Method, device and computer program product for identifying a deterioration state of a surge arrester

Abstract

This application provides a method, device, and computer program product for identifying the deterioration state of a surge arrester. The method includes: acquiring real-time valid operating status data of the surge arrester; inputting the valid operating status data into a surge arrester deterioration state identification model, and outputting the current equipment status of the surge arrester, which includes current status, voltage status, and surge arrester status. The current status is normal current or abnormal current, the voltage status is normal voltage or abnormal voltage, and the surge arrester status is either deteriorated or normal. If the surge arrester status is deteriorated in the current equipment status, the application prompts the user to perform surge arrester maintenance measures. This application solves the problems in the prior art where most research is mainly based on simulation data, the deterioration conditions set deviate significantly from reality, and the identification of surge arrester deterioration state through intelligent algorithms deviates significantly from actual operating conditions, while also having low accuracy in deterioration state identification.

Description

Technical Field

[0001] This invention relates to the field of surge arrester degradation identification technology, and more specifically, to a method, device, computer-readable storage medium, and computer program product for identifying the degradation state of a surge arrester. Background Technology

[0002] During operation, surge arresters are susceptible to various factors such as lightning strikes, short circuits, service life, and environmental changes, which can alter their health status and even lead to deterioration. This deterioration directly impacts their safe and stable operation. Therefore, real-time monitoring and early warning of surge arrester deterioration are crucial technologies for improving equipment operating efficiency, extending equipment lifespan, and reducing maintenance costs. Currently, surge arrester deterioration monitoring largely relies on manual experience, judging based on the arrester's external operating conditions and internal parameters. This method is not only labor-intensive but also inaccurate and no longer meets the needs of surge arrester condition monitoring.

[0003] In recent years, with the development of deep learning technology, deep learning-based surge arrester degradation condition monitoring technology has been widely studied. In surge arrester degradation condition monitoring, signal acquisition devices collect operating status data of the surge arrester through sensors, and then use deep learning methods to extract features and classify the collected data, thereby realizing the identification of the surge arrester's degradation condition. Currently, deep learning-based surge arrester degradation condition monitoring mostly adopts Convolutional Neural Network (CNN) models. For example, CNN-based power equipment degradation condition monitoring methods can be divided into two main categories: single-scale and multi-scale. Single-scale methods extract multiple layers of features from the signal using a single-layer convolutional kernel, enabling high-precision feature extraction, but lacking multi-scale feature extraction, which easily leads to information loss. Multi-scale methods extract multi-scale features from the signal using multi-scale convolutional kernels, extracting multi-scale features, but this method cannot extract high-precision features, easily leading to inaccurate feature extraction.

[0004] It is evident that existing deep learning-based methods for monitoring the degradation status of surge arresters have low accuracy and speed, and cannot cope with the characteristics of high signal sampling rate, numerous sampling points, complex features, and large information volume in surge arrester degradation status monitoring, requiring further research. Current research on surge arrester degradation status identification using intelligent algorithms deviates significantly from actual operating conditions and suffers from low accuracy. Most studies are primarily based on simulation data, with the simulated degradation conditions deviating considerably from reality, while research on identification models based on measured data is scarce. Data acquisition is costly and challenging. Summary of the Invention

[0005] The main objective of this application is to provide a method, device, computer-readable storage medium, and computer program product for identifying the deterioration state of surge arresters, so as to at least solve the problems that most existing research is mainly based on simulation data, the deterioration conditions set are significantly different from reality, and the identification of the deterioration state of surge arresters through intelligent algorithms is significantly different from the actual working conditions, while the accuracy of deterioration state identification is not high.

[0006] To achieve the above objectives, according to one aspect of this application, a method for identifying the deterioration state of a surge arrester is provided, comprising: acquiring effective operating state data of the surge arrester in real time, the effective operating state data including at least effective voltage and current signals; inputting the effective operating state data into a surge arrester deterioration state identification model for analysis and identification using the surge arrester deterioration state identification model, and outputting the current equipment state of the surge arrester, wherein the surge arrester deterioration state identification model includes a data conversion layer, a feature extraction layer, and an output layer, the data conversion layer being used to convert data signals into graphs. For example, the feature extraction layer is used to extract and classify features from the image, and the output layer is used to output the equipment status of the surge arrester. The current equipment status includes current status, voltage status, and surge arrester status. The current status is normal current or abnormal current, the voltage status is normal voltage or abnormal voltage, and the surge arrester status is either surge arrester deterioration or surge arrester normal. If the surge arrester status is surge arrester deterioration in the current equipment status, a prompt is made to perform surge arrester maintenance measures. The surge arrester maintenance measures are at least replacing the surge arrester or repairing the surge arrester.

[0007] Optionally, obtaining the effective operating status data of the surge arrester includes: obtaining the operating status data of the surge arrester; and performing noise reduction processing on the operating status data using an overcomplete wavelet dictionary method to obtain the effective operating status data.

[0008] Optionally, before inputting the effective operating status data into the surge arrester deterioration status identification model, analyzing and identifying it using the surge arrester deterioration status identification model, and outputting the current equipment status of the surge arrester, the method further includes: constructing an initial surge arrester deterioration status identification model, wherein the initial surge arrester deterioration status identification model is a predetermined model to be trained; a first input step, inputting multiple sets of training data into the data transformation layer of the initial surge arrester deterioration status identification model, so as to use the data transformation layer to perform normalization processing and polar coordinate transformation processing on the sample operating status data in sequence, generating two... The initial surge arrester degradation state identification model consists of three training images, each set of training data including sample operating state data and the sample device state corresponding to the sample operating state data; the second input step involves inputting the two-dimensional training images into the feature extraction layer of the initial surge arrester degradation state identification model, so that feature extraction and classification are performed in the feature extraction layer using a densely connected convolutional network combined with a multi-scale channel attention mechanism to generate training output data; the third input step involves transmitting the training output data to the output layer of the initial surge arrester degradation state identification model, so that the output layer can convert the training output into the surge arrester degradation state identification model. The initial surge arrester degradation state identification model is obtained by analyzing the device state during the current training process. The analysis step involves performing bandpass filtering and sliding window slicing analysis based on the training device state to obtain a classification result, which is either correctly classified or incorrectly classified. The repetition step involves sequentially executing the first input step, the second input step, the third input step, and the analysis step at least once until a classification result for the training device state corresponding to each set of training data is obtained. The calculation step involves calculating the accuracy based on the classification results of the training device states corresponding to all training data to obtain the model accuracy. If the model accuracy is less than a set accuracy requirement, all parameters in the initial surge arrester degradation state identification model are adjusted, and the first input step, the second input step, the third input step, the analysis step, and the calculation step are sequentially executed at least once until the model accuracy is greater than or equal to the set accuracy requirement. If the model accuracy is greater than or equal to the set accuracy requirement, the trained initial surge arrester degradation state identification model is determined as the surge arrester degradation state identification model.

[0009] Optionally, multiple sets of training data are input into the data transformation layer of the initial surge arrester degradation state identification model, so that the data transformation layer can be used to perform normalization and polar coordinate transformation on the sample operating state data in sequence to generate a two-dimensional training image, including: normalizing the sample operating state data according to a first formula to obtain normalized operating state data, wherein the first formula is: The normalized operating status data, x i Let X represent the i-th sample running state data, maxX represent the maximum value among the sample running state data, and minX represent the minimum value among the sample running state data; according to the second formula, all the normalized running state data are transformed to obtain the polar coordinates corresponding to each normalized running state data, where the second formula is: θ i The angular cosine value of the polar coordinates corresponding to the i-th normalized running state data. r represents the set of all the normalized runtime state data. i Let t be the radius of the polar coordinates corresponding to the i-th normalized running state data. i x represents i The corresponding timestamp, M represents the normalization factor; the normalized running state data is reconstructed using Gram angle and field algorithms to obtain the two-dimensional training image.

[0010] Optionally, the feature extraction layer is constructed by introducing the multi-scale channel attention mechanism into the densely connected convolutional network. The two-dimensional training image is input into the feature extraction layer of the initial surge arrester degradation state identification model. In the feature extraction layer, feature extraction and classification are performed using the densely connected convolutional network combined with the multi-scale channel attention mechanism to generate training output data. This includes: decomposing and reconstructing the two-dimensional training image using a two-dimensional discrete wavelet algorithm to obtain a reconstructed two-dimensional training image; segmenting the reconstructed two-dimensional training image to obtain multiple segmented reconstructed images; extracting features from the multiple segmented reconstructed images through normalization, linear rectification, and pooling convolution operations in the feature extraction layer to obtain corresponding feature vectors; and classifying and regressing the feature vectors through a fully connected layer in the feature extraction layer to generate the training output data.

[0011] Optionally, bandpass filtering is performed based on the training device state, and slice analysis is performed using a sliding window approach to obtain the classification result of the training device state. This includes: dividing the state data corresponding to the training device state into multiple segments in sequence; obtaining the label value corresponding to each segment, wherein the sum of the label values ​​corresponding to all segments equals n; if the first segment data is greater than the second segment data, the classification result is correct, where the first segment data is the first segment data, and the second segment data is the remaining segment data excluding the first segment data; if the first segment data is less than the second segment data, the classification result is incorrect.

