Transformer life prediction method, device, computer equipment and storage medium

Abstract

The present application relates to a transformer life prediction method, device, computer equipment, storage medium and computer program product. The method inputs the dynamic parameter information of the transformer to be predicted into a feature recognition model to identify the dynamic parameter characteristics of the transformer to be predicted, fuses the dynamic parameter characteristics with the static parameter information of the transformer to be predicted, and inputs them into the life prediction model to obtain the life prediction result of the transformer to be predicted. The above-mentioned transformer life prediction method, device, computer equipment, storage medium and computer program product can evaluate the current life status and remaining life length of the transformer to be predicted through the dynamic parameter information and static parameter information of the transformer to be predicted that affect the transformer life, and use the pre-trained feature recognition model and life prediction model, so as to provide reliable and accurate technical support for the condition maintenance of the transformer.

Description

Technical Field

[0001] The present application relates to the technical field of transformers, and in particular to a transformer life prediction method, device, computer equipment, storage medium and computer program product. Background Art

[0002] In recent years, with the rapid development of my country's economy, the social electricity demand has also shown an increasing trend year by year. As an important voltage conversion device in the power system, when the transformer fails, it will bring huge economic losses. Therefore, it is necessary to effectively maintain the transformer to improve the stability and economy of the power system operation. At present, the normal life of an oil-immersed transformer is about 20 to 25 years. In order to extend the life of the transformer to about 40 years, it is necessary to study the operating reliability and remaining life of the transformer. While improving the operating reliability of the transformer, it will also effectively increase the life of the transformer.

[0003] Traditional maintenance methods for oil-immersed transformers include post-maintenance and regular maintenance methods. The post-maintenance method is to inspect and repair the equipment after an accident occurs in the power equipment. This type of maintenance method is post-compensation in nature. With the improvement of the requirements for the operation of power equipment, it has gradually failed to meet the needs of operation. The regular maintenance method formulates a fixed maintenance cycle based on the operation of the equipment, and then shuts down the power equipment for maintenance according to this cycle. The regular maintenance method can effectively prevent the occurrence of power equipment accidents, but the mechanical execution of the pre-established maintenance cycle often cannot effectively consider the actual operation of the power equipment, and for the purpose of ensuring the safe operation of the power equipment as much as possible, the maintenance cycle established in advance will tend to be conservative; therefore, there may be ineffective shutdowns or excessive maintenance. In more serious cases, it may even cause frequent maintenance of power equipment, resulting in a significant shortening of its life. Summary of the invention

[0004] Based on this, it is necessary to provide a transformer life prediction method, device, computer equipment, computer readable storage medium and computer program product that can accurately and reliably predict the remaining life of the transformer in response to the above technical problems.

[0005] In a first aspect, the present application provides a transformer life prediction method, the method comprising:

[0006] Obtaining dynamic parameter information of the transformer to be predicted, where the dynamic parameter information is data of real-time operating parameters that affect the life of the transformer;

[0007] The dynamic parameter information is input into the feature recognition model to identify the dynamic parameter characteristics of the transformer to be predicted, and the feature recognition model is trained using the historical operation data of the transformer;

[0008] The dynamic parameter characteristics are fused with the static parameter information of the transformer to be predicted and then input into the life prediction model to obtain the life prediction result of the transformer to be predicted. The static parameter information is the data of the transformer attribute parameters that affect the life of the transformer. The life prediction model is trained using the historical operation data of the transformer.

[0009] In one of the embodiments, dynamic parameter information of the transformer to be predicted is obtained, including collecting original operating parameters that affect the life of the transformer during the real-time operation of the transformer to be predicted; using a denoising network to clean the collected original operating parameters to obtain dynamic parameter information, and the denoising network is trained based on a denoising autoencoder using historical operating data of the transformer.

[0010] In one of the embodiments, the dynamic parameter information is input into a feature recognition model to identify the dynamic parameter characteristics of the transformer to be predicted, including: performing discrete Fourier transform processing on the dynamic parameter information to obtain corresponding two-dimensional spectrum image information; inputting the two-dimensional spectrum image information corresponding to each dynamic parameter information into the feature recognition model; and performing inverse Fourier transform processing on the output result of the feature recognition model to obtain the dynamic parameter characteristics.

