Method and device for identifying arc discharge modes in transformer short-gap oil

Abstract

A method and apparatus for identifying arc discharge patterns in short-gap oil of transformers are disclosed. The method involves conducting arc discharge experiments on oil-paper insulation based on different discharge defect models of oil-immersed power transformers; collecting multi-physical feature signals of short-gap oil arc discharge from the oil-paper insulation arc discharge experiments using sensors; processing the multi-physical feature signals using generalized S-transform to generate a multi-physical signal time-frequency spectrum; and inputting the generated multi-physical signal time-frequency spectrum into a multi-channel deep separable convolutional neural network model to achieve arc discharge pattern identification through joint detection of multiple physical quantities.

Description

Technical Field

[0001] This invention relates to the field of arc discharge technology for oil-immersed power transformers, and particularly to a method and device for identifying short-gap arc discharge patterns in oil of transformers. Background Technology

[0002] Short-gap arc discharge in oil can easily damage equipment and even cause fires and explosions. Therefore, fault diagnosis of arc discharge is crucial for maintaining the safe and stable operation of power systems. When an arc discharge fault occurs in a power transformer, the violent reaction generates various physicochemical signals, including acoustic, optical, thermal, and electromagnetic signals. Exploring the development patterns of fault signals for different discharge types and performing pattern recognition using artificial intelligence technology is an important prerequisite for formulating equipment maintenance strategies. In the field of partial discharge, relatively mature discharge pattern recognition algorithms include artificial neural networks, support vector machines, and clustering algorithms. Compared to partial discharge, arc discharge is faster and has greater discharge energy. Its signal characteristics differ significantly from those of partial discharge. Therefore, the research conclusions and methods for partial discharge are not universally applicable to the field of arc discharge, and current research on fault diagnosis and pattern recognition of arc discharge is still relatively lacking. Related technologies all study arc discharge through single physical signals, which is easily affected by objective factors such as noise. Their effectiveness is insufficient to comprehensively address all fault types, and the recognition accuracy is low. There is an urgent need to develop more robust multi-physical quantity joint recognition methods.

[0003] Furthermore, when the arc discharge develops into the long-gap arc discharge stage, the accumulated energy of the arc is already very large, posing a huge threat to the safe operation of power equipment. Therefore, it is essential to carry out fault identification in the short-gap arc discharge stage.

[0004] The information disclosed in the background section is only for enhancing the understanding of the background of this invention, and therefore may contain information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] This invention provides a method and apparatus for identifying arc discharge patterns in transformer short-gap oil. By collecting multiple physical signals such as current, ultrasound, ultra-high frequency, and oil pressure generated by arc discharge in short-gap oil, and based on a joint detection method, a deep separable convolutional neural network model is used to mine the time-frequency spectrum of the multiple physical signals to achieve discharge pattern identification. This solves the contradiction between the limited amount of available information and low recognition accuracy when relying solely on a single physical quantity for discharge pattern identification, and the complexity and variability of arc discharge and low error tolerance.

[0006] A method for identifying short-gap oil arc discharge modes in transformers includes:

[0007] Step 1: Conduct arc discharge experiments on oil-paper insulation based on different discharge defect models of oil-immersed power transformers;

[0008] Step 2: Use sensors to collect multi-physical characteristic signals of short-gap oil arc discharge in the oil-paper insulation arc discharge experiment;

[0009] Step 3: Process the multi-physical feature signals using the generalized S-transform to generate a multi-physical signal time-frequency spectrum;

[0010] Step 4: Input the generated multi-physical signal time-frequency spectrum into a multi-channel deep separable convolutional neural network model to achieve arc discharge pattern recognition through joint detection of multiple physical quantities. The multi-channel deep separable convolutional neural network model has multiple input channels and includes an input layer, a training layer, a stitching layer, and an output layer. The input layer acquires the multi-physical signal time-frequency spectrum and feeds it to each single channel of the convolutional neural network model. Then, the image data input to each single channel undergoes convolution and pooling operations through the training layer within that channel. Finally, the data from each channel after the training layer are fed together into the same stitching layer, where tensor stitching is used to integrate and classify the information from each channel. The recognition result is then output through the output layer.

