Substation equipment intelligent fault diagnosis system and method based on deep learning and knowledge graph

Abstract

The invention discloses a substation equipment intelligent fault diagnosis system and method based on deep learning and a knowledge graph. At present, the fault diagnosis of the substation equipment is generally to monitor the working data of the equipment, locate the position of the abnormal fault data and dispatch maintainers. The method comprises the following steps: acquiring historical fault data of substation equipment and preprocessing; constructing a training substation equipment fault diagnosis knowledge graph based on the preprocessed historical fault data; working data of the substation equipment are collected in real time, and when abnormal fault data are monitored, equipment fault reasons are diagnosed based on the substation equipment fault diagnosis knowledge graph. According to the invention, the fault reason of the substation equipment can be quickly diagnosed, the labor intensity of maintainers is reduced, a large amount of fault diagnosis experience of the substation equipment is integrated, the fault reason diagnosis accuracy is high, and the maintenance efficiency of the substation equipment can be improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of substation equipment fault diagnosis, and particularly relates to an intelligent fault diagnosis system and method for substation equipment based on deep learning and knowledge graph. Background Art

[0002] A substation refers to a place in the power system where the voltage and current of electric energy are transformed, concentrated, and distributed, usually including power distribution equipment, power transformers, control equipment, protection and automation equipment, communication equipment, and compensation equipment, etc. Power distribution equipment, power transformers, control equipment, protection and automation equipment, communication equipment, and compensation equipment, etc. often have faults during long-term operation due to external environmental factors such as temperature, humidity, and dust, as well as internal hardware factors such as vibration.

[0003] Currently, the general method for fault diagnosis of substation equipment is to monitor the working data of substation equipment in real time. When abnormal fault data occurs, locate the position of the abnormal fault data, and dispatch maintenance personnel to that position for fault diagnosis and maintenance.

[0004] With the increasing number of substation equipment, the above method greatly increases the labor intensity of maintenance personnel. Moreover, the diagnosis result is limited by the maintenance experience of maintenance personnel, and in some cases, the cause of the substation equipment fault may not be diagnosed, reducing the efficiency of substation equipment maintenance. Summary of the Invention

[0005] To solve the problems raised in the above background art, the present invention provides an intelligent fault diagnosis system and method for substation equipment based on deep learning and knowledge graph, which can quickly diagnose the cause of substation equipment faults, reduce the labor intensity of maintenance personnel, and at the same time integrate a large amount of fault diagnosis experience of substation equipment, with high accuracy in diagnosing the cause of faults, and can improve the efficiency of substation equipment maintenance.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] An intelligent fault diagnosis method for substation equipment based on deep learning and knowledge graph includes the following steps:

[0008] S1: Collect the historical fault data of substation equipment and perform preprocessing;

[0009] S2: Based on the preprocessed historical fault data, construct a knowledge graph for training substation equipment fault diagnosis, using the working environment of substation equipment, the working fault location of substation equipment, the working fault phenomenon of substation equipment, and the working fault cause of substation equipment as entities of the knowledge graph, and establish the entity relationship between the working environment of substation equipment, the working fault location of substation equipment, the working fault phenomenon of substation equipment, and the working fault cause of substation equipment;

[0010] S3: Collect the working data of substation equipment in real time. When abnormal fault data is detected, diagnose the cause of the equipment fault based on the substation equipment fault diagnosis knowledge graph.

[0011] Furthermore, in the step S2, the specific steps of constructing the knowledge graph for training the substation equipment fault diagnosis based on the preprocessed historical fault data include:

[0012] S201: Take the preprocessed historical fault data of substation equipment as the fault knowledge input;

[0013] S202: Cluster the fault knowledge input;

[0014] Select K historical fault data from the historical fault data of substation equipment in the fault knowledge input as the initial clustering centers;

[0015] Calculate the distance between each historical fault data of substation equipment and each clustering center. The expression is:

[0016]

[0017] In the formula: (x 1 , y 1 ) represents the coordinates of the historical fault data of substation equipment, and (x 2 , y 2 ) represents the coordinates of the clustering center;

[0018] Assign each historical fault data of substation equipment to the cluster of the nearest clustering center;

[0019] Recalculate the center of each cluster according to the mean value of all historical fault data of substation equipment in the current cluster;

[0020] Repeat the calculation and assignment of the distance between the historical fault data of substation equipment and the clustering center until the set number of iterations is reached and stop, and the clustering is completed;

