Visualization method for partial discharge original signal in online partial discharge detection device

Abstract

The invention provides a visualization method for an original partial discharge signal in an online partial discharge detection device. The method comprises the steps of obtaining the original partial discharge signal, preprocessing the signal, converting the format, carrying out visualization processing and carrying out wireless screen projection display. According to the method, a partial discharge signal is collected through a sensor and converted into a digital signal, denoising and normalization processing are carried out, the digital signal is converted into a format suitable for wireless screen projection, a visual image is generated by applying a preset algorithm, and the visual image is displayed through a wireless screen projection technology. According to the invention, the accuracy and real-time performance of partial discharge signal detection can be improved, and the practicability and user interaction experience of a partial discharge monitoring system are enhanced.

Description

Technical Field

[0001] The present invention relates to the field of on-line partial discharge detection devices, and more specifically, to a method for visualizing partial discharge original signals in an on-line partial discharge detection device. Background Art

[0002] Existing on-line partial discharge detection devices are mainly used to monitor partial discharge activities in power equipment, which is crucial for preventing equipment failures and extending equipment life. These devices usually collect analog signals generated by partial discharges through sensors and convert these signals into digital signals through an analog-to-digital conversion module. However, these digital signals often contain noise and are not easily directly observable and analyzable, so preprocessing is required to improve signal quality. In the signal preprocessing stage, common methods include wavelet threshold denoising and normalization processing to ensure the accuracy and consistency of the signals. The preprocessed signals need to be converted into a format suitable for wireless transmission, and this step involves video encoding and encapsulation operations. Nevertheless, existing technologies still have limitations in the visualization processing of signals, especially in converting complex partial discharge signals into intuitive images, lacking effective algorithms and methods to achieve the intuitive display and in-depth analysis of signals.

[0003] In the process of implementing the embodiments of the present invention, the inventors found that there are at least the following problems or defects in the existing technologies: existing partial discharge signal visualization methods often cannot fully capture the complex characteristics of signals, resulting in inaccurate or difficult-to-interpret visualization results; in addition, existing wireless screen mirroring technologies may have problems of delay and low efficiency when transmitting and displaying complexly processed partial discharge signals, affecting the real-time performance and effectiveness of monitoring. Summary of the Invention

[0004] The present invention provides a method for visualizing partial discharge original signals in an on-line partial discharge detection device, including:

[0005] Obtaining partial discharge original signals: collecting partial discharge original signals using the sensors of the on-line partial discharge detection device and converting them into digital signals;

[0006] Signal preprocessing: performing denoising and normalization processing on the obtained digital signals to improve signal quality;

[0007] Format conversion: converting the preprocessed signals into a format suitable for wireless screen mirroring for transmission and display;

[0008] Visualization processing: processing the signals in the converted format using a preset algorithm to generate a visualization image;

[0009] Wireless screen mirroring display: displaying the generated visualization image through wireless screen mirroring technology.

[0010] As a further improvement of the present application, the step of obtaining the original partial discharge signal includes:

[0011] Sensor selection and installation: According to the working environment and detection requirements of the on-line partial discharge detection device, select a sensor with specific performance and install it at the designated position;

[0012] Analog signal acquisition: Use the selected sensor to collect the analog electrical signals generated by partial discharge in real time;

[0013] Analog-to-digital conversion: Use the analog-to-digital conversion module to convert the collected analog signal into a digital signal at a specific sampling frequency and resolution.

[0014] As a further improvement of the present application, the signal preprocessing step includes:

[0015] Wavelet threshold denoising: Perform wavelet transform on the obtained digital signal. Let the original signal be f(t), and after wavelet transform, the wavelet coefficients Wf(a,b) are obtained, where a is the scale parameter and b is the translation parameter; Set the threshold T according to the noise characteristics and perform threshold processing on the wavelet coefficients to obtain the denoised wavelet coefficients Then reconstruct the signal through inverse wavelet transform to obtain the denoised signal

[0016] Normalization processing: Normalize the denoised signal. Let the denoised signal be The normalized signal is f norm (t), and use the formula

[0017]

[0018] Map its amplitude to the interval [0,1], where and are respectively the minimum and maximum values of.

[0019] As a further improvement of the present application, the format conversion step includes:

[0020] Protocol analysis: Analyze the wireless screen mirroring protocol to determine the supported video coding format and encapsulation format;

[0021] Encoding operation: Encode the preprocessed signal using the determined video coding format. The encoding process includes prediction, transformation, quantization, and entropy encoding operations;

[0022] Encapsulation operation: Encapsulate the encoded signal according to the determined encapsulation format and add metadata and index information.

[0023] As a further improvement of the present application, the visualization processing step includes:

[0024] Generative Adversarial Network GAN Pre-training: Construct a generator G and a discriminator D, and use a large amount of image data similar to the partial discharge signal features as the training set to pre-train the GAN; during the training process, the generator G learns to map the random noise z to a pseudo-image G(z), and the discriminator D learns to distinguish between real images x and pseudo-images G(z); the loss function of the discriminator D is:

[0025]

[0026] The loss function of the generator G is:

[0027]

[0028] where E represents expectation, p data (x) is the data distribution of real images, p z (z) is the distribution of random noise z, and by alternately optimizing L D and L G to make the two reach an adversarial balance;

[0029] Topological Data Analysis TDA Feature Extraction: Perform topological data analysis on the signal after converting the format. Let the signal data point set be X = {x 1 , x 2 , …, x n}, calculate the distance d(x i and x j ) between any two points x i and x j to obtain the distance matrix D = [d(x i , x j )] n×n , perform statistical analysis on the elements in the distance matrix D, select an appropriate distance threshold as the distance parameter ∈ according to the signal characteristics and analysis requirements, construct the Vietoris-Rips complex VR(X, ∈), and calculate its persistent homology group H i (VR(X, ∈)), i = 0, 1, … to obtain the signal topological features;

[0030] Fusion and Image Generation: Use the topological features extracted by TDA as additional information to input into the pre-trained generator G, and the generator G generates visual images according to these features and random noise.

