High-voltage switch cabinet fault diagnosis method and system based on deep learning

Abstract

The invention provides a high-voltage switch cabinet fault diagnosis method and system based on deep learning, and the method comprises the steps: dividing a switch cabinet into a plurality of regions based on the operation function of the switch cabinet, and collecting the operation data of the plurality of regions based on a partitioning result; wherein the operation data comprises self-changing data and due-changing data; dividing the operation data of each region based on the operation condition publication of each region; constructing a switch cabinet simulation model, inputting the operation data of each area of the switch cabinet into the simulation model for simulation training to obtain simulation result data, and comparing the simulation result data with the operation data to obtain switch cabinet fault information; and constructing a data set based on the fault information, constructing a neural network model, and training a deep learning model through the fault information to obtain a fault diagnosis model. According to the invention, based on the influence of the fault on different areas, the influence coefficient of the abnormal information of different areas on the fault is reversely deduced, and the accuracy of fault determination of the switch cabinet is further ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of switch cabinet fault diagnosis, and specifically to a high-voltage switch cabinet fault diagnosis method and system based on deep learning. Background Art

[0002] High-voltage switchgear is a key equipment in the power system, used to control, protect and distribute electrical energy. Its normal operation is crucial to the stability, reliability and safety of the power system. However, due to long-term operation, environmental factors and improper operation, high-voltage switchgear is prone to various faults, such as insulation aging, poor contact, partial discharge, mechanical failure, etc. These faults will not only affect the normal operation of the power system, but may also cause serious safety accidents.

[0003] At present, fault diagnosis of high-voltage switchgear mainly relies on regular inspections, offline tests and simple online monitoring. These methods have the following limitations:

[0004] Regular inspections: rely on manual experience, are inefficient, and have difficulty detecting early faults.

[0005] Offline testing: requires power outage, affecting the normal operation of the power system.

[0006] Simple online monitoring: Usually only a single parameter (such as temperature, current) can be monitored, which makes it difficult to fully reflect the operating status of the switch cabinet.

[0007] Insufficient data analysis: There is a lack of fusion and in-depth analysis of multi-source data, and the accuracy and timeliness of fault diagnosis are low.

[0008] Therefore, there is an urgent need for a method for accurately diagnosing switch cabinet faults. Summary of the invention

[0009] In view of the shortcomings of the prior art, the present invention provides a high-voltage switchgear fault diagnosis method and system based on deep learning to solve the problems raised in the above background technology.

[0010] To achieve the above objectives, the present invention is implemented through the following technical solutions: A high-voltage switch cabinet fault diagnosis method based on deep learning, comprising the following steps:

[0011] The switch cabinet is divided into several areas based on the operating functions, and the operating data of several areas are collected based on the partition results;

[0012] Among them, the operation data includes independent variable data and dependent variable data;

[0013] The operating data of each area is divided based on the public operation conditions of each area;

[0014] Build a switch cabinet simulation model, input the operation data of each area of ​​the switch cabinet into the simulation model for simulation training, obtain simulation result data, and obtain switch cabinet fault information by comparing the simulation result data with the operation data;

[0015] Building a data set based on the fault information, building a neural network model, and training the deep learning model through the fault information to obtain a fault diagnosis model;

[0016] The operating data of the switchgear is input into the fault diagnosis model to obtain the fault diagnosis results of the target switchgear.

[0017] As a preferred embodiment of this embodiment, the method for establishing the switch cabinet simulation model includes:

[0018] Establish a physical model of the switch cabinet to ensure that the switch cabinet model parameters are consistent with the switch cabinet reagent properties;

[0019] Set the switch cabinet model environmental parameters according to the switch cabinet working environment;

[0020] After setting all parameters, simulation calculations are performed to calculate the operating parameters and operating status of the switch cabinet. After the calculations are completed, the simulation results are analyzed and evaluated.

