Multi-source fusion carbon emission factor adaptive calculation system and method

Abstract

The invention discloses a multi-source fusion carbon emission factor adaptive calculation system and method, and relates to the technical field of carbon emission management, and the system comprises a data input module, a sub-model module, a total model integration generation module, a model weight adjustment module and a carbon emission factor calculation system module. The sub-model module can generate a calculation model of the four sub-models according to a fuel type, a region, a time sequence and a process type, and the total model integration generation module integrates the four sub-models and generates a Stacking integrated prediction large model. According to the method, four sub-models can be generated according to the known fuel, region, time and process information, the four sub-models can be integrated into a large judgment model, the large judgment model can be applied to calculate the result according to the characteristics of different industries and different scenes through the large judgment model, and the prediction precision can be improved in a large range.

Description

Technical Field

[0001] The present invention relates to the technical field of carbon emission management, and specifically to a multi-source fusion adaptive carbon emission factor calculation system and method. Background Technique

[0002] As a key basic parameter for carbon emission accounting, the accuracy of carbon emission factors not only affects the reliability of greenhouse gas emission inventories, but also serves as an important scientific basis for formulating emission reduction policies, carbon market transactions, and corporate low-carbon transformation decisions. Therefore, improving the accuracy and applicability of carbon emission factors has become a core issue of common concern among current government regulatory departments, industrial enterprises, and research institutions.

[0003] However, the current calculation of carbon emission factors is generally carried out by staff making rough judgments based on known situations and information. However, in this way, only small-scale judgment operations can be performed. When the carbon emission information to be calculated is relatively large or there are many known variables, it is very easy for humans to make judgment errors. Summary of the Invention

[0004] To solve the problem of easy misjudgment mentioned in the above background technique, the purpose of the present invention is to provide a multi-source fusion adaptive carbon emission factor calculation system and method.

[0005] To achieve the above purpose, the present invention provides the following technical solutions: including a data input module, a sub-model module, a total model integration and generation module, a model weight adjustment module, and a carbon emission factor calculation system module; The sub-model module can generate calculation models for four sub-models according to fuel type, region, time series, and process type, namely a fuel sub-model, a region sub-model, a time series sub-model, and a process type sub-model; The total model integration and generation module integrates the four sub-models and generates a Stacking integrated prediction large model: (6); In the above formula (6), is the finally fused carbon emission prediction value, is the total carbon emission based on fuel type, is the total carbon emission based on regional characteristics, is the grid indirect emission factor, is the total carbon emission of a certain production process, is the weight of model i at time t; The model weight adjustment module performs fusion weight adjustment based on the similarity of scenarios and model performance, and its core formula is: (7); In the above formula (7), is the recent performance of the currently calculated model index i, is the scenario matching degree of the currently calculated model index i, is the recent performance of the index j of the previous model, is the scenario matching degree of the index j of the previous model.

[0006] Preferably, the calculation model of the fuel sub-model is: (1); In the above formula (1), is the total carbon emissions based on the fuel type, unit: ton CO2, is the activity of the i-th fuel, unit: ton, is the mass fraction of carbon in the i-th fuel (unit: ton CO2 / ton fuel), is the carbon oxidation ratio of the fuel combustion chamber in the i-th fuel (0 - 1), is the unit energy value of the fuel in the i-th fuel, unit MJ / kg; The calculation model of the regional sub-model is: (2); In the above formula (2), is the total carbon emissions based on regional characteristics, unit: ton CO2, and is the regional energy structure weight, obtained through the proportion of regional energy consumption or regression analysis, and needs to satisfy + = 1, is the indirect emission factor of the power grid, unit: ton CO2 / MWh; The above The calculation formula of is: (3); In the above formula (3), is the total carbon emissions of the regional power grid, is the total regional power generation; The calculation model of the time series sub-model is: (4); In the above formula (4), is the carbon emissions at time t, unit: ton CO2, is the linear long-term trend, is the trend slope, unit: ton CO2 / time unit, is the initial carbon emission base value, unit: ton CO2, is the seasonal adjustment term, m is the season or month number, is a random error, and the base value is 0; The calculation formula of the process type sub-model is: (5); In the above formula (5), is the total carbon emission of a certain production process, unit: ton CO2, is the energy consumption of displacement process j, unit: MWh, is the emission factor of process j, unit: ton CO2 / unit energy consumption.

