A carbon emission management method and system for environmentally friendly cables

Abstract

The present application relates to the technical field of industrial carbon footprint management, in particular to a carbon emission management method and system for environment-friendly cables, comprising: obtaining energy and material flow data of each stage of the whole life cycle of environment-friendly cables, calling a life cycle assessment model to generate a theoretical carbon emission list after normalization processing. Comparing the theoretical list with actual operation data of the production execution system and Internet of Things sensors to obtain a carbon emission difference data set. Inputting the data set into a gradient boosting tree algorithm optimized according to the cable process path dependence relationship to identify key process parameters and abnormal production events that cause the difference. Based on the identification result, a carbon emission optimization strategy for a specific batch is generated and bound with a product digital identifier to form a traceable low-carbon product electronic passport. The method breaks the limitation of independent process parameter processing, accurately locates the root cause of carbon emission abnormalities, eliminates the data gap between theoretical accounting and actual operation, and realizes single-product-level carbon footprint visualization and traceability.

Description

Technical Field

[0001] This invention relates to the field of industrial carbon footprint management technology, and in particular to a carbon emission management method and system for environmentally friendly cables. Background Technology

[0002] Currently, carbon emission management for environmentally friendly cables primarily employs two methods: static life cycle assessment and discrete energy monitoring. Enterprises generally rely on general databases to perform theoretical calculations across the entire life cycle from raw material extraction to recycling, or utilize smart sensors to collect overall energy consumption data for the production process. These two methods are physically isolated when dealing with cable production scenarios. Theoretical calculations depend on fixed coefficients and cannot respond to dynamic impacts such as equipment aging, raw material batch fluctuations, and environmental changes, leading to calculation results that deviate from actual emissions. Discrete monitoring systems can only report total energy consumption and cannot analyze the causal relationship between energy consumption and specific process parameters, making it difficult to uncover the underlying causes of excessive carbon emissions.

[0003] Cable production involves multiple highly coupled processes, including wire drawing, stranding, insulation extrusion, and cabling. The parameter settings of each process directly determine the energy consumption of the next. Existing technologies treat the parameters of each process as independent variables, severing the physical transmission logic of the process path. This often leads to misjudging key control points during anomaly tracing and an inability to pinpoint minute process drifts. Furthermore, existing carbon management data is mostly aggregated at the enterprise level, lacking strong correlation with specific product batches. This results in a lack of targeted optimization strategies and an inability to trace the low-carbon history of individual products, failing to meet the supply chain's requirements for carbon data transparency. This invention aims to address the aforementioned problems of disconnect between theory and practice, missing process correlations, and data traceability gaps. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the existing technology and propose a carbon emission management method and system for environmentally friendly cables.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a carbon emission management method for environmentally friendly cables, comprising:

[0006] Obtain a set of stage data for the entire life cycle of environmentally friendly cables, including energy consumption and material flow data for each stage of raw material mining, production and manufacturing, logistics and transportation, use and operation, and scrapping and recycling.

[0007] The data set of the aforementioned stages is cleaned and normalized to convert data from different sources and dimensions into a unified carbon emission accounting factor.

[0008] Using the converted carbon emission accounting factor, the life cycle assessment model is invoked to calculate the theoretical carbon emission inventory of the environmentally friendly cable at each stage of its entire life cycle;

[0009] The theoretical carbon emission inventory is compared with actual operational data from the production execution system and IoT sensors to obtain a carbon emission difference dataset;

[0010] The carbon emission difference dataset is input into the improved gradient boosting tree algorithm, which is optimized based on the process path dependency of environmentally friendly cable production.

[0011] The improved gradient boosting tree algorithm is used to identify key process parameters and abnormal production events that cause differences in carbon emissions.

[0012] Based on the identified key process parameters and abnormal production events, a set of carbon emission optimization strategies is generated for specific production batches or specific processes.

[0013] The carbon emission optimization strategy set is linked to the digital identification of cable products to form a traceable electronic passport for low-carbon products.

[0014] As a further aspect of the present invention, the stage data set is subjected to data cleaning and normalization processing to convert data from different sources and dimensions into a unified carbon emission accounting factor, including:

[0015] Outlier removal is performed on energy consumption and material flow data in the raw material mining process. Based on the box plot method, energy consumption values ​​that exceed the preset quantile range are identified and removed.

[0016] Standardize the format of multi-source heterogeneous data in the production and manufacturing process, and convert the energy consumption readings reported by different equipment interfaces into a continuous energy consumption sequence under standard timestamps and standard power units.

[0017] A distance-load weighted algorithm is applied to fuel consumption data in the logistics and transportation process to calculate the standardized transportation energy consumption coefficient per unit of goods based on transportation mileage and cargo weight.

[0018] Long-term operating data from the operational phase are filtered using a moving average to eliminate the impact of short-term fluctuations on long-term energy consumption assessment and generate a stable operating energy consumption baseline.

[0019] The processed data from each stage are converted into carbon dioxide equivalents based on a common carbon emission factor database, and the carbon emission accounting factors for each stage are obtained by summarizing the data.

[0020] As a further aspect of the present invention, the step of using the converted carbon emission accounting factor to call the life cycle assessment model to calculate the theoretical carbon emission inventory of the environmentally friendly cable at each stage of its entire life cycle includes:

[0021] Construct a product system boundary model for environmentally friendly cables, which clearly defines the physical and responsibility boundaries from raw material entry to product scrapping and recycling;

[0022] Within the product system boundary model, an input-output material and energy flow list is established for each stage. The material and energy flow list records in detail all materials and energy input to the stage, as well as all products, by-products, waste, and emissions output.

[0023] The carbon emission accounting factors of each stage are allocated to the corresponding input or output items in the material and energy flow list according to the material or energy type to which they belong.

[0024] Using the attribution-based life cycle assessment method, carbon emissions from upstream stages are cumulatively added to downstream stages to calculate the cumulative carbon emissions carried by the finished environmentally friendly cable.

[0025] The cumulative carbon emissions from all stages, from raw material mining to end-of-life recycling, are aggregated to form the theoretical carbon emission inventory broken down by stage. The theoretical carbon emission inventory lists the carbon emission contribution value of each stage in a structured table format.

[0026] As a further aspect of the present invention, the improved gradient boosting tree algorithm is optimized based on the process path dependency relationship of environmentally friendly cable production, and its working principle includes:

[0027] Construct a training dataset with carbon emission differences as the prediction target and multiple production state parameters as feature vectors;

[0028] In the basic iterative framework for constructing gradient boosting trees, process constraints are introduced as regularization terms. These process constraints are defined by the allowable fluctuation range of parameters of adjacent processes in the process path diagram.

[0029] When generating a new decision tree in each iteration, not only is the reduction of the prediction residual considered, but also whether the decision path of the new tree violates the process constraints, and penalty weights are applied to the split nodes that violate the constraints.

[0030] Time series correlation is incorporated into gradient calculation. For samples from the same production batch with a time sequence, their residual gradients are weighted according to the sequence of their processes to strengthen the dependencies on the process path.

[0031] Training stops when the total prediction error of the ensemble tree model no longer decreases or reaches the preset maximum number of iterations. The resulting ensemble model can automatically identify and follow the inherent path dependency logic of the production process when predicting carbon emission differences.

