Carbon data notarization method and system combining blockchain and national secret algorithm

Abstract

The application discloses a carbon data storage method and system combining blockchain and national secret algorithm, and belongs to the technical field of carbon data credible storage. The technical scheme points of the application include: obtaining original carbon data and traceability data, and obtaining a business logic feature set and data variability through preset business rules and carbon flow traceability filtering; obtaining a verification instruction through a pattern recognition confidence model based on the business logic feature set and the data variability; obtaining a verification group list through a quantum consensus model based on the verification instruction and the original carbon data; and obtaining on-chain storage records through a national secret algorithm and a blockchain storage contract based on the verification group list. Through the synergistic effect of the pattern recognition confidence model and the quantum consensus model, the application solves the problems of insufficient data source credibility in carbon data management, easy human intervention in the verification process, and weak tamper-proofing capability of storage results, and effectively improves the source credibility, process fairness and result reliability of carbon data management.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of carbon data credible storage, and more particularly to a carbon data storage method and system combining blockchain and national secret algorithm. BACKGROUND

[0002] Currently, in the field of carbon data management, the traditional carbon data management mode mainly relies on the combination of enterprise self-reporting and periodic auditing by third-party institutions. However, in practical applications, it gradually exposes many limitations: the degree of automation and transparency of data source collection is insufficient, errors are easily introduced into data during generation, transmission and aggregation, and even the data is subject to human tampering, and it is difficult to accurately trace and quantitatively evaluate data anomalies; existing technical means are difficult to perform multi-dimensional and intelligent credibility analysis on massive carbon data, lack of deep mining of business logic and industry correlation behind the data, and cannot effectively identify complex and hidden abnormal patterns; the storage mode of verification results is mainly centralized storage, which has the risk of data tampering or loss by a single node, and the audit traceability chain is weak, so the existing technology has deficiencies. SUMMARY

[0003] In view of the deficiencies of the prior art, the present application aims to provide a carbon data storage method and system combining blockchain and national secret algorithm, which realizes accurate quantitative evaluation of data credibility through cooperative processing of carbon flow traceability filtering and preset business rules, realizes intelligent multi-dimensional verification decision through a pattern recognition confidence model fused with an industry knowledge graph, realizes anti-conspiracy random node assignment based on quantum walking consensus mechanism, and builds a secure and reliable storage chain through national secret algorithm and blockchain storage contract, realizing a complete technical closed loop from data collection, credible evaluation, intelligent decision, consensus verification to secure storage, solving the problems of insufficient data source credibility, questionable verification process fairness and weak storage result tamper resistance in existing carbon data management, and improving the source credibility, process fairness, result reliability and complex abnormal pattern recognition ability of the carbon data storage system.

[0004] To achieve the above-mentioned purpose, the present application provides the following technical scheme:

[0005] The present application provides a carbon data storage method combining blockchain and national secret algorithm, comprising:

[0006] obtaining original carbon data and traceability data;

[0007] Based on the original carbon data and traceability data, business logic feature sets and data variability are obtained through preset business rules and carbon flow traceability filtering;

[0008] Based on the business logic feature set and data variability, a verification instruction is obtained through a pattern recognition confidence model, and the pattern recognition confidence model comprises a feature engineering layer, a graph convolution network aggregation layer, a multi-dimensional confidence generation layer, and a comprehensive decision layer;

[0009] Based on the verification instruction and the original carbon data, a verification group list is obtained through a quantum consensus model;

[0010] Based on the verification group list, a chain storage record is obtained through a national secret algorithm and a blockchain storage contract.

[0011] As a further improvement of the application, based on the original carbon data and traceability data, a business logic feature set and data variability are obtained through a preset business rule and carbon flow traceability filtering, comprising:

[0012] Based on the original carbon data and traceability data, a first state vector is obtained through carbon flow traceability filtering;

[0013] Based on the original carbon data and traceability data, a second state vector and a business logic feature set are obtained through a preset business rule;

[0014] Based on the first state vector and the second state vector, data variability is obtained through residual calculation.

[0015] As a further improvement of the application, based on the business logic feature set and data variability, a verification instruction is obtained through a pattern recognition confidence model, comprising:

[0016] Based on the business logic feature set and data variability, an initial feature vector is obtained through a feature engineering layer;

[0017] An industrial knowledge graph is obtained, and based on the industrial knowledge graph and the initial feature vector, a graph embedding vector is obtained through a graph convolution network aggregation layer;

[0018] Based on the initial feature vector and the graph embedding vector, a confidence vector is obtained through a multi-dimensional confidence generation layer;

[0019] Based on the confidence vector, a verification instruction is obtained through a comprehensive decision layer.

[0020] As a further improvement of the application, based on the business logic feature set and data variability, an initial feature vector is obtained through a feature engineering layer, comprising:

[0021] Based on the business logic feature set and data variability, a normalized feature sequence is obtained through data standardization processing;

[0022] Based on the normalized feature sequence, a composite feature group is obtained through feature cross combination;

[0023] obtaining an initial feature vector of a standard dimension through dimensional alignment based on the composite feature group.

[0024] As a further improvement of the application, the graph embedding vector is obtained through a graph convolution network aggregation layer based on the industrial knowledge graph and the initial feature vector, comprising:

[0025] The target enterprise node is determined through node positioning based on the industrial knowledge graph and the initial feature vector.

[0026] The neighbor node set of the target enterprise node is obtained through neighborhood sampling based on the target enterprise node.

[0027] The graph embedding vector is obtained through a graph convolution neural network based on the neighbor node set and the initial feature vector.

[0028] As a further improvement of the application, the confidence vector is obtained through a multi-dimensional confidence generation layer based on the initial feature vector and the graph embedding vector, comprising:

[0029] The fusion feature vector is obtained through vector splicing based on the initial feature vector and the graph embedding vector.

[0030] The basic confidence score is obtained through a fully connected neural network based on the fusion feature vector.

[0031] The multi-dimensional confidence vector is obtained through a normalization exponential function based on the basic confidence score.

[0032] As a further improvement of the application, the step of obtaining the verification group list through a quantum consensus model based on the verification instruction and the original carbon data, comprising:

[0033] The verification node and the verification node state are obtained through a node management service based on the verification instruction.

[0034] The task attribute is obtained through task analysis based on the verification instruction and the original carbon data.

[0035] The verification group list is obtained through a quantum consensus model based on the verification node state and the task attribute.

[0036] As a further improvement of the application, the verification group list is obtained through a quantum consensus model based on the verification node state and the task attribute, comprising:

[0037] The initial quantum graph is obtained through a quantum state initialization algorithm based on the verification node state.

[0038] The quantum potential configuration is obtained through potential optimization based on the task attribute.

[0039] Based on the initial quantum graph structure and quantum potential energy configuration, a verification group list is obtained through a quantum consensus model.

[0040] As a further improvement of the application, the chain storage record is obtained based on the verification group list, the national secret algorithm and the blockchain storage contract.