[0011] The multi-channel deep separable convolutional neural network model uses Kaiming initialization to set the model parameters, calculates the loss using the cross-entropy function, and uses the loss value as a comparison basis. When the loss value exceeds the manually set threshold, the Adam algorithm is used to find the optimal value of the network hyperparameters to ensure stable convergence of the model.

[0012] In the aforementioned method for identifying arc discharge patterns in transformer short-gap oil, conducting an oil-paper insulation arc discharge experiment includes building an arc discharge experimental platform. An AC power supply is applied to the sealed discharge platform by a voltage regulator to provide a controllable voltage. The discharge platform has electrodes for discharge, as well as ultrasonic sensors, pressure sensors, and ultra-high frequency sensors for measuring characteristic signals of the discharge, voltage sensors for acquiring voltage signals, and Hall coils for acquiring current signals.

[0013] In the aforementioned method for identifying arc discharge patterns in transformer short-gap oil, the discharge defect models tested by the discharge platform are divided into needle plate electrode models, surface electrode models, and inter-turn electrode models.

[0014] In the method for identifying arc discharge patterns in transformer short-gap oil, step 2 involves acquiring voltage, current, ultrasonic, oil pressure, and ultra-high frequency time-domain signals from the arc discharge experiment and saving them to a computer using a data acquisition card.

[0015] In the aforementioned method for identifying arc discharge patterns in transformer short-gap oil, the time-frequency spectrum is a grayscale image or a pseudo-color image.

[0016] In the method for recognizing arc discharge patterns in transformer short-gap oil, step 4 involves inputting the time-frequency spectra of each physical feature signal into each channel of a deep separable convolutional neural network model. The correlation between each physical quantity is obtained through training the deep separable convolutional neural network model, thereby realizing arc discharge pattern recognition through joint detection of multiple physical quantities.

[0017] In the method for identifying arc discharge patterns in transformer short-gap oil, step 3, the algorithm for generating time-frequency spectra using the generalized S-transform, is as follows:

[0018] In this context, GST(τ, f) is the result of the generalized S-transform of the signal, λ is the introduced adjustment factor, τ is the time shift factor, f is the frequency, x(t) is the original signal, and i is the imaginary unit.

[0019] The apparatus for implementing the method of identifying arc discharge patterns in transformer short-gap oil includes,

[0020] The signal acquisition module is used to acquire multi-physical characteristic signals generated by arc discharge in short-gap oil.

[0021] The spectrum generation module, which is connected to the signal acquisition module, is used to convert the multi-physical feature signal into its corresponding time-frequency spectrum.

[0022] The discharge identification module is connected to the spectrum generation module. The discharge identification module includes a deep separable convolutional neural network model that learns the time-frequency spectrum to realize discharge pattern recognition.

[0023] In the aforementioned transformer short-gap oil arc discharge pattern recognition device, the signal acquisition module includes an ultrasonic sensor, a voltmeter, a Hall coil or a high-frequency current sensor, a pressure sensor, and an ultra-high frequency sensor.

[0024] In the aforementioned transformer short-gap oil arc discharge pattern recognition device, the discharge recognition module is integrated into the central processing unit.

[0025] Compared with the prior art, the present invention has the following advantages: The present invention enriches the detailed information for the identification of different discharge models by means of joint detection of multiple physical quantities. When the neural network is working, it can dig out more detailed and deeper discharge features, avoiding the disadvantages of single physical quantity mapping less information, easy to be random, low distinguishability and thus difficult to distinguish, and high false recognition rate. It helps to improve the accuracy of discharge pattern recognition and prevent short gap arc discharge from developing to a more serious stage. Attached Figure Description

[0026] Various other advantages and benefits of the present invention will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiments below. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. It is obvious that the drawings described below are merely some embodiments of the invention, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort. Furthermore, the same reference numerals denote the same parts throughout the drawings.

