Carbon boundary adjustment carbon footprint accounting factor correction method based on Copula model

Abstract

The invention discloses a Copula model-based carbon boundary adjustment carbon footprint accounting factor correction method, which comprises the following steps of: obtaining a carbon footprint accounting sample set and a high-dimensional emission driving variable set of a target product, and recording a data quality score for each variable; marginal distribution fitting is carried out on the high-dimensional emission drive variable set dimension by dimension, normalization is carried out according to a unified quantile mapping rule, and marginal distribution parameters and normalized drive variable samples are obtained; constructing a Vine Copula dependency structure on the basis of the normalized drive variable sample, determining a root node sequence and an edge connection set according to a structure scoring criterion, obtaining the dependency strength of a variable pair, generating a dependency contribution weight in combination with a data quality score, and reducing the weight of the variable of which the dependency strength is lower than a preset dependency threshold value according to a preset attenuation rule; and performing weighting correction on the emission reference quantity according to the dependency contribution weight to generate a carbon footprint accounting correction factor and outputting a correction result. The method improves the stability of the correction factor under the participation of the high-dimensional variable.

Description

Technical Field

[0001] This invention relates to the technical field of product carbon footprint accounting and high-dimensional statistical correction for carbon border adjustment applications, and particularly to a carbon border adjustment carbon footprint accounting factor correction method based on the Copula model. Background Technology

[0002] As the global carbon governance system gradually shifts from an advocacy framework to binding rules, product carbon footprint accounting has evolved from an internal management tool for enterprises to a key evidentiary basis for cross-border trade compliance. Especially under the background of carbon border adjustment mechanisms, exported products need to be linked to multi-source emission driving information such as electricity, fuel, raw materials, transportation, and process efficiency under specified criteria to support the comparability assessment of emission intensity and the unified calibration of accounting factors. Existing product carbon footprint accounting methods generally present an engineered path of data collection, boundary delineation, emission calculation, and result output. However, in scenarios where high-dimensional variables coexist across industries, regions, and supply chains, factors such as differences in variable dimensions, statistical distribution differences, and uneven quality of source data can amplify the fluctuations of accounting factors. At the same time, driving variables often have nonlinear correlations and conditional dependencies. Without interpretable modeling of high-dimensional dependency structures and stable constraints on variable weights, systematic biases due to the omission of key variables or noise amplification due to the superposition of redundant variables can easily occur, thus affecting the repeatability and audit consistency of accounting conclusions under carbon border adjustment scenarios.

[0003] CN116911490A discloses a method for determining the carbon footprint of cables. This method organizes life cycle data and emission calculation processes for specific product objects to support the determination and output of carbon footprint at the product level. However, its technical focus is biased towards the construction of the carbon footprint business chain and the data collection approach. It does not establish an interpretable statistical structure based on the complex dependencies between high-dimensional emission driving variables, nor does it provide a hierarchical constraint and weight correction mechanism for the quality differences of variables from different sources. As a result, the stability of the accounting factor is still limited in scenarios where multiple source variables participate and change with the time window.

[0004] Therefore, existing carbon footprint determination technologies and general high-dimensional dependency modeling technologies suffer from two problems: a lack of stable correction mechanisms for accounting factors under high-dimensional dependency constraints and a lack of variable organization and quality constraints oriented towards the business context of carbon footprint. This invention addresses the problem of accounting factor fluctuations caused by the complex dependencies of high-dimensional emission driving variables, uneven data quality, and variable redundancy / omission in carbon border adjustment scenarios. It proposes a carbon footprint accounting factor correction method based on the Copula model. This method aligns high-dimensional emission driving variables within the same accounting time window and introduces data quality scoring. It then obtains normalized driving variable samples by combining dimensional marginal distribution fitting and unified quantile mapping. Based on the Vine Copula dependency structure and structure scoring criteria, it generates variable pair dependency strength and dependency contribution weights, sets graded attenuation constraints for weakly correlated variables, and finally performs weighted correction on the emission baseline to generate the accounting correction factor. Summary of the Invention

[0005] The purpose of this section is to outline some aspects of the embodiments of the present invention and to briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section, as well as in the abstract and title of the present application, to avoid obscuring the purpose of this section, the abstract and title of the invention. Such simplifications or omissions shall not be used to limit the scope of the present invention.