[0041] Based on the verification group list, the original carbon data and the preset business rule, a group signature is obtained through the national secret SM2 threshold signature algorithm and the national secret SM3 hash algorithm.

[0042] Based on the group signature, a chain storage record is obtained through the blockchain storage contract.

[0043] The application provides a carbon data storage system combining blockchain and national secret algorithm, which comprises:

[0044] A data traceability module is configured to obtain original carbon data and traceability data, and obtain a business logic feature set and data variability through a preset business rule and carbon flow traceability filtering based on the original carbon data and traceability data.

[0045] A confidence decision module is configured to obtain a verification instruction through a pattern recognition confidence model based on the business logic feature set and data variability, wherein the pattern recognition confidence model comprises a feature engineering layer, a graph convolution network aggregation layer, a multi-dimensional confidence generation layer and a comprehensive decision layer.

[0046] A quantum consensus module is configured to obtain a verification group list through a quantum consensus model based on the verification instruction and the original carbon data.

[0047] A national secret storage module is configured to obtain a chain storage record through a national secret algorithm and a blockchain storage contract based on the verification group list.

[0048] The application quantitatively evaluates the data quality and abnormality degree based on the original carbon data and traceability data through a preset business rule and carbon flow traceability filtering, obtains a differentiated verification instruction through a pattern recognition confidence model according to a business logic feature set and data variability, completes the anti-conspiracy random assignment of verification nodes through a quantum consensus model based on the verification instruction and the original carbon data, and finally realizes the secure storage of verification evidence based on a verification group list through a national secret algorithm and a blockchain storage contract, thereby realizing the whole-process closed-loop management of carbon data from collection and evaluation, intelligent decision, consensus verification to credible storage, solving the problems of insufficient data source credibility, easy human intervention in the verification process and weak tamper resistance of storage results in traditional carbon data management, and improving the automation level, verification fairness, system security and abnormality recognition accuracy of carbon data management. BRIEF DESCRIPTION OF DRAWINGS

[0049] Figure 1 A step flow chart of the carbon data storage method combining blockchain and national secret algorithm of the present application;

[0050] Figure 2 A step flow chart of obtaining a verification instruction through a pattern recognition confidence model;

[0051] Figure 3 A step flow chart of obtaining a verification group list through a quantum consensus model;

[0052] Figure 4 A structural schematic diagram of the carbon data storage system combining blockchain and national secret algorithm of the present application. DETAILED DESCRIPTION

[0053] The technical solutions of the present application will be described in detail below through the accompanying drawings and specific embodiments, and it should be understood that the embodiments and specific features in the embodiments are detailed descriptions of the technical solutions of the present application, but not limitations of the technical solutions of the present application.

[0054] The term "and / or" in the following merely describes the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B can represent the three cases of existence of A alone, existence of A and B at the same time, and existence of B alone. In addition, the character " / " generally represents an "or" relationship between the associated objects before and after it.

[0055] As shown in Figure 1 The present application provides a carbon data storage method combining blockchain and national secret algorithm, which comprises:

[0056] obtaining original carbon data and traceability data;

[0057] Based on the original carbon data and the traceability data, the business logic feature set and the data variability are obtained through the preset business rules and the carbon flow traceability filtering;

[0058] Based on the business logic feature set and the data variability, a verification instruction is obtained through a pattern recognition confidence model, and the pattern recognition confidence model comprises a feature engineering layer, a graph convolution network aggregation layer, a multi-dimensional confidence generation layer and a comprehensive decision layer;

[0059] Based on the verification instruction and the original carbon data, a verification group list is obtained through a quantum consensus model;

[0060] Based on the verification group list, a chain storage record is obtained through a national secret algorithm and a blockchain storage contract.

[0061] The original carbon data is time series data reflecting the core quantitative information of carbon emissions and carbon emission reduction of enterprises in production and operation activities, which is obtained through real-time collection of enterprise data interface, energy consumption account statistics and detection of third-party detection agencies, including enterprise identification, carbon yield data, carbon energy consumption data, enterprise basic information and enterprise registration information, etc. The enterprise basic information includes industry classification code and business scale parameters; the traceability data is the data recording the source and flow process of the original carbon data from generation, collection, transmission to processing, which is obtained through data collection terminal log recording, transmission link node tracking and data processing system operation recording, including upstream data source identification, enterprise industry classification code, enterprise business scale parameters, verification node registration information, industry knowledge graph and verification node professional ability label, etc.; the preset business rule is an executable rule library based on national carbon accounting standards, which is obtained through natural language processing technology and used to guide carbon data accounting and verification; the carbon flow traceability filtering is an algorithm model that fuses data traceability certificates and business rules and dynamically perceives and processes the relationship of carbon data, which is used to filter noise interference in data transmission process and correct data deviation; the business logic feature set is an index combination reflecting the correlation between carbon emissions and production and operation, including unit product carbon emissions and energy consumption carbon emission intensity; the data variation degree is an index used to quantify the degree of deviation of the original carbon data and the traceability data from the expected state in the collection, transmission and processing process.

[0062] The pattern recognition confidence model is a multi-dimensional evaluation system based on deep learning fusion of industry knowledge graph and multi-dimensional features, which is used to comprehensively judge the data credibility, including feature engineering layer, graph convolution network aggregation layer, multi-dimensional confidence generation layer and comprehensive decision layer. The feature engineering layer is a data preprocessing module, which is used to standardize and combine the input features; the graph convolution network aggregation layer is a feature extraction module based on graph structure, which is used to obtain associated enterprise information from the industry knowledge graph; the multi-dimensional confidence generation layer is a neural network inference module, which is used to obtain multi-dimensional confidence scores; the comprehensive decision layer is a strategy execution module, which is used to obtain verification instructions according to multi-dimensional confidence scores; the verification instruction is an instruction obtained through the pattern recognition confidence model, which is used to indicate the start of the carbon data verification process; the quantum consensus model is a random assignment algorithm based on quantum walk, which is used to generate an unpredictable and task-matched verification group allocation scheme; the verification group list is a selected set of verification node identifiers; the national encryption algorithm is a cryptographic algorithm approved by the National Cryptography Administration, which is used to realize data signature and encryption; the blockchain storage contract is a smart contract deployed on the blockchain, which is used to realize automatic storage verification; the on-chain storage record is an unalterable verification evidence stored on the blockchain.

[0063] The embodiment realizes dynamic perception of data relationship and noise filtering through carbon flow traceability filtering, realizes multi-dimensional credibility evaluation and verification decision through a pattern recognition confidence model, realizes anti-conspiracy random assignment and professional matching through a quantum consensus model, realizes high-security trusted storage through a national encryption algorithm and a blockchain storage contract, realizes full-process closed-loop management from data collection, trusted evaluation, consensus verification to on-chain storage, and improves the source credibility, process fairness, result reliability and system attack resistance of the carbon data storage system.

