CN-121979980-A - Intent recognition and dynamic routing decision method, system, equipment and storage medium for power grid knowledge graph fusion

CN121979980ACN 121979980 ACN121979980 ACN 121979980ACN-121979980-A

Abstract

The invention discloses an intention recognition and dynamic routing decision method, system, equipment and storage medium for power grid knowledge graph fusion, which relate to the technical field of intelligent question-answering and decision support of power grids; constructing and fusing three layers of knowledge maps covering equipment, topology and states based on multi-source power grid business data to form a structured domain knowledge base; selecting a processing flow based on the intention type obtained by identifying the query intention, the query complexity score determined by the complexity assessment model and the urgency score determined by the urgency assessment model; the method and the system can remarkably improve the recognition precision of the professional query intention of the power grid, support complex reasoning, realize differentiated response, rapidly process emergency faults and have continuous self-optimization capability.

Inventors

YI CHUNFANG
YIN YUAN
HUANG YITING
Ai Xuhua
WANG JIALIN
DONG BIN
CHEN QI
WANG WEI
MENG QI
LIU KAIJIE
MENG ZHIPENG
MENG CHUNZHI

Assignees

广西电网有限责任公司

Dates

Publication Date: 20260505
Application Date: 20251231

Claims (10)

1. The intent recognition and dynamic routing decision method for power grid knowledge graph fusion is characterized by comprising the following steps of: based on the related query of the power grid input by a user, carrying out specialized intention understanding of the power field, identifying the query intention and linking related entities; Constructing and fusing three layers of knowledge maps covering equipment, topology and states based on multi-source power grid business data to form a structured domain knowledge base; Selecting a processing flow based on the intention type obtained by identifying the query intention, the query complexity score determined by the complexity assessment model and the urgency score determined by the urgency assessment model; And generating a response result according to the selected processing flow and outputting the response result to the user.
2. The method for identifying and dynamically determining routing with integrated intent based on power grid knowledge graph as recited in claim 1, wherein the step of performing power domain specialized intent understanding based on the power grid related query input by the user, identifying the query intent and linking the related entities comprises: Extracting basic categories of query intention through a semantic analysis unit, namely segmenting the query to obtain a word sequence, constructing the word and the dependency relationship thereof into word graphs based on dependency syntactic analysis, processing the word graphs by adopting a graph attention network, learning deep semantic representation of the word by aggregating neighbor node information of each word, generating overall semantic representation of the query through pooling operation, and determining the intention category of the query based on the overall semantic representation.
3. The method for identifying intent and dynamically determining routing in combination with power grid knowledge graph as recited in claim 2, wherein the performing power domain specific intent understanding based on the power grid related query input by the user, identifying the query intent and linking the related entities further comprises: the method comprises the steps of linking a professional term and an entity expression in a query to a corresponding node in a power grid knowledge graph through an entity linking unit, identifying an entity mention in the query, retrieving a candidate entity corresponding to the entity mention from the power grid knowledge graph, comprehensively considering semantic matching degree of the candidate entity and the entity mention and topological distance constraint of the candidate entity in the power grid knowledge graph, determining a final entity linking result through weighted calculation, and verifying the entity linking result.
4. The method for intent recognition and dynamic routing decision-making based on power grid knowledge graph fusion as recited in claim 3, wherein said constructing and fusing three layers of knowledge graphs covering devices, topologies and states based on multi-source power grid business data, forming a structured domain knowledge base comprises: based on multi-source power grid business data, a power grid knowledge graph adopting a three-layer architecture is constructed and maintained in a mode of combining automation and manual verification, and the three-layer architecture comprises: The equipment layer is used for storing static attribute information of various power equipment in a node form; the topology layer is used for storing the connection relation and the spatial position relation between the power equipment; The state layer is used for storing dynamic operation state data and historical fault records of the power equipment; And providing knowledge inquiry and reasoning services based on the constructed power grid knowledge graph, wherein the reasoning services comprise deducing a fault influence range according to a connection relation and predicting fault risk according to a historical fault record.
5. The method for power grid knowledge graph fusion intention recognition and dynamic route decision as recited in claim 4, wherein the type of intention obtained based on recognition of query intention, a query complexity score determined by a complexity assessment model and an urgency score determined by an urgency assessment model, the selection process flow comprising: Based on the intention type, the complexity evaluation score and the urgency evaluation score, a processing flow is selected for the query according to preset rules, wherein the processing flow comprises a quick response flow, a standard processing flow and a deep analysis flow, the preset rules are that the fault diagnosis type query is preferentially selected according to the urgency score, the operation and maintenance query type query is preferentially selected according to the complexity score, and the deep analysis flow is started according to the data analysis type query.
6. The method for power grid knowledge graph fusion intention recognition and dynamic route decision as recited in claim 5, wherein the type of intention obtained based on recognition of query intention, the query complexity score determined by the complexity assessment model and the urgency score determined by the urgency assessment model, the selection process further includes: And adopting an optimization mechanism based on reinforcement learning, taking the intention type, the complexity evaluation score, the urgency evaluation score and the current system resource state as state inputs, selecting a processing flow as action output, and calculating a reward signal according to the user satisfaction degree, the response time and the resource consumption after query processing so as to continuously update and optimize the routing decision parameters in a preset rule.
7. The method for intent recognition and dynamic route decision as recited in claim 6, wherein said generating and outputting a response result to a user based on a selected process flow includes: According to the selected processing flow, corresponding knowledge resources and computing resources are called to generate response results, wherein the processing flow comprises a quick response flow, a standard processing flow and a deep analysis flow; When the processing flow is a quick response flow, based on the fault phenomenon and fault equipment information in the query, searching the matched historical fault cases through case reasoning, and quickly identifying the fault type; based on the topological connection relation of the fault equipment in the power grid knowledge graph, evaluating the range of possible influences of the fault; based on the fault type and the influence range evaluation result, retrieving and generating an emergency treatment suggestion comprising immediate action, diagnosis steps and recovery plans from an emergency treatment plan library; When the processing flow is a standard processing flow, related information is retrieved from a multi-source knowledge base according to the query intention, the retrieval results are de-duplicated, sequenced and integrated to form complete answers, and the integrated information is filled into a predefined normalized answer template to generate structured detailed answers; When the processing flow is a deep analysis flow, historical data is extracted from a time sequence database to carry out statistical analysis, correlation analysis and cluster analysis, a time sequence prediction model is adopted to carry out trend prediction analysis, and an analysis result is combined with a chart and characters to generate a comprehensive report; and outputting the generated response result to a user after quality check, wherein the quality check comprises fact check, logic consistency check and language fluency check.
8. An intent recognition and dynamic route decision system for power grid knowledge graph fusion, applying the method of any one of claims 1-7, comprising: The query receiving and intention recognition module is used for carrying out specialized intention understanding in the electric power field based on the power grid related query input by the user, recognizing the query intention and linking related entities; The knowledge map fusion module is used for constructing and fusing three layers of knowledge maps covering equipment, topology and states based on the multi-source power grid business data to form a structured domain knowledge base; The dynamic routing decision module is used for selecting a processing flow based on the intention type obtained by identifying the query intention, the query complexity score determined by the complexity evaluation model and the urgency score determined by the urgency evaluation model; and the response generation module is used for generating a response result according to the selected processing flow and outputting the response result to the user.
9. An electronic device, comprising: A memory and a processor; The memory is for storing computer executable instructions, the processor being for executing the computer executable instructions which when executed by the processor implement the steps of the method of any one of claims 1to 7.
10. A computer-readable storage medium, characterized in that it stores computer-executable instructions which, when executed by a processor, implement the steps of the method of any one of claims 1 to 7.

