A carbon emissions management system and method based on digital twin

Abstract

The present invention discloses a carbon emission management system and method based on digital twins, and specifically relates to the technical field of carbon emission management. By collecting historical carbon emission data from different links of the supply chain and real-time collection of actual carbon emission data from suppliers, combined with the invisible carbon emission labeling and test scenario simulation of the digital twin model, the interference degree of invisible carbon emission data on carbon emission reduction targets is evaluated, and external energy price fluctuation data is introduced for multi-dimensional analysis to accurately evaluate the bias degree of model data. According to the bias degree, the model results are divided into two categories: accuracy reflection and inaccuracy reflection. Dynamic optimization, calibration and adjustment strategies are adopted respectively, which effectively solves the long-term negative feedback problem caused by historical data bias and underestimation of carbon emissions in traditional digital twin models, significantly improves the accuracy of carbon emission prediction and the adaptability of the model to actual scenarios, and provides a scientific basis for enterprises to achieve precise carbon management and emission reduction targets.

Description

Technical Field

[0001] The present invention relates to the technical field of carbon emission management, and in particular to a carbon emission management system and method based on digital twins. Background Art

[0002] Carbon emissions management based on digital twins is a new management method that uses digital twin technology to accurately monitor, analyze and optimize carbon emissions. Digital twins are virtual digital models built through real-time data, simulation models and sensor technology, which correspond one-to-one to real physical systems. It can reflect the state and changes of physical systems in real time, providing high-precision data support and dynamic monitoring capabilities for carbon emissions management. In carbon management, digital twin technology can help companies fully understand the carbon emissions of their production processes and facilities, identify emission sources and evaluate the carbon footprint of different links.

[0003] Through digital twin technology, enterprises can simulate different operating scenarios and emission reduction strategies, predict potential carbon emission impacts, and optimize resource allocation. For example, digital twin models can be used to fine-tune energy consumption, logistics and transportation, or manufacturing processes, thereby achieving continuous optimization and control of carbon emissions. In addition, the technology also supports real-time monitoring and traceability, making it easier for companies to meet carbon emission compliance requirements and promote green and low-carbon transformation. Carbon emission management enabled by digital twins not only improves management efficiency, but also provides technical support for achieving carbon neutrality goals.

[0004] The prior art has the following deficiencies:

[0005] Digital twin models are usually optimized based on historical data and existing rules. If there is an implicit bias in the historical data (such as the long-term underestimated carbon emissions of certain links in the supply chain, or the lack of sufficient records of emissions from non-mainstream suppliers), the model may amplify this bias. For example, the model may give priority to suppliers with high implicit carbon emissions but low labeled carbon emissions, resulting in an increase in carbon emissions instead of a decrease. In addition, if the digital twin model cannot accurately reflect the actual carbon emissions, it may bring long-term negative feedback. For example, after underestimating the carbon emissions of a certain link in the equipment or process, the company may frequently use the process, causing the related equipment to overload and increase the risk of failure and energy consumption. This cumulative effect may further amplify carbon emissions and make the problem difficult to fix. Summary of the invention

[0006] The purpose of the present invention is to provide a carbon emissions management system and method based on digital twins to address the deficiencies in the background technology.

[0007] In order to achieve the above object, the present invention provides the following technical solution: a carbon emissions management method based on digital twins, comprising the following steps:

[0008] S1: Collect historical carbon emission data from different links of the supply chain, including carbon emission data generated during outsourced logistics, raw material processing and energy extraction;

[0009] S2: Collect the actual carbon emission data of suppliers in production and logistics transportation, and compare the collected actual carbon emission data with the historical carbon emission data in the digital twin model. According to the comparison results, mark the invisible carbon emission data in the digital twin model;

[0010] S3: Create two sets of test scenarios, including test scenario A that only uses historical carbon emission data, and test scenario B that embeds invisible carbon emission data. Run the digital twin model to predict the full life cycle carbon emissions under test scenario A and test scenario B respectively.

[0011] S4: Based on the prediction results, calculate the emission differences between test scenarios A and B, and compare the optimization effects before and after the introduction of invisible carbon emission data. Based on the optimization effects, evaluate the degree of interference of invisible carbon emission data on carbon reduction targets.

[0012] S5: If the interference level is high, the energy price fluctuation data in the external environment is compared with the historical energy price data in the digital twin model, and the impact of energy price fluctuations on the enterprise’s energy consumption and carbon emissions is analyzed based on the comparison results;

[0013] S6: According to the degree of interference of hidden carbon emission data on carbon reduction targets and the impact of energy price fluctuations on corporate energy consumption and carbon emissions, the degree of data bias in the digital twin model is evaluated. Based on the evaluation results, the results of the digital twin model reflecting the actual carbon emissions are divided into accurate reflection results and inaccurate reflection results, and corresponding carbon emissions management strategies are adopted.

[0014] Preferably, in S2, the invisible carbon emission data in the digital twin model is marked, specifically:

[0015] Calculate the deviation between the actual carbon emission data and the historical data; classify the deviation into explicit deviation and implicit deviation, and compare the calculated deviation between the actual carbon emission data and the historical data with the deviation threshold set according to the historical data. If the deviation between the actual carbon emission data and the historical data is greater than or equal to the deviation threshold, it indicates that the digital twin model does not accurately capture the high-emission links, and the actual data is marked as invisible carbon emission data; if the deviation between the actual carbon emission data and the historical data is less than the deviation threshold, it indicates that the digital twin model accurately captures the high-emission links, and the actual data is not marked as invisible carbon emission data.

[0016] Preferably, in S3, the full life cycle carbon emissions under test scenario A and test scenario B are predicted as follows:

[0017] Input the data of test scenario A and use historical carbon emission data as model input. When running the digital twin model, the model will predict the carbon emissions of the entire life cycle based on the known historical data, including from raw material procurement to final disposal of the product. The model outputs the prediction of carbon emissions of the entire life cycle. The output prediction results should include the distribution of carbon emissions, classified by each link, record the predicted carbon emissions of the entire life cycle under scenario A, and mark the emission data of each link;

[0018] On the basis of test scenario A, add invisible carbon emission data, integrate the invisible carbon emission data into the model input, run the digital twin model, and predict the full life cycle carbon emissions after adding the invisible emission data. The same as scenario A, the output results will include the carbon emissions of the whole life cycle and its distribution by link, record the predicted full life cycle carbon emissions under scenario B, and mark the emission data of each link.