[0011] In one of the embodiments, the method further includes: using historical operation data corresponding to health factor 1 and historical operation data corresponding to health factor 0 in historical operation data of the transformer under different life states to train a support vector machine neural network, wherein the historical operation data corresponding to health factor 1 is the historical operation data of the transformer in a healthy state, and the historical operation data corresponding to health factor 0 is the historical operation data of the transformer in a failure state, and each group of historical operation data includes dynamic parameter information and static parameter information when the transformer operates under the corresponding life state; inputting the remaining historical operation data into the trained support vector machine neural network to obtain a health factor in the range of 0 to 1 corresponding to each historical operation data; using all historical operation data and their corresponding health factors to train a life prediction model, the output result of the life prediction model is the health factor, and the larger the health factor, the longer the remaining life of the transformer indicated by the corresponding life prediction result.

[0012] In one embodiment, a life prediction model is obtained by training with all historical operating data and their corresponding health factors, including: inputting the dynamic parameter information in each group of historical operating data into a trained feature recognition model, and outputting the corresponding dynamic parameter features; using the static parameter information in each group of historical operating data and the dynamic parameter features identified by the dynamic parameter information as input, and the health factors of the historical operating data as output, to obtain a life prediction model based on BP neural network training.

[0013] In one embodiment, the dynamic parameter information includes at least one of the hot spot temperature of the transformer insulation winding, the transformer load voltage, the transformer load current and the transformer magnetic flux, and the static parameter information includes at least one of the moisture content of the transformer insulation paper, the water-soluble acid component in the transformer oil and the acidity value of the transformer oil.

[0014] In a second aspect, the present application also provides a transformer life prediction device, the device comprising:

[0015] A parameter acquisition module is used to obtain dynamic parameter information of the transformer to be predicted. The dynamic parameter information is the real-time operating parameter that affects the life of the transformer;

[0016] A fault feature extraction module is used to input dynamic parameter information into a feature recognition model to identify dynamic parameter features of the transformer to be predicted. The feature recognition model is trained using historical operating data of the transformer.

[0017] The life prediction module is used to fuse the dynamic parameter characteristics with the static parameter information of the transformer to be predicted and input them into the life prediction model to obtain the life prediction result of the transformer to be predicted. The static parameter information is the transformer attribute parameter that affects the life of the transformer. The life prediction model is trained using the historical operation data of the transformer.

[0018] In a third aspect, the present application further provides a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the transformer life prediction method provided in the first aspect are implemented.

[0019] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the transformer life prediction method provided in the first aspect when the computer program is executed by a processor.

[0020] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the steps of the transformer life prediction method provided in the first aspect when executed by a processor.

[0021] The above-mentioned transformer life prediction method, device, computer equipment, storage medium and computer program product can evaluate the current life status and remaining life of the transformer to be predicted through the dynamic parameter information and static parameter information that affect the transformer life of the transformer to be predicted, using the pre-trained feature recognition model and life prediction model, and provide reliable and accurate technical support for the condition maintenance of the transformer.

[0022] This method fully considers various parameters that affect the life of the transformer for model training and detection. The prediction results are accurate and comprehensive, overcoming the limitations of using only some parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 The figure is a flow chart of a transformer life prediction method in one embodiment.

[0024] Figure 2 FIG. 1 is a flow chart of step 102 in an embodiment.

[0025] Figure 3 FIG. 1 is a flow chart of step 104 in one embodiment.

[0026] Figure 4 A flowchart of another embodiment is shown.

[0027] Figure 5 4 is a structural block diagram of a transformer life prediction device in one embodiment.

[0028] Figure 6 A structural block diagram of a transformer life prediction device in another embodiment.

[0029] Figure 7 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0031] In one embodiment, Figure 1 As shown, a transformer life prediction method is provided. This embodiment uses the method applied to a terminal as an example for illustration. It can be understood that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. The method includes the following steps:

[0032] Step 102: Acquire dynamic parameter information of the transformer to be predicted.

[0033] Among them, the dynamic parameter information is the data of the real-time operating parameters that affect the life of the transformer. It is the data that will change continuously during the real-time operation of the transformer and can be collected and acquired through various sensors or detection circuits. Optionally, the dynamic parameter information includes at least one of the hot spot temperature of the transformer insulation winding, the transformer load voltage, the transformer load current and the transformer magnetic flux. In order to ensure the accuracy and comprehensiveness of the prediction, all of these types of data are generally included.

