A dynamic carbon efficiency evaluation and carbon efficiency code generation method and system

Abstract

The present application belongs to the technical field of industrial carbon emission data governance, and aims to solve the problems of serious distortion of evaluation results, deviation of rating from actual operation level, and poor rating continuity and stability caused by frequent rating reversal in existing carbon efficiency evaluation. The method of the present application comprises: collecting multi-source production and energy carbon data, and aligning the multi-source production and energy carbon data on a unified time axis; obtaining current working condition information, which includes product structure characteristics, production line load rate and energy quality parameters; calculating the equivalent carbon intensity under the standard working condition based on the multi-source production and energy carbon data and the working condition information; generating historical equivalent carbon intensity data based on the equivalent carbon intensity; calculating the carbon efficiency benchmark intensity of the current evaluation period based on the historical equivalent carbon intensity data; calculating the deviation degree of the equivalent carbon intensity and the carbon efficiency benchmark intensity; presetting a trapezoidal membership function; performing fuzzy grade judgment based on the deviation degree and the trapezoidal membership function and outputting the anti-reversal carbon efficiency code.

Description

Technical Field

[0001] This invention belongs to the field of industrial carbon emission data governance technology, specifically relating to a method and system for dynamic carbon efficiency evaluation and carbon efficiency code generation. Background Technology

[0002] Carbon efficiency assessment is a process of comprehensively calculating and evaluating the carbon emission intensity generated by an enterprise or energy-consuming unit during its energy consumption and production and operation processes over a certain period. Its core indicator is carbon emissions per unit of added value, that is, the amount of carbon dioxide emitted per unit of output value. Through carbon efficiency assessment, enterprises can systematically analyze their own carbon emission levels, identify emission reduction potential, and promote technological innovation and production process optimization.

[0003] Currently, carbon efficiency assessments for industrial enterprises largely rely on annual or monthly macroeconomic emission intensity, such as classifying levels based on carbon emissions per unit of product or per unit of output value against fixed policy thresholds. Existing carbon efficiency assessments are mostly limited to data collection, static calculation, and rule lookup, employing static level classification methods that fail to address the issues of comparability across multiple operating conditions and dynamic evolution of benchmarks. Under multi-variety, small-batch production models, product BOM structures, process routes, and equipment loads vary significantly, leading to severe distortion in assessment results when directly comparing original carbon intensity. Fixed benchmark values ​​cannot adaptively adjust to real-time operating conditions such as production line load fluctuations, energy structure switching, and equipment energy efficiency degradation, causing ratings to deviate from actual operating levels. Furthermore, the currently used rigid thresholds are prone to triggering frequent level reversals when data fluctuates even slightly, failing to meet the requirements of green credit contracts or supply chain assessments for rating continuity and stability. Summary of the Invention

[0004] (a) Purpose of the invention The purpose of this invention is to provide a dynamic carbon efficiency evaluation and carbon efficiency code generation method and system, which aims to solve the problems of serious distortion of evaluation results, rating deviating from actual operating levels, and poor rating continuity and stability caused by frequent rating reversals in existing carbon efficiency evaluations.

[0005] (II) Technical Solution To address the above problems, a first aspect of the present invention provides a method for dynamic carbon efficiency evaluation and carbon efficiency code generation, the method comprising: Collect multi-source production and energy carbon data, and align the multi-source production and energy carbon data on a unified time axis; Obtain current operating condition information, including product structure characteristics, production line load rate and energy quality parameters; calculate the equivalent carbon intensity under standard operating conditions based on the multi-source production and energy carbon data and the operating condition information; and generate historical equivalent carbon intensity data based on the equivalent carbon intensity. The carbon efficiency benchmark intensity for the current evaluation period is calculated based on the historical equivalent carbon intensity data. Calculate the deviation between the equivalent carbon intensity and the carbon efficiency benchmark intensity, preset a trapezoidal membership function, perform fuzzy level judgment based on the deviation and the trapezoidal membership function, and output the anti-flip carbon efficiency code.

[0006] Preferably, aligning the multi-source production and energy-carbon data on a unified timeline includes: Using the production work order number as the time anchor, the multi-source production and energy carbon data are sliced ​​and aggregated according to a preset time granularity to form a time-synchronized aligned dataset. The preset time granularity is 15 minutes or 1 hour.

[0007] Preferably, the method further includes: When collecting the multi-source production and energy carbon data, missing data is linearly interpolated using the historical average of similar operating conditions. Work orders with more than 3 consecutive missing time granularities are marked as invalid data and skipped from the current period calculation.

[0008] Preferably, the step of calculating the equivalent carbon intensity under standard operating conditions based on the multi-source production and energy carbon data and the operating condition information includes: A multidimensional normalized correction model is constructed based on the aforementioned working condition information; The equivalent carbon intensity is calculated based on the original unit product carbon emission intensity of the multi-source production and energy carbon data and the modified model.

[0009] Preferably, the equivalent carbon intensity is calculated based on the original unit product carbon emission intensity of the multi-source production and energy carbon data and the modified model, using the following formula: Among them, I raw (t) represents the original unit product carbon emission intensity. The normalization coefficient for the product structural features is... This is a correction factor for the production line load rate. This is a correction factor for the energy quality parameter.

[0010] Preferably, the step of calculating the carbon efficiency benchmark intensity for the current evaluation period based on the historical equivalent carbon intensity data includes: Outlier removal was performed on the historical equivalent carbon intensity data; The dynamically rolling carbon efficiency benchmark intensity is calculated using an exponentially weighted moving average combined with a low quantile truncation strategy.

[0011] Preferably, the carbon efficiency benchmark intensity is calculated using an exponentially weighted moving average combined with a low quantile truncation strategy, and the calculation formula is as follows: ) Where β is the baseline decay factor, B(t-1) is the baseline intensity of the previous evaluation period, W is the rolling time window length, set to 7-30 days according to the production line process cycle, and P 25 This represents the 25th percentile of the equivalent carbon intensity sequence within the window.

[0012] Preferably, the formula for calculating the deviation between the equivalent carbon intensity and the carbon efficiency benchmark intensity is as follows: in, For equivalent carbon intensity, This serves as the carbon efficiency benchmark intensity.

[0013] Preferably, the preset trapezoidal membership function, based on the deviation and the trapezoidal membership function, performs fuzzy level determination, including: The preset trapezoidal membership function consists of five trapezoidal membership function groups corresponding to five carbon efficiency levels. The upper and lower boundary parameters of the trapezoidal membership function groups are set based on the quantiles of the historical deviation distribution in the industry. When the absolute value of the difference in membership degree between two adjacent levels is less than a preset threshold, a weighted centroid decoupling calculation is triggered to output the final level.

