HFCT discharge pulse identification method based on EWT-CNN algorithm

Abstract

The invention discloses an HFCT discharge pulse identification method based on an EWT-CNN algorithm, and relates to the technical field of electrical equipment partial discharge detection, and the method comprises the following steps: S01, carrying out the Fourier transform of a partial discharge pulse signal, and obtaining a frequency diagram; s02, based on a plurality of maximum value points obtained in the frequency diagram, determining the maximum value points and then arranging the maximum value points in sequence from large to small; s03, obtaining a corresponding number of maximum value points according to a preset decomposition number; s04, obtaining a frequency spectrum segmentation point by adopting a mode of taking an envelope minimum value from two adjacent maximum extreme points; and S05, decomposing the signal according to an EWT algorithm to obtain components. According to the method, important information existing in the discharge pulse can be effectively extracted, the inaccuracy of manual feature extraction is effectively avoided by depending on the powerful feature extraction and self-learning function of the CNN in the subsequent feature extraction, and the recognition accuracy of the discharge pulse signal is improved.

Description

Technical Field

[0001] This invention relates to the field of partial discharge detection technology for electrical equipment, and in particular to a method for identifying HFCT discharge pulses based on the EWT-CNN algorithm. Background Technology

[0002] The EWT (Empirical Wavelet Transform) algorithm combines wavelet transform with empirical mode decomposition theory. It constructs corresponding wavelet filter banks through adaptive segmentation in the frequency domain, thereby extracting modal components from different frequency bands. This method effectively suppresses mode aliasing and endpoint effects. Compared to other decomposition algorithms, EWT has a more complete mathematical theoretical foundation; and compared to VMD (Variational Mode Decomposition) algorithms, EWT has higher computational efficiency. Therefore, EWT significantly improves computational speed while accurately extracting important signal features.

[0003] Traditional pattern recognition methods rely heavily on manually selected features, which is not only highly experience-dependent but also makes it difficult to ensure the extraction of all effective features, easily leading to incomplete features. Therefore, these methods often have limited generalization ability when facing partial discharge detection requirements in different scenarios. In contrast, deep learning techniques, represented by CNNs (Convolutional Neural Networks), effectively overcome these shortcomings. With their powerful automatic feature extraction and learning capabilities, CNN algorithms not only significantly improve classification accuracy but also possess stronger model generalization ability.

[0004] Combining the EWT algorithm with CNN leverages EWT's superior feature mining capabilities to first extract crucial information from partial discharge signals (achieving data augmentation), and then combines this with CNN's powerful classification function for model training. This fusion approach effectively addresses the needs of partial discharge identification under complex conditions in various scenarios. Summary of the Invention

[0005] To address the aforementioned technical problems, the present invention provides a method for identifying HFCT discharge pulses based on the EWT-CNN algorithm, the method comprising the following steps:

[0006] S01. Perform Fourier transform on the partial discharge pulse signal to obtain the frequency diagram;

[0007] S02. Based on the number of maxima points obtained from the frequency graph, the maxima points are determined and arranged in descending order.

[0008] S03. Obtain the corresponding number of maximum points according to the preset number of decompositions;

[0009] S04. The spectrum segmentation point is obtained by taking the minimum value of the envelope between two adjacent maximum and extreme points;

[0010] S05. Decompose the signal according to the EWT algorithm to obtain the components;

[0011] S06. Downsample the decomposed components to obtain the feature vector;

[0012] S07. The feature vectors are trained based on the CNN algorithm, and the optimal model is selected by parameter tuning for subsequent partial discharge identification.

[0013] Preferably, the partial discharge pulse signal in step S01 includes, but is not limited to, a tip discharge pulse signal, a floating pulse discharge pulse signal, and a surface pulse discharge pulse signal.

[0014] Preferably, obtaining the corresponding number of maximum points in step S02 refers to obtaining the corresponding number of maximum points in order of sorting from largest to smallest.