[0021] As a preferred embodiment of this embodiment, the comparing simulation result data and operation data to obtain switch cabinet fault information includes:

[0022] Arrange and identify the simulation result data and the independent variable data in the operation data of each area in time sequence and correspond them;

[0023] Extract key parameters from simulation result data and regional operation data;

[0024] Compare the key parameters of the independent variable data corresponding to the dependent variable data in the regional operation data with the key parameters of the corresponding simulation result data;

[0025] Determine the faulty node, analyze the key parameter characteristics of other areas based on the timing characteristics of the faulty node, determine the fault type corresponding to the faulty node, and obtain fault information.

[0026] As a preferred embodiment of this embodiment, the process of constructing a data set based on the fault information, constructing a deep learning model, and training the deep learning model through the fault information includes:

[0027] Extract characteristic data from the fault data to obtain a characteristic data set;

[0028] Divide the feature dataset into training set and validation set;

[0029] Build a deep learning model and input the training set into the deep learning model for training;

[0030] The validation set verifies the trained model. According to the number distribution of the switch cabinet areas that are judged to be faulty in the validation records, all validation records of each area are classified respectively, and the regional combination composed of different areas that can affect the fault diagnosis of the model is extracted, and the fault diagnosis influence index of the regional combination on the deep learning model is evaluated;

[0031] Based on the distribution of regions in the regional combination whose fault diagnosis impact index is greater than the preset threshold, the deep learning model is optimized.

[0032] As a preferred embodiment of this embodiment, the step of evaluating the fault diagnosis influence index of the regional combination on the deep learning model specifically includes:

[0033] Obtain verification records in which only a certain area is a fault area and determine it as fault a;

[0034] Obtain a region combination that is determined to be fault a at least once in the verification record, where the set of region combinations is M, the total number of regions in the region combination is I, and determine a set m in the set M that affects fault a;

[0035] Obtain all area combinations that are judged as fault a in the verification record and whose set contains M. The total number of areas in all area combinations is J;

[0036] Calculate the characteristic influencing factors of the fault area of ​​fault a on the fault diagnosis process of the deep learning model:

[0037] ;

[0038] Among them, γ represents the correction coefficient;

[0039] Evaluation of the impact index of regional combinations on the fault diagnosis of deep learning models:

[0040] ;

[0041] in, They respectively represent the 1st, 2nd, 3rd, ..., ith areas in the area combination of fault a, and i>2.

[0042] Another aspect of the present invention provides a high-voltage switch cabinet fault diagnosis system based on deep learning, which is used to implement the above-mentioned high-voltage switch cabinet fault diagnosis method based on deep learning, including:

[0043] The area division module divides the switch cabinet into several areas based on the operating functions, and collects the operating data of several areas based on the partition results;

[0044] Among them, the operation data includes independent variable data and dependent variable data;

[0045] The area division module is also used to divide the operation data of each area based on the operation conditions of each area;

[0046] A model building module is used to build a switch cabinet simulation model, input the operation data of each area of ​​the switch cabinet into the simulation model for simulation training, obtain simulation result data, and obtain switch cabinet fault information by comparing the simulation result data with the operation data; it is also used to build a data set based on the fault information, build a neural network model, and train the deep learning model through the fault information to obtain a fault diagnosis model;

[0047] The fault diagnosis module is used to input the operating data of the switch cabinet into the fault diagnosis model to obtain the fault diagnosis result of the target switch cabinet.

[0048] As a preferred embodiment of this embodiment, the method for establishing the switch cabinet simulation model includes:

[0049] Establish a physical model of the switch cabinet to ensure that the switch cabinet model parameters are consistent with the switch cabinet reagent properties;

[0050] Set the switch cabinet model environmental parameters according to the switch cabinet working environment;

[0051] After setting all parameters, simulation calculations are performed to calculate the operating parameters and operating status of the switch cabinet. After the calculations are completed, the simulation results are analyzed and evaluated.

