Knowledge graph construction method for complex electricity-carbon market coupling relationship

Abstract

The invention discloses a knowledge graph construction method for a complex electricity-carbon market coupling relationship, and belongs to the field of electricity-carbon coupling. The method comprises the following steps: constructing an electricity-carbon market coupling relation model; constructing a named entity recognition model based on a pointer network and multi-task learning; constructing a multi-head selection relation extraction model based on global entity relative position representation; and constructing a knowledge graph oriented to the complex electricity-carbon market coupling relationship. The method can effectively reflect a complex electricity and carbon market coupling relationship, reveal market dynamics, effectively guide market strategy formulation and resource allocation, and promote collaborative development of electricity and carbon emission markets.

Description

Technical Field

[0001] The present invention relates to a method for constructing a knowledge graph for complex electricity-carbon market coupling relationships, belonging to the field of electricity-carbon coupling. Background Art

[0002] With the increasing global attention to climate change, the collaborative optimization of carbon emission management and electricity markets has become a research hotspot. Electricity-carbon coupling promotes the green and low-carbon development of the whole society by means of the close connection between electricity and carbon emissions, focusing on themes such as "carbon calculation", "carbon reduction", and "carbon governance". As a structured knowledge representation method, a knowledge graph organizes data in the form of triples (subject, predicate, object), which can clearly display the relationships and hierarchical structures between data. In the field of electricity-carbon coupling, constructing a knowledge graph can integrate, correlate, and analyze electricity consumption and carbon emission data, providing strong support for policy-making, market trading, and academic research. Therefore, the method for constructing a knowledge graph for complex electricity-carbon market coupling relationships is one of the important technical means to address climate change and promote green and low-carbon development. Summary of the Invention

[0003] The present invention provides a method for constructing a knowledge graph for complex electricity-carbon market coupling relationships, which is used to accurately capture the coupling mechanism between complex electricity-carbon markets and reveal the internal evolution law of the markets.

[0004] The technical solution of the present invention is as follows:

[0005] According to the first aspect of the present invention, there is provided a method for constructing a knowledge graph for complex electricity-carbon market coupling relationships, including: constructing an electricity-carbon market coupling relationship model; constructing a named entity recognition model based on a pointer network and multi-task learning; constructing a multi-head selection relationship extraction model based on global entity relative position representation; constructing a knowledge graph for complex electricity-carbon market coupling relationships.

[0006] The construction of the electricity-carbon market coupling relationship model includes: analyzing the complex interactions and coupling phenomena caused by common participants in the electricity market and the carbon market from three aspects: market transaction quantity, market transaction price, and market operation time; constructing an electricity-carbon market coupling relationship model including electricity-carbon market quantity coupling relationship, electricity-carbon market price linkage effect, and electricity-carbon market time-scale coupling.

[0007] The construction of the named entity recognition model based on a pointer network and multi-task learning includes: based on a multi-task learning strategy and a pointer network mechanism, with the help of the HowNet knowledge base, constructing a named entity recognition model composed of a shared feature extraction layer, a multi-task learning layer, and an output layer, decomposing the traditional named entity recognition task into two subtasks: entity boundary recognition and entity classification, and accurately identifying electricity-carbon market entities and their categories.

[0008] The constructed multi-head selection relation extraction model based on the relative position representation of global entities includes: proposing a representation method for the relative position of global entities based on BERT and CRF technologies, combining it with the multi-head selection framework, and introducing a global relation classification strategy at the same time. Through joint optimization, entity relations are accurately extracted; the model consists of four parts: a BERT feature extraction layer, a CRF layer, a multi-head selection layer, and global relation classification.

[0009] The constructed knowledge graph for the coupling relationship of the complex electricity-carbon market is specifically: adopting a hierarchical architecture, consisting of four layers: a visualization layer, a service layer, a data access layer, and a data storage layer. The constructed knowledge graph includes a data coupling module, an entity recognition module, a relation extraction module, and a data display module.

[0010] According to the second aspect of the present invention, a processor is provided. The processor is used to run a program, wherein when the program runs, it executes the method for constructing the knowledge graph for the coupling relationship of the complex electricity-carbon market described in any one of the above.

[0011] According to the third aspect of the present invention, a computer-readable storage medium is provided. The computer-readable storage medium includes a stored program, wherein when the program runs, it controls the device where the computer-readable storage medium is located to execute the method for constructing the knowledge graph for the coupling relationship of the complex electricity-carbon market described in any one of the above.

