High-voltage power distribution cabinet AI fault diagnosis method, system and device

Abstract

The invention discloses an AI fault diagnosis method, system and device for a high-voltage power distribution cabinet, and belongs to the technical field of power distribution cabinet diagnosis. A reconstruction error and a prediction error are calculated through a health reference model to obtain a first health deviation degree of each electrical parameter, a first abnormal parameter is selected based on the first health deviation degree, all abnormal parameter combinations are obtained and matched in a library to obtain a first fault type, and one-time diagnosis is completed; and the second fault type is obtained through the diagnosis model to complete secondary diagnosis, and the first fault type and the second fault type are subjected to cross validation to obtain the final fault type of the target branch, so that the accuracy of fault diagnosis is effectively improved.

Description

Technical Field

[0001] This invention relates to the field of power distribution cabinet diagnostic technology, and in particular to an AI fault diagnosis method, system and device for high-voltage power distribution cabinets. Background Technology

[0002] High-voltage switchgear is a key piece of equipment in the power system used to receive and distribute electrical energy, and to control, protect and monitor circuits. It is usually installed at power users such as substations, industrial and mining enterprises, and high-rise buildings. It is responsible for distributing the electrical energy of the high-voltage grid to the downstream electrical equipment. Its importance is self-evident, so fault diagnosis of it is particularly important. However, most existing power distribution cabinet fault diagnosis technologies rely solely on learning models to assess and diagnose fault types based on historical data, resulting in insufficient diagnostic accuracy. Summary of the Invention

[0003] To address the technical problems existing in the prior art, this invention provides an AI fault diagnosis method for high-voltage switchgear, comprising the following steps: S1. Based on the importance of the target branch in the distribution cabinet and combined with the historical fault information of the target branch, adjust the monitoring strategy for each electrical parameter of the target branch. S2. According to the corresponding monitoring strategy, obtain the time sequence vector of each electrical parameter of the target branch within a preset time period. Calculate the reconstruction error and prediction error based on the time sequence vector of the target electrical parameters using a pre-trained health benchmark model, and then calculate the first health deviation of the target electrical parameters. The acquisition of the reconstruction error and prediction error is as follows: the health benchmark model outputs the corresponding reconstruction vector through an autoencoder based on the time sequence vector of the target electrical parameter, and the root mean square error between the two is used as the reconstruction error; the health benchmark model obtains the predicted value of the target electrical parameter at the end of the time sequence vector through a gated recurrent neural network based on the time sequence vector of the target electrical parameter, and the relative error between the predicted value and the corresponding actual value is used as the prediction error. The first health deviation is obtained as follows: P1 = α × CW + β × YW; P1 is the first health deviation, CW and YW are the reconstruction error and prediction error respectively, and α and β are the preset weight coefficients respectively; S3. Take the electrical parameters whose first health deviation is greater than the corresponding preset deviation threshold as the first abnormal parameters, and obtain all abnormal parameter combinations based on all the first abnormal parameters of the target branch. Based on the abnormal parameter combinations, match and retrieve all fault types in the preset fault information database and take them as the first fault types. S4. Using a pre-trained diagnostic model, the probability distribution of each fault type is obtained based on the time sequence vector of each electrical parameter of the target branch, thus obtaining the second fault type. S5. Cross-validate the first fault type and the second fault type to obtain the final fault type of the target branch.

[0004] Furthermore, the historical fault information includes the type of each historical fault, the corresponding historical fault impact index, and the historical major abnormal electrical parameters.

[0005] Furthermore, the strategy for adjusting the monitoring of each electrical parameter of the target branch specifically includes: The initial monitoring frequency of each electrical parameter of the target branch is obtained based on the importance of the branch; Electrical parameters that were previously considered major historical anomalies in the target branch are designated as key electrical parameters. Based on the historical fault impact index corresponding to each historical fault for the key electrical parameters, the key attention index for each key electrical parameter is obtained. The initial monitoring frequency of each key electrical parameter of the target branch is adjusted by frequency boosting to obtain the first monitoring frequency of each key electrical parameter of the target branch. The degree of frequency boosting is determined based on the key attention index.