[0032] As a further aspect of the present invention, the process constraints are defined by the allowable fluctuation range of parameters of adjacent processes in the process path diagram, including:

[0033] Analyze the production process flow chart of environmentally friendly cables and extract all serially connected process pairs and parallel process groups;

[0034] For each serially connected process pair, statistical relationships are obtained from historical normal production data to determine the influence of the key control parameters of the subsequent process on the corresponding parameters of the preceding process, and the allowable fluctuation range is defined based on the confidence interval of the statistical relationships.

[0035] For parallel process groups, define the quality or energy consumption balance equation at their output convergence point, and define the coordination constraint range between parallel processes based on the allowable deviation of the balance equation during normal production.

[0036] The allowable fluctuation range of each serial operation pair and the coordination constraint range of each parallel operation group are encoded into mathematical constraints in the form of inequalities.

[0037] The set of all mathematical constraints in the form of inequalities is used as the process constraints and integrated into the loss function of the gradient boosting tree algorithm.

[0038] As a further aspect of the present invention, the improved gradient boosting tree algorithm is used to identify key process parameters and abnormal production events that lead to differences in carbon emissions, including:

[0039] The carbon emission difference dataset is input into the trained improved gradient boosting tree algorithm model to obtain the carbon emission difference prediction value and prediction contribution decomposition for each sample.

[0040] Based on the predicted contribution decomposition, the Shapley value of each input production state parameter feature on the final prediction result is calculated.

[0041] The Shapley values ​​of all features are sorted, and the set of features whose sum of Shapley values ​​exceeds a preset threshold is selected as the key process parameter.

[0042] During the model prediction process, the production period and equipment status corresponding to the samples that cause a sudden increase in prediction error are recorded;

[0043] By combining equipment logs and production event records, the production period and equipment status are analyzed to identify specific equipment failures, raw material batch switching, or manual intervention operations. The identified specific events are then marked as abnormal production events.

[0044] As a further aspect of the present invention, the step of generating a set of carbon emission optimization strategies for a specific production batch or a specific stage based on the identified key process parameters and abnormal production events includes:

[0045] For each identified key process parameter, its historical optimal operating range is queried. The historical optimal operating range is the common value range of the key process parameter in historical low-carbon emission batches.

[0046] The current real-time value of the key process parameter is compared with the historical optimal operating range. If the current real-time value deviates from the historical optimal operating range, parameter optimization suggestions are generated to adjust the key process parameter to within the historical optimal operating range.

[0047] For each identified abnormal production event, retrieve the standardized contingency plan for the corresponding event type from the knowledge base;

[0048] The parameter optimization suggestions and the standardized disposal plan are categorized and prioritized according to their corresponding production links, impact ranges and expected carbon emission reductions.

[0049] All suggestions and contingency plans that have been categorized and sorted are packaged into a set of carbon emission optimization strategies for a specific production batch or a specific process. The set of carbon emission optimization strategies includes specific operational steps, target values, and execution timing.

[0050] As a further aspect of the present invention, the carbon emission optimization strategy set is bound to the digital identifier of the cable product to form a traceable low-carbon product electronic passport, including:

[0051] A unique digital identification code is generated for each reel or batch of environmentally friendly cable products. The digital identification code encodes the product model, production batch, production line and production time information.

[0052] Establish a product database to associate and store the digital identification code, the calculated theoretical carbon emission inventory, the identified key process parameters and abnormal production events, and the generated set of carbon emission optimization strategies.

[0053] Based on the associated stored data, a structured electronic document is generated, which is the low-carbon product electronic passport. Its content includes at least: the carbon emission footprint of the product throughout its entire life cycle, analysis of the main carbon emission links, and a summary of the implementation of carbon emission optimization strategies in the production process.

[0054] The content of the low-carbon product electronic passport is hashed to generate a digital digest, and the digital digest is signed with a digital certificate to ensure the authenticity and immutability of the passport content.

[0055] The storage address or access index of the signed low-carbon product electronic passport is associated with the physical label or digital identification code of the product entity, so that downstream users or auditors can query and verify it.

[0056] As a further aspect of the present invention, the method also includes a closed-loop verification step for the effectiveness of the carbon emission optimization strategy:

[0057] After the set of carbon emission optimization strategies is applied to the production process, new rounds of production data are continuously collected after the application.

[0058] Based on the new round of production data, the actual carbon emission intensity is recalculated and compared with the historical carbon emission intensity before the application of the strategy.

[0059] If the reduction in carbon emission intensity reaches the expected target, the optimization strategy and its execution parameters will be confirmed as an effective strategy and stored in the optimization strategy knowledge base.

[0060] If carbon emission intensity does not decrease or even increases, analyze the reasons to determine whether it is due to strategy execution deviation, changes in external conditions, or the strategy itself being unsuitable, and generate strategy adjustment suggestions or failure flags.

[0061] The results of closed-loop verification are fed back into the training process of the improved gradient boosting tree algorithm to update model parameters, forming a closed-loop management process from policy generation, execution to verification and optimization.

[0062] As a further aspect of the present invention, the present invention also includes a carbon emission management system for environmentally friendly cables, the system including a processor and a memory, the memory storing a computer program, and the processor executing the computer program to implement the carbon emission management method for environmentally friendly cables as described above.

[0063] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0064] By embedding the process path dependencies of environmentally friendly cable production into the underlying logic of the gradient boosting tree algorithm, the algorithm overcomes the limitations of traditional machine learning that processes parameters independently. When constructing the decision tree node splitting rules and objective function, the algorithm introduces the unique physical production flow sequence of cables as a constraint, enabling the model to capture the cascading effects of fluctuations in insulation extrusion temperature on the energy consumption of subsequent cooling processes. Through this targeted structural optimization, the algorithm can accurately extract minute process parameter drifts or abnormal production events that cause carbon emissions to deviate from theoretical values ​​from high-dimensional and complex production data. This eliminates the interference of irrelevant variables, significantly improving the accuracy of anomaly tracing, and allowing managers to directly adjust specific equipment parameters rather than relying solely on generalized equipment maintenance based on experience.

[0065] Based on a differential comparison mechanism between theoretical carbon emission inventories and actual IoT sensor data, the generated differentiated datasets directly drive the generation of optimization strategies and are physically bound to the digital identifiers of cable products. This mechanism eliminates the data gap between theoretical calculations and actual operation in traditional management, no longer producing general workshop-level energy consumption reports, but outputting carbon emission optimization instructions specific to particular production batches or even specific production lines. By writing the set of optimization strategies into the product's digital passport, the entire carbon footprint of each cable product, from raw materials to end-of-life recycling, becomes visible and traceable. This binding method allows downstream users or regulatory agencies to directly access the product's actual carbon emission performance during the production process and corresponding corrective measures through the digital identifier. Attached Figure Description

[0066] Figure 1 This is a flowchart of a carbon emission management method for environmentally friendly cables according to the present invention;

[0067] Figure 2 A flowchart for data cleaning and normalization processing;

[0068] Figure 3 A flowchart illustrating the optimization principle of the improved gradient boosting tree algorithm. Detailed Implementation

[0069] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0070] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0071] See Figure 1 This invention provides a carbon emission management method for environmentally friendly cables, the overall implementation of which is as follows:

[0072] A comprehensive dataset of data across the entire lifecycle of environmentally friendly cables is acquired, encompassing energy consumption and material flow data at each stage: raw material extraction, manufacturing, logistics, operation, and end-of-life recycling. This dataset undergoes data cleaning and normalization, converting data from different sources and dimensions into a unified carbon emission accounting factor. Using this transformed carbon emission accounting factor, a lifecycle assessment model is invoked to calculate the theoretical carbon emission inventory for each stage of the environmentally friendly cable's lifecycle. This theoretical carbon emission inventory is compared with actual operational data from the production execution system and IoT sensors to obtain a carbon emission variance dataset. This dataset is then input into an improved gradient boosting tree algorithm, optimized based on the process path dependencies in environmentally friendly cable production. The improved algorithm identifies key process parameters and abnormal production events leading to carbon emission variances. Based on these identified parameters and events, a set of carbon emission optimization strategies is generated for specific production batches or stages. These strategies are then linked to the digital identifiers of the cable products, creating a traceable electronic passport for low-carbon products.