[0019] Preferably, in S4, the deviation rate fluctuation situation between the emission of test scenario A and the emission of test scenario B is analyzed to generate a deviation rate fluctuation index, and the method for obtaining the deviation rate fluctuation index is:

[0020] Collect the carbon emission prediction data under test scenario A and test scenario B, and calculate the deviation rate between the emissions of scenario A and scenario B at the i-th moment. The deviation rate formula is: ; In the formula, is the emission of test scenario B at the i-th moment, is the emission of test scenario A at the i-th moment, is the deviation rate at the i-th moment, indicating the difference between scene B and scene A; the weighted moving average is used to calculate the average value of the deviation rate , the calculation expression is: ;in, is the weight given to the deviation rate at the jth moment, k is the window size, and i is the current moment;

[0021] The deviation rate fluctuation index is used to measure the degree of deviation fluctuation between test scenarios A and B, reflecting the degree of interference of invisible carbon emission data on carbon reduction targets. The calculation expression is: ;in, To calculate the maximum value of the weighted moving average within the window period, To calculate the minimum value of the weighted moving average within the window period, is the deviation rate volatility index.

[0022] Preferably, in S5, after analyzing the comparison results of the energy price fluctuation data in the external environment and the energy price historical data in the digital twin model, an energy consumption price elasticity abnormality index is generated, and the method for obtaining the energy consumption price elasticity abnormality index is:

[0023] Collect the energy price data P and energy consumption data Q of the enterprise, as well as the control variables that affect energy consumption, including production volume X, weather factors W and technical parameters T; divide the data into training sets and test sets, and set the relationship between energy consumption Q and price P to follow the power function form: ; convert it to logarithmic form, ; In the formula, is the price elasticity coefficient of energy prices; the loss function of GLM is defined as the log-likelihood function: ; In the formula, is the likelihood function of the ith sample, are model parameters, ; Determine the parameters by maximizing the log-likelihood estimate L ;

[0024] Using the training set data, the GLM model is fitted using an iterative optimization algorithm to obtain parameter estimates , validate the fitted model on the test set and calculate the predicted value With actual value The price elasticity is the sensitivity of energy price changes to energy consumption changes, and its formula is: ; According to the GLM fitting results, the price elasticity is given by Direct expression; calculate the deviation between the predicted price elasticity at each time point and the average price elasticity of the model to generate the price elasticity anomaly index, the expression is: ; Where KML is the abnormal price elasticity index, ; represents the price elasticity calculated based on actual data; , represents the price elasticity predicted by the model.

[0025] Preferably, in S6, the degree of data bias in the digital twin model is evaluated according to the interference degree of invisible carbon emission data on carbon emission reduction targets and the impact of energy price fluctuations on enterprise energy consumption and carbon emissions, specifically:

[0026] The deviation rate volatility index and the price elasticity anomaly index are converted into comprehensive feature vectors, and the comprehensive feature vectors are used as the input of the machine learning model. The machine learning model uses each group of comprehensive feature vectors to predict the data bias degree value label in the digital twin model as the prediction target, and takes minimizing the sum of prediction errors of the data bias degree value labels in all digital twin models as the training target. The machine learning model is trained until the sum of prediction errors converges, then the model training is stopped, and the data bias degree value in the digital twin model is determined according to the model output results, wherein the machine learning model is a polynomial regression model.

[0027] Preferably, the results of the digital twin model reflecting the actual carbon emissions are divided into accuracy reflection results and inaccuracy reflection results. Specifically, the data bias degree value in the acquired digital twin model is compared with a reference threshold of the data bias degree value pre-set according to historical data. If the data bias degree value in the digital twin model is greater than or equal to the reference threshold of the data bias degree value, it means that the data bias degree in the digital twin model is high. At this time, a data bias abnormality signal is generated, and the results of the digital twin model reflecting the actual carbon emissions are divided into inaccuracy reflection results; if the data bias degree value in the digital twin model is less than the reference threshold of the data bias degree value, it means that the data bias degree in the digital twin model is low. At this time, no data bias abnormality signal is generated, and the results of the digital twin model reflecting the actual carbon emissions are divided into accuracy reflection results.

[0028] Preferably, for inaccurate reflection results, the carbon emissions predicted by the real-time monitoring model and actual emissions , calculate the deviation value: ; If ΔE>ΔE threshold, it means that the model underestimates emissions and dynamic management measures need to be initiated; Carbon emissions predicted by the digital twin model, is the actual carbon emission data collected by the monitoring equipment; the ΔE threshold represents the threshold of the allowable deviation range;

[0029] By monitoring the operating data of equipment and processes, the operating parameters of related equipment are dynamically adjusted to avoid overload operation. The adjustment formula is: ; is the adjusted equipment operating parameters, is the current equipment operating parameters; when the emission underestimate of a certain link is identified, resources are reallocated to optimize the process or equipment with high energy consumption; the optimization formula is: ; In the formula, To optimize resources allocated to a certain link, To optimize resources in total; according to the field measurement data, dynamically adjust the emission factors of each link, the adjustment formula is: ; is the updated emission factor, is the current emission factor; the long-term emission risk is calculated through cumulative deviation to prevent the negative cumulative effect caused by underestimation of emissions. The cumulative emission deviation formula is: ; is the emission deviation in the tth time period, T is the total number of monitored time periods, and when the cumulative deviation exceeds the threshold, an early warning signal is triggered.

[0030] The present invention also provides a carbon emission management system based on digital twins, including a data collection module, a comparison module, a test module, an interference assessment module, an external environment analysis module and a classification management module:

[0031] Data collection module: collects historical carbon emission data from different links of the supply chain, including carbon emission data generated during outsourced logistics, raw material processing and energy extraction;

[0032] Comparison module: collects actual carbon emission data from suppliers’ production links and logistics transportation, and compares the collected actual carbon emission data with the historical carbon emission data in the digital twin model. Based on the comparison results, marks the invisible carbon emission data in the digital twin model.

