A product carbon footprint accounting method and system considering green electricity green certificate consumption

Abstract

The application discloses a product carbon footprint accounting method and system considering green electricity and green certificate consumption, and belongs to the field of carbon footprint accounting, and comprises the following steps: S1, constructing a multi-source heterogeneous standardized data set and a credibility weighted carbon accounting rule library; S2, determining green certificate additional emission reduction and green certificate-carbon bidirectional offsetting coefficients; S3, accounting to obtain product basic carbon footprint and electricity link carbon emission; S4, obtaining the carbon footprint intermediate value considering green electricity and green certificate consumption through step difference adjustment; S5, based on the credibility weighted carbon accounting rule library, outputting the final product carbon footprint with a 95% confidence interval and an uncertainty attribution report. The above product carbon footprint accounting method and system considering green electricity and green certificate consumption realize scene adaptive and accurate accounting of product carbon footprint in a green electricity and green certificate coupling scene, quantitative analysis of uncertainty and traceable attribution of error parameters.

Description

Technical Field

[0001] This invention relates to the field of technology carbon footprint accounting, and in particular to a product carbon footprint accounting method and system that takes into account the consumption of green electricity and green certificates. Background Technology

[0002] Product carbon footprint accounting is a core technical support for enterprises' low-carbon information disclosure, supply chain low-carbon management, green electricity and green certificate trading and carbon quota linkage management. In particular, in complex production scenarios that integrate green electricity consumption and green certificate deduction, it puts forward stringent requirements on the accuracy, scenario adaptability, quantifiability of uncertainty and traceability of error in product carbon footprint accounting.

[0003] Existing carbon footprint accounting technologies mostly adopt a manual LCI (Life Cycle Inventory) and fixed empirical parameter assignment accounting model, which has problems such as low accounting efficiency, large human subjective error, and insufficient industry adaptation and standardization. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for calculating the carbon footprint of products that takes into account the consumption of green electricity and green certificates, thereby solving the above-mentioned technical problems.

[0005] To achieve the above objectives, this invention provides a product carbon footprint accounting method considering green electricity and green certificate consumption, comprising the following steps: S1, constructing a multi-source heterogeneous standardized dataset and a credibility-weighted carbon accounting rule base; S2, based on the multi-source heterogeneous standardized dataset, determining the additional emission reduction amount of green certificates and the green certificate-carbon two-way deduction coefficient; S3, based on the multi-source heterogeneous standardized dataset, calculating the product's basic carbon footprint and carbon emissions in the electricity consumption stage; S4, based on the additional emission reduction amount of green certificates and the green certificate-carbon two-way deduction coefficient, combined with the basic carbon footprint and carbon emissions in the electricity consumption stage, obtaining the median carbon footprint considering green electricity and green certificate consumption through tiered differentiated deduction adjustments; S5, based on the credibility-weighted carbon accounting rule base, performing fuzzy rule synthesis, possibility-probability mapping, and three-scenario Monte Carlo simulation on the median carbon footprint, outputting the final product carbon footprint with a 95% confidence interval and an uncertainty attribution report.

[0006] A system for calculating the carbon footprint of products that considers the consumption of green electricity and green certificates includes: A module for constructing a multi-source heterogeneous data acquisition and credibility-weighted carbon accounting rule base is used to build a multi-source heterogeneous standardized dataset and a credibility-weighted carbon accounting rule base. The module for confirming the ownership of green electricity and green certificates and quantifying additional emission reductions is used to determine the additional emission reductions of green certificates and the green certificate-carbon two-way deduction coefficient based on multi-source heterogeneous standardized datasets. A lightweight, large-model-driven module for calculating the basic carbon footprint of a product lifecycle is used to calculate the basic carbon footprint of a product and carbon emissions from electricity consumption based on a multi-source heterogeneous standardized dataset. The Green Electricity and Green Certificate Differentiated Deduction and Carbon Footprint Dynamic Adjustment Module is used to obtain the median carbon footprint considering green electricity and green certificate consumption by adjusting the tiered differentiated deduction based on the additional emission reduction of green certificates and the green certificate-carbon two-way deduction coefficient, combined with the basic carbon footprint and carbon emissions in the electricity consumption process. The fuzzy-probability coupled carbon footprint uncertainty quantification and attribution module is used to synthesize fuzzy rules, perform possibility-probability mapping and three-scenario Monte Carlo simulation on intermediate carbon footprint values ​​based on a credibility-weighted carbon accounting rule base, and output a final product carbon footprint and uncertainty attribution report with a 95% confidence interval.

[0007] Therefore, the present invention employs the above-mentioned product carbon footprint accounting method and system that considers the consumption of green electricity and green certificates, and has the following beneficial effects: 1. The PCF-RWKV carbon accounting-specific language model automatically generates a product lifecycle list, replacing manual sorting of accounting boundaries and activity data, significantly reducing human intervention errors and improving the standardization and intelligence of carbon accounting basic data generation; 2. Based on the multi-dimensional accounting scenario feature dimensions, the activation degree of the rule scenario is quantified. Combined with the rule credibility weight, the weighted Max-Min operator is used to complete the fusion of multi-source rules, which solves the problems of traditional rule one-size-fits-all adaptation and failure to distinguish between scenario adaptability and rule credibility. The fuzzy expression of parameters is more in line with the actual carbon accounting working conditions. 3. By using confidence interval denoising, high confidence interval filtering, and truncated normal distribution construction, qualitative fuzzy rule information is transformed into a quantitative probability distribution that can be used for Monte Carlo simulation, filling the technical gap that fuzzy information in carbon accounting is difficult to use directly for stochastic simulation. 4. The high confidence interval of the uncertainty parameter is divided into three sub-scenarios with low, medium and high parameter values ​​and the simulation results are weighted and fused. Compared with the traditional single full-domain sampling, it can hierarchically represent the differentiated impact of different parameter value intervals on carbon footprint, and the calculation results are more robust. 5. Breaking through the limitations of traditional local sensitivity analysis, it takes into account both the independent effects of parameters and the pairwise interactive coupling effects, quantifies the uncertainty contribution of each parameter such as carbon emission factor, activity data, green certificate extraneousness coefficient, two-way deduction coefficient, and marginal carbon factor, and can accurately locate the core error source, providing a quantitative basis for carbon accounting data optimization. 6. It incorporates exclusive parameters such as the green certificate additionality coefficient, marginal carbon factor, and green certificate-carbon two-way deduction coefficient to achieve integrated uncertainty quantification of conventional carbon emission accounting and green electricity and green certificate emission reduction deduction, and adapts to the complex product carbon footprint accounting needs of green certificate consumption under the dual carbon background.

