Carbon footprint optimization method and system based on LCA-GLSS coupling

Abstract

The invention relates to the technical field of carbon footprint evaluation, and discloses a carbon footprint optimization method and system based on LCA-GLSS coupling. The carbon footprint optimization method comprises the following steps: calculating carbon footprints of all stages of a product in a full life cycle, carbon footprint contribution ratios corresponding to materials, energy sources, resources and wastes involved in all process units in the full life cycle, and carbon footprint sensitivities corresponding to the materials, the energy sources, the resources and the wastes; a high-carbon emission link found in the carbon footprint evaluation process is optimized, waste and variation in the carbon emission process are recognized and eliminated, a carbon footprint prediction model is established, the carbon emission trend in the future is predicted and simulated, and the prediction accuracy is improved through model optimization and parameter adjustment; the evaluation result is comprehensively analyzed and explained, main contribution factors and key links of the carbon footprint are identified, and emission reduction measures and optimization suggestions are formulated. The method has both quantitative analysis and process improvement, and can effectively improve the efficiency and accuracy of carbon footprint management.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of carbon footprint evaluation, and particularly relates to a carbon footprint optimization method and system based on LCA-GLSS coupling. BACKGROUND

[0002] Although some green technologies have been applied to petrochemical chemical production, these technologies have large investment, long cycle and high difficulty in implementation. At the same time, petrochemical chemical enterprises generally lack a systematic carbon footprint evaluation method, and it is difficult to accurately identify the key link of carbon emission, resulting in limited effect of emission reduction measures and lack of pertinence.

[0003] Therefore, it is urgent to provide a carbon footprint optimization scheme which can accurately identify the key link of carbon emission and provide process optimization in pertinence. SUMMARY

[0004] The purpose of the application is to provide a carbon footprint optimization method and system based on LCA-GLSS coupling. The LCA method is used to analyze the carbon emission of the whole life cycle of a product from raw material acquisition, production, use to waste treatment, AI technology is introduced to realize the systematic evaluation of the carbon emission of the whole life cycle of the product, the GLSS is combined to identify the high-carbon emission key node, and the closed-loop management of carbon emission identification-analysis-improvement-control is realized.

[0005] In order to achieve the above purpose, the application provides the following technical scheme: In a first aspect, the application provides a carbon footprint optimization method based on LCA-GLSS coupling, comprising the following steps: S1. Carbon footprint evaluation, specifically including: collecting carbon emission data of each stage of the whole life cycle of a product by using big data technology, and preprocessing, including data cleaning, screening and standardization; based on the preprocessed data, quantitatively analyzing the carbon emission of the whole life cycle of the product according to the whole life cycle evaluation method, calculating the carbon footprint of each stage, the carbon footprint contribution ratio of materials, energy, resources and waste involved in each process unit in the whole life cycle, and the carbon footprint sensitivity of materials, energy, resources and waste; S2. Carbon footprint optimization, specifically including: using the tools and methods of green lean six sigma to optimize the high-carbon emission link found in the carbon footprint evaluation process, identifying and eliminating waste and variation in the carbon emission process through cause-and-effect analysis and statistical process control means; using artificial intelligence algorithm, establishing a carbon footprint prediction model according to historical data, predicting and simulating the future carbon emission trend, and improving the accuracy of prediction through model optimization and parameter adjustment; comprehensively analyzing and explaining the evaluation results, identifying the main contribution factors and key links of carbon footprint, and formulating emission reduction measures and optimization suggestions.

[0006] As one possible approach, carbon footprint assessment includes the following steps: S10. Initialize the evaluation elements, specifically including: determining the evaluation object, evaluation objective, benchmark flow, measurement scale, system boundary, and selection criteria; S11. Data acquisition and data processing, specifically including: dividing the M sequentially executed processes in the process chain into N process units according to preset rules, where N≤M; acquiring and preprocessing the input and output data of each process unit; S12. Construct a lifecycle model, specifically including: using graph neural networks to construct a process chain network model to map the correlation paths between material flow, energy flow and carbon emissions; using GWP100 assessment categories to calculate the lifecycle carbon footprint; and using digital twin technology to simulate the process flow to update carbon footprint data in real time. S13. Calculate the carbon footprint contribution ratio of materials, energy, resources, and waste involved in each process unit throughout the life cycle; S14. Reduce the materials, energy, resources, and waste involved in each process unit according to a preset and identical ratio, obtain the carbon footprint of the materials, energy, resources, and waste involved in each process unit after the reduction, and calculate the carbon footprint sensitivity based on the carbon footprint before and after the reduction.

[0007] As one possible implementation method, data acquisition specifically includes: deploying a multimodal IoT sensor network to collect energy consumption parameters, environmental parameters, and device status parameters; configuring unstructured data including at least enterprise reports and industry reports; and configuring and deploying a data traceability blockchain with smart contracts to record energy consumption parameters, environmental parameters, device status parameters, and unstructured data. Data processing specifically includes data preprocessing and data fusion. Data preprocessing includes: using the Isolation Forest algorithm to identify drifting and breakpoint data, removing drifting data, and using a conditional generative adversarial network to fill in the data gaps and breakpoint data corresponding to drifting data. Data fusion includes: building a scalable data platform based on Hadoop / Spark to achieve real-time processing of multi-source heterogeneous data.

[0008] As one possible implementation, using graph neural networks to construct the process chain network model specifically includes: S120. Directed graph modeling, specifically including: treating each process in the process chain as a node, material flow and energy flow as directed edges, with the direction representing the flow path; the weight of the directed edge is configured as the carbon emission intensity corresponding to the flow amount; S121. Configure a dynamic update mechanism, specifically including: collecting material / energy flow data of each process in real time through IoT sensors, inputting it into a graph neural network to update edge weights; and using a graph attention network to aggregate node features and dynamically calculate path carbon emissions. S122. Correlation path output, specifically including: outputting the carbon emission hotspot paths of the entire process chain through the message passing mechanism of GNN, with a coverage of ≥98%.

[0009] As a possible implementation, the process chain network model is also equipped with a multivariate time-series carbon emission prediction unit based on an Attention-LSTM hybrid architecture; where LSTM is used to capture long-term dependencies in time-series data, and the Attention mechanism enhances attention to important time steps. The multivariate time-series carbon emission prediction unit embeds the SHAP value analysis module algorithm to identify the process that contributes the most to carbon emissions.

[0010] As one possible implementation, the calculation method for lifecycle carbon footprint using the GWP100 assessment categories is as follows: in, For carbon footprint; For the first Such activities; For the first Medium greenhouse gases; Within the system boundary, the first functional unit... Data on GHG emissions and elimination related to various activities, including primary and secondary data; For the first Greenhouse gases corresponding to these activities Emission coefficient; Greenhouse gases of value.