[0006] Furthermore, the key attention index for each key electrical parameter is obtained based on the historical fault impact index corresponding to each historical fault for the key electrical parameter, specifically as follows: ; Where i is the i-th key electrical parameter of the target branch, GZ is the key attention index, C is the number of times the current key electrical parameter has been used as a historical major abnormal electrical parameter, YZ is the corresponding historical fault impact index, n is the index, and K1 and K2 are the first preset weight and the second preset weight, respectively.

[0007] Furthermore, the diagnostic model is a deep learning model, which is trained using multiple sets of different historical fault types of the target branch and the historical time sequence vectors of the corresponding electrical parameters as training samples. After training, the model takes the time sequence vectors of the electrical parameters of the target branch as input and outputs the probability distribution of each fault type. Fault types with a probability greater than or equal to a preset probability threshold are designated as the second fault type.

[0008] Furthermore, the following steps are included after step S5: S6. Based on the preset fault information database, obtain the electrical parameters affected by each final fault type, and determine the number of final fault types matched by each of the first abnormal parameters. S7. Calculate the second key focus index for each of the first abnormal parameters based on the first health deviation degree of each of the first abnormal parameters and the number of the final fault types matched. S8. Based on the current monitoring frequency, perform a second frequency boosting adjustment operation on the electrical parameters corresponding to the first abnormal parameters of the target branch to obtain the second monitoring frequency of the electrical parameters corresponding to the first abnormal parameters of the target branch. The degree of boosting of the second frequency boosting adjustment operation is determined according to the second key attention index.

[0009] Furthermore, the calculation of the second key focus index is as follows: ; Where j is the j-th first abnormal parameter of the target branch, GZ2 is the second key focus index, GS is the number of the final fault types matched, and K3 and K4 are the third and fourth preset weights, respectively.

[0010] The present invention also provides an AI fault diagnosis system for high-voltage switchgear, which applies the AI ​​fault diagnosis method for high-voltage switchgear as described in any of the above claims, including: The monitoring and adjustment module is used to adjust the monitoring strategy for each electrical parameter of the target branch based on the importance of the target branch in the distribution cabinet and in combination with the historical fault information of the target branch. The first analysis module is used to obtain the time sequence vector of each electrical parameter of the target branch within a preset time period according to the corresponding monitoring strategy, and calculate the reconstruction error and prediction error based on the time sequence vector of the target electrical parameters through a pre-trained health benchmark model, and then calculate the first health deviation of the target electrical parameters. The first diagnostic module is used to take electrical parameters whose first health deviation is greater than the corresponding preset deviation threshold as first abnormal parameters, and to obtain all abnormal parameter combinations based on all first abnormal parameters of the target branch, and to match and retrieve all fault types in the preset fault information database based on the abnormal parameter combinations as the first fault type. The second diagnostic module is used to obtain the probability distribution of each fault type based on the time sequence vector of each electrical parameter of the target branch through a pre-trained diagnostic model, thereby obtaining the second fault type. The final diagnosis module is used to cross-validate the first fault type and the second fault type to obtain the final fault type of the target branch.

[0011] Furthermore, the system also includes: The second analysis module is used to obtain the electrical parameters affected by each final fault type according to the preset fault information database, and to determine the number of final fault types matched by each first abnormal parameter. Based on the first health deviation degree of each first abnormal parameter and the number of final fault types matched, the second key attention index of each first abnormal parameter is calculated. The monitoring and adjustment module is further configured to perform a second frequency boosting adjustment operation on the electrical parameters corresponding to each first abnormal parameter of the target branch based on the current corresponding monitoring frequency, thereby obtaining the second monitoring frequency of the electrical parameters corresponding to each first abnormal parameter of the target branch. The degree of boosting of the second frequency boosting adjustment operation is determined according to the second key focus index.