[0034] Step 104: input the dynamic parameter information into a feature recognition model to identify the dynamic parameter features of the transformer to be predicted.

[0035] Among them, the feature recognition model is trained using the historical operation data of the transformer.

[0036] Step 106, the dynamic parameter characteristics are integrated with the static parameter information of the transformer to be predicted and then input into the life prediction model to obtain the life prediction result of the transformer to be predicted. The life prediction result is used to indicate the length of the remaining life of the transformer to be predicted. The life prediction model is trained using the historical operation data of the transformer.

[0037] Among them, static parameter information is the data of transformer attribute parameters that affect the life of the transformer. It is data that is basically stable and unchanged during the real-time operation of the transformer, or data that is basically stable and unchanged within a predetermined period of time. It can be collected and obtained during routine regular maintenance. Optionally, the static parameter information includes at least one of the moisture content of the transformer insulation paper, the water-soluble acid component in the transformer oil, and the acidity value of the transformer oil. In order to ensure the accuracy and comprehensiveness of the prediction, all of these types of data are generally included.

[0038] The above transformer life prediction method can evaluate the life status of the transformer to be predicted through the dynamic parameter information and static parameter information of the transformer to be predicted, using the pre-trained feature recognition model and life prediction model, and provide reliable and accurate technical support for the condition maintenance of the transformer.

[0039] Considering that when obtaining the parameter information of the transformer to be predicted, it is easy to be affected by the external environment or the error of the sensor itself and introduce noise, in order to reduce the influence of noise, in one embodiment, Figure 2 As shown, the above step 102 includes:

[0040] Step 202, during the real-time operation of the transformer to be predicted, original operating parameters that affect the life of the transformer are collected. The original operating data is data of the same information type as the dynamic parameter information in the above step 102, and is the original collected data of the dynamic parameter information that may contain noise.

[0041] Step 204: Use a denoising network to clean the collected original operating parameters to obtain the dynamic parameter information of the above step 102. The denoising network is obtained by training a denoising autoencoder using the historical operating data of the transformer.

[0042] Similarly, optionally, the static parameter information of the transformer to be predicted in the above step 106 is also the information obtained after data cleaning using a noise reduction network.

[0043] In one embodiment, before using the noise reduction network, step 203 is also included, in which the noise reduction network is obtained by training using the historical operation data of the transformer. In one embodiment, when it is necessary to perform data cleaning on the original data of the dynamic parameter information and the static parameter information of the transformer to be predicted, each set of historical operation data of the transformer used to train the noise reduction network includes the dynamic parameter information and the static parameter information during the historical operation of the transformer, and the historical operation data here is generally the statistical data of the transformer of the same model as the transformer to be predicted. In other embodiments, since the static parameter information of the transformer to be predicted is relatively stable and has less noise, while the dynamic parameter information fluctuates greatly and has greater noise, only the original operation data of the dynamic parameter information can be cleaned, and each set of historical operation data used to train the noise reduction network includes the dynamic parameter information of the transformer when it is running under the corresponding life state, without including the static parameter information.

[0044] The denoising network is trained based on the denoising autoencoder. The network learning rate ε is set, and the network parameters of the denoising autoencoder are randomly initialized, including the connection weight W and the offset b. The batch training number and the number of iterations in the forward propagation algorithm are set, and the forward propagation algorithm is executed to calculate the average activation value ρ of any hidden layer neuron j. j , the cost function is calculated using the output of the denoising autoencoder:

[0045]

[0046] Among them, x(i) is the input historical running data, y(i) is the data of x(i) passing through the input layer to the hidden layer, and h W,b (x(i)) is the signal of y(i) output to the output layer after decoding in the hidden layer, i is a parameter and represents a set of historical operation data, and n is the total number of input historical operation data. β is the weight coefficient controlling the penalty term, s2 is the number of neurons in the hidden layer, j is a parameter and represents a neuron, ρ represents the sparsity parameter, and KL() is the relative entropy function.