[0033] Test module: Create two sets of test scenarios, including test scenario A that only uses historical carbon emission data, and test scenario B that embeds invisible carbon emission data. Run the digital twin model to predict the full life cycle carbon emissions under test scenario A and test scenario B respectively;

[0034] Interference assessment module: Based on the prediction results, calculate the emission differences between test scenarios A and B, and compare the optimization effects before and after the introduction of invisible carbon emission data. Based on the optimization effects, evaluate the degree of interference of invisible carbon emission data on carbon reduction targets;

[0035] External environment analysis module: If the interference level is high, the energy price fluctuation data in the external environment will be compared with the historical energy price data in the digital twin model, and the impact of energy price fluctuations on the enterprise's energy consumption and carbon emissions will be analyzed based on the comparison results;

[0036] Classification management module: According to the degree of interference of hidden carbon emission data on carbon reduction targets and the impact of energy price fluctuations on corporate energy consumption and carbon emissions, the degree of data bias in the digital twin model is evaluated, and based on the evaluation results, the results of the digital twin model reflecting the actual carbon emissions are divided into accurate reflection results and inaccurate reflection results, and corresponding carbon emissions management strategies are adopted.

[0037] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0038] 1. The present invention significantly improves the ability of the digital twin model to reflect actual carbon emissions by introducing methods such as invisible carbon emission labeling, test scenario comparison and dynamic optimization, and overcomes the problem of underestimation of carbon emissions caused by historical data bias in traditional methods. By accurately marking the invisible carbon emission data not captured in the supply chain, constructing test scenarios A and B, analyzing the degree of interference of invisible carbon emissions on emission reduction targets, and combining the impact of external energy price fluctuations, the model input and parameters are dynamically adjusted. The present invention also introduces the deviation rate volatility index and the price elasticity anomaly index, integrates them into a comprehensive feature vector, and uses a machine learning model to quantitatively evaluate the degree of data bias to achieve continuous optimization of the model prediction accuracy.

[0039] 2. The present invention can effectively avoid the long-term cumulative effects caused by underestimation of carbon emissions, such as equipment overload, increased energy consumption and emission rebound risk. Through real-time monitoring, dynamic calibration of parameters and abnormal early warning mechanism, it not only improves the model's ability to accurately predict the entire life cycle of carbon emissions, but also provides a scientific basis for corporate carbon emission management. Ultimately, the present invention realizes the refined management of carbon emissions, helps enterprises optimize resource allocation, reduce carbon emission costs, and steadily achieve carbon neutrality goals. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0041] Figure 1 The present invention is a flow chart of the method.

[0042] Figure 2 It is a system module diagram of the present invention. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] Example 1, please refer to Figure 1 and Figure 2 As shown, the carbon emission management method based on digital twin described in this embodiment includes the following steps:

[0045] S1: Collect historical carbon emission data from different links of the supply chain, including carbon emission data generated during outsourced logistics, raw material processing and energy extraction;

[0046] S2: Collect the actual carbon emission data of suppliers in production and logistics transportation, and compare the collected actual carbon emission data with the historical carbon emission data in the digital twin model. According to the comparison results, mark the invisible carbon emission data in the digital twin model;

[0047] S3: Create two sets of test scenarios, including test scenario A that only uses historical carbon emission data, and test scenario B that embeds invisible carbon emission data. Run the digital twin model to predict the full life cycle carbon emissions under test scenario A and test scenario B respectively.

[0048] S4: Based on the prediction results, calculate the emission differences between test scenarios A and B, and compare the optimization effects before and after the introduction of invisible carbon emission data. Based on the optimization effects, evaluate the degree of interference of invisible carbon emission data on carbon reduction targets.

[0049] S5: If the interference level is high, the energy price fluctuation data in the external environment is compared with the historical energy price data in the digital twin model, and the impact of energy price fluctuations on the enterprise’s energy consumption and carbon emissions is analyzed based on the comparison results;

[0050] S6: According to the degree of interference of hidden carbon emission data on carbon reduction targets and the impact of energy price fluctuations on corporate energy consumption and carbon emissions, the degree of data bias in the digital twin model is evaluated. Based on the evaluation results, the results of the digital twin model reflecting the actual carbon emissions are divided into accurate reflection results and inaccurate reflection results, and corresponding carbon emissions management strategies are adopted.

[0051] In S1, historical carbon emission data of different links in the supply chain are collected, including carbon emission data generated during outsourced logistics, raw material processing and energy extraction, specifically:

[0052] Determine the supply chain links that need to be covered (outsourced logistics, raw material processing, energy extraction). Clarify the level of refinement of carbon emissions data (e.g., by region, supplier, product type). Include unit emissions (e.g., CO2 per ton of transport). 2 Emissions), total emissions and emission factors (such as carbon emission coefficients of energy types). Data sources include logistics records and raw material procurement data in the enterprise supply chain management system (SCM).

[0053] Collect data on road, rail, sea and air transport, and calculate emission factors for each mode of transport. Example: Road transport emits about 62 grams of CO per ton-kilometer2 , based on IPCC emission factors. Obtain historical transportation data from logistics service providers, including cargo weight, transportation distance, and fuel type. If detailed data is not available, use industry benchmarks (such as logistics emission factors provided by DEFRA) to estimate.

[0054] Get carbon emission data for different processing steps by material type (e.g. metal, plastic, textile). Example: 1 ton of steel produces approximately 1.85 tons of CO 2 . Require raw material suppliers to provide energy usage (electricity, natural gas, etc.) and corresponding emission factors during processing. Collect historical carbon emission records through procurement contracts or questionnaires. Use public life cycle assessment (LCA) databases, such as Ecoinvent or GaBi database, to obtain emission factors for the processing stage. Obtain carbon emission data for mining and preliminary processing by energy type (coal, oil, natural gas). Example: Coal mining produces about 2 kg CO per ton. 2 . Ask energy suppliers to provide data on energy consumption of mining equipment, fuel types used and mining output. If detailed data is not available, use regional or industry-wide mining carbon emission factors. Refer to the energy supplier's carbon disclosure report or emission data published by an independent certification body (such as CDP, ISO 14064 reports).

[0055] Ensure that data is in consistent units (e.g. all emissions are reported as “tons of CO 2 ”). Compare data from different sources at the same stage and mark outliers. Ensure that the collected data covers multiple years to observe historical trends and seasonal fluctuations. Use standardized tools (such as the emission calculation tool provided by GHG Protocol) to uniformly format the collected data. Integrate into a supply chain carbon emissions database and store it by stage classification.

[0056] S2: Collect actual carbon emission data from suppliers’ production processes and logistics transportation, and compare the collected actual carbon emission data with the historical carbon emission data in the digital twin model. Based on the comparison results, mark the invisible carbon emission data in the digital twin model.

