Methods, electronic devices, media, and program products for identifying associations with electricity trading policies

Abstract

This invention provides a method, electronic device, medium, and program product for identifying policy associations in power trading, relating to the field of power trading technology. The method includes: constructing a policy knowledge graph from power trading policy text data; constructing a business knowledge graph from power trading business data; constructing a market rule knowledge graph from power trading business data; generating a multi-domain fusion association graph from the policy knowledge graph, business knowledge graph, and market rule knowledge graph; performing path reasoning search in the multi-domain fusion association graph, starting from policy nodes and ending at trading nodes, to generate a policy-trading association link set; extracting parameters from the multi-domain fusion association graph for each association link, and calculating a policy impact intensity score for each association link based on the extracted parameters, generating a policy impact intensity result set; and performing trigger condition mapping and structured integration on the policy-trading association link set and the policy impact intensity result set to generate a policy association identification result set.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of power trading technology, and in particular to a method, electronic device, medium and program product for identifying power trading policy associations. Background Technology

[0002] With the continuous development of the electricity market and the diversification of electricity trading activities, electricity trading policies are becoming increasingly complex, involving multi-dimensional information such as policy subjects, transaction types, market rules, and constraints. In the electricity trading process, how to accurately understand and analyze the potential impact of policies on trading behavior, and how to achieve policy implementation and compliance management, have become urgent problems that intelligent electricity trading systems need to solve.

[0003] In existing technologies, the analysis of electricity trading policies, trading data, and market rules typically relies on manual interpretation or keyword matching and rule retrieval to correlate policy provisions with specific trading behaviors. This approach struggles to reveal the intrinsic relationship between policies and transactions. While some existing technologies incorporate knowledge graphs, they are often limited to parsing single policy texts or modeling single transaction data, failing to accurately map policies to trading behaviors and thus undermining the intelligent analysis needs of complex electricity trading scenarios. Summary of the Invention

[0004] This invention provides a method, electronic device, medium, and program product for identifying the correlation between power trading policies. It can model policies, business, and market rules to generate a knowledge graph, and based on this, identify policy-transaction correlation links and quantify the intensity of policy impact, thereby achieving intelligent correlation identification between power trading policies and trading behaviors.

[0005] In a first aspect, the power trading policy association identification method provided in the embodiments of the present invention includes: Based on a pre-configured policy ontology, entity recognition and entity relationship modeling are performed on power trading policy text data to construct a policy knowledge graph; Based on a pre-configured transaction ontology, entity identification and entity relationship modeling are performed on power transaction business data to construct a business knowledge graph; Based on a pre-configured market rule ontology, entity identification and entity relationship modeling are performed on power trading business data to construct a market rule knowledge graph. Align entity nodes in the policy knowledge graph, business knowledge graph, and market rule knowledge graph, and construct cross-domain association edges based on the alignment results to generate a multi-domain fusion association graph; In the multi-domain fusion association graph, path reasoning search is performed with policy nodes as the starting point and transaction nodes as the ending point to generate a set of policy-transaction association links; For each link in the policy transaction link set, path structure parameters and node and relationship weight parameters are extracted from the multi-domain fusion link graph. Based on the extracted parameters, the policy impact intensity score of each link is calculated, and a policy impact intensity result set is generated. Based on the trigger logic elements identified from the power trading policy text data, trigger condition mapping and structured integration are performed on the policy trading association link set and the policy impact intensity result set to generate a policy association identification result set.

[0006] Secondly, the power trading policy association identification device provided in this embodiment of the invention includes: The policy graph construction module is used to perform entity recognition and entity relationship modeling on power trading policy text data based on pre-configured policy ontology, and to construct a policy knowledge graph. The business graph construction module is used to perform entity identification and entity relationship modeling on power transaction business data based on pre-configured transaction ontology, and to construct a business knowledge graph. The market graph construction module is used to perform entity identification and entity relationship modeling on power trading business data based on a pre-configured market rule ontology, and to construct a market rule knowledge graph. The fusion graph generation module is used to align entity nodes in the policy knowledge graph, business knowledge graph, and market rule knowledge graph, and construct cross-domain association edges based on the alignment results to generate a multi-domain fusion association graph. The association link generation module is used to perform path reasoning search in a multi-domain fusion association graph, starting from policy nodes and ending at transaction nodes, to generate a set of policy-transaction association links. The association link scoring module is used to extract path structure parameters and node and relationship weight parameters from the multi-domain fusion association graph for each association link in the policy transaction association link set, and calculate the policy impact intensity score for each association link based on the extracted parameters, generating a policy impact intensity result set; The identification module generation module is used to perform trigger condition mapping and structured integration on the policy transaction association link set and the policy impact intensity result set based on the trigger logic elements identified from the power transaction policy text data, and generate a policy association identification result set.

[0007] Thirdly, the electronic device provided in the embodiments of the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the power trading policy association identification method as described in any embodiment of the present invention.

[0008] Fourthly, the computer-readable storage medium provided in the embodiments of the present invention stores a computer program thereon, which, when executed by a processor, implements the power trading policy association identification method as described in any embodiment of the present invention.

[0009] Fifthly, the computer program product provided in the embodiments of the present invention includes a computer program that, when executed by a processor, implements the power trading policy association identification method as described in any embodiment of the present invention.

[0010] In this embodiment of the invention, entity recognition and entity relationship modeling are performed on power trading policy text data based on a pre-configured policy ontology to construct a policy knowledge graph. This enables a structured representation of policy-related entities and their relationships within the power trading policy text data, providing a data foundation for subsequent cross-domain association and reasoning. Similarly, entity recognition and entity relationship modeling are performed on power trading business data based on a pre-configured trading ontology to construct a business knowledge graph. This enables a structured representation of transaction-related entities and their relationships within the power trading business data, providing a data foundation for subsequent cross-domain association and reasoning. The market rule ontology performs entity identification and entity relationship modeling on power trading business data, constructing a market rule knowledge graph. This graph can structurally represent market rule-related entities and their relationships within power trading business data, providing a data foundation for subsequent cross-domain association and reasoning. It aligns entity nodes in the policy knowledge graph, business knowledge graph, and market rule knowledge graph, and constructs cross-domain association edges based on the alignment results to generate a multi-domain fusion association graph. This enables cross-domain fusion of policy, business, and market rule knowledge, clarifying the correspondence and potential associations between entities in different knowledge graphs, allowing multi-domain information to be integrated. A unified representation and analysis provides a complete and unified knowledge network for subsequent policy transaction correlation and impact analysis. In the multi-domain fusion correlation graph, path reasoning search is performed with policy nodes as the starting point and transaction nodes as the ending point to generate a policy transaction correlation link set. This identifies potential impact paths of policies on transaction behavior and generates a structured policy transaction correlation link set, providing a clear analytical object for calculating policy impact intensity. For each correlation link in the policy transaction correlation link set, path structure parameters and node and relationship weight parameters are extracted from the multi-domain fusion correlation graph. Based on the extracted parameters, a policy impact intensity score is calculated for each correlation link, generating a policy impact intensity result set. This quantifies the degree of impact of different policies on transaction links, providing a basis for policy prioritization and risk assessment. Based on the triggering logic elements identified from power transaction policy text data, trigger condition mapping and structured integration are performed on the policy transaction correlation link set and the policy impact intensity result set to generate a policy correlation identification result set. This integrates policy triggering conditions with transaction behavior correlation links and impact intensity results to form a structured policy correlation identification result, achieving intelligent correlation identification between power transaction policies and transaction behavior, and supporting intelligent policy implementation and compliance analysis. Attached Figure Description

[0011] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a flowchart illustrating a method for identifying the correlation between electricity trading policies provided in an embodiment of the present invention; Figure 2 This is another flowchart illustrating the power trading policy association identification method provided in this embodiment of the invention; Figure 3 This is a schematic diagram of the structure of the power trading policy association identification device provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0013] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0014] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0015] Figure 1This is a flowchart illustrating a method for identifying the correlation between electricity trading policies provided in an embodiment of the present invention. This method is applicable to scenarios involving intelligent correlation identification of policies and trading behaviors in electricity trading. The method can be executed by a device for identifying the correlation between electricity trading policies provided in this embodiment, which can be implemented using software and / or hardware. In one specific embodiment, the device can be integrated into an electronic device, such as a computer or server. The following embodiment illustrates the integration of the electricity trading policy correlation identification device into an electronic device. (See also...) Figure 1 The power trading policy association identification method in this embodiment may include the following steps: Step 101: Based on the pre-configured policy ontology, perform entity identification and entity relationship modeling on the power trading policy text data to construct a policy knowledge graph.

[0016] A pre-configured policy ontology refers to a pre-established and stored knowledge structure for the power trading policy domain, used to uniformly define and standardize the entity types, relationship types, and semantic constraints involved in the policy domain. Power trading policy text data refers to policy document text data describing power trading rules, constraints, applicable subjects, timeframes, and triggering logic, including but not limited to policy texts in the form of measures, notices, rules, and detailed rules. Entity identification refers to the process of automatically identifying and extracting textual information corresponding to the entity types in the policy ontology from power trading policy text data based on predefined entity types. Entity relationship modeling refers to the process of structurally representing and describing the relationships between identified entities based on predefined relationship types in the policy ontology. A policy knowledge graph refers to a graphical data model constructed using entities as nodes and relationships between entities as edges.

[0017] Specifically, semantic analysis of power trading policy texts is performed through a pre-configured policy ontology. First, various entities that conform to the policy ontology definition are identified from the power trading policy texts. Then, the relationships between entities are modeled based on the policy ontology. Finally, a policy knowledge graph is constructed in the form of a graph structure, which can reflect power trading policy knowledge and its inherent relationships.