[0015] Preferably, in step S06, the decomposed components are downsampled and then arranged in order from low frequency to high frequency to form a one-dimensional feature vector.

[0016] Preferably, the EWT algorithm in step S05 decomposes the signal by:

[0017] S51, Calculate the input signal The Fourier transform yields the spectrum.

[0018] S52. Define the center point between two adjacent maxima as... ,by Using the boundary as the boundary, the Fourier spectrum exist The upper part is divided into A series of paragraphs, in which Then, find the local maxima in the spectrum and sort them in descending order. Let there be a total of maxima. One. If Then keep the previous one. A maximum value, if Then keep A maximum value, and for Make corrections.

[0019] Preferably, the EWT algorithm in step S05 further includes the following for decomposing the signal:

[0020] S53. Construct based on spectrum segmentation results An empirical wavelet function ,in:

[0021] The expression for the empirical scaling function:

[0022] ;

[0023] in, , .

[0024] Empirical wavelet function expression:

[0025] ;

[0026] in, ;

[0027] and and The possible values ​​include the following formulas:

[0028] ;

[0029] .

[0030] Preferably, the EWT algorithm in step S05 further includes the following for decomposing the signal:

[0031] S54. Calculate EWT and its detailed correlation coefficient. With approximate correlation coefficient The formulas include the following:

[0032] ;

[0033] ;

[0034] S55, Perform signal reconstruction:

[0035] ;

[0036] ;

[0037] .

[0038] This invention has at least the following beneficial effects: Optimizing the spectrum segmentation method from the maximum or minimum value between adjacent extreme points to selecting the minimum envelope value, based on the EWT algorithm principle, improves signal decomposition performance. Then, by using the EWT algorithm to extract effective information from the partial discharge signal, enhancing data features, and combining this with the self-learning mechanism of the CNN algorithm, the accuracy of identifying different types of partial discharge signals can be effectively improved, which is of great significance for subsequent equipment maintenance. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a flowchart of an HFCT discharge pulse recognition method based on the EWT-CNN algorithm provided in Embodiment 1 of the present invention;

[0041] Figure 2(a) is a schematic diagram of the tip discharge pulse signal provided in Embodiment 1 of the present invention;

[0042] Figure 2(b) is a schematic diagram of the suspended pulse discharge pulse signal provided in Embodiment 1 of the present invention;

[0043] Figure 2(c) is a schematic diagram of the surface pulse discharge pulse signal provided in Embodiment 1 of the present invention;

[0044] Figure 3 This is a schematic diagram of spectrum segmentation provided in Embodiment 1 of the present invention;

[0045] Figure 4 This is an EWT decomposition diagram provided in Embodiment 1 of the present invention. Detailed Implementation

[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention 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 so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or 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 server that includes 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 devices.

[0048] Example 1

[0049] This embodiment provides a method for identifying HFCT discharge pulses based on the EWT-CNN algorithm. The method includes the following steps: Figure 1 As shown:

[0050] S01. Perform Fourier transform on the partial discharge pulse signal to obtain the frequency diagram;

[0051] Specifically, as shown in Figures 2(a), 2(b), and 2(c), the partial discharge pulse signal includes, but is not limited to, the tip discharge pulse signal, the floating pulse discharge pulse signal, and the surface pulse discharge pulse signal.

[0052] S02. Based on the number of maxima obtained from the frequency graph, arrange the maxima in descending order.

[0053] Specifically, obtaining the corresponding number of maximum points mentioned above refers to obtaining the corresponding number of maximum points in order of sorting from largest to smallest.

[0054] S03. Obtain the corresponding number of maximum points according to the preset number of decompositions;

[0055] S04. The spectrum segmentation point is obtained by taking the minimum value of the envelope between two adjacent maximum and extreme points.