[0052] As a preferred embodiment of this embodiment, the comparing simulation result data and operation data to obtain switch cabinet fault information includes:

[0053] Arrange and identify the simulation result data and the independent variable data in the operation data of each area in time sequence and correspond them;

[0054] Extract key parameters from simulation result data and regional operation data;

[0055] Compare the key parameters of the independent variable data corresponding to the dependent variable data in the regional operation data with the key parameters of the corresponding simulation result data;

[0056] Determine the faulty node, analyze the key parameter characteristics of other areas based on the timing characteristics of the faulty node, determine the fault type corresponding to the faulty node, and obtain fault information.

[0057] As a preferred embodiment of this embodiment, the process of constructing a data set based on the fault information, constructing a deep learning model, and training the deep learning model through the fault information includes:

[0058] Extract characteristic data from the fault data to obtain a characteristic data set;

[0059] Divide the feature dataset into training set and validation set;

[0060] Build a deep learning model and input the training set into the deep learning model for training;

[0061] The validation set verifies the trained model. According to the number distribution of the switch cabinet areas that are judged to be faulty in the validation records, all validation records of each area are classified respectively, and the regional combination composed of different areas that can affect the fault diagnosis of the model is extracted, and the fault diagnosis influence index of the regional combination on the deep learning model is evaluated;

[0062] Based on the distribution of regions in the regional combination whose fault diagnosis impact index is greater than the preset threshold, the deep learning model is optimized.

[0063] As a preferred embodiment of this embodiment, the step of evaluating the fault diagnosis influence index of the regional combination on the deep learning model specifically includes:

[0064] Obtain verification records in which only a certain area is a fault area and determine it as fault a;

[0065] Obtain a region combination that is determined to be fault a at least once in the verification record, where the set of region combinations is M, the total number of regions in the region combination is I, and determine a set m in the set M that affects fault a;

[0066] Obtain all area combinations that are judged as fault a in the verification record and whose set contains M. The total number of areas in all area combinations is J;

[0067] Calculate the characteristic influencing factors of the fault area of ​​fault a on the fault diagnosis process of the deep learning model:

[0068] ;

[0069] Among them, γ represents the correction coefficient;

[0070] Evaluation of the impact index of regional combinations on the fault diagnosis of deep learning models:

[0071] ;

[0072] in, They respectively represent the 1st, 2nd, 3rd, ..., ith areas in the area combination of fault a, and i>2.

[0073] The present invention provides a high-voltage switchgear fault diagnosis method and system based on deep learning, which have the following beneficial effects: by dividing the switchgear into areas, and further dividing the areas of the switchgear according to different working conditions, and using the method of establishing a simulation model, a simulation model is performed for different working conditions in different areas of the switchgear, and the simulation results are compared with the actual operation data, and abnormal data is obtained through analysis to ensure the authenticity and accuracy of the data. At the same time, the deep learning model is trained and verified through the obtained abnormal data, and in the verification record, the number distribution of areas in the switchgear that are judged to be faulty is obtained, all verification records of each area are classified respectively, and regional combinations composed of different areas that can affect the fault diagnosis of the model are extracted, and the fault diagnosis influence index of the regional combination on the deep learning model is evaluated, so as to realize the influence coefficient of abnormal information of different areas on the fault based on the influence of the fault on different areas, thereby further ensuring the accuracy of the switchgear fault judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 This is a flow chart of a high-voltage switchgear fault diagnosis method based on deep learning of the present invention;

[0075] Figure 2 This is a block diagram of the high-voltage switchgear fault diagnosis system based on deep learning of the present invention. DETAILED DESCRIPTION

[0076] The embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0077] The disclosure below provides many different embodiments or examples to realize different structures of the present invention. In order to simplify the disclosure of the present invention, the parts and settings of specific examples are described below. Of course, they are only examples, and the purpose is not to limit the present invention. In addition, the present invention can repeat reference numbers and / or reference letters in different examples, and this repetition is for the purpose of simplicity and clarity, which itself does not indicate the relationship between the various embodiments and / or settings discussed. In addition, the present invention provides various specific examples of processes and materials, but those of ordinary skill in the art can be aware of the application of other processes and / or the use of other materials.