[0006] In view of the aforementioned existing problems, the present invention is proposed.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution: As a preferred embodiment of the carbon footprint accounting factor correction method based on the Copula model described in this invention, the carbon footprint accounting sample set of the target product is obtained, each sample includes an emission baseline quantity and a set of high-dimensional emission driving variables aligned with the emission baseline quantity within the same accounting time window, and a corresponding data quality score is recorded for each variable in the set of high-dimensional emission driving variables. The high-dimensional emission driving variable set is fitted with marginal distributions dimension by dimension and normalized according to a unified quantile mapping rule to obtain marginal distribution parameters and normalized driving variable samples. Based on the normalized driving variable samples, a Vine Copula dependency structure is constructed, and the root node order and edge connection set are determined according to the structure scoring criteria to obtain the dependency strength of variable pairs. The dependency contribution weights corresponding to each driving variable are generated by combining the data quality score. The weights of variables whose dependency strength of variable pairs is lower than the preset dependency threshold are reduced according to the preset decay rule. The emission baseline is weighted and corrected according to the dependency contribution weight to generate a carbon footprint accounting correction factor. The carbon footprint accounting correction factor, the marginal distribution parameter, the Vine Copula dependency structure, and the dependency contribution weight are used together as the correction result output.

[0008] The beneficial effects of this invention are as follows: By aligning the emission baseline and the set of high-dimensional emission driving variables within the same accounting time window and introducing a data quality score, this invention reduces sample bias caused by missing data, time misalignment, and inconsistent sources. Furthermore, by fitting dimensionally marginal distributions and using unified quantile mapping to obtain normalized driving variable samples, it reduces the interference of different dimensions and statistical forms on high-dimensional correlation analysis. The Vine Copula structural score is used to determine the root node order and edge connection set, obtaining the dependency strength of variable pairs. Combined with the data quality score, dependency contribution weights are generated, and graded attenuation constraints are applied to weakly dependent variables. This ensures that the weight allocation simultaneously reflects the strength of variable correlation and the reliability of the data, thereby suppressing noise amplification caused by redundant variables and mitigating bias caused by missing key variables. Finally, the emission baseline is weighted and corrected based on the dependency contribution weights to obtain a carbon footprint accounting correction factor, enabling the correction factor to maintain higher stability under conditions of cross-batch, cross-enterprise, and variable combination changes. Attached Figure Description

[0009] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a schematic flowchart of the carbon footprint accounting factor correction method based on the Copula model shown in this invention. Detailed Implementation

[0010] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0011] Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort should fall within the scope of protection of this invention.

[0012] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0013] According to an embodiment of the present invention, in combination Figure 1 The flowchart shown illustrates a method for correcting carbon footprint accounting factors based on the Copula model, specifically including the following steps: S1. Obtain the carbon footprint accounting sample set for the target product. Each sample includes an emission baseline and a set of high-dimensional emission driving variables aligned with the emission baseline within the same accounting time window. Record the corresponding data quality score for each variable in the high-dimensional emission driving variable set. Note that the following should be noted in this step: S1.1 Determine the accounting boundaries corresponding to the target product, set the accounting time window, and generate a unified time index rule based on the accounting time window; As an example, the target product can be an industrial product with stable batch codes and energy consumption metering boundaries, such as hot-rolled coils, aluminum alloy profiles, or cement clinker.

[0014] As an example, the accounting boundaries can be determined based on the enterprise's internal process codes and supply chain document scope, selecting raw material entry, key process segments, and finished product exit as boundary segments, and including the metering systems corresponding to the boundary segments (such as electricity meters, flow meters, fuel meters, raw material warehousing weighing, and transportation documents) in the same boundary list.

[0015] As an example, the accounting time window can be set using a daily cycle.

[0016] For example, for the continuous hot rolling process of steel, a 24-hour period is set as a calculation time window; for the discrete batch aluminum alloy casting process, the start and end times of a single furnace are set as a calculation time window; when generating time index rules based on the calculation time window, a composite index method of time window number + batch number is adopted.

[0017] For example, the time window number is set as ,in Each number corresponds to a continuous time interval. Record the target product batch number as ,in Thus forming a composite index key. This serves as a unified identifier for subsequent alignment, merging, and matching.

[0018] S1.2 Collect the emission baseline quantity corresponding to the target product batch within the accounting time window, and record the time index corresponding to the accounting time window for the emission baseline quantity; It should be noted that the emission baseline is derived from the enterprise's carbon accounting system. The emission baseline includes at least the equivalent amounts of electricity consumption, fuel consumption, and raw material process emissions corresponding to the accounting boundary. When collecting the emission baseline, the original metering records are sliced ​​according to the time window boundary in the time index rule. The metering records falling into the accounting time window are aggregated into the emission baseline of that time window, and the emission baseline is marked as the time index corresponding to the accounting time window.