[0012] The present invention also provides an AI fault diagnosis device for high-voltage switchgear, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the method described in any of the above.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention obtains the time-series vector of electrical parameters within a preset time period by first adjusting the monitoring strategy of each electrical parameter of the target branch, and then calculates the reconstruction error and prediction error through a health benchmark model to obtain the first health deviation of each electrical parameter. Based on the first health deviation, the first abnormal parameter is selected, and all abnormal parameter combinations are obtained and matched with a database to obtain the first fault type to complete the first diagnosis. Then, the second fault type is obtained through the diagnostic model to complete the second diagnosis. The first fault type and the second fault type are cross-validated to obtain the final fault type of the target branch, which effectively improves the accuracy of fault diagnosis. Attached Figure Description

[0014] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart of an AI fault diagnosis method for high-voltage distribution cabinets according to the present invention; Figure 2 This is a structural block diagram of an AI fault diagnosis system for high-voltage distribution cabinets according to the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0018] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indication will also change accordingly.

[0019] Furthermore, the use of terms such as "first" and "second" in this invention is for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined with "first" and "second" may explicitly or implicitly include at least one of those features. Additionally, the technical solutions of the various embodiments can be combined with each other, but only on the basis of being achievable by those skilled in the art. When the combination of technical solutions is contradictory or impossible to implement, such a combination of technical solutions should be considered non-existent and not within the scope of protection claimed by this invention.

[0020] Example 1 See Figure 1 As shown, the present invention provides an AI fault diagnosis method for high-voltage distribution cabinets, which specifically includes the following steps: S1. Based on the importance of the target branch in the distribution cabinet and combined with the historical fault information of the target branch, adjust the monitoring strategy for each electrical parameter of the target branch. S2. According to the corresponding monitoring strategy, obtain the time sequence vector of each electrical parameter of the target branch within a preset time period. Calculate the reconstruction error and prediction error based on the time sequence vector of the target electrical parameters using a pre-trained health benchmark model. Then calculate the first health deviation of the target electrical parameters. In this way, obtain the first health deviation of each electrical parameter of the target branch. S3. Take the electrical parameters whose first health deviation is greater than the corresponding preset deviation threshold as the first abnormal parameters, and obtain all abnormal parameter combinations based on all the first abnormal parameters of the target branch. Based on the abnormal parameter combinations, match and retrieve all fault types in the preset fault information database and take them as the first fault types. S4. Using a pre-trained diagnostic model, the probability distribution of each fault type is obtained based on the time sequence vector of each electrical parameter of the target branch, thereby obtaining the second fault type. S5. Cross-validate the first fault type and the second fault type to obtain the final fault type of the target branch.

[0021] The following is a detailed explanation of each step: S1. Based on the importance of the target branch in the distribution cabinet and combined with the historical fault information of the target branch, adjust the monitoring strategy for each electrical parameter of the target branch: The importance of a branch reflects its criticality in the entire distribution cabinet power supply system. Branches with higher importance usually have a more frequent monitoring frequency, while branches with relatively lower importance can have a lower monitoring frequency.

[0022] The historical fault information includes the type of each historical fault, the corresponding historical fault impact index, and the main historical abnormal electrical parameters. The adjustment strategy for monitoring the electrical parameters of the target branch is as follows: S11. Obtain the initial monitoring frequency of each electrical parameter of the target branch according to the importance of the branch; S12. Record the electrical parameters in the target branch that were previously considered as major historical abnormal electrical parameters as key electrical parameters, and obtain the key attention index for each key electrical parameter based on the historical fault impact index corresponding to each historical fault. S13. The initial monitoring frequency of each key electrical parameter of the target branch is adjusted by frequency boosting to obtain the first monitoring frequency of each key electrical parameter of the target branch. The degree of frequency boosting is determined according to the key attention index.

[0023] The higher the index, the greater the corresponding increase in frequency.

[0024] The historical fault impact index is determined manually based on preset rules, taking into account the scope of impact and losses caused by corresponding historical faults.

[0025] The key attention index for each key electrical parameter is obtained based on the historical fault impact index corresponding to each historical fault for the key electrical parameters. Specifically: ;

[0026] Where i is the i-th key electrical parameter of the target branch, GZ is the key attention index, C is the number of times the current key electrical parameter has been used as a historical major abnormal electrical parameter, YZ is the corresponding historical fault impact index, n is the index, and K1 and K2 are the first preset weight and the second preset weight, respectively.