[0018] Step 102: Based on the pre-configured transaction ontology, perform entity identification and entity relationship modeling on the power transaction business data to construct a business knowledge graph.

[0019] A pre-configured transaction ontology refers to a pre-established and stored knowledge structure oriented towards the power trading business domain, used to uniformly define and standardize the entity types, relationship types, and semantic constraints involved in the power trading business domain. Power trading business data refers to information data generated during the power trading process, used to describe trading entities, trading varieties, trading behaviors, trading periods, trading conditions, trading status, and trading results, including structured and semi-structured data. Entity identification refers to the process of automatically identifying and extracting data items or information units corresponding to the entity types in the transaction ontology from the power trading business data based on predefined entity types. Entity relationship modeling refers to the process of structurally representing and describing the business relationships between identified entities based on predefined relationship types in the transaction ontology. A business knowledge graph refers to a graphical data model constructed using power trading business entities as nodes and the business relationships between entities as edges.

[0020] Specifically, semantic parsing of power trading business data is performed based on a pre-configured transaction ontology. First, various business entities that conform to the definition of the transaction ontology are identified from the power trading business data. Then, the relationships between business entities are modeled according to the pre-defined relationship types in the transaction ontology. Finally, a business knowledge graph is constructed in the form of a graph structure, which can reflect the elements of power trading business and their interrelationships.

[0021] Step 103: Based on the pre-configured market rule ontology, perform entity identification and entity relationship modeling on the power trading business data to construct a market rule knowledge graph.

[0022] A pre-configured market rule ontology refers to a pre-established and stored knowledge structure oriented towards the domain of power trading market rules, used to uniformly define and standardize the rule elements, constraints, restrictions, and compliance logic involved in power trading activities. Entity identification refers to the process of automatically identifying and extracting data items or information units corresponding to entity types in the market rule ontology from power trading business data, based on predefined rule element entity types. Entity relationship modeling refers to the process of structurally representing and describing the constraint, dependency, or restriction relationships between identified rule element entities based on predefined relationship types in the market rule ontology. A market rule knowledge graph refers to a graphical data model constructed using market rule element entities as nodes and constraint or logical relationships between rule elements as edges.

[0023] Specifically, based on the pre-configured market rule ontology, the power trading business data is subjected to rule semantic parsing. Various rule element entities that conform to the definition of the market rule ontology are identified from the power trading business data. Then, according to the predefined relationship types in the market rule ontology, the constraint relationships and logical relationships between rule element entities are modeled. Finally, a market rule knowledge graph is constructed in the form of a graph structure. This market rule knowledge graph can reflect the power trading market rule system and its inherent constraint relationships.

[0024] Step 104: Align the entity nodes in the policy knowledge graph, business knowledge graph, and market rule knowledge graph, and construct cross-domain association edges based on the alignment results to generate a multi-domain fusion association graph.

[0025] Entity nodes refer to entity objects in a knowledge graph. Each node represents a specific entity, such as a transaction entity, policy clause, or market rule item. Alignment refers to matching and associating semantically similar or related entity nodes from different knowledge graphs, enabling cross-domain entities to correspond. Cross-domain association edges are edges connecting semantically related entity nodes from different knowledge graphs, used to represent cross-domain relationships or dependencies. A multi-domain fusion association graph is a unified graphical data model formed by integrating policy knowledge graphs, business knowledge graphs, and market rule knowledge graphs through entity alignment and cross-domain association edges, used to present cross-domain entities and their relationships.

[0026] Specifically, entity nodes in knowledge graphs from three different domains (policy, business, and market rules) are matched to identify entities with the same or related semantics. Then, cross-domain association edges are used to connect them. In this way, the three originally independent knowledge graphs can be integrated into a unified, multi-domain associated graph structure, namely a multi-domain fusion association graph. This multi-domain fusion association graph can reflect the interaction relationship and overall association logic between policy, business, and market rules.

[0027] Step 105: In the multi-domain fusion association graph, perform path reasoning search with policy nodes as the starting point and transaction nodes as the ending point to generate a policy-transaction association link set.

[0028] Policy nodes refer to nodes corresponding to policy entities in the multi-domain integrated association graph, such as specific policy clauses, notices, or rule items. Transaction nodes refer to nodes corresponding to electricity trading business entities in the multi-domain integrated association graph, such as trading entities, trading contracts, or trading events. Path reasoning search refers to exploring possible connection paths along the graph edges from the starting node in the multi-domain integrated association graph to infer potential relationships or impacts between entities. The policy-transaction association link set refers to the set of all feasible paths obtained from policy nodes to transaction nodes in the multi-domain integrated association graph through path reasoning search, used to reveal how policies influence the relationships of specific trading activities.

[0029] Specifically, in the pre-constructed multi-domain integrated relationship graph, starting with the policy entity and ending with the transaction entity, path reasoning search identifies all links that may affect or relate to the transaction through policy. These links form the policy-transaction relationship link set, used to visually display the logical relationship and impact path between policy and electricity trading operations.

[0030] Step 106: For each associated link in the policy transaction associated link set, extract path structure parameters and node and relationship weight parameters from the multi-domain fusion associated graph, and calculate the policy impact intensity score for each associated link based on the extracted parameters to generate a policy impact intensity result set.

[0031] A policy transaction link is a single path within a policy transaction link set. This path consists of a series of entity nodes and relational edges connecting them, representing a specific association between policy and transaction. Path structure parameters describe the overall structural characteristics of the link, including structural attributes such as path hop count, cross-domain jumps, and path interruptions. Node and relation weight parameters characterize the importance of each entity node and the strength of relational edges within the link. Node weights reflect the importance of an entity in the graph, while relation weights reflect the strength of the relationships between entities. The policy impact strength score is a numerical result calculated based on the path structure parameters and node and relation weight parameters of the link, representing the degree of policy influence on transaction behavior within that link. The policy impact strength result set is the collection of policy impact strength scores calculated for each link in the policy transaction link set.

[0032] Specifically, for each link in the policy transaction link set, path structure parameters reflecting the structural characteristics of the link and node and relationship weight parameters reflecting the importance of entities and their relationships are extracted from the multi-domain fusion link graph. Then, through a preset calculation method, the path structure parameters, node and relationship weight parameters of each link are comprehensively calculated to obtain the corresponding policy impact intensity score. The impact degree of different links is distinguished according to the policy impact intensity score to form a policy impact intensity result set. The preset calculation method can be as follows: calculate the path connectivity index based on the path structure parameters of the link, and multiply the path connectivity index by the path importance coefficient to obtain the path score of the link; perform a weighted summation of the node importance coefficients based on the weight parameters of each node in the link to obtain the node score of the link; perform a weighted summation of the relationship importance coefficients based on the weight parameters of each relationship edge in the link to obtain the relationship score of the link; and add the path score, node score and relationship score to obtain the policy impact intensity score of the corresponding link. Specifically, the node importance coefficient can be preset based on the potential impact of policy, transaction, or market rule nodes on target transaction behavior in the entire multi-domain fusion relationship graph, or obtained through scoring by domain experts; the relationship importance coefficient can be preset based on the strength of the relationship and dependence between nodes, as well as the binding force of policies on transactions, or determined by scoring by domain experts; the path importance coefficient can be preset according to the overall weight of the path in the policy transaction impact analysis. Through these preset importance coefficients, it can be ensured that the policy impact intensity score can quantify the relative impact of different links in the entire policy and transaction system.

[0033] Step 107: Based on the trigger logic elements identified from the power trading policy text data, perform trigger condition mapping and structured integration on the policy trading association link set and the policy impact intensity result set to generate a policy association identification result set.

[0034] Triggering logic elements refer to conditions or rules identified from power trading policy text data that can trigger or influence specific trading behaviors. For example, a policy might stipulate that "when power generation enterprises participate in medium- and long-term transactions, the declared electricity volume shall not be less than 70% of the contracted electricity volume." Triggering condition mapping refers to associating the identified triggering logic elements with the policy trading linkage set and its corresponding policy impact intensity result set, clarifying the triggering conditions for each linkage. Structured integration refers to organizing and integrating the mapped triggering conditions, linkages, and impact intensity scores using a unified data structure to form a computable and analyzable result set. The policy linkage identification result set refers to the structured data set generated after triggering condition mapping and structured integration, reflecting the relationship between policy triggering conditions and specific trading behaviors.

[0035] Specifically, conditions (triggering logic elements) that can trigger or influence trading behavior are extracted from the power trading policy text. These conditions are then mapped to the previously generated policy-trading association link set and the impact strength results of each link. Subsequently, this information is organized according to a unified data structure to obtain a complete policy association identification result set. This result set includes both the triggering conditions corresponding to each link and the impact strength of the policy on trading behavior, facilitating subsequent analysis and decision-making.