[0007] Preferably, among them, in the core formula (8); In the above formula (8), is the time decay factor, is the time window length, is the normalized error of model i at time t - k, is the conversion of error to accuracy; In the core formula (9); In the above formula (9), is the current scenario feature vector, is the set of historical similar scenarios, is the scenario similarity metric, is the performance of model i in historical scenario s.

[0008] Preferably, the data input module is communicatively connected to the sub-model module, the sub-model module is communicatively connected to the total model integration and generation module, the total model integration and generation module is communicatively connected to the model weight adjustment module, and the carbon emission factor calculation system module is communicatively connected to both the total model integration and generation module and the model weight adjustment module; The data input module includes a manual information input unit and a network data collection unit; The manual information input unit includes a keyboard and an information entry camera, and the network data collection unit collects data in the network through a web crawler.

[0009] Preferably, the sub-model module includes a fuel sub-model unit, a region sub-model unit, a time series sub-model unit, and a process type sub-model unit; The fuel sub-model unit generates a corresponding fuel sub-model according to the information input by the data input module, the region sub-model unit generates a corresponding region sub-model according to the information input by the data input module, the time series sub-model unit generates a corresponding time series sub-model according to the information input by the data input module, and the process type sub-model unit generates a corresponding process type sub-model according to the information input by the data input module.

[0010] Preferably, the overall model integration and generation module includes a model set unit, an overall model generation unit, and a model output unit; The model set unit is used to integrate the fuel sub-model, the regional sub-model, the time series sub-model, and the process type sub-model.

[0011] Preferably, the model weight adjustment module includes multiple sub-model weight calculation units and a historical performance calculation unit; Through the sub-model weight calculation unit and according to the specific scenario, the dynamic weights of each sub-model are adaptively adjusted. The historical performance calculation unit can calculate the historical performance of each prediction model and compare multiple prediction models to obtain the best prediction data.

[0012] Preferably, the carbon emission calculation system module includes a model confirmation unit, a calculation unit, and a solution benchmark verification unit; The model confirmation unit can verify each sub-model. The calculation unit is used to provide calculation services for this system. The solution benchmark verification unit can further verify the relationship between the obtained overall model and each data and perform calibration and verification.

[0013] The realization of the multi-source fusion carbon emission factor adaptive calculation system specifically includes the following steps: S1. Input the required information through the data input module, where the information includes: fuel type, regional information, time series information, and process type information; S2. Respectively generate a fuel sub-model, a regional sub-model, a time series sub-model, and a process type sub-model through the sub-model module. The four sub-models are independently established and correspond to the information adopted by the data input module respectively; S3. Generate a prediction model through the overall model integration and generation module and the above four sub-models; S4. Adjust the weights of each item of the prediction model through the model weight adjustment module according to the recent performance and scenario matching degree information; S5. Calculate the prediction result through the carbon emission factor calculation system module.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. The present invention proposes an intelligent optimization system for carbon neutralization paths adaptable to multiple scenarios. The core lies in constructing a computational architecture for multi-level feature decoupling and dynamic fusion. The system separates the key influencing factors of carbon emissions and establishes four specialized analysis sub-modules for fuel characteristics, regional attributes, temporal evolution, and process technology respectively: The fuel analysis module deeply analyzes the combustion kinetics of different energy media and establishes a differential emission factor evaluation system; the regional evaluation module integrates geographical space characteristics and industrial layout characteristics to construct a regional-level carbon emission benchmark parameter library; the temporal prediction module uses advanced time series analysis methods to simultaneously capture the long-term evolution trend and short-term fluctuation law of carbon emissions; the process optimization module focuses on modeling carbon reduction technology parameters and quantitatively evaluates the impact of technology iteration on the overall carbon footprint. Through the intelligent fusion engine, the system can dynamically adjust the collaborative weights of each module according to the production characteristics and application scenarios of the target industry, forming an adaptive composite analysis and prediction model. This architecture effectively breaks through the limitations of traditional single models, ensuring both the in-depth analysis ability of specific dimension features and enhancing the adaptability to complex scenarios through cross-module knowledge fusion, providing a dynamically evolving decision support system for differentiated "one enterprise, one policy" carbon management solutions, and significantly improving the scientificity and feasibility of carbon neutralization planning.