[0012] The beneficial effects of the present invention are:

[0013] The coupling relationship model of the electricity-carbon market of the present invention couples the electricity-carbon relationship from three perspectives: market transaction quantity, price, and operation time, and can insight into the interaction and influence mechanism between the two markets; the named entity recognition model based on multi-task learning and pointer network decomposes named entity recognition into two subtasks: boundary recognition and classification, and accurately recognizes the entities and their categories in the electricity-carbon market; the multi-head selection relation extraction model based on the relative position representation of global entities can accurately extract entity relations; finally, the construction of the knowledge graph for the coupling relationship of the complex electricity-carbon market can intuitively reveal the complex connections between various entities in the electricity-carbon market, assist market participants in decision-making, and promote the sustainable development of the electricity-carbon market. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 is a block diagram of a method for constructing a knowledge graph for the coupling relationship of a complex electricity-carbon market according to the present invention;

[0015] Figure 2 is a functional module diagram of the knowledge graph of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0016] The present invention will be further described below in conjunction with the accompanying drawings and embodiments, but the content of the present invention is not limited to the described scope.

[0017] Embodiment 1: As Figure 1 shown, the present invention provides a method for constructing a knowledge graph for complex electricity-carbon market coupling relationships, including: constructing an electricity-carbon market coupling relationship model; constructing a named entity recognition model based on pointer networks and multi-task learning; constructing a multi-head selection relationship extraction model based on global entity relative position representation; and constructing a knowledge graph for complex electricity-carbon market coupling relationships.

[0018] Further, the construction of the electricity-carbon market coupling relationship model includes: analyzing the complex interactions and coupling phenomena caused by common participants in the electricity market and the carbon market from three aspects of market transaction quantity, market transaction price, and market operation time; constructing an electricity-carbon market coupling relationship model including electricity-carbon market quantity coupling relationship, electricity-carbon market price linkage effect, and electricity-carbon market time scale coupling.

[0019] Further, the construction of the named entity recognition model based on pointer networks and multi-task learning includes: constructing a named entity recognition model composed of a shared feature extraction layer, a multi-task learning layer, and an output layer based on multi-task learning strategies and pointer network mechanisms, with the help of the HowNet knowledge base, and decomposing the traditional named entity recognition task into two sub-tasks of entity boundary recognition and entity classification to accurately identify electricity-carbon market entities and their categories.

[0020] Further, the construction of the multi-head selection relationship extraction model based on global entity relative position representation includes: proposing a representation method for the relative position of global entities based on BERT and CRF technologies, combining it with a multi-head selection framework, and introducing a global relationship classification strategy at the same time, and accurately extracting entity relationships through joint optimization; the model consists of four parts: a BERT feature extraction layer, a CRF layer, a multi-head selection layer, and global relationship classification.

[0021] Further, the construction of the knowledge graph for complex electricity-carbon market coupling relationships is specifically: adopting a hierarchical architecture method, consisting of four layers: a visualization layer, a service layer, a data access layer, and a data storage layer, and the constructed knowledge graph includes a data coupling module, an entity recognition module, a relationship extraction module, and a data display module.

[0022] To make the technical means, creative features, achieved purposes, and functions of the present invention easy to understand, the present invention will be further elaborated below in conjunction with specific embodiments:

[0023] I. Construction of the electricity-carbon market coupling relationship model

[0024] 1. Electricity-carbon market quantity coupling relationship

[0025] The correlation between the electricity-carbon market, especially in terms of quantity, is mainly reflected in the interaction between the actual electricity output and the allocated carbon quotas. Specifically, the initial carbon emission permits obtained by power generation enterprises under the carbon trading mechanism and their actual total emissions are both determined based on their respective total electricity production volumes. The relevant quantitative relationship model can be expressed as:

[0026]

[0027] ε = α c - C

[0028] In the formula: α c is the total initial allocation of carbon quotas for the power generator during the period from 0 to N; is the winning electricity output of the power generator at time t (assuming that this output is consistent with the actual power generation output); are the benchmark carbon emission intensity corresponding to the power generation type of the power generator and the actual carbon emission intensity of the power generator respectively; C is the actual carbon emissions of the power generator during the period from 0 to N; ε is the difference between the actual carbon emissions of the power generator and the carbon quotas initially obtained during the period from 0 to N. If it is positive, it indicates that the power generator has surplus carbon quotas and can sell them in the secondary carbon market to obtain benefits. If it is negative, it means that the power generator needs to purchase additional carbon quotas to meet its carbon quota settlement requirements.