[0008] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0009] Figure 1 This is a flowchart illustrating a product carbon footprint accounting method that considers green electricity and green certificate consumption, as described in this invention. Detailed Implementation

[0010] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely illustrative of the embodiments of the present invention and are not intended to limit the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of this application. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout.

[0011] It should be noted that the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, such as a process, method, system, product, or server that includes a series of steps or units, not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such process, method, product, or device.

[0012] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0013] First, it should be noted that the carbon footprint accounting scope of this invention covers the entire life cycle of power products, specifically including: upstream emissions generated from the acquisition of upstream raw materials and supporting equipment, production and processing, and logistics and transportation; midstream emissions generated from equipment manufacturing, power production and operation, and power grid transmission and transformation; and downstream emissions generated from equipment scrapping and dismantling, waste disposal, and material recycling and reuse.

[0014] like Figure 1 As shown, a product carbon footprint accounting method considering green electricity and green certificate consumption includes the following steps: S1, constructing a multi-source heterogeneous standardized dataset and a credibility-weighted carbon accounting rule base; S2, based on the multi-source heterogeneous standardized dataset, determining the additional emission reduction from green certificates and the green certificate-carbon two-way deduction coefficient; S3, based on the multi-source heterogeneous standardized dataset, calculating the product's basic carbon footprint and carbon emissions from the electricity consumption stage; S4, based on the additional emission reduction from green certificates and the green certificate-carbon two-way deduction coefficient, combined with the basic carbon footprint and carbon emissions from the electricity consumption stage, obtaining the median carbon footprint considering green electricity and green certificate consumption through tiered differentiated deduction adjustments; S5, based on the credibility-weighted carbon accounting rule base, performing fuzzy rule synthesis, possibility-probability mapping, and three-scenario Monte Carlo simulation on the median carbon footprint, outputting the final product carbon footprint with a 95% confidence interval and an uncertainty attribution report.

[0015] Step S1 specifically includes the following steps: S11. Collect multi-source heterogeneous raw data; multi-source heterogeneous raw data includes product-side data, green certificate-side data, grid-side data, domain evidence data, and market-side data; among them, product-side data includes production process description texts of the electricity products to be accounted for, raw material consumption records, energy consumption records, and production process parameters; green certificate-side data is the transaction ledger of green electricity and green certificates, which records the transaction time, transaction entity, the installed capacity type of green electricity corresponding to the green certificate, and the transaction binding relationship; grid-side data includes regional grid power generation structure data, carbon emission factors of various types of power sources, and transmission and distribution loss data; domain evidence data includes carbon accounting-related standards, industry reports, academic literature, and enterprise accounting cases; market-side data includes carbon market quota data, transaction prices, green certificate market transaction prices, and supply and demand data.

[0016] S12. Multi-source data standardization preprocessing: The multi-source heterogeneous raw data collected in S11 are sequentially processed by outlier removal, missing value filling, unit unification, and text vectorization.

[0017] S13, Distillation of multi-source evidence.

[0018] S131. Construct domain prompts: Construct a Few-Shot prompt template specifically for carbon accounting, embed a standard example of carbon accounting rule extraction in the prompt, and explicitly require that the rules extracted by the large model must include the applicable industry of the rule, the name of the rule parameter, the range of parameter values ​​and the rule description. At the same time, it automatically marks the source type of the current evidence as the source attribute label of the rule.

[0019] S132. Construct and deploy the PCF-RWKV domain-specific language model. The PCF-RWKV domain-specific language model includes a domain input layer, an RWKV basic encoding layer, a LoRA domain adaptation layer, and a multi-task output layer arranged sequentially. The domain input layer loads a carbon computation-specific enhanced vocabulary, performs terminology alignment and embedding encoding on the vectorized text output from S12, converting the text into a 768-dimensional fixed-length embedding vector sequence, which is then input into the subsequent RWKV basic encoding layer. The RWKV basic encoding layer performs linear temporal feature extraction on the input embedding vectors. The LoRA domain adaptation layer, while freezing the basic weights, trains only the low-rank update matrix to achieve domain knowledge injection, and the weight update formula is: ; In the formula, Weights are applied after domain adaptation; These are the original weights for pre-training; This is the scaling factor, with a value of 16. It is a low-rank dimension, with a value of 8; , All are low-rank adaptation matrices; The multi-task output layer sets up a dedicated output header for evidence distillation to output structured carbon accounting rules, ensuring that the rule format is consistent and the fields are complete.

[0020] S133. Perform batch rule extraction of multi-source evidence based on PCF-RWKV domain-specific language model: Input the domain prompt words generated in S131 and the domain evidence text preprocessed in S12 into PCF-RWKV domain-specific language model to extract four original rules in batch: accounting boundary rules, carbon emission factor value rules, green certificate deduction application rules, and carbon quota linkage rules. Each carbon accounting rule carries industry type, parameter name, value range, and source type attributes, and outputs the original carbon accounting rule set.

