Transformer fault data sample generation method and device and electronic equipment
By acquiring the feature values of transformer fault data, establishing a feature model, and randomly collecting feature values to generate new fault data samples, the problem of unbalanced transformer fault data is solved, and the fault data is expanded and the accuracy of diagnosis is improved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA MOBILE (XIONGAN) ICT CO LTD
- Filing Date
- 2022-05-09
- Publication Date
- 2026-06-23
AI Technical Summary
In existing technologies, the unbalanced distribution of categories in transformer fault data samples makes it difficult for traditional oversampling methods to effectively expand the scope of fault data and generate new fault data samples.
By acquiring the feature values of transformer fault data, a feature model is established. Temporary synthetic data is generated by randomly collecting feature values and inputting them into the feature model for prediction, thereby generating new fault data samples and expanding the scope of fault data.
This effectively expanded the scope of fault data, generated new fault data samples that conform to characteristic relationships, and improved the accuracy and diversity of transformer fault diagnosis.
Smart Images
Figure CN117094084B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of transformer technology, specifically to a method, apparatus, and electronic device for generating transformer fault data samples. Background Technology
[0002] Transformers are fundamental equipment for power transmission and distribution, widely used in industry, agriculture, transportation, and urban communities. As one of the core electrical devices in a power system, the operating status of power transformers directly affects the safety and stability of the power system. Therefore, research on power transformer fault diagnosis methods is of great significance.
[0003] Each power transformer generates a large amount of sample data during operation, but obtaining fault data is very difficult. Because transformer faults are low-probability events, the class distribution of the data samples is imbalanced. Imbalanced data refers to a significant imbalance in the number of samples in each class of a dataset. Taking a binary classification problem as an example, in a dataset, at least one class has significantly more or fewer data than the other class; this type of data is usually called imbalanced data. Therefore, power transformers not only have a small sample size characteristic, but also exhibit a significant imbalance in the quantity of fault data and normal operation data, meaning that the number of normal operation data is significantly greater than the number of fault data.
[0004] In imbalanced datasets, traditional oversampling methods copy minority class samples, resulting in numerous overlapping points in the feature space. Ideally, this enhances the importance of minority class data, but it doesn't expand the scope of minority class data. Transformer fault data is also a minority class data; copying samples does not expand the scope of transformer fault data. Therefore, there is an urgent need for a method that can expand the scope of fault data samples and generate new transformer fault data samples. Summary of the Invention
[0005] This application provides a method, apparatus, and electronic device for generating transformer fault data samples to solve the technical problem that fault data generated by copying samples cannot expand the scope of transformer fault data.
[0006] In a first aspect, embodiments of this application provide a method for generating transformer fault data samples, comprising:
[0007] Obtain transformer fault data, sequentially collect feature values from the value set of each feature of the fault data, and obtain temporary synthetic data based on the feature values;
[0008] Input the temporary synthetic data into the feature model to obtain the predicted fault data output by the feature model, wherein the feature model is trained based on the set of values of other features of the fault data, using one feature of the fault data as a label.
[0009] New fault data samples are generated based on the predicted fault data.
[0010] In one embodiment, the step of sequentially collecting feature values from the value set of each feature of the fault data, and obtaining temporary synthetic data based on the feature values, includes:
[0011] A feature value is randomly selected from the set of values for each feature of the fault data in sequence;
[0012] Based on a feature value corresponding to each feature of the fault data, a synthetic data is obtained;
[0013] Based on the synthesized data, temporary synthesized data is obtained.
[0014] In one embodiment, the number of feature models corresponds to the number of features in the fault data.
[0015] In one embodiment, inputting the temporary synthetic data into the feature model includes:
[0016] The target features of the temporary synthetic data are determined, and the target features of the temporary synthetic data are the labels of the feature model during the training process;
[0017] Input the feature values corresponding to the other features of the temporary synthetic data besides the target feature into the feature model.
[0018] In one embodiment, the number of temporary synthetic data and the number of fault data samples are determined based on the number of fault data samples and the oversampling rate of the fault data.
[0019] In one embodiment, it also includes:
[0020] Obtain non-fault data of the transformer;
[0021] Based on the non-fault data, the fault data, and the fault data samples, a transformer fault diagnosis model is trained using a machine learning algorithm. The transformer fault diagnosis model is used to predict the category of new transformer data and map the new transformer data as fault-type data or non-fault-type data.
[0022] In one embodiment, it also includes:
[0023] A transformer fault data information database is established based on the fault-type data and the non-fault-type data.
