A Dynamic Accounting and Calibration Method for Carbon Emissions of Coal Enterprises Based on Digital Twins
By constructing a digital twin of the entire coal production process and using transfer learning technology, combined with Kalman filtering closed-loop feedback, dynamic calibration of carbon emission accounting for coal enterprises was achieved. This solved the problem that static factors could not adapt to dynamic changes, and improved the accuracy and adaptability of the accounting.
Patent Information
- Application Number
- CN202611354991.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-03
- Publication Date
- 2026-11-17
AI Technical Summary
In existing carbon emission accounting methods, static emission factors cannot capture the dynamic impact of fuel quality fluctuations, changes in combustion equipment efficiency, and the degree of cleanliness of the power grid, resulting in systematic deviations between the accounting results and the actual situation. Furthermore, the difficulty in obtaining measured data and the insufficient data sample size limit dynamic updates.
A digital twin of the entire coal production process is constructed. Dynamic emission factors are generated through digital twin technology. Combined with transfer learning and Kalman filtering closed-loop feedback, the emission factors are corrected and calibrated in real time to adapt to changes in operating conditions.
Without the need to add on-site monitoring equipment, it improves the accuracy and adaptability of carbon emission accounting, ensures the reliability and accuracy of accounting results, and is applicable to different mining methods and data abundance conditions.
Smart Images

Figure CN123117916A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of carbon emission technology, specifically to a method for dynamic accounting and calibration of carbon emissions for coal enterprises based on digital twins. Background Technology
[0002] Carbon emission accounting for coal production enterprises is highly dependent on the accuracy of emission factors. Current mainstream accounting methods (such as IPCC Tier 1 and Tier 2 methods) use fixed default emission factors or industry default factors, which remain unchanged for several years once determined. However, in actual production, fuel quality fluctuates with changes in mining areas, the efficiency of combustion equipment changes with operating conditions, and the cleanliness of the power grid varies from year to year. Static factors cannot capture the impact of these dynamic factors, leading to a systematic deviation between the accounting results and actual conditions.
[0003] In recent years, some studies have attempted to use data-driven methods to dynamically update emission factors, but they face two major bottlenecks: First, it is difficult to obtain actual measurement data, as most coal mines do not have the conditions for online coal quality analysis and online gas flow monitoring; second, the data sample size is insufficient, as the historical energy consumption ledgers of enterprises usually only have 1 to 2 years of data, which is difficult to support the training of complex models. Summary of the Invention
[0004] This invention aims to provide a dynamic carbon emission accounting and calibration method for coal enterprises based on digital twins. By relying on digital twin technology, a dedicated twin model for carbon emission accounting of coal enterprises is constructed. Rich simulation data is generated without relying on a large amount of measured data, and emission factors are dynamically acquired to achieve carbon emission accounting and calibration.
[0005] The basic solution provided by this invention is: a dynamic accounting and calibration method for carbon emissions of coal enterprises based on digital twins, including: S1, construct a digital twin for the entire coal production process, including a set of system twin models corresponding to different coal mining methods; S2, based on real-time simulation operation of digital twins, generates dynamic emission factors of each system under different operating conditions, forming an emission factor library indexed by operating parameters; S3, under preset conditions, uses the emission factor real-time correction model to correct the dynamic emission factor in real time; S4 employs transfer learning technology to construct a mapping bridge from the simulated space to the physical space, transferring the mapping relationship between the emission factors learned by the digital twin under simulated operating conditions and the changes in operating parameters to the physical operating system, so as to perform transfer calibration on the emission factors output by the digital twin. S5 establishes a Kalman filter closed-loop feedback loop between the digital twin and the actual energy consumption data output by the physical operating system. The actual energy consumption data is assimilated into the digital twin through the Kalman filter algorithm, and the parameters of the digital twin are continuously updated. This is used to output the emission factor calibrated by the Kalman filter in the final transfer learning process in S4, so as to dynamically participate in the final carbon emission accounting.
