A corporate carbon emissions analysis method and system based on an ESG comprehensive assessment model

Abstract

The present invention relates to the technical field of enterprise carbon emission analysis, and discloses a method and system for enterprise carbon emission analysis based on an ESG comprehensive assessment model, which includes a data acquisition layer, an analysis and processing layer, and a decision output layer. The data acquisition layer collects carbon emission-related data from the entire production and operation process of the enterprise in real time through multiple devices; the analysis and processing layer builds a trusted data processing environment to achieve data cleaning and verification, secure transmission and classified storage, and dynamic synchronization of data and models; the decision output layer uses pattern recognition technology and integrated learning algorithms to establish a multi-dimensionally correlated ESG comprehensive assessment model to achieve carbon emission intensity calculation, supply chain carbon footprint tracing, and emission reduction path optimization decision-making. This method and system can comprehensively and accurately analyze the carbon emissions of enterprises, provide scientific decision-making support for corporate low-carbon management, and help achieve carbon neutrality goals.

Description

Technical Field

[0001] The present invention relates to the technical field of corporate carbon emission analysis, and specifically to a corporate carbon emission analysis method and system based on an ESG comprehensive assessment model. Background Art

[0002] As the world actively promotes carbon neutrality and carbon peak, businesses, as the main contributors to carbon emissions, face increasingly stringent environmental regulations and social responsibilities for low-carbon development. Traditional corporate carbon emissions analysis methods have many limitations and are unable to meet the current complex carbon emissions management needs.

[0003] From a data collection perspective, traditional methods often rely on manual data entry or single-device monitoring, resulting in limited data coverage and inability to achieve real-time, comprehensive monitoring of a company's entire production and operation process. For example, only partial data on a company's direct emissions is available, while effective means of collecting information on indirect carbon emissions from various links in the supply chain are lacking, resulting in incomplete carbon emissions analysis. Furthermore, manual data collection is prone to errors and data updates are delayed, making it difficult to meet the needs of real-time analysis and decision-making.

[0004] In terms of data processing and storage, traditional systems mostly use centralized storage and simple data processing methods. Centralized storage presents a single point of failure risk, resulting in low data security and reliability. Furthermore, traditional processing methods lack effective cleaning, verification, and integration capabilities for heterogeneous carbon emission data from multiple sources, making it impossible to build a trusted data processing environment. This leads to inconsistent data quality and affects the accuracy of subsequent analysis. Furthermore, data transmission security cannot be effectively guaranteed, making data leaks and other issues more likely to occur.

[0005] In terms of assessment model construction, existing carbon emission assessment models are mostly based on a single dimension or fixed indicator system, lacking the ability to dynamically adjust. A company's production and operation activities are dynamic, and factors such as energy consumption and production processes at different stages will affect carbon emissions. Traditional models are unable to calibrate assessment indicator weights in real time based on a company's actual operational data, making it difficult to accurately reflect a company's true carbon emissions. Furthermore, the lack of effective integration of industry benchmark data and best practices results in a lack of comparability and reference value in the model's assessment results.

[0006] In terms of decision support, traditional methods rely primarily on empirical judgment or simple statistical analysis, failing to provide scientific and accurate decisions on optimizing emission reduction paths. Supply chain carbon footprint tracing is difficult to analyze the contribution of carbon emissions throughout the entire supply chain, and key carbon emission nodes cannot be accurately identified. When formulating emission reduction strategies, there is a lack of comprehensive consideration of information on enterprise equipment energy efficiency, process improvement plans, clean technology applications, and carbon trading markets, making it difficult to find the optimal balance between carbon emission constraints and economic cost limits.

[0007] With the widespread adoption of ESG (environmental, social, and governance) concepts, companies must not only monitor their own carbon emissions but also consider the environmental responsibilities and governance of their supply chains. Traditional carbon emissions analysis methods cannot meet the requirements of comprehensive ESG assessments. There is an urgent need for new methods and systems that can integrate multi-dimensional data, build dynamic assessment models, and enable full-chain carbon emissions analysis and scientific decision-making. Summary of the Invention

[0008] The purpose of the present invention is to provide a method and system for analyzing corporate carbon emissions based on an ESG comprehensive evaluation model to solve the problems raised in the above background technology.

[0009] To achieve the above objectives, the present invention provides the following technical solution: a corporate carbon emissions analysis system based on an ESG comprehensive assessment model, the system comprising:

[0010] Data collection layer, analysis and processing layer, and decision output layer;

[0011] The data collection layer includes environmental monitoring devices, energy metering instruments, supply chain tracking modules, and compliance recording units, which are used to conduct real-time monitoring and collect raw data on carbon emission-related factors throughout the entire production and operation process of the enterprise;

[0012] The analysis and processing layer includes edge computing nodes, distributed storage clusters, and blockchain verification modules. It deploys dynamic optimization algorithms and hybrid encryption protocols to build a trusted data processing environment, perform data cleansing and verification operations, and uses hybrid encryption protocols for secure transmission and classified storage of different business modules and data types. It also achieves dynamic synchronization of enterprise operational data and ESG comprehensive assessment models through standardized interfaces.

[0013] The decision output layer is used to use pattern recognition technology to integrate the company's historical operating data and real-time monitoring data, dynamically calibrate the evaluation indicator system, and establish a multi-dimensional ESG comprehensive evaluation model. Based on this model, an integrated learning algorithm is used to calculate carbon emission intensity, trace the carbon footprint of the supply chain, and make emission reduction path optimization decisions.

[0014] Preferably, the environmental monitoring device includes at least exhaust gas emission sensors, wastewater detection probes, solid waste metering equipment and noise monitors, which are used to obtain environmental parameters of the enterprise's production and operation site; the energy metering instruments include at least electricity meters, gas flow meters, steam meters and fuel consumption recorders, which are used to count the enterprise's energy consumption data; the supply chain tracking module includes at least logistics information collection terminals, raw material traceability databases and transport vehicle positioning devices, which are used to record carbon emission information in each link of the supply chain.

[0015] Preferably, in the analysis and processing layer, the data acquired by each collection terminal is transmitted from the edge computing node to the distributed storage cluster through a dedicated channel. The distributed storage cluster then converts the data format and stores it together with the regulatory information acquired by the compliance recording unit in the blockchain verification module for feature extraction. The data cleaning and verification operation uses outlier detection technology to input the collected raw data into the verification model established by different verification rules for cross-verification.

[0016] The dynamic synchronization of enterprise operating data and ESG comprehensive evaluation models through standardized interfaces includes: establishing standardized data interfaces to achieve interactive communication between business systems and ESG comprehensive evaluation models, synchronizing actual enterprise operating status data, and associating and mapping structured data, while updating model parameter configurations to complete data version management, model iterative updates, and decision instruction verification.

[0017] Preferably, the establishment of a multi-dimensional ESG comprehensive assessment model includes:

[0018] Establish a two-way correlation channel and dynamic matching mechanism between the actual operating data of enterprises and the parameters of the ESG comprehensive assessment model;

[0019] By mapping and correlating corporate production and operation activities through data, and using on-site monitoring data, supply chain records, and compliance information as the foundation, we establish a corporate carbon emissions inventory and an industry benchmark comparison framework. We then initialize model parameters based on business characteristics, simulate multiple operating scenarios, and form a dynamically adjustable comprehensive ESG assessment model for companies.

[0020] Optimize the parameters of the enterprise ESG comprehensive assessment model, import the enterprise's multi-dimensional operational data into the established ESG comprehensive assessment model, and use the Monte Carlo simulation method to conduct sensitivity analysis and correction on the model output results to obtain the optimized enterprise ESG comprehensive assessment model;

[0021] The enterprise ESG comprehensive assessment model includes operating entity data, an assessment rule engine, a dynamic knowledge base, and the interactive relationships between modules;

[0022] The operating entity data is the basic data source of the ESG comprehensive assessment model, including the company's direct emission data and indirect emission data; the assessment rule engine establishes a mapping relationship with the actual business data, and quantitatively assesses the carbon emission characteristics through a multi-dimensional indicator system and a weight distribution mechanism; the dynamic knowledge base integrates industry standard data, policy and regulatory texts and best practice cases, and forms an assessment benchmark database through continuous updating and maintenance; the interactive relationship between the modules realizes information exchange through data pipelines, the operating data and the assessment rule engine are standardized through a data conversion interface, and the assessment rule engine and the dynamic knowledge base realize rule matching through semantic parsing technology.

