A product carbon footprint information management method and system based on knowledge graph verification

Abstract

The application discloses a product carbon footprint information management method and system based on knowledge graph verification, comprising the following steps: firstly, collecting and analyzing carbon footprint data of each stage of the whole life cycle of a product, and constructing a dynamic carbon flow topology graph; then, initializing a knowledge graph entity set carrying candidate management parameters based on the topology graph; iteratively optimizing the entity set by using a verification algorithm, calculating a system carbon footprint index in each iteration, updating individual and global optimal solutions, and driving entity evolution by using a plurality of dynamically adjusted learning factors; when a termination condition is met, extracting a global optimal management parameter vector, generating a real-time optimization control instruction according to the global optimal management parameter vector, and delivering the real-time optimization control instruction to an execution system, so that closed-loop management from intelligent analysis to optimization execution of carbon footprint is realized.

Description

Technical Field

[0001] This invention relates to the field of carbon footprint management, and more specifically, to a method and system for managing product carbon footprint information based on knowledge graph verification. Background Technology

[0002] With increasing global focus on climate change, precise management and continuous optimization of product carbon footprints have become crucial for enterprises to achieve green and low-carbon development. Current technologies for managing carbon emissions throughout a product's lifecycle largely rely on phased static accounting and human experience-based decision-making, lacking the ability to realistically and systematically depict and intelligently optimize the dynamic processes of carbon flow. Traditional methods struggle to effectively integrate the discrete and massive amounts of carbon data from each stage—raw material acquisition, manufacturing, transportation and distribution, use, and waste recycling—and cannot construct network models that reflect the true flow and transformation of carbon in space and time. Furthermore, when seeking carbon reduction optimization solutions, existing methods are often limited to adjusting local parameters, lacking an intelligent decision-making mechanism capable of simulating the coupled effects of multiple parameters, performing global optimization based on swarm intelligence, and ultimately forming executable closed-loop control commands. This results in low efficiency in carbon footprint management, limited optimization effects, and an inability to meet increasingly stringent carbon emission compliance requirements and dynamically changing production and operating environments. Summary of the Invention

[0003] The purpose of this invention is to provide a product carbon footprint information management method and system based on knowledge graph verification.

[0004] In a first aspect, embodiments of the present invention provide a product carbon footprint information management method based on knowledge graph verification, comprising:

[0005] The carbon footprint data stream generated by monitoring nodes in the entire product lifecycle chain within a continuous time period is collected. The carbon footprint data stream is parsed and processed to obtain the instantaneous carbon footprint parameter set of the monitoring node. The instantaneous carbon footprint parameter set includes carbon emission data in the raw material acquisition stage, carbon emission data in the production and manufacturing stage, carbon emission data in the transportation and distribution stage, carbon emission data in the usage stage, and carbon emission data in the waste recycling stage.

[0006] Based on the carbon emission data of the raw material acquisition stage, the carbon emission data of the production and manufacturing stage, the carbon emission data of the transportation and distribution stage, and the carbon emission data of the usage stage, a carbon flow topology map of the monitoring node is constructed in the continuous time period. The carbon flow topology map consists of carbon flow nodes and carbon flow transmission paths connecting the carbon flow nodes. The carbon flow nodes correspond to each stage of the entire life cycle of the monitoring node.

[0007] The knowledge graph initialization module is invoked to generate an initial entity relationship set corresponding to the monitoring node based on the carbon flow topology graph. Each carbon footprint entity in the initial entity relationship set carries a candidate management parameter vector, which includes energy structure adjustment parameters for the production and manufacturing stage, optimization intensity parameters for the process flow, and route scheduling timing adjustment parameters for the transportation and distribution stage.

[0008] The initial entity relationship set is iteratively verified using a knowledge graph verification algorithm. In each iteration, the system carbon footprint index corresponding to the carbon footprint entity is calculated based on the current attribute vector of the carbon footprint entity. The individual historical best attribute of the carbon footprint entity and the global best attribute of the initial entity relationship set are updated based on the system carbon footprint index. The state vector and attribute vector of the carbon footprint entity are updated based on dynamically adjusted semantic weight factor, entity learning factor and relationship learning factor.

[0009] When the knowledge graph verification algorithm reaches the preset iteration termination condition, the target management parameter vector corresponding to the global optimal attribute of the initial entity relationship set is extracted, and a real-time carbon footprint optimization management instruction is generated based on the target management parameter vector and sent to the central processing unit of the product carbon footprint management system.

[0010] In a second aspect, embodiments of the present invention provide a server system, including a server, the server being used to execute the method described in the first aspect.

[0011] Compared to existing technologies, the beneficial effects provided by this invention include: This invention discloses a product carbon footprint information management method and system based on knowledge graph verification, comprising: firstly, collecting and parsing carbon footprint data at each stage of the product's entire life cycle to construct a dynamic carbon flow topology graph. Subsequently, generating a knowledge graph entity set carrying candidate management parameters based on the topology graph. Iteratively optimizing the entity set using a verification algorithm, calculating the system's carbon footprint index and updating individual and global optimal solutions in each iteration, while simultaneously driving entity evolution using dynamically adjusted learning factors. When the termination condition is met, extracting the global optimal management parameter vector and generating real-time optimization control instructions based on it, which are then sent to the execution system, achieving closed-loop management from intelligent carbon footprint analysis to optimized execution. Attached Figure Description

[0012] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as limiting the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 A flowchart illustrating the steps of a product carbon footprint information management method based on knowledge graph verification provided in an embodiment of the present invention;

[0014] Figure 2 A schematic block diagram of the structure of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0015] 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, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0016] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0017] In order to solve the technical problems mentioned in the background art Figure 1 This is a flowchart illustrating the product carbon footprint information management method based on knowledge graph verification provided in this embodiment. The following is a detailed description of the product carbon footprint information management method based on knowledge graph verification.

[0018] Step S201: Collect carbon footprint data streams generated by monitoring nodes in the entire product lifecycle chain within a continuous time period, and parse and process the carbon footprint data streams to obtain the instantaneous carbon footprint parameter set of the monitoring nodes. The instantaneous carbon footprint parameter set includes carbon emission data in the raw material acquisition stage, carbon emission data in the production and manufacturing stage, carbon emission data in the transportation and distribution stage, carbon emission data in the usage stage, and carbon emission data in the waste recycling stage.

[0019] Step S202: Construct a carbon flow topology map of the monitoring node in the continuous time period based on the carbon emission data of the raw material acquisition stage, the carbon emission data of the production and manufacturing stage, the carbon emission data of the transportation and distribution stage, and the carbon emission data of the usage stage. The carbon flow topology map consists of carbon flow nodes and carbon flow transmission paths connecting the carbon flow nodes. The carbon flow nodes correspond to each stage of the entire life cycle of the monitoring node.

[0020] Step S203: Call the knowledge graph initialization module to generate an initial entity relationship set corresponding to the monitoring node based on the carbon flow topology graph. Each carbon footprint entity in the initial entity relationship set carries a candidate management parameter vector. The candidate management parameter vector includes energy structure adjustment parameters for the production and manufacturing stage, optimization intensity parameters for the process flow, and route scheduling timing adjustment parameters for the transportation and distribution stage.

[0021] Step S204: The initial entity relationship set is iteratively verified using a knowledge graph verification algorithm. In each iteration, the system carbon footprint index corresponding to the carbon footprint entity is calculated based on the current attribute vector of the carbon footprint entity. The individual historical best attribute of the carbon footprint entity and the global best attribute of the initial entity relationship set are updated based on the system carbon footprint index. The state vector and attribute vector of the carbon footprint entity are updated based on the dynamically adjusted semantic weight factor, entity learning factor and relationship learning factor.

[0022] Step S205: When the knowledge graph verification algorithm reaches the preset iteration termination condition, extract the target management parameter vector corresponding to the global optimal attribute of the initial entity relationship set, generate a real-time carbon footprint optimization management instruction based on the target management parameter vector, and send the real-time carbon footprint optimization management instruction to the central processing unit of the product carbon footprint management system.

[0023] In this embodiment of the invention, the product carbon footprint information management method based on knowledge graph verification provided by the present invention is executed by a server system deployed on a cloud platform or enterprise data center. The server system is configured with a high-performance processor, large-capacity memory, high-speed network interface, and dedicated data acquisition and processing module. The server collects carbon footprint data streams generated by each monitoring node in the entire product lifecycle chain in real time through industrial IoT gateways, enterprise resource planning system interfaces, supply chain management platforms, and product sensor networks. For example, for the entire lifecycle management of an industrial motor, the server continuously receives carbon emission-related data packets reported within a continuous period (such as the past 24 hours) from monitoring nodes of iron ore mining enterprises (raw material acquisition stage), motor component casting and assembly plants (production and manufacturing stage), logistics company transportation fleets (transportation and distribution stage), motor end-user factories (use stage), and waste motor recycling and processing centers (waste recycling stage), forming the original carbon footprint data stream.

[0024] Upon receiving the aforementioned carbon footprint data stream, the server immediately invokes its internal data parsing engine for processing. The parsing engine first decodes, cleans, and formats the data stream, identifying the lifecycle stage label to which the data packets belong. Subsequently, the engine extracts key parameters from the pre-processed data. For the raw material acquisition stage, it parses the instantaneous numerical sequences of direct carbon emissions (such as fuel consumption) and indirect carbon emissions (such as electricity consumption) generated during the mining and preliminary processing of major raw materials such as iron ore, copper, and aluminum. For the manufacturing stage, it parses the instantaneous carbon emission data corresponding to energy consumption (electricity, natural gas) in each process, including casting, machining, heat treatment, assembly, and testing. For the transportation and distribution stage, it parses the carbon emission data corresponding to fuel consumption of different transportation vehicles (trucks, ships) on different road sections, and associates it with timestamps and geographical location information. For the usage stage, it parses the carbon emission data of electricity consumption converted from the load power and operating time of the motor during operation at the user's factory. For the waste recycling stage, it parses the energy consumption carbon emission data of processing processes such as dismantling, crushing, sorting, and smelting, as well as the carbon emission data offset by recycled materials. Finally, the server categorizes these parsed data by stage and integrates them to form a set of instantaneous carbon footprint parameters for the monitoring node in the current continuous period.

[0025] Next, the server initiates the carbon flow topology map construction module. This module takes the instantaneous carbon footprint parameter set as input and first performs deep feature extraction on carbon emission data from the raw material acquisition and manufacturing stages. For example, through time-frequency analysis, it separates the raw material carbon emission intensity sequence, representing a stable production baseline, and the raw material carbon emission phase sequence, representing intermittent operations (such as blasting and loading / unloading), from the energy consumption fluctuation data of the raw material acquisition stage. Similarly, it extracts the production carbon emission intensity sequence and production carbon emission phase sequence from the manufacturing data. Likewise, it processes the data from the transportation and distribution and usage stages to obtain the transportation carbon emission intensity sequence, transportation carbon emission phase sequence, usage carbon emission intensity sequence, and usage carbon emission phase sequence. Based on these sequences, the server calculates the instantaneous direct carbon emission value, reflecting the intensity of direct emissions from the carbon source; the instantaneous indirect carbon emission value, reflecting indirect emissions from the supply chain; the instantaneous process carbon emission value, reflecting the dynamic emission characteristics of the logistics and usage processes; and the cumulative emissions up to the current moment, i.e., the instantaneous cumulative carbon emission value.

[0026] Simultaneously, the server specifically analyzes data from the waste recycling phase, identifying the set of processing time intervals during which the recycling center actually performs dismantling, smelting, and other operations within a continuous period, as well as the set of idle time intervals for equipment standby and maintenance. Based on these intervals, the server logically infers the state transition moments when the node acts as a "carbon sink" to absorb carbon (e.g., material recycling replacing virgin materials), and the state transition moments when carbon offsetting (e.g., carbon capture) may occur. Based on these transition moments, the server intelligently divides the entire continuous period into multiple continuous carbon flow analysis sub-intervals, ensuring that the carbon source and carbon sink states of the system remain relatively stable within each sub-interval.