[0012] Optionally, a two-dimensional discrete wavelet algorithm is used to decompose and reconstruct the two-dimensional training image to obtain a reconstructed two-dimensional training image. This includes: decomposing the two-dimensional training image according to the decomposition algorithm to obtain multiple image signal components, wherein the image signal components include horizontal components, vertical components, and high-frequency components. The expression for the decomposition algorithm is as follows: This represents the approximation coefficient of the two-dimensional training image at the j-th layer. This represents the horizontal component of the two-dimensional training image at the j-th layer. This represents the vertical component of the two-dimensional training image at the j-th layer. This represents the high-frequency component of the two-dimensional training image in the j-th layer. It means that h l-2k h n-2m Each represents a coefficient of the wavelet low-pass filter used in the corresponding wavelet decomposition process, g l-2k g n-2m All represent the coefficients of the high-pass filter, k and m represent the position coordinates in the structure of the two-dimensional training image, k represents the row index, m represents the column index, and 1 and n both represent the index filter coefficients in the wavelet transform; the reconstructed two-dimensional training image is obtained by reconstructing the image based on multiple image signal components using a reconstruction algorithm, and the expression of the reconstruction algorithm is: These are the reconstructed data points, representing the approximation coefficients at the (j+1)th level of the wavelet transform.

[0013] According to another aspect of this application, a device for identifying the deterioration state of a surge arrester is provided. The device includes: an acquisition unit for acquiring valid operating state data of the surge arrester in real time, the valid operating state data including at least valid voltage and current signals; and an input / output unit for inputting the valid operating state data into a surge arrester deterioration state identification model for analysis and identification using the model, and outputting the current equipment state of the surge arrester. The surge arrester deterioration state identification model includes a data conversion layer, a feature extraction layer, and an output layer. The data conversion layer is used to convert data signals... The image is represented by a feature extraction layer for feature extraction and classification. The output layer is used to output the equipment status of the surge arrester. The current equipment status includes current status, voltage status, and surge arrester status. The current status is normal current or abnormal current, the voltage status is normal voltage or abnormal voltage, and the surge arrester status is either surge arrester deterioration or surge arrester normal. The prompting unit is used to prompt the execution of surge arrester maintenance measures when the surge arrester status is surge arrester deterioration in the current equipment status. The surge arrester maintenance measures are at least replacing the surge arrester or repairing the surge arrester.

[0014] According to another aspect of this application, a computer-readable storage medium is provided, the computer-readable storage medium including a stored program, wherein, when the program is executed, it controls the device on which the computer-readable storage medium is located to perform any of the methods described.

[0015] According to another aspect of this application, a computer program product is provided, including computer instructions that, when executed by a processor, implement any of the methods described.

[0016] Applying the technical solution of this application, the surge arrester degradation state identification model designed in this application is divided into a data conversion layer, a feature extraction layer, and an output layer. The data conversion layer transforms the signals in the effective operating state data into images. Then, the feature extraction layer completes the feature extraction and classification of the image data. Finally, the output layer outputs the equipment status of the surge arrester. For the surge arrester in operation, a signal is collected every set time interval and input into the trained surge arrester degradation state identification model for identification to determine the equipment status of the surge arrester. This application solves the problems in the prior art where most studies are mainly based on simulation data, the deterioration conditions set have a large deviation from reality, and the research on surge arrester degradation state identification through intelligent algorithms has a large deviation from the actual operating conditions, while also having low accuracy in degradation state identification. Attached Figure Description

[0017] Figure 1 A hardware structure block diagram of a mobile terminal for performing a method for identifying the deterioration state of a surge arrester, according to an embodiment of this application, is shown.

[0018] Figure 2 A flowchart illustrating a method for identifying the deterioration state of a surge arrester according to an embodiment of this application is shown.

[0019] Figure 3 The diagram illustrates the training process of a surge arrester degradation state identification model provided according to an embodiment of this application;

[0020] Figure 4 A surge arrester voltage signal diagram according to an embodiment of this application is shown;

[0021] Figure 5 A surge arrester current signal diagram according to an embodiment of this application is shown;

[0022] Figure 6 A GASF feature map of a voltage signal provided according to an embodiment of this application is shown;

[0023] Figure 7 A GASF feature map of a current signal provided according to an embodiment of this application is shown;

[0024] Figure 8 A schematic diagram of the encoding process of a Gram angle and field algorithm provided according to an embodiment of this application is shown;

[0025] Figure 9 A structural block diagram of a surge arrester degradation identification device provided according to an embodiment of this application is shown.

[0026] The above figures include the following reference numerals:

[0027] 102. Processor; 104. Memory; 106. Transmission device; 108. Input / output device. Detailed Implementation

[0028] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0029] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0030] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this application described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0031] As described in the background section, existing technologies include deep learning-based methods for monitoring the deterioration state of surge arresters. However, these methods have low accuracy and speed, and cannot address the challenges of high signal sampling rates, numerous sampling points, complex features, and large amounts of information in surge arrester deterioration state monitoring. To address the issues that most existing research is based on simulation data, which sets deterioration conditions that deviate significantly from reality, and that intelligent algorithms used to study surge arrester deterioration state identification also deviate significantly from actual operating conditions, while simultaneously exhibiting low accuracy in deterioration state identification, this application provides a method, device, computer-readable storage medium, and computer program product for identifying the deterioration state of surge arresters.

[0032] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0033] The methods and embodiments provided in this application can be executed on a mobile terminal, computer terminal, or similar computing device. Taking running on a mobile terminal as an example, Figure 1 This is a hardware structure block diagram of a mobile terminal for a method of identifying the deterioration state of a surge arrester according to an embodiment of the present invention. Figure 1 As shown, a mobile terminal may include one or more ( Figure 1 Only one is shown in the diagram. A processor 102 (which may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.) and a memory 104 for storing data are also shown. The mobile terminal may further include a transmission device 106 for communication functions and an input / output device 108. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the mobile terminal described above. For example, the mobile terminal may also include components that are more... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0034] The memory 104 can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to the lightning arrester degradation state identification method in this embodiment of the invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, thereby implementing the above-described method. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the mobile terminal via a network. Examples of the aforementioned networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof. The transmission device 106 is used to receive or send data via a network. Specific examples of the aforementioned networks may include wireless networks provided by the mobile terminal's communication provider. In one example, the transmission device 106 includes a network interface controller (NIC), which can be connected to other network devices via a base station to communicate with the Internet. In one example, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0035] This embodiment provides a method for identifying the deterioration state of a surge arrester that runs on a mobile terminal, computer terminal, or similar computing device. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0036] Figure 2 This is a flowchart of a method for identifying the deterioration state of a surge arrester according to an embodiment of this application. Figure 2 As shown, the method includes the following steps:

[0037] Step S201: Acquire the effective operating status data of the surge arrester in real time. The effective operating status data includes at least effective voltage and current signals.

[0038] Specifically, it is necessary to monitor and record the surge arrester's operating status data in real time using a signal acquisition device, and preprocess this data to ensure its validity, obtaining the aforementioned effective operating status data. Preprocessing operations may include steps such as noise removal, filtering, and normalization to improve data quality and reduce errors in model analysis. Preprocessed data will be clearer, helping the model to more accurately identify the surge arrester's status. This operating status data mainly includes the surge arrester's voltage and current signals; voltage and current are key parameters for evaluating the surge arrester's status.

[0039] Step S202: Input the above-mentioned valid operating status data into the surge arrester deterioration status identification model, so as to use the above-mentioned surge arrester deterioration status identification model for analysis and identification, and output the current equipment status of the surge arrester. The above-mentioned surge arrester deterioration status identification model includes a data conversion layer, a feature extraction layer and an output layer. The above-mentioned data conversion layer is used to convert the data signal into an image. The above-mentioned feature extraction layer is used to extract and classify the features of the above-mentioned image. The above-mentioned output layer is used to output the equipment status of the surge arrester. The above-mentioned current status includes current status, voltage status and surge arrester status. The above-mentioned current status is normal current or abnormal current. The above-mentioned voltage status is normal voltage or abnormal voltage. The above-mentioned surge arrester status is either surge arrester deterioration or surge arrester normal.

[0040] Specifically, the preprocessed voltage and current signal data (i.e., the aforementioned effective operating state data) are input into the surge arrester degradation state identification model. This model comprises three main parts: a data conversion layer, a feature extraction layer, and an output layer. The main task of the data conversion layer is to convert one-dimensional time-series data (voltage and current signals) into two-dimensional image data. This step is achieved using GASF (Gram Angle and Field Algorithm), encoding the time-series data into a two-dimensional image containing time information, providing input data for subsequent feature extraction and classification. The feature extraction layer uses the MSCAM-DenseNet network model to extract and classify features from the converted image. MSCAM (Multi-Scale Channel Attention) adaptively adjusts the weights of each layer in the convolutional neural network, increasing the weight of useful information and weakening the weight of useless features. DenseNet (Densely Connected Convolutional Network), through its special structure, connects each layer in the network to all preceding layers, effectively utilizing features, reducing the number of parameters, and improving performance. The output layer is responsible for outputting the current equipment state of the surge arrester, including the current state, voltage state, and surge arrester state. The current state can be normal or abnormal, the voltage state can be normal or abnormal, and the surge arrester state can be either degraded or normal. Based on the input data and extracted features, the model will output the current equipment state of the surge arrester. This state includes whether the current is normal, whether the voltage is normal, and whether the surge arrester is degraded.

[0041] Step S203: If the surge arrester is in a state of deterioration in the current equipment state, prompt the user to perform surge arrester maintenance measures. The surge arrester maintenance measures shall at least be replacing the surge arrester or repairing the surge arrester.

[0042] Specifically, when the model identifies a surge arrester as deteriorated, the system will prompt for maintenance measures. These measures include at least replacing the surge arrester or repairing it. This is a crucial step in ensuring the safe operation of the power system and can reduce power accidents caused by surge arrester failures. Through the above steps, real-time identification and timely maintenance of surge arrester deterioration can be achieved, thereby ensuring the stability and safety of the power system.

