A method and system for evaluating the carbon footprint of digital intelligent office consumables and a medium

Abstract

The application discloses a kind of carbon footprint evaluation method, system and medium of digital intelligent office consumables, it is related to carbon footprint accounting technical field, method includes: the multi-source data of digital intelligent office consumables is collected, and carbon footprint value of consumables is obtained by carbon footprint quantization channel calculation;Rely on multi-dimensional green product evaluation index library and obtain the comprehensive weight coefficient containing subjective and objective weight by hierarchical analysis;Carbon footprint deviation degree is calculated to obtain index score data, and carbon footprint evaluation score is obtained by self-adaptive weighting calculation and generates carbon footprint evaluation report.The application solves the technical problems that traditional office consumables carbon footprint accounting relies on manual data processing, fixed parameter calculation and single evaluation weight, data integrity, carbon emission calculation precision and comprehensive evaluation objectivity are insufficient, achieves to improve the precision and dynamic adaptability of carbon footprint evaluation, so that the scoring result can more accurately reflect the real environmental performance of consumables and adapt to the technical effects of green product evaluation requirements.

Description

Technical Field

[0001] This invention relates to the field of carbon footprint accounting technology, and in particular to a method, system and medium for evaluating the carbon footprint of digital intelligent office consumables. Background Technology

[0002] Digital intelligent office consumables are used in various office scenarios, and their full lifecycle carbon footprint is a core indicator for green product evaluation and low-carbon management. Accurate quantification of carbon footprint is crucial for green certification of consumables and supply chain emission reduction management. Current methods for calculating the carbon footprint of office consumables mostly rely on manual collection of material, logistics, and energy consumption data, combined with static emission factors, and then manual scoring using the analytic hierarchy process (AHP) to conduct green evaluations. While this method is usable for basic calculations of standardized and simple products, it reveals significant shortcomings when adapted to digital intelligent office consumables. Existing technologies suffer from multiple defects: manual data collection is prone to data gaps; unclear addresses on logistics documents prevent accurate calculation of transportation carbon emissions; fixed static emission factors fail to differentiate carbon emission differences among multiple supplier levels; total carbon emissions are roughly allocated by weight without considering print lifespan to calculate the carbon footprint per page; and fixed weighting ratios cannot adapt to conflicting indicators and data fluctuations, ultimately resulting in distorted carbon footprint data and biased evaluation results, failing to meet the actual needs of accurate assessment and low-carbon management of green consumables. Summary of the Invention

[0003] This application provides a method, system, and medium for evaluating the carbon footprint of digital intelligent office consumables, which solves the technical problems of traditional office consumables carbon footprint accounting relying on manual data processing, fixed parameter calculation and single evaluation weight, and insufficient data integrity, carbon emission calculation accuracy and comprehensive evaluation objectivity.

[0004] The first aspect of this application provides a method for evaluating the carbon footprint of digital intelligent office consumables. The method includes: intelligently collecting multi-source data from digital intelligent office consumables and automatically capturing multi-source datasets; constructing a carbon footprint quantification channel, calculating the lifecycle carbon footprint of the multi-source datasets through the carbon footprint quantification channel to obtain multiple consumable carbon footprint values; constructing a multi-dimensional green product evaluation index library and performing hierarchical analysis on the multiple consumable carbon footprint values ​​to generate a comprehensive weighting coefficient, the comprehensive weighting coefficient including a subjective weighting coefficient and an objective weighting coefficient; performing deviation calculation based on the multiple consumable carbon footprint values ​​to obtain multiple deviation degrees, matching them with the multi-dimensional green product evaluation index library, and calculating multiple indicator scores; weighting the multiple indicator scores according to the subjective weighting coefficient and the objective weighting coefficient to obtain a carbon footprint evaluation score, and generating a carbon footprint evaluation report based on the carbon footprint evaluation score.

[0005] A second aspect of this application provides a carbon footprint assessment system for digital intelligent office consumables. The system includes: a multi-source dataset acquisition module for intelligently collecting multi-source data from digital intelligent office consumables and automatically capturing multi-source datasets; a consumable carbon footprint value acquisition module for constructing a carbon footprint quantification channel and calculating the lifecycle carbon footprint of the multi-source datasets through the carbon footprint quantification channel to obtain multiple consumable carbon footprint values; a comprehensive weighting coefficient generation module for constructing a multi-dimensional green product evaluation index library and performing hierarchical analysis on the multiple consumable carbon footprint values ​​to generate a comprehensive weighting coefficient, which includes a subjective weighting coefficient and an objective weighting coefficient; an index score data calculation module for performing deviation calculations based on the multiple consumable carbon footprint values ​​to obtain multiple deviation degrees, which are then matched with the multi-dimensional green product evaluation index library to calculate multiple index score data; and a carbon footprint assessment report generation module for weighting the multiple index score data according to the subjective weighting coefficient and the objective weighting coefficient to obtain a carbon footprint assessment score, and generating a carbon footprint assessment report based on the carbon footprint assessment score.

[0006] A third aspect of this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements a carbon footprint assessment method for digital intelligent office consumables provided in this application.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application automatically collects multi-dimensional raw data from the entire lifecycle of office consumables and intelligently cleans and completes it. It dynamically calculates carbon emissions at each stage by combining differences in the upstream and downstream supply chains, and refines the carbon emissions per unit output based on the actual usage characteristics of the products. It integrates standard guidance and sample data characteristics to adaptively generate a comprehensive evaluation score, thereby completely quantifying the low-carbon level of office consumables throughout the entire process. This makes the carbon footprint evaluation results of digital intelligent office consumables more in line with national standards and significantly improves the accuracy and objectivity of the calculation. It achieves the technical effect of improving the accuracy and dynamic adaptability of carbon footprint evaluation, making the scoring results more accurately reflect the real environmental performance of consumables and adapt to the requirements of green product evaluation. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0009] Figure 1 This is a flowchart illustrating a method for evaluating the carbon footprint of digital intelligent office consumables provided in an embodiment of this application.

[0010] Figure 2 This is a single-page carbon footprint trend chart of a carbon footprint evaluation method for digital intelligent office consumables provided in this application embodiment.

[0011] Figure 3 This is a schematic diagram of the structure of a carbon footprint assessment system for digital intelligent office consumables provided in an embodiment of this application.

[0012] Figure labeling: Multi-source dataset acquisition module 1, Consumable carbon footprint value acquisition module 2, Comprehensive weight coefficient generation module 3, Indicator score data calculation module 4, Carbon footprint evaluation report generation module 5. Detailed Implementation

[0013] This application provides a method, system, and medium for evaluating the carbon footprint of digital intelligent office consumables, which solves the technical problems of traditional office consumables carbon footprint accounting relying on manual data processing, fixed parameter calculation and single evaluation weight, and insufficient data integrity, carbon emission calculation accuracy and comprehensive evaluation objectivity.

[0014] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0015] It should be noted that the terms "first," "second," etc., in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to such processes, methods, products, or devices.

[0016] Example 1, as Figure 1 As shown, a method for evaluating the carbon footprint of digital intelligent office consumables is provided, wherein the method includes: Intelligent collection of multi-source data for digital intelligent office supplies, automatically capturing multi-source datasets.

[0017] In this embodiment of the application, digital intelligent office consumables are consumable accessories such as drum and toner cartridges, ink, ribbons, and inks that are compatible with digital printing and copying equipment and have matching functions.

[0018] Specifically, the process involves automatically capturing relevant data on consumables to obtain an initial raw material consumption sequence and removing outliers, filling in missing time-series data to obtain raw material consumption data, scanning and parsing logistics documents to extract transportation origin and destination information, constructing a map distance matrix for fuzzy addresses to obtain distance parameters for multiple candidate paths, and combining this with historical transportation trajectories to obtain transportation distance data. Finally, the raw material consumption data and transportation distance data are integrated to form a multi-source dataset. The specific implementation process of this step will be described in detail later.

[0019] A carbon footprint quantification channel is constructed, and the life cycle carbon footprint of the multi-source dataset is calculated through the carbon footprint quantification channel to obtain multiple consumable carbon footprint values.

[0020] In this embodiment, the carbon footprint quantification channel is a modular program path that integrates multiple types of carbon emission calculation logic, unifies data input and output formats, and automatically completes the calculation of carbon emissions throughout the entire life cycle of digital intelligent office consumables in stages. The carbon footprint value of the consumables is the carbon dioxide equivalent emission value calculated for a single digital intelligent office consumable, with the number of printed pages as the functional unit.

[0021] Optionally, a modular, sequential architecture is adopted when constructing the carbon footprint quantification channel. First, a unified data input specification and output format are defined for the channel, clarifying the field standards and numerical unit requirements for the three core input data categories: raw material consumption data, transportation distance data, and emission factor parameters. The unit for raw material consumption data is kilograms, the unit for transportation distance data is kilometers, and the unit for carbon emission calculation results is kilograms of carbon dioxide equivalent, ensuring that data from different sources can be uniformly accessed and processed within the channel. Second, three independent calculation units are sequentially set within the channel: a raw material carbon emission calculation module, a transportation carbon emission calculation module, and a functional unit allocation module. These three units execute in a fixed sequence according to the lifecycle stages, with the output of the previous unit directly serving as the input parameter for the next unit. Each calculation unit has an independent parameter configuration interface, supporting real-time reading and dynamic updating of external factor library data. A result cache node is also configured to store intermediate calculation data and the final calculation results, ultimately forming a complete carbon footprint quantification calculation pathway.