[0019] For example, the energy consumption and emissions converted from 00:00 to 24:00 on January 1, 2025 are aggregated and recorded as follows: The emission baseline was established and linked to the production ledgers for each batch on that day.

[0020] S1.3 Simultaneously, within the accounting time window, collect variables such as electricity carbon intensity, fuel structure, raw material carbon intensity, transportation intensity, and key process efficiency, and record the time index corresponding to the accounting time window for each variable to form a set of high-dimensional emission driving variables. It should be noted that the electricity carbon intensity variable is derived from the enterprise's green electricity certificate conversion factor; the fuel structure variable is formed by the proportion vector of "coal / natural gas / coke oven gas / biomass" from fuel metering and procurement ledgers; the raw material carbon intensity variable is formed by the supplier's carbon footprint declaration, raw material inspection report and historical accounting database; the transportation intensity variable is converted from transportation mileage, load and transportation mode coding; and the key process efficiency variable is derived from the process control system, such as indicators like "energy consumption per unit output, comprehensive efficiency of key equipment, and recovery heat utilization rate".

[0021] Specifically, when collecting the above variables synchronously within the accounting time window, time window identifiers are added to the original records of each variable according to the time indexing rules, so that the electricity carbon intensity variable, fuel structure variable, raw material carbon intensity variable, transportation intensity variable and key process efficiency variable are classified into the high-dimensional emission driving variable set with a consistent time index.

[0022] S1.4. According to the time indexing rules, the emission baseline quantity and the set of high-dimensional emission driving variables are time-aligned to obtain the aligned data fragments corresponding to the accounting time window; It should be noted that time alignment adopts the order of index matching + gap filling. In this embodiment, according to the time indexing rules, the emission baseline quantity and the set of high-dimensional emission driving variables are indexed and matched with the same time window number. When there are differences in sampling frequency within the same window, the high-frequency variables are aggregated into time window-level variables according to the statistical values ​​(such as mean, weighted mean, or last value) within the time window. When there are local time misalignments that can be aligned according to the time indexing rules, they are merged according to the boundaries of adjacent time windows. This yields aligned data segments corresponding to the accounting time window. Each aligned data segment contains at least the time window-level values ​​of the emission baseline quantity and the set of high-dimensional emission driving variables.

[0023] S1.5 Merge the aligned data fragments according to the target product batch to generate a carbon footprint accounting sample set for the target product, so that each sample contains the emission baseline and a set of high-dimensional emission driving variables aligned with the emission baseline within the same accounting time window. It should be noted that, based on the production ledger of the target product batch, aligned data segments corresponding to the same batch across multiple accounting time windows are aggregated by batch; for example, for batches... exist The aligned data fragments in the sample are merged to form a single sample that simultaneously contains: and The corresponding emission baseline sequence or aggregate value; a set of high-dimensional emission driving variables aligned with the emission baseline within each accounting time window; thereby generating a carbon footprint accounting sample set for the target product.

[0024] S1.6. Based on the accounting time window and time index rules, establish a data quality scoring rule set. The data quality scoring rule set should at least include missing state rules, time consistency rules, and source consistency rules, and set a corresponding scoring level table for each rule; among which: (1) Missing state rules Let variables The number of sampling plan points within the accounting time window is The actual number of valid sampling points is The missing rate is: in, For variables The missing rate; For variables The planned number of sampling points within the accounting time window; For variables The number of valid sampling points within the accounting time window; when At that time, the missing level is high. when Furthermore, if the missing fragment does not cross a critical process segment, the missing level is medium. when If the missing segment covers a critical process segment, the missing level is low.

[0025] (2) Time Consistency Rules Let variables The number of matching points between the time index sequence and the time index rule is The time matching rate is: in, For variables Time matching rate; For variables The number of sampling points whose time index matches the time index rule; when At that time, the time consistency level was high. when Furthermore, when misaligned points can be merged into adjacent time windows according to time indexing rules, the time consistency level is medium. when If there is a misalignment across accounting time windows, the time consistency level is low.

[0026] (3) Source consistency rule Let variables be The three relevant sources of evidence are: measurement records of the same batch. Production ledger records Supply chain documentation records ; Define consistency count: in, For variables Source consistency count; For indicator functions, when the variable If the value and time match the corresponding source of evidence, take 1; otherwise, take 0. This indicates the correspondence between the numerical value and the time. For variables Measurement records of the same batch; For variables Production ledger records; For variables Supply chain documentation records; when At that time, the source consistency level is high; when At that time, the source consistency level was medium. when If any type of evidence source is missing, making it impossible to complete a consistency determination, the source consistency level is low.

