Rail transit power supply system fault diagnosis method based on knowledge graph

Abstract

The invention discloses a rail transit power supply system fault diagnosis method based on a knowledge graph, and relates to the technical field of rail transit, and the method comprises the following steps: collecting real-time monitoring data of a rail transit power supply system; constructing a superior-subordinate power supply relation graph of the rail transit power supply system; and inputting the real-time monitoring data of the rail transit power supply system and the superior and subordinate power supply relation graph into a diagnosis system comprising a fault knowledge graph configuration tool and a correlation analysis configuration tool, and performing fault diagnosis on the rail transit power supply system by the diagnosis system in combination with expert experience and a correlation statistical conclusion. According to the method, the fault analysis process is manually added in the form of a configuration tool, accurate utilization of expert experience can be ensured, and meanwhile, a diagnosis result is more credible under the assistance of a conclusion of statistical correlation analysis. According to the method, expert experience and knowledge are considered, the conclusion of statistical correlation analysis is also considered, and the diagnosis result is more credible.

Description

Technical Field

[0001] The present invention relates to the technical fields of knowledge graphs, power supply systems, fault diagnosis, and rail transit technology. More specifically, it relates to a method for fault diagnosis of rail transit power supply systems based on knowledge graphs. Background Art

[0002] Rail transit refers to a type of transportation vehicle or transportation system where the operating vehicles need to travel on specific tracks. The most typical rail transit is the railway system composed of traditional trains and standard railways. With the diversified development of train and railway technologies, there are more and more types of rail transit, which are not only spread over long-distance land transportation but also widely used in medium and short-distance urban public transportation, such as subway public transportation, playing an important role in various activities related to people's production and life.

[0003] The safe operation of rail transit depends on a safe, standardized, and reliable power supply system. The power supply system is the blood of rail transit transportation and the core system of rail transit. Once a fault or interruption occurs in the power supply system, it will not only cause the paralysis of urban rail transit transportation but also endanger the lives of passengers. At the same time, it will also bring great pressure to the ground public transportation and have an adverse impact on social stability and urban image.

[0004] In the prior art, the patent with the publication number CN111311059A discloses a method for fault diagnosis of a waterwheel chamber based on a knowledge graph, which relates to the technical field of water turbine unit fault diagnosis. This method constructs a knowledge graph of waterwheel chamber faults, which can solve the difficulties in establishing a knowledge base of unstructured texts in the past. By using a Bayes network to autonomously construct a network by learning historical data and learning network parameters, it can express and reason about uncertain knowledge, and the reasoning conclusion is accurate, greatly improving the application effect of waterwheel chamber fault diagnosis in engineering practice. This method can effectively reflect potential problems and hidden dangers during operation, judge and warn about the deterioration trend during operation, and can fundamentally predict and sense abnormal changes in the top cover water level in advance. When a sensor fault or abnormality occurs, it can also diagnose the fault, accurately describe a large number of uncertain factors in the water level change of the waterwheel chamber, have consistent and coherent reasoning, a simple process, and extremely high diagnostic accuracy.

[0005] Defect: This solution mainly uses machine learning algorithms to construct a knowledge graph by extracting knowledge from historical texts. Limited by the accuracy of the model, there may be deviations in the recognition of special names and deviations in the understanding of text semantics. For scenarios with high requirements for analysis accuracy such as fault diagnosis and analysis of rail transit power supply specialties, this solution cannot be effectively implemented.

[0006] The patent with the publication number CN116561302A discloses a fault diagnosis method, device and storage medium based on hybrid knowledge graph reasoning. The method includes: acquiring data related to steel production line faults and equipment, and constructing a fault knowledge graph; constructing a hybrid knowledge graph reasoning model, where the hybrid knowledge graph reasoning model includes a knowledge graph embedding model constructed based on graph attention mechanism and TranSparse, and a knowledge graph reasoning model constructed based on logical rules and reinforcement learning; training the hybrid knowledge graph reasoning model; and using the trained hybrid knowledge graph reasoning model to diagnose faults in steel production line equipment. Compared with the prior art, the present invention has the advantages of high accuracy and stable training.

[0007] Defect: This solution constructs a knowledge graph using multiple models, but still relies on machine learning for knowledge mining and extraction, and the accuracy of knowledge construction is limited. In scenarios such as fault diagnosis in the rail transit power supply specialty where high analysis accuracy is required, this solution cannot be effectively implemented.

[0008] The patent with the publication number CN115619383A discloses a fault diagnosis method, device and computing device based on a knowledge graph, which relates to the technical field of fault handling. The method includes: constructing a knowledge graph according to multi-source knowledge of fault information; where the multi-source knowledge includes expert experience knowledge, case knowledge, and data stream information; the knowledge graph includes entities corresponding to the fault information and the relationships between the entities; acquiring the fault information to be diagnosed; using a pre-trained fault model to diagnose the fault information to be diagnosed to obtain a diagnosis result; where the diagnosis result includes the fault cause and warning information; determining the entities related to the fault cause based on the knowledge graph; where the entities include fault cases, solutions, and warning information. The fault diagnosis method based on the knowledge graph provided by this solution realizes intelligent fault diagnosis and processing based on multi-source heterogeneous information.

[0009] Defect: This solution considers multi-source heterogeneous information such as expert experience knowledge and case knowledge as the basis for fault diagnosis, but the existing experience knowledge may also have biases, and objective statistical results are needed to assist the existing experience knowledge for joint fault diagnosis and analysis. Summary of the Invention

[0010] In order to overcome the defects existing in the above prior art, the present invention discloses a fault diagnosis method for a rail transit power supply system based on a knowledge graph. The present invention is a method for constructing a knowledge graph for fault diagnosis of a rail transit power supply system by combining two elements: expert experience and statistical correlation analysis. To facilitate the collection of expert experience knowledge, the present invention designs a fault knowledge graph configuration tool. To facilitate the absorption of the results of statistical correlation analysis, the present invention designs a correlation analysis configuration tool. The combined use of the two tools makes the diagnosis result more credible.

[0011] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0012] A fault diagnosis method for a rail transit power supply system based on a knowledge graph, comprising the following steps:

[0013] I. Collect real-time monitoring data

[0014] Collect real-time monitoring data of the rail transit power supply system;

[0015] Preferably, the real-time monitoring data of the rail transit power supply system includes real-time action monitoring records and real-time parameter monitoring records of equipment.

[0016] In the present invention, taking the real-time action monitoring records and real-time parameter monitoring records of equipment as inputs, fault diagnosis and analysis are performed. Assume that the equipment monitoring data is uploaded to the diagnosis system in the form of a table. By parsing the table, the diagnosis system obtains the real-time parameters and action information of the equipment.

