Transformer fault diagnosis methods, devices, and systems

Abstract

This application provides a method, apparatus, and system for fault diagnosis of transformers. The method includes: acquiring vibration data of the transformer, wherein the vibration data is data collected through vibration monitoring of the transformer; inputting the vibration data into a vibration feature extraction model to obtain vibration feature vectors, wherein the vibration feature extraction model is used to reduce the dimensionality of the vectors composed of vibration data; sending the vibration feature vectors to a fault diagnosis platform; controlling the fault diagnosis platform to input the vibration feature vectors into a vibration data reconstruction model to obtain reconstructed data, wherein the vibration data reconstruction model is used to increase the dimensionality of the vectors composed of vibration data, and the data in the reconstructed data corresponds one-to-one with the data in the vibration data; and controlling the fault diagnosis platform to analyze the reconstructed data to obtain fault diagnosis results. This method solves the problem of inaccurate fault diagnosis caused by the loss of some data features of vibration data in the prior art.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and more specifically, to a method, apparatus, computer-readable storage medium, processor, and system for diagnosing transformer faults. Background Technology

[0002] Vibration monitoring is a common method for monitoring the operating status of transformers, and it has a good response to abnormal conditions of transformer windings and cores. Typically, transformer vibration monitoring involves deploying multiple vibration sensors on the surface of the transformer tank, with sampling frequencies in the thousands of hertz. Taking a 3200Hz sampling frequency and deploying six three-component vibration sensors as an example, 57,600 data points are collected per second. Each sample is encoded with 4 bytes, resulting in a data acquisition rate of approximately 230 kilobytes per second, generating about 20 gigabytes of data per day. If continuous acquisition is performed over many years, the monitoring system's data communication and storage will face significant pressure. Therefore, currently commonly used vibration monitoring systems typically collect data for a few seconds every few minutes or tens of minutes. This monitoring method is not true real-time online monitoring and will miss the detection of many sudden or intermittent abnormal conditions. Traditional vibration analysis methods usually calculate the mean, root mean square value (intensity), variance, skewness, kurtosis, peak-to-peak value, and other detection indicators for a short segment of vibration waveform data. Then, anomalies are detected according to preset threshold ranges for each indicator. Finally, spectral analysis methods are used for expert analysis to diagnose or locate possible fault types. Therefore, monitoring systems based on traditional vibration analysis, after calculating commonly used statistics and performing spectral analysis of the target frequency band, ultimately retain only some indicators useful in the traditional analysis, while the original data is discarded. Even when the original data is retained, it is usually only data from the most recent few weeks or months, with longer historical data being overwritten by newly acquired data. Because the analytical indicators of traditional vibration analysis methods are manually created by experts, derived from physical mechanism analysis, have clear physical meanings, and are easy to use, they have played an important role in machine fault diagnosis. With the development of big data and artificial intelligence technologies, another completely data-driven machine fault diagnosis method has received extensive research. These methods often start from the original data and extract fault features from the vibration waveform through neural networks; therefore, they must have the original data or data in a form containing the same characteristics as the original data.

[0003] To alleviate the pressure on data communication and storage in monitoring systems, existing technologies use a method of intermittently collecting and transmitting small segments of data, which results in incomplete real-time coverage of detection and missed detections. Alternatively, only traditional vibration analysis index data is retained, which means that the latest artificial intelligence modeling technology can only use existing index features for simple statistical classification, and cannot automatically extract data features from the data that can better achieve fault diagnosis through more advanced modeling techniques.

[0004] The information disclosed above in the background section is only intended to enhance the understanding of the background art of the art described herein. Therefore, the background art may contain certain information that does not constitute prior art known to those skilled in the art in this country. Summary of the Invention

[0005] The main objective of this application is to provide a method, apparatus, computer-readable storage medium, processor, and system for diagnosing transformer faults, in order to solve the problem of inaccurate fault diagnosis caused by the loss of some data features of vibration data in the prior art.

[0006] According to one aspect of the present invention, a fault diagnosis method for a transformer is provided, comprising: acquiring vibration data of the transformer, wherein the vibration data is data collected by vibration monitoring of the transformer; inputting the vibration data into a vibration feature extraction model to obtain a vibration feature vector, wherein the vibration feature extraction model is used to reduce the dimensionality of the vector composed of the vibration data; sending the vibration feature vector to a fault diagnosis platform; controlling the fault diagnosis platform to input the vibration feature vector into a vibration data reconstruction model to obtain reconstructed data, wherein the vibration data reconstruction model is used to increase the dimensionality of the vector composed of the vibration data, wherein the data in the reconstructed data corresponds one-to-one with the data in the vibration data; and controlling the fault diagnosis platform to analyze the reconstructed data to obtain a fault diagnosis result.

