CN-121998430-A - Gravity dam safety risk management and control method based on knowledge graph and large model

CN121998430ACN 121998430 ACN121998430 ACN 121998430ACN-121998430-A

Abstract

The invention discloses a gravity dam safety risk management and control method based on a knowledge graph and a large model, which comprises the steps of constructing a gravity dam safety risk space-time knowledge graph comprising a entity layer, a relation layer, an attribute layer and a space-time mapping layer, realizing multi-source data structural association of a full life cycle, constructing a knowledge enhancement type large model, embedding the space-time knowledge graph into an reasoning process, collecting and preprocessing real-time and historical data of the gravity dam, inputting the large model, calling a graph association relation and an evolution rule through the large model, completing risk abnormality identification, cause positioning and assessment, retrieving historical cases and industry specifications in the graph, generating a management and control decision scheme adapting to the current working condition, collecting scheme execution data, extracting new knowledge and updating the graph. The invention constructs a closed loop control mode and improves the intelligentization and refinement level of risk control.

Inventors

ZHAO JINZHEN
LI YUNJIE
XU HUILIANG

Assignees

云南省水利水电工程有限公司

Dates

Publication Date: 20260508
Application Date: 20260128

Claims (10)

1. The gravity dam safety risk management and control method based on the knowledge graph and the large model is characterized by comprising the following steps of: S1, constructing a gravity dam safety risk space-time knowledge graph, wherein the space-time knowledge graph comprises an entity layer, a relationship layer and an attribute layer, and the structured association and space-time dimension mapping of the gravity dam full life cycle multi-source data are realized; s2, constructing a knowledge enhancement type large model based on a search enhancement generation architecture and field adaptation training, and embedding the space-time knowledge graph as a structured knowledge source into a large model reasoning process; s3, collecting real-time running data and full life cycle historical data of the gravity dam, and inputting the data into the knowledge enhancement type large model after data cleaning, standardization and feature extraction processing; s4, invoking entity association relation and space-time evolution rules of a space-time knowledge graph through the knowledge enhancement type large model, completing risk anomaly identification, association path mining and root cause positioning, and generating a risk assessment result comprising a tracing basis; S5, based on a risk assessment result, searching historical cases, industry specifications and management rules in the space-time knowledge graph by using a knowledge enhancement type large model, and generating an interpretable risk management decision scheme adapting to the current working condition; and S6, collecting execution data and effect evaluation results of the risk management and control decision scheme, extracting new entities, relations and rules through the knowledge enhancement type large model, and dynamically updating the time knowledge graph.
2. The gravity dam security risk management and control method based on a knowledge graph and a large model as claimed in claim 1, wherein in step S1, the entity layer includes a dam structure entity, a monitoring facility entity, a risk factor entity, a history case entity, a management and control measure entity and an industry specification entity; The relation layer comprises a causal relation, a space-time association relation, a subordinate relation and a constraint relation among the entities, the attribute layer comprises static attributes and dynamic attributes of the entities, the static attributes are design parameters, structural features and specification thresholds, and the dynamic attributes are real-time monitoring data, running state parameters and timestamp information.
3. The gravity dam security risk management and control method based on a knowledge graph and a large model according to claim 1, wherein in step S1, the construction process of the space-time knowledge graph comprises: s11, collecting multi-source heterogeneous data of a gravity dam, wherein the multi-source heterogeneous data comprise design documents, construction records, monitoring data, history dangerous cases, maintenance records and industry standard texts; S12, extracting entity, relation and attribute information in the data by adopting a natural language processing technology and a graphic neural network algorithm, and forming a structured knowledge unit after verification and correction; S13, constructing a space-time correlation model based on an OGC standard, and binding a structured knowledge unit with space position information and time sequence information to form a space-time knowledge triplet; And S14, storing the space-time knowledge triples by adopting a graph database, and constructing a gravity dam safety risk space-time knowledge map with space-time query and path traversing capabilities.
4. The gravity dam security risk management and control method based on a knowledge graph and a large model according to claim 1, wherein in step S2, the construction process of the knowledge enhancement type large model includes: S21, converting entities, relations and attributes in the space-time knowledge graph into low-dimensional vectors by adopting a knowledge embedding algorithm, and merging the low-dimensional vectors into a word embedding layer of a large model; s22, constructing a retrieval enhancement generation framework, and establishing real-time association between the space-time knowledge graph and the large model to dynamically retrieve related structured knowledge during large model reasoning; And S23, collecting labeling data in the field of safety control of the gravity dam, and performing field adaptation training on the large model.
5. The gravity dam security risk management and control method based on a knowledge graph and a large model as claimed in claim 1, wherein in step S4, the risk tracing process specifically includes the following steps: S41, comparing the preprocessed data with a normal operation threshold value in a space-time knowledge graph by a knowledge enhancement type large model, and identifying risk abnormal signals and corresponding related entities; s42, traversing a direct relation and indirect relation path of the associated entity in the empty knowledge graph, and mining a key link of risk propagation; S43, analyzing the influence weight of each entity in the key link based on the large model reasoning capacity, and combining with a space-time evolution rule, and positioning the root cause of the risk; and S44, associating professional basis in the space-time knowledge graph to generate an interpretable risk assessment result comprising risk factors, evolution paths and judgment basis components.
6. The gravity dam security risk management and control method based on a knowledge graph and a large model as claimed in claim 1, wherein in step S5, the generation process of the management and control decision scheme includes: S51, searching a historical success case matched in the space-time knowledge graph and related industry specifications according to the risk root cause and the current working condition characteristics by the knowledge enhancement type large model; S52, performing suitability optimization on the retrieved history management and control measures to form a candidate measure set conforming to the structural characteristics and the running state of the current gravity dam; And S53, based on the risk influence range, the management and control targets and the implementation conditions, the candidate measures are prioritized, and a complete decision scheme comprising the management and control targets, the specific measures, the implementation steps, the expected effect and the professional basis is generated.
7. The gravity dam security risk management and control method based on a knowledge graph and a large model according to claim 1, wherein in step S6, the dynamic update process of the space-time knowledge graph specifically comprises the following steps: S61, collecting execution process data, field feedback data and effect verification data of a decision scheme; S62, analyzing deviation of an execution effect and an expected effect through a knowledge enhancement type large model, and extracting new risk factors, management and control measures and association relations; And S63, merging the newly extracted knowledge units into a space-time knowledge graph after verification, and updating entity attributes, relationship paths and evolution rules to realize autonomous evolution of the knowledge graph.
8. The gravity dam security risk management and control method based on a knowledge graph and a large model as claimed in claim 1, wherein the data preprocessing process in step S3 specifically comprises the following steps: and processing abnormal points in the monitoring data by adopting a statistical abnormal value removing algorithm, unifying data dimension by adopting a standardized or normalized algorithm, extracting text features in the unstructured document by adopting an optical character recognition technology, and extracting morphological feature parameters in the monitoring image by adopting an image feature extraction algorithm.
9. The gravity dam security risk management and control method based on a knowledge graph and a large model according to claim 4, wherein the knowledge embedding algorithm in the step S21 specifically adopts one or more combinations of TransE, transH, transR or ComplEx algorithms; The field adaptation training comprises pre-training fine adjustment based on field corpus, prompt engineering optimization and supervised learning training of a small number of labeling samples.
10. The gravity dam security risk management and control method based on a knowledge graph and a large model according to claim 6, wherein the retrieval process in step S51 specifically adopts a dual retrieval mechanism of semantic similarity matching and working condition feature vector comparison, and the priority ranking in step S53 is based on the factors including measure implementation cost, risk alleviation effect speed, construction complexity and influence degree on normal operation of the gravity dam.