[0064] Further, the embodiment provides a step of obtaining a business logic feature set and data variability based on original carbon data and traceability data through preset business rules and carbon flow traceability filtering, comprising:

[0065] Based on the original carbon data and the traceability data, a first state vector is obtained through carbon flow traceability filtering;

[0066] Based on the original carbon data and the traceability data, a second state vector and a business logic feature set are obtained through preset business rules;

[0067] Based on the first state vector and the second state vector, a data variability is obtained through residual calculation.

[0068] The carbon flow traceability filtering is based on a Kalman filtering algorithm, which estimates the state vector of carbon data through state prediction and update steps. The state transition equation and the observation model are constructed based on the dynamic characteristics of carbon flow. The first state vector is the state estimation result obtained by state initialization, prediction and update processing of the original carbon data and the traceability data through carbon flow traceability filtering, including carbon emission estimation value, upstream data source hash value and initial data variability. The second state vector is the state output obtained by logical verification of the original carbon data and the traceability data based on the preset business rules, including carbon emission theoretical value, upstream data source standard hash value and data variability threshold. The business logic feature set is an index combination reflecting the correlation between carbon emission and production and operation, including unit product carbon emission and energy consumption carbon emission intensity. The residual calculation is a data processing method for quantifying the degree of data abnormality by comparing the differences between the first state vector and the second state vector.

[0069] Specifically, first, carbon flow traceability filtering processing is performed on the original carbon data and the traceability data to obtain a first state vector: based on the original carbon data, the enterprise to which the original carbon data belongs is initialized for carbon flow traceability filtering, and an initial state vector is obtained, the initial state vector including an initial carbon emission estimation value, an initial hash value of an upstream data source in the traceability data, and an initial data variability reference value; based on historical carbon data and traceability data of each enterprise, a current time carbon emission value and a state covariance are predicted by a time series prediction algorithm; the actual collected original carbon data and the traceability data are filtered through carbon flow traceability filtering; an initial residual between the actual value and the predicted value is calculated, and an observation noise matrix is dynamically adjusted according to a traceability data calculation model complexity and a data source quantity, to obtain an updated state vector and a covariance matrix, and a first state vector is integrated , wherein is a carbon emission estimation value, is an upstream data source hash value, is an initial data variability; then, the original carbon data is substituted into a carbon accounting formula in a preset business rule library to calculate a carbon emission theoretical value , in combination with a standard hash value of an upstream data source in the preset business rule library and a preset data variability threshold , to obtain a second state vector ; meanwhile, carbon yield and carbon energy consumption data are obtained from the original carbon data, and are associated with the calculated carbon emission estimation value, to filter out associated indexes that meet business rules to obtain a business logic feature set; then, based on the first state vector and the second state vector, a difference value is calculated for elements of the same dimension in the first state vector and the second state vector one by one, each dimension difference value is converted to a [0, 1] interval by a Min-Max normalization method, and a final data variability is obtained by weighted summation.

[0070] The embodiment realizes dynamic tracking and noise filtering of the actual state of carbon data through carbon flow traceability filtering, realizes benchmark determination and associated feature extraction of carbon data compliance through preset business rules, realizes quantification of the deviation degree between the actual state and the expected state of data through residual calculation, realizes accurate evaluation of the credibility of carbon data sources through the synergy of multi-dimensional data processing, solves the problems that the traditional method cannot quantify the data anomaly degree and the business feature extraction is not standardized, improves the pertinence of the business logic feature set and the accuracy of the data variability, and provides high-quality input for subsequent verification decision of the pattern recognition confidence model.

[0071] Further, the embodiment provides a step of obtaining a verification instruction through a pattern recognition confidence model based on a business logic feature set and data variability, including:

[0072] Based on the business logic feature set and data variability, an initial feature vector is obtained through a feature engineering layer;

[0073] Obtain an industry knowledge graph, and based on the industry knowledge graph and initial feature vectors, obtain graph embedding vectors through a graph convolutional network aggregation layer;

[0074] Based on the initial feature vector and graph embedding vector, a confidence vector is obtained through a multi-dimensional confidence generation layer;

[0075] Verification instructions are obtained through a comprehensive decision-making layer based on the confidence vector.

[0076] The pattern recognition confidence model is trained using historical carbon data and multi-source reliable data. The historical carbon data comes from publicly available carbon emission datasets, historically reported data from enterprises, and third-party audit data to ensure the representativeness and reliability of the training data. Cross-validation and outlier filtering methods are used during the training process to improve the robustness of the model. The initial feature vector is a standardized feature representation obtained by standardizing, cross-combining, and dimensional alignment of the business logic feature set and data variability. The industry knowledge graph is a graph structure database including enterprise nodes, product nodes, and their relationships, constructed using publicly available data from industry regulatory authorities and industry chain survey data. The graph embedding vector is a vectorized representation obtained by aggregating the features of the target enterprise and its neighboring nodes in the industry knowledge graph using a graph convolutional network. The confidence vector is a multi-dimensional confidence score set obtained by fusing the initial feature vector and the graph embedding vector using a neural network.

[0077] Specifically, such as Figure 2 As shown, firstly based on the business logic feature set and data variability By using the feature engineering layer of the pattern recognition confidence model, a normalized feature sequence is obtained through the Z-score normalization method. By using the feature cross-combination method, a composite feature group containing interaction terms is obtained. Principal component analysis is used to align dimensions and obtain initial feature vectors of standard dimensions. Then based on the initial feature vector Industry knowledge graph By using a graph convolutional network aggregation layer based on a pattern recognition confidence model, target enterprise nodes are located in the industry knowledge graph through a node matching algorithm. The first-order and second-order neighbor sets of each target enterprise node are obtained through a random walk sampling strategy. Graph embedding vectors are obtained by performing feature propagation and aggregation operations through a graph convolutional neural network. Then, based on the initial feature vector And graph embedding vector The vector splicing operation is performed through the multi-dimensional confidence generation layer of the pattern recognition confidence model to obtain a fusion feature vector The nonlinear transformation is performed through the fully connected neural network to obtain a basic confidence score The normalization processing is performed through the softmax function to obtain a multi-dimensional confidence vector The multi-dimensional confidence vector includes a time sequence smoothness score , a logic consistency score and a peer deviation score ; finally, based on the multi-dimensional confidence vector The verification instruction is obtained through the preset threshold vector through the comprehensive decision layer of the pattern recognition confidence model, the verification instruction includes a data range to be verified and an emergency level, and the emergency level is a quantitative value of the priority of the verification task, wherein the preset threshold vector includes a time sequence smoothness score threshold , a logic consistency score threshold and a peer deviation score threshold , the preset threshold vector is set based on statistical analysis of historical carbon data, and is dynamically adjusted through an online learning mechanism: the system regularly evaluates the accuracy of the verification instruction and the feedback result, optimizes the threshold parameter through reinforcement learning to adapt to data distribution changes and business needs; the threshold is verified through backtracking test and simulation environment to ensure its effectiveness in different scenarios; wherein the emergency level includes: global emergency, local emergency, directional importance and regular sampling; global emergency requires immediate execution and emergency mobilization of the maximum resource quota of the system; local emergency requires high priority execution, i.e. it is started in a short time, usually for a specific enterprise or associated link, and its priority is lower than that of global emergency; directional importance is a standard priority, which is arranged and completed within a certain period, and resources are allocated according to standard quota, and directional verification is performed for specific abnormal indicators, and its priority is lower than that of local emergency; regular sampling is a background or batch processing priority, which is executed when the system load is low, and the resource consumption is controlled at the lowest level, which is mainly used for continuous monitoring and compliance sampling, and its priority is lower than that of directional importance;