[0027] S2. According to the corresponding monitoring strategy, obtain the time series vector of each electrical parameter of the target branch within a preset time period. Calculate the reconstruction error and prediction error based on the time series vector of the target electrical parameters using a pre-trained health benchmark model. Then calculate the first health deviation of the target electrical parameters. This method is used to obtain the first health deviation of each electrical parameter of the target branch. The health benchmark model is a deep neural network model. It is trained unsupervised by using historical data of the corresponding electrical parameters of the corresponding branch under normal operating conditions of the distribution cabinet as sample data. The model outputs a reconstruction vector through the autoencoder and the predicted value of the next time step through the gated recurrent unit network. The total loss is calculated based on the reconstruction vector and the predicted value. The model parameters of the autoencoder and the gated recurrent unit network are adjusted according to the total loss through the backpropagation algorithm. The training is iterated until the total loss converges and stabilizes.

[0028] According to the training method described above, the autoencoder of the health benchmark model can compress and decode the real time sequence vector of electrical parameters to obtain a new vector, namely the reconstructed vector. This reconstructed vector is a feature data that theoretically "should exist" and is reconstructed by the model based on its understanding of "normal health state". It is the approximate parameter value of the target electrical parameter in a fault-free state, which serves as the health reference value of the target electrical parameter.

[0029] The pre-trained health benchmark model calculates the reconstruction error and prediction error based on the time-series vector of the target electrical parameters, specifically as follows: The health baseline model outputs a corresponding reconstruction vector through an autoencoder based on the time sequence vector of the target electrical parameters, and uses the root mean square error between the two as the reconstruction error. The health baseline model obtains the predicted value of the target electrical parameters at the end of the time sequence vector through a gated recurrent neural network based on the time sequence vector of the target electrical parameters, and the relative error between the predicted value and the corresponding actual value is used as the prediction error.

[0030] The first health deviation of the target electrical parameter is specifically calculated as: P1 = α × CW + β × YW; Wherein, P1 is the first health deviation, CW and YW are the reconstruction error and prediction error, respectively, and α and β are the preset weight coefficients.

[0031] S3. Electrical parameters with a first health deviation greater than the corresponding preset deviation threshold are taken as first abnormal parameters. All abnormal parameter combinations are obtained based on all first abnormal parameters of the target branch. All fault types are retrieved from the preset fault information database based on the abnormal parameter combinations and taken as the first fault type. The preset fault information database is constructed based on all historical fault information of the power distribution cabinet.

[0032] For example, the first abnormal parameter of the target branch includes: abnormal parameter 1, abnormal parameter 2, and abnormal parameter 3; its abnormal parameter combination includes: abnormal parameter 1, abnormal parameter 2, abnormal parameter 3, abnormal parameter 1 + abnormal parameter 2, abnormal parameter 1 + abnormal parameter 3, abnormal parameter 2 + abnormal parameter 3, and abnormal parameter 1 + abnormal parameter 2 + abnormal parameter 3. If the main abnormal parameter of a certain fault type A in the preset fault information database is abnormal parameter 1 + abnormal parameter 2, then it can match fault type A. If the main abnormal parameter of each fault type A in the preset fault information database does not include the above abnormal parameter combination, then it does not match fault type B.

[0033] S4. Using a pre-trained diagnostic model, the probability distribution of each fault type is obtained based on the time-series sequence vector of each electrical parameter of the target branch, thereby obtaining the second fault type: The diagnostic model is a deep learning model. It is trained using multiple sets of different historical fault types of the target branch and the historical time sequence vectors of the corresponding electrical parameters as training samples. After training, the model takes the time sequence vectors of the electrical parameters of the target branch as input and outputs the probability distribution of each fault type.

[0034] Fault types with a probability greater than or equal to a preset probability threshold are designated as the second fault type.

[0035] S5. Cross-validate the first fault type and the second fault type to obtain the final fault type of the target branch: The fault type at the intersection of the first fault type and the second fault type is taken as the final fault type of the target branch.

[0036] In some embodiments, the method further includes the following steps: S6. Based on the preset fault information database, obtain the electrical parameters affected by each final fault type, and determine the number of final fault types matched by each of the first abnormal parameters. For example, the final fault types obtained this time are type A, B, and C; the electrical parameters affected by A include parameters a and b, the electrical parameters affected by B include parameters a and c, and the electrical parameters affected by C include parameters b and d; and the first abnormal parameters in this case are a and d. Therefore, the final fault types matched by the first abnormal parameter a are two types, A and B, and the final fault type matched by the first abnormal parameter d is one type, C.

