Carbon asset management method and system based on knowledge graph
Through knowledge graph technology, the processing of diverse and heterogeneous carbon emission data has been solved, and the problems of low data quality and low prediction accuracy in the existing technology have been achieved, efficient carbon emission management and dynamic benchmark adjustments have been achieved, and scientific decision-making support has been provided.
Patent Information
- Application Number
- CN202411120085.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-15
- Publication Date
- 2024-12-03
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing carbon asset management methods rely on manual statistics and simple linear computing models, making it difficult to process diverse and heterogeneous data, resulting in low data quality, low prediction accuracy, inaccurate benchmarks, and undynamic adjustments.
Using a knowledge graph-based method, through entity recognition, attribute extraction, relationship mining and redundant conflict resolution algorithms, a redundant knowledge graph is formed, a carbon emission calculation model is established, and a carbon emission benchmark is dynamically adjusted.
It improves the unified management and utilization efficiency of carbon emission data, improves the prediction accuracy and accuracy of benchmarks, and can make dynamic adjustments based on historical data and policy changes, providing scientific decision-making support.
Smart Images

Figure CN119067254A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of carbon asset management, and specifically relates to a carbon asset management method and system based on knowledge graph. Background Art
[0002] At present, carbon emission management has become a focus of attention for companies with large-scale carbon emissions. Existing carbon asset management methods mainly rely on manual statistics and simple calculation models. These methods have many defects and problems. Since data sources are diverse and heterogeneous, including text, sensor data, financial statements, etc., the formats are diverse and heterogeneous, making it difficult to process them in a unified manner; data from different sources may be redundant and conflicting, and traditional methods are difficult to process effectively, resulting in low data quality; in addition, existing models are mostly linear models, which cannot accurately reflect the complex carbon emission process and have low prediction accuracy. Existing methods are mostly static benchmarks and cannot be dynamically adjusted according to historical data and policy changes, resulting in inaccurate benchmarks.
[0003] Therefore, a carbon asset management approach is needed to improve the assessment and forecasting of corporate carbon emissions and enhance the accuracy and adaptability of benchmarks. Summary of the invention
[0004] The purpose of the present invention is to provide a carbon asset management method and system based on knowledge graph, improve the accuracy of data processing and carbon emission prediction, dynamically adjust the carbon emission benchmark, and provide scientific decision-making support for enterprises.
[0005] In a first aspect, the present invention provides a carbon asset management method based on a knowledge graph, the method comprising the following steps:
[0006] Step S1, collecting basic data related to corporate carbon emissions, extracting knowledge from the basic data, extracting structured information including entities, attributes and relationships between entities, and forming a preliminary knowledge base.
[0007] Step S2, integrate the knowledge from different sources and formats in the preliminary knowledge base, use redundancy elimination and conflict resolution algorithms to integrate the knowledge, and form a non-redundant knowledge graph.
[0008] Step S3, establish a carbon emission calculation model, use the data in the knowledge graph to evaluate and predict the company's carbon emissions, and make a carbon emission benchmark determination.
[0009] Furthermore, knowledge extraction from the basic data includes four steps: entity recognition, attribute extraction, relationship mining and knowledge base establishment, specifically:
[0010] Step S11, identifying entities related to carbon emissions from text and unstructured data.
[0011] Step S12: Using the attribute value embedding method, the attributes associated with each entity are extracted and quantified.
[0012] Step S13, extracting the relationship between entities from the data through the RNA algorithm to form entity-relationship triples.
[0013] Step S14, using a graph database to store the identified entities, attributes and relationships to form a preliminary knowledge base.
[0014] Furthermore, the specific steps of using redundancy elimination and conflict resolution algorithms to perform knowledge fusion include:
[0015] Step S21, using a heterogeneous data integration model to standardize data from different sources and formats.
[0016] The standardization processing formula is: Among them, x is the data value, μ is the mean, σ is the standard deviation, and x′ is the data value after standardization.