[0078] If the time sequence smoothness score is greater than the time sequence smoothness score threshold, the logic consistency score is greater than the logic consistency score threshold, and the peer deviation score is greater than the peer deviation score threshold, it indicates that the overall data reliability is high, and the verification instruction is: execute regular sampling, the priority is P3, and the data range to be verified is randomly extracted not more than original carbon data and traceability data of the enterprise; if any one of the three confidence scores is less than or equal to the threshold value thereof, a key verification is triggered, the key verification includes time sequence anomaly analysis, logic contradiction investigation and peer comparison review, and according to the specific abnormal item, the instruction is subdivided as follows: if only the time sequence smoothness score is less than or equal to the time sequence smoothness score threshold value, the verification instruction is: perform time sequence anomaly analysis, the priority is P2, and the data range to be verified is the full amount of time sequence data of the corresponding enterprise for consecutive periods; if only the logic consistency score is less than or equal to the logic consistency score threshold value, the verification instruction is: perform logic contradiction investigation, the priority is P2, and the data range to be verified is the accounting unit and related traceability data that triggered the logic alarm; if only the peer deviation score is less than or equal to the peer deviation score threshold value, the verification instruction is: perform peer comparison review, the priority is P2, and the data range to be verified is the original carbon data and traceability data of the corresponding enterprise and a random selected household enterprise among peer enterprises of the corresponding enterprise; if any two of the three confidence scores are less than or equal to the threshold value thereof, an emergency verification is triggered, the emergency verification includes data tampering suspicion investigation, systematic concealment suspicion investigation and accounting method violation suspicion investigation, and according to the abnormal combination, the instruction is subdivided as follows: if only the peer deviation score is greater than the peer deviation score threshold value, the verification instruction is: perform data tampering suspicion investigation, the priority is P1, and the data range to be verified is expanded to the full amount of original carbon data and traceability data of the direct upstream data source enterprise of the enterprise and the enterprise; if only the logic consistency score is greater than the logic consistency score threshold value, the verification instruction is: perform systematic concealment suspicion investigation, the priority is P1, and the data range to be verified is expanded to the full amount of original carbon data and traceability data of the main peer competitor enterprise of the enterprise and the enterprise; if only the time sequence smoothness score is greater than the time sequence smoothness score threshold value, the verification instruction is: perform accounting method violation suspicion investigation, the priority is P1, and the data range to be verified is expanded to the full amount of original carbon data and traceability data of all associated companies within the group to which the enterprise belongs; if the time sequence smoothness score, the logic consistency score and the peer deviation score are all less than or equal to the corresponding threshold value, it indicates that the data credibility has collapsed comprehensively and there is a high risk, an emergency verification is triggered, the verification instruction is: perform comprehensive emergency verification and systematic risk assessment, the priority is P0, and the data range to be verified is the full amount of data of all associated enterprises, and the cross-chain verification and supervision early warning process is immediately started. In the embodiment, the priorities of the verification instructions are P0, P1, P2 and P3 from high to low, wherein P0 is prior to P1, P1 is prior to P2, and P2 is prior to P3, and the priority order is used to determine the sequence of the verification instructions in the system resource scheduling and task execution sequence.

[0079] For example, the embodiment assumes that the business logic feature set of a certain enterprise has a unit product carbon emission of and an energy consumption carbon emission intensity of , the data variability is , the initial feature vector is obtained after processing by the feature engineering layer ; the target enterprise node and the neighbor node set thereof are located in the industrial knowledge graph , the graph embedding vector is obtained by graph convolution network aggregation , the confidence vector is calculated by the multi-dimensional confidence generation layer = , wherein is the time series smoothness score, is the logical consistency score, is the industry deviation score, and the preset threshold vector in the comprehensive decision layer is , wherein is the time series smoothness score threshold, is the logical consistency score threshold, is the industry deviation score threshold. In an example, the initial values of the preset threshold vector are set based on the statistical analysis of historical carbon data, the initial value of the time series smoothness score threshold can be , the initial value of the logical consistency score threshold can be , and the initial value of the industry deviation score threshold can be , and the initial values are further dynamically adjusted by an online learning mechanism combined with verification feedback. Since the time series smoothness score is greater than the time series smoothness score threshold , the logical consistency score is greater than the logical consistency score threshold , and the industry deviation score is greater than the industry deviation score threshold , the verification instruction is obtained as follows: execute regular sampling, the priority is P3, and the verification data range is random sampling of no more than of the enterprise original carbon data and traceable data. The initial values of the preset threshold vector in the example are only examples, and a person skilled in the art can set them according to actual conditions, which are not limited by the example. The feature engineering layer in the example realizes the standardization and enhancement of multi-source features, the graph convolution network aggregation layer fuses industrial correlation information, the multi-dimensional confidence generation layer realizes the accurate quantification of multi-dimensional credibility, and the comprehensive decision layer realizes the intelligent execution of verification strategies. The example realizes the deep fusion of carbon data multi-source information and the recognition of complex abnormal patterns, solves the problem of the traditional single index evaluation, and improves the scientificity and pertinence of the verification instruction and the accuracy of the abnormal pattern recognition and the scientificity of the verification decision.

[0080] The feature engineering layer in the example realizes the standardization and enhancement of multi-source features, the graph convolution network aggregation layer fuses industrial correlation information, the multi-dimensional confidence generation layer realizes the accurate quantification of multi-dimensional credibility, and the comprehensive decision layer realizes the intelligent execution of verification strategies. The example realizes the deep fusion of carbon data multi-source information and the recognition of complex abnormal patterns, solves the problem of the traditional single index evaluation, and improves the scientificity and pertinence of the verification instruction and the accuracy of the abnormal pattern recognition and the scientificity of the verification decision.

[0081] ​Furthermore, this embodiment provides a step for obtaining an initial feature vector through a feature engineering layer based on a business logic feature set and data variability, including:

[0082] Based on the business logic feature set and data variability, a normalized feature sequence is obtained through data standardization processing;

[0083] Composite feature groups are obtained by combining normalized feature sequences through feature crossover.

[0084] Based on the composite feature set, the initial feature vector of standard dimensions is obtained through dimension alignment.