[0015] 2. The present invention constructs an intelligent adaptive framework based on dynamic scenario perception and realizes the precise matching and continuous evolution of the prediction model through a double-layer adaptive optimization mechanism. The core of the system adopts a scenario-model dual-driven architecture: In the scenario analysis layer, through a multi-dimensional feature extraction engine, the target environment is deeply deconstructed, and scenario elements such as fuel thermodynamic characteristics, geographical space constraints, temporal evolution laws, and process flow parameters are quantitatively analyzed to construct a dynamic feature fingerprint map; in the model decision layer, a dynamic evaluation matrix of the prediction model performance is established to continuously track the prediction error rate, response speed, and generalization ability indicators of each sub-model under different feature dimensions. Based on the two-way matching mechanism, the system uses the Bayesian model averaging (BMA) technology to transform the scenario adaptation rules of the historical optimal model into a transferable weight allocation strategy. When a new scenario input is detected, the intelligent weight adjustment module dynamically fuses the advantageous components of four types of prediction models, namely fuel-specific, regionally adaptable, temporally sensitive, and process-optimized, according to the matching degree between the real-time scenario fingerprint and the model feature library, generating a scenario-customized composite prediction model. This mechanism breaks through the limitations of traditional static model fusion, and while ensuring the stability of basic predictions, significantly improves the analysis ability for complex working conditions through scenario-driven dynamic recombination of model components. Especially when dealing with typical scenarios such as cross-regional energy allocation, seasonal production capacity fluctuations, and the application of emerging carbon reduction technologies, the system can independently select the optimal model combination strategy to achieve the paradigm upgrade of carbon emission prediction results from "empirical fitting" to "mechanism-driven".

[0016] The parts not involved in this device are the same as those in the prior art or can be implemented using the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a system block diagram of the multi-source fusion adaptive carbon emission factor calculation system of the present invention.

[0018] Figure 2 It is a flowchart of the multi-source fusion adaptive carbon emission factor calculation method of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0019] To make the technical means, creative features, achieved purposes and effects of the present invention easy to understand, the present invention will be further described below in conjunction with specific embodiments.

[0020] As Figure 1 shown, the multi-source fusion adaptive carbon emission factor calculation system and method provided by the present invention include a data input module, a sub-model module, a total model integration and generation module, a model weight adjustment module and a carbon emission factor calculation system module; The sub-model module can generate calculation models of four sub-models according to fuel type (types: natural gas, coal gas, biomass; calorific value, carbon content, oxidation rate), regional data (regional energy structure, grid emission factor, climate condition), time series (historical carbon emissions, seasonal data, economic indicators) and process type (production equipment type (such as blast furnace / electric furnace), process flow parameters (energy consumption, efficiency)), namely fuel sub-model, regional sub-model, time series sub-model and process type sub-model; The calculation model of the fuel sub-model is: (1); In the above formula (1), is the total carbon emission based on fuel type, unit: ton CO2, is the activity of the i-th fuel, unit: ton, is the mass fraction of carbon in the i-th fuel, unit: ton CO2 / ton fuel, is the carbon oxidation ratio (0-1) in the fuel combustion chamber of the i-th fuel, is the unit energy value of the i-th fuel, unit MJ / kg; The calculation model of the regional sub-model is: (2); In the above formula (2), is the total carbon emission based on regional characteristics, unit: ton CO2, and is the weight of the regional energy structure, obtained through the proportion of regional energy consumption or regression analysis, and needs to satisfy + = 1, is the indirect emission factor of the power grid, unit: ton CO2 / MWh; The above The calculation formula is: (3); In the above formula (3), is the total carbon emission of the regional power grid, is the total power generation of the region; The calculation model of the time series sub-model is: (4); In the above formula (4), is the carbon emission at time t, unit: ton CO2, is the linear long-term trend, is the trend slope, unit: ton CO2 / time unit, is the initial carbon emission base value, unit: ton CO2, is the seasonal adjustment term, m is the season or month number, is the random error, with a base value of 0; The calculation formula of the process type sub-model is: (5); In the above formula (5), is the total carbon emission of a certain production process, unit: ton CO2, is the energy consumption of displacement process j, unit: MWh, is the emission coefficient of process j, unit: ton CO2 / unit energy consumption; The total model integration generation module integrates the four sub-models and generates a Stacking integrated prediction large model: (6); In the above formula (6), is the final fused carbon emission prediction value, is the weight of model i at time t; The model weight adjustment module adjusts the weight based on the fusion of scenario similarity and model performance, and its core formula is: (7); In the above formula (7), is the recent performance of the current calculated model index i, is the scenario matching degree of the current calculated model index i, is the recent performance of the index j of the previous model, is the scene matching degree of the index j of the previous model.