[0027] For example, the three levels are mapped to numerical scores: high level is 1, medium level is 0.6, and low level is 0.2; let the missing status score be... The time consistency score is The source consistency score is The data quality score is then: in, For variables The corresponding data quality score; The weight for the missing status score, for example, a value of 0.4; This is used as the weight for the time consistency score, for example, a value of 0.3; This is the weight for the consistency score based on the source, for example, a value of 0.3; For variables The missing state numerical score; For variables Time consistency numerical score; For variables The consistency score of the source.

[0028] S2. Fit marginal distributions to the high-dimensional emission driver variable set dimension by dimension and normalize them according to a unified quantile mapping rule to obtain marginal distribution parameters and normalized driver variable samples. Note that the following should be noted in this step: S2.1 Extract the dimensional observation data of the high-dimensional emission driving variable set from the carbon footprint accounting sample set, and divide the dimensional observation data into segments according to the accounting time window to obtain the time window sample sequence corresponding to each variable; Specifically, when extracting dimension-by-dimensional observation data from the carbon footprint accounting sample set, the accounting time window is used as the segmentation benchmark. The values ​​of the same variable in different accounting time windows are arranged in chronological order to obtain the time window sample sequence corresponding to each variable; for example, the electricity carbon intensity variable is denoted as... Then according to the time window to Extract and organize to obtain the sequence .

[0029] S2.2. Perform marginal distribution fitting on the time window sample sequences to obtain the marginal distribution parameters that correspond one-to-one with each variable; It should be noted that for each variable in the set of high-dimensional emission driving variables, the time window sample sequence corresponding to the variable is extracted and arranged according to the accounting time window order to obtain the marginal fitting sample sequence of the variable; the marginal fitting sample sequence of each variable is fitted to the candidate distribution according to the preset marginal distribution family to obtain the fitting result corresponding to each candidate distribution; the marginal distribution type corresponding to each variable is determined from the fitting result according to the preset goodness of fit evaluation rule, and the marginal distribution parameters corresponding to each variable are output.

[0030] In a preferred embodiment, the preset marginal distribution family includes at least the normal distribution, log-normal distribution, Gamma distribution, Weibull distribution, t distribution, and other distribution families that match the statistical characteristics of variables such as energy consumption, intensity, and efficiency; when fitting candidate distributions to the marginal fitting sample sequences of each variable, the parameter estimation results and goodness-of-fit indices of the candidate distributions are calculated respectively; and the preset goodness-of-fit evaluation rule adopts a combination of AIC, BIC, and KS test.

[0031] For example, the minimum BIC is first used as the main ranking index; when the difference between the BICs of the two candidate distributions is less than the preset difference threshold, the KS statistic is further compared; if the KS both pass the significance threshold, the distribution type with more stable parameter interpretability and monotonic quantile function is selected first; thus, the marginal distribution type corresponding to each variable is determined, and the marginal distribution parameters corresponding to each variable are output.

[0032] S2.3. Construct the quantile mapping relationship of each variable based on the marginal distribution parameters, and apply unified rule constraints to the quantile mapping relationship of each variable to obtain the unified quantile mapping rule; Specifically, based on the one-to-one marginal distribution parameters of each variable, quantile mapping relationships for each variable are generated, and the quantile mapping relationships map the time window sample sequence of the variable to a preset quantile set; uniform rule constraints are applied to the quantile mapping relationships of each variable, and the uniform rule constraints include at least a uniform quantile set, a uniform boundary quantile processing rule, and a uniform interpolation rule; the quantile mapping relationships of each variable that satisfy the uniform rule constraints are summarized into a uniform quantile mapping rule, which stipulates that each variable is mapped to the same quantile set, and samples exceeding the boundary quantiles are processed using consistent boundary merging.

[0033] As an example, the preset quantile set adopts The set of quantiles is unified, and the boundary quantile processing rules adopt the order of boundary truncation + endpoint merging: When the mapped quantile of a sample is lower than the minimum quantile, its mapped value is merged into the value corresponding to the minimum quantile. When the mapped quantile of a sample is higher than the maximum quantile, its mapped value is merged into the value corresponding to the maximum quantile.

[0034] Furthermore, the unified interpolation rule adopts linear interpolation, and monotonic spline interpolation is used as an alternative rule for variables with strong skewness or heavy-tailed characteristics, so as to ensure the monotonicity of quantile mapping and consistency across variables.