[0031] As a further improvement of this application, in the pre-training of the Generative Adversarial Network GAN:

[0032] Network structure construction: The generator G adopts a transposed convolutional neural network structure, and gradually upsamples the low-dimensional noise vector through multiple transposed convolutional layers to generate a high-resolution image; the discriminator D adopts a convolutional neural network structure, and extracts features and discriminates the input image through multiple convolutional layers;

[0033] Optimizer selection: Use the Adam optimizer to update the network parameters of the generator G and the discriminator D, and set the learning rate to α and the momentum parameter to β 1 and β 2 . In each round of training, first fix the generator G, and use the optimizer to update the parameters of the discriminator D according to the loss function L D of the discriminator D; then fix the discriminator D, and use the optimizer to update the parameters of the generator G according to the loss function L G of the generator G.

[0034] As a further improvement of this application, in the topological data analysis TDA feature extraction:

[0035] Simplicial homology calculation: When calculating the persistent homology group H i (VR(X,∈)), use the simplicial homology algorithm. By constructing the boundary matrix B of the simplicial complex, perform elementary transformations on the boundary matrix B to transform it into a row echelon form matrix, and calculate the rank rank(B) of the row echelon form matrix;

[0036] The Betti numbers β i of different dimensions are determined by calculating the ranks of the corresponding dimensional boundary matrices. The formula is:

[0037] β i = rank(B i ) - rank(B i-1 ), where B i and B i-1 are the boundary matrices of dimension i and i-1 respectively. The 0-dimensional Betti number β 0 represents the number of connected components, and the 1-dimensional Betti number β 1 represents the number of holes, etc.;

[0038] Feature screening: According to the specific requirements of signal analysis, screen the extracted topological features; set the screening criteria, retain the topological features that stably exist at a larger scale, and remove the unstable or unimportant topological features caused by noise.

[0039] As a further improvement of this application, the wireless screen mirroring display step includes:

[0040] Connection establishment: Initialize the wireless communication module, search for available wireless screen mirroring receiving devices around, and perform authentication and connection negotiation according to the wireless screen mirroring protocol to establish a stable wireless connection;

[0041] Data transmission: The generated visual image data is fragmented and encapsulated according to the data packet format specified by the wireless screen mirroring protocol, and is sent to the receiving device through the established wireless connection at a specific transmission rate and error correction mechanism;

[0042] Image display: After the receiving device receives the visual image data, it unpacks and verifies according to the protocol, decodes the data, and outputs the decoded image data to the display device for display.

[0043] As a further improvement of this application, after the visualization processing step, it further includes the steps of real-time update and interaction of the visual image:

[0044] Real-time update: Continuously obtain the latest partial discharge raw signal, repeat the steps of obtaining the partial discharge raw signal, signal preprocessing, format conversion, and visualization processing, and generate a new visual image in real time;

[0045] Interaction operation: Set an interaction interface on the display interface of the receiving device, allowing the user to operate on the visual image through touch, gesture or other input methods, and update the visual image in real time according to the user's operation.

[0046] As a further improvement of this application, in the said interaction operation:

[0047] Zoom operation: The user obtains the zoom scale factor k through the input device of the receiving device, and performs a zoom transformation on the pixel matrix of the visual image. Let the pixel matrix of the original visual image be P, and the pixel matrix after zooming be P′. For the pixel point P(x, y), the coordinates (x′, y′) of the pixel point after zooming satisfy Calculate the value of P′(x′, y′) through the bilinear interpolation algorithm, so as to realize the zoom display of the image;

[0048] Translation operation: The user obtains the translation vector (Δx, Δy) through the input device. For the pixel point P(x, y) in the pixel matrix P of the visual image, the coordinates of the pixel point after translation become (x + Δx, y + Δy). If the new coordinates exceed the range of the original image, it is processed according to the set boundary processing method (such as filling the background color, etc.) to realize the translation display of the image.

[0049] The above embodiments of the present invention have at least the following beneficial effects: The visualization method of the partial discharge original signal in the on-line partial discharge detection device provided by the present invention can effectively improve the monitoring efficiency and accuracy of the partial discharge signal. By adopting wavelet threshold denoising and normalization processing technologies, the quality of the signal can be improved, noise interference can be reduced, the signal can be made clearer, and it is convenient for subsequent processing. At the same time, this method uses generative adversarial network (GAN) and topological data analysis (TDA) technologies to accurately extract key features from the partial discharge signal and generate intuitive visualization images, enabling monitoring personnel to quickly identify and analyze partial discharge activities, so as to take preventive measures in time and avoid equipment failures.

[0050] In addition, through wireless screen projection technology, this method can realize the real-time display and interactive operation of the visualization image, enhancing the user experience. Users can observe the changes of the partial discharge signal in real time and perform operations such as zooming and panning on the image through interactive methods such as touch and gestures to obtain more detailed signal information. This real-time update and interactive function makes the monitoring process more flexible and responsive, can improve the real-time performance and interactivity of monitoring, and helps to achieve the efficient management and maintenance of power equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] By referring to the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the drawings, several embodiments of the present invention are shown by way of illustration and not limitation, wherein:

[0052] Figure 1 It is a schematic flowchart of the visualization method of the partial discharge original signal in the on-line partial discharge detection device provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0053] The principles and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided only to enable those skilled in the art to better understand and then implement the present invention, and do not limit the scope of the present invention in any way. On the contrary, these embodiments are provided to make the present invention more thorough and complete, and to be able to convey the scope of the present invention fully to those skilled in the art.