[0007] Optionally, the vibration feature extraction model and the vibration data reconstruction model constitute a neural network. The training process of the neural network includes: inputting historical vibration data into the neural network for training to obtain training reconstruction data, wherein the data in the training reconstruction data corresponds one-to-one with the data in the historical vibration data; calculating the sum of squares of the differences between all first data and the corresponding second data to obtain target parameters, wherein the first data are the data of the training reconstruction function, and the second data are the data of the historical vibration data; and the neural network training is completed when the target parameters are less than a predetermined threshold.

[0008] Optionally, the number of nodes in the input and output layers of the neural network is equal to the number of vibration data in one frame, and the number of vibration data in one frame is the product of the frame length and the sampling rate of the vibration data.

[0009] Optionally, the predetermined threshold is the product of the sum of squares of all the first data and 0.000001.

[0010] Optionally, the neural network includes one input layer, five hidden layers, and one output layer. The five hidden layers include a first hidden layer, a second hidden layer, a third hidden layer, a fourth hidden layer, and a fifth hidden layer. The input layer, the first hidden layer, the second hidden layer, and the third hidden layer constitute the vibration feature extraction model, and the third hidden layer, the fourth hidden layer, the fifth hidden layer, and the output layer constitute the vibration data reconstruction model.

[0011] According to another aspect of the present invention, a transformer fault diagnosis device is also provided, comprising: an acquisition unit for acquiring vibration data of the transformer, wherein the vibration data is data collected by vibration monitoring of the transformer; a first processing unit for inputting the vibration data into a vibration feature extraction model to obtain a vibration feature vector, wherein the vibration feature extraction model is used to reduce the dimensionality of the vector composed of the vibration data; a sending unit for sending the vibration feature vector to a fault diagnosis platform; a second processing unit for controlling the fault diagnosis platform to input the vibration feature vector into a vibration data reconstruction model to obtain reconstructed data, wherein the vibration data reconstruction model is used to increase the dimensionality of the vector composed of the vibration data, wherein the data in the reconstructed data corresponds one-to-one with the data in the vibration data; and an analysis unit for controlling the fault diagnosis platform to analyze the reconstructed data to obtain a fault diagnosis result.

[0012] According to another aspect of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored program, wherein the program executes any one of the methods described.

[0013] According to another aspect of the present invention, a processor is also provided, the processor being configured to run a program, wherein the program, when running, executes any of the methods described.

[0014] According to another aspect of the present invention, a fault diagnosis system for a transformer is also provided, comprising: a data acquisition unit, configured to collect vibration data by monitoring the transformer, input the vibration data into a vibration feature extraction model to obtain a vibration feature vector, and send the vibration feature vector, wherein the vibration feature extraction model is used to reduce the dimensionality of the vector composed of the vibration data; and a fault diagnosis platform, configured to receive the vibration feature vector, input the vibration feature vector into a vibration data reconstruction model to obtain reconstructed data, analyze the reconstructed data to obtain a fault diagnosis result, wherein the vibration data reconstruction model is used to increase the dimensionality of the vector composed of the vibration data, and the data in the reconstructed data corresponds one-to-one with the data in the vibration data.

[0015] Optionally, the collector has edge computing capabilities.

[0016] In this embodiment of the invention, the above-mentioned transformer fault diagnosis method firstly acquires vibration data of the transformer, which is data collected by vibration monitoring of the transformer; then, the vibration data is input into a vibration feature extraction model to obtain a vibration feature vector, which is used to reduce the dimensionality of the vector composed of the vibration data; then, the vibration feature vector is sent to a fault diagnosis platform; then, the fault diagnosis platform is controlled to input the vibration feature vector into a vibration data reconstruction model to obtain reconstructed data, which is used to increase the dimensionality of the vector composed of the vibration data, and the data in the reconstructed data corresponds one-to-one with the data in the vibration data; finally, the fault diagnosis platform is controlled to analyze the reconstructed data to obtain a fault diagnosis result. This method reduces the dimensionality of the vector composed of the vibration data using a vibration feature extraction model to obtain a vibration feature vector. This makes the dimension of the vibration feature vector smaller than that of the vector composed of the vibration data, thus reducing the amount of data, communication data sent to the fault diagnosis platform, and storage data of the fault diagnosis platform. Subsequently, the vibration feature vector is input into the vibration data reconstruction model to obtain reconstructed data. The reconstructed data can then be analyzed to obtain the fault diagnosis result. The reconstructed data consists of vibration waveform data that is highly similar to the vibration data, eliminating the need to lose some data features to reduce communication and storage data. This improves the accuracy of fault diagnosis and solves the problem of inaccurate fault diagnosis caused by the loss of some data features of vibration data in the prior art. Attached Figure Description

[0017] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0018] Figure 1 A flowchart of a transformer fault diagnosis method according to an embodiment of this application is shown;

[0019] Figure 2 A schematic diagram of a transformer fault diagnosis device according to an embodiment of this application is shown. Detailed Implementation

[0020] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0021] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0022] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this application described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0023] It should be understood that when an element (such as a layer, film, region, or substrate) is described as being "on" another element, the element may be directly on the other element, or there may be an intermediate element present. Furthermore, in the specification and claims, when an element is described as being "connected" to another element, the element may be "directly connected" to the other element, or "connected" to the other element via a third element.