[0057] Determine the scope of collection, including covering the supplier's production process (raw material processing, production process energy consumption) and logistics transportation (transportation mode, distance, load). Clarify the carbon emission data indicators collected, including production links: energy use (electricity, fuel consumption), process emissions (such as process gas). Logistics transportation includes transportation distance, cargo weight, and type of transportation tool (truck, railway, ship, etc.).

[0058] Visit supplier production facilities to record energy usage and equipment operating status. Use portable energy consumption monitoring instruments to measure real-time electricity and fuel consumption. Collect process emission data, such as emissions from chemical reactions or industrial processes. Obtain supplier energy bills and carbon emission reports to verify electricity consumption and fuel types. Obtain detailed transportation records from logistics partners, including transportation distance, cargo weight, route selection, and transportation vehicle fuel type. Install GPS devices and fuel sensors to monitor fuel consumption and vehicle emission data during transportation in real time.

[0059] Use the Internet of Things (IoT) platform to upload and integrate the collected data in real time and establish a supplier carbon emissions database. Convert the format and unit of the collected data to ensure consistency with the historical data in the digital twin model (e.g., unify it into "tons of CO 2 ”). Detect and clean up extreme values ​​or unreasonable data, such as records of abnormally high or low energy consumption or emissions. Confirm the rationality of the data by comparing with industry benchmarks. Use estimation methods to fill in missing data, such as filling in the average emission levels of similar production processes or transportation routes. Compare the actual carbon emission data collected with the historical data of the corresponding supplier links in the digital twin model. Group and match by supplier, transportation method and time period to ensure the accuracy of the comparison. Calculate the deviation between the actual carbon emission data and the historical data: ; Draw data comparison charts (such as bar charts, scatter plots) to intuitively display the differences between actual data and historical data.

[0060] The deviations are classified into explicit deviations and implicit deviations, and the deviations between the calculated actual carbon emission data and the historical data are compared with the deviation threshold set based on the historical data. If the deviation between the actual carbon emission data and the historical data is greater than or equal to the deviation threshold, it indicates that the digital twin model does not accurately capture the high-emission links, and the actual data is marked as implicit carbon emission data; if the deviation between the actual carbon emission data and the historical data is less than the deviation threshold, it indicates that the digital twin model accurately captures the high-emission links, and the actual data is not marked as implicit carbon emission data.

[0061] S3: Create two sets of test scenarios, including test scenario A that only uses historical carbon emission data, and test scenario B that embeds invisible carbon emission data. Run the digital twin models separately to predict the full life cycle carbon emissions under test scenario A and test scenario B.

[0062] Test scenario A predicts the carbon emissions of the entire life cycle based on historical data. Only historical carbon emission data of each link in the supply chain is used, usually including emission data from external logistics, raw material processing, energy extraction, etc. These data may include historical emission records from suppliers or existing digital twin models. It is assumed that the emission data has been standardized and cleaned to conform to a unified format (such as tons CO 2 ). Emissions data has not been revised or updated and is based solely on historical records.

[0063] On the basis of test scenario A, the invisible carbon emission data obtained through field audits and data collection are embedded to further improve the accuracy of the model. In addition to historical carbon emission data, scenario B also includes invisible carbon emission data. Invisible carbon emission data include: emissions from the production process that are not reported by suppliers, or emission data that has been underestimated for a long time. Idle emissions in the logistics process, optimization potential of transportation routes, etc. Energy efficiency improvement measures in the supply chain (such as transportation tool updates or energy use optimization). It is assumed that the invisible emission data has been collected through field measurements, field audits, supplier reports, etc., and the data has been verified and integrated with historical data.

[0064] The scenario A model runs, including: inputting the data of test scenario A, using historical carbon emission data as model input. When running the digital twin model, the model will predict the carbon emissions of the entire life cycle based on the known historical data, including from the procurement of raw materials to the final disposal of the product (including transportation, manufacturing, etc.). The model outputs the carbon emission forecast of the entire life cycle (unit: tons CO 2 ). The output prediction results should include the distribution of carbon emissions, classified by each link (raw materials, manufacturing, logistics, etc.). Record the predicted full life cycle carbon emissions under scenario A and mark the emission data of each link.

[0065] The scenario B model runs, including: adding invisible carbon emission data to the test scenario A. Invisible emission data can be obtained through auditing supplier data, real-time monitoring during transportation, etc., and these data are integrated into the model input. Run the digital twin model to predict the life cycle carbon emissions after adding invisible emission data. As with scenario A, the output results will include the carbon emissions of the whole life cycle and their distribution by link. Record the predicted life cycle carbon emissions under scenario B and mark the emission data of each link.

[0066] Collect some actual emission data and compare them with the model output to verify the prediction accuracy of scenario A and scenario B. Based on the comparison results, adjust the model parameters (such as the estimation method of invisible emissions) to improve the model's ability to capture invisible carbon emissions. Establish a regular update mechanism to adjust the estimation of invisible emission data based on new data (such as updated emission reports from suppliers, transportation optimization plans). In the subsequent use process, constantly compare the actual data and model prediction results, and continuously improve the digital twin model to ensure its accuracy.

[0067] S4: Based on the prediction results, calculate the emission differences between test scenarios A and B, and compare the optimization effects before and after the introduction of invisible carbon emission data. Based on the optimization effects, evaluate the degree of interference of invisible carbon emission data on carbon reduction targets.

[0068] Collect the carbon emission data of the whole life cycle, and test the emission of scenario A (historical data): mark the predicted carbon emission of the whole life cycle as EA (unit: tons CO 2 The carbon emissions of the entire life cycle after embedding the invisible carbon emission data are marked as EB (unit: tons CO 2 ). The calculation formula for emission difference is: ;in, represents the difference in carbon emissions over the entire life cycle between the two scenarios. , indicating that carbon emissions increase after the introduction of invisible carbon emission data; if , indicating that carbon emissions are reduced after the introduction of invisible carbon emission data. To further quantify the difference, the deviation rate is calculated: ; The deviation rate can intuitively understand the impact of invisible carbon emission data on the prediction results.

[0069] In test scenario A, the carbon emissions predicted by the model are based only on historical data, which means that carbon emissions in certain supply chain links or logistics processes may be underestimated. In test scenario B, after the invisible carbon emissions data is embedded, the model's estimate of carbon emissions is more accurate.

[0070] Forecast before the introduction of invisible carbon emission data (Scenario A): Some "invisible emissions" links may not be fully considered, such as no-load in the logistics link, underestimated emissions in the production link, etc. This leads to lower carbon emissions and may overestimate the potential for carbon emission reduction.