[0036] In this embodiment, entity recognition and entity relationship modeling are performed on power trading policy text data based on a pre-configured policy ontology to construct a policy knowledge graph. This allows for a structured representation of policy-related entities and their relationships within the power trading policy text data, providing a data foundation for subsequent cross-domain association and reasoning. Similarly, entity recognition and entity relationship modeling are performed on power trading business data based on a pre-configured trading ontology to construct a business knowledge graph. This allows for a structured representation of transaction-related entities and their relationships within the power trading business data, providing a data foundation for subsequent cross-domain association and reasoning. Furthermore, based on a pre-configured market... The market rule ontology performs entity identification and entity relationship modeling on power trading business data, constructing a market rule knowledge graph. This graph provides a structured representation of market rule-related entities and their relationships within the power trading business data, laying the data foundation for subsequent cross-domain association and reasoning. It aligns entity nodes in the policy knowledge graph, business knowledge graph, and market rule knowledge graph, and constructs cross-domain association edges based on the alignment results to generate a multi-domain fusion association graph. This enables cross-domain fusion of policy, business, and market rule knowledge, clarifying the correspondences and potential associations between entities in different knowledge graphs, and allowing for the unified integration of multi-domain information. The system first represents and analyzes the policy transaction links, providing a complete and unified knowledge network for subsequent policy transaction link and impact analysis. In the multi-domain fusion association graph, path reasoning search is performed with policy nodes as the starting point and transaction nodes as the ending point to generate a policy transaction link set. This identifies potential impact paths of policies on transaction behavior and generates a structured policy transaction link set, providing a clear analytical object for calculating policy impact intensity. For each link in the policy transaction link set, path structure parameters and node and relationship weight parameters are extracted from the multi-domain fusion association graph. Based on the extracted parameters, a policy impact intensity score is calculated for each link, generating a policy impact intensity result set. This quantifies the degree of impact of different policies on transaction links, providing a basis for policy prioritization and risk assessment. Based on the triggering logic elements identified from the power transaction policy text data, trigger condition mapping and structured integration are performed on the policy transaction link set and the policy impact intensity result set to generate a policy association identification result set. This integrates policy triggering conditions with transaction behavior link and impact intensity results, forming a structured policy association identification result. This achieves intelligent association identification between power transaction policies and transaction behavior, supporting intelligent policy implementation and compliance analysis.

[0037] Figure 2 This is another flowchart illustrating the power trading policy association identification method provided in this embodiment of the invention, such as... Figure 2 As shown, the power trading policy association identification method in this embodiment may include: Step 201: Preprocess the power trading policy text data into a structured text sequence.

[0038] Preprocessing refers to the process of cleaning and standardizing the original policy text without altering its original semantics. This typically includes removing irrelevant symbols, formatting marks, and redundant descriptions; segmenting the text into sentences, paragraphs, and bullet points; standardizing terminology; and standardizing the expression of time, proportions, and numerical values. Structured text sequence refers to transforming a previously continuous and loosely structured policy text into a sequence of text units that are organized sequentially, have clear semantic boundaries, and can be processed by programs. Each text unit expresses a relatively independent policy semantic, while the original logical order between text units is preserved.

[0039] Specifically, the natural language text in the original power trading policy documents is preprocessed to transform the unstructured policy text into a structured text sequence arranged in a logical order, providing standardized input for subsequent policy element identification.

[0040] Step 202: Use the power policy knowledge extraction model to jointly extract entities and relations from the structured text sequence.

[0041] A power policy knowledge extraction model refers to a knowledge extraction model built or trained specifically for the power trading policy domain. It is used to automatically identify knowledge elements related to power trading from policy texts. This model typically possesses the following capabilities: policy entity identification, policy element semantic modeling, and semantic recognition of relationships between entities. An entity refers to an object element with a clear meaning and business orientation in the semantics of power trading policies, such as policy subjects (e.g., power generation companies, power users, grid companies), trading objects (e.g., medium- and long-term transactions, electricity declarations, ancillary services), rule elements (e.g., minimum declaration quantity, price ceiling, settlement cycle), time elements, and constraint elements. A relationship refers to the semantic association or constraint relationship between different entities in the policy text, such as: applicable / applicable objects, constraints / restrictions, triggers / cause, and related trading varieties. Joint extraction refers to simultaneously identifying entities and the relationships between them within the same model or processing procedure.

[0042] Specifically, through the power policy knowledge extraction model, semantic analysis is performed on the preprocessed structured policy text sequence. In the same processing process, the power transaction-related entities involved in the policy text and the semantic relationships between the entities are identified, thereby forming a structured policy knowledge representation.

[0043] Optionally, the power policy knowledge extraction model includes a sparse label pair selection structure and a semantic constraint mask structure. The power policy knowledge extraction model is used to jointly extract entities and relations from structured text sequences. This includes: using the sparse label pair selection structure to filter the original label pair matrix generated after model encoding according to preset distance filtering rules, semantic relevance filtering rules, and entity type combination rules, resulting in a sparse label pair matrix; wherein, the distance filtering rule limits the relative distance between two labels in a combined label pair to no more than a preset distance threshold, the semantic relevance filtering rule limits the semantic vector similarity of the combined label pairs to no less than a preset similarity threshold, and the entity type combination rule limits the entity type or structural type corresponding to the combined label pair to belong to a preset valid combination list; using the semantic constraint mask structure, based on the structural features and element types of the policy text, a semantic constraint mask matrix is ​​generated, and the semantic constraint mask matrix is ​​used to mask label pairs that cross policy structural boundaries or whose element type combinations do not conform to preset type combination rules, thereby performing a secondary filtering on the sparse label pair matrix to obtain a valid label pair matrix.

[0044] Sparse tag pair selection structure refers to a selection structure that filters tag pairs with potential association value from the original tag pair matrix. Distance filtering rules are rules used to limit the relative positional distance between the two tags participating in a tag pair combination in the original text, excluding tag pairs with excessive text distance and low association probability. A preset distance threshold is a parameter used to limit the maximum allowed relative positional distance between two tags; when the interval between the two tags in a tag pair exceeds this threshold, the tag pair will not be retained. Semantic relevance filtering rules are rules used to measure the degree of relevance between two tags in the semantic space based on the semantic representation of the tags obtained after model encoding. Semantic vector similarity is a numerical value obtained by calculating the similarity between the semantic representation vectors corresponding to two tags, used to reflect the degree of semantic relevance between the two tags. A preset similarity threshold is a parameter used to limit the lower limit of semantic vector similarity; when the semantic vector similarity of a tag pair is lower than this threshold, the tag pair will be filtered. Entity type combination rules are rules that judge the legality of the entity type combination corresponding to a tag pair based on predefined entity type or structure type constraints. A pre-defined list of valid combinations refers to a pre-defined and stored set of combinations between entity types or structural types that are allowed to establish semantic relationships. This is used to exclude semantically unreasonable or meaningless entity type pairings. The original tag pair matrix is ​​a set of tag pairs constructed by combining the encoded structured text sequences in pairs after encoding. It contains all possible tag pair combinations in the text. A tag pair is a combination of two tags arranged according to their sequential position in the text, representing the possible semantic relationships between two text units. The tag refers to the encoded structured text. The sparse tag pair matrix is ​​a matrix constructed from the set of tag pairs retained after structurally filtering the original tag pair matrix using a sparse tag pair selection structure. The semantic constraint mask structure is a masking mechanism used to further impose semantic and structural constraints on tag pairs. It generates a mask matrix to mask tag pairs that do not meet the constraints. The structural features of policy texts refer to the structural information formed during the drafting process of policy documents, including structural features such as clause divisions, chapter levels, and sentence / segment boundaries. Element type refers to the type identification of different tags or tag pairs in a policy text, based on the policy semantic element framework, indicating their semantic category. Semantic constraint mask matrix is ​​a matrix generated based on the structural features and element types of the policy text, where matrix elements indicate whether corresponding tag pairs satisfy semantic and structural constraints. Policy structural boundary refers to the boundary positions between different structural units in the policy text, such as the boundaries between different clauses, chapters, or paragraphs. Element type pairing rules are preset constraint rules used to limit whether semantic associations are allowed between different semantic element types.An effective label pair matrix refers to the label pair matrix that is ultimately retained after the semantic constraint masking process is applied to the sparse label pair matrix, and can be used for subsequent entity recognition and relation modeling.

[0045] Specifically, based on the original label pair matrix obtained after model encoding, a sparse label pair selection structure is introduced. At the label pair construction layer, label pairs in the original matrix are structurally filtered using distance filtering rules, semantic relevance filtering rules, and entity type combination rules. Only label pairs that simultaneously satisfy the above filtering rules are retained, thus constructing a smaller, more semantically focused sparse label pair matrix. Subsequently, based on the sparse label pair matrix, a semantic constraint mask structure is further introduced. A semantic constraint mask matrix is ​​generated based on the structural features and semantic element types of the policy text. Label pairs that cross policy structural boundaries or whose element type combinations do not conform to preset type combination rules are masked, thus performing a secondary filtering of the sparse label pair matrix, ultimately obtaining an effective label pair matrix for entity recognition and relationship modeling. In this embodiment, the power policy knowledge extraction model, which includes a sparse label pair selection structure and a semantic constraint mask structure, is used to jointly extract entities and relations. This effectively filters and constrains label pairs, reduces redundant or erroneous entity relation combinations, and improves the accuracy and reliability of policy text information extraction. The effective label pair matrix generated through secondary filtering can retain key entity relations with high semantic relevance and reasonable type matching, providing high-quality input for subsequent standardization of policy semantic elements and knowledge graph construction, thereby improving the accuracy and efficiency of policy analysis, cross-domain association, and intelligent reasoning.

[0046] For example, let's illustrate the tag pair construction process for the policy text "Regulations on the Price Ceiling for Electricity Sold by Power Generation Enterprises". For ease of representation, let's assume "√" represents a retained tag pair and "—" represents a tag pair not considered (such as a combination of itself). The original tag pair matrix is ​​represented as follows: After the sparse label pair selection structure is filtered, based on distance filtering rules, semantic relevance filtering rules, and entity type combination rules, some label pairs in the original label pair matrix are removed, and the resulting sparse label pair matrix is ​​represented as follows: Further utilizing the semantic constraint mask matrix, marker pairs that do not satisfy the policy structure boundary or element type matching rules are masked, resulting in the final effective marker pair matrix representation as follows: Step 203: Generate a candidate set of policy semantic elements based on the extraction results, and use the policy ontology to standardize the entity type and relation type of the candidate set of policy semantic elements to generate a standard set of policy semantic elements.