[0029] 2. Price linkage effect between the electricity-carbon market

[0030] In the carbon market system, carbon quotas exhibit commodity attributes, and the core of their pricing mechanism lies in the dynamic market supply and demand. From a long-term perspective, the electricity market clearing behavior of power generation enterprises and their carbon emission performance deeply affect the supply and demand balance of carbon quotas. Relatively speaking, in the short term, the fluctuations in carbon quota prices mainly make fine-tuning around this long-term supply and demand baseline, and these fine-tunings usually stem from the immediate supply and demand changes in the secondary carbon market or the expected adjustments to future supply and demand changes.

[0031] 3. Time-scale coupling between the electricity-carbon market

[0032] There are significant differences in the cycle and frequency between electricity trading and carbon trading. Specifically, the carbon market usually sets a one-year compliance period and allows flexible secondary market trading during specific periods within the trading day, requiring power generators to only ensure that they meet the carbon quota delivery conditions at the end of the compliance period. In contrast, the medium- and long-term trading in the electricity market covers different time periods from years to months, and even to weeks, while the spot trading is further refined to the daily and even hourly levels. Therefore, within a carbon market compliance cycle, both the electricity market and the secondary carbon market have the ability to complete transactions and settlements multiple times.

[0033] Based on the difference in the coupling time length, the interaction between the electricity-carbon market is divided into two categories: long-term synergy and short-term interaction. The long-term synergy is based on the compliance cycle of the carbon market, aiming to ensure that the quantity and price relationship between the electricity market and the carbon market reaches coordination and consistency from the current time to the end of the compliance period. The short-term interaction is based on the time period of a single electricity spot transaction. During this period, the electricity spot market and the secondary carbon market quickly reach equilibrium and complete the settlement process simultaneously, thereby determining the electricity spot price and the secondary carbon market price within this specific time period.

[0034] (1) Long-term synergy stage

[0035] To avoid risks, the upper limit of the carbon quota purchased by power generation enterprises each month is set as a certain proportion of the initial government-allocated quota. This aims to prevent enterprises from filling the annual carbon quota gap at one time and ensure that risks can be reasonably dispersed and alleviated. Therefore, a maximum purchase quantity constraint is set:

[0036]

[0037] In the formula: is the carbon quota purchase quantity of power generation enterprise i in month m; μ is the monthly carbon quota trading purchase upper limit coefficient; Q i is the total annual carbon quota allocated to power generation enterprise i according to its power generation volume.

[0038] Similarly, an upper limit for the maximum sales volume is also stipulated, that is, the quantity of carbon quotas that an enterprise can sell shall not exceed the total carbon quota it currently holds. The selling constraint:

[0039]

[0040] In the formula: is the carbon quota selling quantity of power generation enterprise i in month m; is the total surplus of carbon quotas of power generation enterprise i in the current year.

[0041] (2) Short-term interaction stage

[0042] When constructing the short-term interaction stage, it is necessary to take the carbon quota trading strategy in the long-term synergy stage as the cornerstone and regard the long-term decision-making of power generation enterprises as a prerequisite for short-term decision-making to ensure that the trading volume in the secondary carbon market conforms to their established long-term plans. Specifically, the quantity of carbon quotas cumulatively purchased by power generation enterprises in the current month should not be lower than the estimated required purchase quantity for that month according to the long-term strategy, so as to fully ensure that the carbon quotas they purchase are sufficient to fill the existing quota gap. That is:

[0043]

[0044] For the current month, the cumulative amount of carbon quota sold should not exceed the maximum amount for that month preset in the long-term carbon quota trading strategy, so as to ensure that the amount of carbon quota sold by power generation enterprises remains within the scope of their surplus. That is:

[0045]

[0046] In the formula: are the carbon quota purchase volume and sales volume of power generator i at time t, respectively.