[0017] Step S22, eliminating redundant data by jointly applying a multi-layer perceptron MLP and a conditional random field CRF model.
[0018] When MLP and CRF are jointly applied for redundancy elimination, features are first extracted through MLP.
[0019] The linear transformation from the input layer to the hidden layer of the MLP model is expressed as: h = Sigmoid (W1x′+b1); where x′ is the data value after normalization as the input vector; W1 is the weight matrix from the input layer to the hidden layer; b1 is the bias vector of the hidden layer; Sigmoid is the activation function; and h is the output of the hidden layer.
[0020] The linear transformation from the hidden layer to the output layer is expressed as: MLP =ReLU(W2h+b2); where W2 is the weight matrix from the hidden layer to the output layer; b2 is the bias vector of the output layer; ReLU is the activation function; y MLP is the output of the output layer.
[0021] The output y of the output layer MLP As the input of CRF, sequence labeling and redundancy elimination are performed.
[0022] The conditional probability distribution of the CRF model is expressed as:
[0023]
[0024] Among them, P(y|y MLP ) is the given input y MLPThe probability of outputting y under the condition; Z(y MLP ) is a normalization factor to ensure that the probability sum is 1; λ k is the weight of the feature function; f k (y,y MLP ) is the characteristic function.
[0025] By maximizing the conditional probability P(y|y MLP ) to obtain the optimal labeling sequence y * :
[0026] Step S23, using game theory models and Bayesian networks to resolve conflicts and ensure the consistency of the knowledge graph.
[0027] Furthermore, the establishment of a carbon emission calculation model and the use of data in the knowledge graph to evaluate and predict the carbon emissions of an enterprise include:
[0028] A carbon emission calculation model is established to calculate the carbon footprint. The formula is as follows:
[0029]
[0030] Among them, E d is the direct emissions of the enterprise in time t, A i,j (t) is the value of the activity factor of the i-th activity under the j-th attribute at time t, EF i,j is the emission factor, C k (t) is the consumption of the kth category at time t.
[0031] Separate calculation of direct and indirect emissions using matrix decomposition: E d =U·V T +∈; U and V are low-dimensional latent variable matrices, ∈ is the error term; then the indirect emission E f =E(t)-E d ; Where E(t) is the total carbon emissions of the enterprise in time t.
[0032] Furthermore, the total carbon emissions in each time period are calculated based on the total emissions to form a feature set X, and the nonlinear equation of total carbon emissions and time is obtained through nonlinear regression, E(t)=f(X,Θ)+η; wherein E(t) is the total carbon emissions, f is the nonlinear function, Θ is the model parameter, and η is the random error term; then, according to the nonlinear equation of total carbon emissions and time, the predicted value of total carbon emissions E(t+1) at the next moment can be obtained.
[0033] Furthermore, the carbon emission benchmark determination step includes:
[0034] Combine historical data and existing policies to adjust the carbon emission benchmark and set a dynamic benchmark. The formula is as follows:
[0035]
[0036] Among them, P(RID,H) is the baseline probability distribution under the data set D and historical information H, and R is the carbon emission baseline value; P(D|H) is the probability distribution of the data set D under the historical information H, and P(R|H) is the prior probability distribution of the carbon emission baseline value R under the historical information H.
[0037] The carbon emission baseline value is continuously revised through the Bayesian update method:
[0038]
[0039] Among them, P(R new |D, H) is the updated benchmark probability distribution under the data set D and historical information H, R new This is the updated carbon emission baseline value.
[0040] Furthermore, a multi-criteria decision analysis model is used to compare the actual carbon emissions of enterprises with industry standards and policies and regulations. The actual carbon emissions of enterprises are calculated as follows:
[0041]
[0042] Among them, S is the actual carbon emissions of the enterprise, w k is the weight of the kth criterion, f k is the criterion function, E i is the carbon emissions of the i-th activity of the enterprise, B i is the benchmark value for the i-th type of activity.