[0047] Extraction results refer to the initial identification results of entities and relationships obtained from structured text sequences through the power policy knowledge extraction model. These typically include entity text fragments (e.g., power generation companies, medium- and long-term transactions) and descriptions of relationships between entities (e.g., participation, applicability, constraint). Policy semantic elements refer to the semantic abstraction results of the original entities and relationships. These are information units that can represent the core semantics of the policy, including policy subject elements, transaction behavior elements, rule constraint elements, time elements, and effective condition elements. The policy semantic element candidate set refers to policy semantic elements compiled based on the extraction results that have not yet undergone semantic type unification and standardization constraints. The policy ontology refers to a standardized knowledge system describing semantic concepts, entity types, and relationship types in the field of power trading policies. It is used to unify semantic terminology, eliminate ambiguity, and ensure structural consistency. It includes predefined entity types (e.g., policy subjects, trading varieties, rule elements), predefined relationship types (e.g., constraint relationships, applicability relationships, triggering relationships), and the hierarchy and constraint rules between each type. Entity type standardization refers to mapping entities with different expressions among candidate semantic elements to standard entity types defined in the policy ontology. For example, power generation enterprises and power generation-side entities are mapped to policy entity classes, and transaction volume declarations and declared electricity volumes are mapped to transaction behavior classes. Relationship type standardization refers to mapping the extracted diverse relationship expressions to standard relationship types defined in the policy ontology. For example, "must meet" and "must not be lower than" are mapped to constraint relationships, and "participate" and "applicable" are mapped to application relationships. The standard set of policy semantic elements refers to a set of policy semantic elements whose entity types and relationship types have been unified, standardized, and computable after being constrained and mapped by the policy ontology. Its semantic types are clear, its structure is consistent, and it can be directly used for multi-domain fusion association graph construction and subsequent reasoning calculations.

[0048] Specifically, after obtaining the entity and relation extraction results from the policy text, the extraction results are organized to form a candidate set of policy semantic elements. Based on the pre-constructed policy ontology, the entities and relations in the candidate set are type-mapped and semantically constrained to achieve unified standardization of entity types and relation types, thereby generating a standard set of policy semantic elements with consistent structure.

[0049] Step 204: Construct a policy knowledge graph based on the standard set of policy semantic elements.

[0050] Specifically, after extracting and standardizing the semantic elements in the policy text, the resulting standard set of policy semantic elements is organized according to the modeling method of knowledge graph. The entities are used as graph nodes and the relationships between entities are used as graph edges, thereby constructing a policy knowledge graph for systematically expressing the semantic elements and their interrelationships in the power trading policy.

[0051] Step 205: Based on the pre-configured transaction ontology, perform entity identification and entity relationship modeling on the power transaction business data to construct a business knowledge graph.

[0052] Step 206: Based on the pre-configured market rule ontology, perform entity identification and entity relationship modeling on the power trading business data to construct a market rule knowledge graph.

[0053] Step 207: Align the entity nodes in the policy knowledge graph, business knowledge graph, and market rule knowledge graph, and construct cross-domain association edges based on the alignment results to generate a multi-domain fusion association graph.

[0054] Optionally, entity nodes in the policy knowledge graph, business knowledge graph, and market rule knowledge graph are aligned, and cross-domain association edges are constructed based on the alignment results to generate a multi-domain fusion association graph. This includes: extracting the name, code, semantic category, role, variety, timeliness, and constraint information of entities in each knowledge graph to construct a cross-domain entity index table; generating a cross-domain entity mapping set based on the cross-domain entity index table through at least one of name matching, code matching, semantic role matching, and attribute matching; establishing cross-domain association edges between corresponding entities in each knowledge graph based on the cross-domain entity mapping set; and merging the three knowledge graphs with established cross-domain association edges to generate a multi-domain fusion association graph.

[0055] Name refers to the textual identifier or label of an entity, used to uniquely identify it. Encoding is the unique identifier or index number corresponding to an entity in the system. Semantic category refers to the conceptual category or type to which an entity belongs, such as policy elements, transaction behaviors, or rule clauses. Role is the function or identity of an entity in a knowledge graph or specific business scenario, such as subject, object, or constraint conditions. Variety is the specific category or attribute classification to which an entity belongs, such as power generation enterprise type or transaction type. Timeliness refers to the effective time range or applicable period of the information corresponding to an entity. Constraint information refers to the restrictions or conditions imposed on an entity in business or policy, such as minimum declaration volume or settlement cycle restrictions. A cross-domain entity index table is a tabular data structure formed by uniformly organizing entity information from knowledge graphs from different domains, used for quick searching, comparison, and establishing cross-domain mapping relationships. A cross-domain entity mapping set is a set formed by matching corresponding or related entities in different knowledge graphs, used for subsequent association establishment. Cross-domain association edges are connections established between corresponding or related entities in different knowledge graphs, used to represent cross-domain relationships between entities. A multi-domain integrated association graph refers to a graph structure formed by merging policy, business, and market rule knowledge graphs after establishing cross-domain association edges, reflecting a comprehensive network of cross-domain entities and their relationships.

[0056] Specifically, the name, code, semantic category, role, variety, timeliness, and constraint information of each entity are extracted from each knowledge graph, and a unified cross-domain entity index table is constructed. Then, corresponding or related entities in different knowledge graphs are paired by matching names, codes, semantic roles, or attributes to generate a cross-domain entity mapping set. Cross-domain association edges are established between corresponding entities in the knowledge graphs based on the mapping set. Finally, the three knowledge graphs with established cross-domain association edges are merged to form a multi-domain fusion association graph, thereby realizing the unified representation and relationship analysis of cross-domain knowledge.

[0057] For example, suppose the policy knowledge graph contains entity "Power Generation Company A" (name: Power Generation Company A, code: P001, semantic category: subject, role: policy implementing entity, type: thermal power, validity period: from 2026 to 2027, business type: medium- and long-term transaction, constraint information: minimum declaration limit); the business knowledge graph contains entity "Medium- and Long-Term Transaction B" (name: Medium- and Long-Term Transaction B, code: B001, semantic category: transaction behavior, role: transaction object, type: transaction type, medium- and long-term, validity period: first quarter of 2026, constraint information: transaction volume limit); and the market rule knowledge graph contains entity "Minimum Declaration Rule C" (name: Minimum Declaration Rule C, code: M001, semantic category: rule clause, role: constraint condition, type: declaration rule, validity period: first quarter of 2026, constraint information: minimum declaration volume 70%). First, extract the name, code, semantic category, role, type, validity period, and constraint information of the entities in each knowledge graph to construct a cross-domain entity index table, as shown in Table 1. Based on this index table, through name matching, encoding matching, and / or semantic role matching, the "Power Generation Enterprise A" in the policy knowledge graph is mapped to the "Medium- and Long-Term Transaction B" in the business knowledge graph and the "Minimum Declaration Quantity Rule C" in the market rule knowledge graph, respectively, generating a cross-domain entity mapping set. Each mapping element contains a source entity, a target entity, and matching criteria. For example, the business types of "Power Generation Enterprise A" (policy) and "Medium- and Long-Term Transaction B" (business) are consistent, and their timeliness is matched; the business types of "Medium- and Long-Term Transaction B" (business) and "Minimum Declaration Quantity Rule C" (market rule) are consistent, and their constraint information is matched. According to this mapping relationship, cross-domain association edges are established between the corresponding entities in each knowledge graph. For example, an association edge is established between the policy node "Power Generation Enterprise A" and the business node "Medium- and Long-Term Transaction B" to show that policy influences transaction behavior, and an association edge is established between the business node "Medium- and Long-Term Transaction B" and the rule node "Minimum Declaration Quantity Rule C" to show that transaction behavior is constrained by market rules. Finally, the three types of knowledge graphs with established cross-domain association edges are merged to generate a multi-domain fusion association graph, thereby achieving a unified representation and analysis of cross-domain entities and their relationships. Step 208: In the multi-domain fusion association graph, perform path reasoning search with policy nodes as the starting point and transaction nodes as the ending point to generate a policy transaction association link set.

[0058] Optionally, in the multi-domain fusion association graph, path reasoning search is performed with policy nodes as the starting point and transaction nodes as the ending point to generate a set of policy-transaction association links, including: From the multi-domain fusion association graph, nodes belonging to the policy domain are identified to form the starting node set, and nodes belonging to the transaction domain are identified to form the target node set. Search the multi-domain fusion association graph for all reachable paths from the starting set of nodes to the target set of nodes; For each reachable path found, record the node sequence, relation sequence, and path length, and obtain the weight values ​​of each node and relation on the path; Calculate the initial score for each reachable path based on node weights and relationship weights; Based on preset scoring thresholds and path validity constraints, candidate paths are selected from all reachable paths. Path validity constraints include consistency checks on node type, relationship type, and direction. After merging the selected candidate paths, they are categorized semantically to generate policy impact paths and policy constraint paths respectively.