[0037] S7. Based on the first health deviation of each of the first abnormal parameters and the number of matched final fault types, calculate the second key concern index for each of the first abnormal parameters: ; Where j is the j-th first abnormal parameter of the target branch, GZ2 is the second key focus index, GS is the number of the final fault types matched, and K3 and K4 are the third and fourth preset weights, respectively.

[0038] S8. Based on the current monitoring frequency, perform a second frequency boosting adjustment operation on the electrical parameters corresponding to the first abnormal parameters of the target branch to obtain the second monitoring frequency of the electrical parameters corresponding to the first abnormal parameters of the target branch. The degree of boosting of the second frequency boosting adjustment operation is determined according to the second key attention index.

[0039] The second key focus is on the larger the index, the greater the degree of improvement in the corresponding second frequency adjustment operation.

[0040] If the current monitoring frequency corresponding to an electrical parameter is greater than or equal to the preset monitoring frequency threshold, the monitoring frequency of the electrical parameter will be adjusted to the preset monitoring frequency threshold, and no further frequency increase adjustment will be made.

[0041] Example 2 See Figure 2 As shown, the present invention also provides an AI fault diagnosis system for high-voltage switchgear, specifically comprising: The monitoring and adjustment module is used to adjust the monitoring strategy for each electrical parameter of the target branch based on the importance of the target branch in the distribution cabinet and in combination with the historical fault information of the target branch. The first analysis module is used to obtain the time series vector of each electrical parameter of the target branch within a preset time period according to the corresponding monitoring strategy. The reconstruction error and prediction error are calculated based on the time series vector of the target electrical parameters through a pre-trained health benchmark model. Then, the first health deviation of the target electrical parameters is calculated. The first health deviation of each electrical parameter of the target branch is obtained in this way. The first diagnostic module is used to take electrical parameters whose first health deviation is greater than the corresponding preset deviation threshold as first abnormal parameters, and to obtain all abnormal parameter combinations based on all first abnormal parameters of the target branch, and to match and retrieve all fault types in the preset fault information database based on the abnormal parameter combinations as the first fault type. The second diagnostic module is used to obtain the probability distribution of each fault type based on the time sequence vector of each electrical parameter of the target branch through a pre-trained diagnostic model, thereby obtaining the second fault type. The final diagnosis module is used to cross-validate the first fault type and the second fault type to obtain the final fault type of the target branch.

[0042] The system also includes: The second analysis module is used to obtain the electrical parameters affected by each final fault type according to the preset fault information database, and to determine the number of final fault types matched by each first abnormal parameter. Based on the first health deviation degree of each first abnormal parameter and the number of final fault types matched, the second key attention index of each first abnormal parameter is calculated. The monitoring and adjustment module is further configured to perform a second frequency boosting adjustment operation on the electrical parameters corresponding to each first abnormal parameter of the target branch based on the current corresponding monitoring frequency, thereby obtaining the second monitoring frequency of the electrical parameters corresponding to each first abnormal parameter of the target branch. The degree of boosting of the second frequency boosting adjustment operation is determined according to the second key focus index.

[0043] Example 3 The present invention also provides an AI fault diagnosis device for high-voltage distribution cabinets, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of the method as described in Embodiment 1.

[0044] Example 4 The present invention also provides an electronic device, including: a processor, a transmitting device, an input device, an output device, and a memory. The processor may be implemented using a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit, or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory may be implemented using a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM), and is used to store computer program code. The computer program code includes computer instructions. When the processor executes the computer instructions, the electronic device executes a method as described in any of the above possible implementation methods.

[0045] Example 5 The present invention also provides a computer-readable storage medium storing a computer program, the computer program including program instructions, which, when executed by a processor of an electronic device, cause the processor to perform a method as described in any of the above possible implementations.