[0071] Forecast after introducing invisible carbon emission data (Scenario B): The introduction of invisible carbon emission data makes the carbon emission forecast closer to the actual situation, capturing the emissions that are not fully recorded in the production and logistics processes. This change shows that the model is more accurate in carbon emission forecasts, and therefore can provide more reasonable guidance for carbon emission reduction measures.

[0072] If ΔE<0, it means that after the introduction of invisible carbon emission data, the predicted life cycle carbon emissions have decreased, indicating that the optimization effect is good, the actual carbon emissions have been accurately calculated, and carbon reduction measures can play a more effective role. If ΔE>0, it means that the introduction of invisible carbon emission data has led to an increase in carbon emissions, which may indicate that when historical data fails to reflect certain real emissions, the adjustment of the model has instead led to a larger carbon emission estimate.

[0073] The deviation rate fluctuation index is generated by analyzing the deviation rate fluctuation between the emissions of test scenario A and the emissions of test scenario B to evaluate the interference degree of invisible carbon emission data on carbon emission reduction targets. The deviation rate fluctuation index is obtained as follows:

[0074] Collect carbon emission forecast data under test scenario A (historical carbon emission data) and test scenario B (including hidden carbon emission data). The data should be arranged in time series, usually in days, weeks or months.

[0075] Calculate the deviation rate between the emissions of scenario A and scenario B at the i-th moment. The deviation rate formula is: ; In the formula, is the emission of test scenario B at the i-th moment, is the emission of test scenario A at the i-th moment, is the deviation rate at the i-th moment, indicating the relative difference between scene B and scene A.

[0076] In order to smooth out short-term fluctuations in the deviation rate, a weighted moving average (WMA) is used to calculate the average value of the deviation rate. , the calculation expression is: ;in, is the weight given to the deviation rate at the jth moment, which is usually inversely proportional to the time interval (e.g., the latest moment has the largest weight, and the furthest moment has the smallest weight). k is the window size, which indicates how many data points are considered before the current moment (e.g., k=7 means data from the past 7 days). i is the current moment.

[0077] The deviation rate fluctuation index is used to measure the degree of deviation fluctuation between test scenarios A and B, reflecting the degree of interference of invisible carbon emission data on carbon reduction targets. The calculation expression is: ;in, To calculate the maximum value of the weighted moving average within the window period, To calculate the minimum value of the weighted moving average within the window period, is the deviation rate volatility index.

[0078] The larger the deviation rate volatility index, the greater the volatility of the predicted carbon emissions due to the introduction of invisible carbon emission data, indicating that it interferes with the carbon emission reduction target to a high degree. This may mean that there are problems with the quality of invisible carbon emission data, or that its introduction exposes other uncaptured dynamic factors in the model. A high volatility index may make it difficult to clarify carbon emission reduction targets and emission reduction measures cannot be accurately implemented, thus affecting the formulation and implementation of carbon emission reduction.

[0079] The smaller the deviation rate volatility index, the more the introduction of invisible carbon emission data effectively reduces the volatility of the forecast and improves the stability and accuracy of the model. This shows that invisible carbon emission data can supplement the deficiencies of historical data, making the model's prediction of carbon emissions over the life cycle more consistent and reliable. A low volatility index usually means that invisible carbon emission data has little interference with carbon reduction targets, which helps optimize carbon reduction strategies and promote the smooth achievement of targets.

[0080] S5: If the interference level is high, the energy price fluctuation data in the external environment will be compared with the historical energy price data in the digital twin model, and the impact of energy price fluctuations on the enterprise’s energy consumption and carbon emissions will be analyzed based on the comparison results.

[0081] Collect external energy price fluctuation data, including market energy price data (such as electricity, natural gas, coal, fuel oil, etc.). Data acquisition method: energy trading platform and third-party data services (such as Bloomberg, IEA, etc.). Data format: time series format (by day, month, quarter, etc.). Unit: such as electricity price per kilowatt-hour ($ / kWh), coal price per ton ($ / t).

[0082] Obtain the company's energy procurement records, including unit price, purchase volume, and corresponding timestamps. Historical data extracted from the digital twin model should cover the same time range as the external energy price fluctuation data to ensure consistency in comparison. Convert energy prices to comparable units (such as $ / GJ). Use time interpolation or resampling techniques to align external energy prices and data in the digital twin model to the same time granularity (such as daily or monthly). Use the range to calculate the magnitude of price fluctuations and the standard deviation to assess price volatility: If there is a significant difference between the external price and the company's price, analyze the reasons why the company's price lags behind (such as long-term procurement contracts).

[0083] Analyze the share of different energy types in the company's energy consumption (e.g., electricity, natural gas, fuel oil). Determine whether the share of highly volatile energy in total energy consumption is high, as price fluctuations of highly volatile energy sources may lead to significant changes in energy consumption. Assess the energy efficiency of the company's production equipment (e.g., energy consumption per unit of product produced). Rising energy prices may cause companies to reduce the frequency of use of high-energy-consuming equipment or increase the adoption of energy-efficiency technologies.

[0084] According to the carbon emission factors of different energy sources (such as electricity 0.5 tons CO 2 / MWh, coal 2.6 tons CO 2 / ton) to calculate carbon emissions. When energy price fluctuations affect energy consumption, the corresponding carbon emissions will also change. Enterprises may switch energy types (such as from natural gas to coal) due to price fluctuations, resulting in changes in carbon emission intensity. Analyze whether the optimization or deterioration trend of the enterprise's energy structure caused by long-term energy price fluctuations significantly affects the enterprise's emission reduction path.

[0085] After analyzing the comparison results of energy price fluctuation data in the external environment and the historical energy price data in the digital twin model, the energy price elasticity anomaly index is generated to analyze the impact of energy price fluctuations on corporate energy consumption and carbon emissions. The method for obtaining the energy price elasticity anomaly index is as follows:

[0086] Collect the company's energy price data P and energy consumption data Q, as well as other control variables that may affect energy consumption (such as production volume X, weather factors W, technical parameters T). Arrange the data in time series (such as by day, month, quarter, etc.). Normalize or standardize the data so that each variable has a similar scale.