[0023] Preferably, the carbon emission intensity calculation based on the model using an ensemble learning algorithm includes:

[0024] Collect the company's historical annual energy consumption data, production process record data, equipment operation logs, and product carbon footprint reports to build a carbon emission characteristic dataset;

[0025] After filling missing values and correcting outliers in the feature data set, divide it into training set and validation set;

[0026] Configure the initial parameters of the XGBoost-GRU joint model, input the training set into the joint model, use the XGBoost component to sort the feature importance, use the GRU network to perform time series modeling, and use the feature weighting mechanism to eliminate the impact of emission fluctuations in different production cycles until the model error reaches the preset threshold;

[0027] The validation set is input into the trained joint model to calculate the carbon emission intensity prediction value, and the carbon emission intensity level assessment result of the enterprise is generated by combining it with the industry benchmark value;

[0028] Based on the assessment results, the preset improvement strategy library is matched to generate targeted carbon emission management recommendations.

[0029] Preferably, the supply chain carbon footprint tracing using an ensemble learning algorithm based on the model includes:

[0030] Obtain logistics and transportation data, raw material procurement records, production and processing information, and product distribution paths for each link in the supply chain to build a supply chain carbon flow dataset;

[0031] The key features of the supply chain carbon flow are extracted through principal component analysis, and the carbon footprint tracking feature vector is established;

[0032] Construct a random forest regression model and use adaptive particle swarm optimization to optimize the model parameter combination;

[0033] The feature vector is input into the optimized random forest regression model to output the carbon emission contribution ranking of each link in the supply chain;

[0034] According to the contribution ranking results, the key nodes of carbon footprint are identified and the supply chain carbon flow optimization path diagram is generated.

[0035] Preferably, the method of using an ensemble learning algorithm based on the model to make an optimization decision on the emission reduction path includes:

[0036] Integrate enterprise equipment energy efficiency data, process improvement solution library, clean technology application cases and carbon trading market information to build a knowledge graph of emission reduction strategies;

[0037] Perform graph convolutional network feature extraction on the node relationships in the knowledge graph to generate the emission reduction plan feature matrix;

[0038] A multi-objective optimization model is established and the NSGA-II algorithm is used to find the Pareto optimal solution set;

[0039] The characteristic matrix is input into the optimization model, and a set of feasible solutions that meet the carbon emission constraints and economic cost limits is output;

[0040] The final implementation plan is selected based on the decision preferences and a phased emission reduction route planning table is generated.

[0041] Preferably, the specific implementation of the data cleaning and verification operation in the analysis and processing layer includes:

[0042] Establish a data quality detection rule library, including value range verification rules, data integrity verification rules and logical consistency verification rules;

[0043] Use a streaming computing engine to verify real-time data one by one, and trigger an early warning mechanism for data that does not meet the verification rules;

[0044] Perform a full scan of historical batch data and use data repair algorithms to automatically correct abnormal records that can be repaired;

[0045] The data verification process and exception handling logs are recorded through blockchain smart contracts to generate tamper-proof data quality audit reports.

[0046] Preferably, the method for constructing the evaluation rule engine includes:

[0047] Collect domestic and international carbon emission accounting standards, industry technical specifications, and policy and regulatory requirements to build a knowledge base of standard terms;

[0048] Use natural language processing technology to parse standard documents, extract key clauses and form structured evaluation rules;

[0049] Establish a rule matching engine to automatically associate actual enterprise data with standard terms through semantic similarity calculation;

[0050] Set up a dynamic adjustment mechanism for rule weights to automatically optimize rule priorities based on policy update frequency and changes in industry practices.

[0051] Preferably, the present invention further includes a method for analyzing corporate carbon emissions based on an ESG comprehensive assessment model, the method comprising the following steps:

[0052] Environmental monitoring devices deployed at production and operation sites collect real-time data on exhaust gas emission concentrations, wastewater pollutant indicators, solid waste generation, and noise levels. Energy meters are also used to obtain electricity, gas, steam, and fuel oil consumption.

[0053] The collected raw data is filtered for outliers and format standardized by edge computing nodes before being transmitted to a distributed storage cluster for multi-source data fusion. The blockchain verification module is used to encrypt and verify the data integrity and time consistency.

[0054] Using pattern recognition technology, we conduct feature correlation analysis on historical operational data and real-time monitoring data, dynamically adjust the weight parameters of ESG assessment indicators, and build a multi-dimensional ESG comprehensive assessment model that includes environmental compliance, energy efficiency, and supply chain carbon footprint;

[0055] A transfer learning framework is used to transfer knowledge across industries for the ESG comprehensive assessment model, and Monte Carlo simulation is used to generate carbon emission intensity forecast data under different emission reduction scenarios;

[0056] Combined with the logistics routes, raw material sources, and production and processing information recorded by the supply chain tracking module, a full-chain carbon flow mapping relationship is established, and graph neural networks are used to identify key carbon emission nodes;

[0057] Based on the relationship between carbon emission intensity prediction results and carbon flow mapping, a multi-objective optimization algorithm is used to solve a set of emission reduction paths that meet economic constraints, and a decision-making report containing equipment modification plans, process optimization strategies and carbon trading recommendations is output.

[0058] Compared with the prior art, the present invention has the following beneficial effects:

[0059] In terms of data collection and processing, the system uses multi-source equipment such as environmental monitoring devices, energy metering instruments, supply chain tracking modules, and compliance recording units to achieve real-time monitoring and raw data collection of carbon emission-related factors throughout the entire production and operation process of enterprises. Environmental monitoring devices include exhaust gas emission sensors, wastewater detection probes, and other equipment that can comprehensively obtain environmental parameters; energy metering instruments can accurately count various types of energy consumption data; and the supply chain tracking module can record carbon emission information at all links in the supply chain to ensure the comprehensiveness and real-time nature of data collection. The analysis and processing layer builds a trusted data processing environment through edge computing nodes, distributed storage clusters, and blockchain verification modules. It uses outlier detection technology for data cleaning and verification, and uses hybrid encryption protocols to ensure data transmission and storage security. It also uses standardized interfaces to achieve dynamic synchronization of enterprise operating data and ESG comprehensive assessment models, effectively improving data quality and processing efficiency, and providing a reliable data foundation for subsequent analysis.

[0060] In terms of constructing an ESG comprehensive assessment model, the system has established a multi-dimensionally correlated ESG comprehensive assessment model. By building a two-way correlation channel and dynamic matching mechanism between a company's actual operational data and the ESG comprehensive assessment model parameters, based on the company's on-site monitoring data, supply chain record data, and compliance information, a carbon emissions inventory and industry benchmark comparison framework is established. Model parameter initialization and multi-scenario simulation are then performed to form a dynamically adjustable ESG comprehensive assessment model. At the same time, Monte Carlo simulation methods are used to conduct sensitivity analysis and corrections on the model output results to ensure the model's accuracy and adaptability. This model integrates operational entity data, an assessment rule engine, a dynamic knowledge base, and the interactions between various modules. It can comprehensively and dynamically assess a company's carbon emissions status and meet the requirements of ESG comprehensive assessments.

[0061] In terms of data analysis and decision support, the system uses a variety of integrated learning algorithms based on the ESG comprehensive assessment model to achieve carbon emission intensity calculation, supply chain carbon footprint tracing, and emission reduction path optimization decision-making. The XGBoost-GRU joint model is used to calculate carbon emission intensity, effectively eliminating the impact of emission fluctuations in different production cycles, generating accurate carbon emission intensity level assessment results and targeted management recommendations. The random forest regression model and adaptive particle swarm algorithm are used to trace the carbon footprint of the supply chain, identifying key carbon footprint nodes and generating an optimization path map. By constructing an emission reduction strategy knowledge graph and a multi-objective optimization model, the NSGA-II algorithm is used to solve the Pareto optimal solution set, which can output a feasible solution set that meets carbon emission constraints and economic cost limits, and generate a phased emission reduction route planning table, providing scientific and accurate decision support for enterprises.

