adjusted coefficients.

[0109] In one alternative implementation,

[0110] The extended emissions and the dynamic adjustment coefficient are iteratively corrected to obtain the corrected emissions. Based on the core impact path and the factor correlation map, the correction weights are calculated, and the corrected emissions are redistributed to obtain the carbon footprint accounting results, which are then output as follows:

[0111] The extended emissions are arranged into emission vectors according to the process nodes. The dynamic adjustment coefficients are propagated along the core influence path to obtain the node adjustment coefficients. The initial corrected emission vector is obtained by combining the emission vectors. The deviation between the initial corrected emission vector and the emission vector is calculated and iteratively corrected based on a preset convergence threshold to obtain the corrected emissions.

[0112] Edge weights are extracted from the factor association graph and a weighted adjacency matrix is ​​constructed. The weighted adjacency matrix is ​​normalized based on the access frequency of nodes in the core influence path to obtain a transition probability matrix. The steady-state eigenvector of the transition probability matrix is ​​calculated and nonlinearly mapped by combining it with the betweenness centrality of nodes in the core influence path to obtain a corrected weight vector.

[0113] Based on the corrected emissions and the corresponding elements in the corrected weight vector, the allocation benchmark is calculated. The cumulative distribution of the allocation benchmark in each stage of the material conversion chain is calculated and emission concentration areas are identified. The corrected emissions in the emission concentration areas are proportionally redistributed according to the proportion of material flow to obtain a centralized allocation result. Non-concentrated areas are directly allocated according to the corrected weight vector to obtain a discrete allocation result. The discrete allocation result and the centralized allocation result are merged to obtain the carbon footprint accounting result and output.

[0114] Will expand emissions An emission vector is constructed by arranging the nodes of each process in the material transformation chain in the topological order. , of which The component corresponds to the first Extended emissions at each process node. Dynamic adjustment coefficient. Set along the core impact path To propagate along the path, specifically, for each node on the core impact path, [the following will be done]. Multiplying the node's relative position weight in the path yields the node adjustment coefficient for each node. For nodes not on any core impact path, their node adjustment factor is set to a unit value of 1 to maintain constant emissions. The node adjustment factor vector is then used. With emission vector Element-wise multiplication yields the initial revised emission vector. ,Right now ,in This indicates element-wise multiplication.

[0115] Calculate the initial revised emission vector Compared with the original emission vector Deviation between , is defined as the L2 norm of the difference between two vectors, i.e. .Will With preset convergence threshold If a comparison is made, If the current corrected emission vector is used as the new input, the path propagation and element-wise multiplication operations of the node adjustment coefficients are re-executed, and the next iteration begins; if The iteration terminates when the current corrected emission vector is denoted as the corrected emission vector. The iterative process can effectively eliminate the systematic error introduced by the initial estimation deviation of the dynamic adjustment coefficient, so that the corrected emissions can reach a steady state under the constraints of mass conservation and factor response.

[0116] From factor correlation map Extract the edge weights between each key factor node and construct a weighted adjacency matrix. , where matrix elements Key factor nodes With nodes The original correlation strength between them, this value is derived from the Pearson correlation coefficient. The absolute value is directly assigned. Based on the core influence path set. Access frequency vector of each node For the weighted adjacency matrix Frequency-weighted normalization is performed: for each row of the matrix, the access frequency of the corresponding node in that row is calculated. Multiply by each element in the row, then sum all elements in the row and take the reciprocal as the normalization factor to obtain the transition probability matrix. ,make sure The sum of each row is 1, satisfying the properties of a random matrix.