[0017] II. Construct a power supply relationship graph

[0018] Construct an upper and lower level power supply relationship graph of the rail transit power supply system;

[0019] In the present invention, it is necessary to establish a power supply relationship graph to locate the upper-level power supply node of the faulty equipment and achieve accurate troubleshooting of the fault cause, without troubleshooting other irrelevant equipment.

[0020] III. Fault knowledge graph configuration tool and correlation analysis configuration tool

[0021] Input the real-time monitoring data and the upper and lower level power supply relationship graph of the rail transit power supply system into a diagnosis system including a fault knowledge graph configuration tool and a correlation analysis configuration tool. The diagnosis system combines expert experience and correlation statistical conclusions to perform fault diagnosis on the rail transit power supply system;

[0022] Among them, a rail transit power supply system fault knowledge graph is configured in the fault knowledge graph configuration tool, and expert experience is added to the fault knowledge graph. When a fault is triggered, the fault knowledge graph configuration tool automatically executes a diagnosis and analysis task according to expert experience; the correlation analysis configuration tool uses statistical correlation analysis means to assist in judging the cause and effect of the fault and adds the correlation statistical conclusion to the fault knowledge graph.

[0023] 3.1. Fault knowledge graph configuration tool

[0024] Preferably, the fault knowledge graph configuration tool configures the rail transit power supply system fault knowledge graph, including: configuring fault basic information, fault trigger conditions, fault causes, analysis processes, and attribution nodes in sequence.

[0025] 3.1.1. Home Page

[0026] Preferably, on the home page of the fault knowledge graph configuration tool:

[0027] On the left side of the page is the equipment classification tree. Selecting the equipment classification tree switches the device for the main interface.

[0028] At the upper part of the page is a fuzzy query window, which is used for fuzzy keyword query of the fields of fault name, fault phenomenon, fault impact, and emergency measures.

[0029] At the lower part of the page is the fault list that has been configured for the selected device. Each row in the fault list is provided with edit and delete buttons for editing and deleting each fault in the fault list. Above the fault list, there is an add new fault button for adding new faults. Both the add new fault and edit fault pages are provided with fault basic information, fault cause, and troubleshooting process tabs.

[0030] 3.1.2. Add New Fault

[0031] 3.1.2.1. Fault Basic Information

[0032] Preferably, in the fault basic information:

[0033] Fill in the fault name in the fault name input box at the upper part of the page.

[0034] Fill in the trigger condition list in the table at the upper part of the page. The fault knowledge graph configuration tool uses the trigger condition list to judge fault triggering. After the positioning, judgment, and verification links of the fault triggering judgment, if there is an event that meets all conditions, the fault is triggered, and this event is the fault triggering event.

[0035] Click the add button in the trigger condition list, and a new row will be added to the trigger condition list. Click the delete button in the trigger condition list to delete the selected row. All judgment conditions for multiple rows of data in the same table are AND judgments.

[0036] Click the add parallel trigger condition button, and a new trigger condition list will be added below. The trigger conditions in different trigger condition lists are OR judgments.

[0037] Fill in the text content in the fault phenomenon, fault impact, and emergency measures input boxes in the middle of the page.

[0038] In the disposal timeliness requirement input box in the middle of the page, first select the unit, including minutes, hours, and days, and then fill in the number.

[0039] Preferably, the trigger condition list includes a device classification, measuring point name, trigger condition, device number, device name, and installation location filling units. The filling format, filling specification, system response rule, and whether it is mandatory for each filling unit are as follows:

[0040] For device classification, the filling format is: text; the filling specification is: only one measuring point location can be filled; the system response rule is: the system makes a regular expression match in the table name of the measuring point database according to the input text to find the corresponding data table; whether it is mandatory is: yes;

[0041] For the measuring point name, the filling format is: text; the filling specification is: only one measuring point name can be filled; the system response rule is: the system makes a regular expression match in the field names of the found data table according to the input text to find the corresponding field name; whether it is mandatory is: yes;

[0042] For the trigger condition, the filling format is: text; the filling specification is: only one condition can be input; the system response rule is: make a numerical judgment on the measuring point value within the range of greater than, less than, greater than or equal to, less than or equal to, equal to, not equal to. If the measuring point value is within the range, the trigger takes effect; whether it is mandatory is: yes;

[0043] For the device number, the filling format is: text; the filling specification is: multiple numbers are input, separated by commas; the system response rule is: the system uses the input device number as the regular expression matching verification condition for the device number field in the fault trigger event; whether it is mandatory is: no;

[0044] For the device name, the filling format is: text; the filling specification is: multiple device names are input, separated by commas; the system response rule is: the system uses the input device name as the regular expression matching verification condition for the device name field in the fault trigger event; whether it is mandatory is: no;

[0045] For the installation location, the filling format is: text; the filling specification is: multiple installation locations are input, separated by commas; the system response rule is: the system uses the input installation location as the regular expression matching verification condition for the installation location field in the fault trigger event; whether it is mandatory is: no.

[0046] Preferably, the fault trigger judgment includes: sequentially performing device classification positioning, measuring point name positioning, trigger condition judgment, device number regular expression verification, device name regular expression verification, and installation location regular expression verification. After passing the positioning, judgment, and verification links, if there is an event that meets all conditions, the fault is triggered, and this event is the fault trigger event.

[0047] 3.1.2.2. Fault Cause

[0048] Preferably, in the fault cause:

[0049] The fuzzy query box at the upper part of the page is used for fuzzy query of the fields of fault cause and disposal measures;

[0050] The list of added fault causes is presented in the page;

[0051] The button for adding fault causes above the list of fault causes is used to add fault causes;

[0052] The edit button in the list of fault causes is used to edit the fault causes.

[0053] 3.1.2.3, Troubleshooting process

[0054] Preferably, in the troubleshooting process:

[0055] The fuzzy query box at the upper part of the page is used for fuzzy query of the node name and attribution reason fields in the node list at the lower part of the page;

[0056] The created nodes are presented in the form of a node list at the lower part of the page, and the node list is arranged in sequence including level 1 nodes, level 2 nodes, and level 3 nodes;

[0057] A button for adding level 1 nodes is set in the middle of the page to add a row of level 1 nodes in the node list;

[0058] An edit button is set in each row of nodes in the node list to edit the node details;

[0059] A button for viewing the knowledge graph is set in the middle of the page to view the knowledge graph.

[0060] Preferably, the node details of level 1 nodes include:

[0061] Fill in the node name in the input box of the node name field at the upper part of the page;

[0062] Select the troubleshooting type in the troubleshooting type selection box at the upper part of the page, including automatic troubleshooting and manual troubleshooting;

[0063] The added judgment conditions are presented in the form of a list on the left side at the lower part of the page;

[0064] Click the plus button below the judgment condition list to add a new row of judgment conditions in the judgment condition list;

[0065] Select a row in the judgment condition list and click the minus button to delete the selected row;

[0066] Click the configuration button in the judgment condition list to display the judgment rules in the configuration interface on the right side at the lower part of the page;

[0067] The judgment rule list is presented in the configuration interface;

[0068] Enter the difference standard in the difference standard input box above the judgment rule list; if, in the judgment rule list, for the measured point values between rows at the same time point, the pairwise differences are within the input difference standard range, the judgment takes effect.