Description

Intent recognition and dynamic routing decision method, system, equipment and storage medium for power grid knowledge graph fusion Technical Field The invention relates to the technical field of intelligent question and answer and decision support of power grids, in particular to an intention recognition and dynamic routing decision method, system, equipment and storage medium for power grid knowledge graph fusion. Background Along with the continuous improvement of the intelligent level of the power grid, the scenes of power grid operation and maintenance management, fault diagnosis processing, power dispatching command and the like are increasingly urgent in demands of professional, efficient and accurate intelligent question-answering and decision-making support systems. However, when the existing intelligent question-answering system of the power grid faces to the query containing a large number of terms, complex topological association and implicit field semantics, the real intention of the user is often difficult to understand accurately, and a differentiated response strategy cannot be provided, so that the response efficiency and the speciality are insufficient, and the high standard requirements of real-time operation and maintenance and emergency command of the power grid are difficult to meet. However, the prior art has the technical defects that a general purpose recognition framework is difficult to accurately understand the technical terms and implicit semantics of the power grid field under the power grid professional application scene, the purpose recognition accuracy is obviously reduced when the inquiry comprising the professional concepts such as line loss, tide, relay protection and the like is processed, the different graph categories such as power grid operation and maintenance inquiry, fault diagnosis request, scheduling instruction and the like cannot be accurately distinguished, subsequent information retrieval and answer generation deviate from the actual requirements of users, the knowledge graph is not fully integrated with the field characteristics such as a power grid topological structure, equipment association relation and operation state and the like, the depth modeling is not available for the electric connection relation, the spatial position relation and the functional dependency relation among power grid equipment, the accurate relation reasoning support cannot be provided when the inquiry involving multistage voltage grade conversion, trans-regional power grid coordination and complex fault propagation chains is processed, and the mechanism for carrying out dynamic routing decision according to the purpose type, the complexity of the problem and the emergency degree is not available for carrying out uniform processing flow for different types of requests such as power grid emergency processing, conventional operation and maintenance inquiry and historical data analysis, the differential response cannot be realized, and the emergency processing response is influenced, and the safe and stable operation of the power grid is influenced. Disclosure of Invention In view of the above problems, the present invention provides a method, a system, a device and a storage medium for intent recognition and dynamic route decision of power grid knowledge graph fusion. Therefore, the method solves the technical problems of how to accurately understand the professional intention of the power grid, deeply fuse the multidimensional knowledge of the power grid, and dynamically allocate the intention recognition and routing decision of processing resources according to the query characteristics. In order to solve the technical problems, the invention provides the following technical scheme: In a first aspect, the present invention provides an intent recognition and dynamic route decision method for power grid knowledge graph fusion, including: based on the related query of the power grid input by a user, carrying out specialized intention understanding of the power field, identifying the query intention and linking related entities; Constructing and fusing three layers of knowledge maps covering equipment, topology and states based on multi-source power grid business data to form a structured domain knowledge base; Selecting a processing flow based on the intention type obtained by identifying the query intention, the query complexity score determined by the complexity assessment model and the urgency score determined by the urgency assessment model; And generating a response result according to the selected processing flow and outputting the response result to the user. The technical scheme has the beneficial effects that through constructing a complete technical closed loop for query intention understanding, knowledge graph fusion, dynamic routing decision and differential response generation, the end-to-end intelligent processing of complex queries in the power grid field is realized, the