[0021] S134. Fuzzy Rule Standardization Transformation: For the extracted original rules, the carbon accounting uncertainty parameters corresponding to the value range attributes are converted into triangular fuzzy membership degree format: ; In the formula, For the first Membership degree of the carbon accounting rule; , , All are membership parameters; This represents the actual value of the uncertainty parameter in carbon accounting.

[0022] S135. Rule deduplication and conflict pre-verification: Calculate cosine similarity based on rule semantic vectors and merge duplicate rules with cosine similarity greater than 0.95; mark conflicting rules with the same parameter intersecting intervals and retain the conflict attribute without forced correction, for subsequent credibility weighting and differentiation.

[0023] S136. Output the initial carbon accounting rule set and source attribute labels: Use the deduplicated and verified fuzzy rules as the initial carbon accounting rule set, and output the source type label corresponding to each carbon accounting rule simultaneously. Input them into S14 to perform confidence weighting.

[0024] S14. Calculate the credibility weight of the rules based on the entropy weight method and construct a credibility-weighted carbon accounting rule library.

[0025] S141. Construct a rule credibility evaluation matrix: Select three dimensions—source authority, rule consistency, and scenario coverage—to score the initial carbon accounting rule set output in S13 and construct an evaluation matrix. : ; in, ; ; ; In the formula, For the first The source of the carbon accounting rules is authoritative and standardized scoring; For the first The source type of the carbon accounting rule and its original score, and ; The highest score for the source type, and ; For the first Standardized scoring of rule consistency for carbon accounting rules; This represents the total number of rules applicable to the same accounting parameter. It is the intersection interval of all intervals with the same rule; It is the union of all intervals with the same rules; This is a function for calculating the length of an interval. For the first Standardized scoring of scenario coverage for carbon accounting rules; For the first The carbon accounting rules specify the total number of applicable accounting scenarios; The maximum number of accounting scenarios covered across all rules, including industry type, product type, green certificate type, regional power grid type, and accounting boundary.

[0026] S142. Standardize the evaluation matrix using the positive range of maximum and minimum values, and then calculate the dimensional weights: ; in, ; ; In the formula, For the first Dimension weights These respectively represent source authority, rule consistency, and scenario coverage; For the first Dimensional information entropy; For the first Rule No. 1 of the carbon accounting rules The proportion of dimensions; For the first Rule No. 1 of the carbon accounting rules Standardized scores for each dimension; This represents the total number of carbon accounting rules.

[0027] S143. Calculate the overall credibility weight of a single rule: ; In the formula, For the first The overall credibility weight of the carbon accounting rules; S144. Construct a credibility-weighted carbon accounting rule base: Associate each carbon accounting rule with its corresponding credibility weight. Membership parameters are bound together to form a credibility-weighted carbon accounting rule base.

[0028] S15. Output multi-source heterogeneous standardized datasets and a credibility-weighted carbon accounting rule base.

[0029] In step S132, the temporal coding update expression of the RWKV base coding layer is as follows: ; ; In the formula, for Time-bound key vector; for Time value vector; , All are learnable weight matrices of the RWKV basic coding layer; , These are all timing-gated parameters used to balance current input with historical information; for Input the embedding vector at each time step; for Input the embedding vector at each time step; For Hadama accumulation.

[0030] Step S2 specifically includes the following steps: S21. Identify the type of green certificate transaction based on the binding relationship of green certificate transactions: If in the green certificate transaction data, the buyer of green electricity simultaneously obtains the corresponding green certificate, and the green electricity consumption record corresponds one-to-one with the green certificate transaction record, and there is no situation where green certificates and green electricity are traded separately, then it is marked as a certificate-electricity integrated type of green certificate; if the green certificate and the corresponding green electricity are traded separately, that is, the green electricity producer sells green certificates, and the green electricity is traded separately in the electricity market, and there is no binding relationship between the two, then it is marked as a certificate-electricity separated type of green certificate.

[0031] S22, Calculate the marginal carbon reduction factor of green electricity in the region : ; In the formula, The carbon emission factor of marginal thermal power units in the region is taken from the carbon emission factor of the most recently commissioned thermal power units in the region. To correspond to the full life cycle carbon emission factor of green electricity, the corresponding full life cycle carbon emission factor is selected according to the type of green electricity (wind power, photovoltaic, hydropower, etc.).

[0032] S23. Calculate the additional emission reductions from green certificates. : ; In the formula, The trading volume of green certificates is expressed in units of 1 green certificate, with 1 green certificate corresponding to 1 MWh of green electricity. The additionality coefficient for the integrated green certificate for electricity generation is set at 0.95, meaning that 95% of the emission reduction is a real additional emission reduction, while retaining 5% uncertainty. To verify the additionality coefficient of the electricity separation type green certificate, a value of 0.3 is set, meaning that only 30% of the emission reduction is a real additional emission reduction, while the remaining 70% is double-counting or false emission reduction.

[0033] S24. Calculate the green certificate-carbon two-way deduction factor. : ; in, ; In the formula, The company's initial remaining carbon allowance; The total annual carbon allowance allocated to the enterprise; This represents the company's annual baseline carbon emissions.

[0034] This formula means that when a company has sufficient quotas ( When the quota is insufficient, the deduction factor is 0.9 to prevent enterprises from excessively hoarding green certificates; when the enterprise quota is insufficient ( When the green certificate deduction factor is 1.1, it incentivizes enterprises to prioritize the use of green certificates for deduction, thereby alleviating quota pressure. This dynamic factor addresses the irrationality of the traditional static deduction and adapts to the supply and demand changes in the carbon market.

[0035] In this embodiment, attributes such as the type of green certificate, additional emission reduction, and deductible quota are also stored and confirmed, generating a unique confirmation identifier for each green certificate to avoid duplicate deductions. At the same time, the attribute data is synchronized to subsequent deduction steps as the basis for deduction.