[0047] II. Construction of a named entity recognition model based on pointer network and multi-task learning

[0048] 1. Shared feature extraction layer

[0049] In the embedding layer stage, the method of fusing character information embedding and position embedding is adopted to realize the vector representation of the text. The collected information A = {a1, a2,... a l} of sequence length L is divided and word-embedded to obtain the input tensor where b represents the batch size, l represents the sequence length, and d represents the word embedding dimension. Then, the characters in the sequence are encoded through a linear transformation to obtain their position vectors in the collected information That is:

[0050]

[0051] Among them, pos represents the position of the character in the sequence (counting from 0), and i represents a certain dimension of the vector. These encoders are composed of a combination of multi-layer attention mechanisms and feed-forward neural networks, and they can capture relevant information in the position encoding and character embedding. Under the interaction of multiple encoders, the input vector formula of the model is obtained:

[0052] A body = A input + A pos

[0053] Among them, a Transformer encoder based on the multi-head attention mechanism is adopted. This encoder can effectively transform the input text sequence, generate a dataset with diverse and rich features, and has strong unsupervised learning characteristics and representation extraction capabilities, significantly enhancing the overall efficiency of the system and achieving better processing results.

[0054] 2. Multi-task learning layer

[0055] Entity boundary recognition layer:

[0056] By using a shared feature extraction layer, it is possible to efficiently capture the long-distance correlation of the information before and after the entity, and extract valuable information from it. This information is then transformed into a vector representation rich in sample feature information. That is, the probability P of the start and end positions of the i-th data is obtained. i s_r and P i e_r .

[0057]

[0058] In the formula, a i forms a new vector table after being processed by the encoder. s and e represent the start and end respectively, is a weight vector that can be used for training, is the bias, and σ is the sigmoid function.

[0059] Entity classification layer:

[0060] For the entity classification task, the present invention integrates the HowNet knowledge base and the dictionary resources of Wikipedia, aiming to improve the entity classification accuracy in the power carbon market and expand the comprehensive description information of entities; constructs a similarity calculation model based on the comprehensive description features of entities to output the probability distribution of the entity's category.

[0061] A sl =View(A label )W label +b label

[0062] where A label is used as the input; N is the entity category in the sample set, b label is the bias; by using the View reconstruction tensor dimension matrix transformation method, by transforming the dimension originally composed of (N + 1)×d label ×h into the dimension composed of h×(N + 1)×d label , a more accurate entity category description vector can be obtained Then, the dot product method is used for entity classification to obtain the category probability output result corresponding to each input character.

[0063] C a1,a2,...,ad =sofmax(A ss ·A sl )

[0064] where, A ss is the sample feature vector.

[0065] 3. Output layer and loss function calculation

[0066] The present invention integrates the entity recognition probability and the entity classification probability to obtain the output result. Based on the multi-task learning method proposed by this model, the loss functions of the electric carbon market entity recognition task and entity classification are weighted, so as to construct a more comprehensive overall loss function of the model. The specific expression is as follows:

[0067]

[0068] Among them, n represents the length of the input sequence, and and are the known correct classification labels.

[0069]

[0070] Among them, is the entity category label of each character.

[0071] loss=γ·loss s1 +(1-γ)·loss s2

[0072] Among them, γ∈[0,1] is a model hyperparameter. As the training process progresses until it converges to the optimal value, the model can reach the optimal state, thereby significantly improving its performance.

[0073] III. Construct a multi-head selective relationship extraction model based on the global entity relative position representation

[0074] 1. BERT feature extraction layer

[0075] It is stacked based on multiple layers of Transformer Encoder. Each layer contains a multi-head self-attention mechanism and a feed-forward neural network. With its bidirectional encoding characteristics, it simultaneously considers the information relevance before and after the entity, captures the deep features of the entity through a complex attention mechanism, extracts key data, and performs effective preprocessing on it.

[0076] 2. CRF layer

[0077] CRF is built on top of BERT. Assuming that y∈{B-type, I-type, O} is the label, the scoring function s(X, i), y i is the output of BERT at the i-th character, is the trainable parameter, then the objective function for entity recognition is:

[0078] L CRF =-∑P(Y|X)

[0079]

[0080] 3. Long Selection Layer

[0081] The relationship classification task is formulated as a multi-head selection problem, where each head is responsible for extracting a relationship. Given the entity label sequence y=(y o ,y1,...,y n ), each label is mapped to a distributed label embedding as l = (l o ,l1,...,l n ), where is the label embedding size. The most likely tail entity s j With the head entity s i The corresponding relationship r k Predicted to be:

[0082] P(tail=s j ,relation=r k |head=s i )=σ(g(z i ,z j ,r k ))

[0083] g(z i ,z j ,r k )=V r f(U r z j +W r z i +b r )

[0084] Among them, z i =[l i ;o i ], i=0,1,...,N, for each input state; For a given relationship; f(·) is the RELU function, and σ(·) represents the sigmoid function.