[0021] S203: Calculate the similarity between all historical fault data in each cluster, and eliminate the historical fault data with complete similarity. The expression is:

[0022]

[0023] In the formula: (x 1 , y 1 ) represents the coordinates of a historical fault data of substation equipment, and (x 2 , y 2 ) represents the coordinates of another historical fault data of substation equipment, and the complete similarity is represented by D 2 The numerical values are the same;

[0024] S204: Extract the central words of each cluster based on the large language model as tentative entities;

[0025] S205: Calculate the similarity between the tentative entities, disambiguate, and obtain the determined entities, namely the substation equipment working environment, the substation equipment working fault location, the substation equipment working fault phenomenon, and the substation equipment working fault cause. The expression is:

[0026]

[0027] In the formula: c 1 and c 2 respectively represent different entities, and A(c 1 ) and A(c 2 ) respectively represent the attributes of different entities, that is, the attributes of historical fault data;

[0028] S206: Determine the contribution degree of each connection factor between two connected determined entities based on the historical fault data under each determined entity. The expression is:

[0029]

[0030] In the formula: A n and A n-1 represent the historical fault data of the previous entity between two connected determined entities at times n and n - 1, and B n and B n-1 represent the historical fault data of the latter entity between two connected determined entities at times n and n - 1;

[0031] S207: Based on the contribution degree of the connection factors between two connected determined entities, establish the entity relationship between the substation equipment working environment, the substation equipment working fault location, the substation equipment working fault phenomenon, and the substation equipment working fault cause, and the construction of the substation equipment fault diagnosis knowledge graph is completed.

[0032] Further, in step S3, the specific steps for diagnosing the cause of the abnormal fault data of the substation equipment based on the substation equipment fault diagnosis knowledge graph include:

[0033] S301: The large language model compares the detected abnormal fault data with the clustering data of the determined entities, and converts the abnormal fault data into the data to be fault diagnosis analyzed for each determined entity;

[0034] Compare the abnormal fault data with a certain clustering data between the abnormal fault data and each determined entity. The expression is:

[0035]

[0036] Where: (x 1 , y 1 ) represents the coordinates of the abnormal fault data, and (x 2 , y 2 ) represents the coordinates of a certain data for determining entity clustering;

[0037] If the similarity between a certain part of the abnormal fault data and a certain data in the clustering of each determined entity exceeds the preset threshold, then the certain data in the clustering of the determined entity is used as the data to be analyzed for fault diagnosis of the determined entity. Otherwise, continue the comparison until the similarity between a certain part of the abnormal fault data and a certain data in the clustering of each determined entity exceeds the preset threshold, and obtain the data to be analyzed for fault diagnosis of each determined entity;

[0038] S302: Based on the connection relationships established among the determined entities, convert the data to be analyzed for fault diagnosis of each determined entity, and obtain the reasons for the working faults of the substation equipment.

[0039] Furthermore, in step S301, when comparing the abnormal fault data with the clustering data of the determined entity, the weight value of the clustering data with the highest similarity to the abnormal fault data in the clustering data of the determined entity is incremented by one. When the large language model conducts similarity comparison next time, it preferentially uses the clustering data of the determined entity with a high weight value first.

[0040] The intelligent fault diagnosis system for substation equipment based on deep learning and knowledge graph includes:

[0041] A data acquisition module that acquires the historical fault data and real-time working data of the substation equipment and performs preprocessing;

[0042] A knowledge graph construction module that constructs a training knowledge graph for substation equipment fault diagnosis based on the preprocessed historical fault data. Using the working environment of the substation equipment, the working fault location of the substation equipment, the working fault phenomenon of the substation equipment, and the working fault cause of the substation equipment as the entities of the knowledge graph, establish the entity relationships among the working environment of the substation equipment, the working fault location of the substation equipment, the working fault phenomenon of the substation equipment, and the working fault cause of the substation equipment;

[0043] A fault diagnosis module that, when abnormal fault data appears in the real-time working data of the substation equipment, diagnoses the cause of the equipment fault based on the substation equipment fault diagnosis knowledge graph.

[0044] Compared with the prior art, the beneficial effects of the present invention are:

[0045] 1) The present invention constructs a knowledge graph for fault diagnosis of substation equipment based on historical fault data of substation equipment, and conducts intelligent diagnosis of the fault causes of substation equipment with abnormal fault data in the later stage based on the knowledge graph for fault diagnosis of substation equipment. Compared with the prior art, it can quickly diagnose the fault causes of substation equipment, reduce the labor intensity of maintenance personnel, and at the same time integrate a large amount of fault diagnosis experience of substation equipment, with high accuracy in diagnosing fault causes and capable of improving the maintenance efficiency of substation equipment.