[0085] Among them, data standardization is a data preprocessing method that uses the Z-score method to transform features of different dimensions to the same dimension, thereby eliminating differences in dimensions between features; normalized feature sequence is a feature sequence in which each feature value conforms to a standard normal distribution after data standardization, including the standardization results corresponding to the business logic feature set and the data variability; feature cross combination is a feature engineering method that performs product and ratio operations on single features with correlation in the normalized feature sequence to obtain new features, thereby mining implicit relationships between features; composite feature group is a feature set obtained through feature cross combination that includes the original standardized features and the newly generated composite features; and dimension alignment is a data processing method that uses principal component analysis to normalize feature dimensions to a preset standard dimension, thereby ensuring consistency of feature dimensions.

[0086] Specifically, firstly, based on the business logic feature set and data variability The normalized value of each feature is calculated using the Z-score normalization method, resulting in a normalized feature sequence. Then, based on the normalized feature sequence Interactive features are obtained by performing feature product operations through a feature cross-combiner, and ratio features are obtained by performing feature ratio operations. The original normalized feature sequences are then combined. The generated interaction features and ratio features yield a composite feature set. Ultimately based on composite feature groups Principal component analysis is performed using dimension alignment to extract key feature components, and the feature dimensions are then normalized to a predefined standard dimension. This yields the initial feature vectors of standard dimensions. .

[0087] The embodiment is based on the business logic feature set and the data variability, and realizes unified representation of multi-source heterogeneous features through multi-level processing of data standardization, feature cross combination and dimension alignment, improves the integrity of feature expression and the normativity of model input, solves the model calculation deviation problem caused by dimensional difference and dimension confusion of traditional features, improves the accuracy of subsequent graph convolution network aggregation and confidence evaluation, and provides high-quality feature input for the pattern recognition confidence model.

[0088] Further, the embodiment provides a step of obtaining an industrial knowledge graph, obtaining a graph embedding vector through a graph convolution network aggregation layer based on the industrial knowledge graph and an initial feature vector, comprising:

[0089] determining a target enterprise node based on the industrial knowledge graph and the initial feature vector through node positioning;

[0090] obtaining a neighbor node set of the target enterprise node through neighborhood sampling based on the target enterprise node;

[0091] obtaining a graph embedding vector through a graph convolution neural network based on the neighbor node set and the initial feature vector.

[0092] The node positioning is a search positioning method for matching a corresponding graph node in the industrial knowledge graph according to an enterprise identifier, and is used to establish the association between enterprise data and the graph node. The target enterprise node is a graph node in the industrial knowledge graph corresponding to the enterprise to which the original carbon data currently processed belongs. The neighborhood sampling is a graph traversal method for obtaining associated neighbor nodes from the target enterprise node, and is used to construct a local graph structure. The neighbor node set is a node group obtained through neighborhood sampling, which is composed of first-order and second-order neighbor nodes of the target enterprise node. The graph convolution neural network is a deep learning model for feature extraction based on graph structure data, and is used to aggregate feature information of the target node and its neighbor nodes. On the basis of the initial feature vector output in the feature engineering layer, the graph convolution network aggregation layer adopts a two-layer graph convolution structure to realize feature extraction. The input dimension of the first-layer graph convolution structure is consistent with the dimension of the initial feature vector, and the output dimension is 128. The output dimension of the second-layer graph convolution structure is 64, and the activation function is ReLU. Finally, a 64-dimensional graph embedding vector is obtained. When the graph convolution neural network is trained, the learning rate is set to 0.001, the iteration is 200 cycles, the Adam optimizer is used, the contrast loss function is adopted, and the Dropout rate of 0.3 and L2 regularization are used to enhance the generalization ability. In the embodiment, the number of neurons, learning rate and training parameters of the graph convolution network are all examples, and a person skilled in the art can set them according to the actual situation, and the embodiment does not limit them.

[0093] Specifically, first, an initial feature vector is obtained and the enterprise identifier corresponding thereto In the industrial knowledge graph corresponding target enterprise node is obtained through an identity matching algorithm ; then, taking the target enterprise node as a starting point, first-order neighbor nodes directly connected to the target enterprise node and second-order neighbor nodes indirectly connected to the target enterprise node are obtained through a random walk sampling algorithm to obtain a neighbor node set ; finally, based on the neighbor node set and the initial feature vector , feature propagation and aggregation operations are performed through a graph convolutional neural network, inter-layer feature transmission is performed using a graph convolution formula, and after two layers of graph convolution calculation, a graph embedding vector is obtained.

[0094] Based on the industrial knowledge graph and the initial feature vector, the embodiment realizes deep fusion of target enterprise carbon data and industrial background information through the cooperative processing of node positioning, neighborhood sampling and a graph convolutional neural network, solves the problem that traditional evaluation only focuses on single enterprise data and cannot identify collaborative fraud, improves the recognition ability of hidden and associated data anomalies, and provides feature support in the industrial dimension for multi-dimensional confidence evaluation.

[0095] Further, the embodiment provides a step of obtaining a confidence vector through a multi-dimensional confidence generation layer based on an initial feature vector and a graph embedding vector, including:

[0096] obtaining a fusion feature vector through vector splicing based on the initial feature vector and the graph embedding vector;

[0097] obtaining a basic confidence score through a fully connected neural network based on the fusion feature vector;

[0098] obtaining a multi-dimensional confidence vector through a normalized exponential function based on the basic confidence score.

[0099] Vector concatenation is a data fusion method that connects and combines multiple feature vectors along a specified dimension to integrate feature information from different sources. The fused feature vector is a comprehensive feature representation obtained by concatenating the initial feature vector and the graph embedding vector. The fully connected neural network contains three hidden layers with 256, 128, and 64 neurons respectively, all using ReLU activation functions. The output layer has three neurons corresponding to the three confidence scores. During training, the learning rate is set to 0.0005, with 150 iterations, using the AdamW optimizer and mean squared error as the loss function. He initialization and a step-decay learning rate strategy are employed to stabilize the training process. The fully connected neural network is a deep learning model where each layer of neurons is connected to the next fully connected layer for nonlinear transformation of features. The base confidence score is the original confidence score calculated by the neural network. The normalized exponential function is a mathematical function that converts a vector into a probability distribution, used for normalizing the confidence score. In this embodiment, the number of neurons, learning rate, and training parameters of the fully connected neural network are examples; those skilled in the art can set them according to actual conditions, and this embodiment does not impose any limitations.

[0100] Specifically, firstly, based on the initial feature vector And graph embedding vector By concatenating the feature dimensions through vector concatenation, a fused feature vector is obtained. Then based on the fused feature vector The basic confidence score is obtained by calculating using a three-layer fully connected neural network and performing a nonlinear transformation using the ReLU activation function. Ultimately based on the baseline confidence score The multidimensional confidence vector is obtained by calculating the normalized exponential function and the softmax function. Each dimension represents a confidence score for a different aspect, including the time series smoothness score. Logical consistency score Deviation score from peers .