[0087] The data is divided into a training set and a test set to verify the prediction effect of the model. The relationship between energy consumption Q and price P is assumed to follow the power function form: ; convert it to logarithmic form, ; In the formula, is the price elasticity coefficient of energy prices; select the appropriate GLM distribution according to the distribution of energy consumption data: Lognormal distribution: applicable to the case where energy consumption is right-skewed. Gamma distribution: applicable to the case where energy consumption is non-negative and has a right-tail distribution. According to the selected distribution, define the loss function of GLM as the log-likelihood function: ; In the formula, is the likelihood function of the ith sample, are model parameters, ; Determine the parameters by maximizing the log-likelihood estimate L .

[0088] Use the training set data to fit the GLM model using an iterative optimization algorithm (such as gradient descent or Newton's method) to obtain parameter estimates . Validate the fitted model on the test set and calculate the predicted value With actual value Price elasticity is the sensitivity of energy price changes to energy consumption changes, and its formula is: ; According to the GLM fitting results, the price elasticity is given by Direct expression. Calculate the deviation between the predicted price elasticity at each time point and the average price elasticity of the model to generate the price elasticity anomaly index, expressed as: ; Where KML is the abnormal price elasticity index, ; represents the price elasticity calculated based on actual data; , represents the price elasticity predicted by the model.

[0089] The abnormality of energy price elasticity is classified according to the abnormality index value: normal fluctuation: abnormality index <10%; moderate abnormality: abnormality index between 10% and 30%. Severe abnormality: abnormality index >30%. A high abnormality index indicates that the energy consumption response of the enterprise to price fluctuations is abnormal, such as abnormal dependence on certain energy prices or inflexible procurement mechanism. This may lead to uncontrolled carbon emission management. A low abnormality index indicates that the energy consumption of the enterprise responds to price fluctuations relatively stably, and the carbon emission reduction target is less affected.

[0090] S6: According to the degree of interference of hidden carbon emission data on carbon reduction targets and the impact of energy price fluctuations on corporate energy consumption and carbon emissions, the degree of data bias in the digital twin model is evaluated. Based on the evaluation results, the results of the digital twin model reflecting the actual carbon emissions are divided into accurate reflection results and inaccurate reflection results, and corresponding carbon emissions management strategies are adopted.

[0091] The deviation rate volatility index and the price elasticity anomaly index are converted into comprehensive feature vectors, and the comprehensive feature vectors are used as the input of the machine learning model. The machine learning model uses each group of comprehensive feature vectors to predict the data bias degree value label in the digital twin model as the prediction target, and takes minimizing the sum of prediction errors of the data bias degree value labels in all digital twin models as the training target. The machine learning model is trained until the sum of prediction errors converges, then the model training is stopped, and the data bias degree value in the digital twin model is determined according to the model output results, wherein the machine learning model is a polynomial regression model.

[0092] The method for obtaining the data bias value in the digital twin model is to obtain the corresponding function expression from the comprehensive feature vector training data of the trained machine learning model: ; In the formula, is the output function of the model, is the deviation rate volatility index, KML is the price elasticity anomaly index, is the data bias value in the digital twin model.

[0093] The data bias degree value obtained in the digital twin model is compared with the reference threshold of the data bias degree value pre-set according to historical data. If the data bias degree value in the digital twin model is greater than or equal to the reference threshold of the data bias degree value, it means that the data bias degree in the digital twin model is high. At this time, a data bias abnormality signal is generated, and the result of the digital twin model reflecting the actual carbon emissions situation is classified as an inaccurate reflection result; if the data bias degree value in the digital twin model is less than the reference threshold of the data bias degree value, it means that the data bias degree in the digital twin model is low. At this time, no data bias abnormality signal is generated, and the result of the digital twin model reflecting the actual carbon emissions situation is classified as an accurate reflection result.

[0094] Carbon emission management strategies that accurately reflect the results of digital twin models should make full use of the reliability and accuracy of the model to achieve efficient carbon emission reduction management through optimizing resource allocation, dynamic monitoring and adjustment. Specific measures include predicting carbon emission trends based on model results, decomposing carbon emission indicators, optimizing resource allocation in high-emission links, and dynamically adjusting energy consumption strategies to cope with actual production changes. At the same time, with the real-time monitoring capabilities of the model, companies can promptly detect abnormal emission sources and take corrective measures to ensure that carbon emissions are always within a controllable range, providing a scientific basis for decision-making.

[0095] In addition, enterprises can use model data to support carbon trading and policy compliance, generate accurate carbon emission reports, and participate in carbon quota trading, thereby reducing operating costs and improving market competitiveness. At the same time, the accuracy of the model can also be used to evaluate the application effect of low-carbon technologies and promote the introduction of clean energy and the research and development of innovative technologies.

[0096] For inaccurate reflection results, the carbon emissions predicted by the real-time monitoring model and actual emissions , calculate the deviation value: ; If ΔE>ΔE threshold (the deviation value exceeds the preset threshold), it means that the model underestimates emissions and dynamic management measures need to be initiated. Carbon emissions predicted by the digital twin model, It is the actual carbon emission data collected by monitoring equipment. The ΔE threshold represents the threshold of the allowable deviation range, which is set based on historical data and industry benchmarks.

[0097] By monitoring the operating data of equipment and processes, the operating parameters of related equipment (such as power, load, temperature) are dynamically adjusted to avoid overload operation. The adjustment formula is: ; is the adjusted equipment operating parameters, Run parameters for the current device.

[0098] When the underestimated emissions of a certain link are identified, resources are reallocated to optimize the processes or equipment with high energy consumption. The optimization formula is: ; In the formula, To optimize resources allocated to a certain link, For total optimization of resources.

[0099] According to the field measurement data, the emission factors of each link are dynamically adjusted. The adjustment formula is: ; is the updated emission factor, is the current emission factor.

[0100] The long-term emission risk is calculated by cumulative deviation to prevent the negative cumulative effects caused by underestimation of emissions. The cumulative emission deviation formula is: ; is the emission deviation in the tth time period, and T is the total number of monitored time periods. When the cumulative deviation exceeds the threshold, an early warning signal is triggered.