[0027] In the attached diagram:

[0028] Figure 1 This is a flowchart illustrating a method for identifying short-gap arc discharge patterns in oil based on the joint detection of multiple physical quantities.

[0029] Figure 2 This is a schematic diagram of the oil-paper insulated arc discharge experimental platform of the present invention;

[0030] Figure 3 This is a schematic diagram of different discharge defect models of the oil-immersed power transformer of the present invention;

[0031] Figure 4 This is a schematic diagram of the time-frequency spectrum generated by the generalized S-transform of the time-domain signal according to the present invention;

[0032] Figure 5 This is a schematic diagram of the confusion matrix of the discharge pattern recognition method based on joint detection of multiple physical quantities proposed in this invention.

[0033] The present invention will be further explained below with reference to the accompanying drawings and embodiments. Detailed Implementation

[0034] Specific embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While specific embodiments of the invention are shown in the drawings, it should be understood that the invention may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this invention will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

[0035] It should be noted that certain terms are used in the specification and claims to refer to specific components. Those skilled in the art will understand that different terms may be used to refer to the same component. This specification and claims do not distinguish components based on differences in terminology, but rather on differences in function. The terms "comprising" or "including" used throughout the specification and claims are open-ended and should be interpreted as "comprising but not limited to." The following descriptions are preferred embodiments for carrying out the invention; however, these descriptions are for the purpose of understanding the general principles of the specification and are not intended to limit the scope of the invention. The scope of protection of this invention is determined by the appended claims.

[0036] To facilitate understanding of the embodiments of the present invention, further explanations and descriptions will be provided below with reference to the accompanying drawings and specific embodiments. The accompanying drawings do not constitute a limitation on the embodiments of the present invention.

[0037] like Figures 1 to 5 As shown, the method for identifying arc discharge modes in transformer short-gap oil includes the following steps:

[0038] Step 1: Conduct arc discharge experiments on oil-paper insulation based on different discharge defect models of oil-immersed power transformers;

[0039] Step 2: Use sensors to collect multi-physical characteristic signals of short-gap oil arc discharge in the oil-paper insulation arc discharge experiment;

[0040] Step 3: Process the multi-physical feature signals using the generalized S-transform to generate a multi-physical signal time-frequency spectrum;

[0041] Step 4: Input the generated multi-physical signal time-frequency spectrum into a multi-channel deep separable convolutional neural network model to achieve arc discharge pattern recognition through joint detection of multiple physical quantities. The multi-channel deep separable convolutional neural network model has multiple input channels and includes an input layer, a training layer, a stitching layer, and an output layer. The input layer acquires the multi-physical signal time-frequency spectrum and feeds it to each single channel of the convolutional neural network model. Then, the image data input to each single channel undergoes convolution and pooling operations through the training layer within that channel. Finally, the data from each channel after the training layer are fed together into the same stitching layer, where tensor stitching is used to integrate and classify the information from each channel. The recognition result is then output through the output layer.

[0042] The multi-channel deep separable convolutional neural network model uses Kaiming initialization to set the model parameters, calculates the loss using the cross-entropy function, and uses the loss value as a comparison basis. When the loss value exceeds the manually set threshold, the Adam algorithm is used to find the optimal value of the network hyperparameters to ensure stable convergence of the model.

[0043] In a preferred embodiment of the method for identifying arc discharge patterns in transformer short-gap oil, conducting an arc discharge experiment with oil-paper insulation includes setting up an arc discharge experimental platform. An AC power supply is applied to the sealed discharge platform by a voltage regulator to provide a controllable voltage. The discharge platform has electrodes for discharge, an ultrasonic sensor, a pressure sensor, and an ultra-high frequency sensor for measuring characteristic signals of the discharge, a voltage sensor for acquiring voltage signals, and a Hall coil for acquiring current signals.

[0044] In a preferred embodiment of the method for identifying arc discharge patterns in transformer short-gap oil, the discharge defect models tested by the discharge platform are divided into needle plate electrode models, surface electrode models, and inter-turn electrode models.