[0027] Within each predefined carbon flow analysis sub-interval, the server uses previously calculated instantaneous carbon emission values ​​to establish a carbon flow balance equation for that sub-interval. This equation describes the inflow, outflow, accumulation, and offsetting relationships of carbon between various nodes (such as mining sites, factories, warehouses, users, and recycling plants) from raw material input to final product (and waste) output. The server extracts key carbon flow direction parameters (indicating from which node carbon flows to which node) and carbon flow amplitude parameters (indicating the amount of carbon emissions flowing) from these balance equations. The direction parameters are used to determine the start and end points of the carbon flow transmission path, while the amplitude parameters quantify the carbon flow intensity value of that path. Finally, the server correlates and combines all the identified carbon flow nodes (assigned unique identifiers), carbon flow transmission paths (assigned unique identifiers), and the start and end points and intensity information of these paths in chronological order to construct a dynamic carbon flow topology map that reflects the overall spatiotemporal flow of carbon. In this map, nodes represent carbon sources (such as energy-intensive processes) or carbon sinks (such as the recycling stage), and edges represent carbon flow.

[0028] Subsequently, the server invokes its built-in knowledge graph initialization module. This module first parses the aforementioned carbon flow topology graph, statistically analyzing the number of various carbon flow nodes (e.g., 3 raw material acquisition nodes, 5 production and manufacturing nodes, etc.) and their type distribution. Combining the path carbon flow intensity values ​​(i.e., carbon emission flow rates) of each carbon flow transmission path in the graph, the module analyzes and determines which stages contribute the most to carbon emissions or exhibit the most significant fluctuations under the current management status, thereby identifying the types of key carbon footprint parameters that need optimization. In this embodiment, the server identifies the energy structure (ratio of fossil fuels to renewable energy) in the production and manufacturing stage, the optimization intensity of the process flow (e.g., the degree of introduction of energy-saving technologies), and the route scheduling sequence (departure time, route selection) in the transportation and distribution stage as three key adjustable parameters.

[0029] Next, the initialization module obtains the preset value ranges for these key parameters. For example, the renewable energy proportion range for the energy structure adjustment parameter is set to [20%, 80%]; the optimization intensity parameter (representing the investment in technological transformation or the magnitude of process adjustment) ranges to [0, 1]; and the route scheduling timing adjustment parameter is specified as the boundary values ​​of the start and stop time windows of the transportation tasks (e.g., allowing adjustment within the range of [08:00, 22:00]). Then, the module uses its internal uniform random number generator to randomly generate initial values ​​for each "carbon footprint entity" to be created (each entity represents a complete set of management parameter schemes): within the value range of each parameter, an initial energy structure value, an initial optimization intensity value, and an initial route scheduling timing parameter set (containing multiple start and end time suggestions for transportation tasks) are randomly generated. Combining these randomly generated values ​​in a preset order constitutes the initial attribute vector of the carbon footprint entity. At the same time, the server sets this initial attribute vector as the initial individual historical best attribute of the entity. In addition, the module will randomly generate an initial state vector for each entity within another preset state value range (for velocity or momentum calculation in subsequent iterations). Finally, the server aggregates all generated carbon footprint entities (including their initial attribute vectors and initial state vectors) to form an initial entity relationship set, completing the initialization of the knowledge graph.

[0030] Once preparation is complete, the server initiates its core knowledge graph verification algorithm to iteratively verify the initial entity relationship set. In each iteration, the server first traverses each carbon footprint entity in the set. For the currently processing carbon footprint entity, the server reads the specific values ​​contained in its current attribute vector: current energy structure value (e.g., renewable energy share 45%), current optimization intensity value (e.g., 0.6), and current route scheduling time series parameter set (a set of specific transportation plan time points). The server inputs these specific parameter values ​​into a pre-built "product carbon footprint system simulation model" stored in the server's memory. This simulation model contains a full lifecycle topology corresponding to the actual monitoring nodes, capable of simulating the dynamic behavior of carbon flow under given management parameters.

[0031] The simulation model runs for a preset duration (e.g., simulating operations over a future quarter) and outputs detailed simulation results, including carbon emission data for each simulation phase (e.g., emissions during production and transportation) and total simulated carbon emission data. The server performs post-processing on this simulation data: by clustering the phased carbon emission data, it identifies major emission sources (e.g., a specific smelting furnace) and several secondary emission sources, calculates their emission amplitudes, and then summarizes them to obtain the total simulated carbon emission value for the given scenario. Intensity component analysis is performed on the total carbon emission data to calculate carbon emissions per unit of output and per unit of product, and the latter is divided by the former to obtain the simulated carbon emission intensity value. Furthermore, the server calculates the actual carbon emission integral value and the compliance carbon emission threshold integral value set according to regulations or targets within the simulation period, and the ratio of the latter to the former is used as the simulated carbon footprint compliance rate. The total simulated carbon emission value, the simulated carbon emission intensity value, and the simulated carbon footprint compliance rate together constitute the system carbon footprint index corresponding to the carbon footprint entity in the current iteration.

[0032] After obtaining the system carbon footprint index, the server begins updating the optimal attribute. For the current carbon footprint entity, the server compares its newly calculated system carbon footprint index with the historical system carbon footprint index corresponding to its stored historical optimal attribute. The comparison follows a priority rule: first, the simulated carbon footprint compliance rate is compared; if the new value is higher, the entity's individual historical optimal attribute is immediately updated using the current attribute vector. If the compliance rates are equal, the simulated total carbon emissions are compared; if the new total is smaller, it is updated. If the totals are also equal, the simulated carbon emission intensity is compared; if the new intensity is smaller, it is updated. After updating the individual historical optimal attributes of all entities, the server traverses the entire initial entity relation set, finds the individual historical optimal attribute with the highest simulated carbon footprint compliance rate, and sets it as the globally optimal attribute of the entire set after this iteration.

[0033] While updating the optimal attributes, the server also updates the state vector and attribute vector, which is the core of the algorithm's search and optimization. The server obtains the historical state vector of the current carbon footprint entity in the previous iteration. Then, it calculates two difference vectors: the first is the first difference vector between the current attribute vector and its own historical best attribute, and the second is the second difference vector between the current attribute vector and the globally best attribute. To dynamically adjust the update process, the server calculates three key factors in real time: semantic weight factor, entity learning factor, and relation learning factor.

[0034] When calculating the semantic weight factor, the server first calculates the average center of the current attribute vectors of all carbon footprint entities in the current set. Then, it calculates the semantic distance between each entity attribute and this center, and calculates the standard deviation of these distances to obtain the semantic distribution dispersion parameter. Simultaneously, the server calculates the proportion of the current iteration to the total number of iterations as the iteration progress ratio parameter, and calculates the sum of the semantic distances between the current entity attribute and its individual historical best and global best as the individual global deviation parameter. These parameters are input into a preset dynamic adjustment function to correct the initial baseline value of the semantic weight, obtaining the semantic weight factor value for the current iteration. When calculating the entity learning factor and relation learning factor, the server calculates the relative difference (first difference and second difference) between the current entity's individual historical best attribute and global best attribute on the simulated values ​​of total carbon emissions and carbon emission intensity, respectively. Then, these differences are weighted and combined with preset corresponding learning factor baseline values ​​to generate the current iteration calculated values ​​of the entity learning factor and relation learning factor, respectively.

[0035] With these dynamic factors, the server performs vector operations: multiplying the historical state vector by the semantic weight factor yields the first state update component vector. Multiplying the first difference vector by the entity learning factor, and then by a random number (generated by the server's built-in first random number generator), yields the second state update component vector. Multiplying the second difference vector by the relation learning factor, and then by another random number (generated by the second random number generator), yields the third state update component vector. Adding these three component vectors gives the updated state vector for the carbon footprint entity. Finally, adding the current attribute vector to this updated state vector generates the new attribute vector for the entity in the next iteration. This process simulates the complex learning and movement of an entity (solution) in the knowledge graph space, attracted by its own historical experience (individual optimality) and guided by the best role models in the group (global optimality), while maintaining a certain degree of random exploration and inertia (historical state).

[0036] The server continues the above iterative process. After each iteration, the server checks whether a preset iteration termination condition has been met. The termination condition typically includes one of three conditions: First, the number of iterations completed has reached the preset maximum number of iterations (e.g., 1000). Second, the simulated carbon footprint compliance rate corresponding to the current globally optimal attribute has reached or exceeded a preset target threshold (e.g., 95%). Third, in multiple consecutive iterations, the change in the globally optimal attribute has been less than a preset convergence threshold, indicating that the optimization has stabilized. Once any one of these conditions is met, the server determines that the algorithm has converged and stops iterating.

[0037] After the iteration terminates, the server extracts the final globally optimal attribute from the initial entity relationship set. This attribute is essentially an optimal management parameter vector that has been verified and optimized through multiple rounds of testing. The server parses this target management parameter vector to obtain specific target energy structure values ​​(e.g., the proportion of renewable energy needs to be increased to 65%), target optimization intensity values ​​(e.g., the process optimization intensity needs to be set to 0.85), and target route scheduling timing parameter set (a transportation task schedule accurate to the minute).

[0038] Finally, the server needs to convert these optimization results into executable instructions. It converts the target energy structure values ​​into specific energy ratio settings that the central processing unit's energy efficiency management strategy can recognize, and encapsulates these into energy structure control words. It converts the target optimization intensity values ​​into optimization level settings that the process optimization module can understand, and encapsulates these into optimization intensity control words. It parses the target route scheduling timing parameter set into a series of specific instructions such as "start XX transportation task at XX time" and "stop YY task at YY time," forming a route scheduling timing control instruction sequence. Following the communication protocol format specified by the central processing unit, the server encapsulates these control words and instruction sequences into a complete real-time carbon footprint optimization management instruction data packet. Subsequently, the server sends this instruction data packet to the designated instruction register address space of the product carbon footprint management system's central processing unit through its communication interface (such as OPCUA, MQTT, or other industrial protocol interfaces). After transmission, the server triggers an interrupt or signal to notify the central processing unit to read the new instructions. After reading the instructions, the central processing unit can adjust the power supply strategy according to the energy structure control word, adjust the process parameters on the production line or start energy-saving equipment according to the optimization intensity control word, and precisely command the operation of the logistics system according to the route scheduling timing control instruction sequence, thereby realizing real-time optimization management of the carbon footprint of the entire product life cycle. The entire process is executed automatically and cyclically by the server, realizing continuous monitoring, intelligent diagnosis and dynamic optimization of the carbon footprint.

[0039] In this embodiment of the invention, the step of updating the individual historical optimal attribute of the carbon footprint entity and the global optimal attribute of the initial entity relationship set according to the system carbon footprint index can be implemented through the following example.

[0040] Obtain the historical system carbon footprint index corresponding to the individual historical best attribute of the carbon footprint entity. The historical system carbon footprint index includes the historical total carbon emission simulation value, the historical carbon emission intensity simulation value, and the historical carbon footprint compliance rate simulation value.

[0041] The simulated value of carbon footprint compliance rate in the system carbon footprint index corresponding to the carbon footprint entity is compared with the simulated value of historical carbon footprint compliance rate in the historical system carbon footprint index.

[0042] When the simulated value of the carbon footprint compliance rate is greater than the simulated value of the historical carbon footprint compliance rate, the current attribute vector of the carbon footprint entity is set to the updated individual historical best attribute.

[0043] When the simulated value of the carbon footprint compliance rate is equal to the simulated value of the historical carbon footprint compliance rate, a second comparison process is performed between the simulated value of the total carbon emissions in the system carbon footprint index corresponding to the carbon footprint entity and the simulated value of the total historical carbon emissions in the historical system carbon footprint index.

[0044] When the simulated value of the carbon footprint compliance rate is equal to the simulated value of the historical carbon footprint compliance rate and the simulated value of the total carbon emissions is less than the simulated value of the total historical carbon emissions, the current attribute vector of the carbon footprint entity is set to the updated individual historical best attribute.