[0024] As mentioned in the background section, the loss of some data features in vibration data in the prior art leads to inaccurate fault diagnosis. In order to solve the above problems, in a typical embodiment of this application, a fault diagnosis method, device, computer-readable storage medium, processor and fault diagnosis system for transformers are provided.

[0025] According to an embodiment of this application, a method for diagnosing transformer faults is provided.

[0026] Figure 1 This is a flowchart of a transformer fault diagnosis method according to an embodiment of this application. Figure 1 As shown, the method includes the following steps:

[0027] Step S101: Obtain vibration data of the transformer. The vibration data is the data collected by vibration monitoring of the transformer.

[0028] Step S102: Input the above vibration data into the vibration feature extraction model to obtain the vibration feature vector. The vibration feature extraction model is used to reduce the dimensionality of the vector composed of the above vibration data.

[0029] Step S103: Send the above vibration feature vector to the fault diagnosis platform;

[0030] Step S104: Control the above fault diagnosis platform to input the above vibration feature vector into the vibration data reconstruction model to obtain the reconstruction data. The above vibration data reconstruction model is used to increase the dimension of the vector composed of the above vibration data. The data of the above reconstruction data corresponds one-to-one with the data of the above vibration data.

[0031] Step S105: Control the above-mentioned fault diagnosis platform to analyze the above-mentioned reconstructed data to obtain fault diagnosis results.

[0032] In the aforementioned transformer fault diagnosis method, firstly, vibration data of the transformer is acquired, which is data collected through vibration monitoring of the transformer; then, the vibration data is input into a vibration feature extraction model to obtain vibration feature vectors, which are used to reduce the dimensionality of the vectors composed of the vibration data; subsequently, the vibration feature vectors are sent to a fault diagnosis platform; then, the fault diagnosis platform is controlled to input the vibration feature vectors into a vibration data reconstruction model to obtain reconstructed data, which is used to increase the dimensionality of the vectors composed of the vibration data, and the data in the reconstructed data corresponds one-to-one with the data in the original vibration data; finally, the fault diagnosis platform is controlled to analyze the reconstructed data to obtain fault diagnosis results. This method reduces the dimensionality of the vector composed of the vibration data using a vibration feature extraction model to obtain a vibration feature vector. This makes the dimension of the vibration feature vector smaller than that of the vector composed of the vibration data, thus reducing the amount of data, communication data sent to the fault diagnosis platform, and storage data of the fault diagnosis platform. Subsequently, the vibration feature vector is input into the vibration data reconstruction model to obtain reconstructed data. The reconstructed data can then be analyzed to obtain the fault diagnosis result. The reconstructed data consists of vibration waveform data that is highly similar to the vibration data, eliminating the need to lose some data features to reduce communication and storage data. This improves the accuracy of fault diagnosis and solves the problem of inaccurate fault diagnosis caused by the loss of some data features of vibration data in the prior art.

[0033] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0034] In one embodiment of this application, the vibration feature extraction model and the vibration data reconstruction model constitute a neural network. The training process of the neural network includes: inputting historical vibration data into the neural network for training to obtain training reconstruction data, wherein the data in the training reconstruction data corresponds one-to-one with the data in the historical vibration data; calculating the sum of squares of the differences between all first data and the corresponding second data to obtain target parameters, wherein the first data are the data of the training reconstruction function, and the second data are the data in the historical vibration data; and when the target parameters are less than a predetermined threshold, the neural network training is completed. Specifically, the historical vibration data D... obs The above neural network is input for training, and the training reconstruction data D is obtained. rec The above training and reconstruction data D obs The data is consistent with the aforementioned historical vibration data D rec The data corresponds one-to-one, and the above training and reconstruction data D obs Including a1, a2...aN, the above historical vibration data D rec Including b1, b2...bN, the formula for calculating the target parameter C is C=(a1-b1) 2 +(a²-b²) 2 +……(aN-bN) 2 By iteratively training a neural network model with a large number of data samples and adjusting the connection weights of the neural network to minimize the target parameters, the neural network training is completed when the target parameters are less than a predetermined threshold, ensuring that the reconstructed data has a high similarity to the vibration data, and further improving the accuracy of fault diagnosis.

[0035] In one embodiment of this application, the number of nodes in the input and output layers of the neural network is equal to the number of vibration data in one frame, and the number of vibration data in one frame is the product of the frame length and the sampling rate of the vibration data. Specifically, the frame length T can be adjusted as needed, and its value range can be from 0.1 to 1. The number of vibration data in one frame, Ns = T * sr, is the sampling rate of the vibration data. Taking sr = 3200Hz and T = 0.5 as an example, Ns = 1600 is calculated.