[0021] Among them, in the core formula (8); In the above formula (8), is the time decay factor, is the time window length, is the normalized error of model i at time t - k, is the conversion of error to precision; In the core formula (9); In the above formula (9), is the current scene feature vector, is the set of historical similar scenes, is the scene similarity metric, is the performance of model i in the historical scene s.

[0022] The data input module is communicatively connected to the sub - model module, the sub - model module is communicatively connected to the total model integration and generation module, the total model integration and generation module is communicatively connected to the model weight adjustment module, and the carbon emission factor calculation system module is communicatively connected to both the total model integration and generation module and the model weight adjustment module.

[0023] The data input module includes a manual information input unit and a network data collection unit.

[0024] The manual information input unit integrates a high - precision physical input interface and an OCR recognition camera, supporting the digital conversion of paper reports and the direct recording of on - site equipment parameters; the network data collection unit deploys an intelligent crawler engine, configures dynamic IP spoofing and anti - crawling strategies, realizes the scraping of multi - source heterogeneous data from industry databases, meteorological platforms and energy monitoring systems, and performs outlier filtering and format normalization processing through a data cleaning pipeline.

[0025] The sub - model module includes a fuel sub - model unit, a regional sub - model unit, a time - series sub - model unit and a process - type sub - model unit.

[0026] Each professional sub-model module adopts a feature-driven modeling strategy: the fuel sub-model unit constructs a thermodynamic parameter analysis matrix to analyze core indicators such as fuel calorific value and volatile content; the regional sub-model unit integrates a GIS spatial analysis engine to integrate terrain elevation, climate characteristics, and industrial distribution layer data; the time series sub-model unit deploys a time-frequency joint analysis algorithm to capture time series features such as production cycles and equipment maintenance rhythms; the process type sub-model unit establishes a process flow knowledge map to analyze process parameters such as reactor temperature curves and catalyst activity. Each sub-model generates a standardized feature vector through a feature fingerprint extraction layer and transmits it to the overall model integration generation module.

[0027] The overall model integration generation module includes a model collection unit, an overall model generation unit and a model output unit.

[0028] The overall model integration generation module adopts a three-level fusion architecture: the model collection unit realizes the spatial mapping of multi-dimensional vectors through the feature alignment engine and constructs a cross-modal feature association matrix; the overall model generation unit applies the improved Stacking integration technology and designs a two-layer meta-learning architecture - the base learner layer is configured with 6 types of heterogeneous algorithms such as LSTM and XGBoost, and the meta-learner layer deploys a feature weighted aggregator guided by the attention mechanism; the model output unit integrates the confidence assessment module (it can also be achieved through meta large model ai learning technology. In this embodiment, meta large model technology is applied), generates prediction intervals through Bootstrap sampling, and outputs carbon emission prediction results with credibility annotations.

[0029] The model weight adjustment module includes a plurality of sub-model weight calculation units and a historical performance calculation unit.

[0030] The model weight adjustment module builds a dynamic optimization closed loop: the sub-model weight calculation unit deploys a scene-aware neural network, performs similarity matching based on the spectrum characteristics of real-time input data (extracted through wavelet packet transform) and the historical scene library, and generates a dynamic weight allocation scheme in combination with reinforcement learning strategies; the historical performance calculation unit establishes a model performance tracking matrix and performs a comprehensive score in combination with the time decay factor. This module maintains real-time data synchronization with the carbon emission calculation system module through the digital twin interface.

[0031] The carbon emission calculation system module includes a model confirmation unit, a calculation unit and a solution benchmarking verification unit.