[0045] When the simulated value of the carbon footprint compliance rate is equal to the simulated value of the historical carbon footprint compliance rate and the simulated value of the total carbon emissions is equal to the simulated value of the total carbon emissions, a third comparison process is performed between the simulated value of carbon emission intensity in the system carbon footprint index corresponding to the carbon footprint entity and the simulated value of historical carbon emission intensity in the historical system carbon footprint index.

[0046] When the simulated value of the carbon footprint compliance rate is equal to the simulated value of the historical carbon footprint compliance rate, the simulated value of the total carbon emissions is equal to the simulated value of the total historical carbon emissions, and the simulated value of the carbon emission intensity is less than the simulated value of the historical carbon emission intensity, the current attribute vector of the carbon footprint entity is set to the updated individual historical best attribute.

[0047] Traverse all carbon footprint entities in the initial entity relationship set, obtain the updated carbon footprint compliance rate simulation value corresponding to the updated individual historical best attribute of each carbon footprint entity, and set the individual historical best attribute corresponding to the maximum value among all the updated carbon footprint compliance rate simulation values ​​as the global best attribute of the initial entity relationship set.

[0048] In an embodiment of the present invention, for example, during a single iteration of the knowledge graph verification algorithm, the server completes the verification of a certain carbon footprint entity (e.g., numbered Entity). 042 After the system carbon footprint index of an entity (whose current attribute vector represents a specific combination of management parameters) is calculated, the update process of individual and global optimal attributes is immediately initiated.

[0049] First, the server retrieves data from the entity. 042The server retrieves the entity's "individual historical best attribute" from its persistent storage area. This attribute records the entity's optimal solution since the algorithm's iteration began. Simultaneously, the server obtains the historical system carbon footprint index corresponding to this optimal attribute. This index is a data structure containing three key values: historical total carbon emissions simulation value (e.g., 1250.6 tons of CO2e), historical carbon emissions intensity simulation value (e.g., 1.82 tons of CO2e / 10,000 yuan of output value), and historical carbon footprint compliance rate simulation value (e.g., 90.5%).

[0050] Next, the server will calculate the value corresponding to the Entity. 042 The current system carbon footprint metrics of the attribute vector (assumed to be: simulated total carbon emissions of 1180.2 tons of CO2e, simulated carbon emission intensity of 1.75 tons of CO2e per 10,000 yuan of output value, and simulated carbon footprint compliance rate of 92.8%) are compared with the aforementioned historical metrics first, i.e., the simulated carbon footprint compliance rate of the two is compared first. The server determines that the current value of 92.8% is greater than the historical value of 90.5%. According to the preset update logic, compliance improvement has the highest priority. Therefore, the server immediately updates the Entity... 042 The current attribute vector is overwritten and written to the storage location of its individual historical best attribute, completing the update. At the same time, the new system carbon footprint index (1180.2, 1.75, 92.8%) corresponding to the current attribute vector is synchronously recorded as a new historical index.

[0051] If in another iteration, the server discovers the Entity 042 If the current simulated carbon footprint compliance rate (e.g., 92.8%) is equal to its historical value (92.8%), a second comparison is triggered. The server then compares the simulated total carbon emissions of the two. Assume the current simulated total carbon emissions are 1150.0 tons of CO2e, while the historical value is 1180.2 tons of CO2e. The server determines that the current value is less than the historical value. In this case, although the compliance rate has not improved, the total carbon emissions have decreased. Therefore, the server will... 042 The current attribute vector is set to the updated individual historical best attribute, and the historical index is updated to (1150.0, [current intensity value], 92.8%).

[0052] Furthermore, consider a more special case: in subsequent iterations, Entity 042The current simulated carbon footprint compliance rate (92.8%) is equal to the historical value (92.8%), and the current simulated total carbon emissions (1150.0 tons of CO2e) is also equal to the historical value (1150.0 tons of CO2e). At this point, the server initiates the third comparison process, comparing the simulated carbon emission intensity values. Assume the current simulated carbon emission intensity is 1.70 tons of CO2e per 10,000 yuan of output value, while the historical value is 1.72 tons of CO2e per 10,000 yuan of output value. The server determines that the current value is less than the historical value. This means that, under the premise of unchanged compliance and total emissions, the carbon emission efficiency of economic output has been optimized. Therefore, the server still updates the current attribute vector to the individual's historical best attribute and records the new historical indicators (1150.0, 1.70, 92.8%).

[0053] After updating the historical best attributes of all carbon footprint entities in the set (some entities may be updated while others remain unchanged), the server begins updating the globally optimal attributes. The server iterates through the entire initial entity relationship set, reading the "updated carbon footprint compliance rate simulation value" corresponding to the updated historical best attributes of each carbon footprint entity. For example, after iteration, the following is obtained: Entity 001 At 94.2%, Entity 042 92.8%, Entity 103 The compliance rate is 95.1%, ... The server sorts and compares all the simulated compliance rate values ​​it reads, finding the maximum value. In this example, 95.1% is the maximum value, corresponding to Entity 103 The individual's historical best attributes. The server will then send the Entity... 103 The individual historical optimal attributes are set as the globally optimal attributes of the entire initial entity relationship set after this iteration, and stored in global shared memory for all entities to refer to and learn from in the next iteration. This process ensures that collective intelligence always converges towards the currently known highest compliance rate, and seeks further optimization of the total amount and intensity based on this.

[0054] In this embodiment of the invention, updating the state vector and attribute vector of the carbon footprint entity according to the dynamically adjusted semantic weight factor, entity learning factor and relation learning factor can be implemented through the following example.

[0055] Obtain the historical state vector of the carbon footprint entity in the previous iteration. The historical state vector is composed of the state components of the energy structure adjustment parameter, the optimization intensity parameter, and the route scheduling timing adjustment parameter in the previous iteration, arranged in a preset parameter order.

[0056] Obtain a first difference vector between the current attribute vector of the carbon footprint entity and the individual historical best attribute of the carbon footprint entity. The first difference vector is composed of the difference between the current value of the energy structure adjustment parameter and the individual historical best energy structure value, the difference between the current value of the optimization intensity parameter and the individual historical optimal intensity value, and the difference between the current value of the route scheduling timing adjustment parameter and the individual historical best route scheduling timing parameter group, according to a preset parameter arrangement order.

[0057] Obtain a second difference vector between the current attribute vector of the carbon footprint entity and the globally optimal attribute of the initial entity relationship set. The second difference vector is composed of the difference between the current value of the energy structure adjustment parameter and the globally optimal energy structure value, the difference between the current value of the optimization intensity parameter and the globally optimal intensity value, and the difference between the current value of the route scheduling timing adjustment parameter and the globally optimal route scheduling timing parameter set, arranged in a preset parameter order.

[0058] The current iterative calculation value of the semantic weight factor is generated based on the semantic proximity between the current attribute vector of the carbon footprint entity and the individual's historical best attribute, the semantic proximity between the current attribute vector of the carbon footprint entity and the global best attribute, and the semantic distribution dispersion of all carbon footprint entities in the initial entity relationship set.

[0059] The current iterative calculation value of the entity learning factor and the current iterative calculation value of the relation learning factor are generated based on the first difference degree between the simulated total carbon emissions corresponding to the individual historical best attribute of the carbon footprint entity and the simulated total carbon emissions corresponding to the global best attribute, and the second difference degree between the simulated carbon emission intensity corresponding to the individual historical best attribute of the carbon footprint entity and the simulated carbon emission intensity corresponding to the global best attribute.

[0060] Perform a scalar multiplication operation between the historical state vector and the current iterative calculated value of the semantic weight factor to generate the first state update component vector;

[0061] Perform a scalar multiplication operation between the first difference vector and the current iterative calculation value of the entity learning factor to generate a first intermediate vector. Perform a scalar multiplication operation between the first intermediate vector and the first random value output by the first random number generator to generate a second state update component vector.

[0062] Perform a scalar multiplication operation between the second difference vector and the current iterative calculation value of the relation learning factor to generate a second intermediate vector. Perform a scalar multiplication operation between the second intermediate vector and the second random value output by the second random number generator to generate a third state update component vector.

[0063] The first state update component vector, the second state update component vector, and the third state update component vector are added together to generate the updated state vector of the carbon footprint entity.

[0064] The updated attribute vector of the carbon footprint entity is generated by performing a vector addition operation between the current attribute vector of the carbon footprint entity and the updated state vector.

[0065] In this embodiment of the invention, for example, in a single iteration of the knowledge graph verification algorithm, after updating the optimal attributes, the server immediately calculates a new state and attribute vector for each carbon footprint entity to guide its exploration in the solution space. The following uses the carbon footprint entity as an example. 042 Taking this as an example, the execution process of the server is described in detail.

[0066] First, the server from Entity 042 The system retrieves the historical state vector saved at the end of the previous iteration (e.g., the t-th iteration) from its cache. This vector is composed of elements arranged according to a preset parameter order (e.g., [energy structure, optimization intensity, route scheduling sequence]). Assume its value is V. history =[0.12,-0.05,0.03], where 0.12 represents the "velocity" or "momentum" component of the energy structure adjustment parameter in the previous iteration, and -0.05 and 0.03 correspond to the state components of the optimization intensity parameter and the route scheduling timing adjustment parameter.

[0067] Next, the server calculates two key difference vectors. It reads the Entity... 042 The current attribute vector (the vector used when calculating the system's carbon footprint index in the (t+1)th iteration), let's assume it's X. current =[0.55,0.70,{08:30,17:00}] represents a renewable energy share of 55%, an optimization intensity of 0.70, and a set of transportation scheduling times. The server also reads the Entity. 042 Let P be the individual's historical best attribute after its own update. best =[0.60,0.75,{09:00,16:30}]. Then, the server performs vector subtraction to generate the first difference vector: D individual =X current -P best =[-0.05,-0.05,{-00:30,+00:30}]. This vector quantifies the difference between the current solution and its historical best solution in various dimensions.

[0068] Simultaneously, the server reads the globally optimal attribute of the initial entity relation set determined after this iteration from the global shared memory, let's say it's G.best =[0.65,0.80,{08:00,17:30}] (belongs to another, better entity). The server calculates the second difference vector: D global =X current -G best =[-0.10,-0.10,{+00:30,-00:30}]. This vector quantifies the difference between the current solution and the known optimal solution of the population.

[0069] Subsequently, the server enters the dynamic factor calculation phase. To calculate the current iteration value w of the semantic weight factor, the server first evaluates the Entity. 042 The current attribute vector X current respectively with P best and G best The semantic proximity is calculated (e.g., by calculating cosine similarity or the reciprocal of Euclidean distance). Next, the server retrieves the current attribute vectors of all entities in the entire entity relation set, calculates their centroids, and then calculates the standard deviation of the distances from all entity attribute vectors to these centroids, obtaining the semantic distribution dispersion. Finally, the server inputs the "individual global deviation" (a function of the two proximity values), the "semantic distribution dispersion," and the current iteration progress ratio into a preset dynamic weight adjustment function. This function might output a value such as w=0.85, indicating the weight of historical state inertia in the current iteration.

[0070] Next, the server calculates the entity learning factor c1 and the relation learning factor c2. The server queries the Entity. 042 The individual's historical best attribute P best The corresponding historical system carbon footprint indicators are used to obtain the simulated historical total carbon emissions (e.g., 1150.0 tons) and simulated historical carbon emission intensity (e.g., 1.70 tons / 10,000 yuan). Simultaneously, the globally optimal attribute G is queried. best The corresponding simulated total carbon emissions (e.g., 1100.0 tons) and simulated carbon emission intensity (e.g., 1.65 tons / 10,000 yuan) are calculated. The server calculates the first degree of difference: |1150.0-1100.0| / 1100.0≈0.0455; and the second degree of difference: |1.70-1.65| / 1.65≈0.0303. Then, the server compares these two degrees of difference with a preset benchmark coefficient (e.g., c1). base =2.0,c2 base =2.0) to perform weighted combination operation. One possible calculation is: c1=c1 base *(1-0.5*first degree of difference)=2.0*(1-0.5*0.0455)≈1.9545; c2=c2 base*(0.5 + 0.5 * second difference) = 2.0 * (0.5 + 0.5 * 0.0303) ≈ 1.0303. Thus, the values ​​of c1 and c2 are dynamically adjusted based on the difference between the individual optimum and the global optimum in terms of key performance indicators.