[0078] like Figure 1 As shown, an embodiment of the present invention provides a high-voltage switch cabinet fault diagnosis method based on deep learning, comprising the following steps:

[0079] S1: The switch cabinet is divided into several areas based on its operating functions, and the operating data of several areas are collected based on the partitioning results;

[0080] Among them, the operation data includes independent variable data and dependent variable data;

[0081] In this embodiment, several areas are divided according to the operating functions, including:

[0082] Incoming line area: used to connect external power supply or upper power grid to introduce electric energy into switch cabinet;

[0083] Busbar area: As the core part of power distribution, it connects the incoming line area and the outgoing line area to realize the collection and distribution of power;

[0084] Outgoing line area: distributes electrical energy from the switch cabinet to downstream equipment or loads;

[0085] Protection and control area: monitor, protect and control the operating status of the switch cabinet;

[0086] Grounding and safety zone: ensure safe operation of equipment and prevent overvoltage and leakage accidents;

[0087] Auxiliary equipment area: provides auxiliary support for the switch cabinet, such as power supply, communication and heat dissipation;

[0088] Isolation and maintenance area: used for equipment isolation and maintenance to ensure the safety of operators;

[0089] Eight regions were obtained based on executive functions.

[0090] Among them, the operation specifically includes current data, temperature data, humidity data, vibration frequency data, historical maintenance and repair data, etc., as well as environmental physical quantities such as humidity and temperature inside and outside the switch cabinet.

[0091] S2: Divide the operating data of each area based on the public operation conditions of each area;

[0092] It can be understood that according to the operating status classification, the operating status of the switch cabinet can be divided into normal operating conditions, load conditions and standby conditions.

[0093] S3: Build a switch cabinet simulation model, input the operation data of each area of ​​the switch cabinet into the simulation model for simulation training, obtain simulation result data, and obtain switch cabinet fault information by comparing the simulation result data with the operation data;

[0094] The fault information includes the fault information of different operating conditions in each area and the connection between the fault information and each area;

[0095] In this embodiment, the switch cabinet simulation model is established in the following manner:

[0096] Establish a physical model of the switch cabinet to ensure that the switch cabinet model parameters are consistent with the switch cabinet reagent properties;

[0097] Set the switch cabinet model environmental parameters according to the switch cabinet working environment;

[0098] After setting all parameters, simulation calculations are performed to calculate the operating parameters and operating status of the switch cabinet. After the calculations are completed, the simulation results are analyzed and evaluated.

[0099] Among them, by comparing the simulation result data and the operation data, the switch cabinet fault information obtained includes:

[0100] Arrange and identify the simulation result data and the independent variable data in the operation data of each area in time sequence and correspond them;

[0101] Extract key parameters from simulation result data and regional operation data;

[0102] Compare the key parameters of the independent variable data corresponding to the dependent variable data in the regional operation data with the key parameters of the corresponding simulation result data;

[0103] Determine the faulty node, analyze the key parameter characteristics of other areas based on the timing characteristics of the faulty node, determine the fault type corresponding to the faulty node, and obtain fault information.

[0104] Specifically, the fault node can be a single-region fault. By determining the fault nodes of key parameters in other regions of the time series, the changes in key parameters in other regions after the single-region fault can be determined, that is, the correlation of regional faults can be determined.

[0105] It can be understood that the independent variable data of the switchgear include temperature data (the temperature of key parts inside the switchgear such as busbars, circuit breakers, cable joints, etc.), current data (the current value passing through the switchgear, reflecting the load conditions), voltage data (input and output voltages of the switchgear), vibration data (vibration signals of mechanical parts inside the switchgear such as circuit breakers and disconnectors), and the dependent variable data include ambient temperature (the temperature of the environment where the switchgear is located, which affects the internal temperature), ambient humidity (the humidity of the environment where the switchgear is located, which affects the internal humidity), load changes (load changes in the power system affect the current and voltage of the switchgear), and external vibrations (such as vibrations caused by nearby equipment or traffic, which may affect the mechanical parts of the switchgear).