[0069] Fill in the judgment rule list and perform logical judgment. If an event that meets all conditions can be found, the judgment takes effect, and this event is used as the diagnostic event.

[0070] Click the add button in the judgment rule list, and a new row will be added to the list. Click the delete button in the list to delete the selected row. All judgment conditions for multiple rows of data in the same table are AND judged.

[0071] Click the add parallel judgment button, and a new judgment rule list will be added below. The judgment rules in different judgment rule lists are OR judged.

[0072] Preferably, the judgment rule list includes equipment classification, measuring point name, time range, trigger condition, mutation difference, equipment number, equipment name, and installation location filling units. The filling format, filling specification, system response rule, and whether it is required to be filled for each filling unit are as follows:

[0073] In equipment classification, the filling format is: text; the filling specification is: only one measuring point location can be filled; the system response rule is: the system makes a regular expression match in the table name of the measuring point database according to the input text to find the corresponding data table; whether it is required to be filled is: yes.

[0074] In measuring point name, the filling format is: text; the filling specification is: only one measuring point name can be filled; the system response rule is: the system makes a regular expression match in the field names of the found data table according to the input text to find the corresponding field name; whether it is required to be filled is: yes.

[0075] In time range, the filling format is: text; the filling specification is: only one range can be input; the system response rule is: taking the fault trigger event as 0 o'clock, and using the range of t1 seconds before 0 o'clock and t2 seconds after 0 o'clock to screen the time of the event; whether it is required to be filled is: yes.

[0076] In trigger condition, the filling format is: text; the filling specification is: only one condition can be input; the system response rule is: make a numerical judgment on the measured point value within the range of greater than, less than, greater than or equal to, less than or equal to, equal to, not equal to. If the measured point value is within the range, the judgment takes effect; whether it is required to be filled is: if the difference standard is not filled, one of the trigger condition and mutation difference must be filled. If the difference standard is filled, it is not required to be filled.

[0077] For the mutation difference value, the filling format is: numerical value; the filling specification is: only one numerical value can be entered; the system response rule is: for the measured point value, perform a range comparison of the difference according to the input judgment condition. If the difference is within the range, the judgment takes effect; whether it is mandatory: if the difference standard is not filled, one of the trigger condition and the mutation difference value must be filled. If the difference standard is filled, it is not mandatory;

[0078] For the equipment number, the filling format is: text; the filling specification is: enter multiple numbers separated by commas; the system response rule is: the system uses the entered equipment number as the regular matching condition for the equipment number field in the diagnostic event; whether it is mandatory: no;

[0079] For the equipment name, the filling format is: text; the filling specification is: enter multiple equipment names separated by commas; the system response rule is: the system uses the entered equipment name as the regular matching and verification condition for the equipment name field in the diagnostic event; whether it is mandatory: no;

[0080] For the installation location, the filling format is: text; the filling specification is: enter multiple installation locations separated by commas; the system response rule is: the system uses the entered installation location as the regular matching and verification condition for the installation location field in the query event; whether it is mandatory: no.

[0081] Preferably, the details of the 2nd-level node and subsequent nodes include:

[0082] Fill in the node name in the node name field input box at the upper part of the page;

[0083] Click the previous step condition button at the upper part of the page, and the judgment conditions set in the previous node will be presented on the page;

[0084] In the attribute to local fault at the upper part of the page, select the configured fault cause of the current equipment from the drop-down list. The node that attributes to the local fault will no longer be investigated;

[0085] In the attribute to external fault at the upper part of the page, select the configured fault name of other equipment from the drop-down list. For the node that attributes to the external fault, the investigation will be triggered according to the selected fault name.

[0086] Preferably, in the view of the knowledge graph, the outer frames of the nodes with the fault cause checked are displayed in different colors. Click on the nodes in the knowledge graph to enter the node details of the 1st-level node or the details of the 2nd-level node and subsequent nodes.

[0087] 3.2. Correlation analysis configuration tool

[0088] Preferably, the configuration of the correlation analysis configuration tool includes: sequentially configuring the abnormal judgment standard for the monitoring point, the pre-order abnormal judgment standard, the post-order abnormal judgment standard, and the investigation time range.

[0089] 3.2.1, Home Page

[0090] Preferably, on the home page of the correlation analysis configuration tool:

[0091] The drop-down box above the page can be used to select the fault name, which is sourced from the fault names configured in the fault knowledge graph;

[0092] Click the "Add Analysis Item" button above the page to enter the details of a blank analysis item;

[0093] The analysis results are displayed in a graphical interface in the middle of the page, where: the middle block represents the fault name, the upper block represents the antecedents of the fault analyzed by the system, and the lower block represents the consequences of the fault analyzed by the system;

[0094] The system automatically adds the antecedents of the fault to the nodes of the knowledge graph of this fault. When the fault cause cannot be found according to the process of the fault knowledge graph configuration tool, the results of the correlation analysis are pushed to the user;

[0095] The consequences of the system fault are added to the fault impact field of this fault;

[0096] Click the block representing the antecedent or consequence, and the list below shows the event details, including: previous events, subsequent events, out-of-scope events;

[0097] Click the block representing the fault name to enter the details of the configured analysis item.

[0098] Preferably, the details of the analysis item include:

[0099] The fault name selection box above the page, select the fault name by dropping down, the fault name is sourced from the fault names configured in the fault knowledge graph, and the configured fault trigger conditions are followed;

[0100] Enter percentages in the input boxes for the number of subsequent mutations / total number of faults, number of previous mutations / total number of faults, and number of previous mutations / total number of mutations above the page;

[0101] In the trigger interval drop-down box above the page, select 1 month, 2 months, 3 months, or 6 months by dropping down;

[0102] Output text in the time range input box above the page;

[0103] After clicking "OK", the configuration takes effect, and the system performs correlation analysis according to the configured content;

[0104] The results of the most recent correlation analysis are displayed in the list below the page;

[0105] Click on the details in the list to enter the details of the abnormal event;

[0106] Check the confirmation box in the list and click the confirmation button above the list, then the analysis result of this row will take effect and will be added to the graphical interface;

[0107] Click on "Analyze Immediately" above the list to immediately start a correlation analysis.

[0108] 3.2.2, Calculation Logic

[0109] Preferably, the calculation logic of the correlation analysis configuration tool includes a measurement point mutation standard, an event qualification standard, a frequency standard, and an event evaluation standard.