[0071] After factor calculation is completed, the server performs vector composition operation:

[0072] Generate the first state update component vector:

[0073] V1=w*V history =0.85*[0.12,-0.05,0.03]=[0.102,-0.0425,0.0255].

[0074] Generate the second state update component vector: The server calls the first random number generator to obtain a random number rand1 uniformly distributed in the [0,1] range, for example, 0.67. Then calculate: V2 = rand1 * c1 * D individual =0.67*1.9545*[-0.05,-0.05,{-00:30,+00:30}]≈[-0.0655,-0.0655,{-00:20,+00:20}].

[0075] Generating the third state update component vector: The server calls a separate second random number generator to obtain a random number rand2, for example, 0.45. Then it calculates:

[0076] V3=rand2*c2*D global =0.45*1.0303*[-0.10,-0.10,{+00:30,-00:30}]≈[-0.0464,-0.0464,{+00:14,-00:14}].

[0077] Finally, the server performs vector addition to generate Entities. 042 Updated state vector: V new =V1+V2+V3=[0.102-0.0655-0.0464,-0.0425-0.0655-0.0464,{0.0255-00:20+00:14,...}]≈[-0.0099,-0.1544,{-00:06,...}]. This new state vector combines inertia, the tendency to learn from its own experience, and the tendency to learn from the group's optimality, and introduces randomness.

[0078] Ultimately, the server generates Entities through vector addition. 042 The updated attribute vector for the next iteration:

[0079] Xnew =X current +V new =[0.55,0.70,{08:30,17:00}]+[-0.0099,-0.1544,{-00:06,...}]≈[0.5401,0.5456,{08:24,17:14}]. The server will use this X. new Save as an Entity 042 The "current attribute vector" in the next iteration completes a full parameter update and optimization exploration process.

[0080] In this embodiment of the invention, the step of calculating the system carbon footprint index corresponding to the carbon footprint entity based on the current attribute vector of the carbon footprint entity can be implemented through the following example.

[0081] Obtain the current energy structure value of the energy structure adjustment parameter, the current optimization intensity value of the optimization intensity parameter, and the current route scheduling timing parameter group of the route scheduling timing adjustment parameter contained in the current attribute vector of the carbon footprint entity.

[0082] The current energy structure value, the current optimization intensity value, and the current route scheduling timing parameter group are input into the pre-built product carbon footprint system simulation model. The product carbon footprint system simulation model contains the full life cycle topology structure corresponding to the monitoring node.

[0083] The carbon footprint system simulation model is used to simulate and calculate the carbon flow status of the monitoring node under the management of the current energy structure value, the current optimization intensity value, and the current route scheduling time sequence parameter group, and to generate the carbon emission data of the monitoring node in the simulation stage and the overall carbon emission data in the simulation within the preset simulation time.

[0084] Cluster analysis is performed on the carbon emission data of the simulation stage to extract the emission amplitudes of the main emission sources and the emission amplitudes of each secondary emission source in the carbon emission data of the simulation stage. The total simulated value of carbon emission corresponding to the carbon emission data of the simulation stage is calculated based on the emission amplitudes of the main emission sources and all the emission amplitudes of the secondary emission sources.

[0085] Intensity component analysis is performed on the simulated overall carbon emission data to extract the carbon emission per unit output value and carbon emission per unit product from the simulated overall carbon emission data. The simulated carbon emission intensity value corresponding to the simulated overall carbon emission data is calculated based on the ratio of the carbon emission per unit product to the carbon emission per unit output value.

[0086] Based on the carbon emission data of the simulation phase and the overall carbon emission data of the simulation, the actual carbon emission integral value and the compliant carbon emission threshold integral value of the monitoring node within the preset simulation time are calculated, and the ratio of the compliant carbon emission threshold integral value to the actual carbon emission integral value is used as the carbon footprint compliance rate simulation value.

[0087] The simulated values ​​of total carbon emissions, carbon emission intensity, and carbon footprint compliance rate are combined to generate the system carbon footprint index corresponding to the carbon footprint entity.

[0088] In this embodiment of the invention, for example, in a single iteration of the knowledge graph verification algorithm, the server needs to evaluate the merits of the management scheme represented by each carbon footprint entity. The core of this evaluation is calculating the system carbon footprint index corresponding to that entity. The following section describes how the server processes carbon footprint entities (Entities). 042 Taking this as an example, the calculation process will be described in detail.

[0089] First, the server from Entity 042 The specific adjustable parameters are parsed from the current attribute vector. It extracts the current energy structure values ​​(e.g., the proportion of renewable energy r). current =0.55), the current optimization intensity value (e.g., process optimization level o). current =0.70), and the current route scheduling timing parameter set (e.g., a set of specific transportation task schedules s). current ={Task A 08:30-12:00, Task B :13:00-17:00}).

[0090] Subsequently, the server invokes its internally pre-built product carbon footprint system simulation model. This model is a high-fidelity digital twin system that accurately maps the real-world topology of monitoring nodes involved in the entire product lifecycle, including raw material supplier A, component factory B, assembly plant C, regional distribution center D, and end customer E, as well as the material and energy flow relationships between them. The server will then analyze the three parameters (r)... current ,o current ,s current ) is used as a control variable input into the simulation model.

[0091] The simulation model starts, taking the current moment as the starting point, and simulates the operation over a pre-defined future period (e.g., the next quarter). During the simulation, the model operates based on the input r. current =0.55, dynamically adjust the power source composition of each production node (e.g., 55% from wind and solar power, 45% from the grid); based on o current=0.70, and a series of energy-saving processes are implemented proportionally (such as 70% of the production lines using new heat insulation technology); based on s current The model precisely schedules the departure, travel, and stop times of the transport fleet. Based on physical laws, operational rules, and an emissions factor library, the model dynamically calculates and tracks the flow, transformation, and accumulation of carbon at each node and edge in the topological network. After the simulation, the model outputs two sets of core result data: one is time-series data broken down by life cycle stage (raw materials, production, transportation, etc.), i.e., carbon emission data for the simulation stage; the other is the summary data of the entire system during the simulation period, i.e., the overall carbon emission data for the simulation.

[0092] After receiving the raw simulation data, the server begins extracting indicators. For the carbon emission data from the simulation phase, the server performs cluster analysis. It uses a clustering algorithm (such as K-means) to group all emission sources (potentially thousands of instantaneous emission points) by emission magnitude. After analysis, the server identifies a cluster of "major emission sources" (e.g., the baking oven in the painting workshop of assembly plant C), and its emission magnitude (i.e., the emission rate at the center of the cluster) is extracted, denoted as E. main =85kgCO2e / h. Simultaneously, several clusters of "minor emission sources" (such as transport vehicles and auxiliary equipment) were identified, and their respective average emission amplitudes were extracted, for example, E... secondary 1 = 15 kg CO2e / h, E secondary 2 = 8 kg CO2e / h. Then, based on these amplitudes and the simulation duration, the server calculates the simulated total carbon emissions: Total. Sim =(E main +ΣE secondary Simulation duration = (85+15+8)kg / h * 2160h ≈ 233.3 tons of CO2e.

[0093] Next, the server processes the overall carbon emission data from the simulation. It performs intensity component analysis, extracting two key economic and environmental indicators from the aggregated data: carbon emissions per unit of output (e.g., CI). output =0.5 tons CO2e / 10,000 yuan of output value) and carbon emissions per unit of product (e.g., CI) product =1.2 tons of CO2e / unit). Subsequently, the server calculates the simulated carbon intensity value, defined as the ratio of economic output efficiency to product carbon emissions, i.e., Intensity. Sim =CI product / CI output =1.2 / 0.5=2.4 (dimensionless ratio). The lower this value, the higher the carbon efficiency of a single product for the same output value.

[0094] Finally, the server calculates the simulated carbon footprint compliance rate. It first integrates the actual carbon emissions from all monitoring nodes over time during the simulation period to obtain the actual carbon emission integral value, assuming it to be Actual. Integral =240.0 tons of CO2e. Simultaneously, the server generates a compliance carbon emission threshold curve that changes over time based on preset regulations, industry standards, or internal targets, and integrates this threshold curve over the same period to obtain the compliance carbon emission threshold integral value, denoted as Threshold. Integral =260.0 tons of CO2e. The simulated carbon footprint compliance rate is the ratio of the two: Compliance Sim =Threshold Integral / Actual Integral =260.0 / 240.0≈1.083, usually expressed as a percentage as 108.3%, indicating that the actual emissions are below the compliance threshold, with a safety margin of 8.3%.

[0095] At this point, the server has completed the process for configuring the Entity. 042 The comprehensive simulation evaluation of the current management scheme generated three core indicators: simulated total carbon emissions (233.3 tons), simulated carbon emission intensity (2.4), and simulated carbon footprint compliance rate (108.3%). The server combines these three values ​​into a system carbon footprint indicator triple (233.3, 2.4, 108.3%) and stores it in association in the Entity. 042 The records are used for subsequent optimal attribute updates and learning factor calculations. This process is performed once for each entity in the set in each iteration and is the foundation driving the evolution of the entire optimization algorithm.

[0096] In this embodiment of the invention, the updating of the state vector and attribute vector of the carbon footprint entity based on dynamically adjusted semantic weight factor, entity learning factor and relation learning factor is further provided by the following implementation method.

[0097] Obtain the current attribute vectors of all carbon footprint entities in the initial entity relationship set, and calculate the average central attribute vector of all the current attribute vectors;

[0098] Calculate the semantic distance between the current attribute vector and the average central attribute vector for each carbon footprint entity, and perform standard deviation calculation on all semantic distances to generate the semantic distribution dispersion parameter of all carbon footprint entities in the initial entity relation set;

[0099] Obtain the current iteration number and the preset maximum iteration number, and calculate the ratio of the current iteration number to the maximum iteration number as the iteration process ratio parameter;

[0100] Calculate the first semantic distance between the current attribute vector of the carbon footprint entity and the individual's historical best attribute, calculate the second semantic distance between the current attribute vector of the carbon footprint entity and the global best attribute, sum the first semantic distance and the second semantic distance to generate the individual global deviation parameter of the carbon footprint entity;

[0101] The iteration process ratio parameter, the individual global deviation parameter, and the semantic distribution dispersion parameter are input into the semantic weight dynamic adjustment function. The semantic weight dynamic adjustment function is used to correct the preset initial benchmark value of the semantic weight to generate the current iteration calculation value of the semantic weight factor.

[0102] Obtain the first absolute difference between the simulated total carbon emissions value corresponding to the individual's historical best attribute and the simulated total carbon emissions value corresponding to the global best attribute, and divide the first absolute difference by the simulated total carbon emissions value corresponding to the global best attribute to generate a first difference degree.

[0103] Obtain the second absolute difference between the simulated carbon emission intensity value corresponding to the individual's historical best attribute and the simulated carbon emission intensity value corresponding to the global best attribute, and divide the second absolute difference by the simulated carbon emission intensity value corresponding to the global best attribute to generate a second difference degree;

[0104] The first difference degree is multiplied by a preset first learning factor benchmark value to generate a first product result. The second difference degree is multiplied by a preset second learning factor benchmark value to generate a second product result. The first product result and the second product result are summed to generate the current iterative calculation value of the entity learning factor.

[0105] The first difference is multiplied by a preset third learning factor benchmark value to generate a third product result. The second difference is multiplied by a preset fourth learning factor benchmark value to generate a fourth product result. The third product result and the fourth product result are summed to generate the current iterative calculation value of the relation learning factor.

[0106] In this embodiment of the invention, for example, during the iterative process of the knowledge graph verification algorithm, the server needs to calculate the semantic weight factor, entity learning factor, and relation learning factor in real time to dynamically adjust the update behavior of the carbon footprint entity. The following example uses the server in the 50th iteration as an example where the carbon footprint entity is... 042 Taking the calculation of these factors as an example, the process will be described in detail.