[0011] As one possible implementation, a dynamic emission factor library is configured to automatically correct emission factors based on real-time operating conditions; the correction criteria are dynamically adjusted based on the following multi-dimensional parameters: The device load rate is monitored in real time by IoT sensors. When the load rate is ≥80%, a high load correction factor is triggered. Energy structure, matching differentiated emission factors based on energy type and real-time carbon intensity data; The process parameters, combined with the reaction temperature and pressure process parameters, are used to calculate the dynamic compensation amount when the operating conditions deviate from the baseline value through a Bayesian network. Environmental data is used to compensate in real time for the impact of environmental parameters such as temperature, humidity, and atmospheric pressure on combustion efficiency or equipment energy efficiency. The dynamic emission factor database is updated synchronously through digital twin technology to ensure that the error rate between the corrected emission factors and actual carbon emissions is ≤3%.

[0012] As one possible approach, the carbon footprint evaluation results obtained in S1 can be embedded into the analysis and improvement phase of the green lean six sigma method described in S2, that is, the carbon footprint optimization scheme can be determined based on the carbon footprint contribution ratio and carbon footprint sensitivity.

[0013] As one possible implementation approach, the definition phase of the Green Lean Six Sigma methodology decomposes carbon emission targets into key quality characteristics of the production process using the Eco-QFD method, establishing a "carbon emission-process parameter" correlation matrix, specifically including: Based on the carbon footprint assessment results, key emission reduction processes are identified; process-level emission reduction targets are mapped to process parameters, and a multi-objective near-end strategy optimization algorithm is used to generate a carbon footprint optimization scheme, with the reward function being: Through iterative training using digital twin simulation, a set of Pareto optimal solutions is output. In the analysis phase of the Green Lean Six Sigma method, a Bayesian network model is constructed based on the Pareto solution set from the definition phase. The nodes include process parameters, carbon emissions, costs, and outputs. Prior probabilities are assigned through historical data statistics or expert experience. The Monte Carlo sampling method is used to quantify the sensitivity of process parameters, output sensitivity indices, and select the top three key parameters for the improvement phase.

[0014] Secondly, the present invention provides a carbon footprint optimization system based on LCA-GLSS coupling, comprising: a carbon footprint evaluation module and a process optimization module; The carbon footprint assessment module includes an initialization unit, a data collection and inventory analysis unit, a life cycle impact assessment unit, and a results interpretation unit. The initialization unit initializes the assessment elements, specifically defining the assessment object, assessment objective, baseline flow, measurement scale, system boundary, and selection criteria. The data collection and inventory analysis unit collects and processes data, specifically dividing the M sequentially executed processes in the chain into N process units (N≤M) according to preset rules; collecting and preprocessing the input and output data of each process unit; and constructing a life cycle model using the life cycle impact assessment unit. Specifically, this includes: constructing a process chain network model using graph neural networks to map the correlation paths between material flow, energy flow, and carbon emissions; calculating the life cycle carbon footprint using GWP100 assessment categories; simulating the process flow using digital twin technology to update carbon footprint data in real time; calculating the carbon footprint contribution ratio of materials, energy, and waste involved in each process unit during the life cycle using the result interpretation unit; reducing the materials, energy, and waste involved in each process unit by a preset and equal proportion to obtain the carbon footprint of materials, energy, and waste involved in each process unit after reduction; and calculating carbon footprint sensitivity based on the carbon footprint before and after reduction. The process optimization module includes a current performance level assessment unit, an inefficiency factor analysis unit, and an optimal solution determination unit. The current performance level assessment unit, based on the lifecycle carbon footprint, carbon footprint contribution ratio, and carbon footprint sensitivity output by the results interpretation unit, along with configured sustainability value assessment and social indicator assessment, uses a multi-objective reinforcement learning algorithm to determine the carbon footprint optimization scheme. This involves optimizing the balance between carbon emissions, costs, and output. During training, the carbon footprint optimization scheme is continuously adjusted to generate a set of Pareto optimal solutions, achieving the best balance between carbon emissions, costs, and output. The inefficiency factor analysis unit uses a Bayesian network with a prior probability greater than or equal to 0.8 to quantify the sensitivity of process parameters. The optimal solution determination unit then selects the best solution from the Pareto optimal solution set.

[0015] Compared with the prior art, the present invention has the following advantages: 1. The carbon footprint optimization method and system based on LCA-GLSS coupling proposed in this invention innovatively combines LCA and GLSS methods with big data AI technology to achieve a systematic assessment of carbon emissions throughout the product life cycle. It combines quantitative analysis and process improvement, and can effectively improve the efficiency and accuracy of carbon footprint management.

[0016] 2. The carbon footprint optimization method and system based on LCA-GLSS coupling proposed in this invention can adjust carbon emission paths according to real-time data in the constructed process chain network model, thereby achieving accurate carbon footprint tracking and optimization. It achieves dynamic mapping of material / energy flow paths (node ​​coverage greater than or equal to 98%).

[0017] 3. The carbon footprint optimization method and system based on LCA-GLSS coupling proposed in this invention introduces a dynamic emission factor library and automatically corrects the calculation parameters according to real-time operating conditions. This correction mechanism can ensure high accuracy and real-time performance of carbon emission calculations.

[0018] 4. The carbon footprint optimization method and system based on LCA-GLSS coupling proposed in this invention realizes the analysis of carbon emissions throughout the entire life cycle of a product, from raw material acquisition, production, use to waste disposal, through the LCA method, and identifies key nodes with high carbon emissions by combining GLSS, thereby realizing closed-loop management of carbon emissions through "identification-analysis-improvement-control".

[0019] 5. The carbon footprint optimization method and system based on LCA-GLSS coupling proposed in this invention are applicable to typical multi-process, long-cycle, and high-energy-consuming industries such as petrochemical manufacturing, and have good versatility and scalability. Attached Figure Description

[0020] The accompanying drawings, which are provided to further illustrate the invention and constitute a part of this invention, are illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention.