[0047] Execute the back propagation algorithm and use the cost function to update the network parameters W and b of the denoising autoencoder according to the following formula:

[0048]

[0049]

[0050] In one embodiment, the feature recognition model is obtained based on convolutional neural network training. In order to be applicable to the data type of the feature recognition model, in one embodiment, Figure 3 As shown, the above step 104 includes the following steps:

[0051] Step 302: Perform discrete Fourier transform on the dynamic parameter information to obtain corresponding two-dimensional spectrum image information. The acquired dynamic parameter information is one-dimensional data information containing several data points. Any dynamic parameter information containing N data points can be expressed as x n Represents any nth data point, then the discrete Fourier transform processing of the dynamic parameter information is expressed as

[0052] Step 304: input the two-dimensional spectrum image information corresponding to each dynamic parameter information into the trained feature recognition model.

[0053] Step 306, the output result of the feature recognition model is processed by inverse Fourier transform to obtain dynamic parameter features. The inverse Fourier transform of any output result Y[k] is expressed as The dynamic parameter feature representation of M data points in the form of one-dimensional data information can be obtained as

[0054] In one embodiment, before using the feature recognition model, step 303 is also included, using the historical operation data of the transformer to train the feature recognition model. The historical operation data of the transformer used to train the feature recognition model includes dynamic parameter information during the historical operation of the transformer. The feature recognition model is obtained based on convolutional neural network training. The convolutional neural network includes an input layer and a hidden layer. The hidden layer includes a convolution layer, an excitation function and a pooling layer. Similar to the processing process for predicting the dynamic parameter information of the transformer, the historical operation data is first processed by discrete Fourier transform into the corresponding two-dimensional spectrum image information, and then input into the convolutional neural network for training. Since the convolutional neural network uses the gradient descent algorithm for learning, the input of the convolutional neural network needs to be standardized. Therefore, after the historical operation data of the transformer is processed into two-dimensional spectrum image information, it is first normalized, and the original pixel values ​​distributed in [0, 255] are normalized to the [0, 1] interval, and then input into the convolutional neural network for training. The standardization of the input data is conducive to improving the learning efficiency and performance of the convolutional neural network. Optionally, as described in step 102 above, if the dynamic parameter information includes multiple different types of data, a multi-channel convolutional neural network is constructed and used for training to obtain a feature recognition model.

[0055] In one embodiment, before step 106, the service life prediction model is further provided with a step of training to obtain the service life prediction model. Figure 4 As shown:

[0056] Step 402 , a support vector machine neural network is trained using historical operating data corresponding to a health factor of 1 and historical operating data corresponding to a health factor of 0 in historical operating data of the transformer under different life states.

[0057] The life state of the transformer includes a healthy state, a failed state, and various intermediate states between the healthy state and the failed state. The historical operation data corresponding to the health factor 1 is the historical operation data of the transformer in a healthy state, and the historical operation data corresponding to the health factor 0 is the historical operation data of the transformer in a failed state.

[0058] The insulation paper of the oil-immersed transformer is one of the most fragile components inside the transformer. The operating life of the transformer is often mainly determined by the chemical life of the insulation paper, and the most direct parameter to characterize the aging of the insulation paper is the degree of polymerization. Therefore, the degree of polymerization of the insulation paper is optionally used to measure the life state of the transformer, and the state when the degree of polymerization of the insulation paper of the transformer reaches the first predetermined threshold is determined as a healthy state, and corresponds to a health factor of 1, such as the first predetermined threshold is a value in the range of 1000 to 1200. The state when the degree of polymerization of the insulation paper of the transformer is lower than the second predetermined threshold is determined as a failure state, and corresponds to a health factor of 0, such as the second predetermined threshold is a value in the range of 250 to 300.

[0059] Each set of historical operation data includes dynamic parameter information and static parameter information when the transformer is operating under the corresponding life state. The specific parameter meanings of the dynamic parameter information and the static parameter information are as introduced above.

[0060] Specifically, when training the support vector machine neural network, the historical operation data corresponding to the health factor 1 and the historical operation data corresponding to the health factor 0 are divided into a training set and a test set in proportion for training and testing, that is, this step does not use the historical operation data in various intermediate states. Specifically:

[0061] Set the batch training number and iteration number in the forward propagation algorithm, input the training set data into the support vector machine neural network, execute the forward propagation algorithm, and use the output of the support vector machine neural network to calculate the cost function as follows:

[0062]

[0063] Among them, g(i) is the input historical operation data, H W,b (g(i)) is the output after g(i) is input into the support vector machine neural network, q(i) is the health factor of g(i), and P is the total number of groups of input historical operation data.