[0036] Step S3 specifically includes the following steps: S31. Semantic parsing of product manufacturing process: Using a pre-trained PCF-RWKV domain-specific language model, semantic parsing is performed on the text describing the manufacturing process to extract semantic features.

[0037] S32. Multi-Agent Task Decomposition and Collaboration: Based on the Autogen framework, the product carbon footprint accounting task is decomposed into multiple sub-tasks and assigned to different dedicated agents: LCI generation agent: responsible for the generation of lifecycle phase divisions; raw material accounting agent: responsible for extracting carbon accounting activity data in the raw material stage; energy consumption accounting agent: responsible for extracting carbon accounting activity data in the energy consumption stage; carbon emission factor matching agent: responsible for matching carbon emission factors.

[0038] S33. Input the subtask list generated in S32 into the PCF-RWKV domain-specific language model to generate a product lifecycle list.

[0039] S34. Extract all carbon emission factor-related rules from the completed credibility-weighted carbon accounting rule library, standardize them into carbon emission factor entries according to industry and scenario, and construct a structured carbon emission factor library.

[0040] Then, for each carbon accounting activity data in the S33 lifecycle inventory, the corresponding carbon emission factor is matched from the carbon emission factor library through semantic matching, and the matching score is calculated. The calculation formula is as follows: ; in, ; ; In the formula, The semantic similarity between carbon accounting activity data and carbon emission factors; Vocabulary similarity between carbon accounting activity data and carbon emission factors; 768-dimensional word vectors encoded from product lifecycle carbon accounting activity data using a pre-trained PCF-RWKV domain-specific language model; The 768-dimensional word vectors of entries in the carbon emission factor library after being encoded using the PCF-RWKV domain-specific language model; It is an L2 norm; The set of non-repeating valid words is obtained by extracting carbon accounting activity data text from the product lifecycle inventory after Chinese word segmentation, stop word filtering and special symbol cleaning; This is a set of non-repeating valid words extracted from the standardized entry text in the carbon emission factor library after Chinese word segmentation, stop word filtering, and special symbol cleaning.

[0041] S35, Select The carbon emission factor entry with the highest value is selected as the optimal matching carbon emission factor for the current carbon accounting activity data.

[0042] S36. Calculate the basic carbon footprint of products And break down carbon emissions from electricity consumption. : ; ; in, ; In the formula, For the first Carbon emissions at each stage of production; This represents the total number of carbon accounting activity data; For the first Carbon accounting data for each activity, such as raw material consumption and energy consumption, in units corresponding to the activity; For the first The carbon emission factor corresponding to each activity; Electricity consumption during the product manufacturing process; The average carbon emission factor of the regional power grid; This represents the total number of production stages.

[0043] Step S4 specifically includes the following steps: S41. Reduce carbon emissions from electricity consumption. Carbon emissions from the current product's electricity consumption .

[0044] S42. Allocate the company's total green certificate emission reductions to current products according to the proportion of electricity consumption: ; in, ; In the formula, The amount of green certificates that can be allocated to the current product for deduction; For the allocation ratio; This refers to the electricity consumption for the current product production. This represents the company's total annual electricity consumption.

[0045] S43. Introduce a tiered deduction cap and set differentiated deduction ratios: ; In the formula, This represents the effective deduction amount for green certificates.

[0046] As shown in the formula above, the first tier is as follows: when the deduction amount does not exceed 30% of the carbon emissions from electricity consumption, the full amount is deducted to incentivize enterprises to make reasonable use of green certificates; the second tier is as follows: when the deduction amount is between 30% and 60%, the portion exceeding 30% is deducted at 80% to moderately limit excessive deductions; the third tier is as follows: when the deduction amount exceeds 60%, the portion exceeding 60% is deducted at 50% to prevent enterprises from relying entirely on green certificates to reduce carbon emissions to zero without carrying out actual technological emission reductions.

[0047] S44. Based on the valid deduction amount of green certificates Dynamically adjust the carbon quotas of enterprises: ; In the formula, This refers to the adjusted remaining carbon allowance for enterprises.

[0048] S45. Subtract the effective deduction amount of green certificates from the basic carbon footprint to obtain the preliminary carbon footprint result considering green electricity and green certificate consumption: ; In the formula, This is the adjusted median value of the product's carbon footprint.

[0049] Step S5 specifically includes the following steps: S51. Extract all uncertainty parameters affecting carbon footprint accounting and the corresponding rules from the credibility-weighted carbon accounting rule base; among which, uncertainty parameters include carbon emission factors, carbon accounting activity data, and green certificate additionality coefficients. or ), Green Certificate-Carbon Two-Way Deduction Coefficient and Regional Green Electricity Marginal Carbon Emission Reduction Factor.

[0050] S52. Activate the corresponding rule for each uncertainty parameter, and then use the weighted Max-Min operator to synthesize the multi-source rules: ; In the formula, The resulting fuzzy probability distribution; For weighted maxima-minima composition operators; The activation rate of the calculation scenario is determined by the rules, and , , For the first The carbon accounting rule is in the 1st Gaussian similarity under each scene feature dimension. To calculate the total number of scenarios, , For the current scenarios where product carbon footprint accounting is to be carried out, in the first The actual feature values ​​under the feature dimensions of each accounting scenario. In the credibility-weighted carbon accounting rule base, the first The carbon accounting rules are pre-fixed, in the first Anchored standard feature values ​​under the feature dimensions of each accounting scenario. For the first The dimension sensitivity coefficient corresponding to the feature dimension of each accounting scenario; It is a natural exponential function.

[0051] S53. Map the fuzzy probability distribution to a probability distribution that can be used for Monte Carlo simulation.

[0052] S531. Extract the global confidence interval and filter tail noise: ; In the formula, For global confidence intervals; To be the smallest closed interval covering a specified set; This is the confidence level filtering threshold.