[0117] For the transition probability matrix Perform exponential iterations, using a uniformly distributed vector as the initial distribution, and repeat the process. The iteration continues until the L2 norm of the difference between two adjacent iterations is less than the convergence threshold. Finally, the steady-state distribution vector is obtained. Steady-state distribution vector The Each component This reflects the nodes during the random walk of the factor association graph. The long-term probability of being visited, i.e., the structural importance of the node in the overall carbon emission impact transmission network. The steady-state distribution vector... The centrality of each node in the core influence path Perform a nonlinear mapping for each node. Calculate its corrected weight components ,in and The preset nonlinear mapping exponent controls the relative contribution ratios of steady-state probability and betweenness centrality in the correction weights. To prevent extremely small positive numbers with a denominator of zero, for all nodes... Normalization is performed to obtain the corrected weight vector. The sum of its components is 1.

[0118] Based on the modified emission vector With the modified weight vector Multiply the corresponding elements one by one to calculate the allocation base amount for each process node. ,in For the first Corrected emissions at each process node These are the corresponding corrected weight components. The allocation baseline for each node is... By accumulating and summing the values ​​according to the stages of the material transformation chain, the cumulative distribution baseline value distribution curves for each stage are obtained. The cumulative distribution baseline value of each stage is then calculated as a percentage of the total global distribution baseline value. The proportion is used to identify cases where the cumulative proportion exceeds a preset concentration threshold. The process nodes within the defined phase intervals are then designated as emission concentration areas. The remaining node sets are marked as non-centralized regions. .

[0119] For concentrated emission areas The corrected emissions within the region are proportionally redistributed according to the proportion of mass flow at each node to the total mass flow in that region. Let the first node be... The mass flow rate of each node is Then the proportion of material flow of this node in the concentrated area is In the centralized allocation results, the first The carbon footprint allocation obtained by each node is This approach ensures that the carbon footprint allocation within concentrated emission areas aligns with actual material flow intensity, avoiding distortion in the allocation of high-emission areas due to local fluctuations in the correction weights. For non-concentrated areas... The nodes within are directly modified using the weight vector. The corresponding component is the weight, representing the total corrected emissions within that region. The weights are distributed to each node to obtain the discrete allocation result. This approach fully utilizes the topological information of the factor correlation map, giving the carbon footprint allocation of low-emission, dispersed regions a reasonable structural basis.

[0120] Centralized distribution results With discrete allocation results Merging by node affiliation yields the complete carbon footprint accounting result vector. Each component corresponds to the final carbon footprint calculation value of a single process node in the material transformation chain. The system summarizes and statistically analyzes data from multiple dimensions, including stage, substance type, and emission source, to generate and output a structured carbon footprint accounting report. It supports the extraction of analysis results such as the carbon footprint contribution ratio of each stage, the ranking of key emission nodes, and the distribution of indirect emissions across stages, providing refined data support for carbon emission reduction decisions.

[0121] A second aspect of this invention provides a carbon footprint accounting system based on multi-dimensional analysis throughout the entire life cycle, comprising:

[0122] The preliminary calculation unit is used to acquire material flow data of the target object during its complete life cycle, establish a material conversion chain based on the material flow data and extract process nodes, and match and query the process nodes with a preset emission factor library to obtain the preliminary emission amount.

[0123] The path weighting unit is used to trace the initial emissions based on the dependencies between the process nodes to obtain the dependency transmission path, identify the cross-stage indirect contribution paths in the dependency transmission path and calculate the path weights, and perform weighted summation of the initial emissions based on the path weights to obtain the extended emissions.

[0124] The graph source tracing unit is used to associate and label the extended emissions with the attributes of the process nodes to generate a multidimensional emission feature matrix, identify key factors based on the multidimensional emission feature matrix and construct a factor association graph, perform topological analysis on the factor association graph to extract the core influence path, trace back along the core influence path to the source node of the material conversion chain to obtain the source attribute set, and calculate the dynamic adjustment coefficient based on the response relationship between the source attribute set and the extended emissions.

[0125] The iterative correction unit is used to iteratively correct the extended emissions and the dynamic adjustment coefficient to obtain the corrected emissions, calculate the correction weight based on the core influence path and the factor correlation map, and redistribute the corrected emissions to obtain the carbon footprint accounting result and output it.

