CN-121659939-B - Intelligent nuclear power system health assessment method based on large model technology

CN121659939BCN 121659939 BCN121659939 BCN 121659939BCN-121659939-B

Abstract

The application relates to the technical field of fault prediction and health management, in particular to an intelligent nuclear power system health assessment method based on a large model technology, which comprises the steps of carrying out knowledge extraction on historical nuclear power data to obtain data of each triplet; the method comprises the steps of screening similar triplet data of each triplet data from other triplet data, constructing an undirected graph of each triplet data, obtaining path association parameters from an entity in each triplet data to each node in the undirected graph, obtaining pre-filtered triplet data of each triplet data, obtaining association redundancy of each pre-filtered triplet data, filtering redundant triplet data from all triplet data obtained through knowledge extraction, constructing a nuclear power knowledge graph, and evaluating health states of a nuclear power system. The application aims to filter redundant parts in the triplet data, so that the nuclear power knowledge graph is more accurate and simplified, and more accurate nuclear power system health evaluation is realized.

Inventors

XIE BIN
CHEN YINGJIE

Assignees

北京中电昊海科技有限公司

Dates

Publication Date: 20260505
Application Date: 20251210

Claims (9)

1. The intelligent nuclear power system health assessment method based on the large model technology is characterized by comprising the following steps of: knowledge extraction is carried out on the history nuclear power data to obtain each triplet data, and word vectors of words in each triplet data are obtained; Screening each similar triplet data of each triplet data from the rest triplet data through the similarity of word vectors between each word in each triplet data and each word in each rest triplet data, and constructing an undirected graph of each triplet data; obtaining path correlation parameters from entities in the triplet data to each node in the undirected graph by taking the entities in the triplet data as starting points and the shortest path length of each node in the undirected graph and constructing the triplet data quantity of the undirected graph, filtering redundant paths in the undirected graph from the entities in the triplet data as starting points through the distribution of the path correlation parameters, further obtaining each pre-filtered triplet data of each triplet data according to the redundant paths, combining information quantity contained in each pre-filtered triplet data by the distribution of the path correlation parameters from the entities in the pre-filtered triplet data to obtain correlation redundancy of each pre-filtered triplet data, further filtering the redundant triplet data from all the triplet data obtained through knowledge extraction, and constructing a nuclear power map according to filtering results to evaluate the health state of the nuclear power system; the screening each similar triplet data of each triplet data from the rest triplet data and constructing an undirected graph of each triplet data comprises the following steps: acquiring an accumulated similarity coefficient between each triplet data and each other triplet data through the similarity of word vectors between each word in each triplet data and each word in each other triplet data; Acquiring a segmentation threshold value of an accumulated similarity coefficient between each triplet data and all other triplet data, and using the accumulated similarity coefficient which is larger than or equal to the segmentation threshold value as each similar triplet data of each triplet data; and constructing an undirected graph of each triplet data by using each triplet data and the similar triplet data.
2. The method for evaluating the health of an intelligent nuclear power system based on the large model technology as claimed in claim 1, wherein the process of obtaining the accumulated similarity coefficient is as follows: Calculating the average value of the similarity between each word in each triplet data and all words in each other triplet data; the cumulative similarity coefficient is the sum of the mean values between all words in each triplet data and each of the remaining triplet data.
3. The intelligent nuclear power system health assessment method based on the large model technology as claimed in claim 1, wherein the path association parameter obtaining process is as follows: Aiming at each triplet data, obtaining the association rate of each triplet data by comparing the number of similar triplet data participating in building an undirected graph with the total number of similar triplet data; Calculating the product of the association rate and the number of paths; The path-associated parameter is proportional to the product and inversely proportional to the shortest path length.
4. The intelligent nuclear power system health assessment method based on the large model technology as claimed in claim 3, wherein the association rate is the number ratio of similar triplet data participating in building an undirected graph in all similar triplet data.
5. A method of health assessment of an intelligent nuclear power system based on large model technology as claimed in claim 3, wherein said path correlation parameter is the ratio of said product to said shortest path length.
6. The intelligent nuclear power system health assessment method based on the large model technology as claimed in claim 1, wherein the pre-filtering triplet data obtaining process is as follows: Obtaining a segmentation threshold value of path association parameters from an entity in each triplet data to all nodes in the undirected graph, and selecting nodes corresponding to the path association parameters, which are larger than the segmentation threshold value, in the undirected graph of each triplet data; And taking each path except the shortest path in paths from the entity in each triplet data as a starting point to each node selected in the undirected graph as each redundant path, and taking each triplet data formed by every two adjacent nodes and edges between the adjacent nodes on each redundant path as each prefiltered triplet data.
7. The intelligent nuclear power system health assessment method based on the large model technology as claimed in claim 1, wherein the process of obtaining the associated redundancy is as follows: Calculating the average value of path association parameters from the entity in each prefiltered triplet data to all other nodes of the undirected graph; Calculating the entropy of each word in each pre-filtering triplet data according to the occurrence probability of each word in the pre-filtering triplet data in the same position word in all pre-filtering triplet data of each triplet data, and counting the maximum value in the entropy of all words in each pre-filtering triplet data; the associative redundancy is proportional to the average value and inversely proportional to the maximum value.
8. The large model technology based intelligent nuclear power system health assessment method of claim 7, wherein the associative redundancy is a ratio of the maximum values of the average values.
9. The intelligent nuclear power system health assessment method based on the large model technology as claimed in claim 1, wherein the process of filtering redundant triplet data is as follows: And obtaining the segmentation threshold value of the associated redundancy of all pre-filtered triple data of each triple data, taking the pre-filtered triple data with the associated redundancy larger than or equal to the segmentation threshold value as redundant triple data, and filtering.