[0053] S532, Extract the 95% high confidence interval: ; In the formula, The optimal threshold; The maximum threshold required to satisfy the constraints; Let the Lebesgue measure be the interval; The length of the high-confidence interval; The length of the global confidence interval.

[0054] S533, Calculate the standard deviation of the distribution : ; In the formula, and These are the upper and lower limits of the 95% high confidence interval, respectively. It represents the 97.5th percentile of the standard normal distribution, with a value of 1.96, corresponding to a 95% confidence level.

[0055] S534. Constructing a truncated normal probability distribution : ; In the formula, is the probability density function of the standard normal distribution; The value of the uncertainty parameter; The probability density function is the standard normal distribution. This represents the mean of the uncertainty parameters.

[0056] S54, Monte Carlo simulation with three scenarios.

[0057] S541, High confidence interval The scenario is divided into three sub-scenarios with equal parameter value ranges: low, medium, and high. These sub-scenarios simulate the impact of different parameter ranges on carbon footprint. ; ; ; in, ; In the formula, , , These represent the sub-scene ranges for low, medium, and high parameter values, respectively. The parameter value is the length of the equal-width segment in the sub-scene.

[0058] S542. Calculate the weight of each parameter value for each sub-scene: ; In the formula, For the first The membership function of the sub-scene with each parameter value; For the first Each parameter takes a value representing the weight of a sub-scene, and satisfies the following conditions: ; For the first The membership function of each parameter value is a sub-scene.

[0059] S543. Perform conditional Monte Carlo simulations for each parameter value sub-scenario, sampling 10,000 times for each parameter value sub-scenario, to obtain the average carbon footprint for each parameter value sub-scenario. and variance .

[0060] S55. Based on Sobol global variance decomposition, the weighted synthesis and attribution of uncertainty is used to obtain the final carbon footprint by weighted fusion of the simulation results of the three-parameter sub-scenarios and to quantify the contribution of each parameter to the total uncertainty, so as to achieve accurate positioning of the source of uncertainty and interpretable output.

[0061] Step S55 specifically includes the following steps: S551, Weighted Synthetic Global Carbon Footprint Mean Total variance : ; ; S552. Calculate the 95% confidence interval. : ; S553. Uncertainty attribution calculation based on Sobol global variance decomposition.

[0062] S5531. Calculate the first-order contribution of uncertainty parameters. and second-order contribution : ; ; In the formula, For the first The first-order variance of the uncertain parameters, and , To fix all other uncertain parameters hour, The expected value of the condition; For only the first When an uncertain parameter changes conditional variance; For the first The uncertainty parameter and the first The second-order variance of the interaction of several uncertain parameters.

[0063] S5532, Calculate the total sensitivity index of uncertainty parameters : ; S554. Uncertainty Source Ranking and Attribution Report Generation: By Total Sensitivity Index The uncertain parameters are sorted from largest to smallest, the sources of uncertainty are labeled, and an uncertainty attribution report is generated, which includes the contribution of each parameter and the direction of data optimization.

[0064] Step S5 is followed by S6, which dynamically updates the credibility-weighted carbon accounting rule base, PCF-RWKV large model parameters and green certificate deduction coefficient based on the final accounting results and actual feedback data output by S5, and then updates them in reverse to steps S1, S2 and S3 to achieve closed-loop iterative optimization.

[0065] A system for calculating the carbon footprint of products that considers the consumption of green electricity and green certificates includes: The module for constructing a multi-source heterogeneous data acquisition and credibility-weighted carbon accounting rule base is used to build a multi-source heterogeneous standardized dataset and a credibility-weighted carbon accounting rule base.

[0066] The module for confirming the ownership of green electricity and green certificates and quantifying additional emission reductions is used to determine the additional emission reductions of green certificates and the green certificate-carbon two-way deduction coefficient based on multi-source heterogeneous standardized datasets. A lightweight, large-model-driven module for calculating the basic carbon footprint of a product's lifecycle is used to calculate the basic carbon footprint of a product and carbon emissions from electricity consumption based on multi-source heterogeneous standardized datasets.

[0067] The module for differentiated deduction of green electricity and green certificates and dynamic adjustment of carbon footprint is used to obtain the median value of carbon footprint that takes into account the consumption of green electricity and green certificates, based on the additional emission reduction of green certificates and the green certificate-carbon two-way deduction coefficient, combined with the basic carbon footprint and carbon emissions in the electricity consumption process, through tiered differentiated deduction adjustment.

[0068] The fuzzy-probability coupled carbon footprint uncertainty quantification and attribution module is used to synthesize fuzzy rules, perform possibility-probability mapping and three-scenario Monte Carlo simulation on intermediate carbon footprint values ​​based on a credibility-weighted carbon accounting rule base, and output a final product carbon footprint and uncertainty attribution report with a 95% confidence interval.

[0069] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for calculating the carbon footprint of products that considers the consumption of green electricity and green certificates, characterized in that: Includes the following steps: S1. Construct a multi-source heterogeneous standardized dataset and a credibility-weighted carbon accounting rule base; S2. Based on the multi-source heterogeneous standardized dataset, determine the additional emission reduction from green certificates and the green certificate-carbon two-way deduction coefficient; S3. Based on the multi-source heterogeneous standardized dataset, calculate the basic carbon footprint of the product and carbon emissions from the electricity consumption stage; S4. Based on the additional emission reduction from green certificates and the green certificate-carbon two-way deduction coefficient, combined with the basic carbon footprint and carbon emissions from the electricity consumption stage, obtain the median carbon footprint considering green electricity and green certificate consumption through tiered differentiated deduction adjustments; S5. Based on the credibility-weighted carbon accounting rule base, perform fuzzy rule synthesis, possibility-probability mapping, and three-scenario Monte Carlo simulation on the median carbon footprint, and output the final product carbon footprint and uncertainty attribution report with a 95% confidence interval.