[0043] Compare the actual carbon emissions of the enterprise with the industry standard or policy and regulatory value S0. If S<S0, the enterprise's carbon emissions meet the standard, otherwise adjustments and improvements are needed.
[0044] In a second aspect, the present invention provides a carbon asset management system based on a knowledge graph, which is used in the method described in the first aspect, and the system includes a data extraction module, a data fusion module and a carbon emission calculation module connected in sequence.
[0045] Furthermore, the data extraction module is used to collect basic data related to corporate carbon emissions, extract knowledge from the basic data, extract structured information including entities, attributes and relationships between entities, and form a preliminary knowledge base.
[0046] Furthermore, the data fusion module is used to fuse knowledge from different sources and in different formats in the preliminary knowledge base, and use redundancy elimination and conflict resolution algorithms to fuse knowledge to form a knowledge graph without redundancy.
[0047] Furthermore, the carbon emission calculation module is used to establish a carbon emission calculation model, use the data in the knowledge graph to evaluate and predict the carbon emissions of the enterprise, and make carbon emission benchmark determinations.
[0048] Furthermore, the data extraction module includes: an entity recognition unit, an attribute extraction unit, a relationship mining unit and a knowledge base establishment unit.
[0049] The entity recognition unit is used to recognize entities related to carbon emissions from text and unstructured data.
[0050] The attribute extraction unit is used to extract and quantify the attributes associated with each entity by adopting an attribute value embedding method.
[0051] The relationship mining unit is used to extract the relationship between entities from the data through the RNA algorithm to form entity-relationship triples.
[0052] The knowledge base building unit is used to use a graph database to store the identified entities, attributes and relationships to form a preliminary knowledge base.
[0053] Furthermore, the data fusion module includes: a standardization unit, a redundancy elimination unit and a conflict resolution unit.
[0054] The standardization unit adopts a heterogeneous data integration model to standardize data from different sources and formats.
[0055] The redundant data elimination unit eliminates redundant data by jointly applying a multi-layer perceptron MLP and a conditional random field CRF model.
[0056] The conflict resolution unit uses game theory models and Bayesian networks to resolve conflicts and ensure the consistency of the knowledge graph.
[0057] The beneficial effects of the present invention are as follows:
[0058] The present invention realizes the unified management and efficient use of complex and heterogeneous carbon emission data by constructing a carbon asset management method based on knowledge graph; the introduction of knowledge graph makes the correlation and logic between data clearer, facilitates rapid retrieval and analysis, thereby improving the efficiency and accuracy of carbon asset management; helps enterprises understand their own carbon emission status, and can also provide a scientific basis for formulating emission reduction plans and strategies. BRIEF DESCRIPTION OF THE DRAWINGS
[0059] Figure 1The carbon asset management method process based on knowledge graph of the present invention is Figure 2
[0060] Figure 2 This is a block diagram of the knowledge graph-based carbon asset management system of the present invention. DETAILED DESCRIPTION
[0061] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention is described clearly and completely below. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0062] Example 1
[0063] like Figure 1 The following is a flow chart of the agricultural and food product quality control method based on international standards. The following are the detailed steps:
[0064] Step S1, collecting basic data related to corporate carbon emissions, extracting knowledge from the basic data, extracting structured information including entities, attributes and relationships between entities, and forming a preliminary knowledge base.
[0065] In step S1, knowledge extraction from the basic data includes four steps: entity identification, attribute extraction, relationship mining and knowledge base establishment. The basic data at least includes: enterprise production and operation data, energy consumption data, carbon emission data, industry standard data, policy and regulation data; entity (enterprise, facility, equipment), attribute (energy consumption, production volume), and the specific steps are:
[0066] Step S11, identifying entities related to carbon emissions from text and unstructured data; for example, using hybrid natural language processing (Hybrid N LP) and deep learning algorithms.