[0085] During the training process, the candidate tail entity s ij and relationships ij Optimize cross entropy L MHS Loss, given the head entity s j .

[0086]

[0087] Where M is s j The number of relationships.

[0088] 4. Global Relationship Classification

[0089] A global relation classification strategy is introduced, aiming to guide the optimization training of local semantic features. The [CLS] token is used to predict the relation category associated with the whole s. During the training phase, the binary loss of global classification is minimized, that is:

[0090]

[0091] P(relation=r|s)=σ(W g p0+b g )

[0092] where p0 is the hidden state of the relative position layer, σ(·) represents the sigmoid function, and T is the total number of relations.

[0093] 5. Joint training

[0094] Finally, joint training is carried out to obtain the final comprehensive objective function:

[0095] L=L CRF +γL G +L MHS

[0096] where L CRF 、L G and L MHS represent the loss function formulas for head entity recognition, global relation classification, and multi-head selection respectively, and γ∈[ ] is the weight controlling global relation classification.

[0097] IV. Constructing a knowledge graph for the coupling relationship of the complex electricity-carbon market

[0098] To construct a knowledge graph for the coupling relationship of the complex electricity-carbon market, a hierarchical architecture is adopted, which consists of four layers: the visualization layer, the service layer, the data access layer, and the data storage layer.

[0099] The data storage layer of Neo4j can display the relationships between entities in a graphical way, constructing a dynamic knowledge graph. This graph intuitively presents the interconnections between entities, laying a good foundation for subsequent analysis and applications, and effectively supporting the visualization requirements of the electricity-carbon market.

[0100] Based on the Python programming language, the data access function is implemented, achieving a seamless combination of logical operations and data access and storage, improving the compatibility of module expansion, and facilitating the subsequent upgrade work of the system. In addition, this invention also uses the Cypher query language to perform operations such as matching, adding, and deleting relationships between nodes, and finally uses the JSON data format to present more accurate query results.

[0101] The service layer architecture built with Vue components and Flask framework can convert the input query results into the required data format, achieve efficient data transmission and logical operations, and then improve the efficiency and reliability of queries. The Vue framework incorporates many powerful functions, covering form validation, string setting, HTML template generation, and dropdown lists, fully meeting the diverse needs of the electric carbon market. On the other hand, Flask, a Python-based web framework, can handle various complex scenarios with its high flexibility and lightweight architecture.

[0102] The visualization layer utilizes JavaScript and Echarts technologies to fully display data on the Web page. As an interactive visualization chart library based on JS, Echarts can quickly and efficiently display the knowledge graph structure, meet the diverse needs of the electric carbon market, and provide a more convenient service experience for users.

[0103] The knowledge graph constructed in this embodiment includes Figure 2 the four major modules shown as follows. They are the data coupling module, which analyzes the electric carbon market data from three aspects: the market transaction quantity, the market transaction price, and the market operation time; the entity recognition module, which combines multi-task learning and pointer network mechanism to recognize the electric carbon market entities; the relationship extraction module, which combines sequence modeling technology, multi-head selection framework, and global relationship classification to accurately extract entity relationships; and the data display module, which stores and visually displays the data.

[0104] According to the second aspect of the embodiments of the present invention, a processor is provided. The processor is used to run a program, wherein when the program runs, it executes the method for constructing a knowledge graph for complex electric carbon market coupling relationships described in any one of the above.

[0105] According to the third aspect of the embodiments of the present invention, a computer-readable storage medium is provided, characterized in that the computer-readable storage medium includes a stored program, wherein when the program runs, it controls the device where the computer-readable storage medium is located to execute the method for constructing a knowledge graph for complex electric carbon market coupling relationships described in any one of the above.

[0106] The specific implementation manners of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above implementation manners, and various changes can be made without departing from the spirit of the present invention within the scope of knowledge possessed by those of ordinary skill in the art.

Claims

1. A method for constructing a knowledge graph for the coupling relationship of a complex electro-carbon market, characterized in that include: Construct a coupling relationship model between electricity and carbon markets; Build a named entity recognition model based on pointer network and multi-task learning; Construct a multi-head selection relationship extraction model based on global entity relative position representation; Construct a knowledge graph for complex electricity-carbon market coupling relationships.