[0046] The beneficial effects of this invention are as follows: This invention obtains the time-series vector of electrical parameters within a preset time period by first adjusting the monitoring strategy of each electrical parameter of the target branch, then calculates the reconstruction error and prediction error through a health benchmark model to obtain the first health deviation of each electrical parameter, selects the first abnormal parameter based on the first health deviation, obtains all abnormal parameter combinations and matches them with a database to obtain the first fault type to complete the first diagnosis, then obtains the second fault type through a diagnostic model to complete the second diagnosis, and cross-validates the first fault type and the second fault type to obtain the final fault type of the target branch, effectively improving the accuracy of fault diagnosis; By determining the initial monitoring frequency for each electrical parameter of the target branch based on the importance of the branch, the monitoring cost is reduced while ensuring effective monitoring. Furthermore, the monitoring frequency of the corresponding electrical parameters is adjusted according to the key attention index, thereby achieving enhanced monitoring of high-risk parameters. By weighted fusion of reconstruction error and prediction error, the first health deviation index is obtained, which is used to screen out the first abnormal parameter, thereby improving the effectiveness and accuracy of identifying abnormal electrical parameters and thus improving the accuracy of fault type diagnosis. Finally, the fault type is obtained by cross-validation with the fault type diagnosed by the deep learning model based on historical data, which further improves the accuracy of fault type diagnosis. Based on the final fault type analysis, a second key focus index is determined for the first abnormal parameter. This allows for further adjustments to the monitoring frequency corresponding to the first abnormal parameter, thereby enhancing the monitoring of parameters with higher risks.

[0047] In the description of this specification, the references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0048] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0049] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for AI fault diagnosis of high-voltage switchgear, characterized in that, Includes the following steps: S1. Based on the importance of the target branch in the distribution cabinet and combined with the historical fault information of the target branch, adjust the monitoring strategy for each electrical parameter of the target branch. S2. According to the corresponding monitoring strategy, obtain the time sequence vector of each electrical parameter of the target branch within a preset time period. Calculate the reconstruction error and prediction error based on the time sequence vector of the target electrical parameters using a pre-trained health benchmark model, and then calculate the first health deviation of the target electrical parameters. The acquisition of the reconstruction error and prediction error is as follows: the health benchmark model outputs the corresponding reconstruction vector through an autoencoder based on the time sequence vector of the target electrical parameters, and the root mean square error between the two is used as the reconstruction error. The health benchmark model obtains the predicted value of the target electrical parameter at the end of the time sequence vector through a gated recurrent neural network based on the time sequence vector of the target electrical parameter. The relative error between the predicted value and the corresponding actual value is used as the prediction error. The first health deviation is obtained as follows: P1 = α × CW + β × YW; P1 is the first health deviation, CW and YW are the reconstruction error and prediction error respectively, and α and β are the preset weight coefficients respectively; S3. Take the electrical parameters whose first health deviation is greater than the corresponding preset deviation threshold as the first abnormal parameters, and obtain all abnormal parameter combinations based on all the first abnormal parameters of the target branch. Based on the abnormal parameter combinations, match and retrieve all fault types in the preset fault information database and take them as the first fault types. S4. Using a pre-trained diagnostic model, the probability distribution of each fault type is obtained based on the time sequence vector of each electrical parameter of the target branch, thus obtaining the second fault type. S5. Cross-validate the first fault type and the second fault type to obtain the final fault type of the target branch.

2. The high-voltage distribution cabinet AI fault diagnosis method according to claim 1, characterized in that, The historical fault information includes the type of each historical fault, the corresponding historical fault impact index, and the main historical abnormal electrical parameters.

3. The high-voltage distribution cabinet AI fault diagnosis method according to claim 2, characterized in that, The adjustment strategy for monitoring the electrical parameters of the target branch is as follows: The initial monitoring frequency of each electrical parameter of the target branch is obtained based on the importance of the branch; Electrical parameters that were previously considered major historical anomalies in the target branch are designated as key electrical parameters. Based on the historical fault impact index corresponding to each historical fault for the key electrical parameters, the key attention index for each key electrical parameter is obtained. The initial monitoring frequency of each key electrical parameter of the target branch is adjusted by frequency boosting to obtain the first monitoring frequency of each key electrical parameter of the target branch. The degree of frequency boosting is determined based on the key attention index.