[0101] In this embodiment, the carbon emission data in the digital twin model is evaluated and optimized through a systematic process. First, the historical carbon emission data of different links in the supply chain (outsourced logistics, raw material processing and energy mining) are collected, and the actual carbon emission data in the supplier's production links and logistics transportation are collected, and the invisible carbon emission data are marked after comparison; then two sets of test scenarios are created (scenario A using only historical data and scenario B with embedded invisible data), and the digital twin model is run to predict the carbon emissions over the entire life cycle, and the degree of interference of the invisible carbon emission data on the carbon reduction target is evaluated by comparing the emission differences between the two scenarios; if the degree of interference is high, external energy price fluctuation data is introduced, and its impact on the enterprise's energy consumption and carbon emissions is analyzed by comparing with the model's historical data; finally, according to the degree of influence of the invisible carbon emission data and energy price fluctuations, the degree of data bias of the digital twin model is evaluated, the model results are divided into accuracy and inaccuracy reflection results, and a precise carbon emission management strategy is implemented based on the classification results.

[0102] Embodiment 2: A carbon emission management system based on digital twins described in this embodiment includes a data collection module, a comparison module, a test module, an interference assessment module, an external environment analysis module and a classification management module:

[0103] Data collection module: collects historical carbon emission data from different links of the supply chain, including carbon emission data generated during outsourced logistics, raw material processing and energy extraction;

[0104] Comparison module: collects actual carbon emission data from suppliers’ production links and logistics transportation, and compares the collected actual carbon emission data with the historical carbon emission data in the digital twin model. Based on the comparison results, marks the invisible carbon emission data in the digital twin model.

[0105] Test module: Create two sets of test scenarios, including test scenario A that only uses historical carbon emission data, and test scenario B that embeds invisible carbon emission data. Run the digital twin model to predict the full life cycle carbon emissions under test scenario A and test scenario B respectively;

[0106] Interference assessment module: Based on the prediction results, calculate the emission differences between test scenarios A and B, and compare the optimization effects before and after the introduction of invisible carbon emission data. Based on the optimization effects, evaluate the degree of interference of invisible carbon emission data on carbon reduction targets;

[0107] External environment analysis module: If the interference level is high, the energy price fluctuation data in the external environment will be compared with the historical energy price data in the digital twin model, and the impact of energy price fluctuations on the enterprise's energy consumption and carbon emissions will be analyzed based on the comparison results;

[0108] Classification management module: According to the degree of interference of hidden carbon emission data on carbon reduction targets and the impact of energy price fluctuations on corporate energy consumption and carbon emissions, the degree of data bias in the digital twin model is evaluated, and based on the evaluation results, the results of the digital twin model reflecting the actual carbon emissions are divided into accurate reflection results and inaccurate reflection results, and corresponding carbon emissions management strategies are adopted.

[0109] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0110] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0111] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0112] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A carbon emissions management method based on digital twins, characterized by: The following steps are involved: S1: Collect historical carbon emission data from different links of the supply chain, including carbon emission data generated during outsourced logistics, raw material processing and energy extraction; S2: Collect the actual carbon emission data of suppliers in production and logistics transportation, and compare the collected actual carbon emission data with the historical carbon emission data in the digital twin model. According to the comparison results, mark the invisible carbon emission data in the digital twin model; S3: Create two sets of test scenarios, including test scenario A that only uses historical carbon emission data, and test scenario B that embeds invisible carbon emission data. Run the digital twin model to predict the full life cycle carbon emissions under test scenario A and test scenario B respectively. S4: Based on the prediction results, calculate the emission differences between test scenarios A and B, and compare the optimization effects before and after the introduction of invisible carbon emission data. Based on the optimization effects, evaluate the degree of interference of invisible carbon emission data on carbon reduction targets. S5: If the interference level is high, the energy price fluctuation data in the external environment is compared with the historical energy price data in the digital twin model, and the impact of energy price fluctuations on the enterprise’s energy consumption and carbon emissions is analyzed based on the comparison results; S6: According to the degree of interference of hidden carbon emission data on carbon reduction targets and the impact of energy price fluctuations on corporate energy consumption and carbon emissions, the degree of data bias in the digital twin model is evaluated. Based on the evaluation results, the results of the digital twin model reflecting the actual carbon emissions are divided into accurate reflection results and inaccurate reflection results, and corresponding carbon emissions management strategies are adopted.

2. A carbon emissions management method based on digital twins according to claim 1, characterized in that: In S2, the invisible carbon emission data in the digital twin model is marked as follows: Calculate the deviation between the actual carbon emission data and the historical data; classify the deviation into explicit deviation and implicit deviation, and compare the calculated deviation between the actual carbon emission data and the historical data with the deviation threshold set according to the historical data. If the deviation between the actual carbon emission data and the historical data is greater than or equal to the deviation threshold, it indicates that the digital twin model has not accurately captured the high-emission link, and the actual data is marked as implicit carbon emission data; If the deviation between the actual carbon emission data and the historical data is less than the deviation threshold, it means that the digital twin model accurately captures the high-emission links and does not mark the actual data as hidden carbon emission data.

3. A carbon emissions management method based on digital twins according to claim 2, characterized in that: In S3, the full life cycle carbon emissions under test scenario A and test scenario B are predicted as follows: Input the data of test scenario A and use historical carbon emission data as model input. When running the digital twin model, the model will predict the carbon emissions of the entire life cycle based on the known historical data, including from raw material procurement to final disposal of the product. The model outputs the prediction of carbon emissions of the entire life cycle. The output prediction results should include the distribution of carbon emissions, classified by each link, record the predicted carbon emissions of the entire life cycle under scenario A, and mark the emission data of each link; On the basis of test scenario A, add invisible carbon emission data, integrate the invisible carbon emission data into the model input, run the digital twin model, and predict the full life cycle carbon emissions after adding the invisible emission data. The same as scenario A, the output results will include the carbon emissions of the whole life cycle and its distribution by link, record the predicted full life cycle carbon emissions under scenario B, and mark the emission data of each link.

4. The carbon emissions management method based on digital twin according to claim 1 is characterized by: In S4, the deviation rate fluctuation between the emission of test scenario A and the emission of test scenario B is analyzed to generate a deviation rate fluctuation index. The deviation rate fluctuation index is obtained as follows: Collect the carbon emission prediction data under test scenario A and test scenario B, and calculate the deviation rate between the emissions of scenario A and scenario B at the i-th moment. The deviation rate formula is: ; In the formula, is the emission of test scenario B at the i-th moment, is the emission of test scenario A at the i-th moment, is the deviation rate at the i-th moment, indicating the difference between scene B and scene A; Use weighted moving average to calculate the average of the deviation rate , the calculation expression is: ;in, is the weight given to the deviation rate at the jth moment, k is the window size, and i is the current moment; The deviation rate fluctuation index is used to measure the degree of deviation fluctuation between test scenarios A and B, reflecting the degree of interference of invisible carbon emission data on carbon emission reduction targets. The calculation expression is: ;in, To calculate the maximum value of the weighted moving average within the window period, To calculate the minimum value of the weighted moving average within the window period, is the deviation rate volatility index.