[0056] S05. Decompose the signal according to the EWT algorithm to obtain the components;

[0057] Specifically, the EWT algorithm described above decomposes the signal as follows:

[0058] S51, Calculate the input signal The Fourier transform yields the spectrum.

[0059] S52. Define the center point between two adjacent maxima as... ,by Using the boundary as the boundary, the Fourier spectrum exist The upper part is divided into A series of paragraphs, in which Then, find the local maxima in the spectrum and sort them in descending order. Let there be a total of maxima. One. If Then keep the previous one. A maximum value, if Then keep A maximum value, and for Make corrections.

[0060] Preferably, step S05, in which the EWT algorithm decomposes the signal, further includes:

[0061] S53. Construct based on spectrum segmentation results An empirical wavelet function ,in:

[0062] The expression for the empirical scaling function:

[0063] ;

[0064] in, , .

[0065] Empirical wavelet function expression:

[0066] ;

[0067] in, ;

[0068] and and The possible values ​​include the following formulas:

[0069] ;

[0070] .

[0071] Preferably, step S05, in which the EWT algorithm decomposes the signal, further includes:

[0072] S54. Calculate EWT and its detailed correlation coefficient. With approximate correlation coefficient The formulas include the following:

[0073] ;

[0074] ;

[0075] S55, Perform signal reconstruction:

[0076] ;

[0077] ;

[0078] .

[0079] S06. Downsample the decomposed components to obtain the feature vector;

[0080] As mentioned above, downsampling the decomposed components involves downsampling the decomposed components and then constructing a one-dimensional feature vector from the low-frequency components to the high-frequency components.

[0081] S07. The feature vectors are trained based on the CNN algorithm, and the optimal model is selected by parameter tuning for subsequent partial discharge identification.

[0082] Specifically, through training, CNNs can adaptively learn complex discriminative patterns in feature vectors that are highly correlated with different types of partial discharge (PD) signals. The parameter tuning process optimizes hyperparameters such as model structure and learning rate to find the model configuration that performs best on the validation set and has the strongest generalization ability. The resulting "optimal model" is crucial for achieving high accuracy and robustness in subsequent practical PD recognition tasks, effectively handling complex and changing on-site conditions. Its purpose is to optimize the frequency map obtained through the row Fourier transform in step S01.

[0083] In summary, optimizing the spectrum segmentation method from the maximum or minimum value between adjacent extreme points to selecting the minimum envelope value, based on the EWT algorithm principle, improves signal decomposition performance. Then, by using the EWT algorithm to extract effective information from the partial discharge signal, enhancing data features, and combining this with the self-learning mechanism of the CNN algorithm, the recognition accuracy for different types of partial discharge signals can be effectively improved. This is of great significance for subsequent equipment maintenance.

[0084] Example 2

[0085] This invention provides a non-transitory computer-readable storage medium storing at least one instruction or at least one program segment, which is loaded and executed by a processor to implement the following steps:

[0086] The frequency response diagram is obtained by performing a Fourier transform on the partial discharge pulse signal;

[0087] Based on a number of maxima points obtained from the frequency graph, the maxima points are determined and arranged in descending order;

[0088] Obtain the corresponding number of maximum points according to the preset number of decompositions;

[0089] The spectrum segmentation points are obtained by taking the minimum value of the envelope between two adjacent maximum and extreme points;

[0090] The signal is decomposed using the EWT algorithm to obtain its components;

[0091] The decomposed components are downsampled to obtain the feature vector;

[0092] The feature vectors are trained using the CNN algorithm, and the optimal model is selected by parameter tuning for subsequent partial discharge identification.