[0035] S2.4. Normalize the time window sample sequences of each variable according to the unified quantile mapping rule to obtain the normalized driving variable samples corresponding to the accounting time window.

[0036] It should be noted that when normalizing the time window sample sequences of each variable according to the unified quantile mapping rule, the cumulative distribution position of each sample under its marginal distribution is first calculated, and then the cumulative distribution position is mapped to the corresponding interval of the unified quantile set.

[0037] For example, extreme high-value samples of raw material carbon intensity variables are merged into normalized values ​​corresponding to the 0.99 quantile according to a unified boundary quantile processing rule, thereby forming normalized driving variable samples corresponding to the accounting time window.

[0038] S3. Construct a Vine Copula dependency structure based on normalized driving variable samples and determine the root node order and edge connection set according to the structure scoring criteria to obtain the dependency strength of variable pairs. Combine this with the data quality score to generate dependency contribution weights corresponding to each driving variable. Reduce the weights of variables with dependency strength below a preset dependency threshold according to a preset decay rule. Note that the following points should be noted in this step: S3.1 Based on the normalized driving variable samples, extract candidate variable pairs from the high-dimensional emission driving variable set and aggregate them according to the corresponding relationship of the accounting time window to form a candidate variable pair sample sequence; It should be noted that when generating candidate variable pair combinations based on normalized driving variable samples, the high-dimensional emission driving variable set can be combined pairwise to obtain the candidate variable pair set. Subsequently, according to the corresponding relationship of the accounting time window, the normalized values ​​of each candidate variable pair within the same time window are paired to form a candidate variable pair sample sequence.

[0039] S3.2 Calculate the dependency of each candidate variable pair on the sample sequence to obtain a set of candidate dependencies that correspond one-to-one with each candidate variable pair. For example, candidate dependencies are adopted As a nonparametric dependency metric: in, For variables With variables The candidate dependency; n is the number of samples for the candidate variable in the sample sequence; Normalization driving variables in the sample The A normalized value; Normalization driving variables in the sample The Each normalized value is used to obtain a set of candidate dependencies that correspond one-to-one with each candidate variable pair.

[0040] S3.3. Based on the structural scoring criteria, the candidate dependency set is structurally scored to determine the root node order and edge connection set of the Vine Copula dependency structure, thus obtaining the Vine Copula dependency structure. For example, the structural scoring criterion uses a combination of dependency strength and complexity penalty: in, Score the structure of candidate Vine structure G; This is the set of edges connected in the Vine structure; For the edge The structural edge weights; Candidate dependency; This is the complexity penalty coefficient; The number of edge connections is determined by traversing the candidate root node order and the edge connection set. The largest structure determines the root node order and edge connection set of the Vine Copula dependency structure.

[0041] S3.4 On the edge connection set, estimate the dependency parameters of the normalized driving variable samples of the corresponding variable pairs and output the dependency strength of the variable pairs; For example, the dependency parameter estimation of corresponding variable pairs on the edge connection set is performed using maximum likelihood estimation, assuming variable pairs The corresponding binary Copula type is ,but: in, For variable pairs The estimated values ​​of the dependency parameters; For variable pairs The dependency parameters to be estimated; It is the density function of a bivariate copula.

[0042] S3.5 Based on the dependency strength of the variable pairs corresponding to the Vine Copula dependency structure, the variables are aggregated according to the variable dimension of the high-dimensional emission driving variable set to obtain the dependency strength set associated with each variable, and the dependency strength set is associated with the data quality score corresponding to the variable. Specifically, when aggregating the set of dependency strengths of variable pairs by variable dimension, all pairs of variables are included. The dependency strengths of the associated edge connections are summarized into a set. and the set With variables Data quality score Associative storage provides input of the same dimension for subsequent calculation of initial dependency contribution values.

[0043] S3.6. According to the preset weight generation rules, summarize the dependency strength set of each variable to obtain the initial dependency contribution value corresponding to the variable. For example, the preset weight generation rule calculates the initial dependency contribution value corresponding to the variable in the following way: in, For variables The initial dependency contribution value; For variables The set of dependencies of an association; The number of elements in the set; for The dependency strength element in.

[0044] S3.7. Combine the data quality score to perform quality correction on the initial dependency contribution value to obtain the corrected dependency contribution value corresponding to each driving variable. As an example, this embodiment uses a multiplicative correction method to calculate the correction dependency contribution value: in, For variables The corrected dependency contribution value; This is the initial dependency contribution value; Score the data quality.