[0022] Next, a raw material carbon emission factor library is constructed and linked with raw material consumption data to obtain carbon emission data for the raw material acquisition stage. Combined with transportation distance data and transportation mode emission factors, carbon emission data for the transportation stage is calculated. Then, the two types of carbon emission-related data are allocated to functional units through the carbon footprint quantification channel to finally obtain multiple consumable carbon footprint values. The specific implementation process of this step will be described in detail later.

[0023] A multidimensional green product evaluation index library is constructed, and a hierarchical analysis is performed on the carbon footprint values ​​of the multiple consumables to generate a comprehensive weight coefficient, which includes subjective weight coefficients and objective weight coefficients.

[0024] In this embodiment of the application, the green product evaluation index library is a data set that stores low-carbon evaluation indicators, benchmark thresholds, and scoring rules for consumables, such as carbon footprint, recyclability ratio, production energy consumption, and waste output.

[0025] In one embodiment of this application, a hierarchical mapping analysis is conducted based on a multidimensional green product evaluation index library and consumable carbon footprint values, and an indicator importance judgment matrix is ​​constructed and optimized. The subjective weight coefficients of each evaluation index are obtained through consistency analysis and normalization. Then, an evaluation matrix of consumable sample indicators is constructed based on raw material consumption data, and the objective weight coefficients are obtained by combining carbon footprint fluctuation penalty adjustment. Finally, the comprehensive weight coefficients of each evaluation index are obtained by game-theoretic combination of subjective and objective weights. The specific implementation process of this step will also be described in detail later.

[0026] Based on the carbon footprint values ​​of the multiple consumables, deviation calculations are performed to obtain multiple deviation degrees, which are then matched with a multidimensional green product evaluation index library to calculate the scores of multiple indicators.

[0027] Specifically, the evaluation benchmark parameters and scoring rules corresponding to the carbon footprint indicators of a unit product are first retrieved from the multi-dimensional green product evaluation index library. In this embodiment, the evaluation benchmark is selected as the industry average single-page carbon footprint value of digital intelligent office consumables of the same category. "Same category" refers to consumable product categories that are consistent in function, applicable equipment type, and product sub-category. For example, toner cartridge consumables correspond to the industry average level of the toner cartridge category, and inkjet ink consumables correspond to the industry average level of the inkjet ink category. The benchmark reference value is obtained by arithmetically averaging the publicly available industry green product evaluation dataset and the publicly available carbon footprint disclosure data of mainstream consumable products. Simultaneously, an optimal threshold corresponding to the full score and a passing threshold are set. In this embodiment, the optimal threshold is set to 80% of the benchmark reference value, and the passing threshold is set to 120% of the benchmark reference value. The scoring uses a percentage system, with the full score corresponding to the optimal threshold and 0 points corresponding to carbon emission performance exceeding the passing threshold. The threshold boundary determination rule is as follows: when the carbon footprint deviation is exactly equal to the deviation ratio corresponding to the optimal threshold, it is scored as full marks; when the carbon footprint deviation is exactly equal to the deviation ratio corresponding to the passing threshold, it is scored as 0 marks.

[0028] The carbon footprint value for each consumable is extracted individually, and deviation calculations are performed. First, the absolute deviation is calculated by subtracting the benchmark value from the carbon footprint value of the consumable being measured, yielding the absolute difference between the consumable's carbon footprint and the industry benchmark. Next, the relative deviation is calculated by dividing the absolute deviation by the benchmark value, obtaining the relative deviation of a single consumable's carbon footprint relative to the industry benchmark. A negative deviation value indicates that the consumable's carbon footprint is below the industry average; a smaller value indicates better carbon emission performance. A positive deviation value indicates that the consumable's carbon footprint is above the industry average; a larger value indicates worse carbon emission performance. This deviation calculation is performed for all consumable carbon footprint values, resulting in multiple deviation results corresponding to the quantity of consumables.

[0029] The calculated deviation for each consumable is then matched one-to-one with the unit product carbon footprint index in the multidimensional green product evaluation index library, and the index score is calculated according to the preset linear scoring rules of the index library. When the deviation is less than or equal to the deviation ratio corresponding to the optimal threshold, the index score is full. When the deviation is greater than or equal to the deviation ratio corresponding to the passing threshold, the index score is zero. When the deviation is between the optimal threshold and the passing threshold, a linear interpolation method is used to calculate the score. Let the deviation corresponding to the optimal threshold be... The deviation corresponding to the passing threshold is The deviation of the current consumable to be tested is The score is calculated using the following formula: Where S represents the index score on a percentage scale. Using the optimal threshold (corresponding to full score) and the passing threshold (corresponding to zero score) as two calculation endpoints, the score is equal to the full score multiplied by the difference between the deviation corresponding to the passing threshold and the current deviation, then divided by the difference between the deviation corresponding to the passing threshold and the deviation corresponding to the optimal threshold. This process is repeated for all consumables, generating multiple index score data sets.

[0030] By clarifying the sources of benchmark data, boundary determination rules, and linear interpolation calculation methods, the carbon footprint performance can be quantified based on the relative deviation from the benchmark reference. Combined with the scoring rules of the indicator library, the indicator scores can be converted, thus achieving standardized quantitative scoring of carbon footprint indicators and providing a unified scoring basis for subsequent comprehensive evaluation of green products.

[0031] The scores of the multiple indicators are weighted according to the subjective weight coefficient and the objective weight coefficient to obtain a carbon footprint evaluation score, and a carbon footprint evaluation report is generated based on the carbon footprint evaluation score.

[0032] In this embodiment of the application, the carbon footprint evaluation report is a standardized written document that summarizes the scores, weight parameters, comprehensive evaluation scores and optimization suggestions of various low-carbon indicators of consumables, and is used to intuitively display the low-carbon performance of office consumables.

[0033] Specifically, the first step is to calculate the index conflict degree and index variability based on the index score data. The two types of values ​​are then adaptively adjusted to obtain the weight fusion ratio of subjective and objective weights. The combined weight parameters are then weighted and synthesized according to the fusion ratio. Finally, the combined weight parameters are used to weight and sum the index score data to obtain the carbon footprint evaluation score. The specific implementation process of this step will be explained in detail later.

[0034] Next, a carbon footprint assessment report is generated based on the carbon footprint assessment score. For example, in this embodiment, the assessment score grading rules are set as follows: 90 points and above correspond to an excellent grade, 80 to 89 points correspond to a good grade, 70 to 79 points correspond to a satisfactory grade, and below 70 points correspond to a grade requiring improvement. The corresponding grade is matched based on the assessment score of each consumable. The core content of the report is compiled, including basic product model information, detailed scores for each sub-indicator, combined weight parameters, the final assessment score and corresponding grade, comparative data with similar consumables, and suggestions for low-carbon optimization.

[0035] The data includes horizontal comparisons of similar consumables, generated based on the ranking of each consumable sample's scores across various sub-indicators. These scores are presented in tables or bar charts, showing the deviation of each consumable's score from the average score of other similar consumables. Low-carbon optimization suggestions are generated based on the gap between the scores of each sub-indicator and the optimal threshold. Indicators with scores below the acceptable threshold or significantly below the average level of similar consumables are automatically marked, and corresponding targeted optimization suggestions are output. The report supports exporting to PDF and Excel formats. PDF is used for formal archiving and display, while Excel is used for subsequent data statistical analysis. Finally, a complete carbon footprint assessment report is generated according to standardized field layouts.

[0036] Furthermore, the method provided in this application embodiment includes: Based on digital intelligent office supplies, multiple data are automatically captured to obtain an initial raw material consumption sequence. The initial raw material consumption sequence is traversed to identify and remove outliers, obtaining missing location parameters. Based on the missing location parameters, time-series data is filled in to generate raw material consumption data. Optical character scanning and parsing are performed on the digital intelligent office supplies to obtain original logistics document text parameters. The original logistics document text parameters are traversed to extract information, obtaining structured transportation information, including transportation origin data and transportation destination data. When the transportation origin data or the transportation destination data is fuzzy text, an address resolution signal is generated. A map distance matrix is ​​constructed using the address resolution signal, and multiple candidate path distance parameters are determined. Based on the multiple candidate path distance parameters and historical transportation trajectories, dynamic time warping is performed to generate transportation distance data. The raw material consumption data and the transportation distance data are added to the multi-source dataset.

[0037] Specifically, for the production material data of digital intelligent office supplies, a Hypertext Transfer Protocol (HTTP) data interaction interface is used to connect with the enterprise's production material management system via data connection. Using a daily fixed time step, the system extracts all raw material requisition and consumption records for the corresponding batch of digital intelligent office supplies within its production cycle. These raw materials cover all types required for the production of these digital intelligent office supplies, including polypropylene plastic, aluminum photosensitive drum, selenium, and carbon black. The extracted consumption data is arranged chronologically to form an initial raw material consumption sequence with timestamps. This sequence contains the actual consumption quantity of each type of raw material at each time point.

[0038] For the consumption data of each type of raw material, an independent time series subsequence is constructed, and outlier detection is performed for each type. Specifically, for any time series subsequence of a raw material, the arithmetic mean of all data points in the subsequence and the population standard deviation are first calculated. Then, the difference between the value of each data point in the subsequence and the arithmetic mean is calculated. When the absolute difference between the value of a data point and the arithmetic mean is greater than three times the population standard deviation, i.e., when the following condition is met: ,in, The value of the current data point. It is the arithmetic mean of all data points in this time series subsequence. This represents the overall standard deviation of all data points in this time series subsequence.