[0067] Step S12, using an attribute value embedding method to extract and quantify the attributes associated with each entity; for example, using an embedded vector representation technology to determine the semantic representation of the attribute value.
[0068] Step S13, extracting the relationship between entities from the data through RNA (Relation Network Analysis, RNA) algorithm to form entity-relationship triples.
[0069] Step S14, using a graph database (such as GraphDB) to store the identified entities, attributes and relationships to form a preliminary knowledge base.
[0070] Step S2, integrate the knowledge from different sources and formats in the preliminary knowledge base, use redundancy elimination and conflict resolution algorithms to integrate the knowledge, and form a non-redundant knowledge graph.
[0071] In step S2, the specific steps of using redundancy elimination and conflict resolution algorithms to perform knowledge fusion include:
[0072] In step S21, a heterogeneous data integration model (HDI M) is used to standardize data from different sources and formats.
[0073] The standardization processing formula is: Among them, x is the data value, μ is the mean, σ is the standard deviation, and x′ is the data value after standardization.
[0074] Step S22, eliminating redundant data by jointly applying a multi-layer perceptron MLP and a conditional random field CRF model.
[0075] When MLP and CRF are jointly applied for redundancy elimination, features are first extracted through MLP.
[0076] The linear transformation from the input layer to the hidden layer of the MLP model is expressed as: h = Sigmoid (W1x′+b1); where x′ is the data value after normalization as the input vector; W1 is the weight matrix from the input layer to the hidden layer; b1 is the bias vector of the hidden layer; Sigmoid is the activation function; and h is the output of the hidden layer.
[0077] The linear transformation from the hidden layer to the output layer is expressed as: MLP =ReLU(W2h+b2); where W2 is the weight matrix from the hidden layer to the output layer; b2 is the bias vector of the output layer; ReLU is the activation function; y MLP is the output of the output layer.
[0078] The output y of the output layer MLP As the input of CRF, sequence labeling and redundancy elimination are performed.
[0079] The conditional probability distribution of the CRF model is expressed as:
[0080]
[0081] Among them, P(y|y MLP ) is the given input y MLP The probability of outputting y under the condition; Z(y MLP ) is a normalization factor to ensure that the probability sum is 1; λ kis the weight of the feature function; f k (y,y MLP ) is the characteristic function.
[0082] By maximizing the conditional probability P(y|y MLP ) to obtain the optimal labeling sequence y * :
[0083] Step S23, using game theory models and Bayesian networks to resolve conflicts and ensure the consistency of the knowledge graph.
[0084] When using the game theory model, we consider multiple information sources as participants, and these participants play games on the correctness of information in a knowledge graph; the basic elements of the game theory model include: participant set: N = {1, 2, ..., n}, representing the information source; strategy set: each participant i has a strategy set Si, which represents the judgment of the information source on a certain knowledge item; profit function: Ui: S1×S2×…×Sn→R, which represents the profit of participant i when adopting a certain strategy combination; Nash equilibrium: we seek a strategy combination (s1*, s2*, ..., sn*) so that the strategy of each information source is optimal, that is: Ui(si*, si*)≥Ui(si, si*), Among them, si* represents the strategy combination of other information sources except i.
[0085] A Bayesian network is a directed acyclic graph (DAG) that expresses conditional dependencies between random variables. In conflict resolution, we can use Bayesian networks to quantify dependencies and uncertainties between information. The basic elements of a Bayesian network include: a set of nodes representing random variables; directed edges representing dependencies between variables; and a conditional probability table (CPT) representing the conditional probability of a node taking a value given its parent node. When the state of a node in a Bayesian network changes, it can be updated using Bayes' theorem: P(A|B)=P(B|A)P(A)P(B).
[0086] The combined use of game theory models and Bayesian networks can more effectively resolve conflicts and ensure the consistency of the knowledge graph; specifically, the following steps are involved:
[0087] Define participants and strategy sets. Each information source is a participant in the game theory model and has its own strategy set. Establish a Bayesian network, use information sources and information items as nodes of the Bayesian network, and define their dependencies and conditional probability tables.