[0106] In this embodiment, the switch cabinet is divided into different areas according to function, and different areas are further analyzed according to different working conditions, so as to achieve refined control of the switch cabinet. At the same time, by constructing a simulation model of the switch cabinet, the working parameter performance of the switch cabinet under different working conditions can be accurately represented, so that in-depth analysis of different areas can be performed to improve the accuracy of fault diagnosis. Based on the analysis of regional faults, the correlation of key data of each area can be further analyzed, which can further improve the accuracy of switch cabinet fault analysis.

[0107] Among them, in order to realize the diagnosis of fault points in different areas and different working conditions of the switch cabinet, the operating data of the switch cabinet at the preset periodic time is first collected, the switch cabinet is classified by area, and further refined classification is performed based on the operating conditions. The fault point can be independent or caused by the mutual influence of each area. By simulating the switch cabinet, the operating conditions of different areas of the switch cabinet are simulated based on the independent variable data, and the independent faults in different areas of the switch cabinet are reflected. The operating data of other areas under the time series is used to determine the impact of independent faults on other areas. The impact of a certain fault on different areas can be screened out, that is, the important representation of the fault point can be diagnosed. Conversely, the specific cause of the fault point can be diagnosed through these important representations, and the accurate diagnosis and identification of the specific location of the fault point of the switch cabinet can be realized.

[0108] In other embodiments, by identifying the operating data immediately before the fault occurs and analyzing the switch cabinet operating data, the location of the switch cabinet fault can be predicted in advance, so that maintenance can be performed in advance.

[0109] S4: Build a data set based on the fault information, build a neural network model, and train the deep learning model through the fault information to obtain a fault diagnosis model;

[0110] In this embodiment, the process of constructing a data set based on fault information, constructing a deep learning model, and training the deep learning model through the fault information includes:

[0111] S41: extracting feature data from the fault data to obtain a feature data set;

[0112] S42: Divide the feature data set into a training set and a validation set;

[0113] S43: Build a deep learning model and input the training set into the deep learning model for training;

[0114] S44: The validation set verifies the trained model. According to the number distribution of the regions in the switch cabinet that are judged to be faulty in the validation records, all validation records of each region are classified respectively, and the region combination composed of different regions that can affect the fault diagnosis of the model is extracted, and the fault diagnosis influence index of the region combination on the deep learning model is evaluated;

[0115] S45: Optimizing the deep learning model based on the distribution of regions in the regional combination whose fault diagnosis impact index is greater than a preset threshold.

[0116] The steps of S44 specifically include:

[0117] Obtain verification records in which only a certain area is a fault area and determine it as fault a;

[0118] Obtain a region combination that is determined to be fault a at least once in the verification record, the set of region combinations is M, the total number of regions in the region combination is I, and determine the set m that affects fault a in the set M;

[0119] Obtain all area combinations that are judged as fault a in the verification record and whose set contains M. The total number of areas in all area combinations is J;

[0120] Calculate the characteristic influencing factors of the fault area of ​​fault a on the fault diagnosis process of the deep learning model:

[0121] ;

[0122] Among them, γ represents the correction coefficient;

[0123] Evaluation of the impact index of regional combinations on the fault diagnosis of deep learning models:

[0124] ;

[0125] in, They respectively represent the 1st, 2nd, 3rd, ..., ith areas in the area combination of fault a, and i>2.

[0126] It should be noted that before extracting characteristic data from fault data, the fault data is also processed, and the processing process includes data cleaning and noise elimination. The specific process is a conventional technical means and will not be described in detail here.

[0127] S5: Input the operating data of the switch cabinet into the fault diagnosis model to obtain the fault diagnosis result of the target switch cabinet.

