Electric reactor fault diagnosis model training method and electric reactor fault diagnosis method

By building and jointly training diagnostic models of multiple fault types, and using generative adversarial training technology, the problem of inaccurate fault diagnosis of reactors in the existing technology is solved, and higher fault diagnosis accuracy is achieved.

CN118940034AActive Publication Date: 2024-11-12GUANGDONG POWER GRID CO LTD +1
View PDF 15 Cites 0 Cited by

Patent Information

Application Number
CN202410989266.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-23
Publication Date
2024-11-12
Estimated Expiration
2044-07-23

AI Technical Summary

Technical Problem

The existing neural network-based reactor fault diagnosis method has the problem of inaccurate fault diagnosis results.

Method used

By obtaining the initial sample data set, constructing a subsample set of each fault type and training the initial diagnostic model, generating multiple batch sample sets for joint training, obtaining a second diagnostic model, and building a target diagnostic model through generation of adversarial training, Improve the accuracy of fault diagnosis.

Benefits of technology

The fault diagnosis accuracy of the reactor fault diagnosis model is improved, and it can more accurately determine which fault type the reactor associated data belongs to.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118940034A_ABST
    Figure CN118940034A_ABST
Patent Text Reader

Abstract

The invention discloses an electric reactor fault diagnosis model training method and an electric reactor fault diagnosis method. The electric reactor fault diagnosis model training method comprises the steps of obtaining an initial sample data set; constructing a sub-sample set corresponding to each fault type based on the initial sample data set; constructing W initial diagnosis models, and training the W initial diagnosis models based on the sub-sample set corresponding to each fault type to obtain first diagnosis models corresponding to the W fault types; generating a plurality of batch sample sets based on the sample data set; performing joint training on the first diagnosis models corresponding to the W fault types based on the multiple batch sample sets to obtain second diagnosis models corresponding to the W fault types; and constructing a discriminator, training the second diagnosis models respectively corresponding to the W fault types and the discriminator based on the initial sample data set, and obtaining target diagnosis models respectively corresponding to the W fault types under the condition that a training ending condition is satisfied. And the fault diagnosis precision is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of reactor fault diagnosis, and in particular to a reactor fault diagnosis model training method and a reactor fault diagnosis method. Background Art

[0002] Reactors play a dual role of control and protection in the distribution network system, and their operating conditions directly determine the operation of the entire power system. Therefore, fault diagnosis of reactors is of great significance.

[0003] A variety of diagnostic methods have been proposed, involving various artificial intelligence algorithms, such as neural networks, etc. However, the reactor fault diagnosis method based on neural networks has the problem of inaccurate fault diagnosis results. Summary of the invention

[0004] The invention provides a reactor fault diagnosis model training method and a reactor fault diagnosis method, so as to improve the fault diagnosis accuracy of the fault diagnosis model.

[0005] According to one aspect of the present invention, a method for training a fault diagnosis model for a reactor is provided, comprising:

[0006] Acquire an initial sample data set, wherein the initial sample data set includes sample data corresponding to W fault types of the reactor respectively;

[0007] Based on the initial sample data set, construct sub-sample sets corresponding to each fault type respectively;

[0008] Constructing W initial diagnosis models, and training the W initial diagnosis models respectively based on the sub-sample sets corresponding to each of the fault types, to obtain first diagnosis models corresponding to the W fault types respectively;

[0009] Generate multiple batch sample sets based on the initial sample data set, wherein the number of sample data corresponding to any fault type in each batch sample set is at least one;

[0010] Based on multiple batch sample sets, first diagnostic models corresponding to W fault types are jointly trained to obtain second diagnostic models corresponding to the W fault types: in each training batch, sample data of each fault type in the batch sample set is input into the first diagnostic model of the corresponding fault type, the output result of each first diagnostic model of the fault type is converted into vector form data, and the comprehensive output result corresponding to the first diagnostic model of each fault type is determined based on multiple vector form data corresponding to the first diagnostic model of each fault type; the first loss function of the batch is determined based on the comprehensive output results corresponding to the first diagnostic models of the W fault types and a preset vector, and model parameters of the W first diagnostic models of the batch are adjusted based on the first loss function of the batch;

[0011] Construct a discriminator, and perform generative adversarial training on the second diagnostic model corresponding to the W fault types and the discriminator based on the initial sample data set. When the training end conditions are met, the target diagnostic models corresponding to the W fault types are obtained, wherein the target diagnostic model of any fault type is used to determine the probability that the reactor-related data belongs to the fault type.

[0012] According to another aspect of the present invention, a method for diagnosing a fault of a reactor is provided, comprising:

[0013] Acquire reactor-related data of the reactor to be diagnosed;

[0014] The reactor-related data are respectively input into target diagnosis models corresponding to W fault types for fault diagnosis to obtain fault diagnosis results for the W fault types; the fault diagnosis result of the reactor to be diagnosed is determined based on the fault diagnosis results of the W fault types; wherein the target diagnosis model is trained based on the fault diagnosis model training method for the reactor described in any embodiment of the present invention.

[0015] According to another aspect of the present invention, there is provided a reactor fault diagnosis model training device, comprising:

[0016] An initial sample data set acquisition module, used to acquire an initial sample data set, wherein the initial sample data set includes sample data corresponding to W fault types of the reactor respectively;

[0017] A sub-sample set construction module, used to construct sub-sample sets corresponding to each fault type based on the initial sample data set;

[0018] An initial diagnosis model training module, used to construct W initial diagnosis models, and train the W initial diagnosis models respectively based on the sub-sample sets corresponding to each of the fault types, to obtain first diagnosis models corresponding to the W fault types respectively;

[0019] A batch sample set generation module, used to generate multiple batch sample sets based on the initial sample data set, wherein the number of sample data corresponding to any fault type in each batch sample set is at least one;

[0020] A first diagnostic model training module is used to jointly train first diagnostic models corresponding to W fault types based on multiple batch sample sets to obtain second diagnostic models corresponding to the W fault types: in each training batch, sample data of each fault type in the batch sample set is input into the first diagnostic model of the corresponding fault type, the output result of each first diagnostic model of the fault type is converted into vector form data, and the comprehensive output result corresponding to the first diagnostic model of each fault type is determined based on multiple vector form data corresponding to the first diagnostic model of each fault type; the first loss function of the batch is determined based on the comprehensive output results corresponding to the first diagnostic models of the W fault types and a preset vector, and model parameters of the W first diagnostic models of the batch are adjusted based on the first loss function of the batch;

[0021] The target diagnostic model determination module is used to construct a discriminator, and based on the initial sample data set, the second diagnostic model corresponding to the W fault types and the discriminator are subjected to generative adversarial training. When the training end conditions are met, the target diagnostic models corresponding to the W fault types are obtained, wherein the target diagnostic model of any fault type is used to determine the probability that the reactor-related data belongs to the fault type.