[0024] Secondly, embodiments of this application provide an apparatus for generating transformer fault data samples, comprising:
[0025] The data synthesis module is used to acquire transformer fault data, sequentially collect feature values from the value set of each feature of the fault data, and obtain temporary synthesized data based on the feature values;
[0026] The data prediction module is used to input the temporary synthetic data into the feature model to obtain the predicted fault data output by the feature model, wherein the feature model is trained based on the set of values of other features of the fault data, using one feature of the fault data as a label.
[0027] The sample generation module is used to generate new fault data samples based on the predicted fault data.
[0028] In one embodiment, the step of sequentially collecting feature values from the value set of each feature of the fault data, and obtaining temporary synthetic data based on the feature values, includes:
[0029] A feature value is randomly selected from the set of values for each feature of the fault data in sequence;
[0030] Based on a feature value corresponding to each feature of the fault data, a synthetic data is obtained;
[0031] Based on the synthesized data, temporary synthesized data is obtained.
[0032] Thirdly, embodiments of this application provide an electronic device, including a processor and a memory storing a computer program, wherein the processor executes the program to implement the steps of the transformer fault data sample generation method described in the first aspect.
[0033] Fourthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the steps of the method for generating transformer fault data samples as described in the first aspect.
[0034] The method and apparatus for generating transformer fault data samples provided in this application expand the scope of fault data by establishing a model to learn the relationship between fault data features and generating new fault data samples based on the learned feature model. Attached Figure Description
[0035] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0036] Figure 1 This is a flowchart illustrating the method for generating transformer fault data samples provided in an embodiment of this application;
[0037] Figure 2 This is a schematic diagram of the process for generating temporary synthetic data provided in an embodiment of this application;
[0038] Figure 3 This is a schematic diagram of the process for generating fault data samples provided in an embodiment of this application;
[0039] Figure 4 This is a schematic diagram of the structure of the transformer fault data sample generation device provided in the embodiments of this application;
[0040] Figure 5 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation
[0041] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0042] Figure 1 This is a flowchart illustrating the method for generating transformer fault data samples provided in an embodiment of this application. (Refer to...) Figure 1 This application provides a method for generating transformer fault data samples, which may include:
[0043] S1, acquire transformer fault data, sequentially collect feature values from the value set of each feature of the fault data, and obtain temporary synthetic data based on the feature values.
[0044] Specifically, transformer fault data has several characteristics. This application considers the relationship between each characteristic of the fault data, so each characteristic corresponding to the fault data can be classified into a set first. For example, a feature domain can be used to represent the set of values for all characteristics, where the feature domain V = [v1, v2, ..., v...]. mFor transformers, the characteristic domain V here may be [model, production time, manufacturer, place of origin]. If the characteristic v1 is the place of origin, then the characteristic v1 may be {Shandong, Beijing, Xi'an, ...}, that is, all values of the place of origin.
[0045] Optionally, the step of sequentially collecting feature values from the value set of each feature of the fault data, and obtaining temporary synthetic data based on the feature values, includes:
[0046] Feature values are randomly collected sequentially from the value set of each feature of the fault data;
[0047] A composite data is obtained based on a feature value corresponding to each feature of the fault data.
[0048] Based on the synthesized data, temporary synthesized data is obtained.
[0049] Randomly collecting feature values ensures that the final generated fault data samples are random, rather than directly copying the original fault data, thus expanding the range of fault data samples.
[0050] Specifically, feature values collected from the value set of each feature are arranged in the order of all features to obtain a synthetic data. By combining all the synthetic data, temporary synthetic data can be obtained. The specific number of temporary synthetic data can be adjusted according to the actual situation, and this application embodiment does not limit it.
[0051] For example, given v1 = {v 11 ,v 12 ,…,v 1n The first step is to perform permutation sampling on a value from set v1, for the temporary synthetic data sample y. (i) The sampled value is the value of the first feature. Then, the above sampling process can be applied to obtain the second feature value of the temporary synthetic data sample, up to the m-th feature.
[0052] Figure 2 The process of randomly collecting feature values is demonstrated. It is assumed that all 10 features have 7 possible values. Once the feature sampling process is complete, a temporary data sample is obtained. This application does not replicate the original data to generate synthetic data. Instead, it uses a random sampling method, randomly sampling one feature value from each feature set, arranging all feature values according to feature order to obtain a synthetic data set. Combining all synthetic data sets yields temporary synthetic data; that is, based on all the synthetic data, temporary synthetic data is obtained. This random sampling process increases the diversity of the synthetic data.
[0053] S2, input the temporary synthetic data into the feature model to obtain the predicted fault data output by the feature model, wherein the feature model is trained based on the set of values of other features of the fault data, using one feature of the fault data as a label.