[0043] In this embodiment, firstly, the effective operating status data of the surge arrester is acquired in real time. This effective operating status data includes at least effective voltage and current signals. Then, the effective operating status data is input into the surge arrester deterioration status identification model for analysis and identification. The model outputs the current equipment status of the surge arrester. The surge arrester deterioration status identification model includes a data conversion layer, a feature extraction layer, and an output layer. The data conversion layer converts the data signal into an image. The feature extraction layer extracts and classifies the image. The output layer outputs the equipment status of the surge arrester. The current equipment status includes current status, voltage status, and surge arrester status. The current status is either normal or abnormal current. The voltage status is either normal or abnormal voltage. The surge arrester status is either deteriorated or normal. Finally, if the current equipment status indicates that the surge arrester is deteriorated, a surge arrester maintenance measure is prompted. This maintenance measure includes at least replacing the surge arrester or repairing it. The surge arrester degradation state identification model designed in this application consists of a data conversion layer, a feature extraction layer, and an output layer. The data conversion layer transforms the signals in the effective operating status data into images. Then, the feature extraction layer performs feature extraction and classification of the image data. Finally, the output layer outputs the equipment status of the surge arrester. For the surge arrester in operation, signals are collected every set time interval and input into the trained surge arrester degradation state identification model for identification to determine the equipment status of the surge arrester. This application solves the problems of existing technologies, where most studies are mainly based on simulation data, the deterioration conditions set deviate significantly from reality, and the intelligent algorithm research on surge arrester degradation state identification also deviates significantly from actual operating conditions, while the accuracy of degradation state identification is not high.

[0044] To enable those skilled in the art to better understand the technical solution of this application, the implementation process of the method for identifying the deterioration state of surge arresters in this application will be described in detail below with reference to specific embodiments.

[0045] To improve the signal quality of surge arrester operating data, in one optional embodiment, step S201 includes:

[0046] Step S2011: Obtain the operating status data of the surge arrester;

[0047] Step S2012: The above-mentioned running status data is denoised using an overcomplete wavelet dictionary method to obtain the above-mentioned effective running status data.

[0048] In the above embodiments, when the surge arrester begins to deteriorate, some of its operating data also begin to change, differing from the operating data under normal conditions. During the surge arrester deterioration identification process, it is necessary to collect surge arrester operating status data, using the collected voltage and current signals as the basic data for deterioration identification. The surge arrester status information data is processed by denoising data with indistinct signal characteristics to improve signal quality. An overcomplete wavelet dictionary is used to reconstruct and denoise the signal, enhancing its characteristics. The processed signal is represented as follows: Y′=DX′, where Y′ represents the denoised current or voltage signal, D represents the overcomplete wavelet dictionary, and X′ represents the estimated tree coefficient vector.

[0049] To improve the generalization ability of the surge arrester degradation state identification model, in an optional embodiment, before step S202 above, the method further includes:

[0050] Step S301: Construct an initial surge arrester degradation state identification model. The initial surge arrester degradation state identification model is a predetermined model to be trained.

[0051] Step S302, the first input step, inputs multiple sets of training data into the data conversion layer of the initial arrester deterioration state identification model, so as to use the data conversion layer to perform normalization and polar coordinate transformation processing on the sample operating state data in sequence to generate a two-dimensional training image. Each set of the above training data includes sample operating state data and the sample equipment state corresponding to the sample operating state data.

[0052] Step S303, the second input step, inputs the above two-dimensional training image into the feature extraction layer of the above initial lightning arrester deterioration state recognition model, so as to use densely connected convolutional networks combined with multi-scale channel attention mechanism to perform feature extraction and classification processing in the above feature extraction layer, and generate training output data.

[0053] Step S304, the third input step, transmits the above training output data to the above output layer of the above initial surge arrester deterioration state identification model, so as to use the above output layer to convert the above training output into the equipment state of the surge arrester obtained in the current training process, and obtain the training equipment state.

[0054] Step S305, Analysis Step: Based on the above training device status, bandpass filtering is performed and slice analysis is performed in a sliding window manner to obtain the classification result of the above training device status. The classification result is either correct or incorrect.

[0055] Step S306: Repeat the steps, sequentially executing the first input step, the second input step, the third input step, and the analysis step at least once, until the classification result of the training device state corresponding to each group of training data is obtained.

[0056] Step S307, Calculation step: Accuracy calculation is performed based on the classification results of the training device states corresponding to all the above training data to obtain the model accuracy.

[0057] Step S308: If the accuracy of the above model is less than the set accuracy requirement, adjust all parameters in the above initial surge arrester deterioration state identification model, and repeat the above first input step, the above second input step, the above third input step, the above analysis step and the above calculation step at least once in sequence until the accuracy of the above model is greater than or equal to the above set accuracy requirement.

[0058] Step S309: If the accuracy of the above model is greater than or equal to the above-set accuracy requirement, the trained initial arrester deterioration state identification model is determined as the arrester deterioration state identification model.

[0059] In the above embodiments, the classification problem is simplified according to actual application requirements. The surge arrester states correspond to the normal state and the deteriorated state in operation, thus simplifying the machine learning problem into a classification and recognition problem with 6 different operating conditions. Each operating condition corresponds to a label value, as shown in Table 1.

[0060] Table 1. Definitions of 6 Classification Problems

[0061] Tag value signal status Surge arrester status 0 Normal voltage, normal current normal 1 Normal voltage, normal current Deterioration 2 Abnormal voltage, normal current normal 3 Abnormal voltage, normal current Deterioration 4 Normal voltage, abnormal current normal 5 Normal voltage, abnormal current Deterioration

[0062] Design an initial surge arrester degradation state recognition model. This model includes a data transformation layer, a feature extraction layer, and an output layer. Model parameters need to be adjusted using training data to improve its recognition accuracy. Multiple sets of training data are input into the data transformation layer of the initial surge arrester degradation state recognition model. The sample operating state data is normalized to ensure a uniform numerical range for easier model processing. Polar coordinate transformation is performed on the normalized data to generate two-dimensional training images. The generated two-dimensional training images are input into the model's feature extraction layer. A densely connected convolutional network (DenseNet) combined with a multi-scale channel attention mechanism (MSCAM) is used for feature extraction and classification. The training output data is generated through the feature extraction layer. The MSCAM-DenseNet network sub-model uses the two-dimensional images generated by the GASF method as input to the entire model, then enters the feature extraction layer. It passes through a 7×7 convolutional layer followed by a 3×3 max pooling layer to optimize network parameters, and then connects to the DenseNet. An MSCAM attention mechanism is added to each DenseNet for adaptive weight allocation to improve the model's classification accuracy. The training output data is transformed into specific equipment states of the surge arrester using the output layer, i.e., any one of the six states listed in Table 1 above. Bandpass filtering is applied to the training equipment states to remove noise and irrelevant information. Slicing analysis is performed using a sliding window approach to obtain the classification results of the training equipment states. The classification result is either correct or incorrect. This process continues until the classification result for each group of training data is obtained. The overall classification accuracy of the model is defined as follows: if a total of C signal data points are collected, of which C0 are correctly classified and C1 are incorrectly classified, C = C0 + C1, then the model accuracy of the trained surge arrester degradation state identification model is C0 / C. The model accuracy is then checked against the set accuracy requirement. If the model accuracy is insufficient, all parameters in the model are adjusted. The above steps are repeated until the model accuracy is greater than or equal to the set accuracy requirement. When the model accuracy is greater than or equal to the set accuracy requirement, the trained model is determined as the surge arrester degradation state identification model. The training process is illustrated in the diagram below. Figure 3 As shown, through continuous iterative training and optimization of model parameters, a highly accurate arrester degradation state identification model is finally obtained.

[0063] It should be noted that attention mechanisms can adaptively adjust the weights of each layer in a convolutional neural network, increasing the weights of useful information and decreasing the weights of useless features. This reduces the number of computational parameters and improves the model's recognition accuracy. This invention proposes an improved attention mechanism: Multi-scale Channel Attention Mechanism (MSCAM). MSCAM simultaneously focuses on feature information through both local and global feature channels, then fuses the features. The resulting features contain information at different scales, making the extracted features more comprehensive and thus improving the model's recognition accuracy. The multi-scale channel attention mechanism first processes the input data (e.g., image or sequence data) and extracts features through filters of different scales (e.g., convolutional kernels). This can be achieved by using convolutional kernels of different sizes to capture different levels of visual detail, or by applying sliding window techniques with different window sizes to sequence data. The extracted multi-scale features are then fed into a specially designed network module that evaluates the importance of each feature channel and scale. Through the learned weights, the network can determine which features are more important and should be emphasized, and which features are less important and can be suppressed. Weights are automatically learned through backpropagation during training and typically depend on a task-specific loss function. Attention weights reflect the contribution of each feature in performing a specific task. In this way, the model can focus on those features that have the greatest impact on the predicted output. The weighted feature maps are then processed again in subsequent parts of the network to ultimately generate the output required for the task, such as classification labels and detection boxes. Introducing MSCAM into densely connected convolutional networks enables the original DenseNet to have multi-scale feature recognition capabilities. For a densely connected convolutional network, each layer repeatedly implements the nonlinear transformation H. t (·), where t represents the hierarchical index. H t The (·) function includes processes such as normalization (BN), linear rectification (ReLU), pooling, or convolution (Conv). The output of the "t-th layer" is represented as X. t Traditional feedforward neural networks connect the output of layer t as input to layer t+1, resulting in the following layer transformation: X t =H t (X t-1 By combining identity functions, nonlinear transformations can be avoided: X t =H t (X t-1 )+X t-1 Therefore, the t-th layer obtains the feature maps [X0, X1, ..., X] of the previous layers. t-1 ], and use it as input: X t =H t ([X0, X1, ..., Xt-1 Where, [X0, X1, ..., X] t-1 ] represents the concatenation of feature maps 0, 1, ..., t-1 generated in layer 0.