[0128] The present invention divides the switch cabinet into areas, and further divides the areas of the switch cabinet according to different working conditions. The simulation model is established to perform simulation models for different working conditions in different areas of the switch cabinet. The simulation results are compared with the actual operation data, and abnormal data is obtained through analysis to ensure the authenticity and accuracy of the data. The deep learning model is trained and verified through the obtained abnormal data. In the verification records, the number distribution of areas in the switch cabinet that are judged to be faulty is obtained, all verification records of each area are classified respectively, and the regional combination composed of different areas that can affect the fault diagnosis of the model is extracted. The fault diagnosis influence index of the regional combination on the deep learning model is evaluated, and the influence coefficient of abnormal information in different areas on the fault is reversed based on the impact of the fault on different areas, thereby further ensuring the accuracy of the switch cabinet fault judgment.

[0129] like Figure 2 As shown, this embodiment also provides a high-voltage switch cabinet fault diagnosis system based on deep learning, which is used to implement the above-mentioned high-voltage switch cabinet fault diagnosis method based on deep learning, including:

[0130] The area division module divides the switch cabinet into several areas based on the operating functions, and collects the operating data of several areas based on the partition results;

[0131] Among them, the operation data includes independent variable data and dependent variable data;

[0132] The regional division module is also used to divide the operation data of each area based on the operation conditions of each area;

[0133] The model building module builds a switch cabinet simulation model, inputs the operation data of each area of ​​the switch cabinet into the simulation model for simulation training, obtains simulation result data, and obtains switch cabinet fault information by comparing the simulation result data with the operation data; it is also used to build a data set based on the fault information, build a neural network model, and train the deep learning model through the fault information to obtain a fault diagnosis model;

[0134] The fault diagnosis module is used to input the operating data of the switch cabinet into the fault diagnosis model to obtain the fault diagnosis result of the target switch cabinet.

[0135] Furthermore, the method for establishing the switch cabinet simulation model includes:

[0136] Establish a physical model of the switch cabinet to ensure that the switch cabinet model parameters are consistent with the switch cabinet reagent properties;

[0137] Set the switch cabinet model environmental parameters according to the switch cabinet working environment;

[0138] After setting all parameters, simulation calculations are performed to calculate the operating parameters and operating status of the switch cabinet. After the calculations are completed, the simulation results are analyzed and evaluated.

[0139] Furthermore, by comparing the simulation result data with the operation data, the switch cabinet fault information including:

[0140] Arrange and identify the simulation result data and the independent variable data in the operation data of each area in time sequence and correspond them;

[0141] Extract key parameters from simulation result data and regional operation data;

[0142] Compare the key parameters of the independent variable data corresponding to the dependent variable data in the regional operation data with the key parameters of the corresponding simulation result data;

[0143] Determine the faulty node, analyze the key parameter characteristics of other areas based on the timing characteristics of the faulty node, determine the fault type corresponding to the faulty node, and obtain fault information.

[0144] Furthermore, the process of constructing a data set based on the fault information, constructing a deep learning model, and training the deep learning model through the fault information includes:

[0145] Extract characteristic data from the fault data to obtain a characteristic data set;

[0146] Divide the feature dataset into training set and validation set;

[0147] Build a deep learning model and input the training set into the deep learning model for training;

[0148] The validation set verifies the trained model. According to the number distribution of regions in the switch cabinet that are judged to be faulty in the validation records, all validation records of each region are classified, and the regional combination composed of different regions that can affect the fault diagnosis of the model is extracted. The impact index of the regional combination on the fault diagnosis of the deep learning model is evaluated.

[0149] Based on the distribution of regions in the regional combination whose fault diagnosis impact index is greater than the preset threshold, the deep learning model is optimized.