[0126] A third aspect of the present invention provides an electronic device, comprising:

[0127] A processor and a memory for storing processor-executable instructions, wherein the processor is configured to invoke instructions stored in the memory to perform the aforementioned method.

[0128] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0129] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0130] 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 them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A carbon footprint accounting method based on multi-dimensional analysis of the entire life cycle, characterized in that, include: Acquire material flow data of the target object during its complete life cycle, establish a material conversion chain based on the material flow data and extract process nodes, and match and query the process nodes with a preset emission factor library to obtain preliminary emission amounts; Based on the dependencies between process nodes, the initial emissions are traced transitively to obtain the dependency transmission path. Cross-stage indirect contribution paths in the dependency transmission path are identified and path weights are calculated. The initial emissions are weighted and accumulated based on the path weights to obtain the extended emissions. A multidimensional emission feature matrix is ​​generated by associating and labeling the extended emissions with the attributes of the process nodes. Key factors are identified and a factor correlation map is constructed based on the multidimensional emission feature matrix. Topological analysis is performed on the factor correlation map to extract the core influence path. The source node of the material conversion chain is traced back along the core influence path to obtain the source attribute set. The dynamic adjustment coefficient is calculated based on the response relationship between the source attribute set and the extended emissions. The extended emissions and the dynamic adjustment coefficient are iteratively corrected to obtain the corrected emissions. The corrected weights are calculated based on the core impact path and the factor correlation map, and the corrected emissions are redistributed to obtain the carbon footprint accounting result and output.

2. The method according to claim 1, characterized in that, Acquire material flow data of the target object throughout its complete life cycle, establish a material transformation chain based on the material flow data and extract process nodes, and match and query the process nodes with a preset emission factor database to obtain preliminary emission amounts, including: The material input and output records of the target object throughout its entire life cycle are collected and the material flow data is extracted. A directed graph is constructed on the material flow data according to the material flow direction. The circular flow path in the directed graph is identified and the loop is broken to obtain a directed acyclic graph. The predecessor and successor relationship is established based on the directed acyclic graph to obtain the material conversion chain. The material conversion chain is topologically sorted and the node attributes after sorting are extracted to obtain the process node set. The node feature combination associated with each process node is extracted from the set of process nodes. The node feature combination is semantically embedded and mapped to obtain a feature vector. The feature vector is then compared with the factor feature vector in the preset emission factor library to obtain a similarity distribution. Based on the similarity distribution, a set of candidate factors with similarity exceeding a preset similarity threshold is selected. The emission factors in the candidate factor set are then subjected to secondary verification according to applicable boundary conditions to obtain the optimal emission factor. The optimal emission factor is then jointly calculated with the quantity of substances associated with the process node and the conversion efficiency coefficient to obtain the node emission amount. The node emission amounts are then accumulated along the topological order of the substance conversion chain to obtain the preliminary emission amount.

3. The method according to claim 1, characterized in that, Based on the dependencies between process nodes, the initial emissions are traced transitively to obtain dependency propagation paths. Cross-stage indirect contribution paths within these paths are identified and their weights are calculated. The initial emissions are then weighted and accumulated based on these weights to obtain extended emissions, including: The predecessor and successor connections between the process nodes are extracted from the material conversion chain. A dependency graph is constructed based on the predecessor and successor connections, and the stage identifier corresponding to each process node is marked. The dependency graph is traversed in reverse depth-first order. The complete path from each process node back to all predecessor nodes is recorded, and the corresponding material flow and energy conversion loss are extracted to determine the dependency transmission path set. For each path in the dependency transmission path set, the stage identifier of the adjacent nodes is detected along the path direction. When a change in the stage identifier is detected, the current node pair is marked as a cross-stage connection point. The path containing the cross-stage connection point is extracted to obtain the cross-stage indirect contribution path. The loss ratio of mass flow at each node in the cross-stage indirect contribution path is calculated and the cumulative mass transfer efficiency is solved. The number of cross-stage connection points traversed by the cross-stage indirect contribution path is counted and the path weight is calculated in combination with the cumulative mass transfer efficiency. The initial emission amount corresponding to each process node on the cross-stage indirect contribution path is correlated with the path weight to obtain the path indirect emission contribution amount. The path indirect emission contribution amounts of all cross-stage indirect contribution paths are summarized and combined with the initial emission amount to obtain the extended emission amount.