[0039] This data point is then identified as an anomaly. All anomaly-identified data points are removed from the time series subsequence. After removal, the position of the anomaly in the original time series becomes the missing data location. For each missing data location, its corresponding time sequence number, raw material category, and position identifier in the sequence are recorded. These records are stored in an array, with each array corresponding to an independent missing data location, generating a parameter for removing missing data locations. This array is directly called by subsequent time series data imputation steps to achieve precise location and targeted imputation of missing data locations.

[0040] Next, a generator is built by stacking causal temporal convolutional network layers and multi-head self-attention layers to capture time series with long-range dependencies. Then, a convolutional neural network discriminator is built and adversarial training is carried out. The reconstruction error and adversarial loss are alternately optimized until the discriminator can no longer distinguish between the original time series and the imputed time series, thus completing the construction of the generative adversarial network model. Finally, the missing position parameters are input into the model to complete the time series data imputation and obtain complete raw material consumption data. The specific implementation process of this step will be described in detail later.

[0041] Then, images of various logistics documents corresponding to digital intelligent office supplies are acquired using scanning equipment or image acquisition terminals to obtain logistics document images. Optical character recognition (OCR) technology is then used to scan and analyze these images. Specifically: First, preprocessing operations are performed on the acquired logistics document images, including grayscale conversion, binarization, and tilt angle correction to eliminate image noise and deformation interference. Next, text recognition is performed line by line and region by region on the preprocessed document images to extract all text content on the document, resulting in the original logistics document text parameters. These parameters cover all information contained on the document, including shipper information, consignee information, material information, and transportation information.

[0042] Furthermore, optical character recognition employs a technical approach combining traditional image processing and pattern recognition: First, the horizontal projection method is used to locate the text regions of each line based on the pixel density valley value along the row direction. Then, for each text line, the vertical projection method is used to independently segment individual characters based on the pixel density valley value along the column direction. For the segmented character image blocks, their directional gradient histogram features are extracted and compared with the corresponding features of each template character in a pre-built standard printed character template library using cosine similarity. The template character with the largest cosine value of the angle exceeding a preset threshold is selected as the recognition result, with the threshold set to 0.85 in this example. When the maximum similarity is lower than the threshold, topological feature matching based on stroke endpoints and ring number is used for verification and correction. After concatenating the recognition results of each line in row order, neighborhood semantic disambiguation correction is performed on easily confused characters (such as the number 0 and the uppercase letter O) according to the Chinese context, and finally, the complete original logistics document text parameters are output.

[0043] Next, a full traversal of the original logistics document text parameters is performed, employing a combination of keyword matching and regular expression matching to extract transportation information. A pre-defined set of address keywords is established, specifically including matching identifiers such as origin, destination, delivery address, and shipping address. Regular expression matching rules are configured to adapt to address text formats at the provincial, municipal, district, county, and street levels. During the traversal, the text paragraphs corresponding to the keywords are located, and the address text content following the keywords is extracted and organized into transportation origin data and transportation destination data. The extracted origin and destination information are stored according to a unified field format to generate structured transportation information.

[0044] Next, the text completeness of the transportation origin and destination data is verified. When the address text only contains district / county level information, lacks specific road and location information, or contains abbreviations or aliases, the corresponding address is determined to be fuzzy text, and an address resolution signal is generated simultaneously. After triggering the address resolution process, the geocoding interface of the general map service is called to perform multi-candidate address matching on the fuzzy address, returning several semantically accurate geographic coordinate points. All candidate points of the transportation origin and all candidate points of the transportation destination are paired, and the shortest path planning distance of the conventional freight road between the two points is calculated one by one. The path planning adopts the recommended path rule that takes into account both main roads and commonly used freight routes, excluding road sections where freight vehicles are prohibited. All combined planning distance values ​​are arranged in rows and columns to form a map distance matrix, from which multiple candidate path planning distance parameters are extracted.

[0045] Next, historical transportation trajectory data within the same transportation route section is retrieved, and the corresponding mileage change sequences are extracted. For each candidate path's planned distance parameter, the corresponding path mileage sequence is dynamically time-warped with the historical transportation trajectory's mileage sequence. Sequence similarity is measured using Euclidean distance. During warping, step size constraints are set, allowing only three directions for each step: right, up, or upper right. A global path constraint requires the start and end points to correspond to the beginning and end points of both sequences. By elastically stretching and compressing the time dimension of the sequences, the cumulative distance deviation value on the corresponding paths of the two sequences is calculated. The cumulative distance deviation value is the cumulative sum of the Euclidean distances between corresponding points in the two sequences after dynamic warping. A smaller cumulative deviation value indicates a higher similarity between the candidate path and the historical actual transportation path. The planned distance value corresponding to the candidate path with the smallest cumulative deviation value is selected as the final transportation distance data.

[0046] Finally, the raw material consumption data and transportation distance data generated in the above steps are organized in a unified data format and added to a multi-source dataset for unified storage. The two types of data are bound one by one to the product identifier of digital intelligent office consumables, providing a complete basic data source for subsequent life cycle carbon footprint calculation.

[0047] By scanning and collecting optical characters, the system automatically identifies and extracts logistics address information. Combined with fuzzy address multi-candidate matching and dynamic regularization of historical trajectories, it clarifies path calculation rules and sequence measurement standards, thereby achieving automated and accurate calculation of transportation distance and effectively improving the efficiency of logistics data collection and the accuracy of basic data for transportation carbon emission accounting.

[0048] Furthermore, the method provided in this application embodiment includes: A generator is constructed by stacking causal temporal convolutional network layers and multi-head self-attention layers. The generator captures time series containing long-range dependencies. A discriminator is constructed based on a convolutional neural network, and the discriminator is used to discriminate and train the time series: S1: The generator reconstructs the time series according to the long-range dependencies, and performs adversarial loss optimization based on the imputed time series until the discriminator can no longer distinguish between the imputed time series and the original time series. A generative adversarial network model is then constructed. The missing positions are synchronized to the generative adversarial network model to impute the time series data and generate the raw material consumption data.

[0049] Optionally, when constructing the generator, a causal temporal convolutional network layer is first built, using a four-layer causal convolutional structure stacked together. The kernel size of each layer is set to 3, and the dilation rate increases progressively in multiples of 1, 2, 4, and 8. The number of output channels for each layer is set to 16, 32, 64, and 32, respectively. A linear rectified activation function is applied after each convolutional layer. The causal convolution implements temporal constraints through masking, setting the weight positions corresponding to future time steps of each convolutional kernel to zero, ensuring that only data from the current and historical time steps are used during the convolution operation, without introducing information from future time steps. After the causal temporal convolutional network layer, a multi-head self-attention layer is stacked, with 4 attention heads and a single-head attention dimension of 8. When calculating the scaled dot product, the query vector and key vector are first multiplied, then divided by the square root of the attention dimension for scaling. The attention weights are obtained by processing with a flexible maximum function, and finally, the single-head attention output is obtained by weighted summation with the value vector. The multi-head outputs are concatenated along the feature dimension to obtain the complete attention features.

[0050] The outputs of the causal temporal convolutional network layer and the multi-head self-attention layer are fused using a residual connection method. After fusion, the access layer normalization operation completes feature regularization. The generator's input dimension is the temporal length multiplied by the number of raw material category features, and the output dimension is consistent with the input dimension, resulting in a complete time series after imputation. As a specific example of this implementation, the temporal length corresponds to the number of temporal steps for 30 consecutive collection days, and the number of raw material category features corresponds to the total number of all raw material categories required for the production of digital intelligent office supplies, including but not limited to polypropylene plastic, aluminum photosensitive drum, selenium, carbon black, etc. The input tensor dimension is 30 multiplied by the total number of these categories.

[0051] The discriminator employs a three-layer convolutional neural network structure. Each convolutional layer has a kernel size of 3, a stride of 2, and is padded with zeros. After each convolutional layer, a linear rectified activation function (RCF) is applied, followed by a max-pooling layer for feature downsampling. The max-pooling layer has a kernel size of 2 and a stride of 2. The output channels of the three convolutional layers are set to 16, 32, and 64, respectively. Two fully connected layers follow the convolutional layers. The first fully connected layer has an output dimension of 32 and applies the RCF. The second fully connected layer has a single-neuron output and applies the S-curve activation function, outputting a discrimination probability value ranging from 0 to 1, representing the probability that the input sequence is a true sequence. The discriminator's input dimension is consistent with the generator's output dimension; the input is time-series data, and the output is a single-valued discrimination probability.

[0052] When constructing training samples, historically accumulated complete raw material consumption time-series data were selected as the basic dataset, and randomly divided into training and validation sets in a 7:3 ratio. For each complete time-series sequence in the training set, mask positions were randomly generated according to a set missing proportion to simulate data missing states in real-world scenarios. The randomly generated mask positions during the training phase are used to simulate the data missing state represented by the missing position parameter in the aforementioned steps; both are functionally used to identify the specific location of missing data in the time series. The masked time-series sequences are used as input samples for the generator, and the corresponding complete time-series sequences are used as reconstruction supervision labels. Weight initialization uses a normal distribution random initialization method, with the mean set to 0 and the standard deviation set to 0.02.

[0053] Furthermore, the training of the generative adversarial network model employs an alternating optimization strategy. In each iteration, the discriminator parameters are updated twice, followed by the generator parameters once. The optimizer uses an adaptive moment estimation optimizer with an initial learning rate of 0.0002. A fixed-step decay strategy is employed, reducing the learning rate to 0.9 times its original value every 100 iterations. The batch size is set to 32. Multiple overfitting prevention measures are implemented during training. A random deactivation layer with a deactivation probability of 0.2 is added after the fully connected layers of the generator and discriminator. A weight decay mechanism with a decay coefficient of 0.0001 is incorporated into the optimizer, along with an early stopping mechanism based on validation set verification to suppress model overfitting.