2. The product carbon footprint accounting method considering green electricity and green certificate consumption according to claim 1, characterized in that: Step S1 specifically includes the following steps: S11. Collect multi-source heterogeneous raw data; multi-source heterogeneous raw data includes product-side data, green certificate-side data, grid-side data, domain evidence data, and market-side data; among them, product-side data includes production process description texts of the electricity products to be accounted for, raw material consumption records, energy consumption records, and production process parameters; green certificate-side data is the transaction ledger of green electricity and green certificates, which records the transaction time, transaction entity, the installed capacity type of green electricity corresponding to the green certificate, and the transaction binding relationship; grid-side data includes regional grid power generation structure data, carbon emission factors of various types of power sources, and transmission and distribution loss data; domain evidence data includes carbon accounting-related standards, industry reports, academic literature, and enterprise accounting cases; market-side data includes carbon market quota data, transaction prices, green certificate market transaction prices, and supply and demand data; S12. Multi-source data standardization preprocessing: The multi-source heterogeneous raw data collected in S11 are sequentially processed for outlier removal, missing value filling, unit unification, and text vectorization. S13, Distillation of multi-source evidence; S131. Construct domain prompts: Construct a Few-Shot prompt template specifically for carbon accounting, embed a standard example of carbon accounting rule extraction in the prompt, and explicitly require that the rules extracted by the large model must include the applicable industry of the rule, the name of the rule parameter, the range of parameter values ​​and the rule description. At the same time, it automatically marks the source type of the current evidence as the source attribute label of the rule. S132. Construct and deploy the PCF-RWKV domain-specific language model. The PCF-RWKV domain-specific language model includes a domain input layer, an RWKV basic encoding layer, a LoRA domain adaptation layer, and a multi-task output layer arranged sequentially. The domain input layer loads a carbon computation-specific enhanced vocabulary, performs terminology alignment and embedding encoding on the vectorized text output from S12, converting the text into a 768-dimensional fixed-length embedding vector sequence, which is then input into the subsequent RWKV basic encoding layer. The RWKV basic encoding layer performs linear temporal feature extraction on the input embedding vectors. The LoRA domain adaptation layer, while freezing the basic weights, trains only the low-rank update matrix to achieve domain knowledge injection, and the weight update formula is: ； In the formula, Weights are applied after domain adaptation; These are the original weights for pre-training; This is the scaling factor; It is a low-rank dimension; , All are low-rank adaptation matrices; The multi-task output layer is equipped with a dedicated output head for evidence distillation, used to output structured carbon accounting rules; S133. Perform batch rule extraction of multi-source evidence based on PCF-RWKV domain-specific language model: Input the domain prompt words generated in S131 and the domain evidence text preprocessed in S12 into PCF-RWKV domain-specific language model to extract four original rules in batch: accounting boundary rule, carbon emission factor value rule, green certificate deduction application rule and carbon quota linkage rule. Each carbon accounting rule carries industry type, parameter name, value range and source type attributes, and outputs the original carbon accounting rule set. S134. Fuzzy Rule Standardization Transformation: For the extracted original rules, the carbon accounting uncertainty parameters corresponding to the value range attributes are converted into triangular fuzzy membership degree format: ； In the formula, For the first Membership degree of the carbon accounting rule; , , All are membership parameters; The actual values ​​of the uncertainty parameters in carbon accounting; S135. Rule deduplication and conflict pre-validation: Calculate cosine similarity based on rule semantic vectors and merge duplicate rules with cosine similarity greater than 0.95; mark conflicting rules with the same parameter intersecting intervals and retain conflict attributes; S136. Output the initial carbon accounting rule set and source attribute tags: Use the deduplicated and verified fuzzy rules as the initial carbon accounting rule set, and output the source type tag corresponding to each carbon accounting rule simultaneously. Input them into S14 to perform confidence weighting. S14. Calculate the credibility weight of rules based on the entropy weight method and construct a credibility-weighted carbon accounting rule base; S141. Construct a rule credibility evaluation matrix: Select three dimensions—source authority, rule consistency, and scenario coverage—to score the initial carbon accounting rule set output in S13 and construct an evaluation matrix. : ； in, ； ； ； In the formula, For the first The source of the carbon accounting rules is authoritative and standardized scoring; For the first The source type of the carbon accounting rule and its original score, and ; The highest score for the source type, and ; For the first Standardized scoring of rule consistency for carbon accounting rules; This represents the total number of rules applicable to the same accounting parameter. It is the intersection interval of all intervals with the same rule; It is the union of all intervals with the same rules; This is a function for calculating the length of an interval. For the first Standardized scoring of scenario coverage for carbon accounting rules; For the first The carbon accounting rules specify the total number of applicable accounting scenarios; The maximum number of accounting scenarios covered in all rules, including industry type, product type, green certificate type, regional power grid type, and accounting boundary; S142. Standardize the evaluation matrix using the positive range of maximum and minimum values, and then calculate the dimensional weights: ； in, ； ； In the formula, For the first Dimension weights These respectively represent source authority, rule consistency, and scenario coverage; For the first Dimensional information entropy; For the first Rule No. 1 of the carbon accounting rules The proportion of dimensions; For the first Rule No. 1 of the carbon accounting rules Standardized scores for each dimension; The total number of carbon accounting rules; S143. Calculate the overall credibility weight of a single rule: ； In the formula, For the first The overall credibility weight of the carbon accounting rules; S144. Construct a credibility-weighted carbon accounting rule base: Associate each carbon accounting rule with its corresponding credibility weight. Membership parameters are bound together to form a credibility-weighted carbon accounting rule base; S15. Output multi-source heterogeneous standardized datasets and a credibility-weighted carbon accounting rule base.