[0062] In terms of data security and management, the analysis and processing layer uses blockchain smart contracts to record data verification processes and exception handling logs, generating tamper-proof data quality audit reports to ensure data integrity and credibility. The evaluation rules engine uses natural language processing technology to parse standard documents, establish a rule matching engine, and establish a dynamic weight adjustment mechanism. This allows for timely response to policy updates and changes in industry practices, ensuring the scientific nature and timeliness of evaluation rules. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 This is a working principle diagram of the enterprise carbon emission analysis system based on the ESG comprehensive assessment model described in the present invention;

[0064] Figure 2 To analyze the working principle diagram of data transmission and verification in the processing layer;

[0065] Figure 3 Schematic diagram for building a multi-dimensional ESG comprehensive assessment model;

[0066] Figure 4 This is a diagram showing the working principle of supply chain carbon footprint tracing and contribution analysis. DETAILED DESCRIPTION

[0067] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only 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 making creative efforts are within the scope of protection of the present invention.

[0068] See also Figures 1-4 The present invention relates to a corporate carbon emissions analysis system based on an ESG comprehensive assessment model. The system comprises a data acquisition layer, an analysis and processing layer, and a decision output layer. Each layer works together to achieve comprehensive analysis and management of corporate carbon emissions. The specific implementation steps are as follows:

[0069] The data collection layer uses environmental monitoring devices, energy meters, supply chain tracking modules, and compliance recorders to conduct real-time monitoring and collect raw data on carbon emissions throughout a company's production and operations. Environmental monitoring devices capture environmental parameters at the company's production and operations site, energy meters compile energy consumption data, the supply chain tracking module records carbon emissions information at every stage of the supply chain, and the compliance recorder collects the company's compliance information.

[0070] The analysis and processing layer utilizes edge computing nodes, a distributed storage cluster, and a blockchain verification module to build a trusted data processing environment. Edge computing nodes perform preliminary processing on data captured by acquisition terminals and transmit it to the distributed storage cluster via dedicated channels. The distributed storage cluster converts the data format and stores it, along with regulatory information captured by the compliance recorder, in the blockchain verification module for feature extraction. This layer also deploys dynamic optimization algorithms and hybrid encryption protocols to perform data cleansing and verification operations. Hybrid encryption protocols are used for secure transmission and classified storage across different business modules and data types. Standardized interfaces are used to dynamically synchronize enterprise operational data with comprehensive ESG assessment models.

[0071] See attached Figure 2 The data processing flow for each acquisition terminal in the analysis and processing layer first involves transferring data from edge computing nodes via dedicated channels to the distributed storage cluster. As the front end of data processing, edge computing nodes are responsible for performing preliminary processing on the raw data received by the acquisition terminals. This initial processing includes data format conversion, data cleaning and verification using outlier detection technology, and inputting the collected raw data into verification models established with different validation rules for cross-verification.

[0072] Regarding data transmission, the establishment of a dedicated channel ensures security and stability. This channel utilizes multi-layer encryption technology, providing real-time encryption of transmitted data to prevent theft or tampering during transmission. Furthermore, the channel features data transmission monitoring, enabling real-time monitoring of data transmission status. If any anomalies are detected, prompt action can be taken to address them, ensuring secure data transmission.

[0073] After receiving data from edge computing nodes, the distributed storage cluster further converts the data format. This process ensures that the data can be identified and processed by subsequent processing modules. The distributed storage cluster utilizes a distributed file system, which allows data to be stored across multiple nodes, improving storage reliability and availability. Furthermore, the distributed storage cluster features data redundancy, enabling multiple backups of important data to prevent data loss.

[0074] Data and regulatory information obtained by the compliance record unit are stored together in the blockchain verification module for feature extraction. The blockchain verification module utilizes blockchain technology to ensure data immutability and traceability. Before data is stored in the blockchain verification module, it is hashed to generate a unique hash value. This hash value is then stored in the blockchain to ensure data integrity and authenticity. During feature extraction, the blockchain verification module performs multi-dimensional analysis and processing on the data to extract key features and information.

[0075] Data cleaning and validation utilizes outlier detection technology, which cross-checks the collected raw data against validation models based on different validation rules. This process ensures data accuracy and reliability. Outlier detection utilizes a variety of algorithms, including statistical and machine learning-based methods, to effectively detect outliers in the data. Validation models are tailored to the data type and business needs. These models include rules for value range, data integrity, and logical consistency, enabling comprehensive data verification and validation.

[0076] Dynamic synchronization of enterprise operational data with the ESG comprehensive assessment model is achieved through standardized interfaces. Specifically, a standardized data interface is established to enable interactive communication between business systems and the ESG comprehensive assessment model, synchronize the enterprise's actual operating status data, perform correlation mapping of structured data, and simultaneously update model parameter configurations, completing data version management, model iteration updates, and decision instruction verification. The establishment of a standardized interface is key to achieving dynamic synchronization of enterprise operational data with the ESG comprehensive assessment model. This interface utilizes unified standards and specifications to ensure data interaction and sharing between different systems. Regarding data synchronization, the enterprise's actual operating status data is captured in real time and transmitted to the ESG comprehensive assessment model for analysis and processing. Furthermore, correlation mapping of structured data is performed to ensure data consistency and accuracy.

[0077] Regarding model parameter configuration updates, the ESG comprehensive assessment model's parameters are updated in real time based on the company's actual operating conditions and market changes. This process ensures the accuracy and reliability of the ESG comprehensive assessment model and improves the company's decision-making capabilities. Data version management is a key component in achieving dynamic synchronization between a company's operational data and the ESG comprehensive assessment model. Through data version management, historical data versions can be recorded and managed, ensuring data traceability and consistency. Regarding model iteration and updates, the ESG comprehensive assessment model is continuously iterated and updated based on the company's actual needs and market changes. This process ensures the advancement and practicality of the ESG comprehensive assessment model and enhances the company's competitiveness. Decision instruction verification is the final line of defense in achieving dynamic synchronization between a company's operational data and the ESG comprehensive assessment model. Decision instruction verification verifies and audits the decision instructions output by the ESG comprehensive assessment model to ensure their rationality and feasibility.

[0078] The specific implementation method of data cleaning and verification operations is to establish a data quality detection rule library, which includes numerical range verification rules, data integrity verification rules and logical consistency verification rules; use a streaming computing engine to verify real-time data one by one, and trigger an early warning mechanism for data that does not meet the verification rules; perform a full scan of historical batch data, and use a data repair algorithm to automatically correct repairable abnormal records; record the data verification process and exception handling logs through blockchain smart contracts to generate an unalterable data quality audit report.

[0079] The establishment of a data quality check rule library is the foundation for data cleaning and validation operations. This rule library contains various data quality check rules, enabling comprehensive data testing and verification. Numerical range check rules test and verify data ranges, ensuring that data values are within a reasonable range. Data integrity check rules test and verify data integrity, ensuring that data is not missing or duplicated. Logical consistency check rules test and verify data logical consistency, ensuring that logical relationships between data are correct.

[0080] The use of a streaming computing engine is key to enabling line-by-line verification of real-time data. This engine processes and analyzes real-time data in real time, triggering an early warning mechanism for data that doesn't meet verification rules. This early warning mechanism promptly notifies relevant personnel to address abnormal data, ensuring data accuracy and reliability.

[0081] A full scan of historical batch data is performed, and a data repair algorithm is used to automatically correct any abnormal records. This process comprehensively cleans and repairs historical data, improving its quality and usability. The data repair algorithm automatically selects the appropriate repair method based on the data's characteristics and anomalies, ensuring data accuracy and reliability.