[0107] First, to calculate the semantic distribution dispersion parameter, the server retrieves the current attribute vectors of all 100 carbon footprint entities in the initial entity relation set from memory. Each vector is a three-dimensional vector (encoded values ​​for energy structure, optimization intensity, and route scheduling timing). The server calculates the arithmetic mean of these 100 vectors to obtain the average central attribute vector, denoted as C. avg =[0.58,0.72,{08:45,16:45}]. Next, the server calculates the current attribute vector for each entity and C. avg The semantic distance between them. A simplified form of Euclidean distance is used here (e.g., calculated after converting the time parameter to minutes). Assume Entity 042 The current attribute vector is X current =[0.55,0.70,{08:30,17:00}], which is related to C avg The semantic distance is calculated as d current =1.85. The server calculates the standard deviation of the 100 semantic distance values ​​d1, d2, ..., d100 for all 100 entities, and obtains the semantic distribution dispersion parameter σ, which characterizes the degree of dispersion of the group distribution. disperse =0.92.

[0108] Subsequently, the server obtains the iteration progress ratio parameter. The current iteration number t is known. now =50, preset maximum number of iterations T max =200. The server calculates the ratio iter. ratio =t now / T max =50 / 200=0.25, indicating that the optimization process has progressed to one-quarter of the way through.

[0109] Next, the server calculates the Entity. 042 The individual global deviation parameter. It reads the Entity 042 The individual's historical best attribute P best =[0.60,0.75,{09:00,16:30}] and the current globally optimal attribute G best =[0.65,0.80,{08:00,17:30}]. The server calculates X respectively. current With P best First semantic distance d individual =2.10, and X current With G best The second semantic distance d global =3.15. Summing the two yields the individual global deviation parameter D. deviation =d individual +d global=5.25, which reflects the degree to which the current solution is far from both its own experience and the group's best experience.

[0110] Then the server will iter ratio(0.25) D deviation(5.25) and σ disperse(0.92) Input a predefined semantic weight dynamic adjustment function. An example of this function is shown below:

[0111] w=w base *(1-iter ratio )*exp(-D deviation / (10*σ disperse )), where w base The initial baseline value for the semantic weight is set to 1.2. The server calculates the following using numerical values: w = 1.2 * (1 - 0.25) * exp(-5.25 / (10 * 0.92)) ≈ 1.2 * 0.75 * exp(-0.57) ≈ 0.9 * 0.566 ≈ 0.509. This is the current iteration's calculated value for the semantic weight factor, indicating that the influence of historical inertia in the current iteration is approximately 0.509.

[0112] Next, the server calculates the learning factor. It queries the Entity. 042 The individual's historical best attribute P best The corresponding system carbon footprint index was used to obtain its historical total carbon emissions simulation value (Total). P =1150.0 tons and historical carbon emission intensity simulation value Intensity P =1.70. Simultaneously, query the globally optimal attribute G. best The corresponding simulated total carbon emissions value (Total) G =1100.0 tons and carbon emission intensity simulation value Intensity G =1.65.

[0113] The server calculates the first difference:

[0114] Diff1=|Total P -Total G | / Total G =|1150.0-1100.0| / 1100.0=50.0 / 1100.0≈0.0455. Calculate the second degree of difference:

[0115] Diff2=|Intensity P -Intensity G Intensity G =|1.70-1.65| / 1.65=0.05 / 1.65≈0.0303.

[0116] Finally, the server generates learning factors based on the degree of difference. Assume the preset baseline value is: the baseline value c1 for the first learning factor. base =1.8, second learning factor baseline value c2 base =0.5, the baseline value of the third learning factor c3 base =0.4, the baseline value of the fourth learning factor c4 base =1.6. The server calculates the current iteration value of the entity learning factor:

[0117] c1=c1 base *Diff1+c2 base *Diff2 = 1.8 * 0.0455 + 0.5 * 0.0303 ≈ 0.0819 + 0.0152 ≈ 0.0971. This calculates the current iteration value of the relation learning factor.

[0118] c2=c3 base *Diff1+c4 base *Diff2=0.4*0.0455+1.6*0.0303≈0.0182+0.0485≈0.0667.

[0119] At this point, the server is an Entity. 042 In the 50th iteration, three key factors were dynamically calculated to update the entity's state and attribute vectors: semantic weight factor w≈0.509, entity learning factor c1≈0.0971, and relation learning factor c2≈0.0667. These factors will be used in subsequent vector composition operations to ensure that the entity... 042 Its exploratory behavior can adaptively and intelligently adjust itself based on the optimization process, the gap between its own experience and that of the group, and the concentration of the group distribution.

[0120] In this embodiment of the invention, the step of calling the knowledge graph initialization module to generate an initial entity relationship set corresponding to the monitoring node based on the carbon flow topology graph can be implemented through the following example.

[0121] The number of carbon flow nodes in the carbon flow topology graph and the type of carbon flow node corresponding to each carbon flow node are analyzed. The carbon flow node types include raw material acquisition node type, production and manufacturing node type, transportation and distribution node type, usage node type and waste recycling node type.

[0122] Based on the carbon flow node type and the path carbon flow intensity value of the carbon flow transmission path between the carbon flow nodes, the key carbon footprint parameter types that the monitoring node needs to optimize under the current management state are determined. The key carbon footprint parameter types include the energy structure adjustment parameters, the optimization intensity parameters, and the route scheduling timing adjustment parameters.

[0123] The preset value ranges of the energy structure adjustment parameters, the preset value ranges of the optimization intensity parameters, and the preset value ranges of the route scheduling timing adjustment parameters are obtained. The preset value ranges of the route scheduling timing adjustment parameters are determined by the boundary values ​​of the start time interval and the stop time interval of the transportation and distribution stage.

[0124] The uniform random number generator inside the knowledge graph initialization module is invoked to randomly generate an initial energy structure value for each carbon footprint entity within the preset value range of the energy structure adjustment parameter, an initial optimization intensity value for each carbon footprint entity within the preset value range of the optimization intensity parameter, and an initial route scheduling timing parameter set for each carbon footprint entity within the preset value range of the route scheduling timing adjustment parameter.

[0125] The initial energy structure values, the initial optimization intensity values, and the initial route scheduling timing parameter group are combined according to a preset parameter arrangement order to generate the initial attribute vector of each carbon footprint entity.

[0126] Set the initial attribute vector of each carbon footprint entity to the initial individual historical best attribute of that carbon footprint entity;

[0127] The uniform random number generator inside the knowledge graph initialization module is invoked to randomly generate an initial state vector for each carbon footprint entity within a preset range of state values.

[0128] The initial attribute vectors and initial state vectors of all carbon footprint entities are aggregated and processed to generate the initial entity relationship set.

[0129] In this embodiment of the invention, for example, firstly, the server parses the constructed carbon flow topology graph. It counts the number of all carbon flow nodes in the graph, for example, identifying a total of 15 nodes. Simultaneously, the server labels each node with its carbon flow node type. Based on the data source and business logic, the server classifies these nodes into: 3 raw material acquisition node types (corresponding to iron ore, bauxite, and rubber mining sites), 5 production and manufacturing node types (corresponding to steel mills, parts foundries, assembly plants, etc.), 4 transportation and distribution node types (corresponding to ports, logistics centers, and delivery fleets), 2 usage node types (corresponding to customer factories in different regions), and 1 waste recycling node type (corresponding to a recycling and processing center).

[0130] Next, the server performs diagnostic analysis by combining the node types and the carbon flow intensity values ​​(i.e., carbon emission flows) along the carbon flow transmission paths connecting these nodes. The server found that among all carbon flow transmission paths, the path connecting five manufacturing nodes generally had the highest carbon flow intensity values, which were also highly volatile; while the path connecting four transportation and distribution nodes showed significant intermittent peaks in its intensity value. Based on this analysis, the server determined that, under the current management conditions, the key carbon footprint parameters that need optimization focus on: energy structure adjustment parameters that directly affect the energy consumption composition of the manufacturing stage; process optimization intensity parameters used to reduce emissions during the manufacturing process; and route scheduling timing adjustment parameters used to mitigate emission peaks in the transportation stage.

[0131] Then, the server retrieves the preset value ranges for these three key parameters from the configuration database. The energy structure adjustment parameter (renewable energy percentage) is set to a range of [0.20, 0.80]. The optimization intensity parameter (technical transformation degree coefficient) is set to a range of [0, 1]. The route scheduling timing adjustment parameter is more complex, its range determined by the operational time window of the transportation and distribution stage; for example, for a transportation task from a warehouse to a customer, its start time interval is [06:00, 10:00], and its stop time interval is [15:00, 19:00]. These boundary values ​​collectively define the adjustment range of this set of time parameters.

[0132] Once preparation is complete, the initialization module begins creating a large number of candidate solutions (i.e., carbon footprint entities). Assume the server sets the initial population size to 100 entities. It calls the module's internal uniform random number generator to perform the following operations:

[0133] For the first to the 100th carbon footprint entities, an initial energy structure value is randomly generated within the range [0.20, 0.80], for example, Entity. 001 0.35, Entity 002 We get 0.62, ..., Entity 100 The result is 0.71.

[0134] Similarly, within the range [0,1], an initial optimization strength value is randomly generated for each entity, for example, 0.12, 0.88, ..., 0.45 respectively.

[0135] Within the time window defined, an initial set of route scheduling timing parameters is randomly generated for each entity. For example, for Entity 001 Generate {start:08:15,stop:17:30}, which is an Entity. 002 Generate {start:06:45,stop:18:45}, and so on.

[0136] Subsequently, the server combines the three random initial values ​​corresponding to each entity according to a preset parameter arrangement order (such as [energy structure, optimization intensity, route scheduling sequence]) to form the initial attribute vector of that entity. For example, Entity 001 The initial attribute vector is [0.35, 0.12, {08:15, 17:30}].

[0137] Next, the server directly sets this initial attribute vector for each carbon footprint entity as the entity's initial historical best attribute. This is because, at the initial moment, the entity has not yet undergone any iterations, and the first solution it finds is the current historical best solution.

[0138] For subsequent iterations to update the state vector, the server also needs to initialize the state of each entity. It again calls a uniform random number generator to randomly generate an initial state vector for each entity within a preset range of state values ​​(e.g., [-0.1, 0.1]). This vector has the same dimension as the attribute vector, for example, Entity. 001 The initial state vector may be [0.05, -0.03, {+00:02, -00:01}].

[0139] Finally, the server aggregates the 100 carbon footprint entities (each containing its initial attribute vector and initial state vector) into a structured data set. This set constitutes the initial entity relationship set, which forms the starting point for the knowledge graph validation algorithm to perform iterative search and optimization. It contains a large number of randomly generated potential management schemes covering the feasible domains of various parameters.

[0140] In this embodiment of the invention, the construction of the carbon flow topology map of the monitoring node within the continuous time period based on the carbon emission data of the raw material acquisition stage, the carbon emission data of the production and manufacturing stage, the carbon emission data of the transportation and distribution stage, and the carbon emission data of the usage stage can be implemented through the following example.

[0141] Feature extraction processing is performed on the carbon emission data of the raw material acquisition stage to obtain the raw material carbon emission intensity sequence and the raw material carbon emission phase sequence. Feature extraction processing is performed on the carbon emission data of the production and manufacturing stage to obtain the production carbon emission intensity sequence and the production carbon emission phase sequence.

[0142] Feature extraction processing is performed on the carbon emission data of the transportation and distribution stage to obtain the transportation carbon emission intensity sequence and the transportation carbon emission phase sequence. Feature extraction processing is performed on the carbon emission data of the usage stage to obtain the usage carbon emission intensity sequence and the usage carbon emission phase sequence.

[0143] The instantaneous direct carbon emission value and the instantaneous indirect carbon emission value are calculated based on the raw material carbon emission intensity sequence, the production carbon emission intensity sequence, the raw material carbon emission phase sequence, and the production carbon emission phase sequence.

[0144] The instantaneous process carbon emission value and the instantaneous cumulative carbon emission value are calculated based on the transport carbon emission intensity sequence, the usage carbon emission intensity sequence, the transport carbon emission phase sequence, and the usage carbon emission phase sequence.