[0021] Figure 1 A flowchart of a carbon footprint optimization method based on LCA-GLSS coupling provided in an embodiment of the present invention; Figure 2 This is a detailed schematic diagram of the entire life cycle of a drug as described in this embodiment of the invention; Figure 3 This is a schematic diagram illustrating the construction of a supply chain data traceability chain based on blockchain in an embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the steps of constructing a lifecycle model in an embodiment of the present invention; Figure 5 This is a schematic diagram of the carbon footprint optimization system structure based on LCA-GLSS coupling in an embodiment of the present invention; Figure 6 This is a Sus-VSM diagram of mupirocin active pharmaceutical ingredient in an embodiment of the present invention; Figure 7 This is a schematic diagram showing the ten steps involved in the production process of mupirocin active pharmaceutical ingredient in this embodiment of the invention. Figure 8 This is a schematic diagram showing the division of the mupirocin active pharmaceutical ingredient process unit in an embodiment of the present invention; Figure 9 This is a schematic diagram of the production life cycle model of mupirocin active pharmaceutical ingredient in an embodiment of the present invention; Figure 10 This is a Pareto chart of GWP100 for producing 1 kg of mupirocin raw material in an embodiment of the present invention; Figure 11 The image shows a GWP100 Sankey diagram for producing 1 kg of mupirocin raw material in an embodiment of the present invention.

[0022] Figure label: 1-Carbon footprint assessment module, 10-Initialization unit, 11-Data collection and inventory analysis unit, 12-Life cycle impact assessment unit, 13-Result interpretation unit; 2-Process optimization module, 20-Current performance level assessment unit, 21-Inefficiency factor analysis unit, 22-Optimal solution determination unit. Detailed Implementation

[0023] To facilitate a clear description of the technical solutions in the embodiments of the present invention, the terms "first" and "second" are used to distinguish identical or similar items with essentially the same function and effect. For example, the first threshold and the second threshold are merely used to distinguish different thresholds and do not limit their order. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and that the terms "first" and "second" are not necessarily different.

[0024] It should be noted that in this invention, the terms "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in this invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0025] In this invention, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can represent: a, b, c, a combination of a and b, a combination of a and c, a combination of b and c, or a, b, and c, where a, b, and c can be single or multiple.

[0026] The present invention aims to provide a carbon footprint optimization method and system based on LCA-GLSS coupling. The method uses LCA to analyze the carbon emissions of a product throughout its entire life cycle, from raw material acquisition, production, use to waste disposal. It introduces big data AI technology to achieve a systematic assessment of carbon emissions throughout the product's life cycle. Combined with GLSS, it identifies key nodes with high carbon emissions and achieves closed-loop management of carbon emissions through "identification-analysis-improvement-control".

[0027] Traditional Life Cycle Assessment (LCA) methods for carbon footprint assessment suffer from challenges such as data collection difficulties, computational complexity, and difficulty in accurately identifying spatial differences. While Green Lean Six Sigma (GLSS) offers advantages in process optimization and quality improvement, its application alone in carbon footprint assessment struggles to comprehensively consider environmental impacts. This invention combines big data technology to rapidly collect, process, and analyze massive amounts of carbon footprint-related data, improving data accuracy and timeliness. Simultaneously, it utilizes artificial intelligence algorithms, such as machine learning and deep learning, to mine and analyze complex data, enabling accurate prediction and optimization of carbon footprint. This invention combines the comprehensiveness of LCA with the optimization capabilities of GLSS, forming a novel coupled approach that considers carbon emissions throughout the product's entire life cycle while optimizing and improving key aspects of the carbon footprint assessment process using GLSS tools and methods.

[0028] In a first aspect, embodiments of the present invention provide a carbon footprint optimization method based on LCA-GLSS coupling, comprising the following steps: S1. Carbon footprint assessment, specifically including: collecting carbon emission data at each stage of the product's entire life cycle using big data technology, and preprocessing the data, including data cleaning, screening, and standardization; based on the preprocessed data, quantitatively analyzing the carbon emissions of the product's entire life cycle according to the life cycle assessment method, calculating the carbon footprint of each stage, the carbon footprint contribution ratio of materials, energy, resources, and waste involved in each process unit in the entire life cycle, and the carbon footprint sensitivity of materials, energy, resources, and waste; See Figure 1 As one possible approach, carbon footprint assessment specifically includes the following steps: S10. Initialize the evaluation elements, specifically including: determining the evaluation object, evaluation objective, benchmark flow, measurement scale, system boundary, and selection criteria; As an example, evaluation objects could include acarbose API, antibiotics, vaccines, etc. The evaluation objective is to compare the carbon footprint of different production processes (e.g., fermentation versus chemical synthesis), assess carbon reduction optimization measures (e.g., solvent recovery, energy optimization), or provide data for carbon footprint labeling or carbon neutrality certification. A baseline flow is the unit of product required to perform a specified function in a given product system. Typically, a certain number of specific products or services are selected as the baseline flow for quantitative analysis. For example, a functional unit might be defined as 1 kg of drug, where, under national GMP-certified quality standards, a patient taking 500 mg daily for 7 days achieves a 95% cure rate for a certain disease.

[0029] As an example, system boundaries primarily include four forms: "cradle to gate," "cradle to grave," "gate to gate," and "gate to grave." When conducting research on the carbon footprint of pharmaceutical products, the system boundaries can be refined based on the entire lifecycle of the drug. See also... Figure 2 From upstream precursor chemical production to active pharmaceutical ingredient (API) synthesis, it is called "cradle to API production," applicable to API production or pharmaceutical intermediate research; from upstream precursor chemical production to formulation production, it is called "cradle to formulation production"; from upstream precursor chemical production to formulation packaging, it is called "cradle to packaging"; from upstream precursor chemical production through distribution and use to end of life (disposal), it is called "cradle to grave," applicable to pharmaceutical formulation production.

[0030] As an example, data selection rules can ignore factors with minimal impact on evaluation results, thereby simplifying the data collection and evaluation process. For instance, the basis for selection can be the weight ratio of each raw material input to the product's total weight or the total input of the process. Specific rules are as follows: Inputs and outputs for each process in the product lifecycle are screened. If the mass input of a single raw material is less than 1% of the total mass of all raw materials, or a single energy input is less than 1% of the total energy consumption, or a single emission source contributes less than 1% to the total carbon footprint, it can be excluded from the product's carbon footprint calculation. However, the sum of all ignored greenhouse gas emissions must not exceed 5% of the total product carbon footprint.

[0031] S11. Data acquisition and data processing, specifically including: dividing the M sequentially executed processes in the process chain into N process units according to preset rules, where N≤M; acquiring and preprocessing the input and output data of each process unit; It should be explained that the process chain is not limited to production or processing technology, but also includes upstream raw material supply and downstream packaging technology.

[0032] As an example, the pharmaceutical manufacturing process can be divided into process units such as bio-fermentation, refining and purification, packaging and storage, and distribution and transportation. Regarding data collection, data includes primary and secondary data. Primary data is specific data directly related to the research object or system, obtained through on-site measurement, surveys, monitoring, or from supply chain partners. This type of data is directly related to the specific production process, technical parameters, or supply chain activities of the research object, and has high spatiotemporal representativeness and technical relevance. For example, data such as factory electricity consumption and raw material procurement volume can be obtained from company reports or supply chain surveys. Secondary data is non-specific data from public databases, literature, industry reports, or general models, usually based on industry averages or standardized scenarios. This type of data is used to supplement missing primary data or as a substitute when primary data is unavailable. Examples include energy carbon emission factors and transportation emission coefficients from general industry databases (Ecoinvent, GaBi, etc.) or typical emission values ​​from the "China Pharmaceutical Industry Pollutant Emission Standard."