[0014] Another aspect of the present invention provides a dynamic carbon efficiency evaluation and carbon efficiency code generation system, the system comprising: Data acquisition and alignment module: used to acquire multi-source production and energy carbon data, and align the multi-source production and energy carbon data on a unified time axis; Operating condition normalization module: used to obtain current operating condition information, including product structure characteristic parameters, production line load rate and energy quality parameters, calculate the equivalent carbon intensity under standard operating conditions based on the multi-source production and energy carbon data and the operating condition information, and generate historical equivalent carbon intensity data based on the equivalent carbon intensity. Benchmark generation module: used to generate the benchmark intensity for the current evaluation period based on the historical equivalent carbon intensity data; Deviation mapping module: used to calculate the deviation between the equivalent carbon intensity and the reference intensity, preset a trapezoidal membership function, perform fuzzy level judgment based on the deviation and the trapezoidal membership function, and output the anti-flip carbon efficiency code.

[0015] (III) Beneficial Effects The above-described technical solution of the present invention has the following beneficial technical effects: 1. By introducing corrections based on three dimensions—product structure characteristics, production line load rate, and energy quality parameters—the raw carbon intensity measured under different operating conditions is uniformly mapped to a standard operating condition benchmark. This provides a technical basis for horizontal comparability of carbon intensity data across different product batches, load states, and energy structures. After normalization, enterprises can accurately identify the true energy efficiency level of each product in scenarios with multiple products produced in rotation, providing a quantitative basis for refined energy management; they can stably assess operating efficiency under fluctuating production line loads, unaffected by load changes caused by order variations; and they can continuously track energy-using technology levels in the context of dynamic adjustments to the energy structure, separating the impact of energy procurement strategies on emission intensity. The establishment of a unified benchmark upgrades carbon efficiency evaluation from simple absolute value calculation to a refined assessment of relative levels, improving the fairness and reference value of the evaluation results, and providing a technical foundation for industry-level benchmarking and classification.

[0016] 2. A technical solution that dynamically generates benchmark intensity based on historical equivalent carbon intensity data is adopted. This allows the benchmark to slowly drift and evolve with the actual operating level of the production line. The dynamic benchmark avoids frequent jumps due to short-term data fluctuations, maintaining the relative stability of the evaluation standard, while also preventing it from becoming rigid and ineffective due to long-term fixation. It can synchronously reflect baseline shifts caused by changes in equipment status, the effectiveness of process improvements, and adjustments in energy structure. For enterprises continuously promoting energy-saving renovations, the benchmark can steadily shift downwards with energy efficiency improvements, ensuring that carbon efficiency evaluation always anchors to the current advanced level and continues to play a benchmark guiding role. For production lines with naturally aging equipment, the benchmark can also be reasonably adjusted to avoid systematic deviations in rating results.

[0017] 3. A trapezoidal membership function is used for fuzzy level judgment, establishing a continuous transition zone between adjacent levels. This allows the membership degree to change smoothly with the deviation rather than abruptly. When the data is in the transition range, a comprehensive judgment is made through fuzzy logic, outputting an anti-flip carbon efficiency code. This prevents the carbon efficiency level from frequently switching due to measurement fluctuations and short-term disturbances common in industrial settings. Stable and continuous rating results can meet the engineering requirements for rating consistency in application scenarios such as green credit contract performance, daily supply chain assessment, and internal enterprise performance management. Furthermore, the rating results better reflect the gradual changes in energy efficiency levels, which is more in line with the objective law of continuous and gradual changes in industrial production processes. Attached Figure Description

[0018] Figure 1 This is a flowchart of a dynamic carbon efficiency evaluation and carbon efficiency code generation method provided by the present invention; Figure 2 This is a flowchart of a dynamic carbon efficiency evaluation and carbon efficiency code generation method according to one embodiment of the present invention; Figure 3 This is a logic diagram of fuzzy threshold mapping and level determination according to one embodiment of the present invention; Figure 4 This is a schematic diagram of a dynamic carbon efficiency evaluation and carbon efficiency code generation system architecture provided by the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments and the accompanying drawings. It should be understood that these descriptions are merely exemplary and not intended to limit the scope of the invention. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concept of the invention.

[0020] The accompanying drawings illustrate layer structure diagrams according to embodiments of the present invention. These drawings are not to scale, and some details have been enlarged for clarity, and some details may have been omitted. The shapes of the various regions and layers shown in the drawings, as well as their relative sizes and positional relationships, are merely exemplary and may deviate from reality due to manufacturing tolerances or technical limitations. Furthermore, those skilled in the art can design regions / layers with different shapes, sizes, and relative positions as needed.

[0021] Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0022] Combination Figures 1 to 4 The first aspect of the present invention provides a method for dynamic carbon efficiency evaluation and carbon efficiency code generation, the method comprising: Collect multi-source production and energy carbon data, and align the multi-source production and energy carbon data on a unified timeline; Obtain current operating condition information, including product structure characteristics, production line load rate and energy quality parameters. Calculate the equivalent carbon intensity under standard operating conditions based on multi-source production and energy carbon data and operating condition information. Generate historical equivalent carbon intensity data based on the equivalent carbon intensity. The carbon efficiency benchmark intensity for the current evaluation period is calculated based on historical equivalent carbon intensity data. Calculate the deviation between the equivalent carbon intensity and the carbon efficiency benchmark intensity, preset the trapezoidal membership function, perform fuzzy level judgment based on the deviation and the trapezoidal membership function, and output the anti-flip carbon efficiency code.

[0023] Specifically, heterogeneous data sources such as manufacturing execution systems, energy management systems, and green electricity trading platforms are connected via industrial IoT protocols. Due to differences in sampling frequency, timestamp accuracy, and data triggering mechanisms among these systems, the original data suffers from temporal misalignment. This step uses the production work order number as a unified time anchor, slicing, resampling, and aggregating data from different sources according to a preset time granularity. This ensures that energy consumption data, output data, and process parameters correspond one-to-one in the time dimension, providing a spatiotemporally consistent input data foundation for subsequent calculations. The model reads three types of operating parameters corresponding to the current work order: product structure characteristics, production line load rate, and energy quality, and constructs a multi-dimensional normalized correction model. Differences in the material structure of different products lead to different benchmark energy consumption per unit product; production line load rate affects equipment operating efficiency range; and differences in energy type and quality directly affect emission factors. Through coefficient correction in these three dimensions, the original carbon emission intensity measured under actual operating conditions is converted to an equivalent value under a unified standard operating condition, eliminating the incomparability caused by differences in operating conditions. Based on historically accumulated equivalent carbon intensity sequence data, the benchmark intensity value for the current evaluation period is calculated. This benchmark is not a fixed policy threshold, but a dynamically evolving reference baseline based on historical operational data. It tracks the changing trends in production line energy efficiency levels and consistently anchors to a reasonable reference point under current technological conditions. The deviation between the current equivalent carbon intensity and the dynamic benchmark is calculated, and the deviation value is input into a preset trapezoidal membership function group. Unlike the binary decision-making of traditional rigid thresholds, the trapezoidal membership function sets a transition interval at the level boundaries, allowing the membership degree to change continuously. When the data point is near the boundary, a stable carbon efficiency code level is output through a fuzzy decision mechanism, making the rating results robust to data noise within a normal range.