[0054] The discriminator's loss function uses binary cross-entropy loss, with the optimization objective being to maximize the discrimination accuracy between the real and generated sequences. The generator's loss function consists of a weighted sum of reconstruction loss and adversarial loss. The reconstruction loss uses mean squared error loss to calculate the error between the generated and real values ​​at the imputation positions. The adversarial loss uses binary cross-entropy loss to calculate the discriminator's discrimination error against the generated sequences. The weight coefficient for the reconstruction loss is set to 10, and the weight coefficient for the adversarial loss is set to 1. The optimization objective for the generator is to minimize the total loss. During training, the performance is validated using a validation set every 10 epochs. When the average discrimination accuracy of the discriminator on the validation set stabilizes between 0.45 and 0.55, and the generator's reconstruction loss no longer decreases for 30 consecutive epochs, the model is considered to have converged. Training is then stopped, and the generative adversarial network model parameters are saved.

[0055] After the model training is completed, the missing position parameters are synchronously mapped to the initial raw material consumption sequence to be processed. The sequence with missing positions is input into the trained generative adversarial network model. The generator calculates and fills in the missing position values ​​based on the long-range dependencies of the sequence, outputs complete time series data, and finally generates complete raw material consumption data.

[0056] By clarifying the network structure parameters of the generator and discriminator, the rules for constructing training samples, the alternating optimization training process, overfitting prevention measures, and convergence criteria, the model construction and application process for time series data filling is fully disclosed. This enables high-precision automatic filling of missing raw material consumption data, effectively ensuring the integrity and reliability of basic data for carbon footprint accounting.

[0057] Furthermore, the method provided in this application embodiment includes: A raw material carbon emission factor library is constructed, and the raw material consumption data is associated with the raw material carbon emission factor library to calculate the carbon emission data of the raw material acquisition stage. Based on the transportation distance data and the emission factor of the transportation mode, the carbon emission data of the transportation stage is obtained. The carbon footprint quantification channel is used to allocate the carbon emission data of the raw material acquisition stage and the transportation distance data to functional units to calculate the carbon footprint values ​​of the multiple consumables.

[0058] Specifically, when constructing the raw material carbon emission factor library, the basic carbon emission data of all categories of raw materials used in the production of digital intelligent office consumables is first collected, covering major materials such as polypropylene plastics, aluminum photosensitive drums, selenium, and carbon black. The basic data comes from publicly available industry carbon emission statistics sets and publicly disclosed carbon emission information from raw material production enterprises. The unit of measurement for emission factors is uniformly set as kilograms of carbon dioxide equivalent per kilogram of raw material. The basic emission data is classified and organized according to the raw material category, with each category of raw material corresponding to a set of unit mass carbon emission values, forming a static basic emission factor set. A corresponding factor update storage location is set for each category of raw materials to support the replacement and overwriting of subsequent dynamic emission factors. At the same time, a unified factor query interface is built, which can quickly match and call the corresponding emission factor values ​​based on the raw material category name, forming a complete and dynamically updatable raw material carbon emission factor library.

[0059] Next, based on the supplier management logs, material flow relationships are extracted and a directed supply network graph is constructed with suppliers as nodes and material flow relationships as edges. The network graph is traversed to generate initial feature vectors for each node. Then, attention weight coefficients are calculated through graph attention analysis. The node features are weighted and aggregated using these coefficients to obtain node feature vector identifiers. The identifiers are mapped to dynamic upstream emission factors to replace the original static emission factors in the factor library. The updated factor library is matched with raw material consumption data to complete the calculation, and finally, carbon emission data for the raw material acquisition stage is obtained. The specific implementation process of this step will be described in detail later.

[0060] The transport mode emission factor refers to the carbon emissions generated per unit weight of goods transported per unit distance, measured in kilograms of CO2 equivalent per ton per kilometer. Corresponding values ​​are set for three common transport modes: road freight, rail freight, and waterway freight. The emission factor values ​​are derived from publicly available carbon emission statistics in the transportation industry. In actual use, the emission factor is matched to the transport mode specified on the logistics document. If the logistics document does not explicitly indicate the transport mode, the emission factor corresponding to road freight is used by default. The carbon emission data for the transport stage is calculated by multiplying the transport weight and distance of the corresponding raw materials by the corresponding transport mode emission factor to obtain the carbon emission data for a single batch of raw materials during the transport stage.

[0061] Finally, by allocating carbon emission data from the raw material acquisition stage and the transportation stage to functional units through the carbon footprint quantification channel, multiple consumable carbon footprint values ​​can be calculated. The detailed allocation process for this part will be explained in detail in the subsequent functional unit allocation and single-page carbon footprint calculation sections.

[0062] By establishing a modular carbon footprint quantification channel and a standardized emission factor library, we clarify the unified measurement units and calculation logic for transportation carbon emissions, providing a unified calculation framework and basic parameter support for full life-cycle carbon footprint accounting.

[0063] Furthermore, the method provided in this application embodiment includes: Raw material supply relationship data is extracted from supplier management logs. Material flow relationships are then extracted from this data, and multiple supplier data points are located. These supplier data points are used as nodes, and the material flow relationships are used as edges to construct a directed supply network graph. Feature analysis is performed on the directed supply network graph to construct initial feature vectors for multiple nodes. Graph attention analysis is then performed on these initial feature vectors to calculate attention weight coefficients. The initial feature vectors are then weighted and aggregated according to these attention weight coefficients. The weighted aggregation results are used to update the initial feature vectors, generating multiple node feature vector identifiers. These node feature vector identifiers are mapped to dynamic upstream emission factors. These dynamic upstream emission factors are stored in the raw material carbon emission factor library, replacing static emission factors. The raw material carbon emission factor library is then updated, and carbon emission calculations are performed using the raw material consumption data to obtain carbon emission data for the raw material acquisition stage.

[0064] Specifically, when constructing the directed supply network diagram, all raw material supply relationship data is first extracted from the supplier management logs. The supplier hierarchy covers a three-tiered upstream scope, tracing raw material suppliers directly from the production end back to the upstream raw material production suppliers. The supplier entity information for each type of raw material is then compiled, clarifying the flow direction of materials from the upstream supply end to the downstream production end and the corresponding material categories. Each independent supplier entity is treated as a node in the network diagram, and each node is assigned a unique node number and a corresponding raw material category identifier. The material supply relationships between suppliers are represented as directed edges in the network diagram, extending from the material supplier to the material receiver, with the edge attributes including the annual supply quantity of the corresponding material. The network topology is then arranged according to the node numbers and the pointing relationships of the directed edges, forming a complete directed supply network diagram.

[0065] Next, all nodes in the directed supply network graph are traversed, and the basic attribute data corresponding to each node is extracted, including the supplier's annual production capacity, unit product energy consumption, and basic emission factor of the corresponding raw material category. The min-max normalization method is used to process the three types of attribute data separately. Specifically, the original value is subtracted from the minimum sample value of the corresponding attribute, and then divided by the difference between the maximum and minimum sample values ​​of the corresponding attribute, uniformly mapping it to a value range of zero to one. The initial feature vectors of the nodes are concatenated in a fixed order of annual production capacity, unit product energy consumption, and basic emission factor to form a 3-dimensional initial feature vector, with each node corresponding to an independent set of initial feature vectors.

[0066] When calculating attention weight coefficients for graph attention analysis, a graph attention network mechanism is used to perform operations node by node. Specifically: for each target node, the initial feature vectors of all its directly connected neighboring nodes are extracted. The initial feature vector of the target node is concatenated with the initial feature vectors of each neighboring node, resulting in a 6-dimensional feature vector. This concatenated feature vector is then input into a single-layer feedforward neural network for linear transformation. This single-layer feedforward neural network uses a single-weight layer feedforward architecture, consisting of an input layer, a linear transformation layer, and an output layer connected in series. During construction, the dimension of the input layer is set to match the dimension of the concatenated feature vector. The linear transformation layer is configured with a set of trainable weight vectors and a one-dimensional bias parameter, with the weight vector dimension matching the input layer dimension. The output layer dimension is set to 1, and the activation function is a linear rectified activation function. During the operation, the concatenated feature vector and weight vector undergo a dot product operation, and the bias parameter is superimposed. After processing by the activation function, the attention coefficient corresponding to that neighboring node is obtained. The attention coefficients of all neighboring nodes of the target node are normalized using a flexible maximum function to obtain the attention weight coefficient for each neighboring node. The weight coefficients of all neighboring nodes under the same target node are summed to 1. The attention weights of all nodes in the entire graph are calculated for their corresponding neighboring nodes according to the above process to obtain the attention weight coefficients for all supply relationships.

[0067] After calculating the attention weights, a feature weighting and aggregation operation is performed for each target node. The initial feature vectors of all neighboring nodes of the target node are multiplied by their corresponding attention weight coefficients to perform weighting of the single-node features. The weighted feature vectors of all neighboring nodes are then summed element-wise to obtain the aggregated neighborhood feature vector. Finally, the target node's own initial feature vector is summed and fused with the neighborhood feature vector element-wise, and then a linear rectified activation function is applied for a non-linear transformation to generate an updated node feature vector identifier. Each node corresponds to a unique node feature vector identifier.