[0032] The carbon emission calculation system module forms a verification feedback loop: The model confirmation unit adopts an adversarial verification strategy, constructs a virtual test set through a cross-validation generator to evaluate the model robustness; the calculation unit deploys an edge-cloud collaborative architecture, with local nodes performing real-time inference and the cloud cluster conducting large-scale Monte Carlo simulations; the solution benchmark verification unit constructs an industry-level benchmark database, verifies the correlation strength between the prediction results and production process parameters and equipment energy efficiency indicators through Spearman correlation analysis, and generates a multi-dimensional diagnostic report to feedback to the weight adjustment module to form an optimization closed-loop. Industrial-level data interaction is achieved between modules through the OPC-UA protocol to ensure millisecond-level response and data integrity protection.

[0033] By separating the key influencing factors of carbon emissions, four specialized analysis sub-modules are established for fuel characteristics, regional attributes, temporal evolution, and process technology respectively: The fuel analysis module deeply analyzes the combustion kinetics of different energy media and establishes a differential emission factor evaluation system; the regional evaluation module integrates geospatial characteristics and industrial layout characteristics to construct a regional-level carbon emission benchmark parameter library; the temporal prediction module adopts advanced time series analysis methods to synchronously capture the long-term evolution trend and short-term fluctuation law of carbon emissions; the process optimization module focuses on carbon reduction technology parameter modeling and quantitatively evaluates the impact of technology iteration on the overall carbon footprint; through the intelligent fusion engine, the system can dynamically adjust the collaborative weights of each module according to the production characteristics and application scenarios of the target industry, forming an adaptive composite analysis and prediction large model. This architecture effectively breaks through the limitations of traditional single models, not only ensuring the in-depth analysis ability of specific dimension features but also strengthening the adaptability to complex scenarios through cross-module knowledge fusion, providing a dynamically evolving decision support system for differentiated "one enterprise, one policy" carbon management solutions, and significantly improving the scientificity and feasibility of carbon neutrality planning.

[0034] This application also includes a multi-source fusion-based adaptive calculation method for carbon emission factors, which specifically includes the following steps: S1. Implement full-dimensional information capture through the intelligent data input system: ① The fuel type input unit integrates the calorimeter analyzer interface to collect real-time fuel industrial analysis data (types: natural gas, coal gas, biomass; calorific value, carbon content, oxidation rate); ② The regional information collection engine deploys geofencing technology to automatically capture target area data (regional energy structure, grid emission factor, climate condition) and dynamic parameters of the grid carbon emission factor; ③ The temporal information processing module adopts a sliding window mechanism to synchronously analyze the production cycle curve, equipment start-stop log, and energy scheduling temporal characteristics (time series (historical carbon emissions, seasonal data, economic indicators)); ④ The process type identification unit establishes a process knowledge ontology library, and through process flow chart analysis and DCS signal feature matching, automatically labels the key parameters of the reaction path and process type (production equipment type (such as blast furnace / electric furnace), process flow parameters (energy consumption, efficiency)). S2. Generate four sub-models of fuel through the sub-model module respectively: ① For the fuel sub-model, thermodynamic state space modeling is adopted to establish a third-order response surface of fuel components - combustion efficiency - carbon emission coefficient, integrate a chemical kinetics simulator to calculate the free radical reaction path, and then generate the corresponding fuel prediction sub-model; ② For the regional sub-model, a spatial interpolation algorithm is deployed to construct a 10 km × 10 km grid emission benchmark library, integrate the WRF-Chem atmospheric diffusion model to calculate the local transport flux, and generate the corresponding regional prediction sub-model; ③ For the time series sub-model, a time-frequency joint analysis framework is designed to generate the corresponding time series prediction sub-model; ④ For the process sub-model, a process knowledge graph is constructed, a process parameter network (production equipment type (such as blast furnace / electric furnace), process flow parameters (energy consumption, efficiency)) is established, and the node correlation strength is optimized through a process constraint propagation algorithm, and then the corresponding process prediction sub-model is generated; S3. Through the intelligent fusion engine, the system can dynamically adjust the collaborative weights of each module according to the production characteristics and application scenarios of the target industry, and form an adaptive composite analysis and prediction large model; S4. The model weight adjustment system constructs a dual-channel feedback loop: ① The real-time scenario perception channel adopts deep metric learning, calculates the cosine similarity between the current input feature and the historical scenario library through a contrast network, and activates the historical optimal weight configuration with high matching degree; ② The performance evolution channel establishes a sliding time window (window length of 30 days, and the length can be extended to 90 days for special areas with a region larger than 500 hectares) evaluation system, quantifies the marginal contribution of each sub-model to the recent prediction accuracy based on the Shapley value algorithm, and dynamically updates the model weight in combination with a time decay factor (decay coefficient = 0.95); S5. Finally, the carbon emission prediction results of the corresponding regions and processes are calculated through the carbon emission factor calculation system module.