[0082] Blockchain smart contracts record data verification and exception handling logs, generating immutable data quality audit reports. Blockchain smart contracts ensure the immutability and traceability of these logs. Data quality audit reports provide a comprehensive assessment and analysis of data quality, providing strong support for enterprise data management and decision-making.

[0083] The decision-making output layer uses pattern recognition technology to integrate the company's historical operating data and real-time monitoring data, dynamically calibrates the evaluation indicator system, and establishes a multi-dimensional ESG comprehensive evaluation model. Based on this model, it uses an integrated learning algorithm to calculate carbon emission intensity, trace the supply chain carbon footprint, and optimize emission reduction path decisions, providing companies with scientific carbon emission management decisions.

[0084] See attached Figure 3, the establishment of a multi-dimensionally correlated ESG comprehensive assessment model requires the construction of a two-way correlation channel, the establishment of a multi-dimensional model, parameter optimization, and the establishment of a model structure. First, it is necessary to build a two-way correlation channel and a dynamic matching mechanism between the actual operating data of the enterprise and the parameters of the ESG comprehensive assessment model. The establishment of a two-way correlation channel is based on the enterprise's internal data transmission network. Through standardized data interfaces, it enables the real-time transmission of operating data from business systems such as production, energy, and supply chain to the ESG comprehensive assessment model. At the same time, it allows model parameters to act reversely on the business system, forming an interactive closed loop between data and model. The dynamic matching mechanism automatically matches key indicators in the operating data (such as energy consumption, exhaust emissions, supply chain node information, etc.) with model parameters (such as emission factors, weight coefficients, industry benchmark values, etc.) through preset data mapping rules, ensuring that the correspondence between data and parameters is dynamically adjusted as business scenarios change.

[0085] Through data mapping and feature association, a company's production and operation activities are mapped to a carbon emissions inventory and industry benchmark comparison framework based on on-site monitoring data, supply chain records, and compliance information. Model parameters are initialized based on business characteristics, and multiple operating scenarios are simulated to form a dynamically adjustable comprehensive ESG assessment model. During the data mapping process, on-site monitoring data (such as concentration data collected by exhaust emission sensors and energy consumption data recorded by energy meters) is preprocessed by edge computing nodes and mapped to the model's environmental dimension parameters according to a unified data format. Supply chain record data (such as logistics transportation routes and raw material procurement sources) is verified by a blockchain verification module and mapped to the model's supply chain dimension parameters. Compliance information (such as environmental policy compliance records and carbon emission quota implementation status) is accessed through the compliance record unit and mapped to the model's governance dimension parameters. Feature association uses correlation analysis algorithms in machine learning to identify potential connections between data from different dimensions, such as the linear relationship between energy consumption and exhaust emissions, or the positive correlation between supply chain transportation distance and carbon footprint, thereby constructing a multi-dimensional data feature network.

[0086] The establishment of a corporate carbon emissions inventory is based on the principle of full lifecycle carbon emissions accounting, covering both direct emissions (such as those from fossil fuel combustion during production) and indirect emissions (such as those from upstream raw material production and downstream product transportation in the supply chain). Data aggregation generates an inventory table containing the total amount and intensity of carbon emissions from each link. The industry benchmark comparison framework, based on carbon emission benchmark data released by industry associations and public data from benchmarked companies in the same industry, establishes a horizontal comparison coordinate system. Each indicator in the corporate carbon emissions inventory is compared with the industry benchmark to identify the company's carbon emission level within the industry. Model parameter initialization assigns initial values to each dimension of the model based on business characteristics such as the company's industry, production scale, and process level. For example, energy consumption emission factors for high-energy-consuming industries are initially set higher, while those for clean energy industries are set lower. Multi-scenario simulations adjust model parameters (e.g., assuming energy mix optimization, supply chain routing changes, etc.), run the model, and output carbon emission forecasts under different scenarios. This scenario analysis report provides a basis for dynamic model adjustments.

[0087] When optimizing the parameters of a comprehensive ESG assessment model, the company's multi-dimensional operational data is imported into the established ESG comprehensive assessment model. Monte Carlo simulation is then used to perform sensitivity analysis and corrections on the model output, resulting in an optimized model. This multi-dimensional operational data, including three years of historical energy consumption records, carbon flow records for each link in the supply chain, and environmental compliance inspection reports, is imported into the model in batches via the distributed storage cluster interface. Monte Carlo simulation generates a large number of random number sequences to simulate random variations in data parameters, calculates the probability distribution of the model output, and identifies key parameters that significantly influence the model results (e.g., the energy consumption coefficient of a certain type of equipment or the carbon intensity of a specific transportation route). Sensitivity analysis calculates the sensitivity index of each parameter to determine the priority and magnitude of parameter adjustments. For example, parameters with high sensitivity indices are calibrated, and more precise parameter values are obtained through regression analysis of historical data. These values are then re-entered into the model for verification until the stability and accuracy of the model output meet the expected standards.

[0088] The comprehensive enterprise ESG assessment model encompasses operational entity data, an assessment rules engine, a dynamic knowledge base, and the interactions between modules. Operational entity data, serving as the foundational data source for the ESG comprehensive assessment model, includes both direct and indirect enterprise emissions data. Direct emissions data is collected in real time through environmental monitoring devices and energy metering instruments, cleansed and verified by the analysis and processing layer, and stored in the direct emissions database of the distributed storage cluster. Indirect emissions data is collected through the supply chain tracking module, combined with emission factors from the industry database, and stored in the indirect emissions database. The two types of data are linked through timestamps and business process numbers to form a complete enterprise carbon emissions data chain.

[0089] The evaluation rule engine maps business data to actual data. Its construction method involves collecting international and domestic carbon emissions accounting standards, industry technical specifications, and policy and regulatory requirements to build a knowledge base of standard terms. Natural language processing (NLP) technology is used to parse standard documents, extracting key terms and formulating structured evaluation rules. A rule matching engine automatically links actual enterprise data with standard terms through semantic similarity calculations. A dynamic rule weighting mechanism is implemented to automatically prioritize rules based on policy updates and industry practices. The standard terms knowledge base encompasses international standards (such as ISO 14064), national standards (such as GB / T 32151), local standards, and industry association specifications. Web crawlers are used to capture the latest policy documents in real time, which are then manually reviewed and stored in the knowledge base. Natural language processing technology is used to perform word segmentation, part-of-speech tagging, and named entity recognition on the standard documents. Core terms (such as emissions accounting boundaries, data quality requirements, and reporting format specifications) are extracted and converted into structured evaluation rules, which are stored in a rule database. The rule matching engine uses a cosine similarity algorithm to calculate the semantic similarity between a company's actual data (such as carbon emissions accounting scope and data collection frequency) and the assessment rules. When the similarity exceeds a preset threshold, the rule association is automatically triggered, generating a compliance assessment result. The dynamic rule weight adjustment mechanism is based on policy update logs and a library of industry practice cases. When the frequency of policy and regulatory updates increases or new emission reduction technologies are widely adopted within the industry, the weight coefficients of the assessment rules are adjusted accordingly. For example, the weight of rules related to clean energy utilization is increased, while the weight of rules related to outdated processes is reduced.

[0090] The dynamic knowledge base integrates industry standard data, policy and regulatory texts, and best practice cases, and through continuous updating and maintenance, forms an assessment benchmark database. Industry standard data includes indicators such as the industry's average carbon emission intensity, energy efficiency levels, and supply chain carbon efficiency, obtained through industry statistical annual reports and third-party research reports. Policy and regulatory texts are synchronized in real time with national and local carbon emission-related laws, regulations, and policy documents. Best practice cases collect successful emission reduction experiences from domestic and foreign companies (such as specific measures taken by a company to reduce carbon emissions through equipment modifications, or the carbon footprint reduction effects of a supply chain optimization case), and are stored in the case database after expert review. The dynamic knowledge base's update mechanism includes scheduled automatic capture (such as updating policy documents at dawn each day), manual review and entry (such as monthly summary of industry data), and user feedback supplements (such as companies submitting their own emission reduction cases) to ensure the timeliness and practicality of the assessment benchmark data.