[0022] According to another aspect of the present invention, there is provided an electronic device, the electronic device comprising:

[0023] at least one processor; and

[0024] a memory communicatively connected to the at least one processor; wherein,

[0025] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the inductor fault diagnosis model training method described in any embodiment of the present invention, and / or the inductor fault diagnosis method described in any embodiment of the present invention.

[0026] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the reactor fault diagnosis model training method described in any embodiment of the present invention, and / or the reactor fault diagnosis method described in any embodiment of the present invention when executed.

[0027] The technical solution of the embodiment of the present invention is to obtain an initial sample data set, wherein the initial sample data set includes sample data corresponding to W fault types of the reactor; construct sub-sample sets corresponding to each fault type based on the initial sample data set; construct W initial diagnosis models, and train the W initial diagnosis models based on the sub-sample sets corresponding to each fault type to obtain first diagnosis models corresponding to the W fault types; generate multiple batch sample sets based on the sample data set, and the number of sample data corresponding to any fault type in each batch sample set is at least one; jointly train the first diagnosis models corresponding to the W fault types based on the multiple batch sample sets to obtain second diagnosis models corresponding to the W fault types: in each training batch, the sample data of each fault type in the batch sample set is input into the corresponding fault data set; In the first diagnostic model of each fault type, the output result of the first diagnostic model of each fault type is converted into vector form data, and the comprehensive output result corresponding to the first diagnostic model of each fault type is determined based on the multiple vector form data corresponding to the first diagnostic model of each fault type; the first loss function of the batch is determined based on the comprehensive output results corresponding to the first diagnostic models of W fault types and the preset vector, and the model parameters of the W first diagnostic models of the batch are adjusted based on the first loss function of the batch; the discriminator is constructed, and the second diagnostic model and the discriminator corresponding to the W fault types are respectively generated adversarial training based on the initial sample data set. When the training end condition is met, the target diagnostic model corresponding to the W fault types is obtained, wherein the target diagnostic model of any fault type is used to determine the probability that the reactor-related data belongs to the fault type. Through multiple stages of training, the target diagnostic model is obtained to improve the accuracy of fault diagnosis of the fault diagnosis model.

[0028] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0030] Figure 1 is a flow chart of a method for training a reactor fault diagnosis model provided by the first embodiment of the present invention;

[0031] Figure 2 is a flow chart of a reactor fault diagnosis method provided by Embodiment 2 of the present invention;

[0032] Figure 3 It is a structural schematic diagram of a reactor fault diagnosis model training device provided by Embodiment 3 of the present invention;

[0033] Figure 4 is a schematic structural diagram of a reactor fault diagnosis device provided by a fourth embodiment of the present invention;

[0034] Figure 5 It is a structural schematic diagram of an electronic device provided in Embodiment 5 of the present invention. DETAILED DESCRIPTION

[0035] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0036] It should be noted that the terms "first diagnostic model", "second diagnostic model", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0037] Embodiment 1

[0038] Figure 1 This is a flowchart of a method for training a fault diagnosis model for a reactor provided in the first embodiment of the present invention. This embodiment is applicable to the case of training a fault diagnosis model for a reactor. The method can be executed by a training device for a fault diagnosis model for a reactor. The training device for a fault diagnosis model for a reactor can be implemented in the form of hardware and / or software. The training device for a fault diagnosis model for a reactor can be configured in a computer, server or other electronic device for training a fault diagnosis model for a reactor. Figure 1 As shown, the method includes:

[0039] S110 , obtaining an initial sample data set, wherein the initial sample data set includes sample data corresponding to W fault types of the reactor.

[0040] Among them, the sample data refers to the reactor-associated data corresponding to the W fault types of the reactor, and specifically, the sample data includes at least one of the current data, voltage data, temperature data and vibration data of the reactor corresponding to each fault type. In this embodiment, within the preset time range, the edge voltage sensing device is used to collect voltage based on the preset time interval, and the voltage value collected each time is used as a feature; within the preset time range, the edge current sensing device is used to collect current based on the preset time interval, and the current value collected each time is used as a feature; within the preset time range, the edge temperature sensing device is used to collect temperature based on the preset time interval, and the temperature value collected each time is used as a feature; within the preset time range, the acceleration vibration sensor is used to collect real-time vibration data, and the original vibration signal of the reactor is processed by the wavelet packet decomposition method, and the obtained wavelet packet time-frequency energy matrix is ​​used, and each value in the matrix is ​​used as a feature; it can be understood that for the acceleration vibration sensor, during the operation of the reactor, the signal generated by the mechanical vibration of the core and winding structures contains rich equipment status information, and accurate monitoring and analysis of the vibration signal can realize intelligent diagnosis of mechanical faults of the reactor. In addition, the acceleration vibration sensor does not need to be in direct contact with the inductor, which is flexible, convenient and easy to implement detection without power outage.

[0041] It should be noted that the preset time intervals for collecting voltage, current and temperature may be the same or different, and are set by those skilled in the art according to requirements and are not limited here.

[0042] S120: Construct sub-sample sets corresponding to each fault type based on the initial sample data set.

[0043] In this embodiment, for each fault type, a sub-sample set corresponding to the fault type is constructed based on the initial sample data set.