[0036] In one embodiment of this application, the predetermined threshold is the product of the sum of squares of all the first data and 0.000001. Specifically, the training and reconstruction data D... obs Including a1, a2...aN, the predetermined threshold ξ = 0.000001[(a1)] 2 +(a2) 2 +……(aN) 2 Of course, the above-mentioned predetermined thresholds are not limited to these and can be adjusted according to the actual situation.

[0037] In one embodiment of this application, the neural network includes one input layer, five hidden layers, and one output layer. The five hidden layers include a first hidden layer, a second hidden layer, a third hidden layer, a fourth hidden layer, and a fifth hidden layer. The input layer, the first hidden layer, the second hidden layer, and the third hidden layer constitute the vibration feature extraction model, and the third hidden layer, the fourth hidden layer, the fifth hidden layer, and the output layer constitute the vibration data reconstruction model. Specifically, taking 1600 sampling points as an input to the five hidden layers as an example, the number of network nodes in each layer is as follows: input layer: 1600 nodes, first hidden layer: 512 nodes, second hidden layer: 256 nodes, third hidden layer: 128 nodes, fourth hidden layer: 256 nodes, fifth hidden layer: 512 nodes, output layer: 1600 nodes.

[0038] This application also provides a transformer fault diagnosis device. It should be noted that the transformer fault diagnosis device of this application can be used to execute the transformer fault diagnosis method provided in this application. The following describes the transformer fault diagnosis device provided in this application.

[0039] Figure 2 This is a schematic diagram of a transformer fault diagnosis device according to an embodiment of this application. Figure 2 As shown, the device includes:

[0040] The acquisition unit 10 is used to acquire vibration data of the transformer, which is data collected by vibration monitoring of the transformer.

[0041] The first processing unit 20 is used to input the above vibration data into the vibration feature extraction model to obtain a vibration feature vector. The vibration feature extraction model is used to reduce the dimensionality of the vector composed of the above vibration data.

[0042] The transmitting unit 30 is used to transmit the above-mentioned vibration feature vector to the fault diagnosis platform;

[0043] The second processing unit 40 is used to control the fault diagnosis platform to input the vibration feature vector into the vibration data reconstruction model to obtain reconstruction data. The vibration data reconstruction model is used to increase the dimension of the vector composed of the vibration data. The data of the reconstruction data corresponds one-to-one with the data of the vibration data.

[0044] The analysis unit 50 is used to control the fault diagnosis platform to analyze the reconstructed data to obtain fault diagnosis results.

[0045] In the aforementioned transformer fault diagnosis device, the acquisition unit acquires vibration data of the transformer, which is data collected through vibration monitoring of the transformer; the first processing unit inputs the vibration data into a vibration feature extraction model to obtain vibration feature vectors, which are used to reduce the dimensionality of the vectors composed of the vibration data; the sending unit sends the vibration feature vectors to the fault diagnosis platform; the second processing unit controls the fault diagnosis platform to input the vibration feature vectors into a vibration data reconstruction model to obtain reconstructed data, which is used to increase the dimensionality of the vectors composed of the vibration data, and the data in the reconstructed data corresponds one-to-one with the data in the vibration data; the analysis unit controls the fault diagnosis platform to analyze the reconstructed data to obtain fault diagnosis results. This device reduces the dimensionality of the vector composed of the vibration data using a vibration feature extraction model to obtain a vibration feature vector. This makes the dimension of the vibration feature vector smaller than that of the vector composed of the vibration data, thus reducing the amount of data and the amount of communication data sent to the fault diagnosis platform and the amount of data stored on the fault diagnosis platform. Subsequently, the vibration feature vector is input into the vibration data reconstruction model to obtain reconstructed data. The reconstructed data can then be analyzed to obtain the fault diagnosis result. The reconstructed data consists of vibration waveform data that is highly similar to the vibration data, eliminating the need to lose some data features to reduce the amount of communication and storage data. This improves the accuracy of fault diagnosis and solves the problem of inaccurate fault diagnosis caused by the loss of some data features of vibration data in the prior art.

[0046] In one embodiment of this application, the vibration feature extraction model and the vibration data reconstruction model constitute a neural network. The device further includes a training unit, which comprises a training module, a calculation module, and a determination module. The training module is used to input historical vibration data into the neural network for training to obtain training reconstruction data, where the data in the training reconstruction data corresponds one-to-one with the data in the historical vibration data. The calculation module is used to calculate the sum of squares of the differences between all first data and the corresponding second data to obtain a target parameter. The first data is the data of the training reconstruction function, and the second data is the data in the historical vibration data. The determination module is used to complete the training of the neural network when the target parameter is less than a predetermined threshold. Specifically, the historical vibration data D... obs The above neural network is input for training, and the training reconstruction data D is obtained. rec The above training and reconstruction data D obs The data is consistent with the aforementioned historical vibration data D rec The data corresponds one-to-one, and the above training and reconstruction data D obs Including a1, a2...aN, the above historical vibration data D rec Including b1, b2...bN, the formula for calculating the target parameter C is C=(a1-b1)2 +(a²-b²) 2 +……(aN-bN) 2 By iteratively training a neural network model with a large number of data samples and adjusting the connection weights of the neural network to minimize the target parameters, the neural network training is completed when the target parameters are less than a predetermined threshold, ensuring that the reconstructed data has a high similarity to the vibration data, and further improving the accuracy of fault diagnosis.