[0054] The temporary synthetic data is input into the feature model to obtain the predicted fault data output by the feature model, thus ensuring the rationality of the predicted fault data. Before inputting the temporary synthetic data into the feature model, it can also be determined whether there is any unusable data in the temporary synthetic data. If so, the unusable data is removed to ensure the reliability of the predicted fault data.
[0055] Optionally, the number of feature models corresponds to the number of features in the fault data. This application considers all features of transformer fault data and generates a corresponding feature model for each feature, which can further ensure the diversity of generated samples. The feature model corresponding to each feature can be different, thereby enabling more targeted model training and data prediction.
[0056] Optionally, inputting the temporary synthetic data into the feature model includes:
[0057] The target features of the temporary synthetic data are determined, and the target features of the temporary synthetic data are the labels of the feature model during the training process.
[0058] Input the feature values corresponding to the other features of the temporary synthetic data besides the target feature into the feature model.
[0059] Specifically, the types of features in the fault data and the temporary synthetic data are the same. To determine which feature of the fault data serves as the label during the training process of the feature model, i.e., to determine which feature the feature model corresponds to, it is necessary to input the value set of other features besides the first one into the feature model to predict the predicted value of that feature. Because the feature model is trained based on the correlation between features, the predicted fault data output by the feature model can also reflect this correlation, which helps to improve the realism and reasonableness of the generated new fault data samples.
[0060] The feature model also considers the implicit relationships between features during training. An example of its training process is as follows: Let x be the data samples of transformer fault data. (i) Use one of the features (represented as) ) is used as the data label, while the remaining features (represented as The elements of the new feature vector are used to form new training samples. Following this scheme, there can be n training samples, each located in a region of size R. m-1In the space, n is the number of fault data samples, and m is the number of features. Let y be the temporary synthetic data sample. (i) Using the learned feature model and input It can be predicted Let j be the j-th feature of the final synthesized data sample.
[0061] For each feature A model can be trained. Therefore, m feature models can be obtained by following the same process. Assume there are 10 features, with feature values F1, ..., F10. For example, for a transformer, the transformer model, production time, manufacturer, etc., are all features. Furthermore, describing a transformer may require more than 10 features; generally, the number of feature models needed corresponds to the number of features.
[0062] As shown in Table 1, the label for Model 1 is feature F1, while features F2-F10 are used for training. Following the same principle, the label for Model 2 is F2, and features F1 and the remaining features F3-F10 are used for training, and so on. The purpose of this step is to capture the relationships between features in the minority class samples through modeling, and then use the learned feature model to output predicted fault data.
[0063] Table 1. Comparison of Models, Labels, and Features
[0064] Model Label feature Model 1 F1 F2-F10 Model 2 F2 F1, F3-F10 Model 3 F3 F1-F2, F4-F10 Model 4 F4 F1-F3, F5-F10 Model 5 F5 F1-F4, F6-F10 Model 6 F6 F1-F5, F7-F10 Model 7 F7 F1-F6, F8-F10 Model 8 F8 F1-F7, F9-F10 Model 9 F9 F1-F8, F10 Model 10 F10 F1-F9
[0065] Optionally, the feature model can be validated to determine its usability. If it is not usable, the feature model needs to be retrained to ensure the reliability of the predicted fault data.
[0066] S3. Generate new fault data samples based on the predicted fault data. The predicted fault data consists of the output data from each feature model, and further processing is required to generate new fault data samples. Each feature model outputs one predicted fault data point; classifying and aggregating all the predicted fault data yields a new fault data sample, for example, represented in the form of a feature domain. The fault data sample can simulate the feature relationships of real-world transformer fault data.
[0067] The process of generating new fault data samples is as follows Figure 3As shown, taking the generation of 6 transformer fault category data as an example, assuming the transformer has 10 features. Model 1 generates the first feature of the new data, Model 2 generates the second feature, and so on, until Model 10 generates the tenth feature. Once all models 1-10 have run, the complete 6 transformer fault category data are obtained. In other words, the first feature F1 of the final synthesized 6 data is obtained from the prediction of Model 1, and the input of Model 1 is the features of the temporary sampled data other than F1, i.e., F2-F10. Then, the remaining features of the final synthesized data can be obtained using the same process. For example, the second feature F2 of the final newly synthesized 6 data is obtained from the prediction of Model 2, and its input is the features of the temporary sampled data other than F2, i.e., F1 and F3-F10. This process continues until Model 10 is fully trained.