[0059] Nodes in the policy domain refer to nodes in the multi-domain fusion association graph that originate from the policy knowledge graph, representing policy-related entities such as policy implementers or policy clauses. The starting node set is the set of all policy domain nodes in the multi-domain fusion association graph, used as the starting point for path searching. Nodes in the transaction domain refer to nodes in the multi-domain fusion association graph that originate from the business knowledge graph or transaction-related entities, such as transaction behaviors or transaction objects. The target node set is the set of all transaction domain nodes in the multi-domain fusion association graph, used as the endpoint for path searching. A reachable path is a continuous path of nodes and edges from a starting node to a target node in the multi-domain fusion association graph. A node sequence is a list of nodes traversed sequentially along a path. A relation sequence is a list of edges (relationships between nodes) traversed sequentially along a path. Path length is the number of nodes or edges along a path, used to measure the complexity or distance of the path. Node weight value is the importance score of each node, usually obtained from normalized node weight parameters. Relationship weight value is the importance score of each edge, usually obtained from normalized relation weight parameters. The initial score refers to the overall score of the path calculated based on node weights and relationship weights, used to measure the potential policy impact or importance of the path. The scoring threshold is a preset lower limit for the score, used to filter paths with significant policy impact. Path legitimacy constraints refer to the legality requirements for the path, including consistency checks on node type, relationship type, path length, and direction. Candidate paths are the set of paths retained after scoring and legitimacy screening, representing paths with potential policy impact. Semantic classification refers to classifying candidate paths according to semantic meaning, such as distinguishing between policy-impacting paths and policy-constrained paths.

[0060] Specifically, the process involves identifying all policy-related nodes forming the starting node set and transaction-related nodes forming the target node set from the multi-domain fusion association graph. Then, all reachable paths from the starting nodes to the target nodes are searched within the multi-domain fusion association graph. Each path consists of nodes and relationships, and the node sequence, relationship sequence, and path length are recorded. The weight values ​​of each node and relationship on the path are obtained, and an initial score is calculated for each path to measure its policy impact. Meaningful candidate paths are then selected based on a score threshold and path legitimacy constraints, ensuring that the paths are logically consistent in length, type, and direction. These candidate paths are then semantically categorized to generate policy impact paths and policy constraint paths, used to analyze the role and limitations of policies in electricity trading.

[0061] For example, assuming that in a multi-domain fusion association graph, policy node A (power generation enterprise) belongs to the policy domain, and transaction nodes B and C (medium- and long-term transactions, transaction volume declarations) belong to the transaction domain, firstly, policy node A is identified as the starting node set, and transaction nodes B and C are identified as the target node set. Then, all reachable paths from the starting node set to the target node set are searched in the multi-domain fusion association graph, resulting in two paths: path 1 is A→B, and path 2 is A→B→C. The node sequence, relation sequence, and path length of each path are recorded. Simultaneously, the weight values ​​of each node and relation on the path are obtained. After normalization, the node weight values ​​for path 1 are A=0.53 and B=0.47, and for path 2 are A=0.38, B=0.34, and C=0.28. After normalization, the relation weight values ​​for path 1 are A→B=1, and for path 2 are A→B=0.52 and B→C=0.28. C=0.48; Assuming the importance coefficients of each node are A=0.4, B=0.3, and C=0.2, and the importance coefficients of each relationship are A→B=0.6 and B→C=0.5, then the initial score of the node for path 1 is 0.365 (0.53×0.4+0.47×0.3=0.365), the initial score of the relationship is 0.6 (1×0.6=0.6), and the initial score of path 1 is 0.965 (0.365+0.6≈0.965); the initial score of the node for path 2 is 0.269 (0.38×0.4+0.34×0.3+0.28×0.2≈0.269), the initial score of the relationship is 0.554 (0.52×0.6+0.48×0.5≈0.554), and the initial score of path 2 is 0.823 (0.269+0.554=0.823). Assuming a scoring threshold of 0.8, path legality constraints (path length not exceeding 6), node type requirements (starting node must be a policy node, target node a transaction node, intermediate nodes can be policy nodes, transaction nodes, or market rule nodes), path relationship type requirements (each relationship on the path must be a legal policy-business, business-business, or business-rule association), and path direction requirements (pointing from the starting node to the target node in a consistent order), then based on the recorded node sequence, relationship sequence, and path length, both path 1 and path 2 are identified as candidate paths. Subsequently, based on the semantic structure of the paths, path 1 can be labeled as a policy-influenced path, and path 2 can be labeled as a policy-constrained path, thus forming a set of policy-influenced paths and a set of policy-constrained paths, achieving path identification, scoring calculation, candidate path selection, and semantic classification from the starting node to the target node.

[0062] Step 209: For each link in the policy transaction link set, extract path structure parameters and node and relationship weight parameters from the multi-domain fusion link graph, and calculate the policy impact intensity score for each link based on the extracted parameters to generate a policy impact intensity result set.

[0063] Optionally, the path structure parameters include the number of path hops, the number of cross-domain jumps, and the number of path interruptions in the associated links; the node and relationship weight parameters include the set of weight values ​​of all nodes in the associated links and the set of strength values ​​of all relationship edges; the policy impact intensity score of each associated link is calculated based on the extracted parameters, including: calculating the path connectivity index based on the number of path hops, the number of cross-domain jumps, and the number of path interruptions; and inputting the set of weight values ​​of nodes, the set of strength values ​​of relationship edges, and the path connectivity index into a preset scoring function to calculate the policy impact intensity score of the corresponding associated link.

[0064] Path hop count refers to the total number of hops between nodes in an association link from the starting node to the ending node. Cross-domain hop count refers to the number of hops that occur when nodes cross different knowledge graphs in an association link. Path interruption count refers to the number of hop segments that do not meet the continuous validity condition when verifying the continuity of adjacent nodes and relationships in the path during the impact strength calculation phase. Node and relationship weight parameters refer to the set of weight values ​​for each node and the set of strength values ​​for each relationship edge in the association link, used to reflect the importance of nodes and relationships. Node weight value refers to the weight value of a node's overall impact on the path in the association link, usually after normalization. Relationship edge strength value refers to the weight value of each relationship edge in the association link's overall impact on the path, usually also after normalization. Path connectivity index refers to an index reflecting the integrity and continuity of the path structure, calculated based on path hop count, cross-domain hop count, and path interruption count. Policy impact strength score refers to the calculation result that comprehensively considers path structure, node weights, and relationship strength, used to quantify the strength of the policy impact on a certain association link. The preset scoring function refers to a mathematical function or algorithm used to transform the set of node weight values, the set of relationship edge strength values, and the path connectivity index into a score of policy impact intensity.

[0065] Specifically, path structure parameters (including path hop count, cross-domain jump count, and path interruption count), node weight parameters, and relationship weight parameters describing the overall characteristics of the path are extracted from the associated links. A path connectivity index is calculated based on these path structure parameters. Then, the set of node weight values, the set of relationship edge strength values, and the path connectivity index are input into a preset scoring function to comprehensively calculate the policy impact strength score for each associated link, thereby quantifying the degree to which the link is affected by the policy.

[0066] For example, suppose there is a policy transaction link with the following structure: Policy Node A (Power Generation Enterprise) → Business Node B (Medium- and Long-Term Transactions) → Business Node C (Transaction Volume Declaration) → Market Rule Node D (Minimum Declaration Quantity Rule) → Market Rule Node E (Settlement Cycle Rule), totaling 5 nodes. In the multi-domain fusion association graph, the path structure parameters of this link are extracted, where the path hop count is 4, the number of cross-domain hops is 3, and the number of path interruptions is 1; the node weight parameters are A=0.9, B=0.8, C=0.7, D=0.85, and E=0.9, respectively, corresponding to node importance coefficients of A=0.4, B=0.3, C=0.3, D=0.3, and E=0.3; the relationship weight parameters are A→B=0.88, B→C=0.8, C→D=0.86, and D→E=0.88. The importance coefficients for the corresponding relationships are 0.9, A→B=0.35, B→C=0.3, C→D=0.32, and D→E=0.33 respectively; the path importance coefficient is 0.25, and the cross-domain consistency score is 1 (the cross-domain consistency score is determined based on whether the number of cross-domain jumps and cross-domain mapping relationships in the path are consistent. When each cross-domain mapping relationship in the path can be matched one-to-one in the cross-domain entity index table, the cross-domain consistency score is set to 1. When there is no matching or a matching conflict, the cross-domain consistency score is set to 0). The path connectivity index is calculated as 0.75 (3 ÷ 4 × 1 = 0.75) based on (effective consecutive hops (4-1=3) ÷ path hops) × cross-domain consistency score. Multiplying the path connectivity index by the path importance coefficient yields a path score of 0.1875 (0.75 × 0.25 = 0.1875). Normalizing the node weight parameters yields: A≈0.217, B≈0.193, C≈0.169, D≈0.205, E≈0.217. Normalizing the relationship weight parameters yields: A→B≈0.256, B→C≈0.2. 33. C→D≈0.250, D→E≈0.262; The node score calculated based on the normalized node weight parameters is 0.256 (0.217×0.4+0.193×0.3+0.169×0.3+0.205×0.3+0.217×0.3≈0.262), and the relationship score calculated based on the normalized relationship weight parameters is 0.322 (0.256×0.35+0.233×0.3+0.250×0.32+0.262×0.33≈0.322). Adding the path score, node score, and relationship score, the policy impact strength score for the corresponding associated link is 0.7715. Therefore, this policy has a high degree of influence on the target transaction behavior and can be marked as high-impact in the policy impact strength results set.

[0067] Step 210: Based on the trigger logic elements identified from the power trading policy text data, perform trigger condition mapping and structured integration on the policy trading association link set and the policy impact intensity result set to generate a policy association identification result set.