[0064] In order to preserve the time information in the original signal data, in an optional implementation, step S302 above includes:

[0065] Step S3021: Normalize the above sample running status data according to the first formula to obtain normalized running status data. The first formula is: This represents the above normalized operating status data, x i Let X represent the i-th sample running status data, maxX represent the maximum value among the sample running status data, and minX represent the minimum value among the sample running status data.

[0066] Step S3022: Transform all the above-mentioned normalized operating state data according to the second formula to obtain the polar coordinates corresponding to each of the above-mentioned normalized operating state data. The second formula is: θ i Let be the angle cosine value of the polar coordinates corresponding to the i-th normalized running state data. r represents the set of all the above-mentioned normalized running state data. i Let t be the radius of the polar coordinates corresponding to the i-th normalized running state data mentioned above. i x represents i The corresponding timestamp, where M represents the normalization factor;

[0067] Step S3023: The normalized running state data is reconstructed using Gram angle and field algorithms to obtain the two-dimensional training image.

[0068] In the above embodiments, the Gramian Angular Field (GASF) algorithm is used to encode the acquired signal data while preserving the temporal information of the signal. This signal conversion improves the accuracy and performance of downstream tasks. The principle of the GASF image encoding method is to use a Gram matrix to calculate the linear correlation of a set of vectors, encoding the time series into a two-dimensional image while retaining the temporal information contained within the time series. The acquired surge arrester signal is a typical one-dimensional time series, usually in a Cartesian coordinate system, where the horizontal axis represents time and the vertical axis represents the voltage or current signal corresponding to each time point, such as... Figure 4 and Figure 5 As shown, Figure 4 This is a diagram of the surge arrester voltage signal. Figure 5This is a waveform diagram of the surge arrester current signal. The signal is converted into graphical data using GASF, such as... Figure 6 and Figure 7 As shown, Figure 6 This is a GASF feature map of a voltage signal. Figure 7 This is a GASF characteristic map of a current signal. The implementation process is as follows: Figure 8 As shown, in step one, it is assumed that the preprocessed sample running status data is X = {x1, x2, x3, ..., x...} n-1 x n First, normalize the signal using the first formula and scale it to the range [-1, 1]. Second, convert the scaled signal from step one into polar coordinates. Pay attention to two key variables: the signal value and its corresponding timestamp. These variables can be represented by the angle and radius in polar coordinates, respectively. Encode the scaled signal value in polar coordinates using angle cosine, with values ​​between [0, π]. Encode its corresponding timestamp as the radius, as shown in the following formula: Divide the interval [0, 1] into several parts, discard the 0s, and associate the remaining points with the time series data, ensuring each point falls within a unit circle in polar coordinates. As time increases, the acquired signal sequence will continuously distort within this unit circle at different angles and radii. Step three: After the polar coordinate transformation in step two, the signal time series also contains time information, thus allowing the GASF algorithm to reconstruct the signal time series. The expression is as follows: This involves calculating the cosine of the angles between the midpoints of the polar coordinates to generate a matrix or image, namely the two-dimensional training image mentioned above. The resulting GASF feature map (i.e., the two-dimensional training image mentioned above) can be directly used as input data for the feature extraction layer. As can be seen from the equation, as time increases, the encoded position moves from the upper left corner of the matrix to the lower right corner, so the time dimension is encoded into the geometric structure of the matrix through GASF.

[0069] To ensure effective classification and regression prediction of the surge arrester's equipment status, in an optional implementation, step S303 includes:

[0070] Step S3031: The two-dimensional discrete wavelet algorithm is used to decompose and reconstruct the two-dimensional training image to obtain the reconstructed two-dimensional training image.

[0071] Step S3032: Segment the reconstructed two-dimensional training image to obtain multiple segmentation and reconstruction images;

[0072] Step S3033: Through the normalization operation, linear rectification operation, and pooling convolution operation in the above feature extraction layer, feature extraction is performed on multiple above segmentation and reconstruction maps to obtain the corresponding feature vectors;

[0073] Step S3034: Classify and regress the feature vectors through the fully connected layer in the feature extraction layer to generate the training output data.

[0074] In the above embodiment, a two-dimensional discrete wavelet algorithm is used to simultaneously perform one-dimensional wavelet transform on the rows and columns of the two-dimensional training image using a one-dimensional bioorthogonal filter bank, resulting in the reconstructed two-dimensional training image. This image contains the main feature information of the original image. Based on the image's features and structure, an appropriate image segmentation algorithm (such as threshold-based segmentation, region growing segmentation, clustering segmentation, etc.) is used to segment the image, resulting in multiple reconstructed segmented images, each representing a portion of the image's feature regions. The pixel values ​​of the reconstructed segmented images are normalized to reduce scale differences between different features and improve the model's generalization ability. The normalized feature map has each pixel value adjusted to the same scale. A non-linear transformation is applied to the normalized feature map using the ReLU activation function to introduce non-linear features and enhance the model's expressive power. The ReLU-activated feature map has negative values ​​set to 0 and positive values ​​remaining unchanged. A convolution operation is performed on the ReLU-activated feature map using a convolution kernel to extract local features. Pooling operations, such as max pooling or average pooling, are performed on the convolutional feature maps to reduce their spatial dimensionality and extract more abstract features. The resulting feature vectors, after pooling and convolution, contain key information about the image. These feature vectors are then fed into a fully connected layer for further feature integration and classification decisions. In the fully connected layer, the feature vectors are linearly combined using trained weights and biases, and then classified using an activation function (such as the Softmax function) or predicted via linear regression. Classification labels or regression output data are generated, representing the model's final predictions for the input image, enabling the model to extract key features from the original image and perform effective classification and regression predictions.

[0075] To improve the accuracy of the surge arrester degradation state identification model, in an optional implementation, step S305 includes:

[0076] Step S3051: Divide the state data corresponding to the above training device state into multiple segments in sequence;

[0077] Step S3052: Obtain the label value corresponding to each of the above-mentioned segmented data, wherein the sum of the label values ​​corresponding to all the above-mentioned segmented data is equal to n;

[0078] Step S3053: If the first segment data is greater than the second segment data, the above classification result is correct. The first segment data is the first segment data, and the second segment data is the remaining segment data excluding the first segment data.

[0079] Step S3054: If the data in the first segment is smaller than the data in the second segment, the classification result is a classification error.

[0080] In the above embodiments, the collected data is analyzed and classified, with 70% of the data selected as the training set and 30% as the test set. The test data is further analyzed under different operating conditions. Bandpass filtering is applied to the training set data, and it is sliced ​​using a sliding window. The sliced ​​data is divided into training and validation sets at ratios of 0.8 and 0.2, respectively. The trained model is then used to classify the complete training set data. The method for determining correct model classification is defined as follows: if a piece of data with a true label value of 0 (normal voltage signal, normal current signal, undegraded surge arrester) is divided into n segments by a sliding window, the model predicts the m0, m1, m2, m3, m4, and m5 segments of these n segments as having label values ​​of 0, 1, 2, 3, 4, and 5, respectively, where m0 + m1 + m2 + m3 + m4 + m5 = n. If m0 > m i If i = 1, 2, 3, 4, 5, then the classification result of the vibration data is correct; otherwise, the classification result is incorrect.

[0081] In order to extract useful information from the channel to the maximum extent, in an optional implementation, step S3031 above includes:

[0082] Step S30311: Decompose the above two-dimensional training image according to the decomposition algorithm to obtain multiple image signal components. The image signal components include horizontal components, vertical components, and high-frequency components. The expression of the decomposition algorithm is as follows: This represents the approximation coefficients of the above two-dimensional training image at the j-th layer. This represents the horizontal component of the aforementioned two-dimensional training image at layer j. This represents the vertical component of the aforementioned two-dimensional training image at layer j. This represents the high-frequency components of the aforementioned two-dimensional training image at the j-th layer. It means that h l-2k h n-2m Each represents a coefficient of the wavelet low-pass filter used in the corresponding wavelet decomposition process, g l-2k g n-2mAll represent the coefficients of the high-pass filter, k and m represent the position coordinates in the structure of the above two-dimensional training image, k represents the row index, m represents the column index, and 1 and n both represent the index filter coefficients in the wavelet transform;

[0083] Step S30312: Reconstruct the image using a reconstruction algorithm based on the multiple image signal components to obtain the reconstructed two-dimensional training image. The expression for the reconstruction algorithm is as follows: These are the reconstructed data points, representing the approximation coefficients at the (j+1)th level of the wavelet transform.

[0084] In the above embodiments, to maximize the extraction of useful information from each channel, the concept of two-dimensional discrete wavelet transform (2D-DWT) is used to fuse the advantageous information from different channels, thereby improving the utilization rate of two-dimensional feature information. The 2D-DWT method uses multiple wavelet filters for decomposition and reconstruction. Its decomposition algorithm formula is shown below. Here, k and m represent the position coordinates in the two-dimensional data structure, with k representing the row index and m representing the column index. 1 and n are indices, typically used to index filter coefficients in wavelet transform.

[0085] hl-2k, hn- 2m These represent the coefficients of the wavelet low-pass filter used in the corresponding wavelet decomposition process; the coefficients are used to calculate the approximation coefficients. g l-2k g n-2m These represent the coefficients of the high-pass filter, used to calculate the detail coefficients. The reconstruction algorithm formula is shown in the following equation: These are the reconstructed data points, representing the approximation coefficients at the (j+1)th level of the wavelet transform. Part 1 Use two low-pass filters h t-2k and h n-2m This process is used to process the approximation coefficients of the j-th layer to obtain the approximation coefficients of higher layers, preserving the low-frequency information of the data. Part Two By combining low-pass and high-pass filters, the horizontal detail coefficients of layer j are processed to extract and reconstruct high-frequency information in the horizontal direction. Part 3 These are high-pass and low-pass filters used to process the vertical detail coefficients of layer j, with the aim of extracting high-frequency information in the vertical direction. Part Four Two high-pass filters are used to process the diagonal detail coefficients of layer j, obtaining detailed information in the diagonal direction. As can be seen from the reconstruction and decomposition formulas, the essence of the 2D-DWT method is actually to use a one-dimensional bioorthogonal filter bank to simultaneously perform one-dimensional wavelet transform on the rows and columns of the two-dimensional signal.