[0006] The working principle and advantages of this invention are as follows: This invention introduces digital twin technology specifically into the carbon emission accounting scenario of coal enterprises. Based on this technology, a high-fidelity digital image of the physical entity is constructed in virtual space. Rich simulation data is generated without relying on a large amount of measured data, enabling carbon emission accounting to shift from static default factors to dynamic twin calibration. This significantly improves the accuracy and adaptability of carbon emission accounting for coal mining enterprises without the need to add on-site monitoring equipment. The core advantages are: This invention constructs a digital twin of the entire coal production process. Through batch simulation of this digital twin, a dynamic emission factor library covering multiple operating conditions is generated, effectively solving the problem of dynamic emission factors under conditions of insufficient measured data. More importantly, it employs a real-time emission factor correction function based on equipment operating parameters, enabling the calculation model to respond to changes in operating conditions in real time. This overcomes the static limitations of traditional methods where emission factors are adjusted annually or remain unchanged for several years, achieving adaptive adjustment of emission factors under operating conditions without relying on measured data.
[0007] This invention is based on constructing a set of system twin models corresponding to different coal mining methods, forming a unified digital twin accounting technology framework for underground and open-pit mines. It can be selectively loaded for carbon emission accounting calibration under the target coal mining method, supports rapid parameter adaptation under different mining conditions, has good scenario adaptability, and is suitable for promotion and application under different mining methods and different data abundance conditions. It provides key technical support for the digital transformation and upgrading of carbon emission accounting in the coal industry.
[0008] This invention employs transfer learning technology to construct a mapping bridge from simulated space to physical space, transferring the knowledge of the twin model to the actual production scenario. This effectively solves the problem of distribution differences between simulated space and physical space, enabling knowledge to be effectively generalized even under conditions of sparse actual data, and ultimately overcoming the limitations of small sample data on model training.
[0009] This invention constructs a Kalman filter closed-loop feedback loop between the digital twin and actual energy consumption data, enabling continuous self-calibration of the twin model and ensuring that the model parameters remain synchronized with the actual production conditions of the mine. Based on this, the accuracy of the calibration data output by the digital twin is significantly improved, accurately reflecting the actual production situation of the mine and providing a high-confidence data foundation for subsequent carbon emission accounting, fundamentally guaranteeing the reliability of the accounting results. Attached Figure Description
[0010] Figure 1 A flowchart illustrating the dynamic carbon emission accounting and calibration method for coal enterprises based on digital twins provided in this embodiment of the invention. Figure 1 ; Figure 2 A flowchart illustrating the dynamic carbon emission accounting and calibration method for coal enterprises based on digital twins provided in this embodiment of the invention. Figure 2 . Detailed Implementation
[0011] The following detailed explanation illustrates the specific implementation methods: The basic implementation examples are as follows: Figure 1 As shown: A dynamic accounting and calibration method for carbon emissions of coal enterprises based on digital twins, including: S1, construct a digital twin for the entire coal production process, including a set of system twin models corresponding to different coal mining methods; S2, based on real-time simulation operation of digital twins, generates dynamic emission factors of each system under different operating conditions, forming an emission factor library indexed by operating parameters; S3, under preset conditions, uses the emission factor real-time correction model to correct the dynamic emission factor in real time; S4 employs transfer learning technology to construct a mapping bridge from the simulated space to the physical space, transferring the mapping relationship between the emission factors learned by the digital twin under simulated operating conditions and the changes in operating parameters to the physical operating system, so as to perform transfer calibration on the emission factors output by the digital twin. S5 establishes a Kalman filter closed-loop feedback loop between the digital twin and the actual energy consumption data output by the physical operating system. The actual energy consumption data is assimilated into the digital twin through the Kalman filter algorithm, and the parameters of the digital twin are continuously updated. This is used to output the emission factor calibrated by the Kalman filter in the final transfer learning process in S4, so as to dynamically participate in the final carbon emission accounting.
[0012] like Figure 2 As shown, specifically: In S1, a set of corresponding system twin models is constructed based on the coal mining method.