5. A carbon emissions management method based on digital twins according to claim 4, characterized in that: In S5, after analyzing the comparison results of the energy price fluctuation data in the external environment and the energy price historical data in the digital twin model, the energy consumption price elasticity abnormality index is generated. The method for obtaining the energy consumption price elasticity abnormality index is as follows: Collect the energy price data P and energy consumption data Q of the enterprise, as well as the control variables that affect energy consumption, including production volume X, weather factors W and technical parameters T; divide the data into training sets and test sets, and set the relationship between energy consumption Q and price P to follow the power function form: ; Converting it to logarithmic form, ; In the formula, is the price elasticity coefficient of energy price; the loss function of GLM is defined as the log-likelihood function; ; In the formula, is the likelihood function of the ith sample, are model parameters, ; Determine the parameters by maximizing the log-likelihood estimate L ; Using the training set data, the GLM model is fitted using an iterative optimization algorithm to obtain parameter estimates , validate the fitted model on the test set and calculate the predicted value With actual value The price elasticity is the sensitivity of energy price changes to energy consumption changes, and its formula is: ; According to the GLM fitting results, the price elasticity is given by Direct expression; The deviation between the predicted price elasticity at each time point and the average price elasticity of the model is calculated to generate the price elasticity anomaly index, which is expressed as: ; Where KML is the abnormal price elasticity index, ; represents the price elasticity calculated based on actual data; , represents the price elasticity predicted by the model.

6. A carbon emissions management method based on digital twins according to claim 5, characterized in that: In S6, the degree of data bias in the digital twin model is evaluated based on the degree of interference of invisible carbon emission data on carbon reduction targets and the impact of energy price fluctuations on corporate energy consumption and carbon emissions. Specifically: The deviation rate volatility index and the price elasticity anomaly index are converted into comprehensive feature vectors, and the comprehensive feature vectors are used as the input of the machine learning model. The machine learning model uses each group of comprehensive feature vectors to predict the data bias degree value label in the digital twin model as the prediction target, and takes minimizing the sum of prediction errors of the data bias degree value labels in all digital twin models as the training target. The machine learning model is trained until the sum of prediction errors converges, then the model training is stopped, and the data bias degree value in the digital twin model is determined according to the model output results, wherein the machine learning model is a polynomial regression model.

7. The carbon emissions management method based on digital twin according to claim 6 is characterized by: The results of the digital twin model reflecting the actual carbon emissions are divided into accuracy reflection results and inaccuracy reflection results. Specifically, the data bias degree value in the acquired digital twin model is compared with the reference threshold of the data bias degree value pre-set according to historical data. If the data bias degree value in the digital twin model is greater than or equal to the reference threshold of the data bias degree value, it means that the data bias degree in the digital twin model is high. At this time, a data bias abnormality signal is generated, and the results of the digital twin model reflecting the actual carbon emissions are divided into inaccuracy reflection results; if the data bias degree value in the digital twin model is less than the reference threshold of the data bias degree value, it means that the data bias degree in the digital twin model is low. At this time, no data bias abnormality signal is generated, and the results of the digital twin model reflecting the actual carbon emissions are divided into accuracy reflection results.

8. The carbon emissions management method based on digital twin according to claim 7 is characterized by: For inaccurate reflection results, the carbon emissions predicted by the real-time monitoring model and actual emissions , calculate the deviation value: ; If ΔE>ΔE threshold, it means that the model underestimates emissions and dynamic management measures need to be initiated; Carbon emissions predicted by the digital twin model, is the actual carbon emission data collected by the monitoring equipment; the ΔE threshold represents the threshold of the allowable deviation range; By monitoring the operating data of equipment and processes, the operating parameters of related equipment are dynamically adjusted to avoid overload operation. The adjustment formula is: ; is the adjusted equipment operating parameters, is the current equipment operating parameters; when the emission underestimate of a certain link is identified, resources are reallocated to optimize the process or equipment with high energy consumption; the optimization formula is: ; In the formula, To optimize resources allocated to a certain link, To optimize resources in total; according to the field measurement data, dynamically adjust the emission factors of each link, the adjustment formula is: ; is the updated emission factor, is the current emission factor; the long-term emission risk is calculated through cumulative deviation to prevent the negative cumulative effect caused by underestimation of emissions. The cumulative emission deviation formula is: ; is the emission deviation in the tth time period, T is the total number of monitored time periods, and when the cumulative deviation exceeds the threshold, an early warning signal is triggered.

9. A carbon emission management system based on digital twin, used to implement a carbon emission management method based on digital twin according to any one of claims 1 to 8, characterized in that: It includes data collection module, comparison module, test module, interference assessment module, external environment analysis module and classification management module: Data collection module: collects historical carbon emission data from different links of the supply chain, including carbon emission data generated during outsourced logistics, raw material processing and energy extraction; Comparison module: collects actual carbon emission data from suppliers’ production links and logistics transportation, and compares the collected actual carbon emission data with the historical carbon emission data in the digital twin model. Based on the comparison results, marks the invisible carbon emission data in the digital twin model. Test module: Create two sets of test scenarios, including test scenario A that only uses historical carbon emission data, and test scenario B that embeds invisible carbon emission data. Run the digital twin model to predict the full life cycle carbon emissions under test scenario A and test scenario B respectively; Interference assessment module: Based on the prediction results, calculate the emission differences between test scenarios A and B, and compare the optimization effects before and after the introduction of invisible carbon emission data. Based on the optimization effects, evaluate the degree of interference of invisible carbon emission data on carbon reduction targets; External environment analysis module: If the interference level is high, the energy price fluctuation data in the external environment will be compared with the historical energy price data in the digital twin model, and the impact of energy price fluctuations on the enterprise's energy consumption and carbon emissions will be analyzed based on the comparison results; Classification management module: According to the degree of interference of hidden carbon emission data on carbon reduction targets and the impact of energy price fluctuations on corporate energy consumption and carbon emissions, the degree of data bias in the digital twin model is evaluated, and based on the evaluation results, the results of the digital twin model reflecting the actual carbon emissions are divided into accurate reflection results and inaccurate reflection results, and corresponding carbon emissions management strategies are adopted.

Images