[0035] It should be noted that the accurate matching and continuous evolution of the prediction model are achieved through a double-layer adaptive optimization mechanism. The core of the system adopts a scenario-model dual-driven architecture: in the scenario analysis layer, a multi-dimensional feature extraction engine is used to deeply deconstruct the target environment, quantitatively analyze scenario elements such as fuel thermodynamic characteristics, geospatial constraints, temporal evolution laws, and process flow parameters, and construct a dynamic feature fingerprint map; in the model decision layer, a dynamic evaluation matrix of the prediction model performance is established to track in real time the prediction error rate, response speed, and generalization ability indicators of each sub-model under different feature dimensions. Based on the two-way matching mechanism, the system uses the Bayesian model averaging (BMA) technology to transform the scenario adaptation rules of the historical optimal model into a transferable weight allocation strategy. When a new scenario input is detected, the intelligent weight adjustment module dynamically integrates the advantageous components of four types of prediction models, namely fuel-specific type, regionally adaptable type, time-sensitive type, and process-optimized type, according to the matching degree between the real-time scenario fingerprint and the model feature library, and generates a scenario-customized composite prediction model. This mechanism breaks through the limitations of traditional static model fusion. While ensuring the stability of the basic prediction, through the dynamic recombination of scenario-driven model components, it significantly improves the analysis ability of complex working conditions. Especially when dealing with typical scenarios such as cross-regional energy allocation, seasonal production capacity fluctuations, and the application of emerging carbon reduction technologies, the system can independently select the optimal model combination strategy to achieve the paradigm upgrade of the carbon emission prediction result from "empirical fitting" to "mechanism-driven".

[0036] In this document, relational terms such as first and second are used solely to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variation thereof is intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device.

[0037] The foregoing has shown and described the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and the descriptions in the specification are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements fall within the scope of the present invention claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. An adaptive calculation system for carbon emission factors with multi-source fusion, characterized in that It includes a data input module, a sub-model module, a total model integration and generation module, a model weight adjustment module, and a carbon emission factor calculation system module; The sub-model module can generate calculation models of four sub-models according to fuel type, region, time series, and process type, namely a fuel sub-model, a region sub-model, a time series sub-model, and a process type sub-model; The total model integration and generation module integrates the four sub-models and generates a meta large model using the Stacking integrated learning prediction method: （6）； In the above formula (6), is the predicted carbon emission value after final fusion, is the total carbon emission based on fuel type, is the total carbon emission based on regional characteristics, is the indirect emission factor of the power grid, is the total carbon emission of a certain production process, is the weight of model i at time t; The model weight adjustment module adjusts the weights based on the fusion of scenario similarity and model performance, and its core formula is: （7）； In the above formula (7), is the recent performance of the currently calculated model index i, is the scene matching degree of the currently calculated model index i, is the recent performance of the index j of the previous model, is the scene matching degree of the index j of the previous model.