[0088] Calculate the profit function. For each information source, calculate the profit under different strategy combinations. The definition of the profit function is: Ui(s1,s2,…,sn)=∑jP(Kj|s1,s2,…,sn)·Vi(si,Kj) Where, P(Kj|s1,s2,…,sn) is the probability that the knowledge item is Kj under the strategy combination, calculated by Bayesian network; Vi(si,Kj) is the value of information source i when a certain knowledge item is Kj under strategy si;
[0089] Solve Nash equilibrium and find the strategy combination that maximizes the benefits of each information source; use Bayesian network to update information, and after obtaining the optimal strategy combination, use Bayesian theorem to update the node status in the Bayesian network; Bayesian network update rule: P(A=a|B=b)=P(B=b|A=a)·P(A=a)P(B=b);
[0090] Combining Nash equilibrium and Bayesian networks for conflict resolution:
[0091] arg max(s1,s2,…,sn)∑Ui(s1,s2,…,sn)P(Kj|s1=s1*,s2*,...,sn*)=P(s1=s1*,s2=s2*,...,sn=sn*|Kj)·P(Kj)P(s1=s1*,s2=s2*,...,sn=sn*); Through the above steps, using game theory models and Bayesian networks, information conflicts can be effectively resolved and the consistency of the knowledge graph can be ensured.
[0092] Step S3, establish a carbon emission calculation model, use the data in the knowledge graph to evaluate and predict the company's carbon emissions, and make a carbon emission benchmark determination.
[0093] The establishment of a carbon emission calculation model and the use of data in the knowledge graph to evaluate and predict the carbon emissions of an enterprise include:
[0094] A carbon emission calculation model is established to calculate the carbon footprint. The formula is as follows:
[0095]
[0096] Among them, E d is the direct emissions of the enterprise in time t, A i,j (t) is the value of the activity factor of the i-th activity under the j-th attribute at time t, EF i,j is the emission factor, C k (t) is the consumption of the kth category at time t.
[0097] Separate calculation of direct and indirect emissions using matrix decomposition: E d =U·V T+∈; U and V are low-dimensional latent variable matrices, ∈ is the error term; then the indirect emission E f =E(t)-E d ; Where E(t) is the total carbon emissions of the enterprise in time t.
[0098] Based on the total emissions, the total carbon emissions in each time period are calculated to form a feature set X. The nonlinear equation of total carbon emissions and time is obtained through nonlinear regression, E(t)=f(X,Θ)+η; where E(t) is the total carbon emissions, f is the nonlinear function, Θ is the model parameter, and η is the random error term; then the predicted value of the total carbon emissions E(t+1) at the next moment can be obtained according to the nonlinear equation of total carbon emissions and time.
[0099] Here is a specific example of a corporate carbon emissions calculation:
[0100] Enterprise activity factor matrix A i,j (t):
[0101] Emission factor matrix EF i,j :
[0102] Consumption Matrix C k (t):
[0103] Emission Factor EF k :
[0104] Calculate direct emissions E d =(100·0.5+50·0.3+30·0.2)+(60·0.4+40·0.2+20·0.1)+(40·0.3+30·0.1+10·0.05)+(200·0.6+150·0.4+100·0.2)=330.5 tons of CO2.
[0105] Calculate indirect emissions Ef, which is the total emissions E(t) minus direct emissions Ed. If the specific value of total emissions E(t) = 661 tons of CO2, then E f =E(t)-E d =661-330.5=330.5 tons of CO2.
[0106] Use matrix decomposition to calculate the latent variable matrices U and V,
[0107] V T=[0.5 0.3 0.2]; ∈=0
[0108] Nonlinear function f(X, Θ) and model parameters Θ, E(t) = f(X, Θ) + η = a·X + b + η, where X is the feature set and a and b are model parameters.