[0093] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0094] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0095] Example 3

[0096] This invention provides an electronic device, including a processor and a memory, wherein the memory stores at least one instruction or at least one program segment, and the at least one instruction or the at least one program segment is loaded and executed by the processor to implement the following steps:

[0097] The frequency response diagram is obtained by performing a Fourier transform on the partial discharge pulse signal;

[0098] Based on a number of maxima points obtained from the frequency graph, the maxima points are determined and arranged in descending order;

[0099] Obtain the corresponding number of maximum points according to the preset number of decompositions;

[0100] The spectrum segmentation points are obtained by taking the minimum value of the envelope between two adjacent maximum and extreme points;

[0101] The signal is decomposed using the EWT algorithm to obtain its components;

[0102] The decomposed components are downsampled to obtain the feature vector;

[0103] The feature vectors are trained using the CNN algorithm, and the optimal model is selected by parameter tuning for subsequent partial discharge identification.

[0104] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for identifying HFCT discharge pulses based on the EWT-CNN algorithm, characterized in that, The method includes the following steps: S01. Perform Fourier transform on the partial discharge pulse signal to obtain the frequency diagram; S02. Based on the number of maxima points obtained from the frequency graph, the maxima points are determined and arranged in descending order. S03. Obtain the corresponding number of maximum points according to the preset number of decompositions; S04. The spectrum segmentation point is obtained by taking the minimum value of the envelope between two adjacent maximum and extreme points; S05. Decompose the signal according to the EWT algorithm to obtain the components; S06. Downsample the decomposed components to obtain the feature vector; S07. The feature vectors are trained based on the CNN algorithm, and the optimal model is selected by parameter tuning for subsequent partial discharge identification.

2. The HFCT discharge pulse recognition method based on the EWT-CNN algorithm according to claim 1, characterized in that, The partial discharge pulse signal in step S01 includes, but is not limited to, tip discharge pulse signal, suspension pulse discharge pulse signal and surface pulse discharge pulse signal.

3. The HFCT discharge pulse recognition method based on the EWT-CNN algorithm according to claim 1, characterized in that, In step S02, obtaining the corresponding number of maximum points refers to sorting the maximum points sequentially from largest to smallest.

4. The HFCT discharge pulse identification method based on the EWT-CNN algorithm according to claim 1, characterized in that, In step S06, the decomposed components are downsampled and then arranged in order from low frequency to high frequency to form a one-dimensional feature vector.

5. The HFCT discharge pulse identification method based on the EWT-CNN algorithm according to claim 1, characterized in that, The EWT algorithm in step S05 decomposes the signal as follows: S51, Calculate the input signal The Fourier transform yields the spectrum. S52. Define the center point between two adjacent maxima as... ,by Using the boundary as the boundary, the Fourier spectrum exist The upper part is divided into A series of paragraphs, in which Then, find the local maxima in the spectrum and sort them in descending order. Let there be a total of maxima. One. If Then keep the previous one. A maximum value, if Then keep A maximum value, and for Make corrections.

6. The HFCT discharge pulse identification method based on the EWT-CNN algorithm according to claim 5, characterized in that, The EWT algorithm in step S05 further includes the following steps for decomposing the signal: S53. Construct based on spectrum segmentation results An empirical wavelet function ,in: The expression for the empirical scaling function: ； in, , . Empirical wavelet function expression: ； in, ; and and The possible values ​​include the following formulas: ； 。 7. The HFCT discharge pulse recognition method based on the EWT-CNN algorithm according to claim 6, characterized in that, The EWT algorithm in step S05 further includes the following steps for decomposing the signal: S54. Calculate EWT and its detailed correlation coefficient. With approximate correlation coefficient The formulas include the following: ； ； S55, Perform signal reconstruction: ； ； 。 8. A non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores at least one instruction or at least one program segment, characterized in that, The at least one instruction or the at least one program segment is loaded and executed by the processor to implement the steps of the HFCT discharge pulse recognition method based on the EWT-CNN algorithm as described in any one of claims 1-7.

9. An electronic device, characterized in that, The method includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or at least one program is loaded and executed by the processor to implement the steps of the HFCT discharge pulse recognition method based on the EWT-CNN algorithm as described in any one of claims 1-7.

Images