2. The multi-source fusion-based carbon emission factor adaptive calculation system according to claim 1, characterized in that, The calculation model of the fuel sub-model is: （1）； In the above formula (1), is the total carbon emissions based on the fuel type, unit: ton CO2, is the activity of the i-th fuel, unit: ton, is the mass fraction of carbon in the i-th fuel, unit: ton CO2 / ton fuel, is the carbon oxidation ratio (0-1) of the fuel combustion chamber in the i-th fuel, is the unit energy value of the fuel in the i-th fuel, unit: MJ / kg; The calculation model of the region sub-model is: （2）； In the above formula (2), is the total carbon emissions based on regional characteristics, unit: ton CO2, and is the regional energy structure weight, obtained through the proportion of regional energy consumption or regression analysis, and needs to satisfy + = 1, is the indirect emission factor of the power grid, unit: ton CO2 / MWh; The above-mentioned The calculation formula is as follows: （3）； In the above formula (3), is the total carbon emissions of the regional power grid, is the total power generation of the region; The calculation model of the time series sub-model is: （4）； In the above formula (4), is the carbon emission at time t, unit: ton CO2, is the linear long-term trend, is the trend slope, unit: ton CO2 / time unit, is the initial carbon emission base value, unit: ton CO2, is the seasonal adjustment term, where m is the season or month number, is the random error, with a base value of 0; The calculation formula of the process type sub-model is: （5）； In the above formula (5), is the total carbon emissions of a certain production process, unit: ton CO2, is the energy consumption of displacement process j, unit: MWh, is the emission factor of process j, unit: ton CO2 / unit energy consumption.

3. The multi-source fusion carbon emission factor adaptive calculation system according to claim 2, wherein Wherein, In the core formula (8); In the above formula (8), is the time decay factor, is the time window length, is the normalized error of model i at time t - k, is the conversion of error to precision; In the core formula (9); In the above formula (9), is the current scene feature vector, is the set of historical similar scenes, is the scene similarity metric, is the performance of model i in the historical scene s.

4. The multi-source fusion-based carbon emission factor adaptive calculation system according to claim 3, wherein The data input module is communicatively connected to the sub-model module, the sub-model module is communicatively connected to the total model integration and generation module, the total model integration and generation module is communicatively connected to the model weight adjustment module, and the carbon emission factor calculation system module is communicatively connected to both the total model integration and generation module and the model weight adjustment module; The data input module includes a manual information input unit and a network data collection unit; The manual information input unit includes a keyboard and an information entry camera, and the network data collection unit collects data in the network through a web crawler.

5. The multi-source fusion-based carbon emission factor adaptive calculation system according to claim 4, wherein The sub-model module includes a fuel sub-model unit, a region sub-model unit, a time series sub-model unit, and a process type sub-model unit; The fuel sub-model unit generates a corresponding fuel sub-model according to the information input by the data input module, the region sub-model unit generates a corresponding region sub-model according to the information input by the data input module, the time series sub-model unit generates a corresponding time series sub-model according to the information input by the data input module, and the process type sub-model unit generates a corresponding process type sub-model according to the information input by the data input module.

6. The multi-source fusion carbon emission factor adaptive calculation system according to claim 4, characterized in that The total model integration and generation module includes a model set unit, a total model generation unit, and a model output unit; The model set unit is used to integrate the fuel sub-model, the region sub-model, the time series sub-model, and the process type sub-model.

7. The multi-source fusion-based carbon emission factor adaptive calculation system according to claim 4, wherein The model weight adjustment module includes multiple sub-model weight calculation units and a historical performance calculation unit; The dynamic weights of each sub-model are adaptively adjusted through the sub-model weight calculation unit according to the specific scenario, and the historical performance calculation unit can calculate the historical performance of each prediction large model and compare multiple prediction large models to obtain the best prediction data.

8. The multi-source fusion-based carbon emission factor adaptive calculation system according to claim 4, characterized in that The carbon emission calculation system module includes a model confirmation unit, a calculation unit, and a solution benchmark verification unit; The model confirmation unit can verify each sub-model, the calculation unit is used to provide calculation services for this system, and the solution benchmark verification unit can further verify the relationship between the obtained large model and each data and perform calibration and verification.

9. The calculation method for adaptively calculating carbon emission factors with multi-source fusion is implemented by using the multi-source fusion carbon emission factor adaptive calculation system as described in any one of claims 1-8, and is characterized in that, Specifically, it includes the following steps: S1. Input the required information through the data input module, where the information includes: fuel type, regional information, time series information, and process type information; S2. Generate a fuel sub-model, a regional sub-model, a time series sub-model, and a process type sub-model respectively through the sub-model module. The four sub-models are established independently and correspond to the information collected by the data input module respectively; S3. Generate a prediction model through the overall model integration and generation module and the above four sub-models; S4. Adjust the weights of the prediction model through the model weight adjustment module according to the recent performance and scenario matching degree information; S5. Calculate the prediction result through the carbon emission factor calculation system module.

Images