[0044] On the basis of the above embodiments, optionally, a method for constructing a sub-sample set for the i-th fault type includes: for any sample data in the initial sample data set, if the sample data corresponds to the i-th fault type, setting the label of the sample data to a first label; if the sample data corresponds to a fault type other than the i-th fault type, setting the label of the sample data to a second label.

[0045] The first label is used to indicate that the sample data is sample data of the i-th fault type, and the second label is used to indicate that the sample data is not sample data of the i-th fault type; specifically, the first label may be 1, and the second label may be 0; illustratively, the sub-sample set of the first fault type is: The sub-sample set of the second fault type is: By analogy, W sub-sample sets corresponding to W fault types can be obtained. It can be understood that the number of sample data in each sub-sample set is W.

[0046] It is understandable that for each fault type sub-sample set, each sample data in the initial sample data set needs to be labeled, so the number of sample data in the sub-sample set corresponding to each fault type is the same as the number of sample data in the initial sample data set.

[0047] S130, constructing W initial diagnosis models, and training the W initial diagnosis models respectively based on the sub-sample sets corresponding to each of the fault types, to obtain first diagnosis models corresponding to the W fault types respectively.

[0048] Among them, the initial diagnosis model is a generator, and the initial diagnosis model can be a neural network model, such as a BP neural network model, a convolutional neural network model, a fully connected neural network model, etc. In this embodiment, for each fault type, an initial initial diagnosis model among W initial diagnosis models is trained based on the subsample set corresponding to the fault type to obtain the first diagnosis model of the fault type. Among them, the first diagnosis model is used to determine whether it is the fault type corresponding to the first diagnosis model. If it is the fault type corresponding to the first diagnosis model, the first diagnosis model outputs 1, and if it is not the fault type corresponding to the first fault model, the first diagnosis model outputs 0.

[0049] S140 . Generate multiple batch sample sets based on the initial sample data set, wherein the number of sample data corresponding to any fault type in each batch sample set is at least one.

[0050] It can be understood that, for sample data corresponding to any fault type, each batch of sample sets includes at least one sample data corresponding to the fault type; that is, the number of sample data in each batch of sample sets is at least W, and when the number of sample data in the batch sample set is W, the number of sample data corresponding to each fault type in the batch sample set is 1.

[0051] For example, in each batch of samples:

[0052] The sample data of the first fault type can be expressed as:

[0053] The sample data of the second fault type can be expressed as:

[0054] And so on…

[0055] The sample data of the Wth fault type can be expressed as:

[0056] Among them, n1, n2, …, nw ≥ 1.

[0057] On the basis of the above embodiment, optionally, generating multiple batch sample sets based on the initial sample data set includes: setting a unique hot code label for each sample data based on the fault type of each sample data in the initial sample data set; sampling the sample data corresponding to W fault types, and the sample data obtained by sampling and the unique hot code labels corresponding to the sample data form the batch sample set, wherein the sum of the number of sample data corresponding to the W fault types in each batch sample set is a preset batch sample number.

[0058] In this embodiment, the fault type of each sample data in the sample data set is numbered in a one-hot code manner, and a one-hot code label is set for each sample data; sampling is performed in the sample data corresponding to the W fault types respectively by sampling, and the sample data obtained by sampling and the one-hot code label corresponding to the sample data form a batch sample set, wherein the sum of the number of sample data corresponding to the W fault types respectively in each batch sample set is the preset batch sample number. The preset batch sample number refers to the number of sample data in the batch sample set, that is, the number of samples in the sample data corresponding to the W fault types respectively.

[0059] For example, in the batch sample set:

[0060] The sample data vector of the first fault type can be expressed as:

[0061] The sample data vector of the second fault type can be expressed as:

[0062] And so on…

[0063] The sample data vector of the Wth fault type can be expressed as:

[0064] in, is the i-th sample data vector of the W-th fault type, Include sample data and one-hot coded labels It can be understood that the one-hot code labels of the sample data vectors of each fault type are the same.

[0065] S150. Based on multiple batch sample sets, the first diagnostic models corresponding to W fault types are jointly trained to obtain second diagnostic models corresponding to the W fault types: in each training batch, the sample data of each fault type in the batch sample set is input into the first diagnostic model of the corresponding fault type, the output result of the first diagnostic model of each fault type is converted into vector data, and the comprehensive output result corresponding to the first diagnostic model of each fault type is determined based on the multiple vector data corresponding to the first diagnostic model of each fault type; the first loss function of the batch is determined based on the comprehensive output results corresponding to the first diagnostic models of the W fault types and a preset vector, and the model parameters of the W first diagnostic models of the batch are adjusted based on the first loss function of the batch.

[0066] In this embodiment, the training process of each batch of the second diagnostic model is as follows:

[0067] Step 1) Input the sample data of each fault type in the batch sample set into the first diagnostic model of the corresponding fault type, convert the output result of the first diagnostic model of each fault type into vector form data, and determine the comprehensive output result corresponding to the first diagnostic model of each fault type based on the multiple vector form data corresponding to the first diagnostic model of each fault type.

[0068] On the basis of the above embodiment, optionally, the output result of the first diagnostic model of each fault type is converted into vector form data, and the comprehensive output result corresponding to the first diagnostic model of each fault type is determined based on multiple vector form data corresponding to the first diagnostic model of each fault type, including: based on a first function, the output result of the first diagnostic model of the i-th fault type is converted into vector form data, wherein the vector form data includes w data, the i-th data of the vector form data is the output result of the first diagnostic model of the i-th fault type, and other data other than the i-th data of the vector form data are set to random numbers, and the sum of the w data is one; based on a second function, the vector form data is converted into a first intermediate vector, wherein the second function is used to set the maximum value in the vector form data to one and the non-maximum value in the vector form data to zero; the multiple intermediate vectors corresponding to the first diagnostic model of the i-th fault type are averaged to obtain a second intermediate vector; based on a third function, the second intermediate vector is converted into the comprehensive output result corresponding to the first diagnostic model of the i-th fault type, wherein the third function is used to set the non-maximum value in the second intermediate vector to zero and retain the maximum value in the second intermediate vector.