[0068] Optionally, based on the triggering logic elements identified from the power trading policy text data, triggering condition mapping and structured integration are performed on the policy trading association link set and the policy impact intensity result set. This includes: extracting triggering logic elements from the standard set of policy semantic elements obtained during the construction of the policy knowledge graph; parsing the triggering logic elements to obtain triggering condition text, logical connection relationships, and scope of application information; decomposing the triggering condition text into a structured set of condition items; locating the affected trading nodes in the policy trading association link set based on the scope of application information and establishing a mapping relationship between the set of condition items and the corresponding links; encoding each condition item in the set of condition items into executable strategy triggering conditions according to the logical connection relationships; integrating the strategy triggering conditions, the corresponding association link information, and the impact intensity information obtained from the policy impact intensity result set using policy identifiers and trading behavior identifiers as association keys to generate structured policy association identification result records; summarizing all policy association identification result records, establishing an index, and generating a policy association identification result set.

[0069] Triggering logic elements refer to the combination of semantic elements describing the applicable conditions and triggering rules of a policy, obtained based on the standard set of policy semantic elements. These elements characterize the logical conditions under which a policy is activated and generates constraints or effects in a specific transaction scenario. Triggering condition text refers to the original or standardized textual expression directly describing the content of the policy triggering conditions, obtained after parsing the triggering logic elements. Logical connection relationships refer to the relational information describing the logical combination methods between multiple triggering conditions, such as AND, OR, and NOT logical relationships. Scope of application information refers to the semantic information limiting the applicable objects and scenarios of the triggering conditions, such as the transaction type, transaction entity, time interval, or business scope to which the policy applies. A structured set of condition items refers to a set of multiple independently judgeable structured condition items formed after decomposing the triggering condition text; each condition item corresponds to a specific constraint judgment unit. Mapping relationships refer to the association relationship between each condition item and the affected transaction node or link. Strategy triggering conditions refer to the structured rule expression formed by combining and encoding the set of condition items according to logical connection relationships, which can be used to automatically determine whether a policy has been triggered. Policy identifiers refer to the unique identifier information that identifies a specific policy text or policy clause. Transaction behavior identifiers are unique identifiers used to identify specific transaction behaviors or transaction types. Policy association identification result records are structured records formed by integrating policy identifiers and transaction behavior identifiers as association keys, along with information on policy triggering conditions, association links, and the intensity of policy impact. Policy association identification result sets are collections of policy association identification result records that have been aggregated and indexed, used to support unified management and querying of the relationships between policies and transaction behaviors.

[0070] Specifically, standardized policy semantic elements are obtained from the policy knowledge graph, and triggering logic elements are extracted. These are then parsed to obtain the triggering condition text, the logical connections between conditions, and the scope of application. Next, the triggering condition text is decomposed into a set of structured condition items, and affected transaction nodes are located in the policy transaction association chain set according to the scope of application, establishing a mapping relationship between condition items and chains. Then, based on the logical connections, the condition items are combined and encoded into executable policy triggering conditions. Finally, using policy identifiers and transaction behavior identifiers as unified association keys, the policy triggering conditions, association chain information, and corresponding impact strengths are integrated to form a policy association identification result record. All records are then aggregated to create an index, generating a complete policy association identification result set, enabling the system to quickly determine the specific impact of policies on transaction behavior.

[0071] For example, suppose that during the construction of a policy knowledge graph, the following standard set of policy semantic elements is extracted and standardized from a certain power trading policy text, including the subject element "power generation enterprise", the transaction behavior elements "medium- and long-term transactions" and "electricity declaration", and the rule constraint elements "minimum declaration ratio" and "70%". Based on this, the triggering logic element is obtained, and its corresponding original policy text is "When power generation enterprises participate in medium- and long-term transactions, the declared electricity volume shall not be less than 70% of the contract electricity volume". Parsing the triggering logic element, the resulting triggering condition text is "When power generation enterprises participate in medium- and long-term transactions, the declared electricity volume shall not be less than 70% of the contract electricity volume", and the logical connection relationship is a single condition constraint (no parallel or selection logic), indicating that the condition is a mandatory condition that must be met simultaneously, and the scope of application information is "the electricity declaration behavior of power generation enterprises participating in medium- and long-term transactions". Subsequently, the triggering condition text is decomposed into a structured set of condition items {condition item 1: transaction type = medium- and long-term transaction, condition item 2: transaction subject type = power generation enterprise, condition item 3: declared electricity volume ≥ contract electricity volume × 70%}. Based on the scope of application information, the affected transaction nodes in the policy transaction association link set are located as transaction nodes representing "electricity declaration behavior in medium and long-term transactions". Policy transaction association links related to "electricity declaration in medium and long-term transactions" are determined, and a mapping relationship is established between the set of condition items and the corresponding association links. Then, according to the logical connection relationship, the condition items in the condition item set are combined and encoded into executable strategy triggering conditions. For example, when the transaction type is a medium and long-term transaction and the declared electricity is less than 70% of the contracted electricity, the policy constraint is triggered. Next, using the policy identifier corresponding to the policy and the transaction behavior identifier corresponding to the target transaction behavior as association keys, the strategy triggering conditions, the corresponding policy transaction association link information, and the impact intensity score information obtained from the policy impact intensity result set are integrated to generate a structured policy association identification result record. This record simultaneously includes the policy number, triggering condition rule, associated transaction behavior identifier, and corresponding impact intensity level, for example, [Policy number: P2026-01; Associated transaction behavior: medium and long-term transaction electricity declaration; Strategy triggering condition: declared electricity ≥ contracted electricity × 70%; Policy impact intensity level: high]. Finally, the policy association identification results generated above are summarized and indexed to form a policy association identification result set, which is used for subsequent judgment of policy compliance and impact analysis of transaction behavior.

[0072] In this embodiment, by preprocessing the power trading policy text data into a structured text sequence, the policy text format can be unified, providing standardized input for subsequent automated processing. Utilizing a power policy knowledge extraction model to jointly extract entities and relationships from the structured text sequence efficiently identifies key elements and their interrelationships within the policy text, improving the accuracy and completeness of information extraction. Based on the extraction results, a candidate set of policy semantic elements is generated. Furthermore, the policy ontology is used to standardize the entity and relationship types of this candidate set, generating a standard set of policy semantic elements. This standardizes element representation and type definitions, reduces ambiguity, and provides a foundation for knowledge graph construction. This provides a reliable foundation; based on a standard set of policy semantic elements, a policy knowledge graph is constructed, which can represent entities and their relationships in power trading policies in a structured, queryable, and reasonable manner, providing data support for subsequent policy analysis, cross-domain correlation, and intelligent reasoning; based on a pre-configured transaction ontology, entity identification and entity relationship modeling are performed on power trading business data to construct a business knowledge graph, which can represent transaction-related entities and their interrelationships in power trading business data in a structured manner, providing a data foundation for subsequent cross-domain correlation and reasoning; based on a pre-configured market rule ontology, entity identification and entity relationship modeling are performed on power trading business data to construct a market rule ontology. The market rule knowledge graph can structurally represent market rule-related entities and their relationships in power trading business data, providing a data foundation for subsequent cross-domain association and reasoning. It aligns entity nodes in policy, business, and market rule knowledge graphs and constructs cross-domain association edges based on the alignment results to generate a multi-domain fusion association graph. This enables cross-domain fusion of policy, business, and market rule knowledge, clarifying the correspondence and potential associations between entities in different knowledge graphs, allowing for unified representation and analysis of multi-domain information, and providing a complete and unified knowledge association network for subsequent policy transaction association link and impact analysis. In the multi-domain fusion association graph, path reasoning search is performed with policy nodes as the starting point and transaction nodes as the ending point to generate a policy-transaction association link set. This can identify the potential impact paths of policies on transaction behavior and generate a structured policy-transaction association link set, providing a clear analytical object for calculating the policy impact intensity. For each association link in the policy-transaction association link set, path structure parameters and node and relationship weight parameters are extracted from the multi-domain fusion association graph. Based on the extracted parameters, the policy impact intensity score of each association link is calculated, generating a policy impact intensity result set. This can quantify the degree of impact of different policies on transaction links and provide a basis for policy prioritization and risk assessment.Based on the triggering logic elements identified from power trading policy text data, trigger condition mapping and structured integration are performed on the policy trading association link set and the policy impact intensity result set to generate a policy association identification result set. This set integrates policy triggering conditions with trading behavior association links and impact intensity results, forming structured policy association identification results. This enables intelligent association identification between power trading policies and trading behaviors, providing support for intelligent policy implementation and compliance analysis.

[0073] Figure 3 This is a schematic diagram of a power trading policy association identification device provided in an embodiment of the present invention, as shown below. Figure 3 As shown, the device includes: The policy graph construction module 301 is used to perform entity recognition and entity relationship modeling on power trading policy text data based on pre-configured policy ontology, and to construct a policy knowledge graph. The business graph construction module 302 is used to perform entity identification and entity relationship modeling on power transaction business data based on a pre-configured transaction ontology, and to construct a business knowledge graph. The market graph construction module 303 is used to perform entity identification and entity relationship modeling on power trading business data based on a pre-configured market rule ontology, and to construct a market rule knowledge graph. The fusion graph generation module 304 is used to align entity nodes in the policy knowledge graph, business knowledge graph and market rule knowledge graph, and construct cross-domain association edges based on the alignment results to generate a multi-domain fusion association graph. The association link generation module 305 is used to perform path reasoning search in the multi-domain fusion association graph, with policy nodes as the starting point and transaction nodes as the ending point, to generate a set of policy and transaction association links. The association link scoring module 306 is used to extract path structure parameters and node and relationship weight parameters from the multi-domain fusion association graph for each association link in the policy transaction association link set, and calculate the policy impact intensity score of each association link based on the extracted parameters, and generate a policy impact intensity result set. The identification module generation module 307 is used to perform trigger condition mapping and structured integration on the policy transaction association link set and the policy impact intensity result set based on the trigger logic elements identified from the power transaction policy text data, and generate a policy association identification result set.