[0013] The process of constructing twin models for each system is as follows: Mechanism simulation models are established based on the physical mechanism equations (including fluid dynamics equations, thermodynamic equations, and motor efficiency equations) of each coal production system. Historical energy consumption data is used to verify the model parameters, ensuring that the root mean square error between the output of the corresponding system twin model and the actual data of that system is less than a preset threshold. Finally, twin models for each system are obtained. In practical applications, the preset threshold is reasonably determined based on model training and is not limited here.
[0014] In this embodiment, the mining method includes underground and open-pit mining, and the digital twin includes: Twin models of the six major systems in underground mines: ventilation, drainage, hoisting, mining, transportation, and coal preparation. (1) Twin model of ventilation system: Based on the theory of mine ventilation network and fluid dynamics equations, a simulation model of ventilation network with main ventilator and local ventilator as core components is established. Input parameters include roadway wind resistance, natural wind pressure, and ventilator characteristic curves; outputs are the air volume distribution and ventilator power consumption at each air usage location.
[0015] (2) Twin model of drainage system: Based on the water inflow prediction model and pump characteristic curve, a simulation model of drainage system is established. The inputs are mine water inflow, drainage height, and pump operation combination; the output is the energy consumption of drainage system.
[0016] (3) Twin Model of Hoisting System: Based on the dynamic equations of the hoist and the motor efficiency model, a simulation model of the hoisting system of the main and auxiliary shafts is established. The inputs are hoisting load, hoisting speed, and hoisting height; the output is the energy consumption of the hoisting system.
[0017] (4) Twin model of mining system: A simulation model is established based on the cutting resistance model and motor power model of the coal mining machine and the tunneling machine. The inputs are coal seam hardness, cutting depth and advance speed; the output is the energy consumption of the mining system.
[0018] (5) Twin model of transportation system: A simulation model is established based on the running resistance model and driving power model of the belt conveyor. The inputs are transportation volume, transportation distance, and belt speed; the output is the energy consumption of the transportation system.
[0019] (6) Coal preparation system twin model: A simulation model is established based on the energy consumption models of heavy medium separation, flotation, dewatering and other processes. The inputs are the amount of raw coal to be prepared, the washing density and the product structure; the output is the energy consumption of the coal preparation system.
[0020] Twin models of the three major systems of open-pit mines: coal mining, stripping, and drainage: (1) Twin model of coal mining system: A simulation model is established based on the excavation resistance model and fuel consumption model of mining equipment.
[0021] (2) Stripping the twin model of the semi-continuous system: A simulation model is established based on the running resistance model of the belt conveyor system.
[0022] (3) Twin model of drainage system: A simulation model is established based on the prediction model of mine water inflow and the characteristic curve of water pump.
[0023] The general formula for each system twin model is:
[0024] in, These are operating parameters, including measurable operating conditions such as production levels, number of operating units, ambient temperature, and coal quality indicators. This is the model parameter vector, including physical parameters that need to be determined through verification, such as the rated power, efficiency curve fitting coefficient, and drag coefficient of each device. Model parameters The initial values are derived from the equipment nameplate data and design documents, and are continuously updated through Kalman filter closed-loop feedback calibration.
[0025] In practical applications, this method is applicable to different coal mining methods (underground and open-pit mining). In S2, by selecting and loading the system twin model set corresponding to the target coal mining method, a digital twin is created for real-time simulation to complete the dynamic carbon emission accounting and calibration under that target coal mining method. By selecting and loading system twin model sets corresponding to different coal mining methods, mode switching is achieved, forming a unified digital twin accounting technology framework.
[0026] Unlike conventional digital twin construction, the digital twin built in this step involves establishing a parametric model based on the actual physical mechanisms of various equipment systems in coal production (such as fan characteristic curves, pump efficiency curves, and motor efficiency equations). The initial values of the model parameters are derived from equipment nameplate data and design documents, and are continuously updated through subsequent Kalman filter closed-loop feedback, ensuring that the twin model remains synchronized with the actual equipment status. The advantage of this construction method is that it achieves dynamic consistency between the model and the physical system without the need for additional sensors, and it supports rapid parameter adaptation under different mine conditions.