[0033] As one possible implementation method, data acquisition specifically includes: deploying a multimodal IoT sensor network to collect energy consumption parameters, environmental parameters, and device status parameters; configuring unstructured data including at least enterprise reports and industry reports; and configuring and deploying a data traceability blockchain with smart contracts to record energy consumption parameters, environmental parameters, device status parameters, and unstructured data. Data processing specifically includes data preprocessing and data fusion. Data preprocessing includes: using the Isolation Forest algorithm to identify drifting and breakpoint data, removing drifting data, and using a conditional generative adversarial network to fill in the data gaps and breakpoint data corresponding to drifting data. Data fusion includes: building a scalable data platform based on Hadoop / Spark to achieve real-time processing of multi-source heterogeneous data.

[0034] The Isolation Forest algorithm is used to identify drifting and breakpoint data, and to remove drifting data. For example, in a production process, after a temperature sensor has been running for a long time, it may exhibit numerical fluctuations that are significantly different from those of other sensors. The Isolation Forest algorithm can identify this drift through its trained model and issue an alarm signal in a timely manner. Simultaneously, the algorithm can also identify breakpoints in the data, such as the loss of energy consumption data due to equipment downtime, and then process or supplement this data. Conditional Generative Adversarial Networks (CGANs) are used to fill in the data gaps corresponding to drifting data and breakpoint data. For example, in some production processes, data loss may occur due to sensor malfunctions or equipment downtime. CGAN generates reasonable missing values ​​while preserving the data distribution characteristics through adversarial training between the generator and discriminator. Furthermore, in the case of missing energy consumption data for a certain piece of equipment, the generator will generate a reasonable value that matches the data based on historical data and current operating conditions, while the discriminator will determine whether the generated data conforms to the distribution characteristics of the original data.

[0035] As an example, a scalable data platform is built based on Hadoop / Spark, supporting real-time processing of multi-source heterogeneous data (structured and unstructured), with a daily processing capacity of ≥10TB and data latency of less than 5 minutes. A multimodal IoT sensor network is deployed to collect real-time energy consumption data (such as electricity, steam, compressed air, etc.), environmental parameters (such as temperature and humidity, carbon dioxide concentration, VOCs, etc.), and equipment status parameters (such as vibration spectrum, current harmonics, OEE, etc.). A supply chain data traceability chain is built based on blockchain to ensure the credibility of supply chain data. Specifically, smart contracts are deployed on blockchain platforms such as Ethereum or Hyperledger to record data at each link in the supply chain, ensuring data immutability.

[0036] For example, see Figure 3 It showcases the entire data flow of the supply chain, from raw material procurement, production, and transportation to final sales. The data is encrypted and stored at each stage to ensure transparency and traceability at every step.

[0037] S12. Construct a lifecycle model, specifically including: using graph neural networks to construct a process chain network model to map the correlation paths between material flow, energy flow and carbon emissions; using GWP100 assessment categories to calculate the lifecycle carbon footprint; and using digital twin technology to simulate the process flow to update carbon footprint data in real time. As one possible implementation, using graph neural networks to construct the process chain network model specifically includes: S120. Directed graph modeling, specifically including: treating each process step in the process chain as a node, and material and energy flows as directed edges, with the direction representing the flow path (e.g., process A → process B); the weight of the directed edges is configured as the carbon emission intensity corresponding to the flow rate, for example, materials; By treating each process step in the production chain as a node and material and energy flows as directed edges, a directed graph model is constructed. In this directed graph model, the graph neural network dynamically maps the correlation paths between material flow, energy flow, and carbon emissions through the relationships between nodes and the weights of edges.

[0038] S121. Configure a dynamic update mechanism, specifically including: collecting material / energy flow data of each process in real time through IoT sensors, inputting it into a graph neural network to update edge weights; and using a graph attention network to aggregate node features and dynamically calculate path carbon emissions. S122. Output of associated paths, specifically including: outputting the carbon emission hotspot paths of the entire process chain (such as Sankey diagram) through the message passing mechanism of GNN, with node coverage ≥98%.

[0039] The process chain network model constructed in this invention can adjust carbon emission paths based on real-time data, thereby achieving accurate carbon footprint tracking and optimization. It enables dynamic mapping of material / energy flow paths (node ​​coverage greater than or equal to 98%).

[0040] As one possible implementation, the process chain network model also includes a multivariate time-series carbon emission prediction unit based on an Attention-LSTM hybrid architecture. This is a dynamic time-series prediction module embedded in the process chain network model, used to predict the impact of multiple variables (such as energy consumption and process parameters) on carbon emissions in real time. Its function is to supplement the static analysis of the process chain network, enhancing the model's dynamic prediction capabilities through time-series data. LSTM is used to capture long-term dependencies in the time-series data, while the Attention mechanism enhances the focus on important time steps. In this architecture, the input data includes multiple variables, such as equipment energy consumption and production process parameters. Through the Attention mechanism, the model can automatically learn and focus on those time steps that have the greatest impact on carbon emissions. Ultimately, the model can accurately predict carbon emissions with a prediction error rate (MAE) controlled within 3% and a prediction time interval of no more than one hour.

[0041] As an example, lifecycle modeling begins with a process unit dataset; see [link to relevant documentation]. Figure 4 This outlines the steps from establishing a process unit to building a complete lifecycle model. The completeness of a pharmaceutical product lifecycle model encompasses three aspects: complete process unit datasets, complete intermediate flow traceability, and complete underlying database models. The ultimate credibility of pharmaceutical product lifecycle assessment can essentially be decomposed into two aspects: model completeness and data accuracy.

[0042] As one possible implementation, the SHAP value analysis module algorithm is embedded in the multivariate time-series carbon emission prediction unit to identify the processes that contribute the most to carbon emissions. As an interpretive component of the prediction unit, the SHAP algorithm quantifies the contribution of each process to carbon emissions. It relies on the output (prediction result) of the Attention-LSTM and infers key processes through feature importance analysis.

[0043] To improve the interpretability of the model, this invention embeds a SHAP (Shapley Additive Explanations) value analysis module into the multivariate time-series carbon emission prediction unit. SHAP value analysis can accurately identify which processes contribute the most to carbon emissions. Taking a pharmaceutical company's production process as an example, SHAP value analysis revealed that the refining process contributes 38% to carbon emissions, ranking first. This result can help production managers quickly identify and optimize high-carbon-emission processes.