[0068] Based on the above embodiments, as an optional embodiment, the number of temporary synthetic data and the number of fault data samples are determined based on the number of fault data samples and the oversampling rate of the fault data. Determining the number of temporary synthetic data and the number of fault data samples based on the number of fault data samples and the oversampling rate of the fault data ensures the reliability of the fault data samples.
[0069] The steps for generating the fault data sample in the embodiments of this application are illustrated below using pseudocode. Let the fault data be D. min Oversampling rate p, possible feature values of the minority class V = [v1, v2, ..., v m The total number of iterations T in the repeated generation process. The temporary synthetic dataset Y and the synthetic fault dataset S are initialized to a size of n. syn An empty matrix of size ×m. Next, m feature models are trained for m features of the transformer, where x... j For the label, x -j The feature is used to train model j (1≤j≤m).
[0070] The pseudocode for the algorithm that generates new examples of transformer fault data is shown below:
[0071]
[0072]
[0073] In addition to the above embodiments, as an optional embodiment, it further includes:
[0074] Obtain non-fault data of the transformer.
[0075] Based on the non-fault data, the fault data, and the fault data samples, a transformer fault diagnosis model is trained using a machine learning algorithm. The transformer fault diagnosis model is used to predict the category of new transformer data and map the new transformer data as fault-type data or non-fault-type data.
[0076] Once the transformer fault data and non-fault data reach equilibrium, machine learning methods designed for handling balanced data can be used to train the data, such as neural networks, support vector machines, Bayesian classification, and deep learning. The transformer fault data includes both the original fault data and newly generated fault data samples.
[0077] Transformer fault diagnosis models can be used in transformer monitoring systems to identify faults in a timely manner.
[0078] Optionally, a transformer fault data information database is established based on the fault-type data and the non-fault-type data. If the received transformer data matches the transformer fault data information database, an early warning is issued.
[0079] When new data is generated, the classifier trained using the algorithm proposed in this application can predict the category of the data, mapping the new data item to either fault-related or non-fault-related data. This allows for the establishment of a relevant transformer fault data database. When received information matches data in the database, timely warnings can be issued, effectively reducing capital investment and property losses.
[0080] Optionally, based on the fault data and the non-fault data, a transformer early warning database and early warning rules are established. The transformer early warning database and early warning rules are used to compare with the information transmitted in real time by the base station transformer, which can realize the rapid location of transformer faults and nip the harm caused by transformer faults in the bud, thus realizing the identification and location of transformer fault information.
[0081] The transformer fault data sample generation apparatus provided in the embodiments of this application will be described below. The transformer fault data sample generation apparatus described below and the transformer fault data sample generation method described above can be referred to in correspondence.
[0082] Figure 4 This is a schematic diagram of the structure of the transformer fault data sample generation device provided in the embodiments of this application. See also... Figure 4 This application provides an apparatus for generating transformer fault data samples, comprising:
[0083] The data synthesis module 410 is used to acquire transformer fault data, sequentially collect feature values from the value set of each feature of the fault data, and obtain temporary synthesized data based on the feature values.
[0084] The data prediction module 420 is used to input the temporary synthetic data into the feature model to obtain the predicted fault data output by the feature model, wherein the feature model is trained based on the set of values of other features of the fault data, using one feature of the fault data as a label.
[0085] The sample generation module 430 is used to generate new fault data samples based on the predicted fault data.
[0086] The transformer fault data sample generation device provided in this application embodiment can generate new fault data samples that conform to the feature relationships by considering the implicit relationships between transformer features and by establishing and learning a model. The fault data samples can be used in the field of transformer fault identification.
[0087] In one embodiment, the data synthesis module 410 is specifically used for:
[0088] A feature value is randomly selected from the set of values for each feature of the fault data in sequence;
[0089] Based on a feature value corresponding to each feature of the fault data, a synthetic data is obtained;
[0090] Based on the synthesized data, temporary synthesized data is obtained, ensuring the randomness of feature value collection and expanding the range of samples.
[0091] In one embodiment, the data prediction module 420 is specifically used for:
[0092] The number of the feature models is controlled to correspond to the number of features in the fault data.
[0093] In one embodiment, the data prediction module 420 is specifically used for:
[0094] A feature of the temporary synthetic data is determined, wherein the feature of the temporary synthetic data is the label of the feature model during the training process.
[0095] Input the feature values corresponding to other features of the temporary synthetic data into the feature model.
[0096] In one embodiment, the data prediction module 420 is specifically used for:
[0097] The number of temporary synthetic data and the number of fault data samples are determined based on the number of fault data samples and the oversampling rate of the fault data.