[0109] Assume X is the feature at time t, a = 0.1, b = 100, E(t) = 0.1·t+100+η;
[0110] Assuming that E(1)=661 when t=1, then: 661=0.1·1+100+η; η=661-100-0.1=561.9.
[0111] Therefore, the predicted total carbon emissions E(t+1) at the next moment t+1 are: E(t+1)=0.1·(t+1)+100+561.9; E(t+1)=0.1·(t+1)+661.9.
[0112] The carbon emission benchmark determination step includes:
[0113] Combine historical data and existing policies to adjust the carbon emission benchmark and set a dynamic benchmark. The formula is as follows:
[0114]
[0115] Among them, P(RID,H) is the baseline probability distribution under the data set D and historical information H, and R is the carbon emission baseline value; P(D|H) is the probability distribution of the data set D under the historical information H, and P(R|H) is the prior probability distribution of the carbon emission baseline value R under the historical information H.
[0116] The carbon emission baseline value is continuously revised through the Bayesian update method:
[0117]
[0118] Among them, P(R new |D,H) is the updated baseline probability distribution under the combination of data set D and historical information H, R new is the updated carbon emission baseline value.
[0119] A multi-criteria decision analysis model is used to compare the actual carbon emissions of enterprises with industry standards and policies and regulations. The actual carbon emissions of enterprises are calculated as follows:
[0120]
[0121] Among them, S is the actual carbon emissions of the enterprise, w k is the weight of the kth criterion, f kis the criterion function, E i is the carbon emissions of the i-th activity of the enterprise, B i is the benchmark value for the i-th type of activity.
[0122] Compare the actual carbon emissions of the enterprise with the industry standard or policy and regulatory value S0. If S<S0, the enterprise's carbon emissions meet the standard, otherwise adjustments and improvements are needed.
[0123] Example 2
[0124] like Figure 2 The carbon asset management system based on knowledge graph is used to execute the method of Example 1. The system includes a data extraction module, a data fusion module and a carbon emission calculation module connected in sequence.
[0125] The data extraction module is used to collect basic data related to corporate carbon emissions, extract knowledge from the basic data, extract structured information including entities, attributes and relationships between entities, and form a preliminary knowledge base.
[0126] The data fusion module is used to integrate knowledge from different sources and formats in the preliminary knowledge base, and adopts redundancy elimination and conflict resolution algorithms to integrate knowledge and form a knowledge graph without redundancy.
[0127] The carbon emission calculation module is used to establish a carbon emission calculation model, use the data in the knowledge graph to evaluate and predict the carbon emissions of an enterprise, and make carbon emission benchmark determinations.
[0128] The data extraction module includes: an entity recognition unit, an attribute extraction unit, a relationship mining unit and a knowledge base establishment unit; the entity recognition unit is used to identify entities related to carbon emissions from text and unstructured data; the attribute extraction unit is used to extract and quantify the attributes associated with each entity using an attribute value embedding method; the relationship mining unit is used to extract the relationship between entities from the data through an RNA algorithm to form entity-relationship triples; the knowledge base establishment unit is used to use a graph database to store the identified entities, attributes and relationships to form a preliminary knowledge base.
[0129] The data fusion module includes: a standardization unit, a redundancy elimination unit and a conflict resolution unit; the standardization unit adopts a heterogeneous data integration model to standardize data from different sources and formats; the redundancy elimination unit eliminates redundant data by jointly applying a multi-layer perceptron MLP and a conditional random field CRF model; the conflict resolution unit uses a game theory model and a Bayesian network to resolve conflicts and ensure the consistency of the knowledge graph.
[0130] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. A carbon asset management method based on knowledge graph, characterized in that: The method comprises the following steps: Step S1, collecting basic data related to corporate carbon emissions, extracting knowledge from the basic data, extracting structured information including entities, attributes and relationships between entities, and forming a preliminary knowledge base; Step S2, integrating knowledge from different sources and formats in the preliminary knowledge base, using redundancy elimination and conflict resolution algorithms to integrate knowledge and form a non-redundant knowledge graph; Step S3, establish a carbon emission calculation model, use the data in the knowledge graph to evaluate and predict the company's carbon emissions, and make a carbon emission benchmark determination.