[0091] Interactions between modules are facilitated through a data pipeline. Operational data is standardized between the evaluation rule engine and the dynamic knowledge base through a data conversion interface. Semantic parsing technology is used to match rules between the evaluation rule engine and the dynamic knowledge base. The data pipeline utilizes message queue technology to establish a transmission link for operational data from the acquisition layer to the analysis and processing layer and then to the decision output layer, ensuring the orderly flow of data between modules. The data conversion interface between operational data and the evaluation rule engine converts raw operational data (such as analog signals from sensors and structured data from databases) into a format recognizable by the rule engine (such as JSON-formatted rule input parameters). It also converts the evaluation results output by the rule engine into the data format required for visualization reports. Semantic parsing technology between the evaluation rule engine and the dynamic knowledge base uses natural language processing algorithms such as text classification and entity linking to semantically match evaluation rules with standard clauses and cases in the knowledge base. For example, when the rule engine needs to query emission reduction standards for a specific type of equipment, semantic parsing technology automatically retrieves relevant industry standards and best practices from the dynamic knowledge base to provide reference for rule evaluation.

[0092] The steps for calculating carbon emission intensity using an ensemble learning algorithm based on the ESG comprehensive assessment model are as follows:

[0093] First, we collected historical annual energy consumption data, production process records, equipment operation logs, and product carbon footprint reports to construct a carbon emission signature dataset. Energy consumption data covers detailed information on the quantity, time, and department of each energy source (such as coal, electricity, and natural gas) consumed by the company during the historical year. This data is collected through the company's internal energy metering and energy management systems. Production process records include raw material inputs, product outputs, production process parameters, and the operating status of production equipment. This data is obtained from the company's production management system and on-site monitoring equipment. Equipment operation logs record information such as the start and stop times, operating load, and maintenance records of various types of production equipment. These data can be obtained from the equipment control system and maintenance management system. Product carbon footprint reports detail the carbon emissions of products throughout their entire life cycle, from raw material procurement, production and processing, transportation and sales, to final consumption. These reports are generated using a supply chain tracking module and life cycle assessment tools. These multi-source, heterogeneous data are integrated and correlated according to time series and business processes to construct a carbon emission signature dataset encompassing multiple characteristic dimensions.

[0094] After filling missing values and correcting outliers in the feature dataset, the dataset is divided into training and validation sets. Missing value filling utilizes a multiple interpolation approach, estimating and filling missing values using other feature variables based on the data's distribution characteristics and correlations. Outlier correction is based on statistical analysis methods, calculating the mean and standard deviation of the data. Data points that deviate from the mean by more than a certain multiple of the standard deviation are considered outliers and corrected using boundary value replacement or model-based predicted value replacement. When dividing the training and validation sets, the processed feature dataset is randomly divided in an 8:2 ratio. The training set is used for model training and parameter optimization, while the validation set is used to evaluate the performance and generalization capabilities of the ESG comprehensive assessment model.

[0095] The initial parameters of the XGBoost-GRU joint model are configured, and the training set is fed into the joint model. Feature importance is ranked using the XGBoost component, and time series modeling is performed using the GRU network. The feature weighting mechanism eliminates the impact of emission fluctuations across production cycles until the model error reaches a preset threshold. The initial parameter configuration for the XGBoost component includes the maximum tree depth, learning rate, and subsample ratio. The optimal parameter combination is determined through grid search and cross-validation. After the training set is fed into the XGBoost component, the model automatically calculates the importance score of each feature for carbon emission intensity, ranks the features based on the score, and selects the key features that have a significant impact on carbon emission intensity.

[0096] The GRU network is used to process time series data and capture the trends and patterns of carbon emissions data over time. The initial parameter configuration of the GRU network includes the number of hidden layer neurons, time step, dropout rate, etc. The selected key features are input into the GRU network in a time series format, and the network automatically learns the temporal patterns and dependencies in the data. The feature weighting mechanism assigns different weights to the features of each time step based on the characteristics of different production cycles. During the peak production season, energy consumption and carbon emissions are generally higher, and the relevant features are given higher weights at this time; during the off-season, lower weights are assigned. In this way, the impact of emission fluctuations in different production cycles on model predictions can be effectively eliminated. During the model training process, the backpropagation algorithm is used to continuously adjust the model parameters to minimize the error between the predicted value and the actual value until the error reaches the preset threshold.

[0097] The validation set is fed into the trained joint model to calculate the predicted carbon emission intensity. This is then combined with industry benchmarks to generate the company's carbon emission intensity rating. The validation set data is preprocessed using the same preprocessing methods as the training set and then fed into the trained XGBoost-GRU joint model. The model outputs a predicted carbon emission intensity value for each sample. Industry benchmarks are derived from industry carbon emission intensity standards published by industry associations and carbon emission data from benchmarked companies in the same industry. The company's predicted carbon emission intensity is compared with the industry benchmark to generate the company's carbon emission intensity rating. The rating is categorized into multiple levels, such as excellent, good, moderate, poor, and poor, each corresponding to a different carbon emission intensity range.

[0098] Based on the assessment results, the system matches the pre-set improvement strategy library to generate targeted carbon emission management recommendations. The improvement strategy library contains improvement strategies tailored to different carbon emission intensity levels and business scenarios, including energy structure optimization plans, production process improvement measures, equipment upgrade recommendations, and supply chain management optimization strategies. Based on the company's carbon emission intensity level assessment results, the system matches the corresponding improvement strategy from the improvement strategy library. This information is then combined with the company's actual business situation and resource constraints to generate specific, actionable carbon emission management recommendations. These recommendations include short-term, medium-term, and long-term emission reduction targets and implementation paths, providing a scientific basis for the company's carbon emission management decisions.

[0099] In the process of calculating carbon emission intensity, the prediction results of the XGBoost-GRU joint model can be expressed as: in, Indicates time The predicted value of carbon emission intensity is Indicates time The input feature vector contains key features such as energy consumption data and production process data. represents the set of parameters of the model, Represents the mapping function of the XGBoost-GRU joint model, which converts the input feature vector into a carbon emission intensity prediction value through XGBoost feature processing and GRU time series modeling.

[0100] See attached Figure 4The process of tracing the carbon footprint of the supply chain using an integrated learning algorithm based on the ESG comprehensive assessment model is as follows: First, it is necessary to obtain the logistics and transportation data, raw material procurement records, production and processing information, and product distribution paths of each link in the supply chain to construct a supply chain carbon flow data set. Logistics and transportation data include the type of transportation tool (such as trucks, ships, airplanes), transportation route mileage, total fuel consumption, carbon emission factors, etc., which are obtained in real time through logistics information collection terminals and transportation vehicle positioning devices. For example, the GPS positioning system installed on the truck can record the driving route and mileage, and the on-board fuel meter can collect fuel consumption data, and combine the carbon emission factor of diesel to calculate the carbon emissions of the transportation link. Raw material procurement records cover the name of the raw material, purchase quantity, supplier information, raw material production process, etc., and are obtained through the raw material traceability database. The supplier information includes the supplier's region and production scale. The raw material production process information is used to evaluate the carbon emission intensity of the upstream production link. Production and processing information includes equipment model, processing time, energy consumption, and auxiliary material usage. This information is collected from equipment operation logs and energy meters in the production workshop. For example, the total energy consumption for a particular production line can be calculated by multiplying the equipment operation time by the energy consumption per unit time. This information is then combined with the carbon emission factor corresponding to the energy type to calculate the carbon emissions of the production and processing stage. Product distribution path data includes distribution channels (such as distributors and e-commerce platforms), sales regions, and transportation mode conversion points. This data is obtained through product distribution management systems and logistics tracking platforms and is used to analyze the distribution of carbon emissions throughout the entire product distribution process, from factory to consumer.

[0101] This multi-source data is integrated by associating time series and spatial nodes according to the supply chain's business process (procurement → production → transportation → distribution). Duplicate records and invalid data are removed, and the data format is standardized (e.g., using JSON or CSV). This creates a supply chain carbon flow dataset containing fields such as carbon emissions, timestamps, node locations, and business types for each link in the supply chain. The dataset must ensure data integrity and accuracy. For example, missing transportation mileage data is filled with the average value of adjacent nodes, and abnormal raw material purchase quantity data is corrected through verification with suppliers.