4. The high-voltage distribution cabinet AI fault diagnosis method according to claim 3, characterized in that, The key attention index for each key electrical parameter is obtained based on the historical fault impact index corresponding to each historical fault for the key electrical parameters. Specifically: ； Where i is the i-th key electrical parameter of the target branch, GZ is the key attention index, C is the number of times the current key electrical parameter has been used as a historical major abnormal electrical parameter, YZ is the corresponding historical fault impact index, n is the index, and K1 and K2 are the first preset weight and the second preset weight, respectively.

5. The high-voltage distribution cabinet AI fault diagnosis method according to claim 1, characterized in that, The diagnostic model is a deep learning model. It is trained using multiple sets of different historical fault types of the target branch and the historical time sequence vectors of the corresponding electrical parameters as training samples. After training, the model takes the time sequence vectors of the electrical parameters of the target branch as input and outputs the probability distribution of each fault type. Fault types with a probability greater than or equal to a preset probability threshold are designated as the second fault type.

6. The high-voltage distribution cabinet AI fault diagnosis method according to claim 1, characterized in that, The following steps are included after step S5: S6. Based on the preset fault information database, obtain the electrical parameters affected by each final fault type, and determine the number of final fault types matched by each of the first abnormal parameters. S7. Calculate the second key focus index for each of the first abnormal parameters based on the first health deviation degree of each of the first abnormal parameters and the number of the final fault types matched. S8. Based on the current monitoring frequency, perform a second frequency boosting adjustment operation on the electrical parameters corresponding to the first abnormal parameters of the target branch to obtain the second monitoring frequency of the electrical parameters corresponding to the first abnormal parameters of the target branch. The degree of boosting of the second frequency boosting adjustment operation is determined according to the second key attention index.

7. The high-voltage distribution cabinet AI fault diagnosis method according to claim 6, characterized in that, The calculation of the second key focus index is as follows: ； Where j is the j-th first abnormal parameter of the target branch, GZ2 is the second key focus index, GS is the number of the final fault types matched, and K3 and K4 are the third and fourth preset weights, respectively.

8. A high-voltage switchgear AI fault diagnosis system, employing the high-voltage switchgear AI fault diagnosis method as described in any one of claims 1 to 7, characterized in that, include: The monitoring and adjustment module is used to adjust the monitoring strategy for each electrical parameter of the target branch based on the importance of the target branch in the distribution cabinet and in combination with the historical fault information of the target branch. The first analysis module is used to obtain the time sequence vector of each electrical parameter of the target branch within a preset time period according to the corresponding monitoring strategy, and calculate the reconstruction error and prediction error based on the time sequence vector of the target electrical parameters through a pre-trained health benchmark model, and then calculate the first health deviation of the target electrical parameters. The first diagnostic module is used to take electrical parameters whose first health deviation is greater than the corresponding preset deviation threshold as first abnormal parameters, and to obtain all abnormal parameter combinations based on all first abnormal parameters of the target branch, and to match and retrieve all fault types in the preset fault information database based on the abnormal parameter combinations as the first fault type. The second diagnostic module is used to obtain the probability distribution of each fault type based on the time sequence vector of each electrical parameter of the target branch through a pre-trained diagnostic model, thereby obtaining the second fault type. The final diagnosis module is used to cross-validate the first fault type and the second fault type to obtain the final fault type of the target branch.

9. The high-voltage switchgear AI fault diagnosis system according to claim 8, characterized in that, The system also includes: The second analysis module is used to obtain the electrical parameters affected by each final fault type according to the preset fault information database, and to determine the number of final fault types matched by each first abnormal parameter. Based on the first health deviation degree of each first abnormal parameter and the number of final fault types matched, the second key attention index of each first abnormal parameter is calculated. The monitoring and adjustment module is further configured to perform a second frequency boosting adjustment operation on the electrical parameters corresponding to each first abnormal parameter of the target branch based on the current corresponding monitoring frequency, thereby obtaining the second monitoring frequency of the electrical parameters corresponding to each first abnormal parameter of the target branch. The degree of boosting of the second frequency boosting adjustment operation is determined according to the second key focus index.

10. A high-voltage switchgear AI fault diagnosis device, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 7.

Images