[0064] The stochastic gradient descent algorithm is used to perform back propagation calculations, and the cost function is used to update the network parameters W and b of the support vector machine neural network according to the following formula:

[0065]

[0066]

[0067] Step 404, input the remaining historical operation data into the trained support vector machine neural network to obtain a health factor in the range of 0 to 1 corresponding to each historical operation data, that is, to obtain the health factor corresponding to the historical operation data when the transformer is in various other intermediate states.

[0068] Step 406: train a life prediction model using all historical operation data and their corresponding health factors. The output of the life prediction model is the health factor. The larger the health factor, the longer the remaining life of the transformer indicated by the corresponding life prediction result. This step includes:

[0069] Step 406a, input the dynamic parameter information in each set of historical operation data into the trained feature recognition model, and output the corresponding dynamic parameter features. The specific processing flow is similar to steps 302-306, and this embodiment will not expand Miao Sohu.

[0070] Step 406b, taking the static parameter information in each set of historical operation data and the dynamic parameter features identified by the dynamic parameter information as input, and taking the health factor of the historical operation data as output, and obtaining a life prediction model based on BP neural network training.

[0071] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0072] Based on the same inventive concept, the embodiment of the present application also provides a transformer life prediction device for implementing the transformer life prediction method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more transformer life prediction device embodiments provided below can refer to the limitations of the transformer life prediction method above, and will not be repeated here.

[0073] In one embodiment, Figure 5As shown, a transformer life prediction device is provided, including: a parameter acquisition module 510, a fault feature extraction module 520 and a life prediction module 530, wherein:

[0074] The parameter acquisition module 510 is used to acquire dynamic parameter information of the transformer to be predicted. The dynamic parameter information is the real-time operating parameter that affects the life of the transformer.

[0075] The fault feature extraction module 520 is used to input the dynamic parameter information into the feature recognition model to identify the dynamic parameter features of the transformer to be predicted. The feature recognition model is trained using the historical operation data of the transformer.

[0076] The life prediction module 530 is used to fuse the dynamic parameter characteristics with the static parameter information of the transformer to be predicted and input them into the life prediction model to obtain the life prediction result of the transformer to be predicted. The static parameter information is the transformer attribute parameter that affects the life of the transformer. The life prediction model is trained using the historical operation data of the transformer.

[0077] In one embodiment, please refer to Figure 6 , the parameter acquisition module 510 includes a data acquisition unit 511 and a data cleaning unit 512:

[0078] The data acquisition unit 511 is used to acquire original operating parameters that affect the life of the transformer during the real-time operation of the transformer to be predicted.

[0079] The data cleaning unit 512 is used to clean the collected original operating parameters using a denoising network to obtain dynamic parameter information. The denoising network is obtained by training a denoising autoencoder using historical operating data of the transformer.

[0080] In one embodiment, please refer to Figure 6 The fault feature extraction module 520 includes a first transformation unit 521 , an identification unit 522 and a second transformation unit 523 .

[0081] The first transform unit 521 is used to perform discrete Fourier transform processing on the dynamic parameter information to obtain corresponding two-dimensional spectrum image information.

[0082] The recognition unit 522 is used to input the two-dimensional spectrum image information corresponding to each dynamic parameter information into the feature recognition model.

[0083] The second transformation unit 523 is used to perform inverse Fourier transformation on the output result of the feature recognition model to obtain dynamic parameter features.

[0084] In one embodiment, the transformer life prediction device further includes a life prediction model training module 540, and the life prediction model training module 540 includes:

[0085] The first training unit 541 is used to train a support vector machine neural network using historical operation data corresponding to health factor 1 and health factor 0 in historical operation data of the transformer under different life states. The historical operation data corresponding to health factor 1 is the historical operation data of the transformer in a healthy state, and the historical operation data corresponding to health factor 0 is the historical operation data of the transformer in a failed state. Each set of historical operation data includes dynamic parameter information and static parameter information when the transformer is running under the corresponding life state.