[0145] By analyzing the carbon emission data of the waste recycling stage, the set of processing time intervals and the set of idle time intervals of the waste recycling stage within the continuous time period are obtained;

[0146] The set of carbon absorption state switching times and the set of carbon offset state switching times in the offsetting stage within the monitoring node are determined based on the set of processing time intervals and the set of idle time intervals.

[0147] The continuous time period is divided into multiple continuous carbon flow analysis sub-intervals based on the set of carbon absorption state switching times and the set of carbon offset state switching times.

[0148] Within each carbon flow analysis sub-interval, a sub-interval carbon flow balance equation is established based on the instantaneous direct carbon emission value, the instantaneous indirect carbon emission value, the instantaneous process carbon emission value, and the instantaneous cumulative carbon emission value.

[0149] Extract the carbon flow direction parameter and carbon flow amplitude parameter from the carbon flow balance equation of the sub-interval; determine the path start endpoint identifier and path end endpoint identifier of the carbon flow transmission path between the carbon flow nodes based on the carbon flow direction parameter; and determine the path carbon flow intensity value of the carbon flow transmission path based on the carbon flow amplitude parameter.

[0150] The carbon flow node identifiers, carbon flow transmission path identifiers, path start endpoint identifiers, path end endpoint identifiers, and path carbon flow intensity values ​​corresponding to multiple carbon flow analysis sub-intervals are associated and combined in chronological order to generate the carbon flow topology map.

[0151] In an embodiment of the present invention, for example, after receiving raw carbon emission data for a continuous period (e.g., the past 24 hours) reported by each monitoring node throughout the product's lifecycle, the server initiates a carbon flow topology map construction process.

[0152] First, the server performs deep feature extraction on the raw data. For carbon emission data from the raw material acquisition stage, the server applies sliding window analysis and Fourier transform to separate the raw material carbon emission intensity sequence (e.g., emission per minute sequence: [10.2, 10.5, 9.8, ...] kgCO2 / min) representing the average emission level and the raw material carbon emission phase sequence (e.g., timestamp sequence marking the start of high-intensity emission events: [00:05, 01:20, 02:35, ...]) representing periodic operations (e.g., blasting, loading and unloading cycles). The same processing is applied to the carbon emission data from the manufacturing stage to obtain the production carbon emission intensity sequence (e.g., [205.6, 210.3, 198.7, ...] kgCO2 / min) and the production carbon emission phase sequence (e.g., [00:10, 00:40, 01:10, ...]). Similarly, processing data from the transportation and distribution phase yields the transportation carbon emission intensity sequence and the transportation carbon emission phase sequence, while processing data from the usage phase yields the usage carbon emission intensity sequence and the usage carbon emission phase sequence.

[0153] Next, the server uses these feature sequences to calculate four types of instantaneous carbon emission values. It calculates the instantaneous direct carbon emission value (e.g., 315 kg CO2 / min at t=00:30) by weighting the components corresponding to direct fuel consumption in the raw material carbon emission intensity sequence and the production carbon emission intensity sequence, as well as the time alignment determined by the raw material carbon emission phase sequence and the production carbon emission phase sequence. Simultaneously, it calculates the instantaneous direct carbon emission value (e.g., 185 kg CO2 / min at t=00:30) based on the intensity component corresponding to electricity consumption. Based on the dynamic process reflected by the transportation carbon emission intensity sequence and the usage carbon emission intensity sequence and their phase sequences, the server obtains the instantaneous process carbon emission value (characterizing carbon in flow, e.g., 150 kg CO2 at t=00:30) through integration. All emissions are then accumulated to generate the instantaneous cumulative carbon emission value (e.g., 12500 kg CO2 at t=00:30).

[0154] Meanwhile, the server analyzes carbon emission data from the waste recycling phase. It identifies the active operating periods of the recycling center within a 24-hour period, for example, determining the operating times of the processing equipment through energy consumption signals as [02:00-05:00], [10:00-13:00], and [18:00-21:00], forming a set of processing time intervals; the standby times are the remaining times, forming a set of idle time intervals. Based on this, the server determines the set of carbon absorption state switching times for this node as a carbon sink as {02:00, 05:00, 10:00, 13:00, 18:00, 21:00}. Assuming there is no additional carbon offsetting behavior in this stage, the set of carbon offsetting state switching times is empty.

[0155] Based on these switching time sets, the server divides the continuous 24 hours into multiple consecutive carbon flow analysis sub-intervals, for example: [00:00-02:00), [02:00-05:00), [05:00-10:00), [10:00-13:00), [13:00-18:00), [18:00-21:00), [21:00-24:00]. Within each sub-interval, the carbon source / sink status of the system remains consistent.

[0156] Within each carbon flow analysis sub-interval (taking [02:00-05:00] as an example), the server establishes the sub-interval carbon flow balance equation based on the instantaneous direct carbon emissions, instantaneous process carbon emissions, and instantaneous cumulative carbon emissions calculated within that interval, combined with known material balance relationships. This equation describes the mass conservation relationship of carbon from the raw material node, production node to the usage node.

[0157] Subsequently, the server extracts key carbon flow direction parameters (e.g., "flowing from node A to node B") and carbon flow amplitude parameters (e.g., "flow rate of X kg CO2 / min") from the equilibrium equation. Based on the carbon flow direction parameters, the server determines the path start and end point identifiers (e.g., "Node...") for each virtual carbon flow transport path. Prod1 ) and path termination endpoint identifier (such as "Node") Trans2 Based on the carbon flow amplitude parameter, the server determines the path carbon flow intensity value for that path (e.g., "85.3 kg CO2 / min").

[0158] Finally, the server performs a time-series association and combination of the carbon flow node identifiers, carbon flow transmission path identifiers, path start endpoint identifiers, path end endpoint identifiers, and path carbon flow intensity values ​​corresponding to all seven carbon flow analysis sub-intervals. For example, it records the path "Path1" starting from "Node" within the interval [02:00-05:00). Prod1 "To "Node" Trans2 The intensity of the path is 85.3; within the interval [05:00-10:00), the intensity of this path may change to 45.1. By integrating the dynamic information of all sub-intervals and all paths, the server generates a complete, time-varying carbon flow topology map, which depicts the panorama of carbon flow over time in the monitoring node network in a structured manner.

[0159] In this embodiment of the invention, the step of extracting the target management parameter vector corresponding to the globally optimal attribute of the initial entity relationship set when the knowledge graph verification algorithm reaches the preset iteration termination condition can be implemented through the following example.

[0160] The maximum number of iterations preset by the knowledge graph verification algorithm is obtained, and the current number of iterations completed is compared with the maximum number of iterations after each iteration.

[0161] When the current number of iterations completed is equal to the maximum number of iterations, the knowledge graph verification algorithm is determined to have reached the preset iteration termination condition.

[0162] When the number of iterations completed is less than the maximum number of iterations, obtain the global optimal carbon footprint compliance rate simulation value corresponding to the global optimal attribute of the initial entity relationship set, and compare the global optimal carbon footprint compliance rate simulation value with the preset carbon footprint compliance rate target threshold.

[0163] When the simulated value of the globally optimal carbon footprint compliance rate is greater than or equal to the target threshold of the carbon footprint compliance rate, the knowledge graph verification algorithm is determined to have reached the preset iteration termination condition.

[0164] When the simulated value of the global optimal carbon footprint compliance rate is less than the target threshold of the carbon footprint compliance rate, the attribute change of the global optimal attribute in multiple consecutive iterations is obtained, and the attribute change is compared with a preset attribute change convergence threshold.

[0165] When the change in the attribute is less than the convergence threshold of the attribute change, the knowledge graph verification algorithm is determined to have reached the preset iteration termination condition.

[0166] After determining that the knowledge graph verification algorithm has reached the preset iteration termination condition, the target management parameter vector corresponding to the global optimal attribute is extracted from the initial entity relationship set. The target management parameter vector includes the target energy structure value, the target optimization intensity value, and the target route scheduling time sequence parameter group.

[0167] In this embodiment of the invention, for example, during the operation of the knowledge graph verification algorithm, the server continuously monitors the iteration process to determine whether the termination condition has been met. The following describes the server's decision-making and execution process during algorithm operation.

[0168] The server first obtains the maximum number of iterations preset by the algorithm, for example, T. max =1000. After each complete iteration (including metric calculations and attribute updates for all carbon footprint entities), the server will set the current number of iterations completed, t. current (Assuming the value is 850 at this point) and T max A comparison was performed. Since 850 is less than 1000, the server determined that the maximum number of iterations had not yet been reached.

[0169] Next, the server performs a second condition check. It reads the globally optimal attribute of the current initial entity relationship set from global shared memory and obtains the globally optimal carbon footprint compliance rate simulation value corresponding to that attribute. Assume that after this iteration, this value is Compliance. global =108.5%. The server compares this value with a preset carbon footprint compliance target threshold (e.g., Target). Compliance =110.0%) for comparison. Since 108.5% is less than 110.0%, the server determines that the current optimal solution has not yet reached the preset compliance rate target.

[0170] Subsequently, the server initiates a third convergence check. It retrieves the global optimal attribute values ​​from historical records over multiple consecutive iterations (e.g., the most recent 50 iterations). Since the global optimal attribute is a vector (e.g., [energy structure, optimization intensity, route scheduling timing]), the server calculates the change in this vector over these 50 iterations, for example, by calculating the moving average of its Euclidean distance as the attribute change Delta. attr Assuming Delta is calculated... attr =0.015. The server compares this value with the preset attribute change convergence threshold Epsilon=0.02. Since 0.015 is less than 0.02, the server determines that the change in the globally optimal attribute is very small, the optimization process tends to stabilize, and the convergence condition is met.

[0171] Based on this, the server determines that the knowledge graph verification algorithm running this time has reached the preset iteration termination condition (specifically, it has met the third convergence condition mentioned above).

[0172] Upon termination, the server immediately performs a result extraction operation. It locates and extracts the data set marked as having the globally optimal attribute from the initial entity relation set. This data set itself is a parameter vector, specifically the target management parameter vector. The server parses this vector to obtain its specific values: target energy structure values ​​(e.g., renewable energy percentage 0.68%), target optimization intensity values ​​(e.g., process optimization level 0.82), and target route scheduling time sequence parameter sets (e.g., a specific transportation task schedule {Route1}). start 08:15, Route1 stop :16:45;Route2 start 09:30, Route2 stop (18:00) At this point, the server has completed the optimization search process and obtained a set of verified optimal management parameter combinations that can be used for actual production and operational control.

[0173] In this embodiment of the invention, the step of generating real-time carbon footprint optimization management instructions based on the target management parameter vector and sending the real-time carbon footprint optimization management instructions to the central processing unit of the product carbon footprint management system can be implemented through the following example.

[0174] Analyze the target energy structure values, target optimization intensity values, and target route scheduling time sequence parameter groups in the target management parameter vector;

[0175] The target energy structure value is converted into an energy ratio setting value that matches the energy efficiency management strategy of the central processor, and an energy structure control word is generated.

[0176] The target optimization intensity value is converted into an optimization level setting value that matches the adjustment range of the process optimization module of the central processing unit, and an optimization intensity control word is generated.

[0177] The target route scheduling timing parameter group is converted into the start and stop times of the execution instructions of the transportation and distribution stage on the time axis, generating a route scheduling timing control instruction sequence.

[0178] The energy structure control word, the optimization intensity control word, and the route scheduling timing control instruction sequence are encapsulated according to the communication protocol format supported by the central processing unit to generate the real-time carbon footprint optimization management instruction.

[0179] The real-time carbon footprint optimization management instructions are sent to the instruction register address space of the central processing unit through the communication interface in the product carbon footprint management system.

[0180] The central processing unit is triggered to read the energy structure control word, the optimization intensity control word, and the route scheduling timing control instruction sequence stored in the instruction register address space;

[0181] The central processing unit adjusts the output strategy of its internal energy management module according to the energy structure control word, adjusts the output intensity of its internal process optimization module according to the optimization intensity control word, and adjusts the timing of the execution instructions output by its internal scheduling module to the transportation and distribution stage according to the route scheduling timing control instruction sequence.