[0024] This approach, by incorporating corrections across three dimensions—product structure characteristics, production line load rate, and energy quality parameters—maps raw carbon intensity measured under different operating conditions onto a standardized benchmark. This provides a technical basis for horizontally comparable carbon intensity data across different product batches, load states, and energy structures. After normalization, companies can accurately identify the true energy efficiency levels of each product in scenarios involving multiple product lines in rotation, providing a quantitative basis for refined energy management. It also enables stable assessment of operational efficiency under fluctuating production line loads, unaffected by load changes caused by order fluctuations. Furthermore, it allows for continuous tracking of energy technology levels within the context of dynamic energy structure adjustments, separating the impact of energy procurement strategies on emission intensity. The establishment of a unified benchmark upgrades carbon efficiency evaluation from simple absolute value calculation to a refined assessment of relative levels, enhancing the fairness and reference value of the evaluation results and providing a technical foundation for industry-level benchmarking and grading. The technical solution employs a dynamic benchmark intensity generation based on historical equivalent carbon intensity data. This allows the benchmark to slowly drift and evolve with the actual operating level of the production line. The dynamic benchmark avoids frequent jumps due to short-term data fluctuations, maintaining the relative stability of the evaluation standard, while also preventing it from becoming rigid and ineffective due to long-term fixation. It can synchronously reflect baseline shifts caused by changes in equipment status, the effectiveness of process improvements, and adjustments in energy structure. For enterprises continuously promoting energy-saving renovations, the benchmark can steadily shift downwards with energy efficiency improvements, ensuring that carbon efficiency evaluations always anchor to the current advanced level and continuously play a benchmark guiding role. For production lines with naturally aging equipment, the benchmark can also be reasonably adjusted to avoid systematic deviations in rating results. A trapezoidal membership function is used for fuzzy level judgment, establishing a continuous transition zone between adjacent levels. This allows the membership degree to change smoothly with the deviation rather than abruptly. When the data is in the transition range, fuzzy logic is used for comprehensive judgment, outputting a flip-resistant carbon efficiency code. This prevents the carbon efficiency level from frequently switching due to measurement fluctuations and short-term disturbances common in industrial settings. Stable and continuous rating results can meet the engineering requirements for rating consistency in application scenarios such as green credit contract performance, daily supply chain assessment, and internal enterprise performance management. Furthermore, the rating results better reflect the gradual changes in energy efficiency levels, which is more in line with the objective law of continuous and gradual changes in industrial production processes.

[0025] In a preferred embodiment, aligning multi-source production with energy and carbon data on a unified timeline includes: Using the production work order number as the time anchor, multi-source production and energy carbon data are sliced ​​and aggregated according to a preset time granularity to form a time-synchronized aligned dataset. The preset time granularity is 15 minutes or 1 hour.

[0026] Specifically, the production work order number is a unique identifier throughout the entire production batch in the Manufacturing Execution System (MES). Each work order corresponds to the complete production cycle of a specific product from start to finish. Using the work order number as a time anchor means using the start and end times of that work order as the time frame for data alignment. After acquiring the raw data from various data sources, the system first retrieves the corresponding production time period based on the work order number. Then, all relevant data within that time period is sliced ​​at fixed time intervals of 15 minutes or 1 hour. For each time slice, the system integrates and summarizes energy consumption data, statistically accumulates output data, and extracts the mean or feature values ​​of process parameters, ensuring that each time slice contains a set of synchronized multi-dimensional data. In this way, heterogeneous data with varying sampling frequencies and asynchronous timestamps are organized into an aligned dataset with a unified time granularity. The 15-minute granularity is suitable for production lines with fast cycle times and large fluctuations, while the 1-hour granularity is suitable for production lines with stable processes and longer cycles, and can be flexibly configured according to the actual scenario.

[0027] This method, using production work order numbers as time anchors, essentially establishes time alignment based on production event-driven processes. Each work order corresponds to a specific batch of products and a defined process route. All energy consumption data is strictly limited to the actual production time of that work order, ensuring that energy consumption data and output data in carbon intensity calculations physically correspond strictly to the same production object, making the calculation of carbon emission intensity per unit product more accurate and reliable. For production lines with multiple product varieties produced in rotation, the work order number anchor can clearly distinguish the energy consumption of different product batches, avoiding data confusion during production changeovers and providing high-quality data input for subsequent operating condition normalization. After being sliced ​​and aggregated according to a unified preset time granularity, all data sources are normalized into sequence data with the same time resolution. Each time slice corresponds to a structured record containing multiple fields, simplifying the data input logic in normalization models, benchmark calculations, and other processes, reducing intermediate processing of data matching and interpolation, and improving computational efficiency and system stability. At the same time, the standardized time granularity also facilitates direct integration with industry statistical standards and regulatory reporting requirements, reducing the cost and error probability of data format conversion. The normalized time series data also provides a unified structural foundation for subsequent data storage, querying, and visualization, making the entire data chain more concise and efficient. The 15-minute and 1-hour granularity configurations correspond to different application scenarios. For discrete manufacturing production lines with fast process cycles and frequent fluctuations in operating conditions, the fine 15-minute granularity can capture the details of load changes, making normalization calculations more precise and fully reflecting the impact of short-term operating condition fluctuations on energy efficiency. For process industries or large continuous production lines, where operating conditions are relatively stable, the 1-hour granularity is sufficient to characterize the operating status. In this case, using a coarser granularity can significantly reduce the amount of data, reduce storage overhead and computational load, and improve the overall operating efficiency of the system. Enterprises can flexibly choose the appropriate time granularity based on the process characteristics of their own production lines, data infrastructure conditions, and evaluation accuracy requirements.

[0028] In a preferred embodiment, the method further includes: When collecting multi-source production and energy carbon data, missing data is linearly interpolated using the historical average of similar operating conditions. Work orders with more than 3 consecutive missing time granularities are marked as invalid data and skipped from the current period calculation.

[0029] Specifically, data acquisition in industrial settings is inevitably affected by factors such as network fluctuations, sensor malfunctions, and system maintenance, leading to data loss. This claim employs a tiered processing strategy to address data loss of varying degrees. For sporadic, short-term data loss, the system uses linear interpolation based on historical averages of similar operating conditions for repair: first, it retrieves historical operating condition data with the same product type and similar load rate as the current moment, extracts the characteristic average of the corresponding time period as a reference, and then performs linear interpolation based on valid data before and after the missing point, ensuring that the interpolated value conforms to the operating condition characteristics while maintaining the continuity of the time series. For severe cases with continuous missing data exceeding three time granularities, the system no longer forces interpolation but directly marks the work order as invalid data, skipping the current evaluation. The thresholds for the three time windows are empirical engineering values ​​determined based on industrial data sampling patterns and the error characteristics of the interpolation algorithm, balancing data utilization and result reliability.