[0073] In one embodiment of the present invention, the stage data set is cleaned and normalized to convert data from different sources and dimensions into a unified carbon emission accounting factor. The process includes: (See [reference]). Figure 2 Outlier removal is performed on energy consumption and material flow data in the raw material extraction stage. Box plots are used to identify and remove energy consumption values ​​exceeding a preset quantile range. Multi-source heterogeneous data from the manufacturing stage are standardized, converting energy consumption readings reported from different equipment interfaces into continuous energy consumption sequences with standard timestamps and standard power units. For fuel consumption data in the logistics and transportation stage, a distance-load weighted algorithm is applied to calculate the standardized transportation energy consumption coefficient per unit of goods based on transportation mileage and cargo weight. Long-term operational data from the usage and operation stages are filtered using a moving average to eliminate the impact of short-term fluctuations on long-term energy consumption assessment, generating a stable operational energy consumption baseline. Finally, the processed data from each stage are converted into carbon dioxide equivalents item by item using a common carbon emission factor database, and the carbon emission accounting factors for each stage are summarized.

[0074] Using the converted carbon emission accounting factors, a life cycle assessment model is invoked to calculate the theoretical carbon emission inventory of environmentally friendly cables at each stage of their entire life cycle. The process includes: constructing a product system boundary model for the environmentally friendly cable, which clearly defines the physical and responsibility boundaries from raw material entry to product disposal and recycling. Within the product system boundary model, an input-output material and energy flow inventory is established for each stage, detailing all input materials and energy, and all output products, by-products, waste, and emissions. The carbon emission accounting factors for each stage are allocated to the corresponding input or output items in the material and energy flow inventory according to their material or energy type. Using an attribution-based life cycle assessment method, carbon emissions from upstream stages are cumulatively added to downstream stages to calculate the cumulative carbon emissions carried by the finished environmentally friendly cable. The cumulative carbon emissions from all stages from raw material extraction to disposal and recycling are summarized to form a theoretical carbon emission inventory broken down by stage. This theoretical carbon emission inventory lists the carbon emission contribution value of each stage in a structured tabular format.

[0075] In practical implementation, the data set of the entire life cycle of environmentally friendly cables undergoes data cleaning and normalization. Data from different sources and dimensions are converted into a unified carbon emission accounting factor. The processing covers all stages of raw material mining, production and manufacturing, logistics and transportation, use and operation, and scrapping and recycling. For energy consumption and material flow data in the raw material mining stage, energy consumption values ​​exceeding the preset quantile range are identified and removed based on the box plot method. The preset quantile range is set to values ​​outside 1.5 times the interquartile range between the first and third quartiles, eliminating distortion points in the original data caused by recording errors or extreme events. For multi-source heterogeneous data in the production and manufacturing stage, energy consumption readings reported by different equipment interfaces are uniformly converted into continuous energy consumption sequences under standard timestamps and standard power units. The standard timestamps adopt the Coordinated Universal Time format, and the standard power unit adopts kilowatt-hours, solving the problems of inconsistent data collection frequencies and mixed units among different equipment. For fuel consumption data in the logistics and transportation stage, a distance-load weighted algorithm is applied to calculate the standardized transportation energy consumption coefficient per unit of goods based on transportation mileage and cargo weight. The specific expression of the distance-load weighted algorithm is as follows:

[0076]

[0077] in: Indicates the standardized transportation energy consumption coefficient. Indicates fuel consumption. Indicates the distance traveled. This coefficient, representing cargo weight, normalizes the actual energy consumption of different transportation tasks to a baseline value based on mileage and load, making data across batches or routes comparable. For long-term operational data from various stages, a moving average filter is applied to eliminate the impact of short-term fluctuations on long-term energy consumption assessment. The sliding window size is set according to the data sampling interval; for example, averaging hourly energy consumption data over a 24-hour window generates a stable operational energy consumption baseline. The processed data from each stage are then converted into carbon dioxide equivalents based on a common carbon emission factor database. Regional grid emission factors are applied to electricity consumption, fuel type emission factors to fuel consumption, and material production emission factors to material consumption. These conversions are then summarized to obtain the carbon emission accounting factors for each stage.

[0078] In some embodiments, the converted carbon emission accounting factors are used to call the life cycle assessment model to calculate the theoretical carbon emission inventory of environmentally friendly cables at each stage of their entire life cycle, constructing a product system boundary model for environmentally friendly cables. This product system boundary model clearly defines the physical and responsibility boundaries from raw material entry to product scrapping and recycling. The physical boundaries include all facilities and processes related to copper mining, insulation material production, cable stranding manufacturing, finished product transportation, cable laying and operation, and waste cable recycling. The responsibility boundaries cover both direct and indirect carbon emissions. Within the product system boundary model, an input-output material and energy flow inventory is established for each stage. This inventory records in detail all materials and energy inputs and all products, by-products, waste, and emissions output. For example, in the cable manufacturing stage, inputs include copper rods, insulating plastic granules, electricity, and cooling water; outputs include finished cables, waste scraps, and exhaust emissions. The carbon emission accounting factors for each stage are then assigned according to their associated materials or energy sources. The carbon emission accounting factor is assigned to the corresponding input or output item in the material and energy flow inventory. For example, the carbon emission accounting factor corresponding to electricity consumption is assigned to the energy input item in the production and manufacturing process, and the carbon emission accounting factor corresponding to copper consumption is assigned to the material output item in the raw material mining process. Using the attribution-based life cycle assessment method, the carbon emissions of the upstream process are accumulated to the downstream process step by step to calculate the cumulative carbon emissions carried by the finished environmentally friendly cable. The attribution-based life cycle assessment method follows the principle of causal allocation. Starting from the raw material mining, the carbon emissions of each process are fully included and passed to the downstream through the product flow until the final product. The cumulative carbon emissions of all processes from raw material mining to scrapping and recycling are summarized to form a theoretical carbon emission inventory broken down by process. The theoretical carbon emission inventory lists the carbon emission contribution value of each process in the form of a structured table. The table rows include raw material mining, production and manufacturing, logistics and transportation, use and operation, and scrapping and recycling. The table header includes process name, carbon emission amount, and carbon emission intensity per unit product.