[0098] In one embodiment, it also includes:
[0099] The fault diagnosis module is used to acquire non-fault data of the transformer; based on the non-fault data, the fault data, and the fault data samples, a transformer fault diagnosis model is trained using a machine learning algorithm. The transformer fault diagnosis model is used to predict the category of new transformer data and map the new transformer data as fault-type data or non-fault-type data.
[0100] In one embodiment, it also includes:
[0101] The information database determination module is used to establish a transformer fault data information database based on the fault-type data and the non-fault-type data.
[0102] Figure 5 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 5 As shown, the electronic device may include a processor 510, a communication interface 520, a memory 530, and a communication bus 540, wherein the processor 510, the communication interface 520, and the memory 530 communicate with each other via the communication bus 540. The processor 510 can call the computer program in the memory 530 to execute the steps of the method for generating transformer fault data samples.
[0103] Furthermore, the logical instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, 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 a 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 several 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 described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0104] On the other hand, this application also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can perform the steps of the transformer fault data sample generation method provided in the above embodiments.
[0105] On the other hand, embodiments of this application also provide a processor-readable storage medium storing a computer program for causing a processor to perform the steps of the methods provided in the above embodiments.
[0106] The processor-readable storage medium can be any available medium or data storage device that the processor can access, including but not limited to magnetic memory (e.g., floppy disk, hard disk, magnetic tape, magneto-optical disk (MO)), optical memory (e.g., CD, DVD, BD, HVD), and semiconductor memory (e.g., ROM, EPROM, EEPROM, non-volatile memory (NAND FLASH), solid-state drive (SSD)).
[0107] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0108] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.
[0109] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.
Claims
1. A method for generating transformer fault data samples, characterized in that, include: Obtain transformer fault data, sequentially collect feature values from the value set of each feature of the fault data, and obtain temporary synthetic data based on the feature values; Input the temporary synthetic data into the feature model to obtain the predicted fault data output by the feature model, wherein the feature model is trained based on the set of values of other features of the fault data, using one feature of the fault data as a label. New fault data samples are generated based on the predicted fault data; The process of sequentially collecting feature values from the value set of each feature of the fault data, and obtaining temporary synthetic data based on the feature values, includes: A feature value is randomly selected from the set of values for each feature of the fault data in sequence; Based on a feature value corresponding to each feature of the fault data, a synthetic data is obtained; Based on the synthesized data, temporary synthesized data is obtained; The input of the temporary synthetic data into the feature model includes: The target features of the temporary synthetic data are determined, and the target features of the temporary synthetic data are the labels of the feature model during the training process; Input the feature values corresponding to the other features of the temporary synthetic data besides the target feature into the feature model.
2. The method for generating transformer fault data samples according to claim 1, characterized in that, The number of feature models corresponds to the number of features in the fault data.
3. The method for generating transformer fault data samples according to claim 1, characterized in that, The number of temporary synthetic data and the number of fault data samples are determined based on the number of fault data samples and the oversampling rate of the fault data.
4. The method for generating transformer fault data samples according to claim 1, characterized in that, Also includes: Obtain non-fault data of the transformer; Based on the non-fault data, the fault data, and the fault data samples, a transformer fault diagnosis model is trained using a machine learning algorithm. The transformer fault diagnosis model is used to predict the category of new transformer data and map the new transformer data as fault-type data or non-fault-type data.
5. The method for generating transformer fault data samples according to claim 4, characterized in that, Also includes: A transformer fault data information database is established based on the fault-type data and the non-fault-type data.
6. A device for generating transformer fault data samples, characterized in that, include: The data synthesis module is used to acquire transformer fault data, sequentially collect feature values from the value set of each feature of the fault data, and obtain temporary synthesized data based on the feature values; The data prediction module is used to input the temporary synthetic data into the feature model to obtain the predicted fault data output by the feature model, wherein the feature model is trained based on the set of values of other features of the fault data, using one feature of the fault data as a label. The sample generation module is used to generate new fault data samples based on the predicted fault data; The process of sequentially collecting feature values from the value set of each feature of the fault data, and obtaining temporary synthetic data based on the feature values, includes: A feature value is randomly selected from the set of values for each feature of the fault data in sequence; Based on a feature value corresponding to each feature of the fault data, a synthetic data is obtained; Based on the synthesized data, temporary synthesized data is obtained; The input of the temporary synthetic data into the feature model includes: The target features of the temporary synthetic data are determined, and the target features of the temporary synthetic data are the labels of the feature model during the training process; Input the feature values corresponding to the other features of the temporary synthetic data besides the target feature into the feature model.
7. An electronic device comprising a processor and a memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method for generating transformer fault data samples according to any one of claims 1 to 5.
8. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the method for generating transformer fault data samples according to any one of claims 1 to 5.