[0045] In a preferred embodiment of the method for identifying arc discharge patterns in transformer short-gap oil, in step 2, voltage, current, ultrasonic, oil pressure, and ultra-high frequency time domain signals of the arc discharge experiment are collected and saved to a computer using a data acquisition card.

[0046] In a preferred embodiment of the method for identifying arc discharge patterns in transformer short-gap oil, the time-frequency spectrum is a grayscale image or a pseudo-color image.

[0047] In a preferred embodiment of the method for identifying arc discharge patterns in transformer short-gap oil, in step 4, the time-frequency spectra of each physical feature signal are input into each channel of the deep separable convolutional neural network model. The correlation between each physical quantity is obtained through training the deep separable convolutional neural network model, thereby realizing arc discharge pattern identification by joint detection of multiple physical quantities.

[0048] In a preferred embodiment of the method for identifying arc discharge patterns in transformer short-gap oil, step 3, the algorithm for generating time-frequency spectra using the generalized S-transform, is as follows:

[0049] In this context, GST(τ, f) is the result of the generalized S-transform of the signal, λ is the introduced adjustment factor, τ is the time shift factor, f is the frequency, x(t) is the original signal, and i is the imaginary unit.

[0050] An apparatus for identifying arc discharge patterns in transformer short-gap oil includes,

[0051] The signal acquisition module is used to acquire multi-physical characteristic signals generated by arc discharge in short-gap oil.

[0052] The spectrum generation module, which is connected to the signal acquisition module, is used to convert the multi-physical feature signal into its corresponding time-frequency spectrum.

[0053] The discharge identification module is connected to the spectrum generation module. The discharge identification module includes a deep separable convolutional neural network model that learns the time-frequency spectrum to realize discharge pattern recognition.

[0054] In a preferred embodiment of the transformer short-gap oil arc discharge pattern recognition device, the signal acquisition module includes an ultrasonic sensor, a voltmeter, a Hall coil or a high-frequency current sensor, a pressure sensor and an ultra-high frequency sensor.

[0055] In a preferred embodiment of the transformer short-gap oil arc discharge pattern recognition device, the discharge recognition module is integrated into the central processing unit. In one embodiment, the short gap is in the millimeter range. In one embodiment, in step 1, different discharge defect models of oil-immersed power transformers are designed, an arc discharge experimental platform is built, and oil-paper insulation arc discharge experiments are conducted. The arc discharge experimental platform includes a voltage regulator, a test transformer, and an experimental enclosure for simulating oil arc discharge; its data acquisition part includes a high-frequency coil, an ultrasonic sensor, a pressure sensor, and an ultra-high frequency sensor for acquiring discharge characteristic signals. In step 2, the high-frequency coil, ultrasonic sensor, pressure sensor, and ultra-high frequency sensor are used to acquire current, ultrasonic, oil pressure, and ultra-high frequency signals during the short-gap oil arc discharge process, respectively, and the data is saved to a computer using a data acquisition card. The data acquisition card can simultaneously access multiple physical quantity characteristic signals, triggering the acquisition function at the moment of electrode model breakdown. Through the joint detection of multiple physical quantities, richer discharge details can be reflected, and the overall recognition accuracy is as high as 98.730%.

[0056] Although embodiments of the present invention have been described above in conjunction with the accompanying drawings, the present invention is not limited to the specific embodiments and application fields described above. The specific embodiments described above are merely illustrative and instructive, and not restrictive. Those skilled in the art can make many other forms based on the guidance of this specification and without departing from the scope of protection of the claims of the present invention, and all of these are within the scope of protection of the present invention.