[0110] Preferably, in the measurement point mutation standard:

[0111] For DI-type measurement points: For all DI-type measurement points, if the value changes, it is regarded as a mutation;

[0112] For AI-type measurement points: For all AI-type measurement points, with a calculation interval of 3 months, each time a 5-month time window is used as the time window to extract data, calculate the mean value M and variance S. Measurement points where the numerical change amount exceeds 2S and the numerical value exceeds the range of (M - 2S, M + 2S) are regarded as mutations.

[0113] Preferably, in the event qualification standard:

[0114] Preceding mutation: Among the measurement points regarded as mutations, the measurement points that occur within the specified time range before the fault trigger time are used as preceding events;

[0115] Subsequent mutation: Among the measurement points regarded as mutations, the measurement points that occur within the specified time range after the fault trigger time are used as subsequent events;

[0116] Mutation outside the range: Among the measurement points regarded as mutations, the measurement points that occur outside the specified time range are used as events outside the range.

[0117] Preferably, in the frequency standard:

[0118] Fault frequency: The total number of specified fault triggers monitored by the system;

[0119] Total mutation frequency: Taking the equipment classification - measurement point name as the smallest measurement point unit, respectively count the total mutation frequency of each measurement point unit;

[0120] Preceding mutation frequency: Taking the equipment classification - measurement point name as the smallest measurement point unit, respectively count the total preceding mutation frequency of each measurement point unit;

[0121] Number of subsequent mutations: Taking the equipment classification - measurement point name as the minimum measurement point unit, respectively count the total number of subsequent mutations for each measurement point unit.

[0122] Number of mutations outside the range: Taking the equipment classification - measurement point name as the minimum measurement point unit, respectively count the total number of mutations outside the range for each measurement point unit.

[0123] Preferably, in the event evaluation criteria:

[0124] Antecedent evaluation: The number of previous mutations / the total number of faults > the set threshold, and the number of previous mutations / the total number of mutations > the set threshold;

[0125] Consequence evaluation: The number of subsequent mutations / the total number of faults > the set threshold.

[0126] Advantages of the present invention:

[0127] 1. Compared with a fault diagnosis method for a waterwheel room based on a knowledge graph. This solution mainly uses machine learning algorithms to construct a knowledge graph by extracting knowledge from historical texts. Limited by the accuracy of the model, there may be deviations in the recognition of special names and in the understanding of text semantics. For scenarios such as fault diagnosis and analysis in the rail transit power supply specialty where high analysis accuracy is required, this solution cannot be effectively implemented.

[0128] However, in the present invention, the fault analysis process is manually added in the form of a configuration tool, which can ensure the accurate utilization of expert experience. At the same time, supplemented by the conclusions of statistical correlation analysis, the diagnostic results are more credible.

[0129] 2. Compared with a fault diagnosis method, device, and storage medium based on hybrid knowledge graph reasoning. This solution uses multiple models to construct a knowledge graph, but still relies on machine learning for knowledge mining and extraction, and the accuracy of knowledge construction is limited. In scenarios such as fault diagnosis in the rail transit power supply specialty where high analysis accuracy is required, this solution cannot be effectively implemented.

[0130] However, in the present invention, the fault analysis process is manually added in the form of a configuration tool, which can ensure the accurate utilization of expert experience. At the same time, supplemented by the conclusions of statistical correlation analysis, the diagnostic results are more credible.

[0131] 3. Compared with a fault diagnosis method, device, and computing device based on a knowledge graph. This solution considers diverse heterogeneous information such as expert experience knowledge and case knowledge as the basis for fault diagnosis, but existing experience knowledge may also have deviations and needs to be assisted by objective statistical results for existing experience knowledge to jointly conduct fault diagnosis and analysis.

[0132] However, the present invention considers both expert experience knowledge and the conclusions of statistical correlation analysis, and the diagnostic results are more credible. Description of the Drawings

[0133] Figure 1 This is the overall working process of the fault diagnosis method for the rail transit power supply system of the present invention;

[0134] Figure 2 This is the final construction effect of the fault knowledge graph of the present invention;

[0135] Figure 3 This is a schematic diagram of the power supply relationship graph of the present invention;

[0136] Figure 4 This is the usage process of the fault knowledge graph configuration tool of the present invention;

[0137] Figure 5 This is the home page of the fault knowledge graph configuration tool of the present invention;

[0138] Figure 6 This is to add a new fault to the fault knowledge graph configuration tool of the present invention;

[0139] Figure 7 This is the fault trigger judgment logic of the fault knowledge graph configuration tool of the present invention;

[0140] Figure 8 This is the fault cause of the fault knowledge graph configuration tool of the present invention;

[0141] Figure 9 This is the details of the fault cause of the fault knowledge graph configuration tool of the present invention;

[0142] Figure 10 This is the troubleshooting process of the fault knowledge graph configuration tool of the present invention;

[0143] Figure 11 This is the details of the first-level node one of the fault knowledge graph configuration tool of the present invention;

[0144] Figure 12 This is the details of the first-level node two of the fault knowledge graph configuration tool of the present invention;

[0145] Figure 13 This is the logical judgment of the judgment rule list of the fault knowledge graph configuration tool of the present invention;

[0146] Figure 14 This is the details of the second-level and subsequent nodes one of the fault knowledge graph configuration tool of the present invention;

[0147] Figure 15 This is the details of the second-level and subsequent nodes two of the fault knowledge graph configuration tool of the present invention;

[0148] Figure 16 This is to view the knowledge graph of the fault knowledge graph configuration tool of the present invention;

[0149] Figure 17 The usage process of the configuration tool for the relevance analysis of the present invention;

[0150] Figure 18 The home page of the configuration tool for the relevance analysis of the present invention;

[0151] Figure 19 The details of the analysis items of the configuration tool for the relevance analysis of the present invention;

[0152] Figure 20 The details of the abnormal events of the configuration tool for the relevance analysis of the present invention. Detailed implementation manners

[0153] The concept, specific structure and technical effects of the present invention will be clearly and completely described below in conjunction with the embodiments and the drawings, so as to fully understand the purpose, features and effects of the present invention.

[0154] Embodiment 1

[0155] A fault diagnosis method for a rail transit power supply system based on a knowledge graph, as Figure 1 shown, includes the following steps:

[0156] Collect real-time monitoring data of the rail transit power supply system;

[0157] Construct the upper and lower power supply relationship graph of the rail transit power supply system;

[0158] Input the real-time monitoring data and the upper and lower power supply relationship graph of the rail transit power supply system into a diagnosis system including a fault knowledge graph configuration tool and a relevance analysis configuration tool. The diagnosis system combines expert experience and relevance statistical conclusions to diagnose the faults of the rail transit power supply system;

[0159] Among them, a rail transit power supply system fault knowledge graph is configured in the fault knowledge graph configuration tool, and expert experience is added to the fault knowledge graph. When a fault is triggered, the fault knowledge graph configuration tool automatically executes a diagnosis and analysis task according to expert experience; the relevance analysis configuration tool uses statistical relevance analysis means to assist in judging the causes and consequences of faults, and adds the relevance statistical conclusions to the fault knowledge graph.