[0101] This embodiment, based on initial feature vectors and graph embedding vectors, achieves credibility assessment of carbon data in both its own characteristics and industry-related characteristics through cascaded processing of vector concatenation, fully connected neural networks, and normalized exponential functions. This solves the problem that traditional single-dimensional assessment cannot fully reflect data credibility, improves the comprehensiveness and accuracy of confidence assessment, and provides a quantitative basis for generating verification instructions for the comprehensive decision-making level.

[0102] Furthermore, this embodiment provides a step for obtaining a list of verification groups based on verification instructions and raw carbon data using a quantum consensus model, including:

[0103] obtaining the verification node and the verification node state through the node management service based on the verification instruction;

[0104] obtaining the task attribute through task analysis based on the verification instruction and the original carbon data;

[0105] obtaining the verification group list through the quantum consensus model based on the verification node state and the task attribute.

[0106] The node management service is a system service for maintaining and managing the real-time state of all registered verification nodes, and the verification node state is obtained by calling the node registration table and state monitoring interface; the verification node is a server or container deployed with a trusted execution environment technology, used to independently execute verification algorithms in a protected isolated environment; the verification node state is a dynamic state set of all available verification nodes, including node identification, professional ability label, and current resource utilization; task analysis is a process of obtaining key features of verification tasks according to the verification data range in the verification instruction and the enterprise attribute information in the original carbon data, used to determine the required verification algorithm type and resource constraint condition, and the verification task is obtained according to the verification instruction; the task attribute is a structured data object that completely describes the technical features of each verification task, including target enterprise industry, verification algorithm type, urgency level, and resource demand; the quantum consensus model is a random assignment algorithm based on the principle of quantum walk, used to generate an unpredictable and task demand matching verification group allocation scheme.

[0107] Specifically, first, the target enterprise identification is determined based on the verification data range in the verification instruction, the industry classification code and data feature mode of the target enterprise are obtained based on the corresponding enterprise basic information in the original carbon data, and the urgency level in the verification instruction is obtained; then, based on the complete verification instruction, the node management service is called to query the attribute information of all active verification nodes through the node registration table interface, and the current resource utilization and response performance indicators of each verification node are collected in real time through the state monitoring interface, and the verification node state including node identification, professional ability label, and current resource utilization is obtained; then, based on the industry classification code of the target enterprise, the data feature mode, and the urgency level in the verification instruction, the required verification algorithm type and resource demand of the verification task are calculated through the task analyzer to obtain the complete task attribute; finally, based on the verification node state and the task attribute, the quantum consensus model is executed to perform a multi-dimensional optimization node matching and anti-prediction random selection process, and a verification group list that meets the professional requirements of the verification task, load balancing, and unpredictability is obtained.

[0108] For example, the embodiment assumes that the received verification instruction is to perform logical contradiction checking, the priority is P2, the data range to be verified is specifically designated as the current full amount of raw carbon data of enterprise B, and the urgency is P2; based on the registration information of enterprise B in the raw carbon data, it is determined that the industry classification thereof is industry D, and the data characteristic mode is high-frequency sampling; the verification node identifier obtained through the node management service is N201, N202, and N203, and the corresponding verification node state is: {N201: ["perform logical contradiction checking", "high-frequency data processing"], resource utilization rate }、{N202: ["abnormality detection"], resource utilization rate }、{N203: ["perform logical contradiction checking", "traceability analysis"], resource utilization rate };the task attribute obtained through task analysis is {industry: "industry D", verification algorithm type: ["perform logical contradiction checking"], urgency: P2, resource requirement: };based on the verification node state and the task attribute, the verification group list is obtained through quantum consensus model optimization calculation as [N201, N203].

[0109] The embodiment realizes unified monitoring and scheduling of verification node resources through the node management service, converts abstract verification instructions into calculable task feature vectors through task analysis, realizes anti-conspiracy random assignment and multi-objective optimization selection through the quantum consensus model, ensures system data consistency through complete parameter transmission and state maintenance, realizes professional matching of the verification group, resource balance, and unpredictability of the selection process, eliminates the possibility of collusion and human manipulation risk in the verification link from the mechanism, and significantly improves the fairness, security, and resource utilization efficiency of the carbon data verification system.

[0110] Further, the embodiment provides a step of obtaining a verification group list based on verification node states and task attributes through a quantum consensus model, comprising:

[0111] obtaining an initial quantum graph based on the verification node state through a quantum state initialization algorithm;

[0112] obtaining quantum potential configuration through potential optimization based on the task attribute;

[0113] obtaining the verification group list through the quantum consensus model based on the initial quantum graph structure and the quantum potential configuration.

[0114] The quantum state initialization algorithm is a calculation method for mapping the verification node state to a graph structure with a quantum superposition state distribution, and is used to construct an initial quantum graph capable of quantum walk evolution; the initial quantum graph is a graph model taking the verification nodes as vertices, taking the professional similarity between the verification nodes as edges, and giving each verification node an initial quantum state probability amplitude distribution; the potential energy optimization is a calculation process of dynamically adjusting the potential energy value on each edge in the initial quantum graph according to the task attributes, and is used to construct a quantum potential energy landscape adapted to the task to guide the quantum walk path; the quantum potential energy configuration is a set of potential energy values on all edges in the initial quantum graph, and the values are determined by the professional similarity of the verification node and the corresponding verification task, and the resource utilization rate of the verification node.

[0115] Specifically, as shown in the figure, Figure 3 first, based on the professional ability labels of all verification nodes in the verification node state, the professional similarity between any two verification nodes is calculated through the quantum state initialization algorithm, and the edge connection is established according to all professional similarities and a preset professional similarity threshold, that is, the edge connection is established only when the professional similarity is greater than the preset professional similarity threshold, each verification node is mapped to a quantum node in the quantum graph, and the initial quantum state is assigned to each verification node using the uniform superposition principle , obtaining an initial quantum graph with nodes and edges; then, based on the verification algorithm type requirement in the task attribute and the professional ability label in the verification node state, the professional matching degree score of each verification node and the verification task is calculated through the potential energy optimization algorithm, and the current resource utilization rate and network topology information of the verification node are combined to calculate the potential energy value on each edge , wherein are the indexes of the two different verification nodes connected by the edge, respectively, the lower the professional matching degree or the higher the node load, the greater the potential energy value on the corresponding edge, and the complete quantum potential energy configuration is obtained ; finally, based on the initial quantum graph structure and the quantum potential energy configuration, the quantum consensus model is used to project and measure the system quantum state by using continuous-time quantum walk evolution, and the final selected verification node combination is determined by using a weighted random selection algorithm according to the probability distribution obtained by measurement, obtaining a verification group list In the quantum consensus model of the embodiment, each verification node is mapped to a quantum bit, the ground state represents that the verification node is not selected, and the excited state represents that the verification node is selected, and the total state of the system is the tensor product of all verification node quantum bits; the quantum circuit includes an initialization phase, an evolution phase and a measurement phase, and in the initialization phase, all quantum bits are in a uniform superposition state , This indicates an initial equal probability selection. This represents the total number of verification nodes. This serves as the index number of the verification node in the initial quantum graph; the evolution phase employs a continuous-time quantum walk model, with evolution operators... Hamiltonian From the initial quantum diagram The adjacency matrix A and the quantum potential energy-based configuration Constructed diagonal potential matrix Build together For evolution time, The starting node number is used to define the edges in the potential energy matrix. Individual potential energy values ​​in the matrix are dynamically configured based on task attributes and the status of verification nodes. The measurement phase occurs during the evolution time. Then, the quantum state is projected and measured, and the list of verification groups is obtained by weighted random sampling algorithm based on the probability distribution.

