CN-122022465-A - Multi-mode and knowledge graph-based intelligent research and judgment system for environment-friendly compliance of power grid water

CN122022465ACN 122022465 ACN122022465 ACN 122022465ACN-122022465-A

Abstract

The application discloses an environment-friendly and intelligent power grid water compliance research and judgment system based on a multi-mode and knowledge graph, which comprises an environment-friendly rule knowledge graph construction module, a multi-source space-time alignment and data fusion module, a multi-spectrum semantic understanding and violation feature extraction module, a knowledge graph driven compliance reasoning and research and judgment module, and an interpretable rectification report generation and dynamic early warning module, wherein the environment-friendly rule knowledge graph construction module is used for converting unstructured environment-friendly rule texts into structured knowledge graphs, the compliance reasoning and research and judgment module is used for conducting compliance reasoning and research on violation feature facts based on the knowledge graphs, and outputting violation events and risk indexes. Based on further analysis and research on the problems in the prior art, the application converts unstructured rule and regulations into structured constraint rules and entity relations by constructing a calculable and inferable environmental water conservation rule knowledge map, so that a machine can directly understand and apply the rules, the efficiency and accuracy of rule retrieval, association and application are improved, and the manual dependence is reduced.

Inventors

SHI JIANBO
CAI XUAN
ZHANG YING
ZHANG CHI

Assignees

国网湖北省电力有限公司电力科学研究院

Dates

Publication Date: 20260512
Application Date: 20260127

Claims (10)

1. Electric wire netting environmental protection compliance intelligence is ground and is judged system based on multimodality and knowledge graph, its characterized in that, the system includes: the environmental water conservation regulation knowledge graph construction module is used for converting unstructured environmental water conservation regulation texts into structured knowledge graphs; the multi-source space-time alignment and data fusion module is used for carrying out space-time alignment and fusion on multi-source heterogeneous data of the power grid engineering and outputting a space-time aligned fusion data model; The multispectral semantic understanding and violation feature extraction module is used for extracting violation feature facts with environment-friendly semantics based on the fusion data model; the knowledge graph driven compliance reasoning and judging module is used for carrying out compliance reasoning and judging on the violation feature facts based on the knowledge graph and outputting violation events and risk indexes; and the interpretable correction report generation and dynamic early warning module is used for generating a correction report based on the violation event and the risk index and sending out early warning.
2. The intelligent power grid environment-friendly compliance research and judgment system based on the multi-mode and knowledge patterns is characterized in that the environment-friendly rule knowledge pattern construction module is used for preprocessing and vectorizing rule texts to construct a clause-level vector database, extracting entities and relations from the texts based on a large language model, carrying out constraint condition structural analysis, fusing and disambiguating the extracted knowledge, constructing patterns, and carrying out association mapping on semantic constraints in the text rules and remote sensing observable features to form a computable knowledge pattern.
3. The multi-mode and knowledge graph-based intelligent power grid environment-friendly compliance research and judgment system is characterized in that the multi-source space-time alignment and data fusion module is used for carrying out standardized preprocessing on engineering design vector data, remote sensing images and real-time sensor data, carrying out high-precision dynamic registration and space superposition on the preprocessed vector data and the remote sensing images, realizing space coherence alignment of full line data based on a layered topological alignment mechanism of linear engineering, and constructing a space-time data pool containing time, space and attribute multi-dimensional indexes so as to organize the multi-period data after alignment.
4. The multi-modal and knowledge graph-based intelligent research and judgment system for environmental water compliance of a power grid according to claim 1, wherein the multi-spectral semantic understanding and violation feature extraction module comprises: The spatial spectrum combined depth classification network is used for finely classifying ground features of the multispectral remote sensing image; The light model and change detection unit is used for identifying construction behaviors; The violation feature structuring extraction unit is used for carrying out space logic operation on the classification result and the design constraint to generate structured violation features; and the cascaded two-classification fine tuning judging module is used for distinguishing bare soil with spectrum confusion from hardened ground surface, and the judging module synthesizes multi-scale texture characteristics and spectrum characteristics for identification.
5. The system is characterized in that the knowledge-graph-driven compliance reasoning and judging module is used for mapping the violation feature vectors, the spatial relationships and the external geographic data to the knowledge graph, performing space out-of-range detection, site selection compliance judgment and construction behavior compliance judgment on construction activities based on rule constraint rules in the knowledge graph, and dynamically calculating an environment-friendly risk index based on a rule breaking event basic risk value, a time attenuation factor, a space aggregation factor and a responsibility main body historical risk factor.
6. The system is characterized in that the interpretable rectification report generation and dynamic early warning module is used for selecting templates from a template library and automatically filling evidence chain information to generate a structural rectification notice based on the type and severity of a violation event, providing a visual analysis board comprising global situation overview, event detail drilling and time-space track backtracking, implementing hierarchical early warning and closed-loop treatment tracking based on risk indexes and rule severity, integrating a data-driven rectification suggestion personalized recommendation algorithm, and recommending a rectification measure combination for the violation event.
7. The intelligent power grid environment-friendly compliance research and judgment system based on the multi-mode and knowledge patterns is characterized in that the environment-friendly rule knowledge pattern construction module further comprises a rule dynamic update mechanism for monitoring and collecting multi-source rule information, identifying rule change types and preprocessing new texts, performing accurate comparison of clause change based on vector similarity, and performing incremental update, consistency verification and version management on the knowledge patterns according to comparison results.
8. The intelligent research and judgment system for the environmental protection and compliance of the power grid based on the multi-mode and the knowledge graph as claimed in claim 1, wherein a space-time data pool constructed by the multi-source space-time alignment and data fusion module adopts a multi-dimensional index structure of 'time stamp-space grid code-attribute label', and is used for efficiently associating and searching engineering data of different periods, different space positions and different semantic attributes.
9. The system for intelligent research and judgment of environmental protection and water conservation of a power grid based on multi-mode and knowledge-graph according to claim 1, wherein the knowledge-graph driven compliance inference and research and judgment module further comprises a reliability evaluation unit for comprehensively scoring reliability of each offence research and judgment result, wherein the scores are integrated with evidence source reliability, constraint quantization precision and logic consistency scores.
10. The system for intelligently studying and judging the environmental water environmental protection compliance of the power grid based on the multi-mode and the knowledge map according to claim 1 is characterized in that the flow of the system for realizing the data circulation comprises that the knowledge map output by the environmental water environmental protection regulation knowledge map construction module provides a computable regulation constraint for the compliance reasoning and studying and judging module; the fusion data model output by the multi-source space-time alignment and data fusion module provides space-time reference unified image and vector data for the multi-spectrum semantic understanding module; the violation feature facts extracted by the multispectral semantic understanding module provide live-action evidence input for the compliance reasoning and judging module; And the rule breaking event and risk index output by the rule breaking reasoning and judging module provides judging conclusion and risk basis for the correction report generating and early warning module.