[0030] This method uses the historical average of similar operating conditions as the interpolation benchmark. Reference values ​​are selected from historical datasets with the same product structure, similar load rates, and similar energy structures to ensure that the filled values ​​are highly similar to the missing points in terms of operating condition characteristics. Combined with the time continuity constraints of linear interpolation, it considers both the matching of operating condition attributes and the smooth transition of the time series, making the interpolated data closer to the actual operating state. Even with a small amount of missing data, evaluation results continue to be output, effectively maintaining the time coverage and continuity of carbon efficiency evaluation and avoiding evaluation interruptions due to occasional data collection failures. A mechanism is set up to mark work orders as invalid if they have more than three consecutive missing time granularities. Only work orders with data completeness meeting the minimum requirements will enter the evaluation process, ensuring sufficient reliability of the data participating in the evaluation from the source and preventing low-quality data from misleading management decisions and affecting the fairness of the assessment. The two-level processing strategy achieves a reasonable balance between data utilization and result reliability. For short-term and sporadic missing data, it is repaired through condition-based interpolation to make full use of existing data resources. For long-term and continuous missing data, it is decisively abandoned to ensure reliability. While ensuring quality, it maintains a high data utilization rate, enabling the carbon efficiency evaluation system to operate continuously and stably and adapt to the complex operating environment of industrial sites.

[0031] In a preferred embodiment, the equivalent carbon intensity under standard operating conditions is calculated based on multi-source production and energy carbon data and operating condition information, including: A multidimensional normalized correction model is constructed based on operating condition information; Equivalent carbon intensity is calculated based on the original unit product carbon emission intensity and the modified model using multi-source production and energy carbon data.

[0032] Specifically, the multidimensional normalization correction model is the core computational architecture for achieving operating condition normalization in this invention. The process of building the model essentially involves establishing a mapping relationship between actual operating condition parameters and correction coefficients: for each type of operating condition factor affecting carbon intensity, its corresponding correction function or coefficient table is determined. Product structure characteristics correspond to material complexity correction, production line load rate corresponds to equipment efficiency range correction, and energy quality corresponds to emission factor correction. The correction coefficients for the three dimensions are independent of each other, each characterizing the difference between actual and standard operating conditions from different perspectives. When calculating equivalent carbon intensity, the original unit product carbon emission intensity under the current operating condition is first calculated based on the original energy consumption and output data. Then, the original intensity value is input into the multidimensional normalization correction model. Through the combined effect of the three correction coefficients, the equivalent value under the standard operating condition is obtained. The standard operating condition is a predefined reference baseline state, and all actual operating condition data are aligned to this baseline.

[0033] This approach, by simultaneously incorporating three key dimensions—product structure characteristics, production line load rate, and energy quality—and correcting carbon intensity from the material, equipment, and energy ends respectively, systematically isolates the impact of major operating conditions from carbon emission intensity. After three-dimensional normalization, the remaining intensity differences more purely reflect the true gap in an enterprise's energy technology level and management capabilities, providing a basis for fair comparison of evaluation results under different operating conditions. The multi-dimensional parallel correction architecture also possesses good scalability; if other important operating condition influencing factors are discovered in the future, new correction dimensions can be easily added without reconstructing the overall calculation framework. Decomposing the equivalent carbon intensity calculation into two steps—original intensity calculation and correction model conversion—facilitates model verification and auditing. Third-party verification can separately verify the compliance of the original emission calculation and the rationality of the operating condition correction, ensuring clear responsibilities and easy traceability. After normalization, carbon intensity data from all production lines and all time periods are projected onto the same standard operating condition benchmark, forming a unified evaluation scale. This allows companies to benchmark horizontally between different product batches, identify which products have room for improvement in production energy efficiency, conduct vertical comparisons at different time periods, accurately assess the actual effects of energy-saving renovations and process optimizations, and also rank different production lines horizontally, providing a fair quantitative basis for performance evaluation.

[0034] In the preferred embodiment, the equivalent carbon intensity is calculated based on the original unit product carbon emission intensity and the modified model using multi-source production and energy carbon data. The calculation formula is as follows: Among them, I raw (t) represents the original unit product carbon emission intensity. This represents the normalization coefficient for product structural characteristics. This is a correction factor for the production line load rate. This is a correction factor for energy quality parameters.

[0035] The impact of different operating conditions on carbon emission intensity is not a simple additive or subtractive relationship, but rather one of mutual coupling and hierarchical transmission. Complex product structures mean more processes, and the energy consumption of each process is affected by load rate. Simultaneously, energy quality determines the carbon emission factor per unit of energy consumption. The influence of these three factors is proportionally amplified during the transmission process. Using a multiplicative correction structure can accurately reflect this coupling relationship, making the correction results more consistent with the actual physical process. When multiple coefficients deviate from the baseline simultaneously, the total correction magnitude is the product of the factors rather than their sum, enabling a more accurate conversion of actual operating conditions back to standard operating conditions.

[0036] In the preferred embodiment, the normalization coefficient of the product structure characteristics is calibrated by multiple regression analysis using historical multi-variety production data, with a value range of 0.85-1.15, and is used to correct the impact of different product material structures and process routes on unit energy consumption.

[0037] Specifically, the product structure normalization coefficient is used to quantify the differences in baseline energy consumption caused by variations in bill of materials (BOM) structure and process routes among different products. When calibrating this coefficient, the system collects historical production data for each product type within the enterprise, extracting structural characteristics such as the number of material layers, number of parts, number of processes, and processing complexity as independent variables, and using the unit product baseline energy consumption as the dependent variable. A mathematical mapping relationship between structural characteristics and energy consumption baselines is established through multiple regression analysis. The regression model can eliminate the influence of confounding factors such as load factor and energy quality, purely quantifying the contribution of the product structure itself to unit energy consumption. After calibration, each product corresponds to a specific normalization coefficient value, ranging from 0.85 to 1.15. The simpler the structure and the fewer the processes, the smaller the coefficient; the more complex the structure and the longer the processing chain, the larger the coefficient. When calculating equivalent carbon intensity, the original intensity is divided by this coefficient, converting the high baseline energy consumption of complex products to the standard product baseline.

[0038] By employing this method and using multiple regression analysis, statistical calibration based on a large amount of historical production data, the quantitative relationship between various structural factors and energy consumption can be systematically identified. The resulting coefficients have statistical objectivity and accuracy. Regression analysis can also assess the significance of each structural variable, distinguish between primary and secondary influencing factors, and make the coefficient model highly interpretable. The coefficients, calibrated based on real production data, can accurately reflect the energy consumption patterns under the actual product mix of the enterprise, rather than applying industry-standard values ​​or experience-based settings. This directly relates to the accuracy of normalization.

[0039] In the preferred embodiment, the correction factor for the production line load rate is calculated in segments based on the real-time load rate λ: When the load factor λ∈[0.6,1.0], the linear interpolation formula is used. =1-0.2×(1-λ) to calculate; When the load factor λ < 0.6, a fixed compensation value of 1.25 is used to compensate for the increased proportion of basic energy consumption during the standby phase of the equipment under low load.