[0068] Finally, the node feature vector identifier corresponding to each node is input into the pre-trained linear mapping layer. This linear mapping layer is a single-layer fully connected linear transformation structure, consisting of a trainable weight matrix and a 1-dimensional bias term. During construction, the input dimension is set to be consistent with the dimension of the node feature vector identifier, which is 3-dimensional, and the output dimension is 1-dimensional. The weight matrix has a size of 3 rows and 1 column. The weights are initialized randomly using a normal distribution with a mean of 0 and a standard deviation of 0.01. The initial value of the bias term is set to 0. The linear mapping layer is trained using historically accumulated supplier node feature data as input samples and emission factors obtained from actual calculations of the corresponding suppliers. Numerical values ​​serve as supervisory labels. The mean squared error loss function is used to measure the deviation between predicted and actual emission factors. An adaptive moment estimation optimizer is used for iterative parameter updates. The initial learning rate is set to 0.001, and the training batch size is set to 16. During training, the loss change is verified every 5 epochs using a validation set. Model training is considered converged when the validation set loss decreases by less than 0.0001 for 20 consecutive epochs. The final network weights and bias parameters are saved. A linear transformation is used to output the dynamic upstream emission factor values ​​for the corresponding supplier's raw materials. The units of the emission factors are consistent with those in the raw material carbon emission factor library. The dynamic upstream emission factors are synchronously stored in the raw material carbon emission factor library, replacing the original static emission factors for the corresponding raw materials, thus updating the raw material carbon emission factor library. Using the updated raw material carbon emission factor library, the unit product consumption data of the corresponding raw material is multiplied to calculate the carbon emission data for the raw material acquisition stage per unit product.

[0069] By constructing a directed supply network diagram covering three levels of upstream suppliers and combining it with a graph attention-weighted aggregation mechanism with clearly defined parameters, the upstream emission factors can be dynamically updated, accurately reflecting the differences in carbon emissions among different suppliers and improving the accuracy of carbon emission accounting at the raw material stage.

[0070] Furthermore, the method provided in this application embodiment includes: The carbon emissions from the raw material acquisition stage and the transportation stage are summed to obtain the total carbon emission data for consumables. Product attribute data is used to extract nominal print life parameters, which are then combined with historical print page count samples to perform maximum likelihood estimation and Weibull distribution calculation to obtain print shape parameters and print scale parameters. Actual print page count parameters are collected. When the actual print page count parameter reaches the nominal print life parameter, the nominal print life parameter is used as the allocation base. When the actual print page count parameter does not reach the nominal print life parameter, the nominal print life parameter is corrected based on the print shape parameters and the print scale parameters to generate an allocation base. Based on the total carbon emission data for consumables and the allocation base, a single-page print carbon footprint value is calculated and added to the multiple consumables carbon footprint values.

[0071] Specifically, firstly, the carbon emission data of the raw material acquisition stage and the carbon emission data of the transportation stage corresponding to a single digital smart office consumable are summed to obtain the total carbon emission data of the consumable corresponding to a single consumable, with the data unit being kilograms of carbon dioxide equivalent.

[0072] The nominal printing life parameter of this type of consumable is extracted from product attribute data, with the parameter unit being pages. Historical page count data of the same model of consumable is also collected, with a valid sample size of no less than two hundred groups. The maximum likelihood estimation method is used to fit the Weibull distribution. Specifically: first, the probability density function of the Weibull distribution is constructed, which includes two parameters to be solved: the printing shape parameter and the printing scale parameter. A log-likelihood function is constructed based on all historical page count sample data. The natural logarithm of the probability density function is taken, and the results are summed over all samples to obtain the complete expression of the log-likelihood function. The first-order partial derivatives of the printing shape parameter and the printing scale parameter are calculated and set to zero to construct a system of two equations. Newton's iteration method is used to solve this system of equations. The initial value of the printing shape parameter is set to 2, and the initial value of the printing scale parameter is set to 0.8 times the nominal printing life parameter. The iteration termination condition is set to the parameter difference between two iterations being less than 0.001. After the iteration converges, the final printing shape parameter and printing scale parameter are obtained.

[0073] Next, the actual number of pages printed for each consumable is collected, and this number is compared with the nominal printing life parameter. In the green product evaluation system, the nominal printing life parameter of a consumable is the number of pages that the manufacturer can achieve, measured and publicly declared under specified test conditions according to standard testing methods. Its function is positioned as a product performance commitment value, not an absolute upper limit of actual service life. Therefore, in the functional unit allocation logic of carbon footprint evaluation, it is reasonable to use the nominal printing life as the allocation base when the actual number of printed pages reaches or exceeds this commitment value: On the one hand, the number of printed pages exceeding the nominal life reflects the positive performance redundancy of the product in actual use. If the excess is included in the allocation base, it will lead to excessively large differences in carbon footprint values ​​for the same product under the same nominal life due to different actual usage scenarios, weakening the comparability of carbon footprint evaluation results between different consumable products; on the other hand, the allocation rule using the nominal life as the upper limit is a common processing method in the green product evaluation standard system. Its purpose is to unify the carbon footprint accounting benchmark to the publicly declared performance parameters of the product, rather than relying on the actual usage length of specific end users. This rule applies to all categories of digital smart office consumables with nominal print life parameters, including drum and toner cartridges, ink cartridges, and inkjet inks.

[0074] When the actual number of pages printed is greater than or equal to the nominal print life, the nominal print life is directly used as the allocation base for this consumable. When the actual number of pages printed is less than the nominal print life, a Weibull distribution reliability function is used for correction. The reliability function takes the ratio of the nominal print life to the print size parameter as the independent variable, calculates the print shape parameter power of this ratio, takes the negative value, and then takes the natural exponent of the result to obtain the reliability value. The nominal print life is then divided by this reliability value to obtain the corrected allocation base.

[0075] Finally, the total carbon emission data of consumables is divided by the allocation base to calculate the carbon footprint value of a single page of printing. The unit of the value is grams of carbon dioxide equivalent per page. The carbon footprint value of a single page of printing is added to the carbon footprint values ​​of multiple consumables to complete the quantitative calculation of the carbon footprint of a single type of consumable unit function.

[0076] Taking a certain model of drum and toner cartridge with a nominal printing life of 2000 pages as an example, the relationship between the allocated base and the carbon footprint per page under different actual printed pages is shown in Table 1.

[0077] Table 1: Comparison of Carbon Footprint Calculation Results for Different Actual Number of Printed Pages

[0078] Based on the data in the table above, a trend chart of single-page carbon footprint changes was further drawn, as shown below. Figure 2 As shown. Figure 2 With the actual number of printed pages as the horizontal axis and the carbon footprint per page as the vertical axis, the graph shows that as the actual number of printed pages increases, the carbon footprint per page first gradually decreases, and then remains stable after reaching the nominal printing life. This reflects the corrective effect of adjusting the allocation base when the nominal life is not reached.

[0079] By specifying the initial parameters and the Weibull distribution fitting method for calculation, the printing lifespan allocation base can be corrected to adapt to the carbon footprint allocation requirements under different actual use scenarios, thereby improving the rationality and accuracy of the carbon footprint accounting results per unit function.

[0080] Furthermore, the method provided in this application embodiment includes: Based on the multidimensional green product evaluation index library and the carbon footprint values ​​of multiple consumables, a hierarchical mapping analysis is performed to calculate the index importance judgment matrix. The index importance judgment matrix is ​​then optimized and corrected to generate an index importance judgment optimization matrix. Consistency analysis is performed based on the index importance judgment optimization matrix to calculate the judgment matrix eigenvector. This eigenvector is then normalized to generate subjective weight coefficients for multiple evaluation indicators. Multiple consumable sample indicators are extracted based on the raw material consumption data to construct a consumable sample indicator evaluation matrix. Carbon footprint fluctuation penalty adjustments are performed based on the consumable sample indicator evaluation matrix to calculate objective weight coefficients. Finally, a game-theoretic combination weighting is performed based on the subjective weight coefficients and the objective weight coefficients to generate the comprehensive weight coefficient.

[0081] In one embodiment, the construction process of the multi-dimensional green product evaluation index library is as follows: First, based on the evaluation dimensions and index requirements specified in the green product evaluation specifications, all evaluation indicators and their corresponding evaluation criteria required for green product evaluation of digital intelligent office consumables are determined. The evaluation indicators include four core categories: unit product carbon footprint, recyclable raw material ratio, comprehensive production energy consumption, and waste generation, covering the four dimensions of green product evaluation: resource attributes, energy attributes, environmental attributes, and low-carbon attributes. For each category of evaluation indicators, corresponding evaluation benchmark parameters, scoring thresholds, and scoring rules are configured. Specifically, the evaluation benchmark parameters for the unit product carbon footprint indicator include a benchmark reference value, an optimal threshold, and a passing threshold. The benchmark reference value is selected as the industry average single-page carbon footprint value for the same category of digital intelligent office consumables. The optimal threshold is set to 80% of the benchmark reference value, and the passing threshold is set to 120% of the benchmark reference value. The scoring rule adopts a deviation-based linear scoring rule. The three categories of indicators—recyclable raw material ratio, comprehensive production energy consumption, and waste generation—are each configured with their own corresponding evaluation benchmark parameters and scoring rules. All indicators, along with their corresponding evaluation benchmark parameters and scoring rules, are stored in a structured manner according to a unified field format, forming a complete multidimensional green product evaluation indicator library. This indicator library provides a unified indicator query interface and scoring calculation interface for subsequent hierarchical analysis, deviation calculation, and indicator score calculation steps.