[0079] Optionally, during the data cleaning process, the preset quantile range of the box plot method can be dynamically adjusted according to the historical data distribution. For example, the quantile threshold of each stage of data can be determined through statistical learning. The window size of the moving average filter can be selected based on data stability testing to balance the effects of noise filtering and trend preservation. It can be understood that the transportation mileage and cargo weight data in the distance-load weighted algorithm come from the logistics tracking system and warehouse management system to ensure the accuracy of the calculation input.

[0080] In practice, the life cycle assessment model relies on standardized software tools, such as using open-source life cycle assessment software to import material and energy flow inventories and emission factors to achieve automated calculations. In some embodiments, the material and energy flow inventories are established in a modular manner, with each link as an independent module, which facilitates data updates and expansion. Optionally, the attribution-based life cycle assessment method uses allocation coefficients to handle by-products when accumulating carbon emissions. For example, in cable production, the waste recycling portion allocates upstream carbon emissions according to the mass ratio. It can be understood that the structured table output of the theoretical carbon emission inventory is in electronic file format, supporting data interface docking with the production execution system.

[0081] In one embodiment of the present invention, the improved gradient boosting tree algorithm is optimized based on the process path dependency of environmentally friendly cable production. Its working principle includes: (See reference) Figure 3 A training dataset was constructed with carbon emission differences as the prediction target and multiple production state parameters as feature vectors. In the basic iterative framework for building the gradient boosting tree, process constraints were introduced as regularization terms, defined by the allowable fluctuation range of parameters between adjacent processes in the process path diagram. In each iteration generating a new decision tree, not only was the reduction of the prediction residual considered, but also whether the decision path of the new tree violated the process constraints, and penalty weights were applied to split nodes that violated the constraints. Time series correlation was incorporated into the gradient calculation; for samples from the same production batch with a temporal order, their residual gradients were weighted according to their process sequence to strengthen the dependencies on the process path. Training stopped when the total prediction error of the ensemble tree model no longer decreased or when the preset maximum number of iterations was reached. The final ensemble model can automatically identify and follow the inherent path dependency logic of the production process when predicting carbon emission differences.

[0082] The process constraints are defined by the allowable fluctuation range of parameters between adjacent processes in the process flow diagram. The construction process includes: analyzing the production process flow diagram of environmentally friendly cables and extracting all sequentially connected process pairs and parallel process groups. For each sequentially connected process pair, statistical relationships are obtained from historical normal production data regarding the influence of key control parameters of the subsequent process on corresponding parameters of the preceding process. The allowable fluctuation range is defined based on the confidence interval of this statistical relationship. For parallel process groups, the quality or energy consumption balance equation at their output convergence point is defined. The coordination constraint range between parallel processes is defined based on the allowable deviation of the balance equation during normal production. The allowable fluctuation range of each sequential process pair and the coordination constraint range of each parallel process group are encoded into mathematical constraints in the form of inequalities. The set of all mathematical constraints in the form of inequalities is used as the process constraints and integrated into the loss function of the gradient boosting tree algorithm.

[0083] In its implementation, the improved gradient boosting tree algorithm is optimized based on the process path dependencies in the production of environmentally friendly cables. A training dataset is constructed with carbon emission differences as the prediction target and multiple production state parameters as feature vectors. The training dataset is derived from historical production batches, and each sample contains the actual values ​​of process parameters from the production execution system and the carbon emission differences calculated by comparing them with the corresponding batch. In the basic iterative framework for constructing the gradient boosting tree, process constraints are introduced as regularization terms. These constraints are defined by the allowable fluctuation range of parameters between adjacent processes in the process path diagram. The regularization term is added to the algorithm's loss function, ensuring that model training not only pursues prediction accuracy but also forces the model to learn parameter relationships that conform to the actual physical laws of production. In each iteration, when generating a new decision tree, not only is the reduction of prediction residuals considered, but also whether the decision path of the new tree violates the process constraints. Splitting nodes that violate the constraints are penalized. The penalty weight is proportional to the degree of constraint violation. The degree of violation is measured by the difference between the split point parameter value and the boundary value of the allowable fluctuation range of the process constraint. Time series correlation is incorporated into the gradient calculation. For samples from the same production batch with a time sequence, their residual gradients are weighted according to their process sequence to strengthen the dependency relationship on the process path. Specifically, the weighting method is to give greater weight to the gradient of the sample of the preceding process, so that the model pays more attention to the cumulative impact of the parameters of the earlier process in the process flow on the final carbon emission difference. Training stops when the total prediction error of the ensemble tree model no longer decreases or reaches the preset maximum number of iterations. The final ensemble model can automatically identify and follow the inherent path dependency logic of the production process when predicting carbon emission differences. The model outputs not only the predicted value of carbon emission differences, but also the feature importance ranking, which provides a basis for subsequent identification of key parameters.

[0084] In some embodiments, process constraints are defined by the allowable fluctuation range of parameters in adjacent processes in the process flow diagram. The construction process includes analyzing the production process flow diagram of the environmentally friendly cable, extracting all serially connected process pairs and parallel process groups. For example, identifying the serial process sequence of "conductor preheating-extrusion-cooling-cabling-sheathing" and the parallel process group of "parallel stranding of multiple insulated cores" from the process flow diagram. For each serially connected process pair, statistical relationships are calculated from historical normal production data to determine the influence of key control parameters of the subsequent process on the corresponding parameters of the preceding process. The allowable fluctuation range is defined based on the confidence interval of the statistical relationship. For example, there is a linear regression relationship between "extrusion process temperature" and the "conductor temperature after preheating" of the preceding process. The regression model calculates the 95% confidence interval of the predicted value. This interval is defined as the allowable fluctuation range of the "extrusion process temperature" relative to the "preheated conductor temperature". For parallel process groups, the quality or energy consumption balance equation of their output convergence point is defined. Based on the allowable deviation of the balance equation during normal production, the coordination constraint range between parallel processes is defined. For example, in the stranding process, the sum of the tensions of multiple cores needs to be balanced with the take-up tension. Based on historical data, the standard deviation of the difference between the left and right ends of this tension balance equation is determined. A positive or negative three times the standard deviation is used as the coordination constraint range between the tensions of parallel cores. The allowable fluctuation range of each serial process pair and the coordination constraint range of each parallel process group are encoded into mathematical constraints in the form of inequalities. For example, a serial constraint can be expressed as:

[0085]

[0086] in: These are parameters for process i. These are parameters for process j. and It is the regression coefficient. It is the allowable volatility margin. The set of all mathematical constraints in the form of inequalities is used as process constraints and integrated into the loss function of the gradient boosting tree algorithm. The integration method is to add a penalty term to the original loss function. The penalty term is proportional to the sum of the degrees of all constraint violations.

[0087] Optionally, process constraints can be dynamically updated based on production equipment upgrades or process improvements. Statistical relationships and confidence intervals can be recalculated periodically using new normal production data to adjust the allowable fluctuation range. In the improved gradient boosting tree algorithm, the gradient weighting coefficients can be adjusted based on the tightness of coupling between processes, calculated using the mutual information entropy between process parameters. It can be understood that the specific implementation of applying penalty weights to constraint-violation split nodes involves subtracting a penalty value proportional to the degree of constraint violation from the information gain of candidate split points during the gain calculation of decision tree node splits, thereby suppressing the generation of decision paths that violate process rules.