[0045] S3.8. Normalize the modified dependency contribution values ​​of each variable to generate dependency contribution weights that correspond one-to-one with each driving variable. As an example, this embodiment uses a proportional normalization method to calculate the dependency contribution weight: in, For variables Dependency contribution weight; For variables The corrected dependency contribution value; The number of variables in the set of high-dimensional emission driving variables; The contribution value for the correction dependency of the k-th variable.

[0046] S3.9 Compare the dependency strength of each variable with the preset dependency threshold to obtain a set of candidate reduced variables that are lower than the preset dependency threshold; For example, the preset dependency threshold is determined based on the quantiles of the distribution of candidate dependency strengths: in, Preset dependency threshold; This is the set of absolute values ​​of the dependencies of all candidate variables. It is a quantile function; This is the quantile parameter, such as 0.3.

[0047] S3.10. For each variable in the candidate reduction variable set, adjust the dependency contribution weight corresponding to the variable according to the preset decay rule. The preset decay rule includes: when the dependency strength of the variable is less than the preset dependency threshold and not less than half of the preset dependency threshold, the dependency contribution weight of the variable is adjusted to half of the original dependency contribution weight; when the dependency strength of the variable is less than half of the preset dependency threshold, the dependency contribution weight of the variable is adjusted to one-tenth of the original dependency contribution weight. S3.11 For variables that do not belong to the candidate reduction variable set, their original dependency contribution weights are retained, and the dependency contribution weights adjusted by the preset attenuation rule are used as the input for the subsequent emission baseline weighting correction.

[0048] S4. Weight the emission baseline based on the dependency contribution weights to generate a carbon footprint accounting correction factor. Output the carbon footprint accounting correction factor along with the marginal distribution parameters, Vine Copula dependency structure, and dependency contribution weights as the correction result. Note that the following should be noted in this step: S4.1 Obtain the emission baseline corresponding to the accounting time window, and obtain the dependency contribution weights corresponding one-to-one with the set of high-dimensional emission driving variables; It should be noted that when obtaining the emission baseline amount corresponding to the accounting time window, each accounting time window is located using a time indexing rule. emission benchmark Simultaneously, the dependency contribution weights corresponding one-to-one with the set of high-dimensional emission driving variables are obtained from the S3 output. The dependency contribution weights are reorganized into a weight vector that is independent of the time window. To form an executable weighted input, the following time window-level driving variable aggregation is established: in, For accounting time window The weighted driving aggregation amount; For variables Dependency contribution weight; For variables During the accounting time window The normalized values; Number of variables; This is the k-th accounting time window.

[0049] S4.2. The emission baseline is weighted according to the dependency contribution weight to obtain the weighted emission amount corresponding to the accounting time window; For example, the emission baseline is proportionally adjusted using a weighted driving polymerization amount: in, In order to align with the accounting time window The corresponding weighted emissions; In order to align with the accounting time window The corresponding emission baseline; For accounting time window The weighted driving aggregation amount.

[0050] S4.3. The weighted emissions corresponding to each accounting time window are merged according to the target product batch to obtain the corrected emissions for that batch; For example, for those belonging to the same batch Time window set Summation: in, For batch Corrected emissions; In order to be consistent with batch The set of associated accounting time windows; Weighted emissions.

[0051] S4.4 Using the target product batch and accounting time window as the corresponding key, the corrected emissions and the emission baseline are matched in pairs, and the carbon footprint accounting correction factor is generated according to the ratio of the corrected emissions to the emission baseline under the same corresponding key.

[0052] For example, by using the target product batch and the accounting time window as the corresponding key, and pairing the corrected emissions with the emission baseline, a batch-level baseline emissions can be constructed: in, For batch The aggregate value of the emission baseline; This serves as a baseline for emissions within a specific time window. Further generate carbon footprint accounting correction factors: in, For batch The carbon footprint accounting correction factor; For batch Corrected emissions; For batch The aggregate value of the emission baseline.

[0053] As an example, if the baseline emissions for a certain batch within three accounting time windows are 10, 12, and 11 respectively, and the corresponding weighted aggregate emissions are 0.95, 1.05, and 1 respectively, then... .