[0116] This embodiment, based on complete verification node states and task attributes, first constructs a quantum graph structure that accurately reflects the capability relationships of verification nodes through quantum state initialization. Then, it achieves precise matching between task requirements and node capabilities through potential energy optimization. On this basis, it realizes node selection with both randomness and optimization characteristics through quantum walk evolution and subsequent quantum state measurement. The evolution process of quantum walks relies on fundamental properties such as quantum superposition and interference. Its inherent randomness stems from the no-cloning principle and measurement collapse behavior of quantum systems, belonging to true randomness at the physical level. This randomness is fundamentally different from traditional cryptographic pseudo-random numbers generated by deterministic algorithms and are essentially predictable. Thus, at the mechanism level, it significantly improves the anti-collusion capability, professional matching degree, and overall intelligence level of the verification group assignment process.

[0117] Furthermore, this embodiment provides a step for obtaining on-chain evidence records based on a verification group list using national cryptographic algorithms and a blockchain evidence storage contract, including:

[0118] Based on the verification group list, raw carbon data and preset business rules, the group signature is obtained through the national cryptographic SM2 threshold signature algorithm and the national cryptographic SM3 hash algorithm;

[0119] On-chain evidence records are obtained through a blockchain evidence storage contract based on group signatures.

[0120] The SM2 threshold signature algorithm is a threshold signature scheme based on elliptic curve cryptography approved by the National Cryptographic Administration, which is used to share signature power among multiple verification nodes in the verification group list. The distributed key generation and signature collaboration process ensures that only when a preset threshold number of verification nodes agree can a valid signature be generated. The SM3 hash algorithm is a cryptographic hash algorithm approved by the National Cryptographic Administration, which is used to hash the original carbon data and the preset business rules to generate a fixed-length data digest. The one-way hash function ensures data integrity and tamper resistance. The group signature is a digital signature obtained by multiple verification nodes in the verification group list based on the SM2 threshold signature algorithm and the SM3 hash algorithm, which represents the consensus approval of the verification result.

[0121] Specifically, first, based on the identity of each verification node in the verification group list, the public key certificate and private key shard of the corresponding verification node are obtained through the node management service. Meanwhile, based on the original carbon data and the preset business rules, the data hash value is calculated through the SM3 hash algorithm . Then, based on the data hash value and the private key shard of each verification node, the distributed signature protocol is executed through the SM2 threshold signature algorithm, including key agreement, partial signature generation, and signature aggregation steps, to obtain the complete group signature . Subsequently, based on the group signature , the data hash value , and the verification metadata, the storage transaction is obtained through the input interface of the blockchain storage contract, and the verification metadata includes the verification task identifier and the verification node list. Finally, the storage transaction is submitted to the blockchain network by calling the storage method of the blockchain storage contract, and the tamper-proof on-chain storage record is obtained after consensus verification .

[0122] For example, this embodiment assumes that the verification group list , the original carbon data is , the preset business rules are , the data hash value is calculated through the SM3 hash algorithm , the distributed signature is executed through the SM2 threshold signature algorithm based on the nodes in the verification group list, and the group signature is finally aggregated . Subsequently, the storage transaction is uploaded to the blockchain through the blockchain storage contract, and the on-chain storage record is generated .

[0123] The embodiment realizes reliable guarantee of data integrity through the national secret SM3 hash algorithm, realizes anti-repudiation and anti-single point failure of multi-core verification node cooperative signature through the national secret SM2 threshold signature algorithm, realizes automation and transparency of the storage process through the block chain storage contract, realizes high security storage and non-tamperable traceability of the carbon data verification result through a complete technical closed loop, and improves the anti-fake ability, audit credibility and system robustness of the carbon data storage system.

[0124] Further, as shown in Figure 4 The embodiment of the application provides a carbon data storage system combining a block chain and a national secret algorithm, the carbon data storage system combining the block chain and the national secret algorithm comprising a data traceability module, a confidence decision module, a quantum consensus module and a national secret storage module.

[0125] The data traceability module: obtains original carbon data and traceability data; based on the original carbon data and the traceability data, business logic feature sets and data variation degrees are obtained through preset business rules and carbon flow traceability filtering;

[0126] The confidence decision module: based on the business logic feature sets and the data variation degrees, verification instructions are obtained through a pattern recognition confidence model, the pattern recognition confidence model comprising a feature engineering layer, a graph convolution network aggregation layer, a multi-dimensional confidence generation layer and a comprehensive decision layer;

[0127] The quantum consensus module: based on the verification instructions and the original carbon data, a verification group list is obtained through a quantum consensus model;

[0128] The national secret storage module: based on the verification group list, on-chain storage records are obtained through the national secret algorithm and the block chain storage contract.

[0129] The data tracing module, confidence decision module, quantum consensus module, and national cryptographic evidence storage module are all located on the server. The server receives data transmitted from the acquisition devices and performs further analysis. The acquisition devices include enterprise energy metering terminals, production process data collectors, data interfaces of third-party monitoring equipment, and blockchain light node gateways. The data tracing module is used to realize the whole-process trusted processing of carbon data from acquisition to feature extraction. It obtains a feature set reflecting data quality and business logic through carbon stream tracing filtering and business rule execution. The confidence decision module performs multi-dimensional credibility assessment based on the business logic feature set and data variability, through pattern recognition. The confidence model generates differentiated verification instructions; the quantum consensus module dynamically constructs anti-collusion verification groups based on the verification instructions and raw carbon data, and achieves random, fair, and professionally matched assignment of verification nodes through the quantum consensus model; the national cryptographic evidence storage module is used to cryptographically securely store the verification results of the verification groups, and achieves tamper-proof storage of verification evidence through national cryptographic algorithms and blockchain evidence storage contracts. When the blockchain evidence storage contract detects data anomalies or inconsistent verification results, the carbon data evidence storage system automatically triggers a re-verification process, sends a re-evaluation instruction to the confidence decision module, and updates the verification group list, forming a closed-loop control. The modules collaborate through standardized data interfaces and event-driven mechanisms. The output of the data traceability module serves as the input of the confidence decision module. The verification instructions generated by the confidence decision module trigger the node assignment process of the quantum consensus module. The verification group list output by the quantum consensus module is handed over to the national cryptographic evidence storage module for final evidence storage. This forms a complete closed loop from trusted data processing, intelligent decision-making, consensus assignment to secure evidence storage, and together constructs a carbon data evidence storage system with high reliability, anti-collusion, and traceability.