[0160] This embodiment is a method for constructing a knowledge graph for fault diagnosis of a rail transit power supply system by combining two elements of expert experience and statistical relevance analysis. To facilitate the collection of expert experience knowledge, this solution designs a fault knowledge graph configuration tool. To facilitate the absorption of the results of statistical relevance analysis, this solution designs a relevance analysis configuration tool. The combined use of the two tools makes the diagnosis results more credible.

[0161] The final construction effect of the fault knowledge graph is as Figure 2 shown.

[0162] Embodiment 2

[0163] This embodiment is further elaborated on the basis of Embodiment 1. For collecting the real-time monitoring data of the rail transit power supply system, this solution takes the real-time action monitoring records and real-time parameter monitoring records of the equipment as the input for fault diagnosis and analysis. It is assumed that the equipment monitoring data is uploaded to the diagnosis system in the form of Table 1. By parsing the following table, the diagnosis system can obtain the real-time parameters and action information of the equipment.

[0164] Table 1

[0165]

[0166] Embodiment 3

[0167] This embodiment is further elaborated on the basis of Embodiment 2. For the power supply relationship graph, this solution needs to establish a power supply relationship graph to locate the upper-level power supply node of the faulty equipment and achieve accurate troubleshooting of the fault cause. It is assumed that the power supply relationship graph is as Figure 3 shown.

[0168] After establishing the Figure 3 relationship, if the equipment "Station 3 - Equipment No. 8" fails, then through the power supply relationship graph, the upper-level power supply equipment of this equipment, namely "Station 3 - Equipment No. 6", "Station 3 - Equipment No. 3", and "Station 1 - Equipment No. 1", can be found in sequence.

[0169] After finding the upper-level equipment, the diagnosis system will, according to the troubleshooting logic, only check the abnormal situation of the monitoring parameters of these equipment to accurately locate the fault cause, rather than troubleshooting other irrelevant equipment.

[0170] Embodiment 4

[0171] This embodiment further elaborates on the fault knowledge graph configuration tool on the basis of Embodiment 3.

[0172] The fault knowledge graph configuration tool is a software tool that facilitates adding expert experience to the knowledge graph. With the assistance of this tool, users can input the expert experience of fault analysis into the computer with simple operations. When a fault is triggered, the computer automatically executes the diagnostic analysis task according to the expert experience. The usage process of the fault knowledge graph configuration tool is as Figure 4 . The design method of the software is as described in 4.1 - 4.2.

[0173] 4.1 Home Page

[0174] As Figure 5 shown:

[0175] (1) On the left side of the page is the device classification tree. Select the device classification tree to switch devices on the main interface.

[0176] (2) At the upper part of the page is the fuzzy query window. Keywords can be used for fuzzy query in the fields of "fault name, fault phenomenon, fault impact, emergency measures".

[0177] (3) At the lower part of the page is the fault list already configured for the selected device. The list fields are as Figure 5 shown.

[0178] (4) Click the "Add New Fault" button above the list to enter Interface 4.2 (an empty interface).

[0179] (5) Click the "Edit" button in the list to enter Interface 4.2 (the interface with already entered content, which can be modified).

[0180] 4.2 Adding New Faults

[0181] As Figure 6 shown:

[0182] (1) There are three groups of tabs at the upper part of the page. Switch the tab to "Fault Basic Information" to enter 4.2.1; switch the tab to "Fault Cause" to enter 4.2.2; switch the tab to "Troubleshooting Process" to enter 4.2.3.

[0183] (2) Click the "Back" button to return to the previous step.

[0184] 4.2.1 Fault Basic Information

[0185] As Figure 6 shown:

[0186] (1) Fill in the fault name in the "Fault Name" input box at the upper part of the page.

[0187] (2) Fill in the trigger condition list in the table at the upper part of the page. As shown in Table 2. The fault trigger judgment logic is as Figure 7 . If after passing through Figure 7 the search, judgment, and verification steps, there is an event that meets all the conditions, then it is triggered as a fault, and this event is defined as a fault trigger event.

[0188] Table 2

[0189]

[0190]

[0191] (3) Click the "New" button in the list, then a new row will be added to the list. Click the "Delete" button in the list to delete the selected row. All judgment conditions for multiple rows of data in the same table are made with an "AND" judgment.

[0192] (4) Click the "Add Parallel Trigger Condition" button, and a new trigger condition list will be added below. The operation rules are the same as those in (2) and (3) of this chapter. The trigger conditions in different trigger lists are made with an "OR" judgment.

[0193] (5) Fill in the corresponding text content in the "Fault Phenomenon", "Fault Impact", and "Emergency Measures" input boxes in the middle of the page.

[0194] (6) In the "Disposal Timeliness Requirement" input box in the middle of the page, first select the unit, which can be "minutes, hours, days", and then fill in the number.

[0195] 4.2.2 Fault Causes

[0196] As Figure 8 shown:

[0197] (1) The fuzzy query box at the upper part of the page can perform fuzzy queries on the "Fault Causes" and "Disposal Measures" fields.

[0198] (2) A list of already added fault causes is presented on the page. The list fields are as Figure 8 shown.

[0199] (3) Click the "Add Fault Cause" button above the list to enter Interface 4.2.2.1 (an empty interface).

[0200] (4) Click the "Edit" button in the list to enter Interface 4.2.2.1 (the interface with already entered data that can be modified).

[0201] (5) Click the "Back" button to return to the previous step.

[0202] 4.2.2.1 Fault Cause Details

[0203] As Figure 9 shown:

[0204] (1) Fill in the corresponding text in the "Fault Causes" and "Disposal Measures" input / output boxes at the upper part of the page.

[0205] (2) In the "Operation Steps" drop-down selection box at the upper part of the page, multiple selections are allowed.

[0206] (3) Click the "Back" button to return to the previous step.

[0207] 4.2.3 Troubleshooting Process

[0208] As Figure 10As shown below:

[0209] (1) In the fuzzy query box at the upper part of the page, fuzzy queries can be made on the fields of "Node Name" and "Attribution Reason" in the list below the page.

[0210] (2) The created nodes are presented in the form of a list at the lower part of the page. The list fields are as Figure 10 . The list is arranged in the order of level 1 nodes, level 2 nodes, level 3 nodes...

[0211] (3) Click the "Add Level 1 Node" button above the list to add a new row of level 1 nodes in the list.