[0088] In practical implementation, when constructing the training dataset, the production state parameter feature vector includes multi-dimensional data from raw material parameters, equipment operating parameters, and environmental parameters, such as copper rod resistivity, temperature of each section of the extruder, cooling water flow rate, and temperature and humidity of the production environment. The carbon emission difference value is the difference between the theoretical value calculated from the theoretical carbon emission inventory and the actual value calculated through actual energy and material consumption data. The basic iterative framework of the improved gradient boosting tree algorithm adopts the extreme gradient boosting algorithm, and its loss function, after adding the process constraint regularization term, takes the following form:

[0089]

[0090] in: It is the overall loss. It is the loss due to prediction error. It is a regularization term that controls the complexity of the model. It is the process constraint penalty coefficient. It is the sum of all measures that violate process constraints. In some embodiments, the incorporation of time series correlation is achieved by adding process timestamp attributes to the samples in the training dataset. Samples from the same production batch are arranged in the order of their processes. When calculating the gradient, the gradient of the sample with the earlier process order is multiplied by a weight coefficient greater than 1, and the gradient weight of the sample with the later process order is 1, thus reflecting that the influence of the preceding process has a greater weight. Optionally, the model stops training when the total prediction error no longer decreases by more than a minimum threshold in ten consecutive iterations. It can be understood that when the final ensemble model is applied, for new production state parameter inputs, the model can output its carbon emission difference prediction value and give a decomposition of the contribution of each input feature to the prediction result.

[0091] In one embodiment of the present invention, an improved gradient boosting tree algorithm is used to identify key process parameters and abnormal production events that cause carbon emission differences. The process includes: inputting the carbon emission difference dataset into a trained improved gradient boosting tree algorithm model to obtain the predicted carbon emission difference value and prediction contribution decomposition for each sample. Based on the prediction contribution decomposition, the Shapley value of each input production state parameter feature on the final prediction result is calculated. The Shapley values ​​of all features are sorted, and the feature set whose sum of Shapley values ​​exceeds a preset threshold is selected as the key process parameters. During the model prediction process, the production period and equipment status corresponding to the samples that cause a sudden increase in prediction error are recorded. Combining equipment logs and production event records, the production period and equipment status are analyzed to identify specific equipment failures, raw material batch switching, or manual intervention operations, and the identified specific events are marked as abnormal production events.

[0092] In practical implementation, an improved gradient boosting tree algorithm is used to identify key process parameters and abnormal production events leading to carbon emission differences. The carbon emission difference dataset is input into the trained improved gradient boosting tree algorithm model to obtain the predicted carbon emission difference value and prediction contribution decomposition for each sample. Each sample in the carbon emission difference dataset corresponds to a set of production state parameters for a specific production batch within a specific time period. The improved gradient boosting tree algorithm model outputs the predicted carbon emission difference value for that batch. Simultaneously, the model internally calculates the contribution decomposition of each input feature to the predicted value. Based on the prediction contribution decomposition, the Shapley value of each input production state parameter feature is calculated for the final prediction result. The Shapley value calculation is based on cooperative game theory, quantifying the average marginal contribution of each feature to the model's predicted output across all possible feature combinations. For a model containing N features, the Shapley value of feature i is... Calculated using the following formula:

[0093]

[0094] in: It is the set of all features. It is any subset of features that does not contain feature i. The predicted value is obtained by the model based on the feature subset S. The calculation process enumerates the changes in the model's predicted value when feature i is added to different subsets S, and performs a weighted average on all possible combinations. The Shapley values ​​of all features are sorted, and the feature set whose sum of Shapley values ​​exceeds a preset threshold is selected as the key process parameter. The preset threshold is set as a fixed proportion of the sum of the Shapley values ​​of all features, such as 85%. This means that the cumulative contribution of the selected key process parameters accounts for the main part of the model's predicted output.

[0095] In some embodiments, during the model prediction process, the production period and equipment status corresponding to the samples that cause a sudden increase in prediction error are recorded. A sudden increase in prediction error is defined as the absolute error between the model's predicted value and the actual comparison value of carbon emission difference for a certain batch of samples exceeding three times the standard deviation of the historical average error. The system automatically records the precise production timestamp, production line number, and all online monitored equipment status codes corresponding to these abnormal samples. Combining equipment logs and production event records, the system analyzes the production period and equipment status to identify specific equipment failures, raw material batch switching, or manual intervention operations. The identified specific events are marked as abnormal production events. The equipment logs come from the equipment maintenance system and record equipment alarm codes, maintenance records, and start-up and shutdown events. The production event records come from the production execution system and record operations such as material change, machine adjustment, and unplanned shutdown. Through timestamp alignment, the abnormal production periods identified by the model are matched with specific event entries in the logs, such as matching "extruder main motor current over-limit alarm" or "insulation material batch switching record," thereby transforming the abstract prediction error into a specific and interpretable production event.

[0096] Optionally, the Shapley value can be calculated using approximate algorithms to reduce computational complexity. For example, Monte Carlo sampling can be used to estimate the average marginal contribution of features. Preset thresholds can be dynamically configured based on different product models or production stages. For products with high added value or high emissions, higher cumulative contribution thresholds can be set to identify more refined key parameters. It is understood that the identification of samples with sudden increases in prediction error is continuous. Whenever a new batch of production data is input into the model and a prediction is generated, the system calculates its prediction error in real time and compares it with the historical error distribution to achieve dynamic monitoring.

[0097] In practical implementation, taking a production batch of an insulation extrusion process as an example, the carbon emission difference dataset input to the model contains 96 samples collected at 5-minute intervals within an 8-hour production cycle for that batch. Each sample contains five production state parameter features: "preheating temperature", "extrusion pressure", "screw speed", "cooling water flow rate" and "ambient humidity". The model outputs a carbon emission difference prediction value for each sample and calculates the local contribution of each feature on that sample. The local contributions of the 96 samples are averaged to obtain the global Shapley value of each feature at the batch level.

[0098] Table 1: Characteristic Shapley Value Calculation and Key Parameter Identification Table

[0099]

[0100] Referring to Table 1, the cumulative Shapley value ratio of four characteristics—"extrusion pressure," "preheating temperature," "screw speed," and "cooling water flow rate"—reached 94.4%, exceeding the preset threshold of 85%. Therefore, these were identified as key process parameters causing the carbon emission differences in this batch. Simultaneously, the system detected a sudden increase in prediction error for sample number 47 (corresponding to a specific time point). By querying the equipment log at that moment, a maintenance record for "instantaneous drift of the extruder die pressure sensor" was found. The system associated this event and marked it as an abnormal production event of "instantaneous sensor drift." In some embodiments, the correlation analysis not only relies on precise timestamp matching but also allows event correlation within a preset time tolerance window to address situations of inter-system clock asynchrony or event recording delays. Optionally, a confidence score is attached when marking abnormal production events. The confidence score is calculated based on the overlap between the event log and the abnormal time period and the severity of the log type.