2. The carbon asset management method based on knowledge graph according to claim 1 is characterized in that: The knowledge extraction from the basic data includes four steps: entity recognition, attribute extraction, relationship mining and knowledge base establishment, specifically: Step S11, identifying entities related to carbon emissions from text and unstructured data; Step S12, using attribute value embedding method to extract and quantify the attributes associated with each entity; Step S13, extracting the relationship between entities from the data by RNA algorithm to form entity-relationship triples; Step S14, using a graph database to store the identified entities, attributes and relationships to form a preliminary knowledge base.
3. The carbon asset management method based on knowledge graph according to claim 1 is characterized in that: The specific steps of using redundancy elimination and conflict resolution algorithms to perform knowledge fusion include: Step S21, using a heterogeneous data integration model to standardize data from different sources and formats; The standardization processing formula is: Where x is the data value, μ is the mean, σ is the standard deviation, and x′ is the data value after standardization; Step S22, eliminating redundant data by jointly applying a multi-layer perceptron MLP and a conditional random field CRF model; When MLP and CRF are jointly applied for redundancy elimination, features are first extracted through MLP; The linear transformation from the input layer to the hidden layer of the MLP model is expressed as: h = Sigmoid (W1x′ + b1); where x′ is the data value after normalization, as the input vector; W1 is the weight matrix from the input layer to the hidden layer; b1 is the bias vector of the hidden layer; Sigmoid is the activation function; h is the output of the hidden layer; The linear transformation from the hidden layer to the output layer is expressed as: MLP =ReLU(W2h+b2); where W2 is the weight matrix from the hidden layer to the output layer; b2 is the bias vector of the output layer; ReLU is the activation function; y MLP is the output of the output layer; The output y of the output layer MLP As the input of CRF, sequence labeling and redundancy elimination are performed; The conditional probability distribution of the CRF model is expressed as: Among them, P(y|y MLP ) is the given input y MLP The probability of outputting y under the condition; Z(y MLP ) is a normalization factor to ensure that the probability sum is 1; λ k is the weight of the feature function; f k (y,y MLP ) is the characteristic function; By maximizing the conditional probability P(y|y MLP ) to obtain the optimal labeling sequence y * : Step S23, using game theory models and Bayesian networks to resolve conflicts and ensure the consistency of the knowledge graph.
4. The carbon asset management method based on knowledge graph according to claim 3 is characterized in that: The establishment of a carbon emission calculation model and the use of data in the knowledge graph to evaluate and predict the carbon emissions of an enterprise include: A carbon emission calculation model is established to calculate the carbon footprint. The formula is as follows: Among them, E d is the direct emissions of the enterprise in time t, A i,j (t) is the value of the activity factor of the i-th activity under the j-th attribute at time t, EF i,j is the emission factor, C k (t) is the consumption of the kth category at time t. Separate calculation of direct and indirect emissions using matrix decomposition: E d =U·V T +∈; U and V are low-dimensional latent variable matrices, ∈ is the error term; Indirect emissions E f =E(t)-E d ; Where E(t) is the total carbon emissions of the enterprise in time t.
5. The carbon asset management method based on knowledge graph according to claim 4 is characterized in that: Based on the total emissions, the total carbon emissions in each time period are calculated to form a feature set X. The nonlinear equation of total carbon emissions and time is obtained through nonlinear regression, E(t)=f(X,Θ)+η; where E(t) is the total carbon emissions, f is the nonlinear function, Θ is the model parameter, and η is the random error term; then the predicted value of the total carbon emissions E(t+1) at the next moment can be obtained according to the nonlinear equation of total carbon emissions and time.