[0069] In this embodiment, the output result of the first diagnostic model of the i-th fault type is converted into vector data based on the first function; wherein the first function may be:

[0070] Gi represents the output result of the first diagnostic model of the i-th fault type; x 1 、x 2 , ..., x w-2 、x w-1 are all random values ​​in [0,1], and x 1 、x 2 , ..., x w-2 、x w-1 The sum of Gi is 1.

[0071] In this embodiment, the vector form data is converted into the first intermediate vector based on the second function. The second function may be: in, Represents vector form data, that is

[0072] For example, assuming represents the first intermediate vector.

[0073] It should be noted that if there are two maximum values ​​in the vector form data, a maximum value is randomly selected and set to 1.

[0074] In this embodiment, a plurality of intermediate vectors corresponding to the first diagnosis model of the i-th fault type are processed by averaging to obtain a second intermediate vector.

[0075] For example, the output result of the first diagnostic model of the first fault type is taken as an example.

[0076] The calculation formula of the second intermediate vector is:

[0077]

[0078] Among them, Z represents the second intermediate vector, Batch1 represents the data volume of the sample data of the first fault type in the batch sample set, Represents the i-th sample data in the sample data of the first fault type in the batch sample set Output after inputting the first diagnostic model.

[0079] In this embodiment, the second intermediate vector is converted into a comprehensive output result corresponding to the first diagnostic model of the i-th fault type based on the third function. The third function may be:

[0080]

[0081] in, represents the second intermediate vector, x i represents the maximum value in the second intermediate vector, x i =MAX(x 1 ,x 2 ,...,x w-1 ,x w ); Represents the comprehensive output result corresponding to the first diagnostic model.

[0082] Step 2) Determine the first loss function of the batch based on the comprehensive output results corresponding to the first diagnostic models of W fault types and the preset vector.

[0083] On the basis of the above embodiment, optionally, the first loss function of the batch is determined based on the comprehensive output results corresponding to the first diagnostic models of the W fault types and the preset vector, including: summing the comprehensive output results corresponding to the first diagnostic models of the W fault types to obtain the integrated processing results of the W fault types; determining the first loss function of the batch based on the integrated processing results of the W fault types and the spatial distance of the preset vector, wherein the preset vector includes W data, and the W data in the preset vector are 1 respectively.

[0084] The integrated processing result refers to the processing result obtained by summing up the comprehensive output results corresponding to the first diagnostic models of W fault types. 1 +T 2 +...+T w-1 +T w , T 1 、T 2 , ..., T w-1 and T w They are the comprehensive output results corresponding to the first diagnostic models of W fault types.

[0085] The calculation formula of the first loss function is:

[0086] loss1=J(A)

[0087] Among them, loss1 represents the first loss function, and J(A) is used to solve the integration processing result A and the preset vector The spatial distance of the preset vector There are w 1s in total.

[0088] Step 3) Based on the first loss function of the batch, the model parameters of the W first diagnostic models of the batch are adjusted respectively.

[0089] S160. Construct a discriminator, and perform generative adversarial training on the second diagnostic models corresponding to the W fault types and the discriminator based on the initial sample data set. When the training end conditions are met, the target diagnostic models corresponding to the W fault types are obtained, wherein the target diagnostic model of any fault type is used to determine the probability that the reactor-related data belongs to the fault type.

[0090] The discriminator may be a multi-layer perceptron. In this embodiment, generative adversarial training is performed on the second diagnostic model and the discriminator corresponding to the W fault types respectively based on the initial sample data set, and when the training end condition is met, the target diagnostic model corresponding to the W fault types respectively is obtained. The training end condition refers to the condition for the end of generative adversarial training. Specifically, the training end condition may be that the classification accuracy of the second diagnostic model reaches a preset accuracy, and the training end condition may also be that the number of training times reaches a preset number of times.

[0091] On the basis of the above embodiment, optionally, the second diagnostic model and the discriminator corresponding to the W fault types are generated adversarial training based on the initial sample data set, including: setting a unique hot code label for each sample data based on the fault type of each sample data in the initial sample data set; inputting any sample data into the second diagnostic models corresponding to the W fault types to obtain the output results of the second diagnostic models corresponding to the W fault types; integrating the output results of the second diagnostic models corresponding to the W fault types into a third intermediate vector, and inputting the third intermediate vector into the discriminator to obtain the output result of the discriminator; based on the output results of the second diagnostic models corresponding to the W fault types, the third intermediate vector, the output result of the discriminator and one or more of the unique hot code labels of the sample data, respectively generate a second loss function of the second diagnostic model corresponding to the W fault types and a third loss function of the discriminator; adjusting the model parameters of the second diagnostic models corresponding to the W fault types based on the second loss function, and adjusting the model parameters of the discriminator based on the third loss function.

[0092] In this embodiment, the second diagnostic model and the discriminator perform generative adversarial training as follows:

[0093] 1) Number the fault type of each sample data in the initial sample data set in a one-hot code manner, and set a one-hot code label for each sample data;

[0094] 2) Inputting any sample data into the second diagnostic models corresponding to the W fault types, respectively, to obtain the output results of the second diagnostic models corresponding to the W fault types, respectively;

[0095] 3) Integrating the output results of the second diagnostic models corresponding to the W fault types into a third intermediate vector; wherein the third intermediate vector is obtained by integrating the output results of the second diagnostic models corresponding to the W fault types;

[0096] 4) Inputting the third intermediate vector into the discriminator to obtain the output result of the discriminator;

[0097] 5) Based on one or more of the output results of the second diagnostic models corresponding to the W fault types, the third intermediate vector, the output result of the discriminator, and the unique hot code label of the sample data, respectively generate the second loss function of the second diagnostic model corresponding to the W fault types and the third loss function of the discriminator;

[0098] 6) adjusting the model parameters of the second diagnostic models corresponding to the W fault types respectively based on the second loss function, and adjusting the model parameters of the discriminator based on the third loss function;

[0099] 7) When the training end conditions are met, the target diagnosis models corresponding to W fault types are obtained.