[0054] Those skilled in the art know that the embodiments of the present invention can be implemented as a system, device, equipment, method or computer program product. Therefore, the present invention can be specifically implemented in the following forms, namely: completely hardware, completely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0055] It should be noted that the number of any elements in the attached drawings is for illustration rather than limitation, and any naming is only for distinction without any restrictive meaning.

[0056] The following refers to Figure 1 , Figure 1 which is a schematic flowchart of a method for visualizing partial discharge original signals in an on-line partial discharge detection device provided by an embodiment of the present invention. As Figure 1 shown, a method 100 for visualizing partial discharge original signals in an on-line partial discharge detection device includes:

[0057] Obtain partial discharge original signals 101: Use the sensor of the on-line partial discharge detection device to collect partial discharge original signals and convert them into digital signals to provide a data basis for subsequent processing.

[0058] Signal preprocessing 102: Denoise and normalize the obtained digital signals to improve the signal quality and prepare for subsequent visualization processing.

[0059] Format conversion 103: Convert the preprocessed signal into a format suitable for wireless screen mirroring for transmission and display.

[0060] Visualization processing 104: Use a preset algorithm to process the signal in the converted format to generate a visualization image.

[0061] Wireless screen mirroring display 105: Display the generated visualization image through wireless screen mirroring technology.

[0062] It should be noted that this method first involves using the sensor of the on-line partial discharge detection device to collect partial discharge original signals and convert them into digital signals. The partial discharge original signal here refers to the electrical signal generated by partial discharge activities, and the sensor refers to a device capable of detecting these electrical signals. Converting to a digital signal means converting the analog signal into a digital form that can be processed by a computer through an analog-to-digital converter (ADC).

[0063] Specifically, the selection and installation of the sensor need to be determined according to the working environment and detection requirements of the partial discharge detection device. For example, the sensor may need to have high sensitivity and a wide frequency response range to ensure accurate capture of partial discharge signals. Analog signal acquisition involves using these selected sensors to collect the analog electrical signals generated by partial discharge in real time, and analog-to-digital conversion involves converting these analog signals into digital signals using a specific sampling frequency and resolution for subsequent processing.

[0064] Preferably, the analog-to-digital conversion module can use a high-resolution ADC to ensure accurate signal conversion. For example, an ADC with 12 bits or higher can be selected to obtain better signal resolution.

[0065] Furthermore, the selection of the sampling frequency should be based on the Nyquist theorem, that is, at least twice the highest frequency component of the signal, to avoid aliasing. In practical applications, a suitable sampling frequency can be selected according to the characteristics of the PD signal and the monitoring requirements, such as between 1 MHz and 1 GHz, to ensure the integrity and accuracy of the signal.

[0066] In some embodiments, the step of obtaining the original PD signal includes:

[0067] Sensor selection and installation: According to the working environment and detection requirements of the on-line PD detection device, select a sensor with specific performance and install it at the designated position.

[0068] Analog signal acquisition: Use the selected sensor to collect the analog electrical signals generated by PD in real time.

[0069] Analog-to-digital conversion: Use the analog-to-digital conversion module to convert the collected analog signals into digital signals at a specific sampling frequency and resolution.

[0070] It should be noted that this implementation mode details the specific steps of obtaining the original PD signal, including sensor selection and installation, analog signal acquisition, and analog-to-digital conversion. Here, sensor selection refers to selecting a suitable sensor type according to specific working environments and detection requirements. Analog signal acquisition means using these sensors to collect the analog electrical signals generated by PD in real time. Analog-to-digital conversion is the process of converting these analog signals into digital signals.

[0071] Specifically, the selection of the sensor needs to consider its performance parameters, such as sensitivity, frequency response, withstand voltage level, etc., to ensure that the sensor can work stably in a specific working environment and accurately capture the PD signal. During analog signal acquisition, the sensor will monitor PD activities in real time and convert them into analog signals in the form of voltage or current. The analog-to-digital conversion module is responsible for converting these analog signals into digital signals at a specific sampling frequency and resolution. This process usually involves two key steps: sampling and quantization. The sampling frequency determines the signal update rate, while the quantization resolution affects the signal accuracy.

[0072] Preferably, high-sensitivity and wide-band types of sensors can be selected, such as pulse current sensors, ultrasonic sensors, earth current sensors, ultra-high frequency (UHF) sensors, etc., to adapt to different detection environments. During the analog signal acquisition process, a suitable sampling frequency can be set, for example, determined according to the highest frequency component of the signal, to ensure the complete acquisition of the signal. During analog-to-digital conversion, a high-resolution ADC can be selected to obtain higher signal accuracy.

[0073] Furthermore, in order to improve the signal-to-noise ratio of the signal, a filter can be added before analog-to-digital conversion to reduce the interference of high-frequency noise. The optimization of these steps can ensure the high-quality acquisition and conversion of partial discharge signals, providing an accurate data basis for subsequent signal processing and analysis.

[0074] In some embodiments, the signal preprocessing steps include:

[0075] Wavelet threshold denoising: Perform wavelet transform on the acquired digital signal. Let the original signal be f(t), and after wavelet transform, wavelet coefficients Wf(a,b) are obtained, where a is the scale parameter and b is the translation parameter. Set a threshold T according to the noise characteristics, perform threshold processing on the wavelet coefficients, and obtain the denoised wavelet coefficients. Then reconstruct the signal through inverse wavelet transform to obtain the denoised signal.

[0076] Normalization processing: Normalize the denoised signal. Let the denoised signal be The normalized signal is f norm (t), and use the formula

[0077]

[0078] Map its amplitude to the interval [0,1], where and are respectively the minimum and maximum values of.