[0086] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0087] This application also provides a device for identifying the deterioration state of a surge arrester. It should be noted that this device can be used to execute the method for identifying the deterioration state of a surge arrester provided in this application. This device is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0088] The following describes the surge arrester degradation identification device provided in the embodiments of this application.

[0089] Figure 9 This is a structural block diagram of a surge arrester degradation state identification device according to an embodiment of this application. Figure 9 As shown, the device includes:

[0090] The acquisition unit 10 is used to acquire the effective operating status data of the surge arrester in real time. The effective operating status data includes at least effective voltage signals and current signals.

[0091] Specifically, it is necessary to monitor and record the surge arrester's operating status data in real time using a signal acquisition device, and preprocess this data to ensure its validity, obtaining the aforementioned effective operating status data. Preprocessing operations may include steps such as noise removal, filtering, and normalization to improve data quality and reduce errors in model analysis. Preprocessed data will be clearer, helping the model to more accurately identify the surge arrester's status. This operating status data mainly includes the surge arrester's voltage and current signals; voltage and current are key parameters for evaluating the surge arrester's status.

[0092] The input / output unit 20 is used to input the above-mentioned valid operating status data into the surge arrester deterioration status identification model, so as to analyze and identify the surge arrester using the above-mentioned surge arrester deterioration status identification model, and output the current equipment status of the surge arrester. The surge arrester deterioration status identification model includes a data conversion layer, a feature extraction layer and an output layer. The data conversion layer is used to convert the data signal into an image. The feature extraction layer is used to extract and classify the features of the image. The output layer is used to output the equipment status of the surge arrester. The current equipment status includes current status, voltage status and surge arrester status. The current status is normal current or abnormal current. The voltage status is normal voltage or abnormal voltage. The surge arrester status is surge arrester deterioration or surge arrester normal.

[0093] Specifically, the preprocessed voltage and current signal data (i.e., the aforementioned effective operating state data) are input into the surge arrester degradation state identification model. This model comprises three main parts: a data conversion layer, a feature extraction layer, and an output layer. The main task of the data conversion layer is to convert one-dimensional time-series data (voltage and current signals) into two-dimensional image data. This step is achieved using GASF (Gram Angle and Field Algorithm), encoding the time-series data into a two-dimensional image containing time information, providing input data for subsequent feature extraction and classification. The feature extraction layer uses the MSCAM-DenseNet network model to extract and classify features from the converted image. MSCAM (Multi-Scale Channel Attention) adaptively adjusts the weights of each layer in the convolutional neural network, increasing the weight of useful information and weakening the weight of useless features. DenseNet (Densely Connected Convolutional Network), through its special structure, connects each layer in the network to all preceding layers, effectively utilizing features, reducing the number of parameters, and improving performance. The output layer is responsible for outputting the current equipment state of the surge arrester, including the current state, voltage state, and surge arrester state. The current state can be normal or abnormal, the voltage state can be normal or abnormal, and the surge arrester state can be either degraded or normal. Based on the input data and extracted features, the model will output the current equipment state of the surge arrester. This state includes whether the current is normal, whether the voltage is normal, and whether the surge arrester is degraded.

[0094] The prompting unit 30 is used to prompt the execution of surge arrester maintenance measures when the surge arrester is in a state of deterioration in the current equipment state. The surge arrester maintenance measures are at least replacing the surge arrester or repairing the surge arrester.

[0095] Specifically, when the model identifies a surge arrester as deteriorated, the system will prompt for maintenance measures. These measures include at least replacing the surge arrester or repairing it. This is a crucial step in ensuring the safe operation of the power system and can reduce power accidents caused by surge arrester failures. Through the above steps, real-time identification and timely maintenance of surge arrester deterioration can be achieved, thereby ensuring the stability and safety of the power system.

[0096] In this embodiment, the acquisition unit is used to acquire the effective operating status data of the surge arrester in real time. The effective operating status data includes at least effective voltage and current signals. The input / output unit is used to input the effective operating status data into the surge arrester deterioration status identification model for analysis and identification using the model, and output the current equipment status of the surge arrester. The surge arrester deterioration status identification model includes a data conversion layer, a feature extraction layer, and an output layer. The data conversion layer converts the data signal into an image. The feature extraction layer extracts and classifies the image. The output layer outputs the equipment status of the surge arrester. The current equipment status includes current status, voltage status, and surge arrester status. The current status is normal current or abnormal current. The voltage status is normal voltage or abnormal voltage. The surge arrester status is either surge arrester deterioration or surge arrester normal. The prompting unit is used to prompt the execution of surge arrester maintenance measures when the surge arrester status is surge arrester deterioration in the current equipment status. The surge arrester maintenance measures include at least replacing the surge arrester or repairing the surge arrester. The surge arrester degradation state identification model designed in this application consists of a data conversion layer, a feature extraction layer, and an output layer. The data conversion layer transforms the signals in the effective operating status data into images. Then, the feature extraction layer performs feature extraction and classification of the image data. Finally, the output layer outputs the equipment status of the surge arrester. For the surge arrester in operation, signals are collected every set time interval and input into the trained surge arrester degradation state identification model for identification to determine the equipment status of the surge arrester. This application solves the problems of existing technologies, where most studies are mainly based on simulation data, the deterioration conditions set deviate significantly from reality, and the intelligent algorithm research on surge arrester degradation state identification also deviates significantly from actual operating conditions, while the accuracy of degradation state identification is not high.

[0097] To improve the signal quality of surge arrester operating data, in one optional embodiment, the acquisition unit includes:

[0098] The first acquisition module acquires the operating status data of the aforementioned surge arrester;

[0099] The first preprocessing module uses an overcomplete wavelet dictionary method to denoise the above-mentioned running status data to obtain the above-mentioned effective running status data.

[0100] In the above embodiments, when the surge arrester begins to deteriorate, some of its operating data also begin to change, differing from the operating data under normal conditions. During the surge arrester deterioration identification process, it is necessary to collect surge arrester operating status data, using the collected voltage and current signals as the basic data for deterioration identification. The surge arrester status information data is processed by denoising data with indistinct signal characteristics to improve signal quality. An overcomplete wavelet dictionary is used to reconstruct and denoise the signal, enhancing its characteristics. The processed signal is represented as follows: Y′=DX′, where Y′ represents the denoised current or voltage signal, D represents the overcomplete wavelet dictionary, and X′ represents the estimated tree coefficient vector.

[0101] To improve the generalization ability of the surge arrester degradation state identification model, in one optional embodiment, the device further includes:

[0102] The construction unit is used to construct an initial surge arrester deterioration state identification model before inputting the above-mentioned effective operating status data into the surge arrester deterioration state identification model, using the above-mentioned surge arrester deterioration state identification model for analysis and identification, and outputting the current equipment status of the above-mentioned surge arrester. The above-mentioned initial surge arrester deterioration state identification model is a predetermined model to be trained.

[0103] The first input unit is used to perform the first input step, inputting multiple sets of training data into the data conversion layer of the initial arrester deterioration state identification model, so as to use the data conversion layer to perform normalization and polar coordinate transformation on the sample operating state data in sequence to generate a two-dimensional training image. Each set of the above training data includes sample operating state data and the sample equipment state corresponding to the sample operating state data.

[0104] The second input unit is used to perform the second input step, inputting the above two-dimensional training image into the feature extraction layer of the above initial lightning arrester deterioration state recognition model, so as to use a densely connected convolutional network combined with a multi-scale channel attention mechanism to perform feature extraction and classification processing in the feature extraction layer, and generate training output data.

[0105] The third input unit is used to perform the third input step, which transmits the training output data to the output layer of the initial arrester deterioration state identification model, so as to use the output layer to convert the training output into the equipment state of the arrester obtained in the current training process, and obtain the training equipment state.

[0106] The analysis unit is used to perform analysis steps, perform bandpass filtering based on the above training device status and perform slice analysis in a sliding window manner to obtain the classification result of the above training device status, which is either correctly classified or incorrectly classified.

[0107] The first repetition unit is used to perform repetition steps, and sequentially repeats the first input step, the second input step, the third input step and the analysis step at least once, until the classification result of the training device state corresponding to each group of training data is obtained.

[0108] The calculation unit is used to perform calculation steps, calculate the accuracy based on the classification results of the training device states corresponding to all the above training data, and obtain the model accuracy.

[0109] The second repeating unit is used to adjust all parameters in the initial arrester deterioration state identification model when the accuracy of the above model is less than the set accuracy requirement, and to repeat the above first input step, the above second input step, the above third input step, the above analysis step and the above calculation step at least once in sequence until the accuracy of the above model is greater than or equal to the above set accuracy requirement.

[0110] The determining unit is used to determine the trained initial arrester deterioration state identification model as the arrester deterioration state identification model when the accuracy of the above model is greater than or equal to the above-set accuracy requirement.

[0111] In the above embodiments, the classification problem is simplified according to actual application requirements. The surge arrester states correspond to the normal state and the deteriorated state in operation, thus simplifying the machine learning problem into a classification and recognition problem for six different operating conditions, as shown in Table 1.