[0100] Specifically, the second loss function of the second diagnostic model includes an adversarial loss variable function and a cross entropy loss function, and the specific expression is as follows:

[0101]

[0102]

[0103] Among them, loss G Represents the second loss function of the second diagnostic model, loss G1 Represents the adversarial loss variable function, loss G2 Cross entropy loss function, batch represents the amount of sample data contained in each batch during model training, W represents the number of second diagnostic models, b im represents the unique hot label of the mth second diagnostic model in the ith sample data, G im represents the sample data of the mth second diagnostic model in the ith sample data, Indicates that the i-th sample data discriminator accepts the input fault sample data After the output,

[0104] The third loss function of the discriminator includes the real fault label loss function and the generated fault label loss function. The specific expression is as follows:

[0105]

[0106] Among them, loss D Represents the third loss function of the discriminator, loss D1 Represents the true one-hot code label loss function, loss D2 represents the loss function for generating one-hot-coded labels, Indicates the unique hot code label corresponding to the i-th sample data input accepted by the discriminator After the output, The technical solution of this embodiment is to obtain an initial sample data set, which includes sample data corresponding to W fault types of the reactor; construct sub-sample sets corresponding to each fault type based on the initial sample data set; construct W initial diagnosis models, and train the W initial diagnosis models based on the sub-sample sets corresponding to each fault type to obtain first diagnosis models corresponding to the W fault types; generate multiple batch sample sets based on the sample data set, and the number of sample data corresponding to any fault type in each batch sample set is at least one; jointly train the first diagnosis models corresponding to the W fault types based on the multiple batch sample sets to obtain second diagnosis models corresponding to the W fault types: in each training batch, the sample data of each fault type in the batch sample set is input into the sub-sample sets corresponding to the fault type; In the first diagnostic model of the type, the output result of the first diagnostic model of each fault type is converted into vector form data, and the comprehensive output result corresponding to the first diagnostic model of each fault type is determined based on the multiple vector form data corresponding to the first diagnostic model of each fault type; the first loss function of the batch is determined based on the comprehensive output results corresponding to the first diagnostic models of W fault types and the preset vector, and the model parameters of the W first diagnostic models of the batch are adjusted based on the first loss function of the batch; the discriminator is constructed, and the second diagnostic model and the discriminator corresponding to the W fault types are respectively generated adversarial training based on the initial sample data set. When the training end condition is met, the target diagnostic model corresponding to the W fault types is obtained, wherein the target diagnostic model of any fault type is used to determine the probability that the reactor-related data belongs to the fault type. Through multiple stages of training, the target diagnostic model is obtained to improve the accuracy of fault diagnosis of the fault diagnosis model.

[0107] Embodiment 2

[0108] Figure 2 1 is a flow chart of a reactor fault diagnosis method provided by the second embodiment of the present invention. This embodiment performs fault diagnosis on the reactor to be diagnosed based on the target diagnosis model obtained by training the reactor fault diagnosis model training method of the above embodiment. Figure 2 As shown, the method includes:

[0109] S210: Acquire reactor-related data of the reactor to be diagnosed.

[0110] S220. Input the reactor-related data into target diagnosis models corresponding to W fault types respectively for fault diagnosis to obtain fault diagnosis results for the W fault types; determine the fault diagnosis result of the reactor to be diagnosed based on the fault diagnosis results of the W fault types; wherein the target diagnosis model is trained based on the fault diagnosis model training method for the reactor described in any embodiment of the present invention.

[0111] The reactor-related data includes at least one of the current data, voltage data, temperature data and vibration data of the reactor. In this embodiment, an edge voltage sensing device is used to collect the voltage value of the reactor; an edge current sensing device is used to collect the current value of the reactor; an edge temperature sensing device is used to collect the temperature value of the reactor; an acceleration vibration sensor is used to collect the real-time vibration data of the reactor, and the original vibration signal of the reactor is processed using the wavelet packet decomposition method, and the obtained wavelet packet time-frequency energy matrix, each value in the matrix is ​​used as a feature.

[0112] In this embodiment, the reactor-related data are respectively input into the target diagnosis models corresponding to the W fault types for fault diagnosis, and the fault diagnosis results of the W fault types are obtained; wherein the fault diagnosis result of each fault type is the probability that the reactor-related data belongs to the fault type. Further, the fault diagnosis result of the reactor to be diagnosed is determined based on the fault diagnosis results of the W fault types. Specifically, the fault diagnosis results of the W fault types can be sorted, and the fault diagnosis result with the highest sorting is used as the fault diagnosis result of the reactor to be diagnosed.

[0113] The technical solution of this embodiment performs fault diagnosis on the circuit breaker to be diagnosed through target diagnosis models corresponding to W fault types of the reactor, and determines the fault diagnosis result of the reactor to be diagnosed based on the fault diagnosis results of the W fault types, thereby improving the accuracy of fault diagnosis.

[0114] Embodiment 3

[0115] Figure 3 Schematic diagram of the structure of a reactor fault diagnosis model training device provided by the third embodiment of the present invention. Figure 3 As shown, the device comprises:

[0116] An initial sample data set acquisition module 310 is used to acquire an initial sample data set, wherein the initial sample data set includes sample data corresponding to W fault types of the reactor respectively;

[0117] A sub-sample set construction module 320 is used to construct sub-sample sets corresponding to each fault type based on the initial sample data set;

[0118] An initial diagnosis model training module 330 is used to construct W initial diagnosis models, and train the W initial diagnosis models based on the sub-sample sets corresponding to each of the fault types to obtain first diagnosis models corresponding to the W fault types;

[0119] A batch sample set generation module 340 is used to generate multiple batch sample sets based on the initial sample data set, and the number of sample data corresponding to any fault type in each batch sample set is at least one;

[0120] The first diagnostic model training module 350 is used to jointly train the first diagnostic models corresponding to W fault types based on multiple batch sample sets to obtain the second diagnostic models corresponding to the W fault types: in each training batch, the sample data of each fault type in the batch sample set is input into the first diagnostic model of the corresponding fault type, the output result of each first diagnostic model of the fault type is converted into vector form data, and the comprehensive output result corresponding to the first diagnostic model of each fault type is determined based on the multiple vector form data corresponding to the first diagnostic model of each fault type; the first loss function of the batch is determined based on the comprehensive output results corresponding to the first diagnostic models of the W fault types and the preset vector, and the model parameters of the W first diagnostic models of the batch are adjusted based on the first loss function of the batch;

[0121] The target diagnostic model determination module 360 ​​is used to construct a discriminator, and based on the initial sample data set, generative adversarial training is performed on the second diagnostic model corresponding to the W fault types and the discriminator. When the training end conditions are met, the target diagnostic models corresponding to the W fault types are obtained, wherein the target diagnostic model of any fault type is used to determine the probability that the reactor-related data belongs to the fault type.