[0079] It should be noted that this embodiment involves preprocessing the acquired digital signal, including two key steps: wavelet threshold denoising and normalization processing. Wavelet threshold denoising is a method of denoising signals using wavelet transform, and normalization processing refers to mapping the amplitude of the signal to a specific numerical interval, usually the interval [0,1], for subsequent processing. These steps are aimed at improving the signal quality and laying a foundation for visualization processing.

[0080] Specifically, in the process of wavelet threshold denoising, first perform wavelet transform on the digital signal to obtain wavelet coefficients. Wavelet transform is a mathematical method that can decompose the signal into components at different scales, facilitating the identification and processing of noise in the signal. Then, set a threshold according to the characteristics of the noise and process the wavelet coefficients to remove the noise. The setting of the threshold can be based on various strategies, such as hard threshold or soft threshold methods. After denoising, reconstruct the signal through inverse wavelet transform to obtain the denoised signal. Normalization processing maps the amplitude of the denoised signal to the interval [0,1], which involves the calculation of the minimum and maximum values of the signal and the corresponding amplitude adjustment.

[0081] Preferably, in wavelet threshold denoising, Daubechies wavelet or other suitable discrete wavelet transforms can be adopted. Which wavelet and the corresponding parameters (such as scale parameter and translation parameter) to choose can be determined according to the characteristics of the signal and the noise level. In the normalization process, a linear normalization method can be adopted, that is, by calculating the minimum and maximum values of the signal, the amplitude of the signal is linearly mapped to the interval [0,1].

[0082] Furthermore, in order to improve the denoising effect, a signal smoothing step can be added after wavelet threshold denoising, such as using methods like moving average or Gaussian smoothing. The refinement and optimization of these steps can further improve the quality of the signal and provide more accurate data for subsequent visualization processing.

[0083] In some embodiments, the format conversion step includes:

[0084] Protocol analysis: Analyze the wireless screen mirroring protocol to determine the supported video coding format and encapsulation format.

[0085] Encoding operation: Encode the preprocessed signal using the determined video coding format. The encoding process includes operations such as prediction, transformation, quantization, and entropy coding.

[0086] Encapsulation operation: Encapsulate the encoded signal according to the determined encapsulation format, and add necessary metadata and index information.

[0087] It should be noted that this implementation mode describes the process of converting the preprocessed signal into a format suitable for wireless screen mirroring, including three key steps: protocol analysis, encoding operation, and encapsulation operation. Protocol analysis refers to analyzing the video coding format and encapsulation format supported by the wireless screen mirroring protocol. The encoding operation involves video encoding of the signal, and the encapsulation operation is to add necessary metadata and index information to the encoded signal. These steps are to ensure that the signal can be correctly displayed on the wireless screen mirroring device.

[0088] Specifically, protocol analysis needs to determine the video coding format supported by the wireless screen mirroring protocol, such as H.264 / AVC, H.265 / HEVC, etc., and the encapsulation format, such as MP4, MKV, etc. The encoding operation includes steps such as prediction, transformation, quantization, and entropy coding. These steps act together on the preprocessed signal to convert it into a compressed video stream suitable for wireless transmission. The prediction step reduces the data volume by estimating the spatial or temporal redundancy in the signal. The transformation step converts the signal to the frequency domain for quantization. The quantization step reduces the precision of the encoded representation, and the entropy coding step further compresses the data. The encapsulation operation is to encapsulate the encoded video stream into a format that conforms to a specific protocol and add necessary metadata and index information so that the wireless screen mirroring device can identify and play it.

[0089] Preferably, H.265 / HEVC can be selected as the video encoding format during the encoding operation because it provides higher compression efficiency than H.264 / AVC. During the encapsulation operation, MP4 can be selected as the encapsulation format because it is widely supported by various devices and platforms. To improve the efficiency of wireless transmission, a multi-level encoding strategy can be adopted during the encoding process, that is, different levels of encoding are performed on the same video content to adapt to different network conditions and device performances.

[0090] Furthermore, to ensure the stability and compatibility of wireless screen mirroring, an error detection and correction mechanism, such as CRC checksum and Reed-Solomon coding, can be added during the encapsulation process to improve the reliability of data transmission. These refinements and alternatives can further enhance the effect of wireless screen mirroring and the user experience.

[0091] In some embodiments, the visualization processing step includes:

[0092] Generative adversarial network (GAN) pre-training: Construct a generator G and a discriminator D, and use a large amount of image data similar to the partial discharge signal characteristics as the training set to pre-train the GAN. During the training process, the generator G learns to map the random noise z to a pseudo-image G(z), and the discriminator D learns to distinguish between the real image x and the pseudo-image G(z). The loss function of the discriminator D is

[0093]

[0094] The loss function of the generator G is where E represents the expectation, p data (x) is the data distribution of the real image, p z (z) is the distribution of the random noise z, and by alternately optimizing L D and L G the two are made to reach an adversarial balance.

[0095] Topological data analysis (TDA) feature extraction: Perform topological data analysis on the signal after the conversion format. Let the signal data point set be X = {x 1 , x 2 , …, x n}, calculate the distance d(x i and x j ) between any two points x i and x j to obtain the distance matrix D = [d(x i , x j )] n×n, perform statistical analysis on the elements in the distance matrix D, select an appropriate distance threshold as the distance parameter ∈ according to the signal characteristics and analysis requirements, construct the Vietoris-Rips complex VR(X,∈), and calculate its persistent homology group H i (VR(X,∈)), i = 0, 1, … to obtain the topological features of the signal.

[0096] Fusion and image generation: Use the topological features extracted by TDA as additional information to input into the pre-trained generator G. The generator G generates visual images based on these features and random noise.

[0097] It should be noted that this embodiment involves using a generative adversarial network (GAN) and topological data analysis (TDA) to process signals and generate visual images. A generative adversarial network is a deep learning model composed of a generator and a discriminator, which is used to generate synthetic data similar to the real data distribution. Topological data analysis is a mathematical method used to extract the topological features of data. These steps work together to convert signals into visual images for easy analysis and display.