2. The knowledge graph construction method for the complex electro-carbon market coupling relationship according to claim 1, wherein The construction of the electricity-carbon market coupling relationship model includes: From the three aspects of market transaction quantity, market transaction price, and market operation time, the complex interaction and coupling phenomena caused by common participants in the electricity market and the carbon market are analyzed; an electricity-carbon market coupling relationship model is constructed, which includes the quantity coupling relationship of the electricity-carbon market, the price linkage effect of the electricity-carbon market, and the time scale coupling of the electricity-carbon market.

3. The knowledge graph construction method for complex electro-carbon market coupling relationships according to claim 2, characterized in that The quantity coupling relationship of the electricity-carbon market is specifically expressed as follows: ε = α c -C Where: α c is the total initial allocation of carbon quotas for the power generator during the 0 - N periods; is the winning power generation output of the power generator at time t; are respectively the benchmark carbon emission intensity corresponding to the power generation type of the power generator and the actual carbon emission intensity of the power generator's power generation; C is the actual carbon emissions of the power generator during the 0 - N periods; ε is the difference between the actual carbon emissions of the power generator and the carbon quotas initially obtained during the 0 - N periods.

4. The knowledge graph construction method for the complex electro-carbon market coupling relationship according to claim 2, wherein The electricity-carbon market price linkage effect divides the interaction between the electricity and carbon markets into two categories: long-term synergy and short-term interaction. Specifically: (1) Long-term collaboration stage The monthly carbon quota purchase limit for power generation companies is set at a certain proportion of the government's initial allocation of quotas, with a maximum purchase amount constraint: Wherein: is the carbon quota purchase volume of power generator i in month m; μ is the upper limit coefficient of monthly carbon quota trading purchase; Q i is the total annual carbon quota allocated to power generator i according to its power generation volume; The number of carbon quotas that an enterprise can sell must not exceed the total amount of carbon quotas it currently holds, that is, the selling constraint: In the formula: is the carbon quota selling volume of power generator i in month m; is the total surplus of carbon quota of power generator i in the current year; (2) Short-term interaction stage The cumulative amount of carbon allowances purchased by power generation companies in a given month should not be less than the required purchase amount for that month estimated based on the long-term strategy, namely: For the current month, the cumulative sales volume of its carbon quotas should not exceed the maximum sales volume for that month set by the long-term carbon quota trading strategy, that is: Wherein: are respectively the carbon quota purchase volume and sales volume of power generator i in period t.

5. The knowledge graph construction method for the complex electro-carbon market coupling relationship according to claim 1, wherein The construction of a named entity recognition model based on pointer network and multi-task learning includes: Based on the multi-task learning strategy and pointer network mechanism, with the help of the HowNet knowledge base, a named entity recognition model consisting of a shared feature extraction layer, a multi-task learning layer and an output layer is constructed. The traditional named entity recognition task is decomposed into two sub-tasks: entity boundary recognition and entity classification, which can accurately identify the entities and their categories in the electricity carbon market.

6. The method for constructing a knowledge graph for complex electro-carbon market coupling relationships according to claim 1, wherein The multi-head selection relationship extraction model based on global entity relative position representation is constructed, including: Based on BERT and CRF technology, a representation method for the relative positions of global entities is proposed, which is combined with a multi-head selection framework. At the same time, a global relationship classification strategy is introduced. Through joint optimization, entity relationships are accurately extracted. The model consists of four parts: BERT feature extraction layer, CRF layer, multi-head selection layer and global relationship classification.

7. The method for constructing a knowledge graph for complex electro-carbon market coupling relationships according to claim 1, characterized in that The construction of a knowledge graph for complex electricity-carbon market coupling relationships is specifically as follows: It adopts a layered architecture and consists of four levels: visualization layer, service layer, data access layer and data storage layer. The constructed knowledge graph includes data coupling module, entity recognition module, relationship extraction module and data display module.

8. A processor, characterized in that, The processor is used to run a program, wherein when the program is run, the knowledge graph construction method for complex electricity-carbon market coupling relationships described in any one of claims 1 to 7 is executed.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein, when the program is running, the device where the computer-readable storage medium is located is controlled to execute the knowledge graph construction method for complex electricity-carbon market coupling relationships described in any one of claims 1-7.

Images