3. The product carbon footprint accounting method considering green electricity and green certificate consumption according to claim 2, characterized in that: In step S132, the temporal coding update expression of the RWKV base coding layer is as follows: ； ； In the formula, for Time-bound key vector; for Time value vector; , All are learnable weight matrices of the RWKV basic coding layer; , All are timing-gated parameters; for Input the embedding vector at each time step; for Input the embedding vector at each time step; For Hadama accumulation.

4. The product carbon footprint accounting method considering green electricity and green certificate consumption according to claim 3, characterized in that: Step S2 specifically includes the following steps: S21. Identify the type of green certificate transaction based on the binding relationship of green certificate transactions: If the buyer of green electricity also obtains the corresponding green certificate in the green certificate transaction data, and the green electricity consumption record corresponds one-to-one with the green certificate transaction record, then it is marked as a certificate-electricity integrated green certificate; if the green certificate and the corresponding green electricity are traded separately, then it is marked as a certificate-electricity separated green certificate. S22, Calculate the marginal carbon reduction factor of green electricity in the region : ； In the formula, The carbon emission factor for marginal thermal power units within the region; This corresponds to the carbon emission factor throughout the entire life cycle of green electricity; S23. Calculate the additional emission reductions from green certificates. : ； In the formula, The trading volume of green certificates; Additional coefficient for green certificates that integrate electricity and energy; Additional coefficient for verifying the electrical separation type of green certificate; S24. Calculate the green certificate-carbon two-way deduction factor. : ； in, ； In the formula, The company's initial remaining carbon allowance; The total annual carbon allowance allocated to the enterprise; This represents the company's annual baseline carbon emissions.

5. The product carbon footprint accounting method considering green electricity and green certificate consumption according to claim 4, characterized in that: Step S3 specifically includes the following steps: S31. Semantic parsing of product manufacturing process: Using the pre-trained PCF-RWKV domain-specific language model, semantic parsing is performed on the text describing the manufacturing process to extract semantic features; S32. Multi-Agent Task Decomposition and Collaboration: Based on the Autogen framework, the product carbon footprint accounting task is decomposed into multiple sub-tasks and assigned to different dedicated agents: LCI generation agent: responsible for the generation of lifecycle phase divisions; Raw material accounting agent: responsible for extracting carbon accounting activity data in the raw material stage; Energy consumption accounting agent: responsible for extracting carbon accounting activity data in the energy consumption stage; Carbon emission factor matching agent: responsible for matching carbon emission factors. S33. Input the subtask list generated in S32 into the PCF-RWKV domain-specific language model to generate a product lifecycle list; S34. Extract all carbon emission factor-related rules from the completed credibility-weighted carbon accounting rule library, standardize them into carbon emission factor entries according to industry and scenario, and construct a structured carbon emission factor library. Then, for each carbon accounting activity data in the S33 lifecycle inventory, the corresponding carbon emission factor is matched from the carbon emission factor library through semantic matching, and the matching score is calculated. The calculation formula is as follows: ； in, ； ； In the formula, The semantic similarity between carbon accounting activity data and carbon emission factors; Vocabulary similarity between carbon accounting activity data and carbon emission factors; 768-dimensional word vectors encoded from product lifecycle carbon accounting activity data using a pre-trained PCF-RWKV domain-specific language model; The 768-dimensional word vectors of entries in the carbon emission factor library after being encoded using the PCF-RWKV domain-specific language model; It is an L2 norm; The set of non-repeating valid words is obtained by extracting carbon accounting activity data text from the product lifecycle inventory after Chinese word segmentation, stop word filtering and special symbol cleaning; This is a set of non-repeating valid words extracted from the standardized entry text in the carbon emission factor library after Chinese word segmentation, stop word filtering, and special symbol cleaning. S35, Select The carbon emission factor entry with the highest value is selected as the optimal matching carbon emission factor for the current carbon accounting activity data. S36. Calculate the basic carbon footprint of products And break down carbon emissions from electricity consumption. : ； ； in, ； In the formula, For the first Carbon emissions at each stage of production; This represents the total number of carbon accounting activity data; For the first Carbon accounting activity data for each activity; For the first The carbon emission factor corresponding to each activity; Electricity consumption during the product manufacturing process; The average carbon emission factor of the regional power grid; This represents the total number of production stages.

6. The product carbon footprint accounting method considering green electricity and green certificate consumption according to claim 5, characterized in that: Step S4 Specifically, the following steps are included: S41. Reduce carbon emissions from electricity consumption. Carbon emissions from the current product's electricity consumption ; S42. Allocate the company's total green certificate emission reductions to current products according to the proportion of electricity consumption: ； in, ； In the formula, The amount of green certificates that can be allocated to the current product for deduction; For the allocation ratio; This represents the electricity consumption for the current product production. This refers to the company's total annual electricity consumption. S43. Introduce a tiered deduction cap and set differentiated deduction ratios: ； In the formula, The effective deduction amount for green certificates; S44. Based on the valid deduction amount of green certificates Dynamically adjust the carbon quotas of enterprises: ； In the formula, This refers to the adjusted remaining carbon allowance for enterprises. S45. Subtract the effective deduction amount of green certificates from the basic carbon footprint to obtain the preliminary carbon footprint result considering green electricity and green certificate consumption: ； In the formula, This is the adjusted median value of the product's carbon footprint.