[0086] The health factor acquisition unit 542 is used to input the remaining historical operation data into the trained support vector machine neural network to obtain a health factor in the range of 0 to 1 corresponding to each historical operation data.

[0087] The second training unit 543 is used to use all historical operating data and their corresponding health factors to train a life prediction model. The output result of the life prediction model is the health factor, and the larger the health factor, the longer the remaining life of the transformer indicated by the corresponding life prediction result.

[0088] In one embodiment, the second training unit 543 is further used to input the dynamic parameter information in each set of historical operation data into the trained feature recognition model, and output the corresponding dynamic parameter features. The static parameter information in each set of historical operation data and the dynamic parameter features identified by the dynamic parameter information are used as inputs, and the health factors of the historical operation data are used as outputs, and a life prediction model is obtained based on BP neural network training.

[0089] Each module in the above transformer life prediction device can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0090] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 7As shown. The computer device includes a processor, a memory and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the acquired dynamic parameter information of the transformer to be predicted, the static parameter information of the transformer to be predicted, and the pre-trained feature recognition model, life prediction model and noise reduction network. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a transformer life prediction method is implemented.

[0091] Those skilled in the art will understand that Figure 7 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0092] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0093] Obtaining dynamic parameter information of the transformer to be predicted, where the dynamic parameter information is data of real-time operating parameters that affect the life of the transformer;

[0094] The dynamic parameter information is input into the feature recognition model to identify the dynamic parameter characteristics of the transformer to be predicted, and the feature recognition model is trained using the historical operation data of the transformer;

[0095] The dynamic parameter characteristics are fused with the static parameter information of the transformer to be predicted and then input into the life prediction model to obtain the life prediction result of the transformer to be predicted. The static parameter information is the data of the transformer attribute parameters that affect the life of the transformer. The life prediction model is trained using the historical operation data of the transformer.

[0096] In one embodiment, when the processor executes the computer program, the following steps are also implemented: collecting original operating parameters that affect the life of the transformer to be predicted during real-time operation of the transformer to be predicted; using a denoising network to perform data cleaning on the collected original operating parameters to obtain dynamic parameter information, and the denoising network is obtained by training a denoising autoencoder based on the historical operating data of the transformer.

[0097] In one embodiment, when the processor executes the computer program, the following steps are also implemented: the dynamic parameter information is discretely Fourier transformed into corresponding two-dimensional spectrum image information; the two-dimensional spectrum image information corresponding to each dynamic parameter information is input into the feature recognition model; and the output result of the feature recognition model is inversely Fourier transformed to obtain the dynamic parameter feature.

[0098] In one embodiment, when the processor executes the computer program, the following steps are also implemented: using the historical operation data corresponding to the health factor 1 and the historical operation data corresponding to the health factor 0 in the historical operation data of the transformer under different life states to train a support vector machine neural network, wherein the historical operation data corresponding to the health factor 1 is the historical operation data of the transformer in a healthy state, and the historical operation data corresponding to the health factor 0 is the historical operation data of the transformer in a failure state, and each group of historical operation data includes dynamic parameter information and static parameter information when the transformer operates under the corresponding life state; the remaining historical operation data are input into the trained support vector machine neural network to obtain a health factor in the range of 0 to 1 corresponding to each historical operation data; using all the historical operation data and their corresponding health factors to train a life prediction model, the output result of the life prediction model is the health factor, and the larger the health factor, the longer the remaining life of the transformer indicated by the corresponding life prediction result.

[0099] In one embodiment, when the processor executes the computer program, the following steps are also implemented: the dynamic parameter information in each set of historical operation data is input into the trained feature recognition model, and the corresponding dynamic parameter features are output; the static parameter information in each set of historical operation data and the dynamic parameter features identified by the dynamic parameter information are used as input, and the health factor of the historical operation data is used as output, and a life prediction model is obtained based on BP neural network training.

[0100] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0101] Obtaining dynamic parameter information of the transformer to be predicted, where the dynamic parameter information is data of real-time operating parameters that affect the life of the transformer;

[0102] The dynamic parameter information is input into the feature recognition model to identify the dynamic parameter characteristics of the transformer to be predicted, and the feature recognition model is trained using the historical operation data of the transformer;

[0103] The dynamic parameter characteristics are fused with the static parameter information of the transformer to be predicted and then input into the life prediction model to obtain the life prediction result of the transformer to be predicted. The static parameter information is the data of the transformer attribute parameters that affect the life of the transformer. The life prediction model is trained using the historical operation data of the transformer.