[0074] In one embodiment, the policy graph construction module 301 is specifically used for: The power trading policy text data was preprocessed into a structured text sequence; A power policy knowledge extraction model is used to jointly extract entities and relations from structured text sequences. Based on the extraction results, a candidate set of policy semantic elements is generated, and the policy ontology is used to standardize the entity type and relation type of the candidate set of policy semantic elements to generate a standard set of policy semantic elements. A policy knowledge graph is constructed based on a standard set of policy semantic elements.

[0075] In one embodiment, the power policy knowledge extraction model includes a sparse label pair selection structure and a semantic constraint mask structure. The policy graph construction module 301 uses the power policy knowledge extraction model to jointly extract entities and relations from the structured text sequence, including: The sparse label pair selection structure filters the original label pair matrix generated after model encoding according to preset distance filtering rules, semantic relevance filtering rules, and entity type combination rules to obtain a sparse label pair matrix. Among them, the distance filtering rule limits the relative distance between the two labels in the combined label pair to no more than a preset distance threshold, the semantic relevance filtering rule limits the semantic vector similarity of the combined label pair to no less than a preset similarity threshold, and the entity type combination rule limits the entity type or structure type corresponding to the combined label pair to belong to a preset valid combination list. Based on the structural features and element types of policy texts, a semantic constraint mask matrix is ​​generated using a semantic constraint mask structure. This matrix is ​​then used to mask label pairs that cross policy structure boundaries or whose element type combinations do not conform to preset type combination rules. This process performs a secondary screening of the sparse label pair matrix to obtain an effective label pair matrix.

[0076] In one embodiment, the fusion map generation module 304 is specifically used for: Extract the name, code, semantic category, role, type, timeliness and constraint information of entities in each knowledge graph, and construct a cross-domain entity index table; Based on the cross-domain entity index table, a cross-domain entity mapping set is generated through at least one of the following methods: name matching, encoding matching, semantic role matching, and attribute matching. Based on the cross-domain entity mapping set, cross-domain association edges are established between corresponding entities in each knowledge graph; The three knowledge graphs that have established cross-domain association edges are merged to generate a multi-domain fusion association graph.

[0077] In one embodiment, the association link generation module 305 is specifically used for: From the multi-domain fusion association graph, nodes belonging to the policy domain are identified to form the starting node set, and nodes belonging to the transaction domain are identified to form the target node set. Search the multi-domain fusion association graph for all reachable paths from the starting set of nodes to the target set of nodes; For each reachable path found, record the node sequence, relation sequence, and path length, and obtain the weight values ​​of each node and relation on the path; Calculate the initial score for each reachable path based on node weights and relationship weights; Based on preset scoring thresholds and path validity constraints, candidate paths are selected from all reachable paths. Path validity constraints include consistency checks on node type, relationship type, and direction. After merging the selected candidate paths, they are categorized semantically to generate policy impact paths and policy constraint paths respectively.

[0078] In one embodiment, the path structure parameters include the path hop count, cross-domain jump count, and path interruption count of the associated links; the node and relationship weight parameters include the set of weight values ​​for all nodes on the associated links and the set of strength values ​​for all relationship edges; the associated link scoring module 306 calculates the policy impact strength score for each associated link based on the extracted parameters, including: Calculate the path connectivity index based on path hop count, cross-domain jump count, and path interruption count; By inputting the set of node weight values, the set of relation edge strength values, and the path connectivity index into a preset scoring function, the policy impact strength score of the corresponding associated link is calculated.

[0079] In one embodiment, the identification module generation module 307 is specifically used for: Triggering logic elements are extracted from the standard set of policy semantic elements obtained during the construction of the policy knowledge graph; Analyze the trigger logic elements to obtain the trigger condition text, logical connection relationship, and applicable scope information; Decompose the trigger condition text into a structured set of condition items; Based on the scope of application information, the affected transaction nodes are located in the policy transaction linkage, and a mapping relationship between the set of condition items and the corresponding linkage is established. Based on the logical connection relationship, the condition items in the condition item set are combined and encoded into executable strategy trigger conditions; Using policy identifiers and transaction behavior identifiers as association keys, the policy triggering conditions, corresponding association link information, and impact intensity information obtained from the policy impact intensity result set are integrated to generate a structured policy association identification result record; Summarize all policy association identification results records, create an index, and generate a policy association identification result set.

[0080] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional modules is merely an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working process of the functional modules described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0081] The apparatus of this invention performs entity recognition and entity relationship modeling on power trading policy text data based on a pre-configured policy ontology to construct a policy knowledge graph. This enables a structured representation of policy-related entities and their relationships within the power trading policy text data, providing a data foundation for subsequent cross-domain association and reasoning. Similarly, it performs entity recognition and entity relationship modeling on power trading business data based on a pre-configured trading ontology to construct a business knowledge graph. This enables a structured representation of trading-related entities and their relationships within the power trading business data, providing a data foundation for subsequent cross-domain association and reasoning. The system uses a market rule ontology to perform entity identification and entity relationship modeling on electricity trading business data, constructing a market rule knowledge graph. This graph provides a structured representation of market rule-related entities and their relationships within the electricity trading business data, laying the data foundation for subsequent cross-domain association and reasoning. Furthermore, it aligns entity nodes in the policy knowledge graph, business knowledge graph, and market rule knowledge graph, and constructs cross-domain association edges based on the alignment results to generate a multi-domain fusion association graph. This enables cross-domain fusion of policy, business, and market rule knowledge, clarifying the correspondence and potential associations between entities in different knowledge graphs, and allowing multi-domain information to be integrated. This system provides a unified representation and analysis, offering a complete and unified knowledge network for subsequent policy transaction correlation and impact analysis. Within the multi-domain fusion correlation graph, path reasoning searches are performed with policy nodes as the starting point and transaction nodes as the ending point, generating a policy transaction correlation link set. This identifies potential impact paths of policies on transaction behavior, generating a structured policy transaction correlation link set and providing a clear analytical object for calculating policy impact intensity. For each correlation link in the policy transaction correlation link set, path structure parameters and node and relationship weight parameters are extracted from the multi-domain fusion correlation graph. Based on the extracted parameters, a policy impact intensity score is calculated for each correlation link, generating a policy impact intensity result set. This quantifies the degree of impact of different policies on transaction links, providing a basis for policy prioritization and risk assessment. Based on triggering logic elements identified from power transaction policy text data, trigger condition mapping and structured integration are performed on the policy transaction correlation link set and the policy impact intensity result set, generating a policy correlation identification result set. This integrates policy triggering conditions with transaction behavior correlation links and impact intensity results, forming a structured policy correlation identification result. This achieves intelligent correlation identification between power transaction policies and transaction behavior, supporting intelligent policy execution and compliance analysis.

[0082] The following is for reference. Figure 4 It shows a schematic diagram of the structure of a computer system 400 suitable for implementing an electronic device according to embodiments of the present invention. Figure 4 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0083] likeFigure 4 As shown, the computer system 400 includes a central processing unit (CPU) 401, which can perform various appropriate actions and processes based on programs stored in read-only memory (ROM) 402 or programs loaded from storage section 408 into random access memory (RAM) 403. The RAM 403 also stores various programs and data required for the operation of the computer system 400. The CPU 401, ROM 402, and RAM 403 are interconnected via a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0084] The following components are connected to I / O interface 405: input section 406 including keyboard, mouse, etc.; output section 407 including cathode ray tube, liquid crystal display, etc., and speakers, etc.; storage section 408 including hard disk, etc.; and communication section 409 including network interface card, such as modem, etc. Communication section 409 performs communication processing via a network such as the Internet. Drive 410 is also connected to I / O interface 405 as needed. Removable media 411, such as disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 410 as needed so that computer programs read from them can be installed into storage section 408 as needed.

[0085] In particular, according to the embodiments disclosed in this invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 409, and / or installed from removable medium 411. When the computer program is executed by central processing unit (CPU) 401, it performs the functions defined above in the system of this invention.

[0086] It should be noted that the computer-readable medium shown in this invention can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical fiber, a portable compact disk read-only memory, an optical storage device, a magnetic storage device, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this invention, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, etc., or any suitable combination thereof.

[0087] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0088] The modules and / or units described in the embodiments of this invention can be implemented in software or hardware. The described modules and / or units can also be housed in a processor; for example, a processor may include a policy graph construction module, a business graph construction module, a market graph construction module, a fusion graph generation module, a link generation module, a link scoring module, and an identification module generation module. The names of these modules do not necessarily limit the module itself.

[0089] In another aspect, the present invention also provides a computer-readable medium, which may be included in the device described in the above embodiments; or it may exist independently and not assembled into the device. The computer-readable medium carries one or more programs, which, when executed by the device, cause the device to include: Based on a pre-configured policy ontology, entity recognition and entity relationship modeling are performed on power trading policy text data to construct a policy knowledge graph. Based on a pre-configured trading ontology, entity recognition and entity relationship modeling are performed on power trading business data to construct a business knowledge graph. Based on a pre-configured market rule ontology, entity recognition and entity relationship modeling are performed on power trading business data to construct a market rule knowledge graph. Entity nodes in the policy knowledge graph, business knowledge graph, and market rule knowledge graph are aligned, and cross-domain association edges are constructed based on the alignment results to generate a multi-domain fusion association graph. In the multi-domain fusion association graph, path reasoning search is performed with policy nodes as the starting point and trading nodes as the ending point to generate a policy-transaction association link set. For each association link in the policy-transaction association link set, path structure parameters and node and relationship weight parameters are extracted from the multi-domain fusion association graph, and the policy impact intensity score of each association link is calculated based on the extracted parameters to generate a policy impact intensity result set. Based on the trigger logic elements identified from the power trading policy text data, trigger condition mapping and structured integration are performed on the policy-transaction association link set and the policy impact intensity result set to generate a policy association identification result set.