[0122] The technical solution of this embodiment is to obtain an initial sample data set, which includes sample data corresponding to W fault types of the reactor; construct sub-sample sets corresponding to each fault type based on the initial sample data set; construct W initial diagnosis models, and train the W initial diagnosis models based on the sub-sample sets corresponding to each fault type to obtain first diagnosis models corresponding to the W fault types; generate multiple batch sample sets based on the sample data set, and the number of sample data corresponding to any fault type in each batch sample set is at least one; jointly train the first diagnosis models corresponding to the W fault types based on the multiple batch sample sets to obtain second diagnosis models corresponding to the W fault types: in each training batch, the sample data of each fault type in the batch sample set is input into the sub-sample sets corresponding to the fault type; In the first diagnostic model of the type, the output result of the first diagnostic model of each fault type is converted into vector form data, and the comprehensive output result corresponding to the first diagnostic model of each fault type is determined based on the multiple vector form data corresponding to the first diagnostic model of each fault type; the first loss function of the batch is determined based on the comprehensive output results corresponding to the first diagnostic models of W fault types and the preset vector, and the model parameters of the W first diagnostic models of the batch are adjusted based on the first loss function of the batch; the discriminator is constructed, and the second diagnostic model and the discriminator corresponding to the W fault types are respectively generated adversarial training based on the initial sample data set. When the training end condition is met, the target diagnostic model corresponding to the W fault types is obtained, wherein the target diagnostic model of any fault type is used to determine the probability that the reactor-related data belongs to the fault type. Through multiple stages of training, the target diagnostic model is obtained to improve the accuracy of fault diagnosis of the fault diagnosis model.

[0123] On the basis of the above embodiment, optionally, the number of sample data in the sub-sample set corresponding to each fault type is the same as the number of sample data in the initial sample data set; the sub-sample set construction module 320 is specifically used for, for any sample data in the initial sample data set, if the sample data corresponds to the i-th fault type, setting the label of the sample data to the first label; if the sample data corresponds to other fault types other than the i-th fault type, setting the label of the sample data to the second label.

[0124] On the basis of the above embodiment, optionally, the batch sample set generation module 340 is specifically used to set a unique hot code label for each sample data based on the fault type of each sample data in the initial sample data set; sampling is performed in the sample data corresponding to W fault types, and the sample data obtained by sampling and the unique hot code label corresponding to the sample data form the batch sample set, wherein the sum of the number of sample data corresponding to the W fault types in each batch sample set is a preset batch sample number.

[0125] On the basis of the above embodiment, optionally, the first diagnostic model training module 350 includes a comprehensive output result determination unit that converts the output result of the first diagnostic model of the i-th fault type into vector form data based on a first function, wherein the vector form data includes w data, the i-th data of the vector form data is the output result of the first diagnostic model of the i-th fault type, other data other than the i-th data of the vector form data is set to random numbers, and the sum of the w data is one; converts the vector form data into a first intermediate vector based on a second function, wherein the second function is used to set the maximum value in the vector form data to one and the non-maximum value in the vector form data to zero; performs mean processing on multiple intermediate vectors corresponding to the first diagnostic model of the i-th fault type to obtain a second intermediate vector; converts the second intermediate vector into a comprehensive output result corresponding to the first diagnostic model of the i-th fault type based on a third function, wherein the third function is used to set the non-maximum value in the second intermediate vector to zero and retain the maximum value in the second intermediate vector.

[0126] Based on the above embodiments, optionally, the first diagnostic model training module includes a first loss function determination unit, which is used to sum the comprehensive output results corresponding to the first diagnostic models of W fault types to obtain the integrated processing results of the W fault types; determine the first loss function of the batch based on the integrated processing results of the W fault types and the spatial distance of the preset vector, wherein the preset vector includes W data, and the W data in the preset vector are respectively 1.

[0127] On the basis of the above embodiment, optionally, the target diagnostic model determination module is specifically used to set a unique-hot code label for each sample data based on the fault type of each sample data in the initial sample data set; input any sample data into the second diagnostic models corresponding to the W fault types respectively, and obtain the output results of the second diagnostic models corresponding to the W fault types respectively; integrate the output results of the second diagnostic models corresponding to the W fault types respectively into a third intermediate vector, input the third intermediate vector into the discriminator, and obtain the output result of the discriminator; based on the output results of the second diagnostic models corresponding to the W fault types respectively, the third intermediate vector, the output result of the discriminator and one or more of the unique-hot code labels of the sample data, respectively generate a second loss function of the second diagnostic model corresponding to the W fault types and a third loss function of the discriminator; adjust the model parameters of the second diagnostic models corresponding to the W fault types respectively based on the second loss function, and adjust the model parameters of the discriminator based on the third loss function.

[0128] The reactor fault diagnosis model training device provided in the embodiment of the present invention can execute the reactor fault diagnosis model training method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0129] Embodiment 4

[0130] Figure 4 Schematic diagram of the structure of a reactor fault diagnosis device provided by the fourth embodiment of the present invention. Figure 4 As shown, the device comprises:

[0131] A reactor-related data acquisition module 410 is used to acquire reactor-related data of the reactor to be diagnosed;

[0132] A fault diagnosis module 420 is used to input the reactor-related data into target diagnosis models corresponding to W fault types for fault diagnosis, and obtain fault diagnosis results for the W fault types; based on the fault diagnosis results for the W fault types, the fault diagnosis results of the reactor to be diagnosed are determined; wherein the target diagnosis model is trained based on the fault diagnosis model training method for the reactor described in any embodiment of the present invention.