[0101] In one embodiment of the present invention, a set of carbon emission optimization strategies for a specific production batch or a specific stage is generated based on identified key process parameters and abnormal production events. The process includes: for each identified key process parameter, querying its historical optimal operating range, which is the common value range of the key process parameter in historical low-carbon emission batches; comparing the current real-time value of the key process parameter with the historical optimal operating range; if the current real-time value deviates from the historical optimal operating range, generating parameter optimization suggestions to adjust the key process parameter to within the historical optimal operating range; for each identified abnormal production event, retrieving standardized handling plans for the corresponding event type from a knowledge base; classifying and prioritizing the parameter optimization suggestions and standardized handling plans according to their corresponding production stage, scope of impact, and expected carbon emission reduction; and packaging all classified and prioritized suggestions and plans to generate a set of carbon emission optimization strategies for a specific production batch or a specific stage, which includes specific operating steps, target values, and execution timing.

[0102] The process of binding a set of carbon emission optimization strategies with the digital identifiers of cable products to form a traceable low-carbon product electronic passport includes: generating a unique digital identifier for each reel or batch of environmentally friendly cable products, which encodes the product model, production batch, production line, and production time information; establishing a product database to link and store the digital identifiers, the calculated theoretical carbon emission inventory, identified key process parameters and abnormal production events, and the generated set of carbon emission optimization strategies; generating a structured electronic document, the low-carbon product electronic passport, based on the linked data, which includes at least: the product's full life-cycle carbon emission footprint, analysis of major carbon emission stages, and a summary of the implementation status of carbon emission optimization strategies during the production process; generating a digital digest from the content of the low-carbon product electronic passport through hashing, and signing the digital digest using a digital certificate to ensure the authenticity and immutability of the passport content; and associating the storage address or access index of the signed low-carbon product electronic passport with the physical tag or digital identifier of the product entity for downstream users or auditors to query and verify.

[0103] In practical implementation, based on the identified key process parameters and abnormal production events, a set of carbon emission optimization strategies for specific production batches or specific processes is generated. For each identified key process parameter, its historical optimal operating range is queried. The historical optimal operating range is the common value range of the key process parameter in historical low-carbon emission batches. Historical low-carbon emission batches are obtained by screening historical production batches whose carbon emission intensity is lower than the overall average level by a certain percentage. The numerical distribution of the key process parameter in these batches is determined by statistical methods, for example, taking the range between the 10th percentile and the 90th percentile as the historical optimal operating range. The current real-time value of the key process parameter is compared with the historical optimal operating range. If the current real-time value deviates from the historical optimal operating range, parameter optimization suggestions are generated to adjust the key process parameter to within the historical optimal operating range. The parameter optimization suggestions include specific... The system identifies the parameter name, current value, target value range, and adjustment direction. For each identified abnormal production event, a standardized handling plan for the corresponding event type is retrieved from the knowledge base. The knowledge base is a predefined event-plan mapping library. For example, when the abnormal production event is "instantaneous drift of the extruder die head pressure sensor," the retrieved standardized handling plan might include steps such as "immediately check the sensor wiring," "activate the backup sensor channel," and "manually calibrate the reading in the subsequent production cycle." The parameter optimization suggestions and standardized handling plans are categorized and prioritized according to their corresponding production stage, scope of impact, and expected carbon emission reduction. The categorization is based on the production stage coding system, the scope of impact is divided into equipment level, process level, and production line level, and the expected carbon emission reduction is estimated based on the average emission reduction effect of similar historical adjustment cases. Priority ranking is achieved by calculating a priority score. To determine, among which It is a priority score. It is a numerical value that quantifies the scope and level of influence. It is the normalized value of the expected carbon emission reduction. It is the normalized value of the estimated execution cost. , and These are preset weighting coefficients; the higher the score, the higher the priority. All suggestions and plans after being categorized and sorted are packaged into a set of carbon emission optimization strategies for a specific production batch or a specific process. The set of carbon emission optimization strategies is presented in the form of a structured list, which includes specific operation steps, target values, and execution timing. The execution timing is specified according to the strategy type, such as "execute immediately", "execute when the next planned shutdown", or "execute before the next batch of production", see Table 2.

[0104] Table 2: Optimization Strategies for Key Parameters in Insulation Process

[0105]

[0106] In some embodiments, a set of carbon emission optimization strategies is bound to the digital identifier of cable products to form a traceable low-carbon product electronic passport. A unique digital identifier is generated for each reel or batch of environmentally friendly cable products. This digital identifier encodes the product model, production batch, production line, and production time information. The generation algorithm uses a cryptographic hash function to calculate the basic product information and timestamp, converting it into a scannable QR code and a readable string format. A product database is established, storing the digital identifier, the calculated theoretical carbon emission inventory, identified key process parameters and abnormal production events, and the generated set of carbon emission optimization strategies in association. The database table structure uses the digital identifier as the primary key, and stores indexes or specific content of the above-mentioned data in association. Based on the associated stored data, a structured electronic document is generated; this electronic document is the low-carbon product electronic passport, and its content includes at least: carbon emissions throughout the product's entire life cycle. The document includes a carbon footprint analysis, a summary of the implementation of carbon emission optimization strategies during the production process, and a chart showing the carbon emission contribution of each stage. The summary records the implemented strategies, their execution time, and the status of the results. The content of the low-carbon product e-passport is hashed to generate a digital digest, which is then signed with a digital certificate to ensure the authenticity and immutability of the passport content. The hashing uses a secure hash algorithm, and the digital certificate is issued by a trusted certificate authority. The storage address or access index of the signed low-carbon product e-passport is associated with the physical label or digital identification code of the product entity for downstream users or auditors to query and verify. The physical label is printed on the cable reel certificate of conformity or outer packaging. Downstream users can access the passport document stored in the cloud or on the blockchain by scanning a QR code. Auditors can verify the integrity and authenticity of the document by verifying the digital signature.

[0107] Optionally, the query for historical optimal operating intervals can incorporate a sliding time window mechanism, using only production data from the most recent period for analysis. This ensures the interval reflects the latest status of the equipment and processes, and the estimation of expected carbon emission reductions can reference industry benchmarks or simulation results based on physical models. It is understood that the carbon emission optimization strategy set is packaged in a machine-readable markup language, such as JSON or XML, facilitating direct import into production execution systems or automated control systems to drive execution.

[0108] In practical implementation, priority ranking weight coefficient , and The setting is determined by the production management department based on a comprehensive consideration of emission reduction targets, cost control, and production stability requirements. For example, during periods of high emission reduction pressure, [the department may assign specific parameters]. Higher weighting, given during periods of strict cost control With higher weighting, the generation of electronic passports for low-carbon products is automated. Upon product inspection and warehousing, the system automatically assembles and signs all relevant data, then writes the passport's storage index into the product database. In some embodiments, for key parameter optimization strategies, the system can generate draft control instructions directly issued to the distributed control system. These instructions can be executed after operator confirmation. Optionally, the generation of the digital identification code can be combined with blockchain technology, storing the identification code and its hash value on the blockchain to enhance its uniqueness and time reliability. Understandably, when downstream users query the electronic passport, the system will display the original passport document, the digital signature verification status, and a traceability chain of all data within the passport.