[0047] In one embodiment of this application, the number of nodes in the input and output layers of the neural network is equal to the number of vibration data in one frame, and the number of vibration data in one frame is the product of the frame length and the sampling rate of the vibration data. Specifically, the frame length T can be adjusted as needed, and its value range can be from 0.1 to 1. The number of vibration data in one frame, Ns = T * sr, is the sampling rate of the vibration data. Taking sr = 3200Hz and T = 0.5 as an example, Ns = 1600 is calculated.

[0048] In one embodiment of this application, the predetermined threshold is the product of the sum of squares of all the first data and 0.000001. Specifically, the training and reconstruction data D... obs Including a1, a2...aN, the predetermined threshold ξ = 0.000001[(a1)] 2 +(a2) 2 +……(aN) 2 Of course, the above-mentioned predetermined thresholds are not limited to these and can be adjusted according to the actual situation.

[0049] In one embodiment of this application, the neural network includes one input layer, five hidden layers, and one output layer. The five hidden layers include a first hidden layer, a second hidden layer, a third hidden layer, a fourth hidden layer, and a fifth hidden layer. The input layer, the first hidden layer, the second hidden layer, and the third hidden layer constitute the vibration feature extraction model. The third hidden layer, the fourth hidden layer, the fifth hidden layer, and the output layer constitute the vibration data reconstruction model. Specifically, taking 1600 sampling points as an input to 5 hidden layers as an example, the number of network nodes in each layer is as follows: input layer: 1600 nodes, first hidden layer: 512 nodes, second hidden layer: 256 nodes, third hidden layer: 128 nodes, fourth hidden layer: 256 nodes, fifth hidden layer: 512 nodes, output layer: 1600 nodes. The above vibration feature extraction model reduces the vibration data to a 128-dimensional vector, achieving a data compression of 12.5 times. The 128-dimensional vector data of the third hidden layer of the neural network can be used by the above vibration data reconstruction model to reconstruct data that is highly similar to the vibration data.

[0050] This application also provides a transformer fault diagnosis system, including:

[0051] The data acquisition device is used to collect vibration data for vibration monitoring of the above-mentioned transformer, input the vibration data into the vibration feature extraction model to obtain the vibration feature vector, and send the vibration feature vector. The vibration feature extraction model is used to reduce the dimensionality of the vector composed of the vibration data.

[0052] The fault diagnosis platform is used to receive the above-mentioned vibration feature vectors, input the above-mentioned vibration feature vectors into the vibration data reconstruction model to obtain reconstructed data, analyze the above-mentioned reconstructed data to obtain fault diagnosis results, and the above-mentioned vibration data reconstruction model is used to increase the dimensionality of the vector composed of the above-mentioned vibration data, and the data of the above-mentioned reconstructed data corresponds one-to-one with the data of the above-mentioned vibration data.

[0053] In the aforementioned transformer fault diagnosis system, the data acquisition unit obtains vibration data of the transformer, which is data collected through vibration monitoring of the transformer. This vibration data is then input into a vibration feature extraction model to obtain vibration feature vectors. The vibration feature extraction model is used to reduce the dimensionality of the vectors composed of the vibration data. These vibration feature vectors are then sent to the fault diagnosis platform. The fault diagnosis platform inputs these vibration feature vectors into a vibration data reconstruction model to obtain reconstructed data. This vibration data reconstruction model is used to increase the dimensionality of the vectors composed of the vibration data. The data in the reconstructed data corresponds one-to-one with the data in the original vibration data. The fault diagnosis platform analyzes the reconstructed data to obtain the fault diagnosis results. This system reduces the dimensionality of the vector composed of the vibration data using a vibration feature extraction model, resulting in a vibration feature vector. This reduces the dimension of the vibration feature vector compared to the vector composed of the vibration data, thereby decreasing the amount of data transmitted to the fault diagnosis platform and the amount of data stored on the platform. Subsequently, the vibration feature vector is input into the vibration data reconstruction model to obtain reconstructed data. The reconstructed data can then be analyzed to obtain fault diagnosis results. The reconstructed data consists of vibration waveforms that are highly similar to the vibration data, eliminating the need to lose some data features to reduce communication and storage data volume. This improves the accuracy of fault diagnosis and solves the problem of inaccurate fault diagnosis caused by the loss of some data features in vibration data in existing technologies.

[0054] In one embodiment of this application, the aforementioned data collector has edge computing capabilities. Specifically, the data collector must have edge computing capabilities to deploy a vibration feature extraction model.