[0044] This invention employs a multi-objective Proximal Policy Optimization (PPO) reinforcement learning algorithm to optimize the trade-off between carbon emissions, cost, and output. Specifically, the reward function consists of three parts: carbon emissions (weight 0.6), cost (weight 0.3), and output (weight 0.1). During training, the algorithm continuously adjusts the policy to generate a set of Pareto optimal solutions, achieving the best balance between carbon emissions, cost, and output. Through multiple iterations, an optimal optimization strategy suitable for different production scenarios is finally obtained.

[0045] As one possible implementation, the calculation method for lifecycle carbon footprint using the GWP100 assessment categories is as follows: in, For carbon footprint; For the first Such activities; For the first Medium greenhouse gases; Within the system boundary, the first functional unit... Data on GHG emissions and elimination related to various activities, including primary and secondary data; For the first Greenhouse gases corresponding to these activities Emission coefficient; Greenhouse gases of value.

[0046] As an example, see Table 1, which lists the emissions of six gases: carbon dioxide (CO2), methane (CH4), nitrous oxide (N2O), hydrofluorocarbons (HFCs), perfluorocarbons (PFCs), and sulfur hexafluoride (SF6). value: Table 1. Different greenhouse gases value The GWP100 values ​​for hydrofluorocarbons (HFCs) and perfluorocarbons (PFCs) are not listed in Table 1 above because fluorine-containing gases are not single gases and therefore cannot be represented by a single value. This table is only an example of how to obtain the GWP100 values ​​for different greenhouse gases; for specific values, please refer to the IPCC AR6 WG1 report.

[0047] As one possible implementation, a dynamic emission factor library is configured to automatically correct emission factors based on real-time operating conditions; the correction criteria are dynamically adjusted based on the following multi-dimensional parameters: Equipment load rate is monitored in real time by IoT sensors. When the load rate is ≥80%, a high load correction factor is triggered; for example, the emission factor is increased by 5% to 15%. Energy structure, matching differentiated emission factors according to energy type (such as coal, natural gas, green electricity) and real-time carbon intensity data (such as dynamic updates of grid carbon emission factors); The process parameters, combined with the reaction temperature and pressure process parameters, are used to calculate the dynamic compensation amount when the operating conditions deviate from the baseline value through a Bayesian network. Environmental data is used to compensate in real time for the impact of environmental parameters such as temperature, humidity, and atmospheric pressure on combustion efficiency or equipment energy efficiency. The dynamic emission factor database is updated synchronously through digital twin technology to ensure that the error rate between the corrected emission factors and actual carbon emissions is ≤3%.

[0048] S13. Calculate the carbon footprint contribution ratio of materials, energy, resources, and waste involved in each process unit throughout the life cycle; S14. Reduce the materials, energy, resources, and waste involved in each process unit according to a preset and identical ratio, obtain the carbon footprint of the materials, energy, resources, and waste involved in each process unit after the reduction, and calculate the carbon footprint sensitivity based on the carbon footprint before and after the reduction.

[0049] S2. Carbon footprint optimization specifically includes: using tools and methods of green Lean Six Sigma to optimize high-carbon emission links identified during the carbon footprint assessment process; identifying and eliminating waste and variation in the carbon emission process through causal analysis and statistical process control; using artificial intelligence algorithms to build carbon footprint prediction models based on historical data to predict and simulate future carbon emission trends, and improving the accuracy of predictions through model optimization and parameter adjustment; and comprehensively analyzing and interpreting the assessment results to identify the main contributing factors and key links of the carbon footprint, and formulating emission reduction measures and optimization suggestions.

[0050] As one possible approach, the carbon footprint assessment results obtained in S1 can be embedded into the analysis and improvement phase of the green lean six sigma method in S2, that is, the carbon footprint optimization scheme can be determined based on the carbon footprint contribution ratio and carbon footprint sensitivity.

[0051] As one possible implementation approach, the definition phase of the Green Lean Six Sigma methodology decomposes carbon emission targets into key quality characteristics of the production process using the Eco-QFD method, establishing a "carbon emission-process parameter" correlation matrix, specifically including: Based on the carbon footprint assessment results (e.g., processes contributing ≥10%), key emission reduction processes are identified; process-level emission reduction targets (e.g., 15% emission reduction per unit product) are mapped to process parameters (e.g., solvent recovery rate ≥90%, reaction temperature error ±2℃); a carbon footprint optimization scheme is generated using a multi-objective proximal strategy optimization (PPO) algorithm, with the following reward function: Through iterative training using digital twin simulation, a set of Pareto optimal solutions is output. In the analysis phase of the Green Lean Six Sigma method, a Bayesian network model is constructed based on the Pareto solution set from the definition phase. Nodes include process parameters (such as temperature and pressure), carbon emissions, cost, and output. Prior probabilities are assigned through historical data statistics (confidence ≥ 95%) or expert experience (weight ≥ 0.8). Monte Carlo sampling is used to quantify the sensitivity of process parameters, outputting a sensitivity index (such as the Sobol index). The top three ranked key parameters are selected for the improvement phase.

[0052] Secondly, embodiments of the present invention provide a carbon footprint optimization system based on LCA-GLSS coupling, see [link to relevant documentation]. Figure 5 It includes: carbon footprint assessment module 1 and process optimization module 2; The carbon footprint assessment module 1 includes an initialization unit 10, a data collection and inventory analysis unit 11, a life cycle impact assessment unit 12, and a result interpretation unit 13. The initialization unit 10 initializes the assessment elements, specifically including: determining the assessment object, assessment objective, baseline flow, measurement scale, system boundary, and selection criteria. The data collection and inventory analysis unit 11 collects and processes data, specifically including: dividing the M sequentially executed processes in the process chain into N process units according to preset rules, where N ≤ M; collecting and preprocessing the input and output data of each process unit; and using the life cycle impact assessment unit 12 to construct the life cycle impact assessment unit. The lifecycle model specifically includes: constructing a process chain network model using a graph neural network to map the correlation paths between material flow, energy flow, and carbon emissions; calculating the lifecycle carbon footprint using the GWP100 assessment categories; simulating the process flow using digital twin technology to update carbon footprint data in real time; calculating the carbon footprint contribution ratio of materials, energy, and waste involved in each process unit during the lifecycle using the result interpretation unit 13; reducing the materials, energy, and waste involved in each process unit by a preset and identical proportion to obtain the carbon footprint corresponding to the materials, energy, and waste involved in each process unit after the reduction; and calculating the carbon footprint sensitivity based on the carbon footprint before and after the reduction. The process optimization module 2 includes a current performance level assessment unit 20, an inefficiency factor analysis unit 21, and an optimal solution determination unit 22. The current performance level assessment unit 20, based on the lifecycle carbon footprint, carbon footprint contribution ratio, and carbon footprint sensitivity output by the result interpretation unit 13, as well as the configured sustainability value assessment and social indicator assessment, uses a multi-objective reinforcement learning algorithm to determine the carbon footprint optimization scheme. This involves optimizing the balance between carbon emissions, costs, and output. During training, the carbon footprint optimization scheme is continuously adjusted to generate a set of Pareto optimal solutions, achieving the best balance between carbon emissions, costs, and output. The inefficiency factor analysis unit 21 uses a Bayesian network with a prior probability greater than or equal to 0.8 to quantify the sensitivity of process parameters. The optimal solution determination unit 22 then selects the optimal solution from the Pareto optimal solution set.