[0130] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0131] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams.Figure 1 apparatuses that implement the functions specified in the flow or flows and / or blocks. Figure 1 apparatuses that implement the functions specified in the flow or flows and / or blocks.

[0132] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions apparatus implementing the flow or flows and / or blocks. Figure 1 apparatuses that implement the functions specified in the flow or flows and / or blocks. Figure 1 apparatuses that implement the functions specified in the flow or flows and / or blocks.

[0133] The above description is merely that of preferred embodiments of the application, and the protection scope of the application is not limited to the above-mentioned embodiments. Any technical scheme falling within the concept of the application shall fall within the protection scope of the application. It should be noted that, for those of ordinary skill in the art, some improvements and refinements without departing from the principles of the application shall also be considered as falling within the protection scope of the application.

Claims

1. A method for storing carbon data using blockchain and national cryptographic algorithms, characterized in that, include: Obtain raw carbon data and source data; Based on the original carbon data and source tracing data, a set of business logic features and data variability are obtained through preset business rules and carbon stream source tracing filtering. Based on the business logic feature set and data variability, verification instructions are obtained through a pattern recognition confidence model. The pattern recognition confidence model includes a feature engineering layer, a graph convolutional network aggregation layer, a multidimensional confidence generation layer, and a comprehensive decision layer. Based on the verification instructions and the original carbon data, a list of verification groups is obtained through a quantum consensus model. Based on the aforementioned verification group list, on-chain evidence records are obtained through national cryptographic algorithms and blockchain evidence storage contracts; The step of obtaining a business logic feature set and data variability based on the original carbon data and source tracing data through preset business rules and carbon stream source tracing filtering includes: Based on the original carbon data and source tracing data, a first state vector is obtained by carbon flow source tracing filtering. The first state vector is the state estimation result obtained by performing state initialization, prediction and update processing on the original carbon data and source tracing data through carbon flow source tracing filtering. Based on the original carbon data and source traceability data, a second state vector and a business logic feature set are obtained through preset business rules. The second state vector is the state output obtained by logically verifying the original carbon data and source traceability data based on preset business rules. The business logic feature set is a combination of indicators that reflect the correlation between carbon emissions and production and operation. Based on the first and second state vectors, the data variability is calculated using residuals. The step of obtaining verification instructions based on the business logic feature set and data variability through a pattern recognition confidence model includes: Based on the business logic feature set and data variability, an initial feature vector is obtained through the feature engineering layer; Obtain an industry knowledge graph, and based on the industry knowledge graph and the initial feature vector, obtain a graph embedding vector through a graph convolutional network aggregation layer; Based on the initial feature vector and graph embedding vector, a confidence vector is obtained through a multi-dimensional confidence generation layer; Verification instructions are obtained through a comprehensive decision-making layer based on the confidence vector. The step of obtaining the verification group list based on the verification instructions and the original carbon data through a quantum consensus model includes: Based on the verification command, the verification node and its status are obtained through the node management service. Based on the verification instructions and the original carbon data, task attributes are obtained through task parsing; Based on the verification node status and task attributes, a verification group list is obtained through a quantum consensus model.

2. The carbon data storage method combining blockchain and national cryptographic algorithms according to claim 1, characterized in that, The process of obtaining an initial feature vector through a feature engineering layer based on the business logic feature set and data variability includes: Based on the business logic feature set and data variability, a normalized feature sequence is obtained through data standardization processing; Based on the normalized feature sequence, a composite feature group is obtained through feature cross-combination; Based on the composite feature set, an initial feature vector of standard dimensions is obtained through dimension alignment.

3. The carbon data storage method combining blockchain and national cryptographic algorithms according to claim 1, characterized in that, The process of obtaining graph embedding vectors through a graph convolutional network aggregation layer based on the industry knowledge graph and initial feature vectors includes: Based on the industry knowledge graph and initial feature vectors, the target enterprise node is determined through node localization. Based on the target enterprise node, the set of neighboring nodes of the target enterprise node is obtained through neighborhood sampling; Based on the set of neighboring nodes and the initial feature vector, a graph embedding vector is obtained through a graph convolutional neural network.

4. The carbon data storage method combining blockchain and national cryptographic algorithms according to claim 1, characterized in that, The process of obtaining a confidence vector through a multi-dimensional confidence generation layer based on the initial feature vector and graph embedding vector includes: Based on the initial feature vector and the graph embedding vector, a fused feature vector is obtained by vector concatenation; Based on the fused feature vector, a basic confidence score is obtained through a fully connected neural network; A multidimensional confidence vector is obtained based on the basic confidence score through a normalized exponential function.

5. The carbon data storage method combining blockchain and national cryptographic algorithms according to claim 1, characterized in that, The verification group list, obtained through a quantum consensus model based on the verification node status and task attributes, includes: An initial quantum graph is obtained based on the state of the verification node using a quantum state initialization algorithm; Based on the aforementioned task attributes, the quantum potential energy configuration is obtained through potential energy optimization. Based on the initial quantum graph structure and quantum potential energy configuration, a list of verification groups is obtained through a quantum consensus model.

6. The carbon data storage method combining blockchain and national cryptographic algorithms according to claim 1, characterized in that, The on-chain evidence records obtained based on the verification group list through national cryptographic algorithms and blockchain evidence storage contracts include: Based on the verification group list, raw carbon data, and preset business rules, the group signature is obtained through the national cryptographic SM2 threshold signature algorithm and the national cryptographic SM3 hash algorithm; Based on the group signature, an on-chain evidence record is obtained through a blockchain evidence storage contract.

7. A carbon data storage system combining blockchain and national cryptographic algorithms, used to implement the carbon data storage method combining blockchain and national cryptographic algorithms as described in any one of claims 1-6, characterized in that, The carbon data storage system combining blockchain and national cryptographic algorithms includes: The data tracing module acquires raw carbon data and tracing data; based on the raw carbon data and tracing data, it obtains a business logic feature set and data variability through preset business rules and carbon stream tracing filtering. The confidence decision module obtains verification instructions based on the business logic feature set and data variability through a pattern recognition confidence model. The pattern recognition confidence model includes a feature engineering layer, a graph convolutional network aggregation layer, a multi-dimensional confidence generation layer, and a comprehensive decision layer. The quantum consensus module, based on the verification instructions and the original carbon data, obtains a list of verification groups through a quantum consensus model. The national cryptographic evidence storage module obtains on-chain evidence records based on the verification group list using national cryptographic algorithms and blockchain evidence storage contracts.

Images