[0112] Table 26 Definitions of Classification Problems

[0113]

[0114]

[0115] Design an initial surge arrester degradation state recognition model. This model includes a data transformation layer, a feature extraction layer, and an output layer. Model parameters need to be adjusted using training data to improve its recognition accuracy. Multiple sets of training data are input into the data transformation layer of the initial surge arrester degradation state recognition model. The sample operating state data is normalized to ensure a uniform numerical range for easier model processing. Polar coordinate transformation is performed on the normalized data to generate two-dimensional training images. The generated two-dimensional training images are input into the model's feature extraction layer. A densely connected convolutional network (DenseNet) combined with a multi-scale channel attention mechanism (MSCAM) is used for feature extraction and classification. The training output data is generated through the feature extraction layer. The MSCAM-DenseNet network sub-model uses the two-dimensional images generated by the GASF method as input to the entire model, then enters the feature extraction layer. It passes through a 7×7 convolutional layer followed by a 3×3 max pooling layer to optimize network parameters, and then connects to the DenseNet. An MSCAM attention mechanism is added to each DenseNet for adaptive weight allocation to improve the model's classification accuracy. The training output data is transformed into specific equipment states of the surge arrester using the output layer, i.e., any one of the six states listed in Table 1 above. Bandpass filtering is applied to the training equipment states to remove noise and irrelevant information. Slicing analysis is performed using a sliding window approach to obtain the classification results of the training equipment states. The classification result is either correct or incorrect. This process continues until the classification result for each group of training data is obtained. The overall classification accuracy of the model is defined as follows: if a total of C signal data points are collected, of which C0 are correctly classified and C1 are incorrectly classified, C = C0 + C1, then the model accuracy of the trained surge arrester degradation state identification model is C0 / C. The model accuracy is then checked against the set accuracy requirement. If the model accuracy is insufficient, all parameters in the model are adjusted. The above steps are repeated until the model accuracy is greater than or equal to the set accuracy requirement. When the model accuracy is greater than or equal to the set accuracy requirement, the trained model is determined as the surge arrester degradation state identification model. The training process is illustrated in the diagram below. Figure 3 As shown, through continuous iterative training and optimization of model parameters, a highly accurate arrester degradation state identification model is finally obtained.

[0116] It should be noted that attention mechanisms can adaptively adjust the weights of each layer in a convolutional neural network, increasing the weights of useful information and decreasing the weights of useless features. This reduces the number of computational parameters and improves the model's recognition accuracy. This invention proposes an improved attention mechanism: Multi-scale Channel Attention Mechanism (MSCAM). MSCAM simultaneously focuses on feature information through both local and global feature channels, then fuses the features. The resulting features contain information at different scales, making the extracted features more comprehensive and thus improving the model's recognition accuracy. The multi-scale channel attention mechanism first processes the input data (e.g., image or sequence data) and extracts features through filters of different scales (e.g., convolutional kernels). This can be achieved by using convolutional kernels of different sizes to capture different levels of visual detail, or by applying sliding window techniques with different window sizes to sequence data. The extracted multi-scale features are then fed into a specially designed network module that evaluates the importance of each feature channel and scale. Through the learned weights, the network can determine which features are more important and should be emphasized, and which features are less important and can be suppressed. Weights are automatically learned through backpropagation during training and typically depend on a task-specific loss function. Attention weights reflect the contribution of each feature in performing a specific task. In this way, the model can focus on those features that have the greatest impact on the predicted output. The weighted feature maps are then processed again in subsequent parts of the network to ultimately generate the output required for the task, such as classification labels and detection boxes. Introducing MSCAM into densely connected convolutional networks enables the original DenseNet to have multi-scale feature recognition capabilities. For a densely connected convolutional network, each layer repeatedly implements the nonlinear transformation H. t (·), where t represents the hierarchical index. H t The (·) function includes processes such as normalization (BN), linear rectification (ReLU), pooling, or convolution (Conv). The output of the "t-th layer" is represented as X. t Traditional feedforward neural networks connect the output of layer t as input to layer t+1, resulting in the following layer transformation: X t =H t (X t-1 By combining identity functions, nonlinear transformations can be avoided: X t =H t (X t-1 )+X t-1 Therefore, the t-th layer obtains the feature maps [X0, X1, ..., X] of the previous layers. t-1 ], and use it as input: X t =H t ([X0, X1, ..., Xt-1 Where, [X0, X1, ..., X] t-1 ] represents the concatenation of feature maps 0, 1, ..., t-1 generated in layer 0.

[0117] In order to preserve the time information in the original signal data, in one optional embodiment, the first input unit includes:

[0118] The second preprocessing module normalizes the above sample running status data according to the first formula to obtain normalized running status data. The first formula is: This represents the above normalized operating status data, x i Let X represent the i-th sample running status data, maxX represent the maximum value among the sample running status data, and minX represent the minimum value among the sample running status data.

[0119] The data transformation module transforms all the normalized operating state data according to the second formula to obtain the polar coordinates corresponding to each of the normalized operating state data. The second formula is as follows: θ i Let be the angle cosine value of the polar coordinates corresponding to the i-th normalized running state data. r represents the set of all the above-mentioned normalized running state data. i Let t be the radius of the polar coordinates corresponding to the i-th normalized running state data mentioned above. i x represents i The corresponding timestamp, where M represents the normalization factor;

[0120] The reconstruction module uses Gram angle and field algorithms to reconstruct the normalized running state data to obtain the two-dimensional training image.

[0121] In the above embodiments, the Gramian Angular Field (GASF) algorithm is used to encode the acquired signal data while preserving the temporal information of the signal. This signal conversion improves the accuracy and performance of downstream tasks. The principle of the GASF image encoding method is to use a Gram matrix to calculate the linear correlation of a set of vectors, encoding the time series into a two-dimensional image while retaining the temporal information contained within the time series. The acquired surge arrester signal is a typical one-dimensional time series, usually in a Cartesian coordinate system, where the horizontal axis represents time and the vertical axis represents the voltage or current signal corresponding to each time point, such as... Figure 4 and Figure 5 As shown, Figure 4 This is a diagram of the surge arrester voltage signal. Figure 5 This is a waveform diagram of the surge arrester current signal. The signal is converted into graphical data using GASF, such as... Figure 6 and Figure 7 As shown, Figure 6 This is a GASF feature map of a voltage signal. Figure 7 This is a GASF characteristic map of a current signal. The implementation process is as follows: Figure 8 As shown, in step one, it is assumed that the preprocessed sample running status data is X = {x1, x2, x3, ..., x...} n-1 x n First, normalize the signal using the first formula and scale it to the range [-1, 1]. Second, convert the scaled signal from step one into polar coordinates. Pay attention to two key variables: the signal value and its corresponding timestamp. These variables can be represented by the angle and radius in polar coordinates, respectively. Encode the scaled signal value in polar coordinates using angle cosine, with values ​​between [0, π]. Encode its corresponding timestamp as the radius, as shown in the following formula: Divide the interval [0, 1] into several parts, discard the 0s, and associate the remaining points with the time series data, ensuring each point falls within a unit circle in polar coordinates. As time increases, the acquired signal sequence will continuously distort within this unit circle at different angles and radii. Step three: After the polar coordinate transformation in step two, the signal time series also contains time information, thus allowing the GASF algorithm to reconstruct the signal time series. The expression is as follows: This involves calculating the cosine of the angles between the midpoints of the polar coordinates to generate a matrix or image, namely the two-dimensional training image mentioned above. The resulting GASF feature map (i.e., the two-dimensional training image mentioned above) can be directly used as input data for the feature extraction layer. As can be seen from the equation, as time increases, the encoded position moves from the upper left corner of the matrix to the lower right corner, so the time dimension is encoded into the geometric structure of the matrix through GASF.

[0122] To ensure effective classification and regression prediction of the surge arrester's equipment status, in one optional implementation, the second input unit includes:

[0123] The decomposition and reconstruction module uses a two-dimensional discrete wavelet algorithm to decompose and reconstruct the above two-dimensional training image to obtain the reconstructed two-dimensional training image.

[0124] The segmentation module segments the reconstructed two-dimensional training image to obtain multiple segmented reconstructed images.

[0125] The feature extraction module performs feature extraction on multiple segmented and reconstructed images through normalization, linear rectification, and pooling convolution operations in the feature extraction layer to obtain corresponding feature vectors.

[0126] The classification and regression module classifies and regresses the feature vectors through the fully connected layer in the feature extraction layer to generate the training output data.

[0127] In the above embodiment, a two-dimensional discrete wavelet algorithm is used to simultaneously perform one-dimensional wavelet transform on the rows and columns of the two-dimensional training image using a one-dimensional bioorthogonal filter bank, resulting in the reconstructed two-dimensional training image. This image contains the main feature information of the original image. Based on the image's features and structure, an appropriate image segmentation algorithm (such as threshold-based segmentation, region growing segmentation, clustering segmentation, etc.) is used to segment the image, resulting in multiple reconstructed segmented images, each representing a portion of the image's feature regions. The pixel values ​​of the reconstructed segmented images are normalized to reduce scale differences between different features and improve the model's generalization ability. The normalized feature map has each pixel value adjusted to the same scale. A non-linear transformation is applied to the normalized feature map using the ReLU activation function to introduce non-linear features and enhance the model's expressive power. The ReLU-activated feature map has negative values ​​set to 0 and positive values ​​remaining unchanged. A convolution operation is performed on the ReLU-activated feature map using a convolution kernel to extract local features. Pooling operations, such as max pooling or average pooling, are performed on the convolutional feature maps to reduce their spatial dimensionality and extract more abstract features. The resulting feature vectors, after pooling and convolution, contain key information about the image. These feature vectors are then fed into a fully connected layer for further feature integration and classification decisions. In the fully connected layer, the feature vectors are linearly combined using trained weights and biases, and then classified using an activation function (such as the Softmax function) or predicted via linear regression. Classification labels or regression output data are generated, representing the model's final predictions for the input image, enabling the model to extract key features from the original image and perform effective classification and regression predictions.