Claims

1. A method for identifying arc discharge patterns in transformer short-gap oil, characterized in that, Includes the following steps: Step 1: Conduct arc discharge experiments on oil-paper insulation based on different discharge defect models of oil-immersed power transformers; Step 2: Use sensors to collect multi-physical characteristic signals of short-gap oil arc discharge in the oil-paper insulation arc discharge experiment; Step 3: Process the multi-physical feature signals using the generalized S-transform to generate a multi-physical signal time-frequency spectrum; Step 4: Input the generated multi-physical signal time-frequency spectrum into a multi-channel deep separable convolutional neural network model to achieve arc discharge pattern recognition through joint detection of multiple physical quantities. The multi-channel deep separable convolutional neural network model has multiple input channels and includes an input layer, a training layer, a stitching layer, and an output layer. The input layer acquires the multi-physical signal time-frequency spectrum and feeds it to each single channel of the convolutional neural network model. Then, the image data input to each single channel undergoes convolution and pooling operations through the training layer within that channel. Finally, the data from each channel after the training layer are fed together into the same stitching layer, where tensor stitching is used to integrate and classify the information from each channel. The recognition result is then output through the output layer. The multi-channel deep separable convolutional neural network model uses Kaiming initialization to set the model parameters, calculates the loss using the cross-entropy function, and uses the loss value as a comparison basis. When the loss value exceeds the manually set threshold, the Adam algorithm is used to find the optimal value of the network hyperparameters to ensure stable convergence of the model.

2. The method for identifying arc discharge patterns in transformer short-gap oil according to claim 1, characterized in that, Conducting arc discharge experiments on oil-paper insulation involves building an arc discharge experimental platform. An AC power supply is regulated by a voltage regulator to apply a controllable voltage to the sealed discharge platform. The discharge platform has electrodes for discharge, as well as ultrasonic sensors, pressure sensors, and ultra-high frequency sensors for measuring characteristic signals of the discharge, voltage sensors for acquiring voltage signals, and high-frequency current sensors for acquiring current signals.

3. The method for identifying arc discharge patterns in transformer short-gap oil according to claim 1, characterized in that, The discharge defect models for discharge platform testing are divided into needle plate electrode model, surface electrode model, and inter-turn electrode model.

4. The method for identifying arc discharge patterns in transformer short-gap oil according to claim 1, characterized in that, In step 2, voltage, current, ultrasonic, oil pressure, and ultra-high frequency time domain signals of the arc discharge experiment are collected and saved to the computer using a data acquisition card.

5. The method for identifying arc discharge patterns in transformer short-gap oil according to claim 1, characterized in that, The time-frequency spectrum is a grayscale image or a pseudo-color image.

6. The method for identifying arc discharge patterns in transformer short-gap oil according to claim 1, characterized in that, In step 4, the time-frequency spectra of each physical feature signal are input into each channel of the deep separable convolutional neural network model. The correlation between each physical quantity is obtained through training the deep separable convolutional neural network model, so as to realize the arc discharge pattern recognition by joint detection of multiple physical quantities.

7. The method for identifying arc discharge patterns in transformer short-gap oil according to claim 1, characterized in that, In step 3, the algorithm for generating time-frequency spectra using the generalized S-transform is as follows: ,in, GST ( τ , f The result of the generalized S-transform of the signal is shown below. λ For the introduced regulatory factor, τ For time shift factor, f For frequency, x ( t ) represents the original signal, and i represents the imaginary unit.

8. An apparatus for implementing the method for identifying arc discharge patterns in transformer short-gap oil according to any one of claims 1-7, characterized in that, It includes, The signal acquisition module is used to acquire multi-physical characteristic signals generated by arc discharge in short-gap oil. The spectrum generation module, which is connected to the signal acquisition module, is used to convert the multi-physical feature signal into its corresponding time-frequency spectrum. The discharge identification module is connected to the spectrum generation module. The discharge identification module includes a deep separable convolutional neural network model that learns the time-frequency spectrum to realize discharge pattern recognition.

9. The apparatus according to claim 8, characterized in that, The signal acquisition module includes an ultrasonic sensor, a voltage sensor, a high-frequency current sensor, a pressure sensor, and an ultra-high frequency sensor.

10. The apparatus according to claim 8, characterized in that, The discharge identification module is integrated into the central processing unit.

Images