[0082] The multidimensional green product evaluation index library includes four core evaluation indicators: carbon footprint per unit product, recyclable raw material ratio, comprehensive energy consumption in production, and waste generation. It uses the comprehensive green product evaluation target as the target layer of a hierarchical structure and the four evaluation indicators as the criterion layer, completing a hierarchical mapping analysis between indicators and evaluation targets. A 1-9 scale is used to compare the pairwise importance of indicators. Professional evaluators assign values ​​to the relative importance of each pair of indicators according to the green product evaluation requirements. The importance level ranges from equally important to extremely important, corresponding to integer scores from 1 to 9. All pairwise comparison scores are arranged according to row-column correspondence, with the rows and columns of the matrix corresponding to all evaluation indicators, forming an initial indicator importance determination matrix.

[0083] Next, the initial indicator importance judgment matrix is ​​optimized and corrected. First, the maximum eigenvalue of the matrix is ​​solved using the power method iteratively. Then, a consistency index is calculated based on the matrix order. The consistency index equals the maximum eigenvalue minus the matrix order, divided by the matrix order minus 1. A random consistency index corresponding to a 4th-order matrix is ​​matched, with a value of 0.89. This initial consistency ratio is calculated by dividing the initial consistency index by the random consistency index. When the consistency ratio is greater than 0.1, the matrix iterative correction process is initiated. The deviation of each off-diagonal element in the matrix from its corresponding position in the consistency matrix is ​​calculated. The element with the largest absolute deviation is selected as the adjustment target. The value of this element is adjusted by gradually fine-tuning the score towards adjacent integers. After each adjustment, the consistency ratio is recalculated, and this iteration is repeated until the consistency ratio is less than or equal to 0.1, ultimately generating the optimized indicator importance judgment matrix.

[0084] Consistency analysis is performed based on the indicator importance determination optimization matrix, and the sum-product method is used to calculate the eigenvector of the judgment matrix. Specifically: First, the values ​​in each column of the matrix are summed. Then, each element in the matrix is ​​divided by the sum of its column to complete the column normalization process. The normalized matrix is ​​then summed row-wise to obtain a row sum vector. Finally, each element of the row sum vector is divided by the sum of all elements in the vector to complete the second normalization process, resulting in the eigenvector of the judgment matrix. Each element in this vector corresponds to the weight percentage of an evaluation indicator, and the sum of all elements is 1, thus generating the subjective weight coefficients of multiple evaluation indicators.

[0085] Subsequently, multiple sets of indicator data corresponding to consumable samples were extracted from raw material consumption data and multiple consumable carbon footprint values. For each set of samples, actual values ​​of four indicators were extracted: unit product carbon footprint, raw material recyclability ratio, comprehensive production energy consumption, and waste generation. The unit product carbon footprint was directly taken from the previously calculated consumable carbon footprint value. The raw material recyclability ratio was obtained by calculating the proportion of the total mass of recyclable raw materials to the total mass of all raw materials from the product bill of materials data, sourced from raw material property parameter tables and product design material documents. The comprehensive production energy consumption was calculated from the energy consumption statistics of the production process, converting the electricity, heat, and other energy sources corresponding to the production of a single unit of consumable into standard coal equivalent. The waste generation was obtained by calculating the total mass of solid waste, waste liquid, and other waste generated per unit of consumable from the waste ledger data of the production process. Finally, an initial consumable sample indicator evaluation matrix was constructed with consumable samples as rows and evaluation indicators as columns.

[0086] Next, the indicator data within the matrix are standardized. For the three negative indicators—carbon footprint per unit product, comprehensive energy consumption in production, and waste generation—range standardization is used. The standardized value is equal to the maximum value of the corresponding indicator minus the current indicator value, then divided by the difference between the maximum and minimum values, mapping the value to the 0-1 range. The smaller the indicator value, the higher the standardized score. The positive indicator, the proportion of recyclable raw materials, is simultaneously mapped using range standardization. The standardized value is equal to the current indicator value minus the minimum value of the corresponding indicator, then divided by the difference between the maximum and minimum values, mapping to the same numerical range. This results in a standardized evaluation matrix for consumable sample indicators.

[0087] Carbon footprint fluctuation penalty adjustment is performed based on the consumable sample indicator evaluation matrix. First, the initial objective weight of each indicator is calculated using the entropy weight method. Then, the proportion of each sample value under each indicator is calculated sequentially. The proportion of a single sample is equal to the sample's indicator value divided by the sum of all sample values ​​for that indicator. Next, the information entropy value of each indicator is calculated based on the proportion value. Information entropy is equal to the reciprocal of the negative natural logarithm of the number of samples, multiplied by the sum of the products of the proportions and the natural logarithms of the proportions. The smaller the information entropy, the higher the sample discrimination of the indicator, and the larger the corresponding initial weight. A fluctuation penalty mechanism is introduced for the unit product carbon footprint indicator. The coefficient of variation of all sample values ​​for this indicator is calculated, which is the ratio of the sample standard deviation to the sample mean. After normalizing the coefficient of variation, the fluctuation penalty coefficient is obtained, with a value ranging from 0 to 1. The initial objective weight of the carbon footprint indicator is multiplied by 1 and the fluctuation penalty coefficient is subtracted to adjust the weight. Finally, the weights of all indicators are normalized again to obtain the final objective weight coefficient.

[0088] Finally, a game-theoretic combination weighting method is used based on subjective and objective weight coefficients. The subjective and objective weight vectors are used as two sets of basic weight vectors to construct a linear combined weight model. Weighting coefficients are set for each set of weights, and these coefficients satisfy a normalization constraint, meaning the sum of the two weighting coefficients equals 1. An objective function is established with the goal of minimizing the sum of squared deviations between the combined weights and the two sets of basic weights. Substituting the normalization constraint into the objective function to eliminate one weight coefficient, the first-order partial derivatives of the remaining weight coefficients are calculated and set to zero to obtain the optimal numerical solutions for the two weight coefficients. The two optimal weight coefficients are multiplied by their corresponding basic weight vectors and then summed to obtain the comprehensive weight coefficients. The sum of the comprehensive weight coefficients for each index is 1.

[0089] Subjective weights are generated by the analytic hierarchy process (AHP), and objective weights are generated by the entropy weight method with carbon footprint fluctuation penalty. The comprehensive weight coefficient is then obtained by game theory combination weighting. This approach can take into account both expert evaluation experience and actual sample data characteristics, effectively improving the rationality and objectivity of the green product evaluation weight system.

[0090] Furthermore, the method provided in this application embodiment includes: Based on the indicator score data, indicator conflict analysis is performed to calculate the conflict degree of multiple indicators; based on the indicator score data, internal dispersion analysis is performed to calculate the variability of multiple indicators; the subjective weight coefficient and the objective weight coefficient are adaptively adjusted according to the conflict degree and variability of multiple indicators to calculate the weight fusion ratio; the subjective weight coefficient and the objective weight coefficient are weighted and synthesized according to the weight fusion ratio to generate a combined weight parameter; based on the combined weight parameter, the multiple indicator score data are weighted and summed to obtain the carbon footprint evaluation score.

[0091] Optionally, an index conflict analysis can be conducted based on the index score data of all consumable samples. Specifically: for any two evaluation indicators, extract the score sequences of all samples on the two indicators. First, calculate the covariance of the two sets of score data, then calculate the sample standard deviation of each set of data. Divide the covariance by the product of the two standard deviations to obtain the Pearson correlation coefficient between the two indicators. The correlation coefficient ranges from -1 to 1. Take the absolute value of the calculated correlation coefficient. The closer the value is to 1, the higher the information overlap between the two indicators; the closer the value is to 0, the stronger the independence between the two indicators. For each evaluation indicator, calculate the arithmetic mean of the absolute values ​​of its correlation coefficients with all other indicators to obtain the average correlation coefficient for that indicator. Subtract the average correlation coefficient from 1 to obtain the conflict degree of the corresponding indicator. The larger the conflict degree value, the lower the information overlap between that indicator and other indicators, and the stronger its independent distinguishing value. Complete the calculation for all indicators in sequence to obtain multiple index conflict degrees.

[0092] Internal dispersion analysis is conducted based on the indicator score data. Specifically, for each evaluation indicator, the score data of all consumable samples for that indicator are extracted. First, the arithmetic mean of the data set is calculated, and then the sample standard deviation is calculated, which is the square root of the sum of the squares of the differences between each sample score and the mean, divided by the sample size minus 1. The sample standard deviation is then divided by the arithmetic mean to obtain the variability of that indicator. The variability is a dimensionless value; a larger value indicates a higher degree of dispersion of scores among different samples and a stronger ability to distinguish between different consumables. This calculation is performed for all evaluation indicators to obtain the variability of multiple indicators.

[0093] Next, the subjective and objective weight coefficients are adaptively adjusted according to the conflict and variability of multiple indicators, and the weight fusion ratio is calculated. First, min-max normalization is performed on the conflict degree of all indicators, mapping all conflict degree values ​​to a uniform range of 0 to 1, resulting in a normalized conflict degree. Simultaneously, min-max normalization is performed on the variability of all indicators, resulting in a normalized variability. As an optional implementation, for example, for each evaluation indicator, the influence ratio of conflict degree is set to 0.6, and the influence ratio of variability is set to 0.4. This ratio setting is based on common practices in green product evaluation scenarios, prioritizing the independent information value of evaluation indicators while also considering the discriminative power of sample data. The normalized conflict degree is multiplied by 0.6 and added to the result of multiplying the normalized variability by 0.4 to obtain the objective weight adjustment coefficient for that indicator, with a value ranging from 0 to 1. The objective weight adjustment coefficient is used as the fusion ratio of objective weights, and 1 minus the objective weight adjustment coefficient is used as the fusion ratio of subjective weights. This process is repeated to calculate the weight fusion ratio for all indicators.