[0102] Principal component analysis (PCA) is used to extract key features of supply chain carbon flows and establish a carbon footprint tracking feature vector. Principal component analysis (PCA) is a dimensionality reduction technique used to convert high-dimensional supply chain carbon flow data into low-dimensional key features. First, the dataset is normalized to eliminate the influence of different variable dimensions and numerical ranges. The calculation formula is:

[0103] in, Indicates the The first sample The original eigenvalues, For the The mean of the features, For the The standard deviation of each feature is calculated. After standardization, the feature covariance matrix is calculated, and the eigenvalues and eigenvectors of the covariance matrix are solved. The number of principal components is determined based on the cumulative contribution rate of the eigenvalues (usually, principal components with a cumulative contribution rate exceeding 85% are selected). The original features are linearly combined to form new principal component features. For example, if the original features include 10 variables such as transportation mileage, fuel consumption, and raw material procurement volume, PCA can extract three principal components, representing comprehensive characteristics such as transportation intensity, production scale, and distribution complexity. The principal component eigenvalues of each sample are arranged in order to form a carbon footprint tracking feature vector, which is used as input for subsequent models.

[0104] A random forest regression model was constructed and the adaptive particle swarm algorithm was used to optimize the model parameter combination. The random forest regression model consists of multiple decision trees. The base learner is constructed through bootstrap sampling and random feature selection. It has strong anti-overfitting ability and good fitting effect for nonlinear data. The initial parameters of the model include the number of decision trees ( ), the maximum number of features considered when each node splits ( ), the maximum depth of the decision tree ( ), etc. The adaptive particle swarm algorithm (APSO) is used to optimize model parameters. It simulates the foraging behavior of a flock of birds to find the optimal solution. Each particle represents a set of parameter combinations, and the algorithm iteratively updates its velocity and position in the search space to find the parameter combination that minimizes the model's loss function. The adaptive mechanism dynamically adjusts the inertia weight and learning factor based on the particle's fitness. For example, when a particle approaches the optimal solution, the inertia weight is reduced to improve local search accuracy; when a particle is trapped in a local optimum, the learning factor is increased to enhance global search capabilities.

[0105] During the model training process, the feature vector (independent variable) and the corresponding carbon footprint (dependent variable) of the supply chain carbon flow dataset are input into the random forest model. Each decision tree makes an independent prediction for the sample, and the average of all decision tree predictions is taken as the model output. The APSO algorithm is used to continuously adjust the parameters to minimize the mean square error (MSE) of the model on the training set. The optimized parameter combination is as follows: 、 、 wait

[0106] The feature vectors are input into the optimized random forest regression model, which outputs a ranking of the carbon emission contributions of each supply chain link. After model training, the random forest's feature importance assessment mechanism is used to calculate the contribution of each principal component feature to the total carbon footprint. Feature importance is measured by calculating the sum of the purity improvements caused by each feature across all decision trees. Features with higher contributions have a greater impact on the overall carbon footprint of the supply chain. For example, if the principal component feature corresponding to transportation mileage has the highest importance score, then transportation is the primary contributor to the supply chain's carbon footprint. The feature importance scores for each link (procurement, production, transportation, and distribution) are normalized and converted into percentage contribution values. These scores are then sorted from high to low to form a list of the carbon emission contributions of each supply chain link.

[0107] Based on the contribution ranking results, key carbon footprint nodes are identified and a supply chain carbon flow optimization path map is generated. Key nodes are identified by links or specific business nodes whose contribution exceeds a preset threshold (e.g., 20%). For example, if the contribution of the transportation link is 35%, the transportation link is identified as a key node. The contribution of sub-nodes within this link, such as specific transportation routes and transportation vehicle types, is further analyzed to identify the transportation route with the highest contribution (e.g., a long-distance trucking route) as the key sub-node. Optimization strategies are developed for key nodes, combining actual supply chain business processes and industry best practices. For example, for high-contribution transportation routes, routes can be optimized to shorten mileage or replaced with new energy vehicles to reduce carbon emissions.

[0108] When generating a supply chain carbon flow optimization path diagram, supply chain nodes are used as vertices and carbon flow directions as edges, with different colors and thicknesses used to represent the contribution of each node and the size of the carbon flow. For example, key nodes are represented by red circles, non-key nodes by blue circles, edges with large carbon flows are drawn with thick lines, and edges with small carbon flows are drawn with thin lines. The optimization path diagram can also be labeled with recommended optimization measures, such as "Replace electric trucks" next to nodes in the transportation link and "Upgrade energy-saving equipment" next to nodes in the production link, providing companies with a visual decision-making basis for optimizing the carbon footprint of the supply chain. The entire process uses a data-driven approach to achieve accurate traceability and scientific optimization of the supply chain's carbon footprint, ensuring that companies can take effective emission reduction measures for high-carbon emission links and improve the environmental sustainability of the supply chain.

[0109] The process for optimizing emission reduction paths using an ensemble learning algorithm based on an ESG comprehensive assessment model is as follows: First, the company's equipment energy efficiency data, process improvement solution library, clean technology application cases, and carbon trading market information are integrated to construct a knowledge graph for emission reduction strategies. This equipment energy efficiency data covers the model, operating age, rated power, and actual energy efficiency values of various production equipment. This data is collected in real time through equipment management systems and energy metering instruments. For example, the actual energy efficiency value of an injection molding machine can be calculated by comparing its energy consumption per unit of output with its rated energy consumption. The process improvement solution library stores the company's accumulated process optimization solutions, including adjustments to injection molding process parameters and heat treatment process improvements. Each solution includes implementation steps, expected emission reduction results, and cost inputs, and is jointly compiled and entered by the production and technical R&D departments. The clean technology application case library collects successful experiences of domestic and international companies in the same industry using clean technologies (such as solar photovoltaic power generation and wastewater recycling systems), including technical principles, implementation cycles, and return on investment. These cases are obtained through industry reports, technical seminars, and other channels, and are reviewed by experts before inclusion in the case library. Carbon trading market information is synchronized in real time with the price trends, quota supply and demand, policy dynamics, etc. of the carbon emission trading platform, and is automatically captured and stored in the system database through the data interface.

[0110] This multi-source, heterogeneous information is structured and converted into nodes and edges in a knowledge graph. Equipment energy efficiency data, such as equipment model and energy efficiency value, serves as entity nodes; process improvement plans, such as optimization measures, costs, and expected results, serve as attribute nodes; clean technology cases, such as technology type and applying companies, serve as case nodes; and carbon trading market prices and quotas serve as market nodes. Relationships between nodes are represented by edges. For example, the equipment node and process improvement plan node are connected by an "applicable process" edge; the clean technology case node and equipment node are connected by a "technology adaptation" edge; and the carbon trading market node and the enterprise's total carbon emissions node are connected by a "quota impact" edge. These nodes and edges are visualized using knowledge graph building tools (such as Neo4j) to form a knowledge graph of emission reduction strategies that encompasses multiple dimensions, including a company's internal resources, external technologies, and market environment.

[0111] A graph convolutional network (GCN) extracts features from node relationships in the knowledge graph to generate a feature matrix for emission reduction strategies. A graph convolutional network (GCN) is a neural network specifically designed for processing graph-structured data. It learns hidden representations of nodes by aggregating feature information from nodes and their neighbors. First, the knowledge graph is normalized, converting node features into numeric vectors (e.g., using one-hot encoding for equipment models). An adjacency matrix is constructed to represent the connectivity between nodes. The adjacency matrix and node feature matrix are then input into the GCN model. Through multiple convolutional layers, high-order correlation features of the nodes are extracted, such as synergistic effects between equipment energy efficiency and process improvement solutions, and linkages between clean technology applications and carbon trading quotas. The output of each convolutional layer serves as the input to the next layer, where a nonlinear transformation is introduced using an activation function (e.g., ReLU). Finally, a low-dimensional feature vector for each node is obtained at the output layer. The feature vectors of all nodes are arranged in sequence to form an emission reduction strategy feature matrix. This matrix contains the potential features and correlations of various emission reduction strategies in the knowledge graph, providing input data for the subsequent optimization model.