[0133] The technical solution of this embodiment performs fault diagnosis on the circuit breaker to be diagnosed through target diagnosis models corresponding to W fault types of the reactor, and determines the fault diagnosis result of the reactor to be diagnosed based on the fault diagnosis results of the W fault types, thereby improving the accuracy of fault diagnosis.

[0134] The reactor fault diagnosis device provided in the embodiment of the present invention can execute the reactor fault diagnosis method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0135] Embodiment 5

[0136] Figure 5 1 is a schematic diagram of the structure of an electronic device provided in Embodiment 5 of the present invention. The electronic device 10 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0137] like Figure 5As shown, the electronic device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0138] A number of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0139] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The processor 11 executes the various methods and processes described above, such as a fault diagnosis model training method for a reactor, and / or a reactor fault diagnosis method.

[0140] In some embodiments, the fault diagnosis model training method for the reactor, and / or, the fault diagnosis method for the reactor may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, the fault diagnosis model training method for the reactor described above, and / or, one or more steps of the fault diagnosis method for the reactor may be executed. Alternatively, in other embodiments, the processor 11 may be configured to execute the fault diagnosis model training method for the reactor, and / or, the fault diagnosis method for the reactor by any other appropriate means (e.g., by means of firmware).

[0141] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0142] The computer program for implementing the reactor fault diagnosis model training method of the present invention, and / or the reactor fault diagnosis method can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or the block diagram are implemented. The computer program can be executed entirely on the machine, partially on the machine, partially on the machine as an independent software package and partially on a remote machine, or entirely on a remote machine or server.

[0143] Embodiment 6

[0144] Embodiment 6 of the present invention further provides a computer-readable storage medium, the computer-readable storage medium storing computer instructions, the computer instructions being used to cause a processor to execute a method for training a reactor fault diagnosis model, the method comprising:

[0145] Acquire an initial sample data set, wherein the initial sample data set includes sample data corresponding to W fault types of the reactor respectively;

[0146] Based on the initial sample data set, construct sub-sample sets corresponding to each fault type respectively;

[0147] Constructing W initial diagnosis models, and training the W initial diagnosis models respectively based on the sub-sample sets corresponding to each of the fault types, to obtain first diagnosis models corresponding to the W fault types respectively;

[0148] Generate multiple batch sample sets based on the initial sample data set, wherein the number of sample data corresponding to any fault type in each batch sample set is at least one;

[0149] Based on multiple batch sample sets, first diagnostic models corresponding to W fault types are jointly trained to obtain second diagnostic models corresponding to the W fault types: in each training batch, sample data of each fault type in the batch sample set is input into the first diagnostic model of the corresponding fault type, the output result of each first diagnostic model of the fault type is converted into vector form data, and the comprehensive output result corresponding to the first diagnostic model of each fault type is determined based on multiple vector form data corresponding to the first diagnostic model of each fault type; the first loss function of the batch is determined based on the comprehensive output results corresponding to the first diagnostic models of the W fault types and a preset vector, and model parameters of the W first diagnostic models of the batch are adjusted based on the first loss function of the batch;

[0150] Construct a discriminator, and perform generative adversarial training on the second diagnostic model corresponding to the W fault types and the discriminator based on the initial sample data set. When the training end conditions are met, the target diagnostic models corresponding to the W fault types are obtained, wherein the target diagnostic model of any fault type is used to determine the probability that the reactor-related data belongs to the fault type.

[0151] And / or, performing a reactor fault diagnosis method, the method comprising:

[0152] Acquire reactor-related data of the reactor to be diagnosed;

[0153] The reactor-related data are respectively input into target diagnosis models corresponding to W fault types for fault diagnosis to obtain fault diagnosis results for the W fault types; the fault diagnosis result of the reactor to be diagnosed is determined based on the fault diagnosis results of the W fault types; wherein the target diagnosis model is trained based on the fault diagnosis model training method for the reactor described in any embodiment of the present invention.

[0154] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0155] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0156] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0157] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0158] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0159] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for training a reactor fault diagnosis model, characterized in that: include: Acquire an initial sample data set, wherein the initial sample data set includes sample data corresponding to W fault types of the reactor respectively; Based on the initial sample data set, construct sub-sample sets corresponding to each fault type respectively; Constructing W initial diagnosis models, and training the W initial diagnosis models respectively based on the sub-sample sets corresponding to each of the fault types, to obtain first diagnosis models corresponding to the W fault types respectively; Generate multiple batch sample sets based on the sample data set, wherein the number of sample data corresponding to any fault type in each batch sample set is at least one; Based on multiple batch sample sets, first diagnostic models corresponding to W fault types are jointly trained to obtain second diagnostic models corresponding to the W fault types: in each training batch, sample data of each fault type in the batch sample set is input into the first diagnostic model of the corresponding fault type, the output result of each first diagnostic model of the fault type is converted into vector form data, and the comprehensive output result corresponding to the first diagnostic model of each fault type is determined based on multiple vector form data corresponding to the first diagnostic model of each fault type; the first loss function of the batch is determined based on the comprehensive output results corresponding to the first diagnostic models of the W fault types and a preset vector, and model parameters of the W first diagnostic models of the batch are adjusted based on the first loss function of the batch; Construct a discriminator, and perform generative adversarial training on the second diagnostic model corresponding to the W fault types and the discriminator based on the initial sample data set. When the training end conditions are met, the target diagnostic models corresponding to the W fault types are obtained, wherein the target diagnostic model of any fault type is used to determine the probability that the reactor-related data belongs to the fault type.

2. The method according to claim 1, characterized in that The number of sample data in the sub-sample set corresponding to each fault type is the same as the number of sample data in the initial sample data set; The construction method of the sub-sample set of the i-th fault type includes: For any sample data in the initial sample data set, if the sample data corresponds to the i-th fault type, setting the label of the sample data to the first label; If the sample data corresponds to a fault type other than the i-th fault type, the label of the sample data is set to the second label.