[0053] The technical solution of this invention will be further illustrated below using the carbon footprint evaluation of mupirocin raw material as an example.

[0054] See Figure 6 This is a Sus-VSM diagram for mupirocin API, including the following data: traditional value stream mapping parameters such as process flow, number of workers, changeover time, and equipment utilization rate, as well as data on raw material consumption, water consumption, energy consumption, and carbon footprint data (percentage by process). The process flow for mupirocin API includes: seed culture, bio-fermentation, resin adsorption (fermentation broth), desorption and concentration, extraction, dehydration and decolorization, concentration and crystallization, recrystallization, cleaning and drying, pulverization and packaging, etc. Among them, the carbon footprint contribution of the seed culture and bio-fermentation processes is 28.505%, the carbon footprint contribution of the pulverization and packaging process is 2.299%, and the carbon footprint contribution of the other processes is 69.196%. Figure 6 This is just an example and omits the processes of analysis, concentration, extraction, dehydration and decolorization, concentration and crystallization, and recrystallization.

[0055] The following is the process of obtaining the life cycle carbon footprint of mupirocin active pharmaceutical ingredient using the life cycle assessment method: First, the evaluation object was determined to be mupirocin active pharmaceutical ingredient (API). The evaluation objective was to assess the carbon footprint of mupirocin API production, identify processes with high carbon footprint contributions based on the results, provide a basis for subsequent green lean Six Sigma process optimization, and re-evaluate after improvements by comparing the carbon footprint before and after improvements to measure the environmental performance of the improvements. The functional unit was mupirocin API with a purity of ≥98% under the quality level of national pharmaceutical GMP certification, with 1 kg of mupirocin API (excluding packaging) as the baseline flow. The system boundary was "cradle to API packaging". The exclusion rules were as follows: the input and output of each process in the product lifecycle were screened. If the mass input of a single raw material was less than 1% of the total mass of all raw materials, or the energy input of a single source was less than 1% of the total energy consumption, or the contribution of a single emission source to the total carbon footprint was less than 1%, it could be excluded from the product carbon footprint accounting. However, the sum of all ignored greenhouse gas emissions must not exceed 5% of the total product carbon footprint.

[0056] Secondly, see Figure 7 The production process flow chart for mupirocin active pharmaceutical ingredient is divided into ten steps, including seed culture and bio-fermentation. (See also...) Figure 8 Based on the production process, the ten steps of the mupirocin API production process are divided into four process units: "fermentation," "refining," "packaging," and "wastewater treatment." "Fermentation" corresponds to the first two steps of the production process, "refining" to the third through ninth steps, and "packaging" to the last step. The input and output data for these four process units are foreground data, while the upstream data for raw materials, energy, and waste comes from the background data in the Ecoinvent 3.9.1 database. The carbon footprint calculation for this project divides the process units according to the steps. Wastewater treatment is defined here as a "unit process," as each stage generates wastewater. Since the wastewater treatment process is completed within the plant (not from an external manufacturer), an additional downstream unit process is traced. Figure 8 In other words, the wastewater treatment unit process corresponds to the wastewater treatment process. In practical applications, enterprises usually produce multiple varieties of active pharmaceutical ingredients (APIs) simultaneously, and the water and energy used in their production processes need to be allocated and processed. In this embodiment, the "weight allocation method" is used to divide the system's input and output data into different products or functions. After statistical processing, the list data of the four process units with 1 kg of mupirocin API as the baseline flow are shown in Tables 2 to 5.

[0057] Table 2 Fermentation process inventory data Table 3. Refining Process Checklist Data Table 4 Packaging Process Checklist Data Table 5 Wastewater Treatment Process Inventory Data Next, the OpenLCA lifecycle modeling platform will be used to model the lifecycle of mupirocin active pharmaceutical ingredient. See the model diagram below. Figure 9 The impact assessment method selected was the Ecoinvent-IPCC 2021 method built into the Ecoinvent database, and the impact assessment category was GWP100. The LCA results for 1 kg of mupirocin active pharmaceutical ingredient were calculated. See [link / reference]. Figure 6 According to the LCA results, when the impact assessment category is GWP100, the lifecycle carbon footprint of producing 1 kg of mupirocin API is 435.42 kgCO2eq. The following is a contribution analysis and sensitivity analysis. Contribution analysis identifies the stage, substance, or process with the greatest environmental impact. The raw materials and excipients are mapped to their respective production processes, yielding the following carbon footprint contribution percentages. The contribution percentages for producing 1 kg of mupirocin API under GWP100 are shown in Table 6: Table 6. Contribution of GWP100 to the production of 1kg of mupirocin API The cumulative Pareto chart of GWP100 contribution during the production of mupirocin active pharmaceutical ingredient is shown below. Figure 10 As shown in the Sankey diagram, a maximum of 25 processes are set to produce 1 kg of mupirocin active pharmaceutical ingredient (API). Figure 11 As shown.

[0058] According to Table 6, based on the different processes in the production process, the carbon footprint contribution of the mupirocin raw material extraction process is more than twice that of the fermentation stage, accounting for 69.20%, due to the more complex production process involving a large amount of solvent use, energy consumption, and wastewater treatment. The carbon footprint contribution of the fermentation process is 28.51%, and the packaging process has the lowest contribution at 2.3%.