[0112] A multi-objective optimization model was established, using the NSGA-II algorithm to find the Pareto optimal solution set. This multi-objective optimization model aims to maximize carbon emission reduction, minimize economic costs, and minimize implementation timelines, while also considering the company's production capacity, technical feasibility, and policy compliance constraints. The carbon emission reduction target was measured by calculating the impact of different emission reduction options on the company's carbon emission intensity and total volume. Economic cost targets included equipment procurement costs, technological transformation costs, and personnel training costs. The implementation timeline target was determined based on the solution's complexity and resource requirements. Constraints included ensuring that equipment modifications could not impact normal production schedules and that the application of clean technologies complied with environmental protection policies.

[0113] The NSGA-II algorithm is a non-dominated sorting genetic algorithm that searches for Pareto optimal solutions in the solution space by simulating the selection, crossover, and mutation operations of biological evolution. The algorithm first generates an initial population, with each individual representing a combination of emission reduction options (e.g., simultaneous equipment upgrades and process improvements). The genetic code of each individual contains parameters such as the choice of option, the order of implementation, and resource allocation. The population is then non-dominated sorted, with individuals divided into different Pareto levels. The crowding distance of each individual is calculated to maintain population diversity. The next generation of the population is generated through selection (e.g., roulette wheel selection), crossover (e.g., single-point crossover), and mutation (e.g., randomly changing the order of implementation of a particular option). This process is repeated until a predetermined number of iterations or convergence criteria are met. The resulting Pareto optimal solution set consists of multiple non-dominated combinations of emission reduction options, each of which achieves an optimal balance between carbon emissions, costs, and time constraints.

[0114] The feature matrix is input into the optimization model, which outputs a set of feasible solutions that meet carbon emission constraints and economic cost limits. The feature matrix, as an input parameter to the optimization model, characterizes the potential performance and associated characteristics of each emission reduction option. Based on the preset objective function and constraints, the model evaluates and screens each option in the feature matrix, eliminating options that do not meet the constraints (e.g., options with costs exceeding the company's budget). The model retains options that meet the constraints and calculates their scores for each objective. Each option in the feasible solution set meets the company's minimum carbon emission reduction threshold and maximum economic cost cap, for example, requiring a carbon emission reduction of at least 10% and a cost not exceeding 5 million yuan.

[0115] The final implementation plan is selected based on the decision preferences, and a phased emission reduction route plan is generated. Decision preferences are determined by corporate management based on their own strategic goals and resource availability. For example, some companies prioritize short-term emission reductions and are willing to bear higher costs; others prioritize cost control and prefer to implement low-cost solutions in phases. The system provides a visual solution comparison tool that displays the performance of the feasible solution set in terms of carbon emissions, costs, and cycles in graphical form (such as radar charts and Gantt charts), helping decision makers to intuitively compare the pros and cons of different solutions. Decision makers use an interactive interface to select their preferred solution combination (e.g., a solution with a 15% carbon emissions reduction, a cost of 4.5 million yuan, and a cycle of 12 months). The system automatically generates a phased emission reduction route plan based on the solution's implementation steps and resource requirements.

[0116] The planning table is divided into chronological phases (such as preparation, implementation, and acceptance), with each phase clearly defining specific tasks, responsible departments, timelines, and resource requirements. For example, the preparation phase includes equipment procurement bidding and technical team formation, with a timeline of one to two months and the responsible departments being the Procurement and Technology departments. The implementation phase includes dismantling old equipment, installing and commissioning new equipment, and adjusting process parameters, with a timeline of three to ten months and the responsible departments being the Production and Engineering departments. The acceptance phase includes carbon emissions monitoring and assessment, and economic benefit analysis, with a timeline of eleven to twelve months and the responsible departments being the Environmental Protection and Finance departments. The planning table also links to relevant documents (such as equipment procurement contracts and technical implementation plans) to facilitate cross-departmental review and implementation, ensuring the implementation of optimized emission reduction path decisions. The entire process integrates multi-source data, applies graph neural networks, and utilizes optimization algorithms to intelligently generate and scientifically plan emission reduction strategies, providing enterprises with emission reduction solutions that deliver both environmental and economic benefits.

[0117] It should be noted that, in this document, relational terms such as first and second, etc., are used only 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 terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0118] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An enterprise carbon emission analysis system based on an ESG comprehensive assessment model, characterized by: It includes: Data collection layer, analysis and processing layer, and decision output layer; The data collection layer includes environmental monitoring devices, energy metering instruments, supply chain tracking modules, and compliance recording units, which are used to conduct real-time monitoring and collect raw data on carbon emission-related factors throughout the entire production and operation process of the enterprise; The analysis and processing layer includes edge computing nodes, distributed storage clusters, and blockchain verification modules. It deploys dynamic optimization algorithms and hybrid encryption protocols to build a trusted data processing environment, perform data cleansing and verification operations, and uses hybrid encryption protocols for secure transmission and classified storage of different business modules and data types. It also achieves dynamic synchronization of enterprise operational data and ESG comprehensive assessment models through standardized interfaces. The decision output layer is used to integrate the company's historical operating data and real-time monitoring data using pattern recognition technology, dynamically calibrate the evaluation indicator system, and establish a multi-dimensional ESG comprehensive evaluation model. Based on this model, an integrated learning algorithm is used to calculate carbon emission intensity, trace the carbon footprint of the supply chain, and optimize emission reduction path decisions; In the analysis and processing layer, the data acquired by each collection terminal is transmitted from the edge computing node to the distributed storage cluster via a dedicated channel. The distributed storage cluster then converts the data format and stores it together with the regulatory information acquired by the compliance recording unit in the blockchain verification module for feature extraction. The data cleaning and verification operation uses outlier detection technology to input the collected raw data into the verification model established by different verification rules for cross-verification. The dynamic synchronization of enterprise operational data and the ESG comprehensive assessment model through standardized interfaces includes: establishing standardized data interfaces to enable interactive communication between business systems and the ESG comprehensive assessment model, synchronizing the enterprise's actual operating status data, and performing correlation mapping on structured data, while also updating model parameter configurations, completing data version management, model iteration updates, and decision instruction verification; The multi-dimensional ESG comprehensive assessment model includes: Establish a two-way correlation channel and dynamic matching mechanism between the actual operating data of enterprises and the parameters of the ESG comprehensive assessment model; By mapping and correlating corporate production and operation activities through data, and using on-site monitoring data, supply chain records, and compliance information as the foundation, we establish a corporate carbon emissions inventory and an industry benchmark comparison framework. We then initialize model parameters based on business characteristics, simulate multiple operating scenarios, and form a dynamically adjustable comprehensive ESG assessment model for companies. Optimize the parameters of the enterprise ESG comprehensive assessment model, import the enterprise's multi-dimensional operational data into the established ESG comprehensive assessment model, and use the Monte Carlo simulation method to conduct sensitivity analysis and correction on the model output results to obtain the optimized enterprise ESG comprehensive assessment model; The enterprise ESG comprehensive assessment model includes operational entity data, an assessment rule engine, a dynamic knowledge base, and the interactive relationships between modules; The operating entity data is the basic data source of the ESG comprehensive assessment model, including the company's direct emission data and indirect emission data; the assessment rule engine establishes a mapping relationship with the actual business data, and quantitatively assesses the carbon emission characteristics through a multi-dimensional indicator system and a weight distribution mechanism; the dynamic knowledge base integrates industry standard data, policy and regulatory texts and best practice cases, and forms an assessment benchmark database through continuous updating and maintenance; the interactive relationship between the modules realizes information exchange through data pipelines, the operating data and the assessment rule engine are standardized through a data conversion interface, and the assessment rule engine and the dynamic knowledge base realize rule matching through semantic parsing technology.