[0055] More specifically, the vibration feature extraction model is deployed on a data acquisition unit with edge computing capabilities. Each frame of vibration data collected by the vibration sensor consists of Ns sampling points. This data is input into the vibration feature extraction model deployed on the data acquisition unit, which outputs a 128-dimensional vector. This vector is then transmitted to the fault diagnosis cloud platform via a communication module. With only 128 dimensions compared to the 1600 dimensions of the vibration data, the amount of communication data is significantly reduced. After receiving the 128-dimensional vibration feature vector from the data acquisition unit, the fault diagnosis cloud platform can store all 128 dimensions, greatly reducing the amount of data stored. Directly using these 128-dimensional vibration features as input data for subsequent fault diagnosis models significantly reduces the computational burden on the cloud platform for extracting vibration features. Simultaneously, a vibration data reconstruction model consisting of the last four layers of a neural network is deployed on the cloud platform. This model can reconstruct vibration waveform data with very similar vibration features to the original vibration data, enabling subsequent algorithmic calculations that must be based on the vibration data.

[0056] The aforementioned transformer fault diagnosis device includes a processor and a memory. The aforementioned acquisition unit, first processing unit, transmission unit, second processing unit, and analysis unit are all stored in the memory as program units. The processor executes the aforementioned program units stored in the memory to achieve the corresponding functions.

[0057] The processor contains a kernel, which retrieves the corresponding program units from memory. One or more kernels can be configured, and adjusting kernel parameters can address the problem of inaccurate fault diagnosis caused by the loss of some vibration data features in existing technologies.

[0058] The memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0059] This invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the above-described method.

[0060] This invention provides a processor for running a program, wherein the program executes the method described above when it runs.

[0061] This invention provides a device including a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, it performs at least the following steps:

[0062] Step S101: Obtain vibration data of the transformer. The vibration data is the data collected by vibration monitoring of the transformer.

[0063] Step S102: Input the above vibration data into the vibration feature extraction model to obtain the vibration feature vector. The vibration feature extraction model is used to reduce the dimensionality of the vector composed of the above vibration data.

[0064] Step S103: Send the above vibration feature vector to the fault diagnosis platform;

[0065] Step S104: Control the above fault diagnosis platform to input the above vibration feature vector into the vibration data reconstruction model to obtain the reconstruction data. The above vibration data reconstruction model is used to increase the dimension of the vector composed of the above vibration data. The data of the above reconstruction data corresponds one-to-one with the data of the above vibration data.

[0066] Step S105: Control the above-mentioned fault diagnosis platform to analyze the above-mentioned reconstructed data to obtain fault diagnosis results.

[0067] The devices mentioned in this article can be servers, PCs, tablets, mobile phones, etc.

[0068] This application also provides a computer program product, which, when executed on a data processing device, is suitable for executing an initialization program having at least the following method steps:

[0069] Step S101: Obtain vibration data of the transformer. The vibration data is the data collected by vibration monitoring of the transformer.

[0070] Step S102: Input the above vibration data into the vibration feature extraction model to obtain the vibration feature vector. The vibration feature extraction model is used to reduce the dimensionality of the vector composed of the above vibration data.

[0071] Step S103: Send the above vibration feature vector to the fault diagnosis platform;

[0072] Step S104: Control the above fault diagnosis platform to input the above vibration feature vector into the vibration data reconstruction model to obtain the reconstruction data. The above vibration data reconstruction model is used to increase the dimension of the vector composed of the above vibration data. The data of the above reconstruction data corresponds one-to-one with the data of the above vibration data.

[0073] Step S105: Control the above-mentioned fault diagnosis platform to analyze the above-mentioned reconstructed data to obtain fault diagnosis results.

[0074] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0075] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units described above can be a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0076] The units described above 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 units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0077] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0078] If the aforementioned integrated units are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a computer-readable 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 the present invention. The aforementioned computer-readable storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0079] As can be seen from the above description, the embodiments of this application achieve the following technical effects:

[0080] 1) In the transformer fault diagnosis method of this application, firstly, vibration data of the transformer is acquired, wherein the vibration data is data collected by vibration monitoring of the transformer; then, the vibration data is input into a vibration feature extraction model to obtain a vibration feature vector, wherein the vibration feature extraction model is used to reduce the dimensionality of the vector composed of the vibration data; then, the vibration feature vector is sent to a fault diagnosis platform; then, the fault diagnosis platform is controlled to input the vibration feature vector into a vibration data reconstruction model to obtain reconstructed data, wherein the vibration data reconstruction model is used to increase the dimensionality of the vector composed of the vibration data, wherein the data of the reconstructed data corresponds one-to-one with the data of the vibration data; finally, the fault diagnosis platform is controlled to analyze the reconstructed data to obtain a fault diagnosis result. This method reduces the dimensionality of the vector composed of the vibration data using a vibration feature extraction model to obtain a vibration feature vector. This makes the dimension of the vibration feature vector smaller than that of the vector composed of the vibration data, thus reducing the amount of data, communication data sent to the fault diagnosis platform, and storage data of the fault diagnosis platform. Subsequently, the vibration feature vector is input into the vibration data reconstruction model to obtain reconstructed data. The reconstructed data can then be analyzed to obtain the fault diagnosis result. The reconstructed data consists of vibration waveform data that is highly similar to the vibration data, eliminating the need to lose some data features to reduce communication and storage data. This improves the accuracy of fault diagnosis and solves the problem of inaccurate fault diagnosis caused by the loss of some data features of vibration data in the prior art.