Description

Intelligent nuclear power system health assessment method based on large model technology Technical Field The application relates to the technical field of fault prediction and health management, in particular to an intelligent nuclear power system health assessment method based on a large model technology. Background The health assessment of the intelligent nuclear power system has great influence on the safety and the economical efficiency of the nuclear power plant, and in recent years, along with the development of large model technology, the health assessment of the nuclear power system tends to be intelligent, and particularly, a knowledge base is formed by combining expert experience and historical data by adopting the large model technology, and a more complex and accurate health assessment task is realized by combining multi-mode data in the nuclear power operation process. Because nuclear power equipment is extremely easy to corrode and crack under extreme environments such as high temperature, high radiation and the like for a long time, accurate realization of health assessment of an intelligent nuclear power system based on a large model technology has important significance. Because the nuclear power system data is complex and various, the problem of 'illusion' easily occurs when the large model technology is used for realizing the health assessment of the nuclear power system, although the existing research combines the knowledge graph to relieve the illusion problem of the large model technology, knowledge in the nuclear power system is usually accompanied with knowledge redundancy condition, a large number of redundancy triples occur, so that the knowledge graph useful information is diluted, the built knowledge graph lacks accuracy, the relative importance degree between nuclear power fault data and nuclear power normal data is difficult to distinguish when the large model is used for carrying out the health assessment of the nuclear power system, and the final health assessment effect is poor. Disclosure of Invention In view of the above, it is necessary to provide a method for evaluating the health of an intelligent nuclear power system based on a large model technology, which, compared with the traditional method for evaluating the health of an intelligent nuclear power system based on a large model technology, filters redundant parts in triplet data to make a nuclear power knowledge graph more accurate and simplified, thereby realizing more accurate health evaluation of the nuclear power system: the intelligent nuclear power system health assessment method based on the large model technology adopts the following technical scheme: the embodiment of the application provides an intelligent nuclear power system health assessment method based on a large model technology, which comprises the following steps: knowledge extraction is carried out on the history nuclear power data to obtain each triplet data, and word vectors of words in each triplet data are obtained; The method comprises the steps of screening each similar triplet data of each triplet data from the rest triplet data according to the similarity of word vectors between each word in each triplet data and each word in each other triplet data, constructing an undirected graph of each triplet data, obtaining path association parameters from each triplet data to each node in the undirected graph when an entity in each triplet data is used as a starting point by combining the number of paths and the shortest path length from each entity in each triplet data to each node in the undirected graph when the entity in each triplet data is used as the starting point and constructing the triplet data of the undirected graph, screening redundant paths in the undirected graph when the entity in each triplet data is used as the starting point through the distribution of the path association parameters, obtaining each prefiltered triplet data according to the distribution of the path association parameters from each entity in each triplet data to the nodes in the undirected graph, combining the information quantity contained in each prefiltered triplet data, obtaining the association of each prefiltered triplet data, filtering the redundant state from the nuclear power system according to the redundancy state, and filtering the redundant state of each nuclear power system according to the filter state, and obtaining the filter knowledge map. In one embodiment, the filtering each similar triplet data of each triplet data from the rest of triplet data and constructing an undirected graph of each triplet data includes: acquiring an accumulated similarity coefficient between each triplet data and each other triplet data through the similarity of word vectors between each word in each triplet data and each word in each other triplet data; Acquiring a segmentation threshold value of an accumulated similarity coefficient between each triplet data and al