[0046] 2) The abnormal fault data of substation equipment in the later stage of the present invention can be used as the basis for improving the weight value of entity clustering data determined by the knowledge graph, enabling the entities determined by the knowledge graph to continuously learn, complementing the learning of large language models, and capable of improving the speed of the system in diagnosing the fault causes of substation equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 is the flowchart of the method of the present invention;

[0048] Figure 2 is the system framework diagram of the present invention;

[0049] In the figure: 1. Data acquisition module; 2. Knowledge graph construction module; 3. Fault diagnosis module. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0050] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0051] As Figure 1 shown, the intelligent fault diagnosis method for substation equipment based on deep learning and knowledge graph of the present invention includes the following steps:

[0052] S1: Collect historical fault data of substation equipment and perform preprocessing;

[0053] The preprocessing includes deleting missing data values, deleting abnormal data values, deleting duplicate data values, data standardization, and data normalization;

[0054] S2: Based on the preprocessed historical fault data, construct a training knowledge graph for fault diagnosis of substation equipment, using the working environment of substation equipment, the working fault location of substation equipment, the working fault phenomenon of substation equipment, and the working fault cause of substation equipment as entities of the knowledge graph, and establish entity relationships among the working environment of substation equipment, the working fault location of substation equipment, the working fault phenomenon of substation equipment, and the working fault cause of substation equipment;

[0055] The working environment of substation equipment includes deserts, plateaus, mountains, etc.;

[0056] The working fault locations of substation equipment include display and control screens, brakes, etc. The working fault locations of each substation equipment are determined by its own structure;

[0057] The working fault phenomena of substation equipment include electric leakage, oil leakage, etc. The working fault phenomena of each substation equipment are determined by its own structure;

[0058] The working fault causes of substation equipment include line short - circuit, open - circuit, etc. The working fault causes of each substation equipment are determined by its own structure;

[0059] S3: Real - time collect the working data of substation equipment. When abnormal fault data is detected, diagnose the equipment fault cause based on the substation equipment fault diagnosis knowledge graph;

[0060] Preset the normal working data of substation equipment, compare the working data of the real - time collected substation equipment with the preset normal working data, and detect abnormal fault data.

[0061] Specifically, in step S2, the specific steps of constructing a training substation equipment fault diagnosis knowledge graph based on the pre - processed historical fault data include:

[0062] S201: Use the pre - processed historical fault data of substation equipment as fault knowledge input;

[0063] S202: Cluster the fault knowledge input;

[0064] Select K historical fault data from the historical fault data of substation equipment in the fault knowledge input as the initial clustering centers;

[0065] Calculate the distance between each historical fault data of substation equipment and each clustering center. The expression is:

[0066]

[0067] In the formula: (x 1 , y 1 ) represents the coordinates of the historical fault data of substation equipment, and (x 2 , y 2 ) represents the coordinates of the clustering center;

[0068] Assign each historical fault data of substation equipment to the cluster of the nearest clustering center;

[0069] Recalculate the center of each cluster according to the mean value of all historical fault data of substation equipment in the current cluster;

[0070] Repeat the calculation and assignment of the distances between the historical fault data of substation equipment and the clustering centers until the set number of iterations is reached and the clustering is completed;

[0071] S203: Calculate the similarity between all historical fault data in each cluster, and eliminate the historical fault data with complete similarity. The expression is:

[0072]

[0073] In the formula: (x 1 , y 1 ) represents the coordinates of a piece of historical fault data of substation equipment, (x 2 , y 2 ) represents the coordinates of another piece of historical fault data of substation equipment, and the complete similarity is expressed as D 2 The numerical values are the same;

[0074] S204: Extract the central words of each cluster based on the large language model as tentative entities;

[0075] S205: Calculate the similarity between the tentative entities, disambiguate, and obtain the definite entities, namely the working environment of substation equipment, the working fault location of substation equipment, the working fault phenomenon of substation equipment, and the working fault cause of substation equipment. The expression is:

[0076]

[0077] In the formula: c 1 and c 2 respectively represent different entities, and A(c 1 ) and A(c 2 ) respectively represent the attributes of different entities, that is, the attributes of historical fault data;

[0078] S206: Determine the contribution degree of each connection factor between two connected definite entities based on the historical fault data under each definite entity. The expression is:

[0079]

[0080] In the formula: A n and A n-1 represent the historical fault data of the previous entity between two connected definite entities at times n and n - 1, and B n and B n-1 represent the historical fault data of the latter entity between two connected definite entities at times n and n - 1;

[0081] S207: Based on the connection factor contribution between the two connected entities, the entity relationship between the substation equipment working environment, the substation equipment working fault location, the substation equipment working fault phenomenon and the substation equipment working fault cause is established, and the substation equipment fault diagnosis knowledge graph is constructed.