[0109] In one embodiment of the present invention, the closed-loop verification step for the execution effect of carbon emission optimization strategies includes: continuously collecting new rounds of production data after the carbon emission optimization strategy set is applied to the production process; recalculating the actual carbon emission intensity based on the new round of production data and comparing it with the historical carbon emission intensity before the application of the strategy; if the carbon emission intensity decreases to the expected target, the optimization strategy and its execution parameters are confirmed as effective strategies and stored in the optimization strategy knowledge base; if the carbon emission intensity does not decrease or even increases, the reasons are analyzed to determine whether it is due to strategy execution deviation, changes in external conditions, or the strategy itself being unsuitable, and strategy adjustment suggestions or failure markers are generated. The results of the closed-loop verification are fed back to the training process of the improved gradient boosting tree algorithm to update the model parameters, forming a closed-loop management process from strategy generation, execution to verification and optimization.

[0110] In practice, the closed-loop verification step for the effectiveness of carbon emission optimization strategies is initiated after the set of carbon emission optimization strategies is applied to the production process. A new round of production data is continuously collected after the application. This new round of production data comes from production batches of the same or similar product models produced after the application of the optimization strategies on the same production line. The dimensions, frequency, and sources of data collection remain consistent with those before optimization to ensure data comparability. The data includes process parameters from the production execution system, real-time energy consumption readings from IoT sensors, material consumption records, and corresponding product output information. Based on the new round of production data, the actual carbon emission intensity is recalculated and compared with the historical carbon emission intensity before the application of the strategy. The calculation of the actual carbon emission intensity uses the same methodology and emission factors as the generation of the theoretical carbon emission inventory. For the optimized production batch, the carbon dioxide equivalent is calculated based on its actual energy consumption and material consumption data, and then divided by the output of qualified products in that batch to obtain the actual carbon emission intensity per unit product. The historical carbon emission intensity is selected as the benchmark based on the average carbon emission intensity of similar products within a certain period before the application of the optimization strategy, such as the past three months.

[0111] In some embodiments, if the carbon emission intensity reduction reaches the expected target, the optimization strategy and its execution parameters are confirmed as effective strategies and stored in the optimization strategy knowledge base. The expected target is set before applying the strategy and can be a specific percentage reduction value, such as a reduction of no less than 5% in carbon emission intensity per unit product, or an absolute reduction. The optimization strategy knowledge base is a structured storage system used to archive verified effective optimization strategies. Each record contains strategy content, applicable product models and process conditions, execution time, measured carbon emission reduction effect, and related production context data. If the carbon emission intensity does not decrease or even increases, the cause is analyzed to determine whether it is due to strategy execution deviation, changes in external conditions, or the strategy itself being inapplicable. Strategy adjustment suggestions or failure flags are generated. The process of analyzing the cause involves reviewing the execution log, checking the production environment parameters, and re-evaluating the optimization strategy logic. For example, checking whether the parameter adjustment settings were accurately executed, or whether unforeseen changes occurred in the raw material properties during the production process. Strategy adjustment suggestions may include correcting the target parameter range, adjusting the strategy execution order, or adding new constraints. Failure flags are associated with the corresponding optimization strategy and accompanied by a cause analysis report to prevent future inappropriate calls.

[0112] In practical implementation, the results of closed-loop verification are fed back into the training process of the improved gradient boosting tree algorithm to update model parameters, forming a closed-loop management process from policy generation, execution to verification and optimization. Feedback is achieved by adding new production data samples containing the effects of the optimization policy application to the model's training dataset. The feature vectors of the new samples contain production state parameters after the policy application, and the label value is the difference between the actual and theoretical carbon emission intensity of that batch. When a sufficient number of new samples are accumulated, incremental learning or full retraining of the improved gradient boosting tree algorithm is triggered, enabling the model to learn the updated correlation between the new features of the production process after the new policy application and the carbon emission differences. This allows for more accurate evaluation and recommendation of optimization measures in future policy generation. After introducing new training data, the penalty weights for violating process constraints in the loss function of the improved gradient boosting tree algorithm may be adaptively adjusted by reassessing the distribution differences between historical and latest data. The calculation can be based on the following formula:

[0113]

[0114] in: The KL divergence represents the distribution of features between the old and new training data, and is used to measure the degree of change in the data distribution. It is a smoothing hyperparameter used to prevent the denominator from being zero. It is the basic process constraint penalty coefficient. Through this adjustment, the model can dynamically adjust the degree of emphasis on process constraints according to changes in the distribution of production process data.

[0115] Optionally, the expected target can be set based on statistical process control methods. Control limits are calculated using historical carbon emission intensity time-series data, and the optimized intensity value falling below the lower control limit is defined as achieving the expected target. It can be understood that effective strategies stored in the optimization strategy knowledge base can be preferentially recommended or automatically applied in subsequent similar production scenarios, forming knowledge accumulation and reuse. In some embodiments, the analysis of the causes of strategy failure will initiate an independent incident investigation process. This process may involve reviewing the original sensor data, conducting in-depth inspections of equipment status, and interviewing operators to ensure the comprehensiveness of the cause determination. Optionally, the retraining timing of the improved gradient boosting tree algorithm can be set to periodic triggering, such as once a month, or it can be set to trigger after accumulating a specific number of new samples. It can be understood that the continuous operation of the closed-loop management process enables the carbon emission management method to have self-evolution capabilities, adapting to the slow drift of production processes, equipment aging, and changes in raw material sources.

[0116] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for carbon emission management of environmentally friendly cables, characterized in that, The method includes: Obtain a set of stage data for the entire life cycle of environmentally friendly cables, including energy consumption and material flow data for each stage of raw material mining, production and manufacturing, logistics and transportation, use and operation, and scrapping and recycling. The data set of the aforementioned stages is cleaned and normalized to convert data from different sources and dimensions into a unified carbon emission accounting factor. Using the converted carbon emission accounting factor, the life cycle assessment model is invoked to calculate the theoretical carbon emission inventory of the environmentally friendly cable at each stage of its entire life cycle; The theoretical carbon emission inventory is compared with actual operational data from the production execution system and IoT sensors to obtain a carbon emission difference dataset; The carbon emission difference dataset is input into the improved gradient boosting tree algorithm, which is optimized based on the process path dependency of environmentally friendly cable production. The improved gradient boosting tree algorithm is used to identify key process parameters and abnormal production events that cause differences in carbon emissions. Based on the identified key process parameters and abnormal production events, a set of carbon emission optimization strategies is generated for specific production batches or specific processes. The aforementioned set of carbon emission optimization strategies is linked to the digital identification of cable products to form a traceable electronic passport for low-carbon products. The improved gradient boosting tree algorithm is optimized based on the process path dependency of environmentally friendly cable production, and its working principle includes: Construct a training dataset with carbon emission differences as the prediction target and multiple production state parameters as feature vectors; In the basic iterative framework for constructing gradient boosting trees, process constraints are introduced as regularization terms. These process constraints are defined by the allowable fluctuation range of parameters of adjacent processes in the process path diagram. When generating a new decision tree in each iteration, not only is the reduction of the prediction residual considered, but also whether the decision path of the new tree violates the process constraints, and penalty weights are applied to the split nodes that violate the constraints. Time series correlation is incorporated into gradient calculation. For samples from the same production batch with a time sequence, their residual gradients are weighted according to the sequence of their processes to strengthen the dependencies on the process path. Training stops when the total prediction error of the ensemble tree model no longer decreases or reaches the preset maximum number of iterations. The resulting ensemble model can automatically identify and follow the inherent path dependency logic of the production process when predicting carbon emission differences. The process constraints are defined by the allowable fluctuation range of parameters between adjacent processes in the process path diagram, including: Analyze the production process flow chart of environmentally friendly cables and extract all serially connected process pairs and parallel process groups; For each serially connected process pair, statistical relationships are obtained from historical normal production data to determine the influence of the key control parameters of the subsequent process on the corresponding parameters of the preceding process, and the allowable fluctuation range is defined based on the confidence interval of the statistical relationships. For parallel process groups, define the quality or energy consumption balance equation at their output convergence point, and define the coordination constraint range between parallel processes based on the allowable deviation of the balance equation during normal production. The allowable fluctuation range of each serial operation pair and the coordination constraint range of each parallel operation group are encoded into mathematical constraints in the form of inequalities. The set of all mathematical constraints in the form of inequalities is used as the process constraints and integrated into the loss function of the gradient boosting tree algorithm.