[0104] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: collecting original operating parameters that affect the life of the transformer to be predicted during real-time operation of the transformer to be predicted; using a denoising network to perform data cleaning on the collected original operating parameters to obtain dynamic parameter information, and the denoising network is obtained by training a denoising autoencoder based on the historical operating data of the transformer.

[0105] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: the dynamic parameter information is discretely Fourier transformed into corresponding two-dimensional spectrum image information; the two-dimensional spectrum image information corresponding to each dynamic parameter information is input into the feature recognition model; the output result of the feature recognition model is inversely Fourier transformed to obtain the dynamic parameter feature.

[0106] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: a support vector machine neural network is trained using historical operation data corresponding to health factor 1 and health factor 0 in historical operation data of the transformer under different life states, wherein the historical operation data corresponding to health factor 1 is the historical operation data of the transformer in a healthy state, and the historical operation data corresponding to health factor 0 is the historical operation data of the transformer in a failure state, and each group of historical operation data includes dynamic parameter information and static parameter information when the transformer operates under the corresponding life state; the remaining historical operation data are input into the trained support vector machine neural network to obtain a health factor in the range of 0 to 1 corresponding to each historical operation data; a life prediction model is trained using all historical operation data and their corresponding health factors, and the output result of the life prediction model is the health factor, and the larger the health factor, the longer the remaining life of the transformer indicated by the corresponding life prediction result.

[0107] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: the dynamic parameter information in each set of historical operation data is input into the trained feature recognition model, and the corresponding dynamic parameter features are output; the static parameter information in each set of historical operation data and the dynamic parameter features identified by the dynamic parameter information are used as input, and the health factor of the historical operation data is used as output, and a life prediction model is obtained based on BP neural network training.

[0108] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0109] Obtaining dynamic parameter information of the transformer to be predicted, where the dynamic parameter information is data of real-time operating parameters that affect the life of the transformer;

[0110] The dynamic parameter information is input into the feature recognition model to identify the dynamic parameter characteristics of the transformer to be predicted, and the feature recognition model is trained using the historical operation data of the transformer;

[0111] The dynamic parameter characteristics are fused with the static parameter information of the transformer to be predicted and then input into the life prediction model to obtain the life prediction result of the transformer to be predicted. The static parameter information is the data of the transformer attribute parameters that affect the life of the transformer. The life prediction model is trained using the historical operation data of the transformer.

[0112] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: collecting original operating parameters that affect the life of the transformer to be predicted during real-time operation of the transformer to be predicted; using a denoising network to perform data cleaning on the collected original operating parameters to obtain dynamic parameter information, and the denoising network is obtained by training a denoising autoencoder based on the historical operating data of the transformer.

[0113] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: the dynamic parameter information is discretely Fourier transformed into corresponding two-dimensional spectrum image information; the two-dimensional spectrum image information corresponding to each dynamic parameter information is input into the feature recognition model; the output result of the feature recognition model is inversely Fourier transformed to obtain the dynamic parameter feature.

[0114] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: a support vector machine neural network is trained using historical operation data corresponding to health factor 1 and health factor 0 in historical operation data of the transformer under different life states, wherein the historical operation data corresponding to health factor 1 is the historical operation data of the transformer in a healthy state, and the historical operation data corresponding to health factor 0 is the historical operation data of the transformer in a failure state, and each group of historical operation data includes dynamic parameter information and static parameter information when the transformer operates under the corresponding life state; the remaining historical operation data are input into the trained support vector machine neural network to obtain a health factor in the range of 0 to 1 corresponding to each historical operation data; a life prediction model is trained using all historical operation data and their corresponding health factors, and the output result of the life prediction model is the health factor, and the larger the health factor, the longer the remaining life of the transformer indicated by the corresponding life prediction result.

[0115] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: the dynamic parameter information in each set of historical operation data is input into the trained feature recognition model, and the corresponding dynamic parameter features are output; the static parameter information in each set of historical operation data and the dynamic parameter features identified by the dynamic parameter information are used as input, and the health factor of the historical operation data is used as output, and a life prediction model is obtained based on BP neural network training.