[0182] In an embodiment of the present invention, for example, after the knowledge graph verification algorithm reaches the iteration termination condition and extracts the target management parameter vector, the server immediately starts the instruction generation and issuance process to implement the optimization strategy.

[0183] The server first parses the target management parameter vector. It extracts from the vector the target energy structure value (e.g., 0.68), the target optimization intensity value (e.g., 0.82), and the target route scheduling timing parameter set (e.g., {Route...}). A [08:15, 16:45], Route B [09:30, 18:00]}).

[0184] Next, the server performs instruction translation. It converts the target energy structure value of 0.68 into an energy allocation setting that can be directly interpreted by the downstream central processing unit (CPU)'s energy efficiency management strategy. Assuming the CPU requires a percentage-based power allocation for the three energy channels, the server calculates and generates a structured energy structure control word, such as ENERGY. CFG :PV=50,WIND=18,GRID=32 indicates that the power supplied by photovoltaics, wind power, and grid power accounts for 50%, 18%, and 32%, respectively.

[0185] Simultaneously, the server converts the target optimization intensity value of 0.82 into an optimization level setting that can be recognized by the CPU's internal process optimization module. Assuming the optimization intensity is mapped to levels 1-5, the server generates an optimization intensity control word, such as OPTIM, through linear mapping and rounding. LVL :4;PARAM:HEAT TEMP =850,FAN SPD =60 indicates that the Level 4 optimization scheme is enabled, with the specific parameters being a heat treatment temperature of 850°C and a fan speed of 60%.

[0186] For the target route scheduling timing parameter set, the server converts it into a series of specific start and stop time point instructions on the timeline. It generates a route scheduling timing control instruction sequence, such as SCHEDULE:TASK. A ,START=08:15,STOP=16:45;TASK B ,START=09:30,STOP=18:00

[0187] Subsequently, the server encapsulates the three control words and instruction sequence into a complete data packet, including a frame header, address code, data body, checksum, and frame trailer, according to the communication protocol format specified by the central processing unit (such as an industrial control protocol based on TCP / IP). This data packet is the real-time carbon footprint optimization management instruction.

[0188] After encapsulation, the server uses its equipped industrial Ethernet card (communication interface) and standard Socket communication to send the encapsulated instruction data packet to the instruction register address space of the central processing unit. This address space is predefined in the communication protocol, for example, a specific memory area mapped to port 502 of IP address 192.168.1.100.

[0189] After the data is successfully written, the server triggers the CPU to read the newly stored data from its instruction register address space by sending a specific hardware interrupt signal or setting a flag. The CPU's firmware then responds, reading and parsing the energy structure control word, optimization intensity control word, and route scheduling timing control instruction sequence.

[0190] Finally, the central processing unit (CPU) executes control actions based on the parsed instructions. Based on the energy structure control word, it adjusts the output strategy of its internal energy management module, such as issuing new power scheduling instructions to the plant's microgrid controller. Based on the optimization intensity control word, it adjusts the output intensity of its internal process optimization module, such as sending new process parameter settings to the production line PLC (Programmable Logic Controller). Based on the route scheduling timing control instruction sequence, it adjusts the timing of the execution instructions output by its internal scheduling module to the transportation fleet management system, such as precisely updating the daily transportation task schedule. Thus, the server completes a full closed loop from intelligent optimization decision-making to physical execution control.

[0191] This invention provides a computer device 100, which includes a processor and a non-volatile memory storing computer instructions. When the computer instructions are executed by the processor, the computer device 100 executes the aforementioned product carbon footprint information management method based on knowledge graph verification. Figure 2 As shown, Figure 2 This is a structural block diagram of a computer device 100 provided in an embodiment of the present invention. The computer device 100 includes a memory 111, a processor 112, and a communication unit 113. To enable data transmission or interaction, the memory 111, processor 112, and communication unit 113 are electrically connected to each other directly or indirectly. For example, these components can be electrically connected to each other through one or more communication buses or signal lines.

[0192] For illustrative purposes, the foregoing description has been made with reference to specific embodiments. However, the foregoing illustrative discussions are not intended to be exhaustive or to limit the present disclosure to the precise forms disclosed. Numerous modifications and variations are possible in accordance with the foregoing teachings. These embodiments were chosen and described in order to best illustrate the principles of the present disclosure and its practical application, thereby enabling those skilled in the art to best utilize the disclosure and to employ various embodiments with different modifications to suit a particular intended application.

Claims

1. A product carbon footprint information management method based on knowledge graph verification, characterized in that, include: The carbon footprint data stream generated by monitoring nodes in the entire product lifecycle chain within a continuous time period is collected. The carbon footprint data stream is parsed and processed to obtain the instantaneous carbon footprint parameter set of the monitoring node. The instantaneous carbon footprint parameter set includes carbon emission data in the raw material acquisition stage, carbon emission data in the production and manufacturing stage, carbon emission data in the transportation and distribution stage, carbon emission data in the usage stage, and carbon emission data in the waste recycling stage. Based on the carbon emission data of the raw material acquisition stage, the carbon emission data of the production and manufacturing stage, the carbon emission data of the transportation and distribution stage, and the carbon emission data of the usage stage, a carbon flow topology map of the monitoring node is constructed in the continuous time period. The carbon flow topology map consists of carbon flow nodes and carbon flow transmission paths connecting the carbon flow nodes. The carbon flow nodes correspond to each stage of the entire life cycle of the monitoring node. The knowledge graph initialization module is invoked to generate an initial entity relationship set corresponding to the monitoring node based on the carbon flow topology graph. Each carbon footprint entity in the initial entity relationship set carries a candidate management parameter vector, which includes energy structure adjustment parameters for the production and manufacturing stage, optimization intensity parameters for the process flow, and route scheduling timing adjustment parameters for the transportation and distribution stage. The initial entity relationship set is iteratively verified using a knowledge graph verification algorithm. In each iteration, the system carbon footprint index corresponding to the carbon footprint entity is calculated based on the current attribute vector of the carbon footprint entity. The individual historical best attribute of the carbon footprint entity and the global best attribute of the initial entity relationship set are updated based on the system carbon footprint index. The state vector and attribute vector of the carbon footprint entity are updated based on dynamically adjusted semantic weight factor, entity learning factor and relationship learning factor. When the knowledge graph verification algorithm reaches the preset iteration termination condition, the target management parameter vector corresponding to the global optimal attribute of the initial entity relationship set is extracted, and a real-time carbon footprint optimization management instruction is generated based on the target management parameter vector and sent to the central processing unit of the product carbon footprint management system.

2. The method according to claim 1, characterized in that, The step of updating the individual historical optimal attribute of the carbon footprint entity and the globally optimal attribute of the initial entity relationship set based on the system carbon footprint index includes: Obtain the historical system carbon footprint index corresponding to the individual historical best attribute of the carbon footprint entity. The historical system carbon footprint index includes the historical total carbon emission simulation value, the historical carbon emission intensity simulation value, and the historical carbon footprint compliance rate simulation value. The simulated value of carbon footprint compliance rate in the system carbon footprint index corresponding to the carbon footprint entity is compared with the simulated value of historical carbon footprint compliance rate in the historical system carbon footprint index. When the simulated value of the carbon footprint compliance rate is greater than the simulated value of the historical carbon footprint compliance rate, the current attribute vector of the carbon footprint entity is set to the updated individual historical best attribute. When the simulated value of the carbon footprint compliance rate is equal to the simulated value of the historical carbon footprint compliance rate, a second comparison process is performed between the simulated value of the total carbon emissions in the system carbon footprint index corresponding to the carbon footprint entity and the simulated value of the total historical carbon emissions in the historical system carbon footprint index. When the simulated value of the carbon footprint compliance rate is equal to the simulated value of the historical carbon footprint compliance rate and the simulated value of the total carbon emissions is less than the simulated value of the total historical carbon emissions, the current attribute vector of the carbon footprint entity is set to the updated individual historical best attribute. When the simulated value of the carbon footprint compliance rate is equal to the simulated value of the historical carbon footprint compliance rate and the simulated value of the total carbon emissions is equal to the simulated value of the total carbon emissions, a third comparison process is performed between the simulated value of carbon emission intensity in the system carbon footprint index corresponding to the carbon footprint entity and the simulated value of historical carbon emission intensity in the historical system carbon footprint index. When the simulated value of the carbon footprint compliance rate is equal to the simulated value of the historical carbon footprint compliance rate, the simulated value of the total carbon emissions is equal to the simulated value of the total historical carbon emissions, and the simulated value of the carbon emission intensity is less than the simulated value of the historical carbon emission intensity, the current attribute vector of the carbon footprint entity is set to the updated individual historical best attribute. Traverse all carbon footprint entities in the initial entity relationship set, obtain the updated carbon footprint compliance rate simulation value corresponding to the updated individual historical best attribute of each carbon footprint entity, and set the individual historical best attribute corresponding to the maximum value among all the updated carbon footprint compliance rate simulation values ​​as the global best attribute of the initial entity relationship set.

3. The method according to claim 1, characterized in that, The step of updating the state vector and attribute vector of the carbon footprint entity based on dynamically adjusted semantic weight factors, entity learning factors, and relation learning factors includes: Obtain the historical state vector of the carbon footprint entity in the previous iteration. The historical state vector is composed of the state components of the energy structure adjustment parameter, the optimization intensity parameter, and the route scheduling timing adjustment parameter in the previous iteration, arranged in a preset parameter order. Obtain a first difference vector between the current attribute vector of the carbon footprint entity and the individual historical best attribute of the carbon footprint entity. The first difference vector is composed of the difference between the current value of the energy structure adjustment parameter and the individual historical best energy structure value, the difference between the current value of the optimization intensity parameter and the individual historical optimal intensity value, and the difference between the current value of the route scheduling timing adjustment parameter and the individual historical best route scheduling timing parameter group, according to a preset parameter arrangement order. Obtain a second difference vector between the current attribute vector of the carbon footprint entity and the globally optimal attribute of the initial entity relationship set. The second difference vector is composed of the difference between the current value of the energy structure adjustment parameter and the globally optimal energy structure value, the difference between the current value of the optimization intensity parameter and the globally optimal intensity value, and the difference between the current value of the route scheduling timing adjustment parameter and the globally optimal route scheduling timing parameter set, arranged in a preset parameter order. The current iterative calculation value of the semantic weight factor is generated based on the semantic proximity between the current attribute vector of the carbon footprint entity and the individual's historical best attribute, the semantic proximity between the current attribute vector of the carbon footprint entity and the global best attribute, and the semantic distribution dispersion of all carbon footprint entities in the initial entity relationship set. The current iterative calculation value of the entity learning factor and the current iterative calculation value of the relation learning factor are generated based on the first difference degree between the simulated total carbon emissions corresponding to the individual historical best attribute of the carbon footprint entity and the simulated total carbon emissions corresponding to the global best attribute, and the second difference degree between the simulated carbon emission intensity corresponding to the individual historical best attribute of the carbon footprint entity and the simulated carbon emission intensity corresponding to the global best attribute. Perform a scalar multiplication operation between the historical state vector and the current iterative calculated value of the semantic weight factor to generate the first state update component vector; Perform a scalar multiplication operation between the first difference vector and the current iterative calculation value of the entity learning factor to generate a first intermediate vector. Perform a scalar multiplication operation between the first intermediate vector and the first random value output by the first random number generator to generate a second state update component vector. Perform a scalar multiplication operation between the second difference vector and the current iterative calculation value of the relation learning factor to generate a second intermediate vector. Perform a scalar multiplication operation between the second intermediate vector and the second random value output by the second random number generator to generate a third state update component vector. The first state update component vector, the second state update component vector, and the third state update component vector are added together to generate the updated state vector of the carbon footprint entity. The updated attribute vector of the carbon footprint entity is generated by performing a vector addition operation between the current attribute vector of the carbon footprint entity and the updated state vector.