2. The carbon emission management method for environmentally friendly cables according to claim 1, characterized in that, The aforementioned data set undergoes data cleaning and normalization processing to convert data from different sources and dimensions into a unified carbon emission accounting factor, including: Outlier removal is performed on energy consumption and material flow data in the raw material mining process. Based on the box plot method, energy consumption values ​​that exceed the preset quantile range are identified and removed. Standardize the format of multi-source heterogeneous data in the production and manufacturing process, and convert the energy consumption readings reported by different equipment interfaces into a continuous energy consumption sequence under standard timestamps and standard power units. A distance-load weighted algorithm is applied to fuel consumption data in the logistics and transportation process to calculate the standardized transportation energy consumption coefficient per unit of goods based on transportation mileage and cargo weight. Long-term operating data from the operational phase are filtered using a moving average to eliminate the impact of short-term fluctuations on long-term energy consumption assessment and generate a stable operating energy consumption baseline. The processed data from each stage are converted into carbon dioxide equivalents based on an internationally recognized carbon emission factor database, and the carbon emission accounting factors for each stage are then summarized.

3. The carbon emission management method for environmentally friendly cables according to claim 1, characterized in that, The process of using the converted carbon emission accounting factor and calling the life cycle assessment model to calculate the theoretical carbon emission inventory of the environmentally friendly cable at each stage of its entire life cycle includes: Construct a product system boundary model for environmentally friendly cables, which clearly defines the physical and responsibility boundaries from raw material entry to product scrapping and recycling; Within the product system boundary model, an input-output material and energy flow list is established for each stage. The material and energy flow list records in detail all materials and energy input to the stage, as well as all products, by-products, waste, and emissions output. The carbon emission accounting factors of each stage are allocated to the corresponding input or output items in the material and energy flow list according to the material or energy type to which they belong. Using the attribution-based life cycle assessment method, carbon emissions from upstream stages are cumulatively added to downstream stages to calculate the cumulative carbon emissions carried by the finished environmentally friendly cable. The cumulative carbon emissions from all stages, from raw material mining to end-of-life recycling, are aggregated to form the theoretical carbon emission inventory broken down by stage. The theoretical carbon emission inventory lists the carbon emission contribution value of each stage in a structured table format.

4. The carbon emission management method for environmentally friendly cables according to claim 1, characterized in that, Using the improved gradient boosting tree algorithm, key process parameters and abnormal production events that cause differences in carbon emissions are identified, including: The carbon emission difference dataset is input into the trained improved gradient boosting tree algorithm model to obtain the carbon emission difference prediction value and prediction contribution decomposition for each sample. Based on the predicted contribution decomposition, the Shapley value of each input production state parameter feature on the final prediction result is calculated. The Shapley values ​​of all features are sorted, and the set of features whose sum of Shapley values ​​exceeds a preset threshold is selected as the key process parameter. During the model prediction process, the production period and equipment status corresponding to the samples that cause a sudden increase in prediction error are recorded; By combining equipment logs and production event records, the production period and equipment status are analyzed to identify specific equipment failures, raw material batch switching, or manual intervention operations. The identified specific events are then marked as abnormal production events.

5. The carbon emission management method for environmentally friendly cables according to claim 1, characterized in that, Based on the identified key process parameters and abnormal production events, a set of carbon emission optimization strategies is generated for specific production batches or specific stages, including: For each identified key process parameter, its historical optimal operating range is queried. The historical optimal operating range is the common value range of the key process parameter in historical low-carbon emission batches. The current real-time value of the key process parameter is compared with the historical optimal operating range. If the current real-time value deviates from the historical optimal operating range, parameter optimization suggestions are generated to adjust the key process parameter to within the historical optimal operating range. For each identified abnormal production event, retrieve the standardized contingency plan for the corresponding event type from the knowledge base; The parameter optimization suggestions and the standardized disposal plan are categorized and prioritized according to their corresponding production links, scope of impact and expected carbon emission reductions. All suggestions and plans, after being categorized and sorted, are packaged into a set of carbon emission optimization strategies for a specific production batch or a specific process. The set of carbon emission optimization strategies includes specific operational steps, target values, and execution timing.

6. The carbon emission management method for environmentally friendly cables according to claim 5, characterized in that, The aforementioned set of carbon emission optimization strategies is linked to the digital identifiers of cable products to form a traceable electronic passport for low-carbon products, including: A unique digital identification code is generated for each reel or batch of environmentally friendly cable products. The digital identification code encodes the product model, production batch, production line and production time information. Establish a product database to associate and store the digital identification code, the calculated theoretical carbon emission inventory, the identified key process parameters and abnormal production events, and the generated set of carbon emission optimization strategies. Based on the associated stored data, a structured electronic document is generated, which is the low-carbon product electronic passport. Its content includes at least: the carbon emission footprint of the product throughout its entire life cycle, analysis of the main carbon emission links, and a summary of the implementation of carbon emission optimization strategies in the production process. The content of the low-carbon product electronic passport is hashed to generate a digital digest, and the digital digest is signed with a digital certificate to ensure the authenticity and immutability of the passport content. The storage address or access index of the signed low-carbon product electronic passport is associated with the physical label or digital identification code of the product entity, so that downstream users or auditors can query and verify it.

7. The carbon emission management method for environmentally friendly cables according to claim 1, characterized in that, The method also includes a closed-loop verification step for the effectiveness of the carbon emission optimization strategy: After the set of carbon emission optimization strategies is applied to the production process, new rounds of production data are continuously collected after the application. Based on the new round of production data, the actual carbon emission intensity is recalculated and compared with the historical carbon emission intensity before the application of the strategy. If the reduction in carbon emission intensity reaches the expected target, the optimization strategy and its execution parameters will be confirmed as an effective strategy and stored in the optimization strategy knowledge base. If carbon emission intensity does not decrease or even increases, analyze the reasons to determine whether it is due to strategy execution deviation, changes in external conditions, or the strategy itself being unsuitable, and generate strategy adjustment suggestions or failure flags. The results of closed-loop verification are fed back into the training process of the improved gradient boosting tree algorithm to update model parameters, forming a closed-loop management process from policy generation, execution to verification and optimization.

8. A carbon emission management system for environmentally friendly cables, characterized in that, The system includes a processor and a memory, the memory storing a computer program, and the processor executing the computer program to implement a carbon emission management method for environmentally friendly cables according to any one of claims 1 to 7.

Images