4. The method according to claim 1, characterized in that, The process involves associating and labeling the extended emissions with the attributes of the process nodes to generate a multidimensional emission feature matrix. Based on this multidimensional emission feature matrix, key factors are identified and a factor correlation map is constructed, including: Extract the material type and process parameters corresponding to the process node as node attributes, map the node attributes to the corresponding extended emissions and construct an attribute emission association table, expand each node in the attribute emission association table according to the dimension of the node attribute and generate a node feature vector, arrange the node feature vectors into row vectors according to the topological order of the material conversion chain and combine them into a matrix to obtain a multidimensional emission feature matrix. Singular value decomposition is performed on the multidimensional emission feature matrix to obtain singular values ​​in the direction of the principal components in the feature space. Based on the distribution of the singular values, principal components whose contribution exceeds a preset contribution threshold are identified and their corresponding feature vector components are extracted. Loading coefficients of each dimension in the feature vector components are calculated, and a set of key factors is determined based on the loading coefficients. Attribute columns corresponding to the set of key factors are extracted from the multidimensional emission feature matrix, and Pearson correlation coefficients between different attribute columns are calculated. Edge weights between key factors are constructed based on the Pearson correlation coefficients. The set of key factors is used as nodes, and connection relationships are established by combining the edge weights to obtain a factor correlation graph.

5. The method according to claim 1, characterized in that, Topological analysis is performed on the factor correlation map to extract the core influence path. The core influence path is then traced back to the source node of the material transformation chain to obtain the source attribute set. Based on the response relationship between the source attribute set and the extended emissions, a dynamic adjustment coefficient is calculated, including: The edge weights in the factor association graph are normalized and a weighted adjacency matrix is ​​constructed. The betweenness centrality of each node is calculated based on the weighted adjacency matrix. Nodes whose betweenness centrality exceeds a preset centrality threshold are identified as hub nodes. Random walk sampling is performed on the weighted adjacency matrix starting from the hub nodes and the access frequency is recorded. Power iteration is performed based on the access frequency until convergence is obtained to obtain a steady-state distribution vector. The set of core influence paths is determined based on the steady-state distribution vector. The path-end nodes are extracted from the core impact path set and mapped to the material conversion chain. The preceding connections of the material conversion chain are traced back to the source nodes, and the corresponding material types and process parameters are extracted as source attributes to construct a source attribute set. For the source nodes, the loss ratio of material flow in each conversion stage is extracted along the forward path, and the cumulative attenuation coefficient is calculated. Based on the cumulative attenuation coefficient and the extended emissions of each process node on the corresponding path, the transmission contribution is obtained. A linear fit is performed on the source attributes and the transmission contribution to obtain regression coefficients. A perturbation is applied to the regression coefficients, and the rate of change of the transmission contribution is calculated to obtain the marginal response intensity. Based on the marginal response intensity and the transmission contribution, the dynamic adjustment coefficient is obtained.