[0098] Specifically, during the pre-training of the generative adversarial network, the generator learns to map random noise to pseudo-images, while the discriminator learns to distinguish real images from pseudo-images. In this process, the loss functions of the generator and the discriminator respectively define their learning objectives. The loss function of the generator is based on the discriminator's judgment of the generated images, while the loss function of the discriminator is based on its classification accuracy for real and generated images. In topological data analysis feature extraction, the signal data point set is used to calculate the distance matrix and construct the Vietoris-Rips complex to obtain the topological features of the signal.

[0099] Preferably, the generator of the generative adversarial network can adopt the structure of multiple transposed convolutional layers, which is suitable for gradually upsampling low-dimensional noise vectors to generate high-resolution images. The discriminator can adopt the structure of multiple convolutional layers to extract the features of the input images and make judgments. In topological data analysis, the simplicial homology algorithm can be used to calculate the persistent homology group, which is a method to determine topological features by analyzing the boundary matrix of the simplicial complex.

[0100] Furthermore, in order to improve the accuracy of feature extraction, a feature screening step can be adopted. According to the specific requirements of signal analysis, retain the features that significantly contribute to signal visualization, such as topological features that stably exist at larger scales, and remove unstable or unimportant topological features caused by noise. These refinement and alternative solutions can further enhance the effect of signal visualization and the depth of analysis.

[0101] In some embodiments, during the pre-training of the generative adversarial network (GAN):

[0102] Network structure construction: The generator G adopts a transposed convolutional neural network structure, and gradually upsamples the low-dimensional noise vector through multiple transposed convolutional layers to generate a high-resolution image; the discriminator D adopts a convolutional neural network structure, and extracts features and discriminates the input image through multiple convolutional layers.

[0103] Optimizer selection: Use the Adam optimizer to update the network parameters of the generator G and the discriminator D, and set the learning rate to α and the momentum parameters to β 1 and β 2 . In each round of training, first fix the generator G, and use the optimizer to update the parameters of the discriminator D according to the loss function L D of the discriminator D; then fix the discriminator D, and use the optimizer to update the parameters of the generator G according to the loss function L G of the generator G.

[0104] It should be noted that this embodiment details the specific steps of the pre-training of the generative adversarial network (GAN), including network structure construction and optimizer selection. Here, network structure construction refers to constructing the specific architectures of the generator and the discriminator, and optimizer selection refers to determining the algorithm for updating network parameters. These steps are the key links for realizing GAN pre-training and directly affect the quality of the generated images and the efficiency of network training.

[0105] Specifically, in network structure construction, the generator can adopt a transposed convolutional neural network structure, which gradually upsamples the low-dimensional noise vector through multiple transposed convolutional layers to generate a high-resolution image. The discriminator adopts a convolutional neural network structure, and extracts features and discriminates the input image through multiple convolutional layers. In optimizer selection, the Adam optimizer can be used to update the network parameters of the generator and the discriminator. The Adam optimizer is an adaptive learning rate optimization algorithm that combines the advantages of the RMSProp and Momentum optimization methods and adjusts the learning rate by calculating the first-order moment estimate and the second-order moment estimate of the gradient.

[0106] Preferably, the network parameters of the generator and the discriminator can be further optimized by setting the learning rate, the momentum parameters β1 and β2. The learning rate determines the step size of network parameter update, and the momentum parameters β1 and β2 respectively control the exponential decay rates of the first-order moment and the second-order moment of the gradient. In each round of training, the generator can be fixed first, and the optimizer is used to update the parameters of the discriminator according to the loss function of the discriminator; then the discriminator is fixed, and the optimizer is used to update the parameters of the generator according to the loss function of the generator. This alternating optimization strategy can enable the GAN to achieve the ability to generate high-quality images and effectively discriminate.

[0107] Furthermore, a learning rate decay strategy can be considered to gradually reduce the learning rate as the training progresses, so as to improve the stability and convergence of the training. These refinements and alternative solutions can further enhance the effect of GAN pre-training and network performance.

[0108] In some embodiments, in the topological data analysis (TDA) feature extraction:

[0109] Simplicial homology calculation: Calculate the persistent homology group H i (VR(X,∈)) using the simplicial homology algorithm. By constructing the boundary matrix B of the simplicial complex and performing elementary transformations on the boundary matrix B to transform it into a row echelon form matrix, calculate the rank rank(B) of the row echelon form matrix. The Betti numbers β i in different dimensions are determined by calculating the ranks of the boundary matrices corresponding to the respective dimensions, β i = rank(B i ) - rank(B i-1 ), where B i and B i-1 are the boundary matrices in the i-th and (i - 1)-th dimensions respectively, and the 0-dimensional Betti number β 0 represents the number of connected components, and the 1-dimensional Betti number β 1 represents the number of holes, etc.

[0110] Feature screening: According to the specific requirements of signal analysis, screen the extracted topological features. Set screening criteria, for example, only retain the features that make significant contributions to signal visualization, such as topological features that stably exist at a larger scale, and remove unstable or unimportant topological features caused by noise.

[0111] It should be noted that this implementation details the specific steps of topological data analysis (TDA) feature extraction, including simplicial homology calculation and feature screening. Topological data analysis is a mathematical framework for analyzing the topological structure of data, and simplicial homology calculation is a method in TDA for identifying topological features such as holes and connectivity in data. Feature screening is the process of selecting important features based on analysis requirements.