[0059] according to Figure 10 Pareto chart and Figure 11The Sankey diagram, categorized by system output type, shows that, firstly, the carbon footprint of raw material refining is 110.28 kg CO2 equivalent, contributing the most at 25.33% of the total. The top five raw materials are ethyl acetate, acetone, isobutyl acetate, resin, and n-heptane, primarily solvents used in the refining of mupirocin API, accounting for 10.92%, 8.65%, 1.96%, 1.09%, and 1.02% of the total carbon footprint per kg of mupirocin API production, respectively. Secondly, wastewater treatment from mupirocin API production contributes significantly, accounting for 21.23%. Most of this wastewater originates from the refining stage of mupirocin API. The two largest contributors to wastewater treatment are liquid alkali and electricity from the power grid. Liquid alkali serves as the primary neutralizing agent in wastewater treatment, ensuring that the neutralized wastewater meets the pH requirements for discharge. In terms of energy use, the carbon footprint contributions of grid electricity, steam, and photovoltaic power generation accounted for 15.57%, 12.63%, and 0.51% respectively, totaling 28.71% of the overall contribution. This indicates that the energy consumption carbon footprint contribution of the mupirocin raw material production process is also relatively high. Next are solid waste treatment (12.84%), fermentation raw materials (6.54%), transportation of purchased raw materials (5.16%), and packaging raw materials (0.18%). Among these, solid waste treatment is incineration, which has a relatively high carbon footprint contribution. The top three fermentation raw materials are ammonia (2.90%), glucose (1.85%), and glycerol (1.02%). Finally, the carbon footprint contribution of industrial water and dust generated during the packaging process is relatively small, both less than 0.01%.

[0060] Sensitivity analysis is used to assess the impact of different parameters and assumptions on LCA results to identify which factors are most sensitive to carbon footprint results. In this example, a 10% reduction was implemented to conduct sensitivity analysis on fermentation raw materials, refining raw materials, packaging raw materials, grid electricity, photovoltaic power, steam, industrial water, transportation of purchased raw materials, wastewater treatment, and solid waste in the production of 1 kg of mupirocin API, observing their impact on the GWP100 result of 1 kg of mupirocin API. The top 15 factors with the highest sensitivity were selected, and the results are shown in Table 7. Table 7. Sensitivity analysis results of GWP100 for the production of 1 kg of mupirocin API. Table 7 shows that wastewater treatment has the highest sensitivity; a 10% reduction in wastewater treatment volume reduces the overall carbon footprint by 2.12%. The consumption of steam and electricity also has a relatively large impact on the lifecycle carbon footprint of 1 kg of mupirocin raw material. A 10% reduction in grid electricity consumption reduces the carbon footprint by 1.56%, and a 10% reduction in steam consumption reduces the carbon footprint by 1.26%. The sensitivity for solid waste treatment is 1.28%, and for the transportation of purchased raw materials it is 0.52%. The percentages of the raw materials ethyl acetate, acetone, ammonia, isobutyl acetate, glucose, resin, n-heptane, glycerol, activated carbon, and bicarbonate are 1.09%, 0.86%, 0.29%, 0.20%, 0.19%, 0.11%, 0.10%, 0.10%, 0.06%, and 0.05%, respectively.

[0061] In the subsequent GLSS production process optimization and improvement phase, referring to the above results of the carbon footprint evaluation of mupirocin API, improvements can be made from the following aspects: First, in terms of energy, reduce overall energy consumption or adjust the energy structure to increase the use of green energy; in terms of solid waste treatment, reduce the generation of solid waste or reasonably change the solid waste treatment method; in terms of wastewater treatment, reduce wastewater discharge or optimize the wastewater treatment process to reduce the raw materials or energy used in wastewater treatment; in terms of raw materials used in fermentation and refining, improve the utilization rate of raw materials, such as increasing solvent recovery rate and resin recovery rate, and reduce waste through lean improvement; in terms of the transportation of purchased raw materials, reduce transportation distance or change the transportation method.

[0062] Although the invention has been described herein in conjunction with various embodiments, those skilled in the art will understand and implement other variations of the disclosed embodiments by reviewing the accompanying drawings, disclosure, and other materials. In this specification, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude multiple components. A single processor or other unit can implement several functions listed in the specification. While certain measures are described in different embodiments, this does not mean that these measures cannot be combined to produce good results.

[0063] Although the invention has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made therein without departing from the spirit and scope of the invention. Accordingly, this specification and drawings are merely illustrative of the invention and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if such modifications and modifications fall within the scope of the invention and its equivalents, the invention is also intended to include such modifications and modifications.

Claims

1. A carbon footprint optimization method based on LCA-GLSS coupling, characterized in that, Includes the following steps: S1. Carbon footprint assessment, specifically including: collecting carbon emission data at each stage of the product's entire life cycle using big data technology, and preprocessing the data, including data cleaning, screening, and standardization; based on the preprocessed data, quantitatively analyzing the carbon emissions of the product's entire life cycle according to the life cycle assessment method, calculating the carbon footprint of each stage, the carbon footprint contribution ratio of materials, energy, resources, and waste involved in each process unit in the entire life cycle, and the carbon footprint sensitivity of materials, energy, resources, and waste; S2. Carbon footprint optimization specifically includes: using tools and methods of green Lean Six Sigma to optimize high-carbon emission links identified during the carbon footprint assessment process; identifying and eliminating waste and variation in the carbon emission process through causal analysis and statistical process control; using artificial intelligence algorithms to build carbon footprint prediction models based on historical data to predict and simulate future carbon emission trends, and improving the accuracy of predictions through model optimization and parameter adjustment; and comprehensively analyzing and interpreting the assessment results to identify the main contributing factors and key links of the carbon footprint, and formulating emission reduction measures and optimization suggestions.

2. The carbon footprint optimization method based on LCA-GLSS coupling according to claim 1, characterized in that, Carbon footprint assessment specifically includes the following steps: S10. Initialize the evaluation elements, specifically including: determining the evaluation object, evaluation objective, benchmark flow, measurement scale, system boundary, and selection criteria; S11. Data acquisition and data processing, specifically including: dividing the M sequentially executed processes in the process chain into N process units according to preset rules, where N≤M; acquiring and preprocessing the input and output data of each process unit; S12. Construct a lifecycle model, specifically including: using graph neural networks to construct a process chain network model to map the correlation paths between material flow, energy flow and carbon emissions; using GWP100 assessment categories to calculate the lifecycle carbon footprint; and using digital twin technology to simulate the process flow to update carbon footprint data in real time. S13. Calculate the carbon footprint contribution ratio of materials, energy, resources, and waste involved in each process unit throughout the life cycle; S14. Reduce the materials, energy, resources, and waste involved in each process unit according to a preset and identical ratio, obtain the carbon footprint of the materials, energy, resources, and waste involved in each process unit after the reduction, and calculate the carbon footprint sensitivity based on the carbon footprint before and after the reduction.