[0040] Specifically, the load rate correction factor is designed based on the energy efficiency characteristics of industrial equipment. Most industrial motors, pumps, fans, and other power equipment have an economic operating range, which typically corresponds to a load rate of 60% or higher. Within the high-efficiency range, the equipment efficiency changes relatively smoothly with the load rate, so linear interpolation is used for approximation. =1-0.2×(1-λ). When the load rate is 100%, the coefficient is 1, meaning no correction is needed under rated operating conditions. As the load rate decreases, the coefficient increases slowly, indicating that the equipment efficiency has decreased and a larger correction is needed. When the load rate falls below 60% and enters the inefficient zone, the proportion of no-load loss of the equipment increases significantly. Continuing to extrapolate linearly will underestimate the increase in energy consumption. Therefore, a fixed compensation value of 1.25 is set to reflect the characteristics of increased standby energy consumption and basic loss proportion under low load. The dividing point of 0.6 refers to the high-efficiency zone definition in the industrial pump energy-saving monitoring method and the motor economic operation standard. The fixed value of 1.25 comprehensively considers the proportion of basic load such as equipment no-load loss and auxiliary system energy consumption.

[0041] The load efficiency curve of equipment is inherently non-linear, and a single linear correction is insufficient to cover the entire load range. In the high-efficiency operating range, efficiency changes are relatively gradual and can be approximated well by a linear relationship. However, in the low-load range, due to the presence of no-load losses, efficiency deteriorates rapidly as the load decreases, and the linear relationship no longer applies. A two-stage correction function, using linear interpolation in the high-efficiency range and fixed compensation in the low-efficiency range, can more accurately fit the actual variation law of equipment energy efficiency. In particular, the selection of the 60% load rate cutoff point is consistent with the definition of the economic operating range in industrial standards. High-precision load rate correction is especially effective on production lines with large load fluctuations, ensuring the accuracy of normalized results across the entire range from full load to low load, and enabling carbon efficiency evaluation to provide fair conclusions under various operating loads.

[0042] In the preferred embodiment, the formula for calculating the correction factor for the energy quality parameter is: Where, η green ΔLHV represents the proportion of green electricity and green certificates in total energy consumption, and is the deviation of the lower heating value of current fossil energy used relative to the benchmark coal type.

[0043] By systematically separating energy structure factors from carbon intensity through the green electricity ratio correction item, the normalized equivalent intensity primarily reflects the technical efficiency and management level of energy consumption. After correction, companies with a high green electricity ratio will not receive excessively high ratings solely based on their energy structure advantages, while companies with a more traditional energy structure will not have their ratings excessively lowered. The evaluation results are more focused on the company's own energy efficiency management capabilities, enabling the carbon efficiency code to accurately measure the company's level of energy utilization efficiency. This guides companies to focus their emission reduction efforts on improving energy utilization efficiency and optimizing production processes, rather than simply relying on adjustments to energy procurement strategies. The evaluation orientation focused on energy efficiency is highly consistent with the original intention of carbon efficiency evaluation to promote technological progress and production optimization, and can better leverage the incentive and guidance role of the evaluation system.

[0044] In a preferred embodiment, the carbon efficiency baseline intensity for the current evaluation period is calculated based on historical equivalent carbon intensity data, including: Outlier removal from historical equivalent carbon intensity data; The dynamic rolling carbon efficiency benchmark intensity is calculated using an exponentially weighted moving average combined with a low quantile truncation strategy.

[0045] Specifically, the dynamic benchmark is calculated based on historical equivalent carbon intensity sequences. However, the original historical data may contain outliers caused by abnormal operating conditions, such as the restart process after equipment failure, abnormal operation during process debugging, temporary sampling inspections, or trial production. These data do not represent normal production levels. The first step is to remove outliers, filtering out data points from abnormal operating conditions to ensure that the benchmark is established on valid operating data. The second step uses an exponentially weighted moving average combined with a low quantile truncation strategy to calculate the dynamic benchmark. The exponentially weighted moving average assigns higher weights to recent data and decreasing weights to older data, allowing the benchmark to slowly track changes in operating levels. The low quantile truncation uses the lower quantile value within each rolling time window as the reference benchmark, representing the optimal operating level of the production line during that period.

[0046] This method filters out data points from abnormal operating conditions such as equipment failures, unplanned shutdowns, process debugging, and new product trial production. It ensures that the baseline is established on valid data from the company's normal production status. The clean data foundation allows the baseline to truly reflect the energy efficiency level under normal production conditions, without being distorted or disturbed by occasional abnormal events. For discrete manufacturing industries with large fluctuations, the proportion of abnormal data is relatively high, making the elimination process particularly crucial. The purified baseline is not only more accurate in value, but also has a smoother evolution trajectory, without abrupt jumps due to the entry or exit of individual abnormal data, thus improving the representativeness and stability of the baseline. Furthermore, the exponentially weighted moving average achieves an ideal balance between stability and adaptability through its weighting mechanism. Recent data has a higher weight, enabling the benchmark to respond promptly to real changes in production line energy efficiency. For example, after energy-saving renovations are completed, new lower energy consumption data will gradually lower the benchmark, reflecting the results of technological progress in a timely manner. The weight of long-term data decreases, avoiding the long-term dragging down of the benchmark by historical data and preventing the benchmark from lagging behind the actual level. At the same time, the smoothing characteristic of the weighted average ensures that the benchmark will not change drastically due to data fluctuations in individual periods, and always maintains a slow and continuous evolution, avoiding the problem of a fixed benchmark being out of touch with reality and preventing the instability of evaluation standards caused by frequent benchmark changes.

[0047] In the preferred embodiment, the dynamically rolling carbon efficiency benchmark intensity is calculated using an exponentially weighted moving average combined with a low quantile cutoff strategy. The calculation formula is as follows: ) Where β is the baseline decay factor, B(t-1) is the baseline intensity of the previous evaluation period, W is the rolling time window length, set to 7-30 days according to the production line process cycle, and P 25 This represents the 25th percentile of the equivalent carbon intensity sequence within the window.

[0048] The core advantage of the exponentially weighted moving average lies in its ability to precisely control the benchmark update speed through the decay factor β. For production lines with mature technology and stable operation, energy efficiency levels change slowly, allowing for a smaller β value to maintain strong benchmark inertia and ensure the stability of the evaluation standard. For newly commissioned production lines, those in the commissioning or renovation phase, energy efficiency levels change rapidly, allowing for a larger β value to enable the benchmark to keep up with changes in actual operating levels more quickly. This adjustable dynamic characteristic allows the benchmark mechanism to adapt to production lines at different stages of their life cycle, and the controllable evolutionary inertia provides a stable expectation for carbon efficiency assessment. Enterprises know that the benchmark will not change suddenly and significantly, enabling them to rationally plan energy-saving targets and improvement paths. The 25th percentile is in a relatively good but not extreme position, indicating that the production line can reach or exceed this level for a considerable proportion of the time during that period. This suggests that the level is actually achievable and is better than the average level, providing clear guidance. For most industrial production lines, energy efficiency levels follow an approximately normal distribution, and the 25th percentile is precisely located in the middle between the mean and the optimum. This reflects a benchmark effect, allowing companies to improve their operating levels above the baseline for more periods and obtain better carbon efficiency ratings by optimizing operating parameters, strengthening equipment maintenance, and standardizing operations. The length of the rolling time window determines the data basis for quantile calculation. A range of 7 to 30 days corresponds to different process cycles. For production lines with short batch cycles and frequent changes, about a week's worth of data already contains enough samples to stably reflect the distribution of operating levels. In this case, using a shorter window allows the benchmark to respond to changes in operating conditions more quickly. For process industries with long production cycles and continuous and stable processes, a longer period of data accumulation is needed to obtain statistically stable quantile values. In this case, a 30-day window is more appropriate.