[0112] Specifically, simplicial homology calculation involves constructing the boundary matrix of the simplicial complex and performing elementary transformations on it to calculate the persistent homology group. A simplicial complex is a geometric structure composed of basic elements such as points, lines, and surfaces, and the boundary matrix is a matrix that describes the relationships between these elements. By performing row echelon transformations on the boundary matrix, the Betti numbers in different dimensions can be determined, and these Betti numbers provide important information about the topological structure of the data. For example, the 0-dimensional Betti number represents the number of connected components, and the 1-dimensional Betti number represents the number of holes, etc.

[0113] Preferably, when performing simplicial homology calculations, efficient algorithms can be employed to handle large-scale datasets, such as using compressive sensing techniques to reduce computational complexity. During the feature screening process, screening criteria can be set according to the specific requirements of the signal. For example, topological features that stably exist at larger scales can be retained, and these features usually have an important impact on the global structure of the signal.

[0114] Furthermore, machine learning methods can be introduced to assist in feature screening by training a model to identify which topological features are most contributive to signal classification or prediction tasks. These refinements and alternatives can improve the efficiency and accuracy of feature extraction, thereby enhancing the effect of signal visualization.

[0115] In some embodiments, the wireless screen mirroring display step includes:

[0116] Connection establishment: Initialize the wireless communication module, search for available wireless screen mirroring receiving devices in the vicinity, and perform authentication and connection negotiation according to the wireless screen mirroring protocol to establish a stable wireless connection.

[0117] Data transmission: Fragment and encapsulate the generated visualization image data according to the data packet format specified by the wireless screen mirroring protocol, and send it to the receiving device through the established wireless connection at a specific transmission rate and error correction mechanism.

[0118] Image display: After the receiving device receives the visualization image data, it unpacks and verifies according to the protocol, decodes the data, and outputs the decoded image data to the display device for display.

[0119] It should be noted that this embodiment describes the specific steps of wireless screen mirroring display, including connection establishment, data transmission, and image display. Wireless screen mirroring refers to the process of transmitting image data from the sending end to the receiving end through wireless communication technology and displaying it on the display device at the receiving end. This process involves the initialization, authentication, connection negotiation of the wireless communication module, as well as the encapsulation, transmission, decoding, and display of data.

[0120] Specifically, in the connection establishment step, it is necessary to initialize the wireless communication module, search for available wireless screen mirroring receiving devices in the vicinity, and perform authentication and connection negotiation according to the wireless screen mirroring protocol to establish a stable wireless connection. The data transmission step involves fragmenting and encapsulating the generated visualization image data according to the data packet format specified by the wireless screen mirroring protocol and sending it to the receiving device through the established wireless connection at a specific transmission rate and error correction mechanism. The image display step is that after the receiving device receives the visualization image data, it unpacks and verifies according to the protocol, decodes the data, and outputs the decoded image data to the display device for display.

[0121] Preferably, in the connection establishment step, Wi-Fi Direct or Bluetooth technology can be adopted to achieve fast connection between devices. In the data transmission step, appropriate transmission rates and error correction mechanisms can be set, such as using Automatic Repeat reQuest (ARQ) or Forward Error Correction (FEC) to improve the reliability of data transmission. In the image display step, hardware acceleration decoding technology can be adopted to improve the decoding efficiency and ensure that the image data can be displayed quickly and accurately on the screen of the receiving device.

[0122] Furthermore, multiplexing technology can also be considered to be introduced, allowing multiple senders to transmit image data to the same receiving device simultaneously to support scenarios of multi-user collaboration or multi-view display. These refinements and alternative solutions can further enhance the performance and user experience of wireless screen mirroring display.

[0123] In some embodiments, after the visualization processing step, there is also a step of real-time update and interaction of the visualization image:

[0124] Real-time update: Continuously obtain the latest partial discharge raw signal, and repeat the steps of obtaining the partial discharge raw signal, signal preprocessing, format conversion, and visualization processing to generate a new visualization image in real time.

[0125] Interaction operation: Set an interaction interface on the display interface of the receiving device, allowing users to operate the visualization image through touch, gestures or other input methods, such as zooming, panning, switching display modes, etc., and update the visualization image in real time according to the user operation.

[0126] It should be noted that this implementation method involves real-time update and interaction operations after the visualization processing step. Real-time update means continuously obtaining the latest partial discharge raw signal and repeating the previous signal processing steps to generate a new visualization image. Interaction operation refers to allowing users to operate the visualization image through various input methods, such as zooming and panning, to meet the analysis needs of users. These steps enhance the interactivity and real-time nature of the system, enabling users to interact with the data more intuitively.

[0127] Specifically, in the real-time update step, the system will continuously monitor the partial discharge signal, collect new signal data in real time, and generate the latest visualization image according to the signal preprocessing, format conversion, and visualization processing steps described above. In the interaction operation step, users can operate the visualization image through touching the screen, gesture recognition or other input devices. For example, users can zoom in or out of the image through gestures, or pan the image by dragging to view information in a specific area in more detail.

[0128] Preferably, in the real-time update step, the system can adopt an event-driven or time-driven strategy to determine when to obtain new signal data. For example, the system can set a timer to automatically collect new signal data and update the image at regular time intervals (such as every second). In the interactive operation step, the system can provide various interaction methods, including multi-touch, voice control, etc., to adapt to different user operation habits.

[0129] Furthermore, the system can also provide some advanced interactive functions, such as multi-user collaborative editing, real-time data annotation, etc., to enhance the user experience and analysis efficiency. These refinement and alternative solutions can further improve the real-time performance and interactivity of the system, making it more suitable for complex monitoring and analysis tasks.

[0130] In some embodiments, in the said interactive operation:

[0131] Zoom operation: The user obtains the zoom scale factor k through the input device of the receiving device and performs a zoom transformation on the pixel matrix of the visualization image. Let the pixel matrix of the original visualization image be P, and the pixel matrix after zooming be P'. For the pixel point P(x, y), the coordinates (x', y') of the pixel point after zooming satisfy Calculate the value of P'(x', y') through the bilinear interpolation algorithm, thereby realizing the zoom display of the image.