[0090] The technical solution of this invention, by performing entity recognition and entity relationship modeling on power trading policy text data based on a pre-configured policy ontology, constructs a policy knowledge graph. This enables a structured representation of policy-related entities and their relationships within the power trading policy text data, providing a data foundation for subsequent cross-domain association and reasoning. Similarly, by performing entity recognition and entity relationship modeling on power trading business data based on a pre-configured trading ontology, this constructs a business knowledge graph. This enables a structured representation of trading-related entities and their relationships within the power trading business data, providing a data foundation for subsequent cross-domain association and reasoning. The first step involves configuring a market rule ontology to perform entity identification and entity relationship modeling on electricity trading business data, constructing a market rule knowledge graph. This graph provides a structured representation of market rule-related entities and their relationships within the electricity trading business data, laying the data foundation for subsequent cross-domain association and reasoning. Entity nodes in the policy knowledge graph, business knowledge graph, and market rule knowledge graph are aligned, and cross-domain association edges are constructed based on the alignment results to generate a multi-domain fusion association graph. This enables cross-domain fusion of policy, business, and market rule knowledge, clarifying the correspondence and potential associations between entities in different knowledge graphs, and facilitating multi-domain information integration. It can uniformly represent and analyze policies, providing a complete and unified knowledge network for subsequent policy transaction link and impact analysis. In the multi-domain fusion association graph, path reasoning search is performed with policy nodes as the starting point and transaction nodes as the ending point to generate a policy transaction link set. This identifies potential impact paths of policies on transaction behavior, generating a structured policy transaction link set and providing a clear analytical object for calculating policy impact intensity. For each link in the policy transaction link set, path structure parameters and node and relationship weight parameters are extracted from the multi-domain fusion association graph. Based on the extracted parameters, a policy impact intensity score is calculated for each link, generating a policy impact intensity result set. This quantifies the degree of impact of different policies on transaction links, providing a basis for policy prioritization and risk assessment. Based on the triggering logic elements identified from power transaction policy text data, trigger condition mapping and structured integration are performed on the policy transaction link set and the policy impact intensity result set to generate a policy association identification result set. This integrates policy triggering conditions with transaction behavior link and impact intensity results to form a structured policy association identification result, achieving intelligent association identification between power transaction policies and transaction behavior, and supporting intelligent policy implementation and compliance analysis.

[0091] This invention also provides a computer program product, including a computer program that, when executed by a processor, implements the power trading policy association identification method as provided in any embodiment of this invention.

[0092] In the implementation of a computer program product, computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages ​​as well as conventional procedural programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including local area networks (LANs) or wide area networks (WANs), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0093] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0094] It should be noted that the collection, use, storage, sharing, and transfer of user personal information involved in the technical solution of this invention all comply with the provisions of relevant laws and regulations, and require notification to the user and obtaining the user's consent or authorization. Where applicable, user personal information has undergone de-identification and / or anonymization and / or encryption technical processing. In addition, a corresponding operation entry is provided for the user to choose to agree to or reject the automated decision result; if the user chooses to reject, the process proceeds to the expert decision-making process.

[0095] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can occur depending on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for identifying the correlation between electricity trading policies, characterized in that, include: Based on a pre-configured policy ontology, entity recognition and entity relationship modeling are performed on power trading policy text data to construct a policy knowledge graph; Based on a pre-configured transaction ontology, entity identification and entity relationship modeling are performed on power transaction business data to construct a business knowledge graph; Based on a pre-configured market rule ontology, entity identification and entity relationship modeling are performed on the power trading business data to construct a market rule knowledge graph. The entity nodes in the policy knowledge graph, the business knowledge graph, and the market rule knowledge graph are aligned, and cross-domain association edges are constructed based on the alignment results to generate a multi-domain fusion association graph. In the multi-domain fusion association graph, a path reasoning search is performed with policy nodes as the starting point and transaction nodes as the ending point to generate a policy-transaction association link set. For each link in the policy transaction link set, path structure parameters and node and relationship weight parameters are extracted from the multi-domain fusion link graph. Based on the extracted parameters, the policy impact intensity score of each link is calculated to generate a policy impact intensity result set. Based on the trigger logic elements identified from the power trading policy text data, trigger condition mapping and structured integration are performed on the policy trading association link set and the policy impact intensity result set to generate a policy association identification result set.

2. The method according to claim 1, characterized in that, Based on a pre-configured policy ontology, entity identification and entity relationship modeling are performed on power trading policy text data to construct a policy knowledge graph, including: The power trading policy text data is preprocessed into a structured text sequence; The structured text sequence is jointly extracted using a power policy knowledge extraction model; Based on the extraction results, a candidate set of policy semantic elements is generated, and the policy ontology is used to standardize the entity type and relation type of the candidate set of policy semantic elements to generate a standard set of policy semantic elements. The policy knowledge graph is constructed based on the standard set of policy semantic elements.

3. The method according to claim 2, characterized in that, The power policy knowledge extraction model includes a sparse label pair selection structure and a semantic constraint mask structure. It utilizes this model to jointly extract entities and relations from the structured text sequence, including: The sparse label pair selection structure uses preset distance filtering rules, semantic relevance filtering rules, and entity type combination rules to filter the original label pair matrix generated after model encoding, resulting in a sparse label pair matrix. The distance filtering rules limit the relative distance between two labels in a combined label pair to no more than a preset distance threshold; the semantic relevance filtering rules limit the semantic vector similarity of the combined label pairs to no less than a preset similarity threshold; and the entity type combination rules limit the entity type or structure type corresponding to the combined label pairs to belong to a preset valid combination list. Based on the structural features and element types of the policy text, the semantic constraint mask structure is used to generate a semantic constraint mask matrix. The semantic constraint mask matrix is ​​then used to shield the marker pairs that cross the policy structure boundary or whose element type combinations do not conform to the preset type combination rules, so as to perform a secondary screening of the sparse marker pair matrix and obtain an effective marker pair matrix.

4. The method according to claim 1, characterized in that, Aligning entity nodes in the policy knowledge graph, the business knowledge graph, and the market rule knowledge graph, and constructing cross-domain association edges based on the alignment results to generate a multi-domain fusion association graph, including: Extract the name, code, semantic category, role, type, timeliness and constraint information of entities in each knowledge graph, and construct a cross-domain entity index table; Based on the cross-domain entity index table, a cross-domain entity mapping set is generated through at least one of the following methods: name matching, encoding matching, semantic role matching, and attribute matching. Based on the cross-domain entity mapping set, cross-domain association edges are established between corresponding entities in each knowledge graph; The three knowledge graphs that have established cross-domain association edges are merged to generate the multi-domain fusion association graph.

5. The method according to claim 1, characterized in that, In the multi-domain fusion association graph, path reasoning search is performed with policy nodes as the starting point and transaction nodes as the ending point to generate a set of policy-transaction association links, including: From the multi-domain fusion association graph, nodes belonging to the policy domain are identified to form a starting node set, and nodes belonging to the transaction domain are identified to form a target node set. Search the multi-domain fusion association graph for all reachable paths from the starting node set to the target node set; For each reachable path found, record the node sequence, relation sequence, and path length, and obtain the weight values ​​of each node and relation on the path; Calculate the initial score for each reachable path based on node weights and relationship weights; Based on a preset scoring threshold and path validity constraints, candidate paths are selected from all reachable paths. The path validity constraints include consistency checks on node type, relationship type, and direction. After merging the selected candidate paths, they are categorized semantically to generate policy impact paths and policy constraint paths respectively.

6. The method according to claim 1, characterized in that, The path structure parameters include the number of path hops, cross-domain jumps, and path interruptions of the associated links; the node and relationship weight parameters include the set of weight values ​​for all nodes on the associated links and the set of strength values ​​for all relationship edges. Based on the extracted parameters, a policy impact strength score is calculated for each related link, including: Based on the path hop count, cross-domain jump count, and path interruption count, calculate the path connectivity index; The set of weight values ​​of the nodes, the set of strength values ​​of the relational edges, and the path connectivity index are input into a preset scoring function to calculate the policy impact strength score of the corresponding associated link.

7. The method according to claim 1, characterized in that, Based on the triggering logic elements identified from the power trading policy text data, triggering condition mapping and structured integration are performed on the policy trading association link set and the policy impact intensity result set, including: Triggering logic elements are extracted from the standard set of policy semantic elements obtained during the construction of the policy knowledge graph. The trigger logic elements are analyzed to obtain the trigger condition text, logical connection relationship, and applicable scope information; The trigger condition text is decomposed into a structured set of condition items; Based on the aforementioned scope of application information, the affected transaction nodes are located in the policy transaction association chain, and a mapping relationship between the set of condition items and the corresponding chain is established. Based on the logical connection relationship, each condition item in the condition item set is combined and encoded into an executable strategy triggering condition; Using policy identifiers and transaction behavior identifiers as association keys, the policy triggering conditions, corresponding association link information, and impact intensity information obtained from the policy impact intensity result set are integrated to generate a structured policy association identification result record; All policy association identification result records are aggregated, an index is created, and the policy association identification result set is generated.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the power trading policy association identification method as described in any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the power trading policy association identification method as described in any one of claims 1 to 7.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the power trading policy association identification method as described in any one of claims 1 to 7.

Images