[0212] (4) Click the "Edit" button in the list to enter the interface 4.2.3.1 (for level 1 nodes) or interface 4.2.3.2 (for nodes after level 2).

[0213] (5) Click the "View Knowledge Graph" button above the list to enter 4.2.3.3.

[0214] 4.2.3.1 Details of Level 1 Nodes

[0215] As Figure 11 and 12 shown below:

[0216] (1) Enter the node name in the input box of the "Node Name" field at the upper part of the page.

[0217] (2) Select the troubleshooting type in the "Troubleshooting Type" selection box at the upper part of the page. The options are "Automatic Troubleshooting" and "Manual Troubleshooting". Currently, only automatic troubleshooting is available. The manual troubleshooting function is reserved.

[0218] (3) On the left side at the lower part of the page, the added judgment conditions are presented in the form of a list. The list fields are as Figure 11 .

[0219] (4) Click the "+" button below the judgment condition list to add a new row of judgment conditions in the list.

[0220] (5) Select a row in the judgment condition list and click the "—" button to delete the selected row.

[0221] (6) Click the "Configure" button in the judgment condition list to display the judgment rules in the configuration interface on the right side at the lower part of the page.

[0222] (7) The judgment rule list is presented in the configuration interface. The list fields are as Figure 11 .

[0223] (8) In the difference standard input box above the judgment rule list, you can enter: "<4", "<=5", ">6", "==1", "!=1", "(-3, 1)", "[2, 5]"...... If, in the list, for the measured point values between rows at the same time point, the pairwise differences are within the entered range, the judgment takes effect.

[0224] (9) Fill in the judgment rule list. The field rules are as shown in Table 3, and the logical judgment is as Figure 13 . If, after Figure 13 going through the process, an event that meets all the conditions in the figure can be found, the judgment takes effect, and this event is used as the diagnostic event.

[0225] Table 3

[0226]

[0227]

[0228] (10) Click the "Add New" button in the list, and a new row will be added to the list. Click the "Delete" button in the list to delete the selected row. All judgment conditions for multiple rows of data in the same table are made with an "AND" judgment.

[0229] (11) Click the "Add Parallel Judgment" button, and a new judgment rule list will be added below. The operation rules are the same as those in (7)(8)(9)(10) in this chapter. The judgment rules in different lists are made with an "OR" judgment.

[0230] (12) Click the "Return" button to return to the previous step.

[0231] 4.2.3.2 Details of Nodes at Level 2 and Above

[0232] As Figure 14 and 15 shown:

[0233] (1) Fill in the node name in the "Node Name" field input box at the upper part of the page.

[0234] (2) Click the "Previous Step Conditions" button at the upper part of the page to switch to Figure 14 . The judgment conditions set in the previous node will be presented on the page.

[0235] (3) In the "Attribute to Local Fault" at the upper part of the page, you can select the configured fault causes of the current device from the drop-down list. The options for fault causes are from the content configured in 4.2.2. Nodes that have been attributed to local faults will no longer be investigated.

[0236] (4) In the "Attribute External Fault" at the top of the page, you can select the fault name configured for other devices from the dropdown. The options for the fault name are sourced from the content configured in 4.2.1. For the node where the external fault is attributed, troubleshooting is triggered according to the selected fault name.

[0237] (5) Click "Troubleshoot" at the top of the page to switch to Figure 15 . Other rules are the same as those in 4.2.3.1.

[0238] 4.2.3.3 View Knowledge Graph

[0239] As Figure 16 shown:

[0240] (1) The outer frame of the node for which the fault cause is selected turns red.

[0241] (2) Click on the node in the graph to enter 4.2.3.1 or 4.2.3.2.

[0242] Example 5

[0243] Based on Example 4, this example further elaborates on the correlation analysis configuration tool.

[0244] The correlation analysis configuration tool is a software tool that, to make up for the lack of expert experience, uses statistical correlation analysis to assist in judging the "antecedents" and "consequences" of faults and can add the analysis conclusions to the fault knowledge graph. The usage process of the correlation analysis configuration tool is as Figure 17 shown. The design method of the software is as described in 5.1 - 5.2.

[0245] 5.1 Home Page

[0246] As Figure 18 shown:

[0247] (1) The dropdown box at the top of the page can be used to select the fault name. The fault names are sourced from the fault names configured in 4.2 in the fault knowledge base.

[0248] (2) Click the "New Analysis Item" button at the top of the page to enter Section 5.1.1, a blank page.

[0249] (3) The analysis results are displayed in a graphical interface in the middle of the page. The middle block represents the fault name (including the counted number of times), the upper block represents the fault "antecedents" (including the counted number of times) analyzed by the system (and confirmed by the user), and the lower block represents the fault "consequences" (including the counted number of times) analyzed by the system (and confirmed by the user).

[0250] (4) The system automatically adds the "antecedent" of the fault to the first-level node of the knowledge graph of this fault. When the cause of the fault cannot be found according to the process of the fault knowledge graph configuration tool, the result of the correlation analysis is pushed to the user.

[0251] (5) The "consequence" of the system fault is added to the "fault impact" field of this fault, which is convenient for users to identify the fault risk.

[0252] (6) Click on the tile representing the antecedent or consequence, and the list below shows the event details. The fields are as in this section Figure 18 , and the event classifications include: previous event, subsequent event, out-of-scope event.

[0253] (7) Click on the tile representing the fault name to enter Section 5.1.1, the configured page.

[0254] 5.1.1 Analysis Project Details

[0255] As Figure 19 shown:

[0256] (1) In the "fault name" selection box at the top of the page, the fault name can be selected by dropping down. The fault name comes from the fault names configured in 4.2 of the fault knowledge base and follows the fault trigger conditions configured in 4.2.1.

[0257] (2) Enter percentages in the input boxes of "subsequent mutation times / fault total times", "previous mutation times / fault total times", and "previous mutation times / mutation total times" at the top of the page.

[0258] (3) In the "trigger interval" drop-down box at the top of the page, "1 month, 2 months, 3 months, 6 months" can be selected by dropping down.

[0259] (4) Enter text in the "time range" input box at the top of the page. For example, "(-4, 5)" represents taking the fault trigger event as 0 o'clock, with the first 4 seconds and the subsequent 5 seconds as the time range. In the time selection bar, "seconds, minutes, hours" can be selected by dropping down.

[0260] (5) After clicking "OK", the configuration takes effect. The system performs correlation analysis according to the configured content.

[0261] (6) The result of the most recent correlation analysis is displayed in the list at the bottom of the page. The list fields are as in this section Figure 19 .

[0262] (7) Click "Details" in the list to enter Section 5.1.1.1.

[0263] (8) Check the "OK" box in the list and click the "OK" button above the list, then the analysis result of this line will take effect and will be added to the graphical interface in Section 5.1 of this section.