[0128] To improve the accuracy of the surge arrester degradation state identification model, in one optional implementation, the analysis unit includes:

[0129] The segmentation module divides the state data corresponding to the above training device states into multiple segments in sequence.

[0130] The second acquisition module acquires the label values ​​corresponding to each of the above-mentioned segmented data, wherein the sum of the label values ​​corresponding to all the above-mentioned segmented data equals n.

[0131] In the first determination module, if the first segment data is greater than the second segment data, the above classification result is considered correct. The first segment data refers to the first segment data, and the second segment data refers to the remaining segment data excluding the first segment data.

[0132] The second determination module determines that if the data in the first segment is less than the data in the second segment, the classification result is incorrect.

[0133] In the above embodiments, the collected data is analyzed and classified, with 70% of the data selected as the training set and 30% as the test set. The test data is further analyzed under different operating conditions. Bandpass filtering is applied to the training set data, and it is sliced ​​using a sliding window. The sliced ​​data is divided into training and validation sets at ratios of 0.8 and 0.2, respectively. The trained model is then used to classify the complete training set data. The method for determining correct model classification is defined as follows: if a piece of data with a true label value of 0 (normal voltage signal, normal current signal, undegraded surge arrester) is divided into n segments by a sliding window, the model predicts the m0, m1, m2, m3, m4, and m5 segments of these n segments as having label values ​​of 0, 1, 2, 3, 4, and 5, respectively, where m0 + m1 + m2 + m3 + m4 + m5 = n. If m0 > m i If i = 1, 2, 3, 4, 5, then the classification result of the vibration data is correct; otherwise, the classification result is incorrect.

[0134] In order to extract useful information from the channel to the maximum extent, in one optional implementation, the above-mentioned decomposition and reconstruction module includes:

[0135] The decomposition submodule decomposes the aforementioned two-dimensional training image according to the decomposition algorithm, obtaining multiple image signal components. These image signal components include horizontal components, vertical components, and high-frequency components. The expression for the decomposition algorithm is as follows: This represents the approximation coefficients of the above two-dimensional training image at the j-th layer. This represents the horizontal component of the aforementioned two-dimensional training image at layer j. This represents the vertical component of the aforementioned two-dimensional training image at layer j. This represents the high-frequency components of the aforementioned two-dimensional training image at the j-th layer. It means that h l-2k h n-2m Each represents a coefficient of the wavelet low-pass filter used in the corresponding wavelet decomposition process, g l-2k g n-2m All represent the coefficients of the high-pass filter, k and m represent the position coordinates in the structure of the above two-dimensional training image, k represents the row index, m represents the column index, and 1 and n both represent the index filter coefficients in the wavelet transform;

[0136] The reconstruction submodule reconstructs the image signal from multiple components using a reconstruction algorithm to obtain the reconstructed two-dimensional training image. The expression for the reconstruction algorithm is as follows: These are the reconstructed data points, representing the approximation coefficients at the (j+1)th level of the wavelet transform.

[0137] In the above embodiments, to maximize the extraction of useful information from each channel, the concept of two-dimensional discrete wavelet transform (2D-DWT) is used to fuse the advantageous information from different channels, thereby improving the utilization rate of two-dimensional feature information. The 2D-DWT method uses multiple wavelet filters for decomposition and reconstruction. Its decomposition algorithm formula is shown below. Here, k and m represent the position coordinates in the two-dimensional data structure, with k representing the row index and m representing the column index. 1 and n are indices, typically used to index filter coefficients in wavelet transforms. l-2k h n-2m These represent the coefficients of the wavelet low-pass filter used in the corresponding wavelet decomposition process; the coefficients are used to calculate the approximation coefficients. g l-2k g n-2m These represent the coefficients of the high-pass filter, used to calculate the detail coefficients. The reconstruction algorithm formula is shown in the following equation: These are the reconstructed data points, representing the approximation coefficients at the (j+1)th level of the wavelet transform. Part 1 Use two low-pass filters h t-2k and h n-2m This process is used to process the approximation coefficients of the j-th layer to obtain the approximation coefficients of higher layers, preserving the low-frequency information of the data. Part Two By combining low-pass and high-pass filters, the horizontal detail coefficients of layer j are processed to extract and reconstruct high-frequency information in the horizontal direction. Part 3 These are high-pass and low-pass filters used to process the vertical detail coefficients of layer j, with the aim of extracting high-frequency information in the vertical direction. Part Four Two high-pass filters are used to process the diagonal detail coefficients of layer j, obtaining detailed information in the diagonal direction. As can be seen from the reconstruction and decomposition formulas, the essence of the 2D-DWT method is actually to use a one-dimensional bioorthogonal filter bank to simultaneously perform one-dimensional wavelet transform on the rows and columns of the two-dimensional signal.

[0138] The aforementioned surge arrester degradation status identification device includes a processor and a memory. All the aforementioned units are stored as program units in the memory, and the processor executes these program units to achieve the corresponding functions. All the aforementioned modules are located in the same processor; alternatively, the aforementioned modules may be located in different processors in any combination.

[0139] The processor contains a kernel, which retrieves the corresponding program units from memory. One or more kernels can be configured, and their parameters can be adjusted to address issues arising from existing technologies. Most research in this field is based on simulation data, which sets deterioration conditions that deviate significantly from reality. Furthermore, intelligent algorithms used to study arrester deterioration status identification also show significant discrepancies with actual operating conditions, and the accuracy of deterioration status identification is not high.

[0140] The memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0141] This invention provides a computer-readable storage medium including a stored program, wherein, when the program is executed, it controls the device containing the computer-readable storage medium to perform the method for identifying the deterioration state of the surge arrester.

[0142] This invention provides a processor for running a program, wherein the program executes the method for identifying the deterioration state of a surge arrester.

[0143] This invention provides a surge arrester management system, which includes a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, it performs at least the following steps:

[0144] Step S201: Real-time acquisition of effective operating status data of the surge arrester, wherein the effective operating status data includes at least effective voltage and current signals;

[0145] Step S202: Input the above-mentioned valid operating status data into the surge arrester deterioration status identification model, so as to use the above-mentioned surge arrester deterioration status identification model for analysis and identification, and output the current equipment status of the surge arrester. The above-mentioned surge arrester deterioration status identification model includes a data conversion layer, a feature extraction layer and an output layer. The above-mentioned data conversion layer is used to convert the data signal into an image. The above-mentioned feature extraction layer is used to extract and classify the features of the above-mentioned image. The above-mentioned output layer is used to output the equipment status of the surge arrester. The above-mentioned current status includes current status, voltage status and surge arrester status. The above-mentioned current status is normal current or abnormal current. The above-mentioned voltage status is normal voltage or abnormal voltage. The above-mentioned surge arrester status is either surge arrester deterioration or surge arrester normal.

[0146] Step S203: If the surge arrester is in a state of deterioration in the current equipment state, prompt the user to perform surge arrester maintenance measures. The surge arrester maintenance measures shall at least be replacing the surge arrester or repairing the surge arrester.

[0147] This application also provides a computer program product, which, when executed on a data processing device, is suitable for executing an initialization program having at least the following method steps:

[0148] Step S201: Real-time acquisition of effective operating status data of the surge arrester, wherein the effective operating status data includes at least effective voltage and current signals;

[0149] Step S202: Input the above-mentioned valid operating status data into the surge arrester deterioration status identification model, so as to use the above-mentioned surge arrester deterioration status identification model for analysis and identification, and output the current equipment status of the surge arrester. The above-mentioned surge arrester deterioration status identification model includes a data conversion layer, a feature extraction layer and an output layer. The above-mentioned data conversion layer is used to convert the data signal into an image. The above-mentioned feature extraction layer is used to extract and classify the features of the above-mentioned image. The above-mentioned output layer is used to output the equipment status of the surge arrester. The above-mentioned current status includes current status, voltage status and surge arrester status. The above-mentioned current status is normal current or abnormal current. The above-mentioned voltage status is normal voltage or abnormal voltage. The above-mentioned surge arrester status is either surge arrester deterioration or surge arrester normal.

[0150] Step S203: If the surge arrester is in a state of deterioration in the current equipment state, prompt the user to perform surge arrester maintenance measures. The surge arrester maintenance measures shall at least be replacing the surge arrester or repairing the surge arrester.

[0151] It is obvious to those skilled in the art that the modules or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. They can be implemented using computer-executable program code, and thus can be stored in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those described herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0152] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0153] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0154] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0155] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0156] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0157] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0158] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0159] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0160] As can be seen from the above description, the embodiments of this application achieve the following technical effects:

[0161] 1) The method for identifying the deterioration state of a surge arrester according to this application firstly acquires the effective operating state data of the surge arrester in real time, wherein the effective operating state data includes at least effective voltage and current signals; then, the effective operating state data is input into a surge arrester deterioration state identification model for analysis and identification, and the current equipment state of the surge arrester is output. The surge arrester deterioration state identification model includes a data conversion layer, a feature extraction layer, and an output layer. The data conversion layer is used to convert data signals into images, and the feature extraction layer... The above-mentioned image is used for feature extraction and classification. The above-mentioned output layer is used to output the equipment status of the above-mentioned surge arrester. The current equipment status includes current status, voltage status, and surge arrester status. The current status is normal current or abnormal current, the voltage status is normal voltage or abnormal voltage, and the surge arrester status is either surge arrester deterioration or surge arrester normal. Finally, if the surge arrester status in the above-mentioned current equipment status is surge arrester deterioration, the surge arrester maintenance measures are prompted to be performed. The surge arrester maintenance measures are at least replacement of the surge arrester or repair of the surge arrester. The surge arrester deterioration status identification model designed in this application is divided into a data conversion layer, a feature extraction layer, and an output layer. The data conversion layer is used to transform the signal in the effective operating status data into an image. Then, the feature extraction layer is used to complete the feature extraction and classification of the image data. Finally, the equipment status of the surge arrester is output through the output layer. For the surge arrester in operation, a signal is collected once every set time period and input into the trained surge arrester deterioration status identification model for identification to determine the equipment status of the surge arrester in operation. This application addresses the problems in existing technologies where most studies are based on simulation data, the deterioration conditions set up deviate significantly from reality, and the identification of arrester deterioration status through intelligent algorithms deviates significantly from actual operating conditions, while also exhibiting low accuracy in deterioration status identification.