[0054] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. 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 be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for correcting carbon footprint accounting factors based on the Copula model, characterized in that, include: Obtain a carbon footprint accounting sample set for the target product. Each sample contains an emission baseline and a set of high-dimensional emission driving variables aligned with the emission baseline within the same accounting time window. Record the corresponding data quality score for each variable in the set of high-dimensional emission driving variables. The high-dimensional emission driving variable set is fitted with marginal distributions dimension by dimension and normalized according to a unified quantile mapping rule to obtain marginal distribution parameters and normalized driving variable samples. Based on the normalized driving variable samples, a Vine Copula dependency structure is constructed, and the root node order and edge connection set are determined according to the structure scoring criteria to obtain the dependency strength of variable pairs. The dependency contribution weights corresponding to each driving variable are generated by combining the data quality score. The weights of variables whose dependency strength of variable pairs is lower than the preset dependency threshold are reduced according to the preset decay rule. The emission baseline is weighted and corrected according to the dependency contribution weight to generate a carbon footprint accounting correction factor. The carbon footprint accounting correction factor, the marginal distribution parameter, the Vine Copula dependency structure, and the dependency contribution weight are used together as the correction result output.

2. The method for correcting carbon footprint accounting factors based on the Copula model according to claim 1, characterized in that, The method for generating the carbon footprint accounting sample set and the high-dimensional emission driving variable set includes: Determine the accounting boundary corresponding to the target product, set the accounting time window, and generate a unified time index rule based on the accounting time window; Within the accounting time window, the emission baseline quantity corresponding to the target product batch is collected, and the time index corresponding to the accounting time window is recorded for the emission baseline quantity. Simultaneously, within the accounting time window, variables such as electricity carbon intensity, fuel structure, raw material carbon intensity, transportation intensity, and key process efficiency are collected, and a time index corresponding to the accounting time window is recorded for each variable to form the set of high-dimensional emission driving variables. According to the time indexing rules, the emission baseline quantity and the set of high-dimensional emission driving variables are time-aligned to obtain an aligned data segment corresponding to the accounting time window; The aligned data fragments are merged according to the target product batch to generate a carbon footprint accounting sample set for the target product, such that each sample contains the emission baseline and the set of high-dimensional emission driving variables aligned with the emission baseline within the same accounting time window.

3. The method for correcting carbon footprint accounting factors based on the Copula model according to claim 1 or 2, characterized in that, The data quality score is obtained by: Based on the accounting time window and the time indexing rule, a data quality scoring rule set is established. The data quality scoring rule set includes at least missing state rules, time consistency rules and source consistency rules, and a corresponding scoring level table is set for each rule. For each variable in the set of high-dimensional emission driving variables, the missing status of the variable is counted within the accounting time window, and the missing level is given according to the missing status rules. Where no missing is recorded as high level, sporadic missing does not cross the key process section as medium level, and continuous missing or covering the key process section is recorded as low level. For each variable in the set of high-dimensional emission driving variables, verify the consistency between the time index of the variable and the time index rule, and give the time consistency level according to the time consistency rule, where full window matching is recorded as high level, local misalignment that can be aligned according to the time index rule is recorded as medium level, and misalignment across accounting time windows is recorded as low level. For each variable in the set of high-dimensional emission driving variables, the source consistency is verified based on the same batch of measurement records, production ledger records and supply chain voucher records within the accounting time window, and the source consistency level is given according to the source consistency rules. The variable is recorded as high level when it is consistent with at least two types of records in terms of both value and time; it is recorded as medium level when it is consistent with only one type of record in terms of both value and time; and it is recorded as low level when it is inconsistent with all three types of records or when any type of record is missing. Based on the variable's missing level, temporal consistency level, and source consistency level, a data quality score is generated for the variable according to the scoring level table.

4. The method for correcting carbon footprint accounting factors based on the Copula model according to claim 3, characterized in that, Obtaining the marginal distribution parameters and the normalized driving variable samples includes: The dimensional observation data of the high-dimensional emission driving variable set are extracted from the carbon footprint accounting sample set, and the dimensional observation data are segmented and aggregated according to the accounting time window to obtain the time window sample sequence corresponding to each variable. Marginal distribution fitting is performed on the sample sequences of the time window to obtain marginal distribution parameters that correspond one-to-one with each variable; Based on the marginal distribution parameters, the quantile mapping relationship of each variable is constructed respectively, and the quantile mapping relationship of each variable is constrained by a unified rule to obtain a unified quantile mapping rule. The time window sample sequences of each variable are normalized according to the unified quantile mapping rule to obtain the normalized driving variable samples corresponding to the accounting time window.