[0094] Subsequently, the subjective and objective weight coefficients are weighted and synthesized according to the weight fusion ratio. For each evaluation indicator, the subjective weight coefficient corresponding to the indicator is multiplied by the subjective weight fusion ratio, and the objective weight coefficient corresponding to the indicator is multiplied by the objective weight fusion ratio. The two calculation results are added together to obtain the initial combined weight of the indicator. Normalization is performed on the initial combined weights of all indicators so that the sum of the combined weight parameters of all indicators is 1, completing the weight calibration, and finally obtaining the combined weight parameters corresponding to each indicator.

[0095] Finally, the carbon footprint evaluation score is obtained by weighting and summing the scores of multiple indicators based on the combined weight parameters. For a single digital smart office consumable, all corresponding indicator scores are extracted, and the score of each indicator is multiplied by the corresponding combined weight parameters. The products of all indicators are then summed to obtain the carbon footprint evaluation score for that consumable. The score is based on a percentage system, with a higher score indicating better overall green and low-carbon performance. This process is repeated for all the consumables tested to obtain the carbon footprint evaluation scores for all samples.

[0096] By adaptively adjusting the fusion ratio of subjective and objective weights through absolute value processing of the degree of conflict between indicators and internal dispersion variability, the basis for parameter setting and report output format are clarified. By combining weighted scoring to output evaluation results and standardized reports, the independent characteristics between indicators and the distinguishing features of sample data can be taken into account, effectively improving the scientificity and rationality of the comprehensive carbon footprint evaluation results.

[0097] In summary, the carbon footprint assessment method for digital intelligent office consumables provided in this application has the following technical effects: This application collects multi-source monitoring data from the entire process of consumables, and obtains carbon footprint and green indicator scores through graph attention networks, entropy weight game weighting, and other calculations. Combining the index conflict degree and variability, the weights are adaptively optimized, and a comprehensive weighted score is obtained to obtain the carbon footprint evaluation score. This accurately quantifies the low-carbon performance of office consumables, making the carbon footprint evaluation results of digital intelligent office consumables more objective. It achieves the technical effect of improving the accuracy and dynamic adaptability of carbon footprint evaluation, making the scoring results more accurately reflect the real environmental performance of consumables and adapt to the requirements of green product evaluation.

[0098] Example 2, as Figure 3 As shown, based on the same inventive concept as in Embodiment 1 above, this application provides a carbon footprint assessment system for digital intelligent office consumables, the system comprising: Multi-source dataset acquisition module 1 is used to intelligently collect multi-source data from digital intelligent office consumables and automatically capture multi-source datasets.

[0099] Consumable carbon footprint value acquisition module 2 is used to construct a carbon footprint quantification channel, and to calculate the life cycle carbon footprint of the multi-source dataset through the carbon footprint quantification channel to obtain multiple consumable carbon footprint values.

[0100] The comprehensive weight coefficient generation module 3 is used to construct a multi-dimensional green product evaluation index library and perform hierarchical analysis by combining the carbon footprint values ​​of the multiple consumables to generate a comprehensive weight coefficient, which includes a subjective weight coefficient and an objective weight coefficient.

[0101] The indicator score data calculation module 4 performs deviation calculations based on the multiple consumable carbon footprint values ​​to obtain multiple deviation degrees, which are then matched with a multidimensional green product evaluation indicator library to calculate multiple indicator score data.

[0102] The carbon footprint assessment report generation module 5 is used to weight the scores of the multiple indicators according to the subjective weight coefficient and the objective weight coefficient to obtain a carbon footprint assessment score, and generate a carbon footprint assessment report based on the carbon footprint assessment score.

[0103] Furthermore, the multi-source dataset acquisition module 1 is used to perform the following steps: Based on digital intelligent office supplies, multiple data are automatically captured to obtain an initial raw material consumption sequence. The initial raw material consumption sequence is traversed to identify and remove outliers, obtaining missing location parameters. Based on the missing location parameters, time-series data is filled in to generate raw material consumption data. Optical character scanning and parsing are performed on the digital intelligent office supplies to obtain original logistics document text parameters. The original logistics document text parameters are traversed to extract information, obtaining structured transportation information, including transportation origin data and transportation destination data. When the transportation origin data or the transportation destination data is fuzzy text, an address resolution signal is generated. A map distance matrix is ​​constructed using the address resolution signal, and multiple candidate path distance parameters are determined. Based on the multiple candidate path distance parameters and historical transportation trajectories, dynamic time warping is performed to generate transportation distance data. The raw material consumption data and the transportation distance data are added to the multi-source dataset.

[0104] Furthermore, the multi-source dataset acquisition module 1 is used to perform the following steps: A generator is constructed by stacking causal temporal convolutional network layers and multi-head self-attention layers. The generator captures time series containing long-range dependencies. A discriminator is constructed based on a convolutional neural network, and the discriminator is used to discriminate and train the time series: S1: The generator reconstructs the time series according to the long-range dependencies, and performs adversarial loss optimization based on the imputed time series until the discriminator can no longer distinguish between the imputed time series and the original time series. A generative adversarial network model is then constructed. The missing positions are synchronized to the generative adversarial network model to impute the time series data and generate the raw material consumption data.

[0105] Furthermore, the consumable carbon footprint value acquisition module 2 is used to perform the following steps: A raw material carbon emission factor library is constructed, and the raw material consumption data is associated with the raw material carbon emission factor library to calculate the carbon emission data of the raw material acquisition stage. Based on the transportation distance data and the emission factor of the transportation mode, the carbon emission data of the transportation stage is obtained. The carbon footprint quantification channel is used to allocate the carbon emission data of the raw material acquisition stage and the transportation distance data to functional units to calculate the carbon footprint values ​​of the multiple consumables.

[0106] Furthermore, the consumable carbon footprint value acquisition module 2 is used to perform the following steps: Raw material supply relationship data is extracted from supplier management logs. Material flow relationships are then extracted from this data, and multiple supplier data points are located. These supplier data points are used as nodes, and the material flow relationships are used as edges to construct a directed supply network graph. Feature analysis is performed on the directed supply network graph to construct initial feature vectors for multiple nodes. Graph attention analysis is then performed on these initial feature vectors to calculate attention weight coefficients. The initial feature vectors are then weighted and aggregated according to these attention weight coefficients. The weighted aggregation results are used to update the initial feature vectors, generating multiple node feature vector identifiers. These node feature vector identifiers are mapped to dynamic upstream emission factors. These dynamic upstream emission factors are stored in the raw material carbon emission factor library, replacing static emission factors. The raw material carbon emission factor library is then updated, and carbon emission calculations are performed using the raw material consumption data to obtain carbon emission data for the raw material acquisition stage.

[0107] Furthermore, the consumable carbon footprint value acquisition module 2 is used to perform the following steps: The carbon emissions from the raw material acquisition stage and the transportation stage are summed to obtain the total carbon emission data for consumables. Product attribute data is used to extract nominal print life parameters, which are then combined with historical print page count samples to perform maximum likelihood estimation and Weibull distribution calculation to obtain print shape parameters and print scale parameters. Actual print page count parameters are collected. When the actual print page count parameter reaches the nominal print life parameter, the nominal print life parameter is used as the allocation base. When the actual print page count parameter does not reach the nominal print life parameter, the nominal print life parameter is corrected based on the print shape parameters and the print scale parameters to generate an allocation base. Based on the total carbon emission data for consumables and the allocation base, a single-page print carbon footprint value is calculated and added to the multiple consumables carbon footprint values.

[0108] Furthermore, the comprehensive weight coefficient generation module 3 is used to perform the following steps: Based on the multidimensional green product evaluation index library and the carbon footprint values ​​of multiple consumables, a hierarchical mapping analysis is performed to calculate the index importance judgment matrix. The index importance judgment matrix is ​​then optimized and corrected to generate an index importance judgment optimization matrix. Consistency analysis is performed based on the index importance judgment optimization matrix to calculate the judgment matrix eigenvector. This eigenvector is then normalized to generate subjective weight coefficients for multiple evaluation indicators. Multiple consumable sample indicators are extracted based on the raw material consumption data to construct a consumable sample indicator evaluation matrix. Carbon footprint fluctuation penalty adjustments are performed based on the consumable sample indicator evaluation matrix to calculate objective weight coefficients. Finally, a game-theoretic combination weighting is performed based on the subjective weight coefficients and the objective weight coefficients to generate the comprehensive weight coefficient.

[0109] Furthermore, the carbon footprint assessment report generation module 5 is used to perform the following steps: Based on the indicator score data, indicator conflict analysis is performed to calculate the conflict degree of multiple indicators; based on the indicator score data, internal dispersion analysis is performed to calculate the variability of multiple indicators; the subjective weight coefficient and the objective weight coefficient are adaptively adjusted according to the conflict degree and variability of multiple indicators to calculate the weight fusion ratio; the subjective weight coefficient and the objective weight coefficient are weighted and synthesized according to the weight fusion ratio to generate a combined weight parameter; based on the combined weight parameter, the multiple indicator score data are weighted and summed to obtain the carbon footprint evaluation score.

[0110] The carbon footprint assessment system for digital intelligent office consumables provided in this embodiment of the invention can execute the carbon footprint assessment method for digital intelligent office consumables provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0111] In embodiment three, this application also provides a computer-readable storage medium that can be used to store software programs, computer-executable programs, and modules, such as program instructions / modules corresponding to a carbon footprint assessment system for digital intelligent office consumables in this application embodiment, thereby realizing the aforementioned carbon footprint assessment method for digital intelligent office consumables.