[0082] Specifically, in step S3, the specific steps of diagnosing the cause of abnormal fault data of substation equipment based on the substation equipment fault diagnosis knowledge graph include:

[0083] S301: the large language model compares the detected abnormal fault data with the cluster data of the determined entities for similarity, and converts the abnormal fault data into data to be diagnosed and analyzed for the faults of the determined entities;

[0084] The similarity between abnormal fault data and a certain data of clusters between determined entities is compared, and the expression is:

[0085]

[0086] Where: (x 1 ,y 1 ) represents the coordinates of abnormal fault data, (x 2 ,y 2 ) represents the coordinates of a data for determining entity clustering;

[0087] If the similarity between a certain part of the abnormal fault data and a certain data of the cluster between each determined entity exceeds a preset threshold, the certain data of the cluster of the determined entity is used as the data to be diagnosed and analyzed for the determined entity; otherwise, the comparison is continued until the similarity between a certain part of the abnormal fault data and a certain data of the cluster between each determined entity exceeds a preset threshold, and the data to be diagnosed and analyzed for the determined entity is obtained;

[0088] S302: converting the to-be-determined fault diagnosis and analysis data of each determined entity to obtain the cause of the working fault of the substation equipment based on the connection relationship established between the determined entities.

[0089] Specifically, in step S301, when the abnormal fault data is compared with the cluster data of the determined entity, the weight value of the cluster data of the determined entity with the highest similarity to the abnormal fault data is increased by one, and the cluster data of the determined entity with the highest weight value is given priority when the large language model performs similarity comparison next time.

[0090] like Figure 2 As shown, the intelligent fault diagnosis system for substation equipment based on deep learning and knowledge graph of the present invention includes:

[0091] Data acquisition module 1, collects historical fault data and real-time working data of substation equipment and performs preprocessing;

[0092] The knowledge graph construction module 2 constructs a training substation equipment fault diagnosis knowledge graph based on the preprocessed historical fault data, uses the working environment of substation equipment, the working fault location of substation equipment, the working fault phenomenon of substation equipment, and the working fault cause of substation equipment as entities of the knowledge graph, and establishes the entity relationships among the working environment of substation equipment, the working fault location of substation equipment, the working fault phenomenon of substation equipment, and the working fault cause of substation equipment;

[0093] The fault diagnosis module 3 diagnoses the equipment fault cause based on the substation equipment fault diagnosis knowledge graph when abnormal fault data appears in the real-time working data of substation equipment.

[0094] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principle and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. Intelligent fault diagnosis method for substation equipment based on deep learning and knowledge graph, characterized in that: The following steps are involved: S1: Collect historical fault data of substation equipment and pre-process it; S2: Based on the pre-processed historical fault data, a training substation equipment fault diagnosis knowledge graph is constructed. The substation equipment working environment, substation equipment working fault location, substation equipment working fault phenomenon and substation equipment working fault cause are used as entities of the knowledge graph, and the entity relationship between the substation equipment working environment, substation equipment working fault location, substation equipment working fault phenomenon and substation equipment working fault cause is established; S3: Collect substation equipment working data in real time. When abnormal fault data is detected, diagnose the cause of equipment failure based on the substation equipment fault diagnosis knowledge graph.