[0162] 2) The surge arrester degradation state identification device of this application is designed with a surge arrester degradation state identification model consisting of a data conversion layer, a feature extraction layer, and an output layer. The data conversion layer transforms the signals in the effective operating state data into images. Then, the feature extraction layer completes the feature extraction and classification of the image data. Finally, the output layer outputs the equipment status of the surge arrester. For the surge arrester in operation, a signal is collected every set time interval and input into the trained surge arrester degradation state identification model for identification to determine the equipment status of the surge arrester. This application solves the problems in the prior art where most studies are mainly based on simulation data, the deterioration conditions set have a large deviation from reality, and the identification of surge arrester degradation state through intelligent algorithms also has a large deviation from the actual operating conditions, as well as low accuracy in deterioration state identification.

[0163] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for identifying the deterioration state of a surge arrester, characterized in that, include: Real-time acquisition of effective operating status data of surge arresters, wherein the effective operating status data includes at least effective voltage and current signals; The effective operating status data is input into the surge arrester deterioration status identification model for analysis and identification. The current equipment status of the surge arrester is output. The surge arrester deterioration status identification model includes a data conversion layer, a feature extraction layer, and an output layer. The data conversion layer is used to convert data signals into images. The feature extraction layer is used to extract and classify features from the images. The output layer is used to output the equipment status of the surge arrester. The current equipment status includes current status, voltage status, and surge arrester status. The current status is normal current or abnormal current. The voltage status is normal voltage or abnormal voltage. The surge arrester status is either surge arrester deterioration or surge arrester normal. If the surge arrester is in a deteriorated state in the current equipment status, a prompt is made to perform surge arrester maintenance measures, which shall at least be replacing the surge arrester or repairing the surge arrester.

2. The method according to claim 1, characterized in that, Obtain valid operating status data of the surge arrester, including: Obtain the operating status data of the surge arrester; The overcomplete wavelet dictionary method is used to denoise the running status data to obtain the effective running status data.

3. The method according to claim 1, characterized in that, Before inputting the effective operating status data into the surge arrester deterioration status identification model, analyzing and identifying it using the surge arrester deterioration status identification model, and outputting the current equipment status of the surge arrester, the method further includes: Construct an initial surge arrester degradation state identification model, wherein the initial surge arrester degradation state identification model is a predetermined model to be trained; The first input step involves inputting multiple sets of training data into the data transformation layer of the initial surge arrester deterioration state identification model, so as to use the data transformation layer to perform normalization and polar coordinate transformation processing on the sample operating state data in sequence to generate a two-dimensional training image. Each set of training data includes sample operating state data and the sample equipment state corresponding to the sample operating state data. The second input step involves inputting the two-dimensional training image into the feature extraction layer of the initial lightning arrester degradation state identification model, so as to use a densely connected convolutional network combined with a multi-scale channel attention mechanism to perform feature extraction and classification processing in the feature extraction layer, and generate training output data. The third input step is to transmit the training output data to the output layer of the initial surge arrester degradation state identification model, so as to use the output layer to convert the training output into the equipment state of the surge arrester obtained in the current training process, and obtain the training equipment state. The analysis steps involve performing bandpass filtering based on the training device status and then performing slice analysis using a sliding window approach to obtain the classification result of the training device status, which is either a correct classification or an incorrect classification. Repeat the steps, sequentially executing the first input step, the second input step, the third input step, and the analysis step at least once, until the classification result of the training device state corresponding to each group of training data is obtained; The calculation step involves calculating the accuracy based on the classification results of the training device states corresponding to all the training data, thereby obtaining the model accuracy. If the model accuracy is less than the set accuracy requirement, adjust all parameters in the initial surge arrester deterioration state identification model, and repeat the first input step, the second input step, the third input step, the analysis step, and the calculation step at least once in sequence until the model accuracy is greater than or equal to the set accuracy requirement; If the accuracy of the model is greater than or equal to the set accuracy requirement, the trained initial arrester deterioration state identification model is determined as the arrester deterioration state identification model.

4. The method according to claim 3, characterized in that, Multiple sets of training data are input into the data transformation layer of the initial surge arrester degradation state identification model. The data transformation layer then performs normalization and polar coordinate transformation on the sample operating state data sequentially to generate a two-dimensional training image, including: The sample running status data is normalized according to the first formula to obtain normalized running status data. The first formula is: , This represents the normalized operating status data. This represents the running status data of the i-th sample. This represents the maximum value in the sample's running status data. This represents the minimum value in the sample's operational status data; According to the second formula, all the normalized operating state data are transformed to obtain the polar coordinates corresponding to each normalized operating state data. The second formula is: , The angular cosine value of the polar coordinates corresponding to the i-th normalized running state data. This represents the set of all the normalized running state data. The radius of the polar coordinates corresponding to the i-th normalized running state data. express The corresponding timestamp, where M represents the normalization factor; The normalized running state data is reconstructed using Gram angle and field algorithms to obtain the two-dimensional training image.

5. The method according to claim 3, characterized in that, The feature extraction layer is constructed by introducing the multi-scale channel attention mechanism into the densely connected convolutional network. The two-dimensional training image is input into the feature extraction layer of the initial lightning arrester degradation state identification model. In this layer, feature extraction and classification are performed using a densely connected convolutional network combined with the multi-scale channel attention mechanism to generate training output data, including: The two-dimensional training image is decomposed and reconstructed using a two-dimensional discrete wavelet algorithm to obtain a reconstructed two-dimensional training image. The reconstructed two-dimensional training image is segmented to obtain multiple segmentation and reconstruction images; The feature extraction layer performs normalization, linear rectification, and pooling convolution operations to extract features from multiple segmented and reconstructed maps, resulting in corresponding feature vectors. The feature vectors are classified and regressed through the fully connected layer in the feature extraction layer to generate the training output data.

6. The method according to claim 3, characterized in that, Based on the training device status, bandpass filtering is performed and slice analysis is conducted using a sliding window approach to obtain the classification results of the training device status, including: The state data corresponding to the state of the training device is divided into multiple segments in sequence; Obtain the label value corresponding to each of the segmented data, wherein the sum of the label values ​​corresponding to all the segmented data is equal to n; If the first segment data is greater than the second segment data, the classification result is considered correct. The first segment data is the first segment data, and the second segment data is the remaining segment data excluding the first segment data. If the data in the first segment is smaller than the data in the second segment, the classification result is a classification error.

7. The method according to claim 5, characterized in that, The two-dimensional training image is decomposed and reconstructed using a two-dimensional discrete wavelet algorithm to obtain a reconstructed two-dimensional training image, including: The two-dimensional training image is decomposed according to a decomposition algorithm to obtain multiple image signal components, including horizontal components, vertical components, and high-frequency components. The expression of the decomposition algorithm is as follows: , This represents the approximation coefficient of the two-dimensional training image at the j-th layer. This represents the horizontal component of the two-dimensional training image at the j-th layer. This represents the vertical component of the two-dimensional training image at the j-th layer. This represents the high-frequency component of the two-dimensional training image in the j-th layer. This represents the approximation coefficient of the two-dimensional training image at the (j+1)th layer. Each of these represents a coefficient of the wavelet low-pass filter used in the corresponding wavelet decomposition process. , All represent the coefficients of the high-pass filter, k and m represent the position coordinates in the structure of the two-dimensional training image, k represents the row index, m represents the column index, and l and n both represent the index filter coefficients in the wavelet transform; The reconstructed two-dimensional training image is obtained by reconstructing the image using a reconstruction algorithm based on multiple image signal components. The expression of the reconstruction algorithm is as follows: , These are the reconstructed data points, representing the approximation coefficients at the (j+1)th level of the wavelet transform.

8. A device for identifying the deterioration state of a surge arrester, characterized in that, The device includes: The acquisition unit is used to acquire the effective operating status data of the surge arrester in real time, wherein the effective operating status data includes at least effective voltage signals and current signals; The input / output unit is used to input the effective operating status data into the surge arrester deterioration status identification model, so as to analyze and identify the surge arrester using the surge arrester deterioration status identification model, and output the current equipment status of the surge arrester. The surge arrester deterioration status identification model includes a data conversion layer, a feature extraction layer and an output layer. The data conversion layer is used to convert the data signal into an image. The feature extraction layer is used to extract and classify the features of the image. The output layer is used to output the equipment status of the surge arrester. The current equipment status includes current status, voltage status and surge arrester status. The current status is normal current or abnormal current. The voltage status is normal voltage or abnormal voltage. The surge arrester status is either surge arrester deterioration or surge arrester normal. The prompting unit is used to prompt the execution of surge arrester maintenance measures when the surge arrester status is deteriorated in the current equipment status. The surge arrester maintenance measures are at least replacing the surge arrester or repairing the surge arrester.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein, when the program is executed, it controls the device on which the computer-readable storage medium is located to perform the method according to any one of claims 1 to 7.

10. A computer program product comprising computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images