[0116] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0117] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0118] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0119] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A transformer life prediction method, characterized in that: The method comprises: Acquire dynamic parameter information of the transformer to be predicted, wherein the dynamic parameter information is data of real-time operating parameters that affect the life of the transformer; Inputting the dynamic parameter information into a feature recognition model to identify the dynamic parameter features of the transformer to be predicted, wherein the feature recognition model is trained using historical operation data of the transformer; The dynamic parameter features are integrated with the static parameter information of the transformer to be predicted and then input into the life prediction model to obtain the life prediction result of the transformer to be predicted, wherein the static parameter information is the data of the transformer attribute parameters that affect the life of the transformer, and the life prediction result is used to indicate the length of the remaining life of the transformer to be predicted. The life prediction model is trained using the historical operation data of the transformer; Among them, the dynamic parameter information includes the hot spot temperature of the transformer insulation winding, the transformer load voltage, the transformer load current and the transformer magnetic flux, and the static parameter information includes the moisture content of the transformer insulation paper, the water-soluble acid components in the transformer oil and the transformer oil acidity value; The method further comprises: A support vector machine neural network is obtained by training the historical operation data corresponding to health factor 1 and health factor 0 in the historical operation data of the transformer under different life states, wherein the historical operation data corresponding to health factor 1 is the historical operation data of the transformer in a healthy state, and the historical operation data corresponding to health factor 0 is the historical operation data of the transformer in a failed state, and each group of historical operation data includes dynamic parameter information and static parameter information when the transformer is operating under the corresponding life state; The rest of the historical operation data is input into the trained support vector machine neural network to obtain the health factor in the range of 0 to 1 corresponding to each historical operation data; The dynamic parameter information in each group of historical operation data is input into the feature recognition model obtained by training, and the corresponding dynamic parameter features are output; the static parameter information in each group of historical operation data and the dynamic parameter features obtained by identifying the dynamic parameter information are used as input, and the health factor of the historical operation data is used as output, and the life prediction model is obtained based on BP neural network training; the output result of the life prediction model is the health factor, and the larger the health factor, the longer the remaining life of the transformer indicated by the corresponding life prediction result.

2. The method according to claim 1, characterized in that: The step of obtaining dynamic parameter information of the transformer to be predicted includes: During the real-time operation of the transformer to be predicted, original operating parameters that affect the life of the transformer are collected; The collected original operating parameters are cleaned using a denoising network to obtain the dynamic parameter information. The denoising network is trained based on a denoising autoencoder using historical operating data of the transformer.

3. The method according to claim 1, characterized in that: The step of inputting the dynamic parameter information into a feature recognition model to identify and obtain the dynamic parameter features of the transformer to be predicted includes: Performing discrete Fourier transform on the dynamic parameter information to obtain corresponding two-dimensional spectrum image information; Inputting the two-dimensional spectrum image information corresponding to each dynamic parameter information into the feature recognition model; The output result of the feature recognition model is subjected to inverse Fourier transform processing to obtain the dynamic parameter feature.

4. A transformer life prediction device, characterized in that: The transformer life prediction device is used to execute the transformer life prediction method according to any one of claims 1 to 3, and the transformer life prediction device includes: A parameter acquisition module is used to acquire dynamic parameter information of the transformer to be predicted, wherein the dynamic parameter information is a real-time operating parameter that affects the life of the transformer; A fault feature extraction module, used for inputting the dynamic parameter information into a feature recognition model to identify the dynamic parameter features of the transformer to be predicted, wherein the feature recognition model is trained using historical operation data of the transformer; A life prediction module, used for fusing the dynamic parameter characteristics with the static parameter information of the transformer to be predicted and inputting them into a life prediction model to obtain a life prediction result of the transformer to be predicted, wherein the static parameter information is a transformer attribute parameter that affects the life of the transformer, and the life prediction model is trained using historical operation data of the transformer; Among them, the dynamic parameter information includes the hot spot temperature of the transformer insulation winding, the transformer load voltage, the transformer load current and the transformer magnetic flux, and the static parameter information includes the moisture content of the transformer insulation paper, the water-soluble acid component in the transformer oil and the acidity value of the transformer oil.

5. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 3 are implemented.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 3 are implemented.

7. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 3 are implemented.

Images