2. According to claim 1, the intelligent fault diagnosis method for substation equipment based on deep learning and knowledge graph is characterized in that: In step S2, the specific steps of constructing a training substation equipment fault diagnosis knowledge graph based on the preprocessed historical fault data include: S201: using the pre-processed historical fault data of substation equipment as fault knowledge input; S202: clustering fault knowledge input; S203: Calculate the similarity between all historical fault data in each cluster, and remove historical fault data with complete similarity. The expression is: Where: (x1, y1) represents the coordinates of the historical fault data of one substation equipment, (x2, y2) represents the coordinates of the historical fault data of another substation equipment, and the complete similarity is represented by the same D2 value; S204: extracting the central word of each cluster as a tentative entity based on the large language model; S205: Calculate the similarity between the tentative entities, eliminate ambiguity, and obtain the determined entities, namely, the substation equipment working environment, the substation equipment working fault location, the substation equipment working fault phenomenon and the substation equipment working fault cause, which are expressed as: Where: c1 and c2 represent different entities, A(c1) and A(c2) represent the attributes of different entities, that is, the attributes of historical fault data; S206: Determine the contribution of each connection factor between two connected determination entities based on the historical fault data of each determination entity, and the expression is: Where: A n and A n-1 It is represented as the historical fault data of the previous entity at time n and n-1 between two connected entities, B n and B n-1 It is represented as the historical fault data of the latter entity between two connected entities at time n and n-1; S207: Based on the connection factor contribution between the two connected entities, the entity relationship between the substation equipment working environment, the substation equipment working fault location, the substation equipment working fault phenomenon and the substation equipment working fault cause is established, and the substation equipment fault diagnosis knowledge graph is constructed.

3. According to claim 2, the intelligent fault diagnosis method for substation equipment based on deep learning and knowledge graph is characterized in that: In step S202, clustering the fault knowledge input is specifically performed as follows: Select K historical fault data from the historical fault data of substation equipment input by fault knowledge as the initial clustering centers; Calculate the distance between each substation equipment historical fault data and each cluster center. The expression is: Where: (x1, y1) represents the coordinates of the historical fault data of the substation equipment, (x2, y2) represents the coordinates of the cluster center; Assign the historical fault data of each substation equipment to the cluster closest to the cluster center; Recalculate the center of each cluster according to the mean of the historical fault data of all substation equipment in the current cluster; Repeat the distance calculation and allocation between the historical fault data of substation equipment and the cluster center until the set number of iterations is reached and the clustering is completed.

4. According to claim 3, the intelligent fault diagnosis method for substation equipment based on deep learning and knowledge graph is characterized in that: In step S3, the specific steps of diagnosing the cause of abnormal fault data of substation equipment based on the substation equipment fault diagnosis knowledge graph include: S301: the large language model compares the detected abnormal fault data with the cluster data of the determined entities for similarity, and converts the abnormal fault data into data to be diagnosed and analyzed for the faults of the determined entities; The similarity between abnormal fault data and a certain data of clusters between determined entities is compared, and the expression is: Where: (x1, y1) represents the coordinates of abnormal fault data, (x2, y2) represents the coordinates of a certain data of the determined entity cluster; If the similarity between a certain part of the abnormal fault data and a certain data of the cluster between each determined entity exceeds a preset threshold, the certain data of the cluster of the determined entity is used as the data to be diagnosed and analyzed for the determined entity; otherwise, the comparison is continued until the similarity between a certain part of the abnormal fault data and a certain data of the cluster between each determined entity exceeds a preset threshold, and the data to be diagnosed and analyzed for the determined entity is obtained; S302: converting the to-be-determined fault diagnosis and analysis data of each determined entity to obtain the cause of the working fault of the substation equipment based on the connection relationship established between the determined entities.

5. According to claim 4, the intelligent fault diagnosis method for substation equipment based on deep learning and knowledge graph is characterized in that: In step S301, when the abnormal fault data is compared with the cluster data of the determined entity, the weight value of the cluster data of the determined entity with the highest similarity to the abnormal fault data is increased by one, and the cluster data of the determined entity with the highest weight value is given priority when the large language model performs similarity comparison next time.

6. Intelligent fault diagnosis system for substation equipment based on deep learning and knowledge graph, characterized by: It includes a data acquisition module (1) for collecting historical fault data and real-time working data of substation equipment and performing pre-processing; The knowledge graph construction module (2) constructs a training substation equipment fault diagnosis knowledge graph based on the preprocessed historical fault data, takes the substation equipment working environment, the substation equipment working fault location, the substation equipment working fault phenomenon and the substation equipment working fault cause as entities of the knowledge graph, and establishes the entity relationship between the substation equipment working environment, the substation equipment working fault location, the substation equipment working fault phenomenon and the substation equipment working fault cause; The fault diagnosis module (3) detects abnormal fault data in the real-time working data of the substation equipment and diagnoses the cause of the equipment fault based on the substation equipment fault diagnosis knowledge graph.

Images