[0027] In S2, the dynamic emission factor library is generated through batch simulations of digital twins under various operating conditions (including different output levels, different equipment operating combinations, and different seasonal conditions). The unit product energy consumption is obtained by dividing the theoretical energy consumption of each system output by the simulated output, then multiplying this by the grid emission factor to obtain the dynamic emission factor, which is then stored in a factor library indexed by operating parameters. Specifically: After the digital twin is constructed, a batch simulation is performed. Multiple sets of operating condition parameter combinations are set, as shown in Table 1.
[0028] Table 1. Schematic diagram of parameter combinations for batch simulation of digital twins
[0029] The digital twin operates under each combination of operating parameters and outputs the theoretical energy consumption of each system. Meanwhile, the digital twin outputs simulated production based on the cutting resistance model of the mining equipment (inputting coal seam hardness, cutting depth, and advance speed). This is used as the denominator in the calculation of energy consumption per unit product. Therefore, the dynamic emission factor is:
[0030] The dynamic emission factors for all operating condition combinations are stored in a factor library indexed by the operating condition parameters:
[0031] in, For the first Various working condition parameter vectors For the first The emission factors corresponding to each operating condition.
[0032] In S3, emission factors are corrected in real time: When the actual operating conditions in production do not match the operating conditions in the emission factor database, an interpolation method is used to interpolate several emission factors corresponding to several operating conditions in the emission factor database to obtain theoretical emission factors that match the actual operating conditions. Considering the equipment operating status, the theoretical emission factors are corrected in real time based on the multivariate nonlinear function of the equipment performance curve. Specifically: S31, the actual operating conditions in production do not match the operating conditions in the factor library, i.e., comparing the actual operating condition parameter vector. The working condition parameter vector corresponding to each discrete working condition point in the factor library If the minimum distance is greater than a preset threshold (e.g., 5% of the range of parameters in each dimension), it is considered an incomplete match. In this case, the nearest discrete operating point (e.g., the four nearest points) in the factor library to the actual operating condition is used as interpolation nodes. Inverse distance weighted interpolation or linear interpolation methods are used to interpolate the emission factors corresponding to the nearest discrete operating points. Interpolation calculations are performed to obtain the theoretical emission factor that matches the actual operating conditions. .
[0033] Furthermore, considering real-time correction based on equipment operating status, the theoretical emission factor obtained from interpolation is corrected in real time using the equipment's real-time load rate, actual operating time, and equipment efficiency decay coefficient as input variables. The correction formula is as follows:
[0034] in, It is a multivariate nonlinear function based on the device performance curve.
[0035] In practical applications, the following can be adopted: (1) Equipment load rate correction: For motor-driven equipment, when the actual load rate deviates from the rated load rate, the efficiency deviates from the rated efficiency. Based on the equipment efficiency curve... The emission factor correction coefficient is:
[0036] in This is a function curve showing the change in equipment efficiency with load rate. This represents the efficiency value of equipment (determined based on energy consumption statistics, not limited to a single device) at its rated load rate. This represents the efficiency value under the actual load rate. Both are scalar values, and dividing them yields the equipment load rate correction factor.
[0037] (2) Equipment aging correction: As the equipment's operating years increase, its efficiency tends to decline. Based on the equipment degradation curve, the equipment aging correction factor is:
[0038] in, The degradation coefficient is... The cumulative service life of the equipment. Degradation coefficient. The value is obtained by fitting historical operating data of the equipment (such as efficiency test data in regular maintenance records) or by estimating based on the life decay curve provided by the equipment manufacturer. The initial value can be set as the reciprocal of the equipment's design life, and then adaptively updated through Kalman filter closed-loop feedback.