[0150] Furthermore, the steps of evaluating the influence index of the regional combination on the fault diagnosis of the deep learning model specifically include:

[0151] Obtain verification records in which only a certain area is a fault area and determine it as fault a;

[0152] Obtain a region combination that is determined to be fault a at least once in the verification record, the set of region combinations is M, the total number of regions in the region combination is I, and determine the set m that affects fault a in the set M;

[0153] Obtain all area combinations that are judged as fault a in the verification record and whose set contains M. The total number of areas in all area combinations is J;

[0154] Calculate the characteristic influencing factors of the fault area of ​​fault a on the fault diagnosis process of the deep learning model:

[0155] ;

[0156] Among them, γ represents the correction coefficient;

[0157] Evaluation of the impact index of regional combinations on the fault diagnosis of deep learning models:

[0158] ;

[0159] in, Respectively represent the 1st, 2nd, 3rd, ..., ith regions in the region combination of fault a, and i>2

[0160] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A high-voltage switchgear fault diagnosis method based on deep learning, characterized in that: The following steps are involved: The switch cabinet is divided into several areas based on the operating functions, and the operating data of several areas are collected based on the partition results; Among them, the operation data includes independent variable data and dependent variable data; The operating data of each area is divided based on the public operation conditions of each area; Build a switch cabinet simulation model, input the operation data of each area of ​​the switch cabinet into the simulation model for simulation training, obtain simulation result data, and obtain switch cabinet fault information by comparing the simulation result data with the operation data; Building a data set based on the fault information, building a neural network model, and training the deep learning model through the fault information to obtain a fault diagnosis model; The operating data of the switchgear is input into the fault diagnosis model to obtain the fault diagnosis results of the target switchgear.

2. A high-voltage switchgear fault diagnosis method based on deep learning according to claim 1, characterized in that: The method for establishing the switch cabinet simulation model includes: Establish a physical model of the switch cabinet to ensure that the switch cabinet model parameters are consistent with the switch cabinet reagent properties; Set the switch cabinet model environmental parameters according to the switch cabinet working environment; After setting all parameters, simulation calculations are performed to calculate the operating parameters and operating status of the switch cabinet. After the calculations are completed, the simulation results are analyzed and evaluated.

3. A high-voltage switchgear fault diagnosis method based on deep learning according to claim 2, characterized in that: The comparison of simulation result data and operation data to obtain switch cabinet fault information includes: Arrange and identify the simulation result data and the independent variable data in the operation data of each area in time sequence and correspond them; Extract key parameters from simulation result data and regional operation data; Compare the key parameters of the independent variable data corresponding to the dependent variable data in the regional operation data with the key parameters of the corresponding simulation result data; Determine the faulty node, analyze the key parameter characteristics of other areas based on the timing characteristics of the faulty node, determine the fault type corresponding to the faulty node, and obtain fault information.

4. A high-voltage switchgear fault diagnosis method based on deep learning according to claim 1, characterized in that: The process of constructing a data set based on the fault information, constructing a deep learning model, and training the deep learning model through the fault information includes: Extract characteristic data from the fault data to obtain a characteristic data set; Divide the feature dataset into training set and validation set; Build a deep learning model and input the training set into the deep learning model for training; The validation set verifies the trained model. According to the number distribution of the switch cabinet areas that are judged to be faulty in the validation records, all validation records of each area are classified respectively, and the regional combination composed of different areas that can affect the fault diagnosis of the model is extracted, and the fault diagnosis influence index of the regional combination on the deep learning model is evaluated; Based on the distribution of regions in the regional combination whose fault diagnosis impact index is greater than the preset threshold, the deep learning model is optimized.

5. A high-voltage switchgear fault diagnosis method based on deep learning according to claim 4, characterized in that: The step of evaluating the impact index of the regional combination on the fault diagnosis of the deep learning model specifically includes: Obtain verification records in which only a certain area is a fault area and determine it as fault a; Obtain a region combination that is determined to be fault a at least once in the verification record, where the set of region combinations is M, the total number of regions in the region combination is I, and determine a set m in the set M that affects fault a; Obtain all area combinations that are judged as fault a in the verification record and whose set contains M. The total number of areas in all area combinations is J; Calculate the characteristic influencing factors of the fault area of ​​fault a on the fault diagnosis process of the deep learning model: ； Among them, γ represents the correction coefficient; Evaluation of the impact index of regional combinations on the fault diagnosis of deep learning models: ； in, They respectively represent the 1st, 2nd, 3rd, ..., ith areas in the area combination of fault a, and i>

2.