[0049] In the preferred embodiment, the deviation between the equivalent carbon intensity and the carbon efficiency benchmark intensity is calculated using the following formula: in, For equivalent carbon strength, This serves as the benchmark intensity for carbon efficiency.

[0050] This method, employing relative deviation, normalizes the difference to a proportional scale relative to the benchmark value. It naturally adapts to the continuously evolving nature of dynamic benchmarks. Regardless of how the benchmark dynamically shifts, the deviation is always expressed as a percentage of the current benchmark. The threshold for grade classification can maintain a stable relative proportion, eliminating the need for frequent adjustments with benchmark changes. The benchmark tracks changes in absolute levels, while the deviation provides a standardized measure of relative position. Both functions work seamlessly together. This relative measurement also allows production lines at different benchmark levels to share a common set of grade mapping rules, reducing the complexity of system configuration and maintenance, and facilitating the establishment of a unified carbon efficiency grading system. The deviation is symmetrically distributed around zero, with positive values ​​indicating performance worse than the benchmark and negative values ​​indicating performance better. This allows managers to easily determine whether current energy efficiency is above or below the benchmark, and by approximately how much it deviates. It enables the identification of inefficient zones for improvement and the summarization and promotion of best practices in efficient zones.

[0051] In a preferred embodiment, a trapezoidal membership function is preset, and fuzzy level determination is performed based on the deviation degree and the trapezoidal membership function, including: The preset trapezoidal membership functions are five trapezoidal membership function groups corresponding to the five carbon efficiency levels. The upper and lower boundary parameters of the trapezoidal membership function groups are set based on the quantiles of the historical deviation distribution in the industry. When the absolute value of the difference in membership degree between two adjacent levels is less than a preset threshold, a weighted centroid decoupling calculation is triggered to output the final level.

[0052] Specifically, the five trapezoidal membership functions correspond to carbon efficiency levels 1 to 5, with each function occupying a segment on the deviation axis. The two ends of the segment are the sloping transition zones of the trapezoid, and the flat-topped segment in the middle is the complete membership zone. The membership functions of two adjacent levels overlap in the transition zone, meaning that a data point belongs to two levels simultaneously within the overlapping interval, but with different membership degrees. The inflection points of each function are determined based on the quantiles of the historical deviation distribution in the industry, ensuring that the level classification matches the overall industry distribution. When determining the level, the membership degree values ​​of the current deviation under each of the five membership functions are first calculated. Normally, the level with the highest membership degree is taken as the determination result. However, when a data point is located near the boundary between two adjacent levels, and their membership degrees are very close, it indicates that the data point is in a sensitive area of ​​level transition. In this case, weighted center decoupling is triggered: the weighted center position is calculated using the ordinal number of the five levels as the x-axis and the membership degree as the weight, and rounded to obtain the final level.

[0053] By using this method, the trapezoidal membership function transforms the step-like change in membership degree into a continuous and gradual change by setting a sloping transition zone at the level boundary. Within the transition zone, data points simultaneously belong to two adjacent levels, and the membership degrees smoothly transition with one increasing and the other decreasing. There are no abrupt change points where it is either one or the other. The continuous processing eliminates the discontinuity of rigid boundaries, making the level determination naturally robust to data fluctuations. Even if the deviation fluctuates slightly near the boundary, the change in membership degree is still gradual. The boundary parameters of the membership function are determined based on the quantiles of the historical deviation distribution of the industry, ensuring that the five-level classification matches the overall energy efficiency distribution of the industry. Each level corresponds to a certain proportion of enterprises or operating conditions, and the level distribution is relatively balanced. The quantile calibration makes the sample size of each level relatively uniform. The carbon efficiency code can effectively distinguish enterprises at different levels, give full play to the evaluation role of the five-level classification, and avoid situations where some levels are too crowded and some levels are almost empty. At the same time, the industry quantile benchmark also makes the level have a clear industry benchmark meaning. Enterprises can know their approximate position in the industry by seeing their level. The reference value and management significance of the evaluation results are significantly improved.

[0054] In a preferred embodiment, the step of outputting the anti-reversal carbon code further includes: The carbon efficiency code level, deviation value, base version number and timestamp are combined and then used to generate a data fingerprint using the national cryptographic SM3 hash algorithm. This fingerprint is then encapsulated into a structured data packet and output through the application programming interface.

[0055] Specifically, the carbon efficiency code is not merely a level number, but a set of structured data containing complete verification information. During output, the system combines and concatenates key fields such as the carbon efficiency code level, deviation value, baseline version number, and generation timestamp, then inputs them into the national cryptographic SM3 hash algorithm to generate a fixed-length hash data fingerprint. The SM3 algorithm possesses unidirectional irreversible, collision-resistant, and tamper-proof cryptographic properties; any minor modification to the original data will lead to a significant change in the hash value. The generated hash fingerprint, along with the original data fields, is encapsulated into a structured data packet, forming the complete carbon efficiency code data body. When external systems obtain the carbon efficiency code through a standardized application programming interface, they can obtain not only the level result but also complete detailed data and the verification fingerprint. The receiver can verify whether the carbon efficiency code data has been tampered with during transmission and storage by recalculating the hash value and comparing it with the attached fingerprint.

[0056] By introducing the national cryptographic SM3 hash algorithm to generate data fingerprints, the carbon code data packet possesses cryptographic-level integrity verification capabilities. Any tampering with fields such as level, deviation, and timestamp will cause the hash verification to fail, allowing the recipient to immediately identify data alteration. This ensures that the original data cannot be deduced from the fingerprint, preventing the disclosure of sensitive details while conducting public verification. Trustworthy carbon code data is a prerequisite for its application in serious scenarios such as financial credit and policy subsidies. The anti-tampering mechanism enhances the authority and practical value of the carbon code, upgrading it from a simple internal management indicator to a trustworthy data asset that can be proven externally.

[0057] In a preferred embodiment, the method further includes: Regularly receive third-party verification data, compare the relative error between the equivalent carbon intensity output by the system and the verification value. When the absolute value of the error exceeds 8% for two consecutive evaluation periods, use the finite difference method to approximate the partial derivative, update the normalization correction coefficient and the benchmark attenuation factor through gradient descent, and write the updated parameters into the parameter version library and automatically take effect in the next evaluation period.