5. The method for correcting carbon footprint accounting factors based on the Copula model according to claim 4, characterized in that, Marginal distribution fitting is performed on the sample sequences within the time window, including: For each variable in the set of high-dimensional emission driving variables, the time window sample sequence corresponding to the variable is extracted and arranged according to the accounting time window order to obtain the marginal fitting sample sequence of the variable; For each variable, the marginal fitting sample sequence is fitted with a candidate distribution according to a preset marginal distribution family to obtain the fitting result corresponding to each candidate distribution; Based on the preset goodness-of-fit evaluation rules, the marginal distribution type corresponding to each variable is determined from the fitting results, and the marginal distribution parameters corresponding to each variable are output.

6. The method for correcting carbon footprint accounting factors based on the Copula model according to claim 5, characterized in that, The unified quantile mapping rule is obtained by: Based on the marginal distribution parameters that correspond one-to-one with each variable, quantile mapping relationships are generated for each variable, and the quantile mapping relationships map the time window sample sequence of the variable to a preset set of quantiles. A unified rule constraint is applied to the quantile mapping relationship of each variable. The unified rule constraint includes at least a unified set of quantiles, a unified boundary quantile processing rule, and a unified interpolation rule. The quantile mapping relationships of each variable that satisfy the unified rule constraints are summarized into a unified quantile mapping rule. The unified quantile mapping rule stipulates that each variable is mapped to the same set of quantiles, and samples that exceed the boundary quantiles are processed by consistent boundary merging.

7. The method for correcting carbon footprint accounting factors based on the Copula model according to claim 4, characterized in that, The strength of the dependency between the variables is obtained, including: Based on the normalized driving variable sample, candidate variable pairs are extracted from the high-dimensional emission driving variable set and aggregated according to the corresponding relationship of the accounting time window to form a candidate variable pair sample sequence. The dependency of each candidate variable pair is calculated on the sample sequence of the candidate variable pair to obtain a candidate dependency set that corresponds one-to-one with each candidate variable pair. The candidate dependency set is structurally scored according to the structural scoring criteria to determine the root node order and edge connection set of the Vine Copula dependency structure, thereby obtaining the Vine Copula dependency structure. On the set of edges, the dependency parameters of the normalized driving variable samples of the corresponding variable pairs are estimated, and the dependency strength of the variable pairs is output.

8. The method for correcting carbon footprint accounting factors based on the Copula model according to claim 7, characterized in that, Generating the dependency contribution weights includes: Based on the dependency strength of the variable pairs corresponding to the Vine Copula dependency structure, the variables are aggregated according to the variable dimensions of the high-dimensional emission driving variable set to obtain the dependency strength set associated with each variable, and the dependency strength set is associated with the data quality score corresponding to the variable. According to the preset weight generation rules, the set of dependency strengths for each variable is summarized to obtain the initial dependency contribution value corresponding to that variable. The initial dependency contribution value is corrected by combining the data quality score to obtain the corrected dependency contribution value corresponding to each driving variable. The modified dependency contribution values ​​of each variable are uniformly normalized to generate dependency contribution weights that correspond one-to-one with each driving variable.

9. The method for correcting carbon footprint accounting factors based on the Copula model according to claim 8, characterized in that, The reduction of weights according to a preset attenuation rule includes: The dependency strength of each variable is compared with a preset dependency threshold to obtain a set of candidate reduced variables that are lower than the preset dependency threshold. For each variable in the candidate reduction variable set, the dependency contribution weight corresponding to the variable is adjusted according to a preset decay rule. The preset decay rule includes: when the dependency strength of the variable is less than a preset dependency threshold and not less than half of the preset dependency threshold, the dependency contribution weight of the variable is adjusted to half of the original dependency contribution weight; when the dependency strength of the variable is less than half of the preset dependency threshold, the dependency contribution weight of the variable is adjusted to one-tenth of the original dependency contribution weight. For variables that do not belong to the candidate reduction variable set, their original dependency contribution weights are retained, and the dependency contribution weights adjusted by the preset attenuation rule are used as inputs for subsequent emission baseline weighting correction.

10. The method for correcting carbon footprint accounting factors based on the Copula model according to claim 8, characterized in that, Generating the carbon footprint accounting correction factor includes: Obtain the emission baseline amount corresponding to the accounting time window, and obtain the dependency contribution weight corresponding one-to-one with the set of high-dimensional emission driving variables; The emission baseline is weighted according to the dependent contribution weight to obtain the weighted emission corresponding to the accounting time window; The weighted emissions corresponding to each accounting time window are merged according to the target product batch to obtain the corrected emissions for that batch; Using the target product batch and the accounting time window as corresponding keys, the corrected emissions and the emission baseline are matched in pairs, and the carbon footprint accounting correction factor is generated according to the ratio of the corrected emissions to the emission baseline under the same corresponding key.

Images