[0132] Translation operation: The user obtains the translation vector (Δx, Δy) through the input device. For the pixel point P(x, y) in the pixel matrix P of the visualization image, the coordinates of the pixel point after translation become (x + Δx, y + Δy). If the new coordinates exceed the range of the original image, they are processed according to the set boundary processing method (such as filling the background color, etc.) to realize the translation display of the image.

[0133] It should be noted that this implementation mode describes the specific technical details in the interactive operation, especially the zoom operation and the translation operation. The zoom operation refers to the user adjusting the size of the visualization image through the input device, while the translation operation refers to the user moving the image to view the details of different regions. These operations enable the user to view and analyze the visualization image more flexibly.

[0134] Specifically, in the zoom operation, the user obtains the zoom scale factor through an input device such as a touch screen or a mouse and performs a zoom transformation on the pixel matrix of the visualization image. This means that the position of each pixel point will be adjusted according to the zoom scale factor to display the enlarged or reduced image. In the translation operation, the user obtains the translation vector through the input device, and this vector indicates the direction and distance of the image moving on the screen. Each pixel point in the image will be repositioned according to this vector.

[0135] Preferably, in the zoom operation, a bilinear interpolation algorithm can be used to calculate the values of the pixels after zooming to achieve a smooth image zoom effect. Bilinear interpolation is an image zooming technique that determines the values of new pixels by calculating the weighted average of four nearest neighbor pixels, thereby reducing image distortion during the zooming process. In the translation operation, if the coordinates of the pixels after translation exceed the range of the original image, boundary handling methods such as filling with the background color or repeating the edge pixels can be adopted to maintain the continuity of the image.

[0136] Furthermore, shortcuts or gesture recognition can also be provided to quickly perform zooming and translation operations, improving the convenience of user operations. These refinements and alternatives can further enhance the efficiency and experience of user interaction.

[0137] The above-mentioned various embodiments of the present invention have the following beneficial effects: The visualization method of the partial discharge original signal in the on-line partial discharge detection device described in the present invention can improve the efficiency and accuracy of partial discharge signal processing. Through steps such as sensor selection and installation, analog signal acquisition, and analog-to-digital conversion, this method can ensure the acquisition of high-quality partial discharge original signals from the source, laying a solid foundation for subsequent signal processing and visualization. The wavelet threshold denoising and normalization processing in the signal preprocessing step can further optimize the signal quality, making the signal more suitable for subsequent visualization processing. The format conversion step can ensure that the signal is compatible with wireless screen mirroring technology, facilitating wireless transmission and display of the signal.

[0138] Furthermore, the visualization processing step, including generative adversarial network (GAN) pre-training and topological data analysis (TDA) feature extraction, can effectively extract key features from the signal and generate intuitive visualization images, which can greatly enhance the readability and analysis depth of the signal. This method can not only improve the quality and efficiency of signal visualization, but also enable monitoring personnel to monitor and interact with partial discharge signals in real time through wireless screen mirroring display and real-time update and interaction functions, thereby realizing the efficient management and maintenance of power equipment.

[0139] Further, the storage medium according to the embodiments of the present application stores program instructions capable of implementing all the above methods. Among them, the program instructions can be stored in the above storage medium in the form of a software product, including several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods described in the various embodiments of the present application. The foregoing storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs, or terminal devices such as computers, servers, mobile phones, and tablets.

[0140] The above description is only some preferred embodiments of the present invention and an explanation of the technical principles applied. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present invention is not limited to the technical solutions formed by the specific combination of the above technical features, and should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above inventive concept. For example, the technical solutions formed by mutually replacing the above features with the technical features (but not limited to) having similar functions disclosed in the embodiments of the present invention.

Claims

1. A method for visualizing the original partial discharge signal in an online partial discharge detection device, characterized in that: The following steps are involved: Obtaining the original partial discharge signal: using the sensor of the online partial discharge detection device to collect the original partial discharge signal and convert it into a digital signal; Signal preprocessing: denoising and normalizing the acquired digital signal to improve signal quality; Format conversion: convert the pre-processed signal into a format suitable for wireless projection for transmission and display; Visualization processing: Use the preset algorithm to process the converted signal to generate a visual image; Wireless screen projection display: The generated visual images are displayed through wireless screen projection technology.

2. The method for visualizing the original partial discharge signal in the online partial discharge detection device according to claim 1, characterized in that: The step of obtaining the original partial discharge signal comprises: Sensor selection and installation: According to the working environment and detection requirements of the online partial discharge detection device, select sensors with specific performance and install them in the specified location; Analog signal acquisition: Use the selected sensor to collect the analog electrical signal generated by partial discharge in real time; Analog-to-digital conversion: The analog-to-digital conversion module is used to convert the collected analog signal into a digital signal at a specific sampling frequency and resolution.

3. The method for visualizing the original partial discharge signal in the online partial discharge detection device according to claim 1, characterized in that: The signal preprocessing step comprises: Wavelet threshold denoising: Perform wavelet transform on the acquired digital signal, set the original signal as f(t), and obtain the wavelet coefficient Wf(a,b) after wavelet transform, where a is the scale parameter and b is the translation parameter; set the threshold T according to the noise characteristics, perform threshold processing on the wavelet coefficient, and obtain the denoised wavelet coefficient Then reconstruct the signal through inverse wavelet transform to get the denoised signal Normalization: Normalize the denoised signal. Suppose the denoised signal is The normalized signal is f norm (t), using the formula Map its amplitude to the interval [0,1], where and They are The minimum and maximum values ​​of .