3. The carbon footprint optimization method based on LCA-GLSS coupling according to claim 2, characterized in that, Data acquisition specifically includes: deploying a multimodal IoT sensor network to collect energy consumption parameters, environmental parameters, and device status parameters; configuring unstructured data, including at least enterprise reports and industry reports; and configuring and deploying a data traceability blockchain with smart contracts to record energy consumption parameters, environmental parameters, device status parameters, and unstructured data. Data processing specifically includes data preprocessing and data fusion. Data preprocessing includes: using the Isolation Forest algorithm to identify drifting and breakpoint data, removing drifting data, and using a conditional generative adversarial network to fill in the data gaps and breakpoint data corresponding to drifting data. Data fusion includes: building a scalable data platform based on Hadoop / Spark to achieve real-time processing of multi-source heterogeneous data.

4. The carbon footprint optimization method based on LCA-GLSS coupling according to claim 2, characterized in that, The specific methods for constructing a process chain network model using graph neural networks include: S120. Directed graph modeling, specifically including: treating each process in the process chain as a node, material flow and energy flow as directed edges, with the direction representing the flow path; the weight of the directed edge is configured as the carbon emission intensity corresponding to the flow amount; S121. Configure a dynamic update mechanism, specifically including: collecting material / energy flow data of each process in real time through IoT sensors, inputting it into a graph neural network to update edge weights; and using a graph attention network to aggregate node features and dynamically calculate path carbon emissions. S122. Correlation path output, specifically including: outputting the carbon emission hotspot paths of the entire process chain through the message passing mechanism of GNN, with a coverage of ≥98%.

5. The carbon footprint optimization method based on LCA-GLSS coupling according to claim 4, characterized in that, The process chain network model is also equipped with a multivariate time-series carbon emission prediction unit based on the Attention-LSTM hybrid architecture; LSTM is used to capture long-term dependencies in time-series data, and the Attention mechanism enhances attention to important time steps. The multivariate time-series carbon emission prediction unit embeds the SHAP value analysis module algorithm to identify the process that contributes the most to carbon emissions.

6. The carbon footprint optimization method based on LCA-GLSS coupling according to claim 2, characterized in that, The calculation method for lifecycle carbon footprint using the GWP100 assessment categories is as follows: in, For carbon footprint; For the first Such activities; For the first Medium greenhouse gases; Within the system boundary, the first functional unit... Data on GHG emissions and elimination related to various activities, including primary and secondary data; For the first Greenhouse gases corresponding to these activities Emission coefficient; Greenhouse gases of value.

7. The carbon footprint optimization method based on LCA-GLSS coupling according to claim 6, characterized in that, Configure a dynamic emission factor library to automatically correct emission factors based on real-time operating conditions; dynamically adjust the correction criteria based on the following multi-dimensional parameters: The device load rate is monitored in real time by IoT sensors. When the load rate is ≥80%, a high load correction factor is triggered. Energy structure, matching differentiated emission factors based on energy type and real-time carbon intensity data; The process parameters, combined with the reaction temperature and pressure process parameters, are used to calculate the dynamic compensation amount when the operating conditions deviate from the baseline value through a Bayesian network. Environmental data is used to compensate in real time for the impact of environmental parameters such as temperature, humidity, and atmospheric pressure on combustion efficiency or equipment energy efficiency. The dynamic emission factor database is updated synchronously through digital twin technology to ensure that the error rate between the corrected emission factors and actual carbon emissions is ≤3%.

8. The carbon footprint optimization method based on LCA-GLSS coupling according to claim 1, characterized in that, The carbon footprint evaluation results obtained in S1 are embedded into the analysis and improvement phase of the green lean six sigma method described in S2, that is, the carbon footprint optimization scheme is determined based on the carbon footprint contribution ratio and carbon footprint sensitivity.

9. The carbon footprint optimization method based on LCA-GLSS coupling according to claim 8, characterized in that: In the definition phase of the Green Lean Six Sigma methodology, the carbon emission target is decomposed into key quality characteristics of the production process using the Eco-QFD method, establishing a "carbon emission-process parameter" correlation matrix. Specifically, this includes: identifying key emission reduction processes based on carbon footprint assessment results; mapping process-level emission reduction targets to process parameters; and generating a carbon footprint optimization scheme using a multi-objective near-end strategy optimization algorithm, with the following reward function: Through iterative training using digital twin simulation, a set of Pareto optimal solutions is output. In the analysis phase of the Green Lean Six Sigma method, a Bayesian network model is constructed based on the Pareto solution set from the definition phase. The nodes include process parameters, carbon emissions, costs, and outputs. Prior probabilities are assigned through historical data statistics or expert experience. The Monte Carlo sampling method is used to quantify the sensitivity of process parameters, output sensitivity indices, and select the top three key parameters for the improvement phase.

10. A carbon footprint optimization system based on LCA-GLSS coupling, characterized in that, include: Carbon footprint assessment module and process optimization module; The carbon footprint assessment module includes an initialization unit, a data collection and inventory analysis unit, a life cycle impact assessment unit, and a results interpretation unit. The initialization unit initializes the assessment elements, specifically defining the assessment object, assessment objective, baseline flow, measurement scale, system boundary, and selection criteria. The data collection and inventory analysis unit collects and processes data, specifically dividing the M sequentially executed processes in the chain into N process units (N≤M) according to preset rules; collecting and preprocessing the input and output data of each process unit; and constructing a life cycle model using the life cycle impact assessment unit. Specifically, this includes: constructing a process chain network model using graph neural networks to map the correlation paths between material flow, energy flow, and carbon emissions; calculating the life cycle carbon footprint using GWP100 assessment categories; simulating the process flow using digital twin technology to update carbon footprint data in real time; calculating the carbon footprint contribution ratio of materials, energy, and waste involved in each process unit during the life cycle using the result interpretation unit; reducing the materials, energy, and waste involved in each process unit by a preset and equal proportion to obtain the carbon footprint of materials, energy, and waste involved in each process unit after reduction; and calculating carbon footprint sensitivity based on the carbon footprint before and after reduction. The process optimization module includes a current performance level assessment unit, an inefficiency factor analysis unit, and an optimal solution determination unit. The current performance level assessment unit, based on the lifecycle carbon footprint, carbon footprint contribution ratio, and carbon footprint sensitivity output by the results interpretation unit, along with configured sustainability value assessment and social indicator assessment, uses a multi-objective reinforcement learning algorithm to determine the carbon footprint optimization scheme. This involves optimizing the balance between carbon emissions, costs, and output. During training, the carbon footprint optimization scheme is continuously adjusted to generate a set of Pareto optimal solutions, achieving the best balance between carbon emissions, costs, and output. The inefficiency factor analysis unit uses a Bayesian network with a prior probability greater than or equal to 0.8 to quantify the sensitivity of process parameters. The optimal solution determination unit then selects the best solution from the Pareto optimal solution set.

Images