[0058] Specifically, after the system is put into operation, the model parameters may gradually deviate from the optimal values ​​due to factors such as equipment aging, changes in product structure, and process adjustments, leading to discrepancies between the calculated results and the actual situation. The closed-loop self-calibration mechanism uses authoritative data verified by a third party as a calibration reference to periodically check the accuracy of the system output. In each calibration cycle, the system compares the equivalent carbon intensity output in the same period with the authoritative value obtained from the third-party verification, calculating the relative error. If the absolute value of the error exceeds 8% for two consecutive cycles, it indicates that the system deviation has exceeded the acceptable range, requiring parameter updates. Parameter updates employ the gradient descent method: using the verification error as the objective function and each parameter to be calibrated as the independent variable, the partial derivatives of the objective function with respect to each parameter are approximately solved using the finite difference method to determine the direction and magnitude of parameter adjustment. The normalized correction coefficients and the baseline attenuation factor are updated along the direction of error reduction at a certain learning rate. The updated parameters are stored in the parameter version library, along with a version number and effective date; the new version is automatically loaded in the next evaluation cycle.

[0059] This method, using third-party verification data as an objective reference, establishes a feedback loop of periodic inspection and automatic correction. This ensures that model parameters always closely match the actual operating characteristics of the production line. Third-party verification typically has higher detection accuracy, and using it as a calibration benchmark ensures the correctness of the correction direction. The rule of triggering an update only after two consecutive cycles of exceeding tolerance avoids parameter misadjustments caused by occasional errors, ensuring the prudence of parameter updates. Through continuous calibration, carbon efficiency calculation results remain within an acceptable accuracy range over the long term, and will not gradually become invalid due to parameter drift caused by factors such as equipment wear, product iteration, and process adjustments. The continuous assurance of model accuracy means that the carbon efficiency evaluation system has long-term effectiveness and vitality, and can operate stably year-round without frequent manual recalibration, reducing the system's total lifecycle maintenance costs.

[0060] Another aspect of the present invention provides a dynamic carbon efficiency evaluation and carbon efficiency code generation system, the system comprising: Data Acquisition Alignment Module: Used to acquire multi-source production and energy carbon data and align the multi-source production and energy carbon data on a unified time axis; Operating condition normalization module: used to obtain current operating condition information, including product structure characteristic parameters, production line load rate and energy quality parameters. Based on multi-source production and energy carbon data and operating condition information, it calculates the equivalent carbon intensity under standard operating conditions and generates historical equivalent carbon intensity data based on the equivalent carbon intensity. Benchmark generation module: Used to generate the benchmark intensity for the current evaluation period based on historical equivalent carbon intensity data; Deviation Mapping Module: Used to calculate the deviation between equivalent carbon intensity and benchmark intensity, preset trapezoidal membership function, perform fuzzy level judgment based on deviation and trapezoidal membership function and output anti-flip carbon efficiency code.

[0061] In a preferred embodiment, the system further includes: The encapsulation and output module combines the carbon efficiency code level, deviation value, baseline version number, and timestamp, generates a data fingerprint using the national cryptographic SM3 hash algorithm, encapsulates it into a structured data packet, and outputs it externally through the application programming interface. The calibration iteration module is used to periodically receive third-party verification data and compare the relative error between the equivalent carbon intensity output by the system and the verification value. When the absolute value of the error exceeds 8% for two consecutive evaluation cycles, the partial derivative is approximately solved using the finite difference method. The normalized correction coefficient and the benchmark attenuation factor are updated by gradient descent. The updated parameters are written to the parameter version library and automatically take effect in the next evaluation cycle.

[0062] In one specific implementation, a method for dynamic carbon efficiency evaluation and carbon efficiency code generation includes the following steps: S1: Multi-source production and energy carbon data acquisition and time axis alignment.

[0063] Interact with enterprise Manufacturing Execution Systems (MES), Energy Management Systems (EMS), quality inspection systems, and green electricity trading platforms via Industrial Internet of Things (IIoT) protocols such as OPC UA, MQTT, or HTTP. Using production work order numbers as time anchors, align energy consumption data, product batches, and process parameters to a unified time window, configured as 15 minutes or 1 hour. Perform linear interpolation using historical means for similar operating conditions on missing data; work orders with more than three consecutive missing time windows are marked as invalid and skipped from current evaluation. Output a standardized time series dataset D. align (t) serves as the input for subsequent calculations, ensuring the spatiotemporal consistency of the data.

[0064] S2: Adaptive carbon intensity normalization calculation under operating conditions.

[0065] Read the product BOM structure coefficients of the current work order Production line real-time load rate λ, energy quality factor A multidimensional normalized correction model is constructed to map the original carbon intensity to the equivalent carbon intensity under standard operating conditions. The calculation logic is as follows: Among them, I raw (t) represents the original unit product carbon emission intensity, in units of tCO2 / ton of product, calculated according to the emission factor method in the GB / T 32151 series of standards.

[0066] The normalization coefficient for the BOM structure is determined by multiple regression analysis using historical production data of multiple varieties, with a typical value range of 0.85-1.15. This is the load factor correction factor, applied when λ∈[0.6,1.0]. =1-0.2×(1-λ) linear interpolation, when λ<0.6, the fixed value of 1.25 is taken according to the equipment standby energy consumption benchmark compensation.

[0067] The energy quality factor is calculated using the following formula: Where η green The green electricity / green certificate ratio is represented by ΔLHV, which is the relative deviation of the lower heating value of fossil fuels (relative to the benchmark coal type). All coefficients are dimensionless values, and the calculated output is I. eq (t), the unit is tCO2 / ton of standard product.

[0068] Step S3: Quantile rolling baseline generation.

[0069] History I eqOutlier removal is performed on the sequence, specifically removing outliers that deviate from the median by more than 3σ. A dynamic baseline strength is generated using an exponentially weighted moving average (EWMA) combined with a low quantile truncation strategy. The calculation logic is as follows: }) Where β is the baseline decay factor, ranging from 0.1 to 0.3; 0.1 is used for mature production lines, and 0.3 is used for new production lines or during the commissioning period; W is the rolling time window length, set to 7-30 days according to the production line's process cycle time; P 25 This represents the 25th percentile of the equivalent carbon intensity sequence within the window, used to filter high-value interference caused by high-load shutdowns, process adjustments, or abnormal operating conditions. The units of B(t) and I... eq (t) is consistent and drifts slowly with the optimal operating level, avoiding the fixed threshold from deviating from the actual production.

[0070] Step S4: Calculation of carbon efficiency deviation and mapping of fuzzy threshold.

[0071] Calculate the relative deviation between the current equivalent carbon intensity and the dynamic benchmark, input a preset trapezoidal membership function for fuzzy level determination, and output an anti-flip carbon efficiency code. The calculation logic is as follows: First, calculate the relative deviation: Define 5 trapezoidal membership functions μ k (D) (k=1,2,3,4,5), the upper and lower boundaries of each function are set according to the quantiles (e.g., P10, P30, P70, P90) of the historical deviation distribution of the industry. The initial determination of the carbon efficiency code level is as follows: When the membership degree difference between adjacent levels When this occurs, the weighted core decoupling logic is triggered: To prevent the rating from frequently flipping at the threshold boundary due to data noise, the final output L∈{1,2,3,4,5} is obtained.