2. The enterprise carbon emission analysis system based on the ESG comprehensive assessment model according to claim 1 is characterized in that: The environmental monitoring device includes at least exhaust gas emission sensors, wastewater detection probes, solid waste metering equipment and noise monitors, which are used to obtain environmental parameters of the enterprise's production and operation site; the energy metering instruments include at least electricity meters, gas flow meters, steam meters and fuel consumption recorders, which are used to count the enterprise's energy consumption data; the supply chain tracking module includes at least logistics information collection terminals, raw material traceability databases and transport vehicle positioning devices, which are used to record carbon emission information at each link of the supply chain.

3. The enterprise carbon emission analysis system based on the ESG comprehensive assessment model according to claim 1 is characterized in that: The carbon emission intensity calculation based on the model using an integrated learning algorithm includes: Collect the company's historical annual energy consumption data, production process record data, equipment operation logs, and product carbon footprint reports to build a carbon emission characteristic dataset; After filling missing values and correcting outliers in the feature data set, divide it into training set and validation set; Configure the initial parameters of the XGBoost-GRU joint model, input the training set into the joint model, use the XGBoost component to sort the feature importance, use the GRU network to perform time series modeling, and use the feature weighting mechanism to eliminate the impact of emission fluctuations in different production cycles until the model error reaches the preset threshold; The validation set is input into the trained joint model to calculate the carbon emission intensity prediction value, and the carbon emission intensity level assessment result of the enterprise is generated by combining it with the industry benchmark value; Based on the assessment results, the preset improvement strategy library is matched to generate targeted carbon emission management recommendations.

4. The enterprise carbon emission analysis system based on the ESG comprehensive assessment model according to claim 1 is characterized in that: The model-based integrated learning algorithm is used to trace the carbon footprint of the supply chain, including: Obtain logistics and transportation data, raw material procurement records, production and processing information, and product distribution paths for each link in the supply chain to build a supply chain carbon flow dataset; The key features of the supply chain carbon flow are extracted through principal component analysis, and the carbon footprint tracking feature vector is established; Construct a random forest regression model and use adaptive particle swarm optimization to optimize the model parameter combination; The feature vector is input into the optimized random forest regression model to output the carbon emission contribution ranking of each link in the supply chain; According to the contribution ranking results, the key nodes of carbon footprint are identified and the supply chain carbon flow optimization path diagram is generated.

5. The enterprise carbon emission analysis system based on the ESG comprehensive assessment model according to claim 1 is characterized in that: The method of using an integrated learning algorithm based on the model to make an optimization decision on the emission reduction path includes: Integrate enterprise equipment energy efficiency data, process improvement solution library, clean technology application cases and carbon trading market information to build a knowledge graph of emission reduction strategies; Perform graph convolutional network feature extraction on the node relationships in the knowledge graph to generate the emission reduction plan feature matrix; A multi-objective optimization model is established and the NSGA-II algorithm is used to find the Pareto optimal solution set; The characteristic matrix is input into the optimization model, and a set of feasible solutions that meet the carbon emission constraints and economic cost limits is output; The final implementation plan is selected based on the decision preferences and a phased emission reduction route planning table is generated.

6. The enterprise carbon emission analysis system based on the ESG comprehensive assessment model according to claim 1 is characterized in that: The specific implementation of the data cleaning and verification operation in the analysis and processing layer includes: Establish a data quality detection rule library, including value range verification rules, data integrity verification rules and logical consistency verification rules; Use a streaming computing engine to verify real-time data one by one, and trigger an early warning mechanism for data that does not meet the verification rules; Perform a full scan of historical batch data and use data repair algorithms to automatically correct abnormal records that can be repaired; The data verification process and exception handling logs are recorded through blockchain smart contracts to generate tamper-proof data quality audit reports.

7. The enterprise carbon emission analysis system based on the ESG comprehensive assessment model according to claim 1 is characterized in that: The method for constructing the evaluation rule engine includes: Collect domestic and international carbon emission accounting standards, industry technical specifications, and policy and regulatory requirements to build a knowledge base of standard terms; Use natural language processing technology to parse standard documents, extract key clauses and form structured evaluation rules; Establish a rule matching engine to automatically associate actual enterprise data with standard terms through semantic similarity calculation; Set up a dynamic adjustment mechanism for rule weights to automatically optimize rule priorities based on policy update frequency and changes in industry practices.

8. A corporate carbon emissions analysis method based on an ESG comprehensive assessment model, characterized by: The following steps are involved: Environmental monitoring devices deployed at production and operation sites collect real-time data on exhaust gas emission concentrations, wastewater pollutant indicators, solid waste generation, and noise levels. Energy meters are also used to obtain electricity, gas, steam, and fuel oil consumption. The collected raw data is filtered for outliers and format standardized by edge computing nodes before being transmitted to a distributed storage cluster for multi-source data fusion. The blockchain verification module is used to encrypt and verify the data integrity and time consistency. Using pattern recognition technology, we conduct feature correlation analysis on historical operational data and real-time monitoring data, dynamically adjust the weight parameters of ESG assessment indicators, and build a multi-dimensional ESG comprehensive assessment model that includes environmental compliance, energy efficiency, and supply chain carbon footprint; A transfer learning framework is used to transfer knowledge across industries for the ESG comprehensive assessment model, and Monte Carlo simulation is used to generate carbon emission intensity forecast data under different emission reduction scenarios; Combined with the logistics routes, raw material sources, and production and processing information recorded by the supply chain tracking module, a full-chain carbon flow mapping relationship is established, and graph neural networks are used to identify key carbon emission nodes; Based on the relationship between carbon emission intensity prediction results and carbon flow mapping, a multi-objective optimization algorithm is used to solve a set of emission reduction paths that meet economic constraints, and a decision-making report containing equipment modification plans, process optimization strategies, and carbon trading recommendations is output; In the analysis and processing layer, the data obtained by each collection terminal is transmitted from the edge computing node to the distributed storage cluster through a dedicated channel. After the distributed storage cluster converts the data format, it is stored together with the regulatory information obtained by the compliance record unit in the blockchain verification module for feature extraction. The data cleaning and verification operation uses outlier detection technology, and the collected raw data is input into the verification model established by different verification rules for cross-verification. Dynamically synchronize enterprise operational data with ESG comprehensive assessment models through standardized interfaces, including: establishing standardized data interfaces to enable interactive communication between business systems and ESG comprehensive assessment models, synchronizing actual enterprise operating status data, and mapping structured data, while also updating model parameter configurations, completing data version management, model iteration updates, and decision instruction verification; Establish a multi-dimensional ESG comprehensive assessment model, including: Establish a two-way correlation channel and dynamic matching mechanism between the actual operating data of enterprises and the parameters of the ESG comprehensive assessment model; By mapping and correlating corporate production and operation activities through data, and using on-site monitoring data, supply chain records, and compliance information as the foundation, we establish a corporate carbon emissions inventory and an industry benchmark comparison framework. We then initialize model parameters based on business characteristics, simulate multiple operating scenarios, and form a dynamically adjustable comprehensive ESG assessment model for companies. Optimize the parameters of the enterprise ESG comprehensive assessment model, import the enterprise's multi-dimensional operational data into the established ESG comprehensive assessment model, and use the Monte Carlo simulation method to conduct sensitivity analysis and correction on the model output results to obtain the optimized enterprise ESG comprehensive assessment model; The enterprise ESG comprehensive assessment model includes operational entity data, an assessment rule engine, a dynamic knowledge base, and the interactive relationships between modules; The operating entity data is the basic data source of the ESG comprehensive assessment model, including the company's direct emission data and indirect emission data; the assessment rule engine establishes a mapping relationship with the actual business data, and quantitatively assesses the carbon emission characteristics through a multi-dimensional indicator system and a weight distribution mechanism; the dynamic knowledge base integrates industry standard data, policy and regulatory texts and best practice cases, and forms an assessment benchmark database through continuous updating and maintenance; the interactive relationship between the modules realizes information exchange through data pipelines, the operating data and the assessment rule engine are standardized through a data conversion interface, and the assessment rule engine and the dynamic knowledge base realize rule matching through semantic parsing technology.

Images