6. The method according to claim 5, characterized in that, The loss ratio of mass flow in each stage of conversion is extracted from the source node along the forward path, and the cumulative attenuation coefficient is calculated. Based on the cumulative attenuation coefficient and the extended emissions of each process node on the corresponding path, the transmission contribution is obtained, including: Starting from the source node, extract the forward material flow path along the subsequent connection relationship of the material conversion chain, calculate the node entropy increase loss corresponding to each process node in the forward material flow path, calculate the effective energy loss ratio of material flow in single-stage conversion based on the material input-output relationship between adjacent nodes and the entropy increase loss to obtain the single-stage loss coefficient, and perform geometric mean aggregation on the single-stage loss coefficient and combine it with a preset path length correction factor to obtain the cumulative attenuation coefficient. The cumulative attenuation coefficient is quantified using the Markov chain Monte Carlo sampling method to obtain the attenuation coefficient distribution range. The expected value of the attenuation coefficient distribution range is correlated with the extended emissions element by element to obtain the attenuation contribution component. The attenuation contribution component is accumulated and aggregated along the positive material flow path and solved by combining a preset time discount factor to obtain the path transmission contribution. All positive material flow paths originating from the source node are counted and the corresponding path transmission contributions are extracted to construct a contribution set. The path transmission contributions in the contribution set are nonlinearly mapped and aggregated according to the path topology complexity and the number of material cycles to obtain the transmission contribution of the source node.

7. The method according to claim 1, characterized in that, The extended emissions and the dynamic adjustment coefficient are iteratively corrected to obtain the corrected emissions. Based on the core impact path and the factor correlation map, the correction weights are calculated, and the corrected emissions are redistributed to obtain the carbon footprint accounting results, which are then output as follows: The extended emissions are arranged into emission vectors according to the process nodes. The dynamic adjustment coefficients are propagated along the core influence path to obtain the node adjustment coefficients. The initial corrected emission vector is obtained by combining the emission vectors. The deviation between the initial corrected emission vector and the emission vector is calculated and iteratively corrected based on a preset convergence threshold to obtain the corrected emissions. Edge weights are extracted from the factor association graph and a weighted adjacency matrix is ​​constructed. The weighted adjacency matrix is ​​normalized based on the access frequency of nodes in the core influence path to obtain a transition probability matrix. The steady-state eigenvector of the transition probability matrix is ​​calculated and nonlinearly mapped by combining it with the betweenness centrality of nodes in the core influence path to obtain a corrected weight vector. Based on the corrected emissions and the corresponding elements in the corrected weight vector, the allocation benchmark is calculated. The cumulative distribution of the allocation benchmark in each stage of the material conversion chain is calculated and emission concentration areas are identified. The corrected emissions in the emission concentration areas are proportionally redistributed according to the proportion of material flow to obtain a centralized allocation result. Non-concentrated areas are directly allocated according to the corrected weight vector to obtain a discrete allocation result. The discrete allocation result and the centralized allocation result are merged to obtain the carbon footprint accounting result and output.

8. A carbon footprint accounting system based on multi-dimensional analysis of the entire life cycle, used to implement the method of any one of claims 1-7, characterized in that, include: The preliminary calculation unit is used to acquire material flow data of the target object during its complete life cycle, establish a material conversion chain based on the material flow data and extract process nodes, and match and query the process nodes with a preset emission factor library to obtain the preliminary emission amount. The path weighting unit is used to trace the initial emissions based on the dependencies between the process nodes to obtain the dependency transmission path, identify the cross-stage indirect contribution paths in the dependency transmission path and calculate the path weights, and perform weighted summation of the initial emissions based on the path weights to obtain the extended emissions. The graph source tracing unit is used to associate and label the extended emissions with the attributes of the process nodes to generate a multidimensional emission feature matrix, identify key factors based on the multidimensional emission feature matrix and construct a factor association graph, perform topological analysis on the factor association graph to extract the core influence path, trace back along the core influence path to the source node of the material conversion chain to obtain the source attribute set, and calculate the dynamic adjustment coefficient based on the response relationship between the source attribute set and the extended emissions. The iterative correction unit is used to iteratively correct the extended emissions and the dynamic adjustment coefficient to obtain the corrected emissions, calculate the correction weight based on the core influence path and the factor correlation map, and redistribute the corrected emissions to obtain the carbon footprint accounting result and output it.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images