Description

Gravity dam safety risk management and control method based on knowledge graph and large model Technical Field The invention relates to the technical field of hydraulic engineering safety management and control, in particular to a gravity dam safety risk management and control method based on a knowledge graph and a large model. Background The gravity dam is used as a water retaining structure with huge volume and key functions in hydraulic engineering, is widely applied to the construction of important infrastructures such as flood control, power generation and irrigation, and the safe and stable operation of the gravity dam is directly related to the stable development of people life and property safety, ecological environment sustainability and regional economy and society in a river basin. Along with the extension of service life, the gravity dam faces the natural factors such as dam body concrete aging, basic geological condition evolution, water flow scouring erosion and the like, and meanwhile, artificial and environmental factors such as extreme climate event frequency, operation load fluctuation and the like are superimposed, so that potential safety hazards such as seepage flow abnormality, dam body displacement exceeding and crack expansion and the like are continuously accumulated. Under the background, the multi-source heterogeneous data generated in the whole life cycle of the gravity dam is in explosive growth, and covers structural parameters of a design stage, process records of a construction stage, real-time monitoring data of an operation stage, historical dangerous cases, maintenance records, industry standard texts and the like, and the traditional management and control mode which depends on manual experience judgment and single monitoring index analysis is difficult to meet the high requirements of modern safety management and control on data integration, risk prejudgment and decision scientificity, so that an intelligent and systematic technical means is required to be introduced to realize the risk management and control upgrading. The key problems to be solved in the current gravity dam safety risk management and control field are that firstly, multisource data lacks effective structural association and space-time dimension mapping, information such as design parameters, real-time monitoring data, historical cases and the like are stored in a scattered mode, formats are heterogeneous, logic links such as causal relation and space-time association relation among entities cannot be established, risk analysis lacks comprehensive and coherent data support and is difficult to capture risk evolution rules, secondly, accuracy and suitability of risk factor positioning and management and control decision are insufficient, the conventional method relies on a single threshold to judge and identify risks, key links and root causes of risk propagation are difficult to mine, and the decision scheme is multi-view and carries general experience or industry specifications, personalized optimization cannot be carried out in combination with specific working conditions such as current dam structural features and running states, and meanwhile a knowledge system cannot be iterated along with management and control practice, so that decision effect is limited. Disclosure of Invention The technical problems to be solved by the invention are that the multi-source data structure association is insufficient, the risk cause positioning is inaccurate, the management and control decision adaptability is poor and the knowledge system cannot be iterated dynamically in the management and control of the safety risk of the gravity dam, and the gravity dam safety risk management and control method based on the knowledge graph and the large model is provided for solving the problems. In order to solve the technical problems, the technical scheme of the invention is that the gravity dam safety risk management and control method based on a knowledge graph and a large model comprises the following steps: S1, constructing a gravity dam safety risk space-time knowledge graph, wherein the space-time knowledge graph comprises an entity layer, a relationship layer and an attribute layer, and the structured association and space-time dimension mapping of the gravity dam full life cycle multi-source data are realized; s2, constructing a knowledge enhancement type large model based on a search enhancement generation architecture and field adaptation training, and embedding the space-time knowledge graph as a structured knowledge source into a large model reasoning process; s3, collecting real-time running data and full life cycle historical data of the gravity dam, and inputting the data into the knowledge enhancement type large model after data cleaning, standardization and feature extraction processing; s4, invoking entity association relation and space-time evolution rules of a space-time knowledge graph through the knowle