[0081] 2) In the transformer fault diagnosis device of this application, the acquisition unit acquires the vibration data of the transformer, the vibration data being the data collected by vibration monitoring of the transformer; the first processing unit inputs the vibration data into a vibration feature extraction model to obtain a vibration feature vector, the vibration feature extraction model being used to reduce the dimensionality of the vector composed of the vibration data; the sending unit sends the vibration feature vector to the fault diagnosis platform; the second processing unit controls the fault diagnosis platform to input the vibration feature vector into a vibration data reconstruction model to obtain reconstructed data, the vibration data reconstruction model being used to increase the dimensionality of the vector composed of the vibration data, the data of the reconstructed data corresponding one-to-one with the data of the vibration data; the analysis unit controls the fault diagnosis platform to analyze the reconstructed data to obtain fault diagnosis results. This device reduces the dimensionality of the vector composed of the vibration data using a vibration feature extraction model to obtain a vibration feature vector. This makes the dimension of the vibration feature vector smaller than that of the vector composed of the vibration data, thus reducing the amount of data and the amount of communication data sent to the fault diagnosis platform and the amount of data stored on the fault diagnosis platform. Subsequently, the vibration feature vector is input into the vibration data reconstruction model to obtain reconstructed data. The reconstructed data can then be analyzed to obtain the fault diagnosis result. The reconstructed data consists of vibration waveform data that is highly similar to the vibration data, eliminating the need to lose some data features to reduce the amount of communication and storage data. This improves the accuracy of fault diagnosis and solves the problem of inaccurate fault diagnosis caused by the loss of some data features of vibration data in the prior art.

[0082] 3) In the transformer fault diagnosis system of this application, the data acquisition unit acquires the vibration data of the transformer, which is the data collected by vibration monitoring of the transformer; the vibration data is input into the vibration feature extraction model to obtain the vibration feature vector, which is used to reduce the dimensionality of the vector composed of the vibration data; the vibration feature vector is sent to the fault diagnosis platform; the fault diagnosis platform inputs the vibration feature vector into the vibration data reconstruction model to obtain the reconstructed data, which is used to increase the dimensionality of the vector composed of the vibration data, and the data of the reconstructed data corresponds one-to-one with the data of the vibration data; the fault diagnosis platform analyzes the reconstructed data to obtain the fault diagnosis result. This system reduces the dimensionality of the vector composed of the vibration data using a vibration feature extraction model, resulting in a vibration feature vector. This reduces the dimension of the vibration feature vector compared to the vector composed of the vibration data, thereby decreasing the amount of data transmitted to the fault diagnosis platform and the amount of data stored on the platform. Subsequently, the vibration feature vector is input into the vibration data reconstruction model to obtain reconstructed data. The reconstructed data can then be analyzed to obtain fault diagnosis results. The reconstructed data consists of vibration waveforms that are highly similar to the vibration data, eliminating the need to lose some data features to reduce communication and storage data volume. This improves the accuracy of fault diagnosis and solves the problem of inaccurate fault diagnosis caused by the loss of some data features in vibration data in existing technologies.

[0083] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for diagnosing transformer faults, characterized in that, include: Obtain vibration data of the transformer, wherein the vibration data is data collected by vibration monitoring of the transformer; The vibration data is input into the vibration feature extraction model to obtain a vibration feature vector. The vibration feature extraction model is used to reduce the dimensionality of the vector composed of the vibration data. The vibration feature vector is sent to the fault diagnosis platform; The fault diagnosis platform is controlled to input the vibration feature vector into the vibration data reconstruction model to obtain reconstructed data. The vibration data reconstruction model is used to increase the dimensionality of the vector composed of the vibration data. The data in the reconstructed data corresponds one-to-one with the data in the vibration data. The fault diagnosis platform is controlled to analyze the reconstructed data to obtain fault diagnosis results. The vibration feature extraction model and the vibration data reconstruction model constitute a neural network. The training process of the neural network includes: inputting historical vibration data into the neural network for training to obtain training reconstructed data, wherein the data in the training reconstructed data corresponds one-to-one with the data in the historical vibration data; calculating the sum of squares of the differences between all first data and the corresponding second data to obtain target parameters, wherein the first data are the training reconstructed data and the second data are the data in the historical vibration data; and completing the training of the neural network when the target parameters are less than a predetermined threshold, wherein the predetermined threshold is the product of the sum of squares of all first data and 0.000001. The number of nodes in the input and output layers of the neural network is equal to the number of vibration data in one frame, and the number of vibration data in one frame is the product of the frame length and the sampling rate of the vibration data. The neural network includes one input layer, five hidden layers, and one output layer. The five hidden layers include a first hidden layer, a second hidden layer, a third hidden layer, a fourth hidden layer, and a fifth hidden layer. The input layer, the first hidden layer, the second hidden layer, and the third hidden layer constitute the vibration feature extraction model. The third hidden layer, the fourth hidden layer, the fifth hidden layer, and the output layer constitute the vibration data reconstruction model.