[0112] It should be understood that the embodiments disclosed in this application and the above description enable those skilled in the art to implement this application. However, this application is not limited to the embodiments mentioned above. It should be understood that those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this invention, and should all be included within the protection scope of this application.

Claims

1. A method for evaluating the carbon footprint of digital intelligent office consumables, characterized in that, The method includes: Intelligent multi-source data collection for digital intelligent office supplies, automatically capturing multi-source datasets; A carbon footprint quantification channel is constructed, and the life cycle carbon footprint of the multi-source dataset is calculated through the carbon footprint quantification channel to obtain multiple consumable carbon footprint values. A multidimensional green product evaluation index library is constructed, and a hierarchical analysis is performed on the carbon footprint values ​​of the multiple consumables to generate a comprehensive weight coefficient, which includes subjective weight coefficient and objective weight coefficient. Based on the carbon footprint values ​​of the multiple consumables, deviation calculations are performed to obtain multiple deviation degrees, which are then matched with a multidimensional green product evaluation index library to calculate multiple index scores. The scores of the multiple indicators are weighted according to the subjective weight coefficient and the objective weight coefficient to obtain a carbon footprint evaluation score, and a carbon footprint evaluation report is generated based on the carbon footprint evaluation score.

2. The carbon footprint assessment method for digital intelligent office consumables as described in claim 1, characterized in that, Multi-source intelligent data collection for digital intelligent office supplies, automatically capturing multi-source datasets, includes the following methods: Based on digital intelligent office consumables, multiple data are automatically captured to obtain the initial raw material consumption sequence; The initial raw material consumption sequence is traversed to identify and remove outliers, and parameters for the missing removal locations are obtained. Based on the missing location parameters, time-series data is filled in to generate raw material consumption data; Optical character scanning and parsing based on digital intelligent office supplies were used to obtain the original logistics document text parameters. The original logistics document text parameters are traversed to extract information and obtain transportation structured information, which includes transportation origin data and transportation destination data. When the transportation origin data or the transportation destination data is fuzzy text, an address resolution signal is generated. A map distance matrix is ​​constructed using the address resolution signal to calculate and determine the distance parameters of multiple candidate paths. Based on the distance parameters of the multiple candidate paths and combined with historical transportation trajectories, dynamic time warping is performed to generate transportation distance data; The raw material consumption data and the transportation distance data are added to the multi-source dataset.

3. The carbon footprint assessment method for digital intelligent office consumables as described in claim 2, characterized in that, Based on the missing location parameters, time-series data is filled in to generate raw material consumption data. The method includes: A generator is constructed by stacking causal temporal convolutional network layers and multi-head self-attention layers. The generator captures time series containing long-range dependencies; A discriminator is constructed based on a convolutional neural network, and the discriminator is used to train the time series data for discrimination. S1: The generator reconstructs the time series according to the long-range dependency, and performs adversarial loss optimization based on the imputed time series until the discriminator can no longer distinguish between the imputed time series and the time series, thus constructing a generative adversarial network model. The missing positions are synchronized to the generative adversarial network model to fill in the time-series data and generate the raw material consumption data.

4. The carbon footprint assessment method for digital intelligent office consumables as described in claim 2, characterized in that, The carbon footprint quantification channel is used to calculate the lifecycle carbon footprint of the multi-source dataset to obtain multiple consumable carbon footprint values. The method includes: Construct a raw material carbon emission factor library, associate the raw material consumption data with the raw material carbon emission factor library, and calculate the carbon emission data during the raw material acquisition stage; Carbon emission data for the transportation phase is obtained by combining the transportation distance data with the emission factor of the transportation mode. The carbon footprint quantification channel is used to allocate carbon emission data from the raw material acquisition stage and transportation distance data to functional units, and to calculate the carbon footprint values ​​of the multiple consumables.

5. The carbon footprint assessment method for digital intelligent office consumables as described in claim 4, characterized in that, The method involves associating the raw material consumption data with the raw material carbon emission factor database to calculate the carbon emission data during the raw material acquisition stage. Raw material supply relationship data is extracted from supplier management logs. Material flow relationships are extracted from the raw material supply relationship data. Multiple supplier data are located, and the multiple supplier data are used as nodes. The material flow relationships are used as edges to construct a directed supply network graph. The directed supply network graph is traversed to perform feature analysis and construct initial feature vectors for multiple nodes; Graph attention analysis is performed based on the initial feature vectors of the multiple nodes, and attention weight coefficients are calculated. The initial feature vectors of the multiple nodes are weighted and aggregated according to the attention weight coefficients, and the initial feature vectors of the multiple nodes are updated according to the weighted aggregation result to generate multiple node feature vector identifiers. The multiple node feature vector identifiers are mapped to dynamic upstream emission factors. The dynamic upstream emission factors are stored in the raw material carbon emission factor library to replace the static emission factors. The raw material carbon emission factor library is updated and carbon emission calculation is performed in combination with the raw material consumption data to obtain the carbon emission data of the raw material acquisition stage.

6. The carbon footprint assessment method for digital intelligent office consumables as described in claim 4, characterized in that, The carbon footprint quantification channel is used to allocate carbon emission data from the raw material acquisition stage and transportation distance data to functional units to calculate multiple consumable carbon footprint values. The method includes: The carbon emissions from the raw material acquisition stage and the carbon emissions from the transportation stage are added together to obtain the total carbon emissions data for consumables. By introducing product attribute data to extract nominal printing life parameters and combining them with historical printed page count sample parameters, maximum likelihood estimation is performed to fit a Weibull distribution and obtain printing shape parameters and printing scale parameters. Collect the actual number of pages printed. When the actual number of pages printed reaches the nominal printing life parameter, the nominal printing life parameter is used as the amortization base. When the actual number of printed pages parameter does not reach the nominal printing life parameter, the nominal printing life parameter is corrected based on the printing shape parameter and the printing scale parameter to generate an amortization base. Based on the total carbon emission data of the consumables and the allocation base, a single-page printing carbon footprint value is calculated, and the single-page printing carbon footprint value is added to the multiple consumables carbon footprint values.

7. The carbon footprint assessment method for digital intelligent office consumables as described in claim 2, characterized in that, A multidimensional green product evaluation index library is constructed, and a hierarchical analysis is performed on the carbon footprint values ​​of the aforementioned consumables to generate a comprehensive weighting coefficient. The method includes: Based on the multidimensional green product evaluation index library and the carbon footprint values ​​of the multiple consumables, a hierarchical mapping analysis is performed to calculate the index importance judgment matrix; Based on the aforementioned indicator importance determination matrix, optimization and correction are performed to generate an indicator importance determination optimization matrix; Consistency analysis is performed based on the indicator importance judgment optimization matrix, the eigenvector of the judgment matrix is ​​calculated, the eigenvector of the judgment matrix is ​​normalized, and the subjective weight coefficients of multiple evaluation indicators are generated. Based on the raw material consumption data, multiple consumable sample indicators are extracted, and a consumable sample indicator evaluation matrix is ​​constructed. Based on the evaluation matrix of the consumable sample indicators, carbon footprint fluctuation penalty adjustment is performed, and objective weight coefficients are calculated. The comprehensive weight coefficient is generated by combining the subjective weight coefficient and the objective weight coefficient through a game theory approach.

8. The carbon footprint assessment method for digital intelligent office consumables as described in claim 1, characterized in that, The carbon footprint evaluation score is obtained by weighting the scores of the multiple indicators according to the subjective weight coefficient and the objective weight coefficient, and the method includes: Based on the index score data, an index conflict analysis is performed to calculate the conflict degree of multiple indicators. Internal dispersion analysis is performed based on the index score data to calculate the variability of multiple indicators; The subjective weight coefficient and the objective weight coefficient are adaptively adjusted according to the conflict degree of the multiple indicators and the variability of the multiple indicators, and the weight fusion ratio is calculated. The subjective weight coefficient and the objective weight coefficient are weighted and synthesized according to the weight fusion ratio to generate a combined weight parameter. The carbon footprint evaluation score is obtained by weighting and summing the scores of the multiple indicators based on the combined weight parameters.

9. A carbon footprint assessment system for digital intelligent office consumables, characterized in that, A carbon footprint assessment method for implementing a digital intelligent office consumable as described in any one of claims 1-8, the system comprising: The multi-source dataset acquisition module is used to intelligently collect multi-source data on digital intelligent office consumables and automatically capture multi-source datasets. The consumable carbon footprint value acquisition module is used to construct a carbon footprint quantification channel, and to perform life cycle carbon footprint calculation on the multi-source dataset through the carbon footprint quantification channel to obtain multiple consumable carbon footprint values. The comprehensive weight coefficient generation module is used to construct a multi-dimensional green product evaluation index library and perform hierarchical analysis by combining the carbon footprint values ​​of the multiple consumables to generate a comprehensive weight coefficient, which includes subjective weight coefficients and objective weight coefficients. The indicator score data calculation module performs deviation calculations based on the multiple consumable carbon footprint values, obtains multiple deviation degrees, matches them with the multidimensional green product evaluation indicator library, and calculates multiple indicator score data. The carbon footprint assessment report generation module is used to weight the scores of the multiple indicators according to the subjective weight coefficient and the objective weight coefficient to obtain a carbon footprint assessment score, and generate a carbon footprint assessment report based on the carbon footprint assessment score.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements a method for evaluating the carbon footprint of digital intelligent office consumables as described in any one of claims 1 to 8.

Images