[0264] (9) Click "Analyze Immediately" above the list to start a correlation analysis immediately.

[0265] 5.1.1.1 Details of abnormal events, such as Figure 20 shown.

[0266] 5.2 Calculation Logic

[0267] (1) Standard for measuring point mutation

[0268]

[0269] (2) Standard for event qualification

[0270]

[0271] (3) Standard for number of times

[0272]

[0273] (4) Standard for event evaluation

[0274]

[0275] The above has specifically described the implementation manners of the present invention, but the present invention is not limited to the described embodiments. Those skilled in the art can also make various equivalent variations or substitutions without departing from the spirit of the present invention, and these equivalents or substitutions are all included within the scope defined by the claims of the present invention.

Claims

1. A rail transit power supply system fault diagnosis method based on knowledge graph, characterized in that: The following steps are involved: Collect real-time monitoring data of rail transit power supply system; Construct a relationship diagram of upper and lower power supply for the rail transit power supply system; The real-time monitoring data of the rail transit power supply system and the upper and lower power supply relationship map are input into the diagnosis system including the fault knowledge map configuration tool and the correlation analysis configuration tool. The diagnosis system combines expert experience and correlation statistical conclusions to perform fault diagnosis on the rail transit power supply system; Among them, the fault knowledge graph configuration tool is configured with a rail transit power supply system fault knowledge graph, and expert experience is added to the fault knowledge graph. When a fault is triggered, the fault knowledge graph configuration tool automatically performs diagnostic analysis tasks according to expert experience; the correlation analysis configuration tool uses statistical correlation analysis to assist in determining the causes and consequences of the fault, and adds the correlation statistical conclusions to the fault knowledge graph.

2. The rail transit power supply system fault diagnosis method according to claim 1, characterized in that: On the homepage of the fault knowledge graph configuration tool: The left side of the page is the device classification tree. Select the device classification tree to switch the device to the main interface; The upper part of the page is a fuzzy query window, which is used to perform keyword fuzzy query on the fault name, fault phenomenon, fault impact, and emergency measures fields; The lower part of the page is a list of faults that have been configured for the selected device. Each row in the fault list is provided with an edit and delete button for editing and deleting each row of the fault in the fault list. Above the fault list is an add new fault button for adding a new fault. The add new fault and edit fault pages are both provided with tabs for basic fault information, fault cause, and troubleshooting process.

3. The rail transit power supply system fault diagnosis method according to claim 2, characterized in that: In the basic fault information: Fill in the fault name in the fault name input box at the top of the page; Fill in the trigger condition list in the table at the top of the page. The fault knowledge graph configuration tool uses the trigger condition list to make fault trigger judgments. After the fault trigger judgment is located, judged, and verified, if there is an event that meets all the conditions, the fault is triggered. This event is the fault trigger event. Click the Add button in the trigger condition list to add a row to the trigger condition list. Click the Delete button in the trigger condition list to delete the selected row. All judgment conditions for multiple rows of data in the same table are judged together. Click the Add Parallel Trigger Condition button to add a new trigger condition list below. The trigger conditions in different trigger condition lists are judged by OR. Fill in the text content in the Fault Phenomenon, Fault Impact, and Emergency Measures input boxes in the middle of the page; In the disposal timeliness requirement input box in the middle of the page, first select the unit, including minutes, hours, and days, and then fill in the number.

4. The rail transit power supply system fault diagnosis method according to claim 3, characterized in that: The trigger condition list includes equipment classification, measurement point name, trigger condition, equipment number, equipment name and installation location filling unit. The filling format, filling specification, system response rule and whether it is required for each filling unit include: In the equipment classification, the filling format is: text; the filling specification is: only one measuring point location can be filled in; the system response rule is: the system performs regular matching on the table name in the measuring point database according to the input text, and finds the corresponding data table; whether it is required is: yes; In the measurement point name, the filling format is: text; the filling specification is: only one measurement point name can be filled in; the system response rule is: the system performs regular matching on the field names in the found data table based on the input text, and finds the corresponding field name; whether it is required is: yes; In the trigger condition, the filling format is: text; the filling specification is: only one condition can be entered; the system response rule is: make a numerical judgment on the measurement point value greater than, less than, greater than or equal to, less than or equal to, equal to, and not equal to the range. If the measurement point value is within the range, the trigger will take effect; whether it is required is: yes; In the device number, the filling format is: text; the filling specification is: enter multiple numbers separated by commas; the system response rule is: the system uses the entered device number as the regular matching verification condition of the device number field in the fault trigger event; whether it is required is: no; In the device name, the filling format is: text; the filling specification is: enter multiple device names separated by commas; the system response rule is: the system uses the entered device name as the regular matching verification condition of the device name field in the fault trigger event; whether it is required is: no; In the installation location, the filling format is: text; the filling specification is: enter multiple installation locations, separated by commas; the system response rule is: the system uses the entered installation location as the regular matching verification condition of the installation location field in the fault trigger event; whether it is required is: no.

5. The rail transit power supply system fault diagnosis method according to claim 2, characterized in that: During the troubleshooting process: The fuzzy query box at the top of the page is used to perform fuzzy queries on the node name and attribution reason fields in the node list at the bottom of the page; The created nodes are presented in the form of a node list at the bottom of the page, and the node list is arranged in the order of level 1 nodes, level 2 nodes, and level 3 nodes; In the middle of the page, there is an Add Level 1 Node button for adding a row of Level 1 nodes to the node list; Each row of nodes in the node list has an edit button for editing the node details; There is a View Knowledge Graph button in the middle of the page for viewing the knowledge graph.

6. The rail transit power supply system fault diagnosis method according to claim 5, characterized in that: Node details for level 1 nodes include: Fill in the node name in the Node Name field input box at the top of the page; In the troubleshooting type selection box at the top of the page, select the troubleshooting type, including automatic troubleshooting and manual troubleshooting; On the lower left side of the page, the added judgment conditions are displayed in the form of a list; Click the Add button below the judgment condition list to add a new row of judgment conditions to the judgment condition list; Select the row in the judgment condition list and click the minus button to delete the selected row; Click the Configure button in the judgment condition list to display the judgment rules in the configuration interface on the right side of the page; Present a list of judgment rules in the configuration interface; Enter the difference standard in the difference standard input box above the judgment rule list; if the pairwise difference between the measured point values ​​between rows at the same time point in the judgment rule list is within the input difference standard range, the judgment takes effect; Fill in the judgment rule list and perform logical judgment. If an event that meets all the conditions can be found, the judgment takes effect and the event is used as a diagnostic event. Click the Add button in the judgment rule list to add a row to the list, and click the Delete button in the list to delete the selected row. All judgment conditions for multiple rows of data in the same table are judged together; Click the Add Parallel Judgment button to add a new judgment rule list below. The judgment rules in different judgment rule lists are made into OR judgments.