4. The method for visualizing the original partial discharge signal in the online partial discharge detection device according to claim 1, characterized in that: The format conversion step comprises: Protocol analysis: Analyze the wireless screen projection protocol to determine the video encoding format and packaging format it supports; Coding operation: using a certain video coding format to encode the preprocessed signal. The coding process includes prediction, transformation, quantization and entropy coding operations. Encapsulation operation: Encapsulate the encoded signal according to the determined encapsulation format, and add metadata and index information.

5. The method for visualizing the original partial discharge signal in the online partial discharge detection device according to claim 1, characterized in that: The visualization processing step comprises: Generative Adversarial Network (GAN) pre-training: Construct a generator G and a discriminator D, and use a large amount of image data with similar characteristics to partial discharge signals as a training set to pre-train the GAN. During the training process, the generator G learns to map random noise z to a pseudo image G(z), and the discriminator D learns to distinguish between real images x and pseudo images G(z). The loss function of the discriminator D is: The loss function of the generator G is: Among them, E represents expectation, p data (x) is the data distribution of real images, p z (z) is the distribution of random noise z, and by alternately optimizing L D and L G To achieve a counterbalancing balance between the two; Topological data analysis TDA feature extraction: Perform topological data analysis on the converted signal. Suppose the signal data point set is X = {x1, x2, ..., x n }, calculate any two points x i and x j The distance d(x i ,x j ), and get the distance matrix D = [d(x i ,x j )] n×n , perform statistical analysis on the elements in the distance matrix D, select an appropriate distance threshold as the distance parameter ∈ according to the signal characteristics and analysis requirements, construct the Vietoris-Rips complex VR(X,∈), and calculate its persistent homology group H i (VR(X,∈)), i=0,1,… to obtain signal topological characteristics; Fusion and image generation: The topological features extracted by TDA are input into the pre-trained generator G as additional information. The generator G generates visual images based on these features and random noise.

6. The method for visualizing the original partial discharge signal in the online partial discharge detection device according to claim 5, characterized in that: In the generative adversarial network GAN pre-training: Network structure construction: The generator G adopts a transposed convolutional neural network structure, which gradually upsamples the low-dimensional noise vector through multiple layers of transposed convolutional layers to generate a high-resolution image; the discriminator D adopts a convolutional neural network structure, which extracts and discriminates the input image through multiple layers of convolutional layers; Optimizer selection: Use Adam optimizer to update the network parameters of generator G and discriminator D, set the learning rate to α, momentum parameters β1 and β2. In each round of training, first fix the generator G, and use the optimizer to calculate the loss function L of the discriminator D. D Update the parameters of the discriminator D; then fix the discriminator D and use the optimizer to calculate the loss function L of the generator G G Update the parameters of the generator G.

7. The method for visualizing the original partial discharge signal in the online partial discharge detection device according to claim 5, characterized in that: In the topological data analysis TDA feature extraction: Simple homology computation: Computing the persistent homology group H i (VR(X,∈)), the simple homology algorithm is used. By constructing the boundary matrix B of the simplicial complex, the boundary matrix B is transformed into a row echelon matrix by elementary transformation, and the rank of the row echelon matrix rank(B) is calculated; Betti number β in different dimensions i It is determined by calculating the rank of the boundary matrix of the corresponding dimension, and the formula is: β i =rank(B i )-rank(B i-1 ), where B i and B i-1 are the i-dimensional and i-1-dimensional boundary matrices respectively, the 0-dimensional Betti number β0 represents the number of connected components, the 1-dimensional Betti number β1 represents the number of holes, and so on; Feature screening: Screen the extracted topological features according to the specific needs of signal analysis; The screening criteria are set to retain the topological features that are stable at a larger scale and remove the unstable or unimportant topological features caused by noise.

8. The method for visualizing the original partial discharge signal in the online partial discharge detection device according to claim 1, characterized in that: The wireless screen projection display step includes: Connection establishment: Initialize the wireless communication module, search for available wireless screen projection receiving devices in the surrounding area, and perform identity authentication and connection negotiation according to the wireless screen projection protocol to establish a stable wireless connection; Data transmission: The generated visual image data is fragmented and encapsulated according to the data packet format specified by the wireless projection protocol, and sent to the receiving device through the established wireless connection at a specific transmission rate and error correction mechanism; Image display: After receiving the visual image data, the receiving device unpacks and verifies it according to the protocol, decodes the data, and outputs the decoded image data to the display device for display.

9. The method for visualizing the original partial discharge signal in the online partial discharge detection device according to claim 1, characterized in that: After the visualization processing step, the real-time updating and interaction step of the visualization image is also included: Real-time update: continuously obtain the latest partial discharge original signal, repeat the steps of obtaining the partial discharge original signal, signal preprocessing, format conversion, and visualization processing, and generate new visualization images in real time; Interactive operation: An interactive interface is set on the display interface of the receiving device to allow the user to operate the visualized image through touch, gesture or other input methods, and the visualized image is updated in real time according to the user operation.

10. The method for visualizing the original partial discharge signal in the online partial discharge detection device according to claim 9, characterized in that: In the interaction: Scaling operation: The user obtains the scaling factor k through the input device of the receiving device and scales the pixel matrix of the visualized image. Let the pixel matrix of the original visualized image be P, and the scaled pixel matrix be P′. For the pixel point P(x, y), the scaled pixel point coordinates (x′, y′) satisfy The value of P′(x′,y′) is calculated by bilinear interpolation algorithm, so as to realize the zoom display of the image; Translation operation: The user obtains the translation vector (Δx, Δy) through the input device. For the pixel point P(x, y) in the visualized image pixel matrix P, the coordinates of the pixel point after translation become (x+Δx, y+Δy). If the new coordinates exceed the range of the original image, they are processed according to the set boundary processing method (such as filling the background color, etc.) to achieve image translation display.

Images