6. A high-voltage switchgear fault diagnosis system based on deep learning, used to implement the high-voltage switchgear fault diagnosis method based on deep learning according to any one of claims 1 to 5, characterized in that: include: The area division module divides the switch cabinet into several areas based on the operating functions, and collects the operating data of several areas based on the partition results; Among them, the operation data includes independent variable data and dependent variable data; The area division module is also used to divide the operation data of each area based on the operation conditions of each area; A model building module is used to build a switch cabinet simulation model, input the operation data of each area of ​​the switch cabinet into the simulation model for simulation training, obtain simulation result data, and obtain switch cabinet fault information by comparing the simulation result data with the operation data; it is also used to build a data set based on the fault information, build a neural network model, and train the deep learning model through the fault information to obtain a fault diagnosis model; The fault diagnosis module is used to input the operating data of the switch cabinet into the fault diagnosis model to obtain the fault diagnosis result of the target switch cabinet.

7. A high-voltage switchgear fault diagnosis system based on deep learning according to claim 6, characterized in that: The method for establishing the switch cabinet simulation model includes: Establish a physical model of the switch cabinet to ensure that the switch cabinet model parameters are consistent with the switch cabinet reagent properties; Set the switch cabinet model environmental parameters according to the switch cabinet working environment; After setting all parameters, simulation calculations are performed to calculate the operating parameters and operating status of the switch cabinet. After the calculations are completed, the simulation results are analyzed and evaluated.

8. A high-voltage switchgear fault diagnosis system based on deep learning according to claim 7, characterized in that: The comparison of simulation result data and operation data to obtain switch cabinet fault information includes: Arrange and identify the simulation result data and the independent variable data in the operation data of each area in time sequence and correspond them; Extract key parameters from simulation result data and regional operation data; Compare the key parameters of the independent variable data corresponding to the dependent variable data in the regional operation data with the key parameters of the corresponding simulation result data; Determine the faulty node, analyze the key parameter characteristics of other areas based on the timing characteristics of the faulty node, determine the fault type corresponding to the faulty node, and obtain fault information.

9. A high-voltage switchgear fault diagnosis system based on deep learning according to claim 6, characterized in that: The process of constructing a data set based on the fault information, constructing a deep learning model, and training the deep learning model through the fault information includes: Extract characteristic data from the fault data to obtain a characteristic data set; Divide the feature dataset into training set and validation set; Build a deep learning model and input the training set into the deep learning model for training; The validation set verifies the trained model. According to the number distribution of the switch cabinet areas that are judged to be faulty in the validation records, all validation records of each area are classified respectively, and the regional combination composed of different areas that can affect the fault diagnosis of the model is extracted, and the fault diagnosis influence index of the regional combination on the deep learning model is evaluated; Based on the distribution of regions in the regional combination whose fault diagnosis impact index is greater than the preset threshold, the deep learning model is optimized.

10. A high-voltage switchgear fault diagnosis system based on deep learning according to claim 9, characterized in that: The step of evaluating the impact index of the regional combination on the fault diagnosis of the deep learning model specifically includes: Obtain verification records in which only a certain area is a fault area and determine it as fault a; Obtain a region combination that is determined to be fault a at least once in the verification record, where the set of region combinations is M, the total number of regions in the region combination is I, and determine a set m in the set M that affects fault a; Obtain all area combinations that are judged as fault a in the verification record and whose set contains M. The total number of areas in all area combinations is J; Calculate the characteristic influencing factors of the fault area of ​​fault a on the fault diagnosis process of the deep learning model: ； Among them, γ represents the correction coefficient; Evaluation of the impact index of regional combinations on the fault diagnosis of deep learning models: ； in, They respectively represent the 1st, 2nd, 3rd, ..., ith areas in the area combination of fault a, and i>2.

Images