7. The rail transit power supply system fault diagnosis method according to claim 6, characterized in that: The judgment rule list includes equipment classification, measurement point name, time range, trigger condition, mutation difference, equipment number, equipment name, and installation location filling units. The filling format, filling specification, system response rules, and whether each filling unit is required include: In the equipment classification, the filling format is: text; the filling specification is: only one measuring point location can be filled in; the system response rule is: the system performs regular matching on the table name in the measuring point database according to the input text, and finds the corresponding data table; whether it is required is: yes; In the measurement point name, the filling format is: text; the filling specification is: only one measurement point name can be filled in; the system response rule is: the system performs regular matching on the field names in the found data table based on the input text, and finds the corresponding field name; whether it is required is: yes; In the time range, the filling format is: text; the filling specification is: only one range can be entered; the system response rule is: take the fault trigger event as 0 o'clock, and filter the event time by t1 seconds before 0 o'clock and t2 seconds after 0 o'clock; whether it is required is: yes; In the trigger condition, the filling format is: text; the filling specification is: only one condition can be entered; the system response rule is: make a numerical judgment on the measurement point value greater than, less than, greater than or equal to, less than or equal to, equal to, and not equal to the range. If the measurement point value is within the range, the judgment takes effect; whether it is required is: if the difference standard is not filled in, one of the trigger condition and the sudden change difference must be filled in. If the difference standard is filled in, it is not required; In the Sudden Difference, the filling format is: Numerical value; Filling specification is: Only one numerical value can be entered; System response rule is: For the measured point value, the difference range is compared according to the input judgment condition. If the difference is within the range, the judgment takes effect; Whether it is required is: If the difference standard is not filled in, one of the trigger condition and the sudden difference must be filled in. If the difference standard is filled in, it is not required; In the device number, the filling format is: text; the filling specification is: enter multiple numbers separated by commas; the system response rule is: the system uses the entered device number as the regular matching condition of the device number field in the diagnostic event; whether it is required is: no; In the device name, the filling format is: text; the filling specification is: enter multiple device names separated by commas; the system response rule is: the system uses the entered device name as the regular matching verification condition of the device name field in the diagnostic event; whether it is required is: no; In the installation location, the filling format is: text; the filling specification is: enter multiple installation locations, separated by commas; the system response rule is: the system uses the entered installation location as the regular matching verification condition of the installation location field in the query event; whether it is required is: no.

8. The rail transit power supply system fault diagnosis method according to claim 5, characterized in that: Details of Level 2 nodes and above include: Fill in the node name in the Node Name field input box at the top of the page; Click the Previous Condition button at the top of the page, and the judgment condition set in the previous node will be displayed on the page; In the attributed local fault at the top of the page, pull down and select the fault cause configured for the current device. Select the node attributed to the local fault and do not perform troubleshooting. In the Attribution External Fault area at the top of the page, pull down and select the fault name configured for other devices, select the node attributed to the external fault, and trigger troubleshooting according to the selected fault name.

9. The rail transit power supply system fault diagnosis method according to claim 1, characterized in that: On the home page of the Correlation Analysis Configuration Tool: In the drop-down box at the top of the page, you can select the fault name. The fault name comes from the fault name configured in the fault knowledge graph. Click the Add Analysis Project button at the top of the page to enter the details of the blank analysis project; The middle of the page displays the analysis results in a graphical interface, where the middle block represents the fault name, the upper block represents the fault cause analyzed by the system, and the lower block represents the fault consequence analyzed by the system; The system automatically adds the cause of the fault to the node of the knowledge graph of the fault. If the cause of the fault cannot be found according to the process of the fault knowledge graph configuration tool, the result of the correlation analysis is pushed to the user. The consequences of a system failure are added to the Failure Impact field of that failure; Click on the tile representing the cause or consequence, and the list below will show the event details, including: preceding events, subsequent events, and out-of-scope events; Click the tile representing the fault name to enter the details of the configured analysis project; The analysis project details include: In the fault name selection box at the top of the page, pull down to select the fault name. The fault name is derived from the fault name configured in the fault knowledge graph and uses the configured fault trigger condition. Enter the percentage in the input boxes of Post-sequence mutation number / Total number of faults, Pre-sequence mutation number / Total number of faults, and Pre-sequence mutation number / Total number of mutations at the top of the page; In the trigger interval drop-down box at the top of the page, select 1 month, 2 months, 3 months, or 6 months; Output the text in the time range input box at the top of the page; After clicking OK, the configuration takes effect and the system performs correlation analysis based on the configured content; The list at the bottom of the page shows the results of the most recent correlation analysis; Click on the details in the list to enter the abnormal event details; Check the OK box in the list and click the OK button above the list, and the analysis results of that row will take effect and be added to the graphical interface; Click Analyze Now above the list to start a correlation analysis immediately.

10. The rail transit power supply system fault diagnosis method according to claim 1, characterized in that: The calculation logic of the correlation analysis configuration tool includes a measurement point mutation standard, an event qualitative standard, a frequency standard and an event judgment standard; Among them, in the test point mutation standard: DI type measuring points: For all DI type measuring points, if the value changes, it is considered a mutation; AI measurement points: For all AI measurement points, the calculation interval is 3 months, and the most recent 5 months are used as the time window each time to extract data, calculate the mean M and variance S, and the measurement points with a value change exceeding 2S and a value outside the range of (M-2S, M+2S) are considered to be mutations; The event characterization criteria include: Precursor mutation: Among the measurement points regarded as mutations, the measurement points that occur before the fault trigger time and within the specified time range are regarded as the preceding events; Post-sequence mutation: Among the measurement points regarded as mutations, the measurement points that occur after the fault trigger time and within the specified time range are regarded as post-sequence events; Out-of-range mutation: Among the measurement points considered as mutations, the measurement points that occur outside the specified time range are regarded as out-of-range events; In the above number of standards: Fault times: the total number of times the specified fault is triggered detected by the system; Total number of mutations: Take equipment classification-measuring point name as the smallest measuring point unit, and count the total number of mutations for each measuring point unit; Pre-sequence mutation times: Take equipment classification-measuring point name as the smallest measuring point unit, and count the total number of pre-sequence mutations for each measuring point unit; Post-sequence mutation times: Take equipment classification-measuring point name as the smallest measuring point unit, and count the total number of post-sequence mutations for each measuring point unit; Out-of-range mutation times: Take equipment classification-measuring point name as the smallest measuring point unit, and count the total out-of-range mutation times of each measuring point unit; The event evaluation criteria include: Cause judgment: the number of preceding mutations / total number of faults> the set threshold, and the number of preceding mutations / total number of mutations> the set threshold; Consequence judgment: number of subsequent mutations / total number of failures > set threshold.

Images