2. A fault diagnosis device for a transformer, characterized in that, include: The acquisition unit is used to acquire vibration data of the transformer, wherein the vibration data is data collected by vibration monitoring of the transformer; The first processing unit is used to input the vibration data into the vibration feature extraction model to obtain a vibration feature vector. The vibration feature extraction model is used to reduce the dimensionality of the vector composed of the vibration data. The transmitting unit is used to transmit the vibration feature vector to the fault diagnosis platform; The second processing unit is used to control the fault diagnosis platform to input the vibration feature vector into the vibration data reconstruction model to obtain reconstructed data. The vibration data reconstruction model is used to increase the dimension of the vector composed of the vibration data. The data of the reconstructed data corresponds one-to-one with the data of the vibration data. The analysis unit is used to control the fault diagnosis platform to analyze the reconstructed data to obtain fault diagnosis results; The vibration feature extraction model and the vibration data reconstruction model form a neural network. The device further includes a training unit, which comprises a training module, a calculation module, and a determination module. The training module inputs historical vibration data into the neural network for training to obtain training reconstructed data, which corresponds one-to-one with the historical vibration data. The calculation module calculates the sum of squares of the differences between all first data points and their corresponding second data points to obtain a target parameter. The first data points are the training reconstructed data, and the second data points are the historical vibration data. The determination module determines that the neural network training is complete when the target parameter is less than a predetermined threshold, where the predetermined threshold is the product of the sum of squares of all first data points and 0.000001. The number of nodes in the input and output layers of the neural network is equal to the number of vibration data in one frame, and the number of vibration data in one frame is the product of the frame length and the sampling rate of the vibration data. The neural network includes one input layer, five hidden layers, and one output layer. The five hidden layers include a first hidden layer, a second hidden layer, a third hidden layer, a fourth hidden layer, and a fifth hidden layer. The input layer, the first hidden layer, the second hidden layer, and the third hidden layer constitute the vibration feature extraction model. The third hidden layer, the fourth hidden layer, the fifth hidden layer, and the output layer constitute the vibration data reconstruction model.

3. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein the program performs the method of claim 1.

4. A processor, characterized in that, The processor is used to run a program, wherein the program executes the method of claim 1 when it runs.

5. A fault diagnosis system for a transformer, characterized in that, include: The data acquisition device is used to collect vibration data of the transformer, input the vibration data into the vibration feature extraction model to obtain the vibration feature vector, and send the vibration feature vector. The vibration feature extraction model is used to reduce the dimensionality of the vector composed of the vibration data. The fault diagnosis platform is used to receive the vibration feature vector, input the vibration feature vector into the vibration data reconstruction model to obtain reconstructed data, analyze the reconstructed data to obtain fault diagnosis results, and the vibration data reconstruction model is used to increase the dimensionality of the vector composed of the vibration data, and the data of the reconstructed data corresponds one-to-one with the data of the vibration data. The vibration feature extraction model and the vibration data reconstruction model constitute a neural network. The training process of the neural network includes: inputting historical vibration data into the neural network for training to obtain training reconstructed data, wherein the data in the training reconstructed data corresponds one-to-one with the data in the historical vibration data; calculating the sum of squares of the differences between all first data and the corresponding second data to obtain target parameters, wherein the first data are the training reconstructed data and the second data are the data in the historical vibration data; and completing the training of the neural network when the target parameters are less than a predetermined threshold, wherein the predetermined threshold is the product of the sum of squares of all first data and 0.000001. The number of nodes in the input and output layers of the neural network is equal to the number of vibration data in one frame, and the number of vibration data in one frame is the product of the frame length and the sampling rate of the vibration data. The neural network includes one input layer, five hidden layers, and one output layer. The five hidden layers include a first hidden layer, a second hidden layer, a third hidden layer, a fourth hidden layer, and a fifth hidden layer. The input layer, the first hidden layer, the second hidden layer, and the third hidden layer constitute the vibration feature extraction model. The third hidden layer, the fourth hidden layer, the fifth hidden layer, and the output layer constitute the vibration data reconstruction model.

6. The system according to claim 5, characterized in that, The data collector has edge computing capabilities.

Images