7. The product carbon footprint accounting method considering green electricity and green certificate consumption according to claim 6, characterized in that: Step S5 specifically includes the following steps: S51. Extract all uncertain parameters affecting carbon footprint accounting and the corresponding rules in the credibility-weighted carbon accounting rule base; among which, uncertain parameters include carbon emission factors, carbon accounting activity data, green certificate additionality coefficient, green certificate-carbon two-way deduction coefficient, and regional green electricity marginal carbon emission reduction factor. S52. Activate the corresponding rule for each uncertainty parameter, and then use the weighted Max-Min operator to synthesize the multi-source rules: ； In the formula, The resulting fuzzy probability distribution; For weighted maxima-minima composition operators; The activation rate of the calculation scenario is determined by the rules, and , , For the first The carbon accounting rule is in the 1st Gaussian similarity under each scene feature dimension. To calculate the total number of scenarios, , For the current scenarios where product carbon footprint accounting is to be carried out, in the first The actual feature values ​​under the feature dimensions of each accounting scenario. In the credibility-weighted carbon accounting rule base, the first The carbon accounting rules are pre-fixed and in the first Anchored standard feature values ​​under the feature dimensions of each accounting scenario. For the first The dimension sensitivity coefficient corresponding to the feature dimension of each accounting scenario; It is a natural exponential function; S53. Map the fuzzy probability distribution to a probability distribution that can be used for Monte Carlo simulation; S531. Extract the global confidence interval and filter tail noise: ； In the formula, For global confidence intervals; To be the smallest closed interval covering a specified set; The confidence level filtering threshold; S532, Extract the 95% high confidence interval: ； In the formula, The optimal threshold; The maximum threshold required to satisfy the constraints; Let the Lebesgue measure be the interval; The length of the high-confidence interval; The length of the globally trusted interval; S533, Calculate the standard deviation of the distribution : ； In the formula, and These are the upper and lower limits of the 95% high confidence interval, respectively. It is the 97.5th percentile of the standard normal distribution; S534. Constructing a truncated normal probability distribution : ； In the formula, is the probability density function of the standard normal distribution; The value of the uncertainty parameter; is the probability density function of the standard normal distribution; The mean of the uncertainty parameter; S54, Monte Carlo simulation under three scenarios; S541, High confidence interval The scenario is divided into three sub-scenarios with equal parameter value ranges: low, medium, and high. These sub-scenarios simulate the impact of different parameter ranges on carbon footprint. ； ； ； in, ； In the formula, , , These represent the sub-scene ranges for low, medium, and high parameter values, respectively. The parameter value is the length of the equal-width segment in the sub-scene; S542. Calculate the weight of each parameter value for each sub-scene: ； In the formula, For the first The membership function of the sub-scene with each parameter value; For the first Each parameter takes a value representing the weight of a sub-scene, and satisfies the following conditions: ; For the first The membership function of the sub-scene with each parameter value; S543. Perform conditional Monte Carlo simulations for each parameter value sub-scenario, sampling 10,000 times for each parameter value sub-scenario, to obtain the average carbon footprint for each parameter value sub-scenario. and variance ; S55. Based on Sobol global variance decomposition, the uncertainty weighted synthesis and attribution are used to obtain the final carbon footprint by weighted fusion of the simulation results of the three-parameter sub-scenarios and to quantify the contribution of each parameter to the total uncertainty.

8. The product carbon footprint accounting method considering green electricity and green certificate consumption according to claim 7, characterized in that: Step S55 specifically includes the following steps: S551, Weighted Synthetic Global Carbon Footprint Mean Total variance : ； ； S552. Calculate the 95% confidence interval. : ； S553, Uncertainty attribution calculation based on Sobol global variance decomposition; S5531. Calculate the first-order contribution of uncertainty parameters. and second-order contribution : ； ； In the formula, For the first The first-order variance of the uncertain parameters, and , To fix all other uncertain parameters hour, The expected value of the condition; For only the first When an uncertain parameter changes conditional variance; For the first The uncertainty parameter and the first The second-order variance of the interaction of several uncertain parameters; S5532, Calculate the total sensitivity index of uncertainty parameters : ； S554. Uncertainty Source Ranking and Attribution Report Generation: By Total Sensitivity Index The uncertain parameters are sorted from largest to smallest, the sources of uncertainty are labeled, and an uncertainty attribution report is generated, which includes the contribution of each parameter and the direction of data optimization.

9. A product carbon footprint accounting method considering green electricity and green certificate consumption according to claim 8, characterized in that: Step S5 is followed by S6, which dynamically updates the credibility-weighted carbon accounting rule base, PCF-RWKV large model parameters and green certificate deduction coefficient based on the final accounting results and actual feedback data output by S5, and then updates them in reverse to steps S1, S2 and S3 to achieve closed-loop iterative optimization.

10. A system for calculating the carbon footprint of a product considering green electricity and green certificate consumption as described in any one of claims 1-9, characterized in that: include: A module for constructing a multi-source heterogeneous data acquisition and credibility-weighted carbon accounting rule base is used to build a multi-source heterogeneous standardized dataset and a credibility-weighted carbon accounting rule base. The module for confirming the ownership of green electricity and green certificates and quantifying additional emission reductions is used to determine the additional emission reductions of green certificates and the green certificate-carbon two-way deduction coefficient based on multi-source heterogeneous standardized datasets. A lightweight, large-model-driven module for calculating the basic carbon footprint of a product lifecycle is used to calculate the basic carbon footprint of a product and carbon emissions from electricity consumption based on a multi-source heterogeneous standardized dataset. The Green Electricity and Green Certificate Differentiated Deduction and Carbon Footprint Dynamic Adjustment Module is used to obtain the median carbon footprint considering green electricity and green certificate consumption by adjusting the tiered differentiated deduction based on the additional emission reduction of green certificates and the green certificate-carbon two-way deduction coefficient, combined with the basic carbon footprint and carbon emissions in the electricity consumption process. The fuzzy-probability coupled carbon footprint uncertainty quantification and attribution module is used to synthesize fuzzy rules, perform possibility-probability mapping and three-scenario Monte Carlo simulation on intermediate carbon footprint values ​​based on a credibility-weighted carbon accounting rule base, and output a final product carbon footprint and uncertainty attribution report with a 95% confidence interval.

Images