6. The carbon asset management method based on knowledge graph according to claim 4 is characterized in that: The carbon emission benchmark determination step includes: Combine historical data and existing policies to adjust the carbon emission benchmark and set a dynamic benchmark. The formula is as follows: Among them, P(R|D, H) is the baseline probability distribution under the data set D and historical information H, where R is the carbon emission baseline value; P(D|H) is the probability distribution of the data set D under the historical information H, and P(R|H) is the prior probability distribution of the carbon emission baseline value R under the historical information H; The carbon emission baseline value is continuously corrected through the Bayesian update method: Among them, P(R new |D, H) is the updated benchmark probability distribution under the data set D and historical information H, R new This is the updated carbon emission baseline value.
7. The carbon asset management method based on knowledge graph according to claim 5 is characterized in that: A multi-criteria decision analysis model is used to compare the actual carbon emissions of an enterprise with industry standards and policies and regulations. The calculation formula for the actual carbon emissions of the enterprise is: Among them, S is the actual carbon emissions of the enterprise, w k is the weight of the kth criterion, f k is the criterion function, E i is the carbon emissions of the i-th activity of the enterprise, B i is the benchmark value for the i-th type of activity; Compare the actual carbon emissions of the enterprise with the industry standard or policy and regulation value S0. If S < S0, the carbon emissions of the enterprise meet the standard; otherwise, adjustments and improvements are required.
8. A carbon asset management system based on knowledge graph, used to execute the method according to any one of claims 1 to 7, characterized in that: The system includes a data extraction module, a data fusion module, and a carbon emission calculation module that are connected in sequence; The data extraction module is used to collect basic data related to the carbon emissions of an enterprise, perform knowledge extraction from the basic data, extract structured information including entities, attributes, and relationships between entities, and form a preliminary knowledge base; The data fusion module is used to fuse knowledge from different sources and in different formats in the preliminary knowledge base, and perform knowledge fusion using redundancy elimination and conflict resolution algorithms to form a non-redundant knowledge graph; The carbon emission calculation module is used to establish a carbon emission calculation model, use the data in the knowledge graph to evaluate and predict the carbon emissions of the enterprise, and perform carbon emission baseline determination.
9. The carbon asset management system based on knowledge graph according to claim 8 is characterized in that: The data extraction module includes: an entity recognition unit, an attribute extraction unit, a relationship mining unit, and a knowledge base establishment unit; The entity recognition unit is used to identify entities related to carbon emissions from text and unstructured data; The attribute extraction unit is used to extract and quantify the attributes associated with each entity using the attribute value embedding method; The relationship mining unit is used to extract the relationships between entities from the data through the RNA algorithm to form entity-relationship triples; The knowledge base establishment unit is used to store the identified entities, attributes, and relationships using a graph database to form a preliminary knowledge base.
10. The carbon asset management method based on knowledge graph according to claim 9 is characterized in that: The data fusion module includes: a standardization unit, a redundancy elimination unit, and a conflict resolution unit; The standardization unit uses a heterogeneous data integration model to standardize data from different sources and formats; The redundancy elimination unit jointly applies a multi-layer perceptron MLP and a conditional random field CRF model to eliminate redundant data; The conflict resolution unit uses a game theory model and a Bayesian network to resolve conflicts to ensure the consistency of the knowledge graph.
Citation Information
Patent Citations
Carbon asset trading method and system
CN109272405A
Carbon policy knowledge graph construction system
CN117371531A
Carbon policy knowledge graph construction system based on text analysis
CN117407535A
Building block type product carbon footprint intelligent analysis system and method based on knowledge graph
CN117591679A
Cited By
Intelligent tumor marker data processing system
CN119763806A
Enterprise carbon asset digitalization method and device based on multi-modal knowledge graph
CN120012905A
Enterprise data optimization processing method based on cloud computing data fusion
CN120068146A
A method for optimizing enterprise data processing based on cloud computing data fusion
CN120068146B
Engineering project carbon emission prediction method based on knowledge graph
CN120197786A