[0039] (3) Comprehensive correction formula:
[0040] in, For other environmental factors (such as the effect of temperature on fan efficiency). The value is determined by consulting the environmental correction table in the equipment manual or by calibrating based on field test data, and is usually in the range of 0.95 to 1.05.
[0041] In S4, the transfer learning mapping bridge employs an instance-based transfer learning method. By calculating the similarity of feature distributions between operating conditions in the factor library and actual operating conditions, it weights and adjusts the emission factors output by the digital twin, enabling the emission response patterns trained in the simulated space to effectively generalize under conditions of sparse real-world data. Specifically: The emission response patterns learned by the digital twin under various operating conditions in the simulated space (i.e., the mapping relationship between the real-time corrected emission factor and the changes in operating parameters). It needs to be migrated to physical space. Let the characteristic distribution of the simulated space operating conditions be... ,in It is a feature vector of operating conditions (including output, number of operating equipment, ambient temperature, etc.). Probability distribution representing operating condition characteristics in the simulated space; distribution of operating condition characteristics in the physical space. Similarly, this is obtained from statistical analysis of actual production data. The difference in distribution between the two is measured by the Maximum Mean Discrepancy (MMD):
[0042] in, For kernel function mapping, For the regenerating nucleus Hilbert space, For samples taken from simulated space conditions or physical space conditions, MMD is used to calculate the distance between two mean vectors in a high-dimensional space. It measures the distance between the data distribution in the simulated space (source domain) and the data distribution in the physical space (target domain / actual mine). The smaller the MMD value, the closer the two distributions are, and the better the transfer learning effect.
[0043] Instance-based transfer learning is used to weight and adjust the emission factors of the digital twin output. For the first factor in the factor library... For each simulated sample, the transfer weights are:
[0044] in, This is the vector of operating condition parameters corresponding to the kth simulation sample in the factor library; This is the current actual operating condition parameter vector. This is a bandwidth parameter (a scale-adjusting hyperparameter in kernel functions, especially Gaussian / RBF kernels), which determines the sensitivity of the kernel function to the distance between data points. The larger the value, the more lenient the judgment. The smaller the value, the stricter the judgment. The median of the Euclidean distances between all sample pairs can be used as the threshold. ).
[0045] Then the migration-calibrated emission factor for:
[0046] That is, for a single system, the current period basic emission factor calculated and corrected in S1 and S2 is used, and then the migration weight set determined by each working condition and the actual working condition in the factor library is used to calculate the current period migration-calibrated emission factor of the system by weighted average.
[0047] In S5, the Kalman filter closed-loop feedback uses the residual between the model state vector of the digital twin and the observation vector representing the actual energy consumption data as the feedback signal. The Kalman gain matrix is used to optimally estimate and update the model state. The recursive formula is as follows:
[0048] in Here is the Kalman gain matrix. These are actual observed values. This is the observation matrix.
[0049] Specifically: S51 constructs a Kalman filter closed-loop feedback loop between the digital twin and the actual energy consumption data.
[0050] Model state vector ,in, This includes the efficiency coefficient of the ventilation fan, the efficiency coefficient of the water pump, and the efficiency coefficient of the hoist. This includes monthly production levels, the number of equipment operating combinations, and average ambient temperature; Observation vector These data come from the monthly energy consumption ledgers of the enterprise's energy management system.
[0051] State transition matrix The observation matrix describes the slow change of model parameters as the equipment ages. Map the state to the predicted energy consumption values of each system.
[0052] The state transition model and the observation model are as follows:
[0053]
[0054] in, and These are process noise and observation noise, respectively.
[0055] The Kalman filter recursive process is as follows: (1) Prior state estimation:
[0056] That is, to calculate the preliminary prediction value at the current time. . Based on the kth The kth time estimated from all measurement data up to time 1 The optimal state value at time 1. Let be the state transition matrix.
[0057] (2) Prior error covariance:
[0058] This involves calculating the confidence level (covariance P) of the initial prediction. The error from the previous time step... The data is transmitted over, along with process noise. (Random disturbances within the system itself) will increase the uncertainty of the predicted values.