Description

Multi-mode and knowledge graph-based intelligent research and judgment system for environment-friendly compliance of power grid water Technical Field The application belongs to the technical field of geospatial information, and particularly relates to a power grid environment-friendly compliance intelligent research and judgment system based on a multi-mode and knowledge graph. Background With the increasing strictness of the requirements of rapid promotion and ecological environment protection of the construction of the power grid infrastructure in China, the whole process of planning, construction and operation of the power grid engineering is required to strictly fulfill environmental protection and water and soil conservation (hereinafter referred to as environmental protection) regulations. The traditional environment-friendly supervision mode is highly dependent on manual field inspection, paper document inspection and post report inspection, and faces a plurality of challenges: Firstly, massive, scattered and frequently updated environment-friendly rule texts are mostly in unstructured forms, manual review and understanding are low in efficiency, and accurate and dynamic association of the rule texts and specific engineering scenes is difficult to achieve, so that compliance judgment standards are different and lag is caused. Secondly, the supervision data sources are various, including engineering design drawings (CAD/GIS), multi-period remote sensing images, sensors of the Internet of things and the like, and the data have isomerism on space-time reference, format and semantic level, and lack of effective automatic alignment and fusion means, so that the coherent analysis and change tracking of 'design-construction-live-action' are difficult. Thirdly, interpretation of earth surface observation data such as remote sensing images and the like is mostly dependent on visual interpretation or traditional image processing, and identification of construction disturbance (such as bare soil and water turbidity) is easily interfered by 'same-spectrum foreign matters' (such as confusion of spectrum of bare soil and hardened earth surface), so that false alarm rate is high, and automation and intellectualization levels are insufficient. Fourth, the links of rule violation discovery, research and judgment, report generation and correction tracking are disjointed, a data-driven and knowledge-assisted intelligent closed loop cannot be formed, the supervision efficiency and the accuracy are required to be improved, and the dynamic supervision requirement of long-distance multi-scale power grid linear engineering is difficult to deal with. Therefore, there is a need for an intelligent research and judgment system that can deeply fuse multi-source data, intelligently interpret regulations, automatically and accurately identify violations, and support closed-loop treatment. Disclosure of Invention The application provides a multi-mode and knowledge graph-based intelligent system for studying and judging environmental water compliance of a power grid, and aims to solve the problems that in the prior art, the accurate and dynamic association of a strip document and a specific engineering scene is difficult, and an effective automatic alignment and fusion means is lacking. Electric wire netting environmental protection compliance intelligence is ground and is judged system based on multimodality and knowledge graph, the system includes: the environmental water conservation regulation knowledge graph construction module is used for converting unstructured environmental water conservation regulation texts into structured knowledge graphs; the multi-source space-time alignment and data fusion module is used for carrying out space-time alignment and fusion on multi-source heterogeneous data of the power grid engineering and outputting a space-time aligned fusion data model; The multispectral semantic understanding and violation feature extraction module is used for extracting violation feature facts with environment-friendly semantics based on the fusion data model; the knowledge graph driven compliance reasoning and judging module is used for carrying out compliance reasoning and judging on the violation feature facts based on the knowledge graph and outputting violation events and risk indexes; and the interpretable correction report generation and dynamic early warning module is used for generating a correction report based on the violation event and the risk index and sending out early warning. The environment-friendly rule knowledge graph construction module is used for preprocessing and vectorizing rule texts to construct a term-level vector database, extracting entities and relations from the texts based on a large language model, carrying out constraint condition structural analysis, fusing, disambiguating and graph construction on the extracted knowledge, and carrying out associated mapping on semantic constraints