3. The method according to claim 1, characterized in that The generating a plurality of batch sample sets based on the initial sample data set comprises: Based on the fault type of each sample data in the initial sample data set, setting a unique hot code label for each sample data; Sampling is performed on the sample data corresponding to the W fault types respectively, and the sample data obtained by sampling and the one-hot code labels corresponding to the sample data form the batch sample set, wherein the sum of the number of sample data corresponding to the W fault types respectively in each batch sample set is the preset batch sample number.

4. The method according to claim 1, characterized in that The step of converting the output result of each of the first diagnostic models of the fault type into vector data, and determining the comprehensive output result corresponding to each of the first diagnostic models of the fault type based on a plurality of vector data corresponding to each of the first diagnostic models of the fault type, comprises: Based on the first function, converting the output result of the first diagnostic model of the i-th fault type into vector form data, wherein the vector form data includes w data, the i-th data of the vector form data is the output result of the first diagnostic model of the i-th fault type, and other data except the i-th data of the vector form data are set to random numbers, and the sum of the w data is one; converting the vector form data into a first intermediate vector based on a second function, wherein the second function is used to set the maximum value in the vector form data to one and set the non-maximum value in the vector form data to zero; Performing mean processing on a plurality of intermediate vectors corresponding to the first diagnostic model of the i-th fault type to obtain a second intermediate vector; The second intermediate vector is converted into a comprehensive output result corresponding to the first diagnostic model of the i-th fault type based on a third function, wherein the third function is used to set the non-maximum value in the second intermediate vector to zero and retain the maximum value in the second intermediate vector.

5. The method according to claim 1, characterized in that The step of determining the first loss function of the batch based on the comprehensive output results respectively corresponding to the first diagnostic models of the W fault types and the preset vector comprises: Sum up the comprehensive output results corresponding to the first diagnostic models of W fault types to obtain the integrated processing results of the W fault types; The first loss function of the batch is determined based on the integrated processing results of the W fault types and the spatial distance of the preset vector, wherein the preset vector includes W data, and the W data in the preset vector are 1 respectively.

6. The method according to claim 1, characterized in that The performing generative adversarial training on the second diagnostic model and the discriminator respectively corresponding to the W fault types based on the initial sample data set includes: Based on the fault type of each sample data in the initial sample data set, setting a unique hot code label for each sample data; Inputting any sample data into the second diagnostic models corresponding to the W fault types, respectively, to obtain the output results of the second diagnostic models corresponding to the W fault types, respectively; Integrate the output results of the second diagnostic models corresponding to the W fault types into a third intermediate vector, input the third intermediate vector into the discriminator, and obtain the output result of the discriminator; Based on one or more of the output results of the second diagnostic models respectively corresponding to the W fault types, the third intermediate vector, the output result of the discriminator, and the one-hot code label of the sample data, respectively generate a second loss function of the second diagnostic model respectively corresponding to the W fault types and a third loss function of the discriminator; The model parameters of the second diagnostic models respectively corresponding to the W fault types are adjusted based on the second loss function, and the model parameters of the discriminator are adjusted based on the third loss function.

7. A reactor fault diagnosis method, characterized in that: include: Acquire reactor-related data of the reactor to be diagnosed; The reactor-related data are respectively input into target diagnosis models corresponding to W fault types for fault diagnosis to obtain fault diagnosis results for the W fault types; the fault diagnosis result of the reactor to be diagnosed is determined based on the fault diagnosis results of the W fault types; wherein the target diagnosis model is trained based on the fault diagnosis model training method for the reactor described in any one of claims 1-6.

8. A reactor fault diagnosis model training device, characterized in that: include: An initial sample data set acquisition module, used to acquire an initial sample data set, wherein the initial sample data set includes sample data corresponding to W fault types of the reactor respectively; A sub-sample set construction module, used to construct sub-sample sets corresponding to each fault type based on the initial sample data set; An initial diagnosis model training module, used to construct W initial diagnosis models, and train the W initial diagnosis models respectively based on the sub-sample sets corresponding to each of the fault types, to obtain first diagnosis models corresponding to the W fault types respectively; A batch sample set generation module, used to generate multiple batch sample sets based on the initial sample data set, wherein the number of sample data corresponding to any fault type in each batch sample set is at least one; A first diagnostic model training module is used to jointly train first diagnostic models corresponding to W fault types based on multiple batch sample sets to obtain second diagnostic models corresponding to the W fault types: in each training batch, sample data of each fault type in the batch sample set is input into the first diagnostic model of the corresponding fault type, the output result of each first diagnostic model of the fault type is converted into vector form data, and the comprehensive output result corresponding to the first diagnostic model of each fault type is determined based on multiple vector form data corresponding to the first diagnostic model of each fault type; the first loss function of the batch is determined based on the comprehensive output results corresponding to the first diagnostic models of the W fault types and a preset vector, and model parameters of the W first diagnostic models of the batch are adjusted based on the first loss function of the batch; The target diagnostic model determination module is used to construct a discriminator, and based on the initial sample data set, the second diagnostic model corresponding to the W fault types and the discriminator are subjected to generative adversarial training. When the training end conditions are met, the target diagnostic models corresponding to the W fault types are obtained, wherein the target diagnostic model of any fault type is used to determine the probability that the reactor-related data belongs to the fault type.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the reactor fault diagnosis model training method described in any one of claims 1 to 6, and / or the reactor fault diagnosis method described in claim 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the reactor fault diagnosis model training method described in any one of claims 1 to 6 and / or the reactor fault diagnosis method described in claim 7 when executed.

Citation Information

Patent Citations

  • Oil immersed type reactor fault diagnosis method based on two-way deep network

    CN108170994A

  • Transformer fault diagnosis method and system based on conditional generative adversarial network

    CN108680807A

  • Intelligent monitoring and identification method for state of substation equipment

    CN110175571A

  • Model training method and device, fault prediction method and device and electronic equipment

    CN112149301A

  • Fault diagnosis and early warning method, system and equipment for multi-mode operation of slurry circulating pump

    CN113392912A