[0059] (3) Kalman gain:
[0060] That is, gain K The decision is whether to place more trust in the prediction or the measurement. If the measurement noise R is small (the system itself has little random interference), the gain increases, and the measurement is more trusted. If the prediction covariance P is small (the model is very accurate), the gain decreases, and the prediction is more trusted.
[0061] (4) Posterior state estimation:
[0062] That is, to calculate the final estimated value. .
[0063] (5) Posterior error covariance:
[0064] This refers to the uncertainty of the current estimate. Due to the introduction of measurement information, the new covariance... Usually more accurate than predicted Smaller (more certain).
[0065] Through the above recursion, new monthly actual energy consumption observations for each system are obtained each time. Then, the Kalman filter algorithm is combined with the current twin model for state prediction. and observation matrix Calculate the optimal state estimate The model parameters The system is updated to minimize the deviation between the energy consumption predicted by the twin model and the actual energy consumption, thus achieving continuous synchronization between the model and the actual system.
[0066] S6, Dynamic Carbon Emission Accounting Output Based on the dynamic emission factor that has been corrected in real time and calibrated by Kalman filtering, combined with the enterprise's monthly energy consumption ledger, the final carbon emission accounting result is output.
[0067] The calculation formulas for direct emissions (Scope 1) and indirect emissions (Scope 2) for the accounting cycle (monthly) are as follows:
[0068]
[0069] in, Let be the direct emissions for period t (month), where For the t-th period (month) The consumption of various fuels, For the t-th period (month) after digital twin calibration The dynamic emission factor of the fuel (it should be noted that this emission factor is not specifically limited to the emission factor after real-time correction, but can also be the emission factor after migration calibration, and the emission factor of the final processing output of the digital twin at the time of current accounting shall prevail). Let be the indirect emissions for period t (month), where For the first Purchased electricity consumption per cycle (month), For the first The grid emission factor over a period of time (month).
[0070] Total carbon emissions in period t (month) ;for:
[0071] The dynamic carbon emission accounting and calibration method for coal enterprises based on digital twins provided in this embodiment introduces digital twin technology specifically into the carbon emission accounting scenario of coal enterprises. Relying on this technology, a high-fidelity digital image of the physical entity is constructed in virtual space, generating rich simulation data without relying on a large amount of measured data. This realizes the transformation of carbon emission accounting from static default factors to dynamic twin calibration, and significantly improves the accuracy and adaptability of carbon emission accounting for coal mining enterprises without the need to add on-site monitoring equipment.
[0072] The above descriptions are merely embodiments of the present invention. Commonly known structures and characteristics of the solutions are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent.
Claims
1. A method for dynamic accounting and calibration of carbon emissions for coal enterprises based on digital twins, characterized in that, include: S1, construct a digital twin for the entire coal production process, including a set of system twin models corresponding to different coal mining methods; S2, based on real-time simulation operation of digital twins, generates dynamic emission factors of each system under different operating conditions, forming an emission factor library indexed by operating parameters; S3, under preset conditions, uses the emission factor real-time correction model to correct the dynamic emission factor in real time; S4 employs transfer learning technology to construct a mapping bridge from the simulated space to the physical space, transferring the mapping relationship between the emission factors learned by the digital twin under simulated operating conditions and the changes in operating parameters to the physical operating system, so as to perform transfer calibration on the emission factors output by the digital twin. S5 establishes a Kalman filter closed-loop feedback loop between the digital twin and the actual energy consumption data output by the physical operating system. The actual energy consumption data is assimilated into the digital twin through the Kalman filter algorithm, and the parameters of the digital twin are continuously updated. This is used to output the emission factor calibrated by the Kalman filter in the final transfer learning process in S4, so as to dynamically participate in the final carbon emission accounting.