[0072] Where D(t) is the relative deviation of carbon efficiency at time t, I eq B(t) represents the standard operating condition equivalent carbon strength at time t, B(t) represents the dynamic rolling reference strength at time t, and μ k (D) represents the membership degree of the k-th level carbon code corresponding to the current deviation degree D, μ k+1 (D) represents the membership degree of the (k+1)th level carbon efficiency code corresponding to the current deviation degree D, L represents the final output carbon efficiency code level, and k represents the carbon efficiency code level number.

[0073] Step S5: Trusted encapsulation and interface output of carbon efficiency code.

[0074] The carbon efficiency code level, deviation details, baseline version number, and timestamp are combined and used to generate a tamper-proof data fingerprint using the national cryptographic SM3 hash algorithm. The calculation logic is as follows: Among them, V rule This represents the current mapping rule and parameter version number; t is the timestamp, accurate to the second. The system automatically encapsulates this into a JSON data packet, with fields including: Enterprise Unified Social Credit Code, Carbon Efficiency Code Level L, Deviation D(t), Dynamic Benchmark B(t), Hash Fingerprint H, and Rule Version V. rule Generation time t. Pushed to external systems via a standardized RESTful API, end-to-end processing latency is controlled within 2 seconds.

[0075] Step S6: Closed-loop parameter self-calibration.

[0076] Regularly receive third-party verification reports or audit logs, and compare the relative error E between the system output and the verification value. audit When two consecutive evaluation cycles At that time, the parameter gradient update mechanism is activated: The learning rate η is set to 0.05; the partial derivatives are approximated on the local server using the finite difference method. The updated parameters are written to the version control library and automatically take effect in the next evaluation cycle, realizing dynamic self-optimization of model parameters based on actual production performance.

[0077] It should be understood that the specific embodiments described above are merely illustrative or explanatory of the principles of the invention and do not constitute a limitation thereof. Therefore, any modifications, equivalent substitutions, improvements, etc., made without departing from the spirit and scope of the invention should be included within the protection scope of the invention. Furthermore, the appended claims are intended to cover all variations and modifications falling within the scope and boundaries of the appended claims, or equivalent forms of such scope and boundaries.

Claims

1. A method for dynamic carbon efficiency evaluation and carbon efficiency code generation, characterized in that, The method includes: Collect multi-source production and energy carbon data, and align the multi-source production and energy carbon data on a unified time axis; Obtain current operating condition information, including product structure characteristics, production line load rate and energy quality parameters; calculate the equivalent carbon intensity under standard operating conditions based on the multi-source production and energy carbon data and the operating condition information; and generate historical equivalent carbon intensity data based on the equivalent carbon intensity. The carbon efficiency benchmark intensity for the current evaluation period is calculated based on the historical equivalent carbon intensity data. Calculate the deviation between the equivalent carbon intensity and the carbon efficiency benchmark intensity, preset a trapezoidal membership function, perform fuzzy level judgment based on the deviation and the trapezoidal membership function, and output the anti-flip carbon efficiency code.

2. The method according to claim 1, characterized in that, Aligning the multi-source production and energy carbon data on a unified timeline includes: Using the production work order number as the time anchor, the multi-source production and energy carbon data are sliced ​​and aggregated according to a preset time granularity to form a time-synchronized aligned dataset. The preset time granularity is 15 minutes or 1 hour.

3. The method according to claim 2, characterized in that, The method further includes: When collecting the multi-source production and energy carbon data, missing data is linearly interpolated using the historical average of similar operating conditions. Work orders with more than 3 consecutive missing time granularities are marked as invalid data and skipped from the current period calculation.

4. The method according to claim 1, characterized in that, The calculation of the equivalent carbon intensity under standard operating conditions based on the multi-source production and energy carbon data and the operating condition information includes: A multidimensional normalized correction model is constructed based on the aforementioned working condition information; The equivalent carbon intensity is calculated based on the original unit product carbon emission intensity of the multi-source production and energy carbon data and the modified model.

5. The method according to claim 4, characterized in that, The equivalent carbon intensity is calculated based on the original unit product carbon emission intensity of the multi-source production and energy carbon data and the modified model, using the following formula: Among them, I raw (t) represents the original unit product carbon emission intensity. The normalization coefficient for the product structural features is... This is a correction factor for the production line load rate. This is a correction factor for the energy quality parameter.

6. The method according to claim 1, characterized in that, The calculation of the carbon efficiency benchmark intensity for the current evaluation period based on the historical equivalent carbon intensity data includes: Outlier removal was performed on the historical equivalent carbon intensity data; The dynamically rolling carbon efficiency benchmark intensity is calculated using an exponentially weighted moving average combined with a low quantile truncation strategy.

7. The method according to claim 6, characterized in that, The carbon efficiency benchmark intensity is calculated using an exponentially weighted moving average combined with a low quantile truncation strategy. The calculation formula is as follows: ) Where β is the baseline decay factor, B(t-1) is the baseline intensity of the previous evaluation period, W is the rolling time window length, set to 7-30 days according to the production line process cycle, and P 25 This represents the 25th percentile of the equivalent carbon intensity sequence within the window.

8. The method according to claim 7, characterized in that, The deviation between the equivalent carbon intensity and the carbon efficiency benchmark intensity is calculated using the following formula: in, For equivalent carbon intensity, This serves as the benchmark intensity for carbon efficiency.

9. The method according to any one of claims 1-8, characterized in that, The preset trapezoidal membership function, based on the deviation and the trapezoidal membership function, performs fuzzy level determination, including: The preset trapezoidal membership function consists of five trapezoidal membership function groups corresponding to five carbon efficiency levels. The upper and lower boundary parameters of the trapezoidal membership function groups are set based on the quantiles of the historical deviation distribution in the industry. When the absolute value of the difference in membership degree between two adjacent levels is less than a preset threshold, a weighted centroid decoupling calculation is triggered to output the final level.

10. A dynamic carbon efficiency evaluation and carbon efficiency code generation system, characterized in that, The system includes: Data acquisition and alignment module: used to acquire multi-source production and energy carbon data, and align the multi-source production and energy carbon data on a unified time axis; Operating condition normalization module: used to obtain current operating condition information, including product structure characteristic parameters, production line load rate and energy quality parameters, calculate the equivalent carbon intensity under standard operating conditions based on the multi-source production and energy carbon data and the operating condition information, and generate historical equivalent carbon intensity data based on the equivalent carbon intensity. Benchmark generation module: used to generate the benchmark intensity for the current evaluation period based on the historical equivalent carbon intensity data; Deviation mapping module: used to calculate the deviation between the equivalent carbon intensity and the reference intensity, preset a trapezoidal membership function, perform fuzzy level judgment based on the deviation and the trapezoidal membership function, and output the anti-flip carbon efficiency code.

Images