4. The method according to claim 1, characterized in that, The step of calculating the system carbon footprint index corresponding to the carbon footprint entity based on the current attribute vector of the carbon footprint entity includes: Obtain the current energy structure value of the energy structure adjustment parameter, the current optimization intensity value of the optimization intensity parameter, and the current route scheduling timing parameter group of the route scheduling timing adjustment parameter contained in the current attribute vector of the carbon footprint entity. The current energy structure value, the current optimization intensity value, and the current route scheduling timing parameter group are input into the pre-built product carbon footprint system simulation model. The product carbon footprint system simulation model contains the full life cycle topology structure corresponding to the monitoring node. The carbon footprint system simulation model is used to simulate and calculate the carbon flow status of the monitoring node under the management of the current energy structure value, the current optimization intensity value, and the current route scheduling time sequence parameter group, and to generate the carbon emission data of the monitoring node in the simulation stage and the overall carbon emission data in the simulation within the preset simulation time. Cluster analysis is performed on the carbon emission data of the simulation stage to extract the emission amplitudes of the main emission sources and the emission amplitudes of each secondary emission source in the carbon emission data of the simulation stage. The total simulated value of carbon emission corresponding to the carbon emission data of the simulation stage is calculated based on the emission amplitudes of the main emission sources and all the emission amplitudes of the secondary emission sources. Intensity component analysis is performed on the simulated overall carbon emission data to extract the carbon emission per unit output value and carbon emission per unit product from the simulated overall carbon emission data. The simulated carbon emission intensity value corresponding to the simulated overall carbon emission data is calculated based on the ratio of the carbon emission per unit product to the carbon emission per unit output value. Based on the carbon emission data of the simulation phase and the overall carbon emission data of the simulation, the actual carbon emission integral value and the compliant carbon emission threshold integral value of the monitoring node within the preset simulation time are calculated, and the ratio of the compliant carbon emission threshold integral value to the actual carbon emission integral value is used as the carbon footprint compliance rate simulation value. The simulated values ​​of total carbon emissions, carbon emission intensity, and carbon footprint compliance rate are combined to generate the system carbon footprint index corresponding to the carbon footprint entity.

5. The method according to claim 1, characterized in that, The step of updating the state vector and attribute vector of the carbon footprint entity based on dynamically adjusted semantic weight factors, entity learning factors, and relation learning factors further includes: Obtain the current attribute vectors of all carbon footprint entities in the initial entity relationship set, and calculate the average central attribute vector of all the current attribute vectors; Calculate the semantic distance between the current attribute vector and the average central attribute vector for each carbon footprint entity, and perform standard deviation calculation on all semantic distances to generate the semantic distribution dispersion parameter of all carbon footprint entities in the initial entity relation set; Obtain the current iteration number and the preset maximum iteration number, and calculate the ratio of the current iteration number to the maximum iteration number as the iteration process ratio parameter; Calculate the first semantic distance between the current attribute vector of the carbon footprint entity and the individual's historical best attribute, calculate the second semantic distance between the current attribute vector of the carbon footprint entity and the global best attribute, sum the first semantic distance and the second semantic distance to generate the individual global deviation parameter of the carbon footprint entity; The iteration process ratio parameter, the individual global deviation parameter, and the semantic distribution dispersion parameter are input into the semantic weight dynamic adjustment function. The semantic weight dynamic adjustment function is used to correct the preset initial benchmark value of the semantic weight to generate the current iteration calculation value of the semantic weight factor. Obtain the first absolute difference between the simulated total carbon emissions value corresponding to the individual's historical best attribute and the simulated total carbon emissions value corresponding to the global best attribute, and divide the first absolute difference by the simulated total carbon emissions value corresponding to the global best attribute to generate a first difference degree. Obtain the second absolute difference between the simulated carbon emission intensity value corresponding to the individual's historical best attribute and the simulated carbon emission intensity value corresponding to the global best attribute, and divide the second absolute difference by the simulated carbon emission intensity value corresponding to the global best attribute to generate a second difference degree; The first difference degree is multiplied by a preset first learning factor benchmark value to generate a first product result. The second difference degree is multiplied by a preset second learning factor benchmark value to generate a second product result. The first product result and the second product result are summed to generate the current iterative calculation value of the entity learning factor. The first difference is multiplied by a preset third learning factor benchmark value to generate a third product result. The second difference is multiplied by a preset fourth learning factor benchmark value to generate a fourth product result. The third product result and the fourth product result are summed to generate the current iterative calculation value of the relation learning factor.

6. The method according to claim 1, characterized in that, The invocation of the knowledge graph initialization module generates an initial entity relationship set corresponding to the monitoring node based on the carbon flow topology graph, including: The number of carbon flow nodes in the carbon flow topology graph and the type of carbon flow node corresponding to each carbon flow node are analyzed. The carbon flow node types include raw material acquisition node type, production and manufacturing node type, transportation and distribution node type, usage node type and waste recycling node type. Based on the carbon flow node type and the path carbon flow intensity value of the carbon flow transmission path between the carbon flow nodes, the key carbon footprint parameter types that the monitoring node needs to optimize under the current management state are determined. The key carbon footprint parameter types include the energy structure adjustment parameters, the optimization intensity parameters, and the route scheduling timing adjustment parameters. The preset value ranges of the energy structure adjustment parameters, the preset value ranges of the optimization intensity parameters, and the preset value ranges of the route scheduling timing adjustment parameters are obtained. The preset value ranges of the route scheduling timing adjustment parameters are determined by the boundary values ​​of the start time interval and the stop time interval of the transportation and distribution stage. The uniform random number generator inside the knowledge graph initialization module is invoked to randomly generate an initial energy structure value for each carbon footprint entity within the preset value range of the energy structure adjustment parameter, an initial optimization intensity value for each carbon footprint entity within the preset value range of the optimization intensity parameter, and an initial route scheduling timing parameter set for each carbon footprint entity within the preset value range of the route scheduling timing adjustment parameter. The initial energy structure values, the initial optimization intensity values, and the initial route scheduling timing parameter group are combined according to a preset parameter arrangement order to generate the initial attribute vector of each carbon footprint entity. Set the initial attribute vector of each carbon footprint entity to the initial individual historical best attribute of that carbon footprint entity; The uniform random number generator inside the knowledge graph initialization module is invoked to randomly generate an initial state vector for each carbon footprint entity within a preset range of state values. The initial attribute vectors and initial state vectors of all carbon footprint entities are aggregated and processed to generate the initial entity relationship set.

7. The method according to claim 1, characterized in that, The step of constructing a carbon flow topology map of the monitoring node within the continuous time period based on carbon emission data from the raw material acquisition stage, the production and manufacturing stage, the transportation and distribution stage, and the usage stage includes: Feature extraction processing is performed on the carbon emission data of the raw material acquisition stage to obtain the raw material carbon emission intensity sequence and the raw material carbon emission phase sequence. Feature extraction processing is performed on the carbon emission data of the production and manufacturing stage to obtain the production carbon emission intensity sequence and the production carbon emission phase sequence. Feature extraction processing is performed on the carbon emission data of the transportation and distribution stage to obtain the transportation carbon emission intensity sequence and the transportation carbon emission phase sequence. Feature extraction processing is performed on the carbon emission data of the usage stage to obtain the usage carbon emission intensity sequence and the usage carbon emission phase sequence. The instantaneous direct carbon emission value and the instantaneous indirect carbon emission value are calculated based on the raw material carbon emission intensity sequence, the production carbon emission intensity sequence, the raw material carbon emission phase sequence, and the production carbon emission phase sequence. The instantaneous process carbon emission value and the instantaneous cumulative carbon emission value are calculated based on the transport carbon emission intensity sequence, the usage carbon emission intensity sequence, the transport carbon emission phase sequence, and the usage carbon emission phase sequence. By analyzing the carbon emission data of the waste recycling stage, the set of processing time intervals and the set of idle time intervals of the waste recycling stage within the continuous time period are obtained; The set of carbon absorption state switching times and the set of carbon offset state switching times in the offsetting stage within the monitoring node are determined based on the set of processing time intervals and the set of idle time intervals. The continuous time period is divided into multiple continuous carbon flow analysis sub-intervals based on the set of carbon absorption state switching times and the set of carbon offset state switching times. Within each carbon flow analysis sub-interval, a sub-interval carbon flow balance equation is established based on the instantaneous direct carbon emission value, the instantaneous indirect carbon emission value, the instantaneous process carbon emission value, and the instantaneous cumulative carbon emission value. Extract the carbon flow direction parameter and carbon flow amplitude parameter from the carbon flow balance equation of the sub-interval; determine the path start endpoint identifier and path end endpoint identifier of the carbon flow transmission path between the carbon flow nodes based on the carbon flow direction parameter; and determine the path carbon flow intensity value of the carbon flow transmission path based on the carbon flow amplitude parameter. The carbon flow node identifiers, carbon flow transmission path identifiers, path start endpoint identifiers, path end endpoint identifiers, and path carbon flow intensity values ​​corresponding to multiple carbon flow analysis sub-intervals are associated and combined in chronological order to generate the carbon flow topology map.

8. The method according to claim 1, characterized in that, When the knowledge graph verification algorithm reaches the preset iteration termination condition, the step of extracting the target management parameter vector corresponding to the globally optimal attribute of the initial entity relationship set includes: The maximum number of iterations preset by the knowledge graph verification algorithm is obtained, and the current number of iterations completed is compared with the maximum number of iterations after each iteration. When the current number of iterations completed is equal to the maximum number of iterations, the knowledge graph verification algorithm is determined to have reached the preset iteration termination condition. When the number of iterations completed is less than the maximum number of iterations, obtain the global optimal carbon footprint compliance rate simulation value corresponding to the global optimal attribute of the initial entity relationship set, and compare the global optimal carbon footprint compliance rate simulation value with the preset carbon footprint compliance rate target threshold. When the simulated value of the globally optimal carbon footprint compliance rate is greater than or equal to the target threshold of the carbon footprint compliance rate, the knowledge graph verification algorithm is determined to have reached the preset iteration termination condition. When the simulated value of the global optimal carbon footprint compliance rate is less than the target threshold of the carbon footprint compliance rate, the attribute change of the global optimal attribute in multiple consecutive iterations is obtained, and the attribute change is compared with a preset attribute change convergence threshold. When the change in the attribute is less than the convergence threshold of the attribute change, the knowledge graph verification algorithm is determined to have reached the preset iteration termination condition. After determining that the knowledge graph verification algorithm has reached the preset iteration termination condition, the target management parameter vector corresponding to the global optimal attribute is extracted from the initial entity relationship set. The target management parameter vector includes the target energy structure value, the target optimization intensity value, and the target route scheduling time sequence parameter group.

9. The method according to claim 1, characterized in that, The step of generating real-time carbon footprint optimization management instructions based on the target management parameter vector and sending the real-time carbon footprint optimization management instructions to the central processing unit of the product carbon footprint management system includes: Analyze the target energy structure values, target optimization intensity values, and target route scheduling time sequence parameter groups in the target management parameter vector; The target energy structure value is converted into an energy ratio setting value that matches the energy efficiency management strategy of the central processor, and an energy structure control word is generated. The target optimization intensity value is converted into an optimization level setting value that matches the adjustment range of the process optimization module of the central processing unit, and an optimization intensity control word is generated. The target route scheduling timing parameter group is converted into the start and stop times of the execution instructions of the transportation and distribution stage on the time axis, generating a route scheduling timing control instruction sequence. The energy structure control word, the optimization intensity control word, and the route scheduling timing control instruction sequence are encapsulated according to the communication protocol format supported by the central processing unit to generate the real-time carbon footprint optimization management instruction. The real-time carbon footprint optimization management instructions are sent to the instruction register address space of the central processing unit through the communication interface in the product carbon footprint management system. The central processing unit is triggered to read the energy structure control word, the optimization intensity control word, and the route scheduling timing control instruction sequence stored in the instruction register address space; The central processing unit adjusts the output strategy of its internal energy management module according to the energy structure control word, adjusts the output intensity of its internal process optimization module according to the optimization intensity control word, and adjusts the timing of the execution instructions output by its internal scheduling module to the transportation and distribution stage according to the route scheduling timing control instruction sequence.

10. A server system, characterized in that, Includes a server, the server being used to perform the method according to any one of claims 1-9.

Images