2. The method for dynamic accounting and calibration of carbon emissions of coal enterprises based on digital twins according to claim 1, characterized in that, In S1, the digital twin includes a set of twin models of the six major systems of underground mines: ventilation, drainage, hoisting, mining, transportation, and coal preparation, as well as a set of twin models of the three major systems of open-pit mines: coal mining, stripping, and drainage.
3. The method for dynamic accounting and calibration of carbon emissions of coal enterprises based on digital twins according to claim 1, characterized in that, In S1, a mechanism simulation model is established based on the physical mechanism equations of each coal production system, and the model parameters are checked using historical energy consumption data to ensure that the root mean square error between the output of the corresponding system twin model and the actual data of the system is less than a preset threshold. Finally, twin models of each system are obtained, and the parameters of each model are continuously updated through Kalman filter closed-loop feedback.
4. The method for dynamic accounting and calibration of carbon emissions of coal enterprises based on digital twins according to claim 1, characterized in that, In S2, by selecting and loading the system twin model set corresponding to the target coal mine mining method, a digital twin is created under the target coal mine mining method and simulated in real time to complete the dynamic accounting and calibration of carbon emissions under the target coal mine mining method.
5. The method for dynamic accounting and calibration of carbon emissions of coal enterprises based on digital twins according to claim 1, characterized in that, In S2, the dynamic emission factor is obtained by batch simulation operation of digital twins under various operating conditions, dividing the theoretical energy consumption benchmark value of each system output by the simulation output by the simulation output, and then multiplying by the grid emission factor.
6. The method for dynamic accounting and calibration of carbon emissions of coal enterprises based on digital twins according to claim 1, characterized in that, In S3, when the actual operating conditions in production do not match the operating conditions in the emission factor library, an interpolation method is used to interpolate several emission factors corresponding to several operating conditions in the emission factor library to obtain theoretical emission factors that match the actual operating conditions; considering the equipment operating status, the theoretical emission factors are corrected in real time based on the multivariate nonlinear function of the equipment performance curve.
7. The method for dynamic accounting and calibration of carbon emissions of coal enterprises based on digital twins according to claim 6, characterized in that, The real-time correction adopts the following model: in, This is the theoretical emission factor after real-time correction; Theoretical emission factor; This is the equipment load rate correction factor, where This represents the efficiency value of the equipment at its rated load rate. This is the efficiency value under the actual load rate; This is the equipment aging correction factor, where The degradation coefficient is... The cumulative service time of the equipment; This is a correction factor for other environmental factors.
8. The method for dynamic accounting and calibration of carbon emissions of coal enterprises based on digital twins according to claim 1, characterized in that, In S4, the mapping bridge uses an instance-based transfer learning method. By calculating the difference in feature distribution between the operating conditions in the emission factor library and the actual operating conditions in production, the emission factors output by the digital twin are weighted for transfer calibration. Migration-calibrated emission factors for: in, The emission factor corresponding to the kth operating condition in the emission factor library; For migration weights; MMD is the maximum mean difference. The parameter vector for the kth operating condition in the emission factor library; This is the current actual operating condition parameter vector; This is the bandwidth parameter.
9. The method for dynamic accounting and calibration of carbon emissions of coal enterprises based on digital twins according to claim 1, characterized in that, In S5, after obtaining the observed values representing the actual energy consumption data of each system in a new cycle, the Kalman filter algorithm combines the current state prediction and observation matrix of the twin model of each system to calculate the optimal state estimate and update the parameters of the twin model of each system, so as to minimize the deviation between the energy consumption data of each system predicted by the twin model and the actual energy consumption data of the system.
10. The method for dynamic accounting and calibration of carbon emissions of coal enterprises based on digital twins according to claim 1, characterized in that, In S5, the process of dynamic carbon emission accounting is as follows: in, Let t be the total carbon emissions in period t; Let be the direct emissions in period t, where For the t-th period The consumption of this type of fuel For the t-th period, the calibrated digital twin is... Dynamic emission factors of various fuels; Let be the indirect emissions in period t, where For the first Cyclic external electricity consumption For the first Periodic grid emission factors.
