High-frequency partial discharge signal intelligent identification method

Abstract

The invention relates to the technical field of electrical variable measurement, in particular to a high-frequency partial discharge signal intelligent identification method, which comprises the following steps of: monitoring a partial discharge signal, and performing real-time calculation on signal frequency and energy to obtain frequency distribution data and energy data; and determining a main frequency band according to the peak value of the frequency distribution data, and generating main frequency band data. According to the invention, by monitoring the partial discharge signal in real time and calculating the frequency and energy of the partial discharge signal, frequency distribution data and energy data are provided, so that the signal processing is more accurate. Through peak analysis of frequency distribution data, a main frequency band can be determined more accurately, and the filter and transformation parameters are further optimized. The dynamic parameter adjustment enhances the response capability to the signal change, and improves the accuracy of signal classification. A new signal processing strategy can keep efficient operation in a constantly changing operation environment, the accuracy of fault prediction is enhanced, and health monitoring and maintenance decision making of power equipment are supported.

Description

Technical Field

[0001] The invention relates to the technical field of measuring electrical variables, and in particular to a method for intelligently identifying high-frequency partial discharge signals. Background Art

[0002] The high-frequency partial discharge signal recognition method is a technology specifically used to detect and analyze partial discharge phenomena in power equipment. Partial discharge is an electrical discharge phenomenon in the insulation material of power equipment, which usually indicates potential equipment failure.

[0003] However, existing technologies may not be sensitive enough or may have response delays when processing signals with high-frequency changes. Due to the lack of real-time data analysis capabilities and dynamic parameter adjustment, it is difficult to capture subtle changes in power equipment, which may lead to the omission of early fault signals. In addition, fixed parameter settings limit the adaptability of the method under different grid load changes, affecting the accuracy of signal analysis and the timeliness of fault judgment. These limitations increase the operation and maintenance costs of the power system and affect the stable operation of the power system. Summary of the invention

[0004] The purpose of the present invention is to solve the shortcomings in the prior art and to propose a high-frequency partial discharge signal intelligent recognition method.

[0005] In order to achieve the above object, the present invention adopts the following technical solution, a high-frequency partial discharge signal intelligent identification method, comprising the following steps: Monitor partial discharge signals, calculate the signal frequency and energy in real time, and obtain frequency distribution data and energy data; determine the main frequency band according to the peak value of the frequency distribution data, and generate the main frequency band data; Based on the main frequency band data, determining matching parameter settings, selecting filter types and transformation parameters, adjusting the filters and transformation parameters, and generating adjustment parameter data; applying the adjustment parameter data to the signal, recalculating the frequency distribution and energy level of the adjusted signal, and obtaining new spectrum data; According to the new spectrum data, performing detail feature extraction, calculating signal changes, and generating detail feature data; performing signal classification according to the detail feature data, analyzing classification accuracy, and obtaining classification effect data; The classification effect data is used to update the parameter adjustment strategy, optimize the parameter selection process, and obtain the recognition result.

[0006] Preferably, the steps of acquiring the frequency distribution data are: Monitor partial discharge signals through sensors and capture frequency points in real time and energy value , get the original signal data set; Based on the original signal data set, the energy level of each frequency point is calculated to form frequency distribution data and energy data. The calculation formula is: ; in, is the frequency point, is the corresponding energy value, is the total number of data points, for energy levels; Based on the frequency distribution data and energy data, the frequency distribution data and energy data are integrated and reconstructed to obtain a comprehensive data report.

[0007] Preferably, the steps of acquiring the main frequency band data are: Analyzing the frequency distribution data, identifying the energy peak of each frequency point, and obtaining a detailed peak data set; Based on the detailed peak data set, frequency points whose energy exceeds a preset threshold are selected, and preliminary main frequency band data are formed through the frequency points; Based on the preliminary main frequency band data, the energy and frequency characteristics of adjacent frequency points are analyzed to establish the main frequency band and obtain the main frequency band data.

[0008] Preferably, the step of acquiring the adjustment parameter data is: Extracting frequency characteristics from the main frequency band data, matching filter types according to the frequency characteristics, and obtaining a list of candidate filter types; Based on the candidate filter type list, by comparing the performance parameters of each type of filter, selecting an adaptive filter and transformation parameters to obtain a selected filter and parameter setting; According to the selected filter and parameter settings, the configuration of the filter is adjusted and the transformation parameters are optimized to generate adjustment parameter data.

[0009] Preferably, the steps of acquiring the new spectrum data are: Applying the adjustment parameter data to the original electrical signal, adjusting the parameter configuration of the filter so that the parameter configuration matches the adjustment parameter data, and obtaining an adjusted signal output; recalculating frequency distribution and energy level based on the adjusted signal output to obtain revised frequency and energy data; Information is extracted from the modified frequency and energy data to construct new spectrum data.

[0010] Preferably, the steps of acquiring the detailed feature data are: Analyze the new spectrum data, extract each frequency point and corresponding energy value, and generate a preliminary feature extraction report; Based on the preliminary feature extraction report, the overall change index between each frequency point is calculated, and the calculation formula is: ; in, Indicates the overall change index of the signal, is the energy value of the ith frequency point, is the average value of energy at all frequency points, is the frequency point, is the total number of frequency points; Based on the overall change index of the signal, the change of the signal is analyzed to generate detailed feature data.

[0011] Preferably, the steps of obtaining the classification effect data are: Analyze the detailed feature data, use a machine learning model to perform signal classification, and obtain classification effect data; Based on the classification effect data, the classification accuracy is calculated using the following formula: ; in, represents the overall classification accuracy, Indicates whether the i-th data point is correctly classified. is the weight of each data point, is the total number of data points; The classification accuracy is used to analyze and evaluate the classification effect and generate a classification effect data report.

[0012] Preferably, the steps of obtaining the recognition result are: Analyze the classification effect data report, analyze the efficiency and accuracy of the current parameter settings, and obtain a parameter evaluation report; Based on the parameter evaluation report, update the parameter adjustment strategy, reconfigure and optimize the parameters of the algorithm, and obtain an optimized parameter configuration; Apply the optimized parameter configuration, execute the recognition process, and obtain recognition result data.

[0013] Compared with the prior art, the advantages and positive effects of the present invention are: The present invention provides frequency distribution data and energy data by real-time monitoring of partial discharge signals and calculating their frequency and energy, thereby making signal processing more accurate. By analyzing the peak value of the frequency distribution data, the main frequency band can be determined more accurately, and the filter and transformation parameters can be further optimized. This dynamic parameter adjustment enhances the responsiveness to signal changes and improves the accuracy of signal classification. The new signal processing strategy can maintain efficient operation in a constantly changing operating environment, enhance the accuracy of fault prediction, and support health monitoring and maintenance decisions of power equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a schematic diagram of the steps of the present invention. DETAILED DESCRIPTION

[0015] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0016] See also Figure 1 The present invention provides a technical solution, a high-frequency partial discharge signal intelligent identification method, comprising the following steps: Monitor partial discharge signals, calculate the signal frequency and energy in real time, and obtain frequency distribution data and energy data; determine the main frequency band based on the peak value of the frequency distribution data, and generate the main frequency band data; Based on the main frequency band data, determine the matching parameter settings, select the filter type and the transformation parameters, adjust the filter and the transformation parameters, and generate the adjustment parameter data; apply the adjustment parameter data to the signal, recalculate the frequency distribution and energy level of the adjusted signal, and obtain new spectrum data; According to the new spectrum data, perform detail feature extraction, calculate the signal change, and generate detail feature data; perform signal classification according to the detail feature data, analyze the classification accuracy, and obtain classification effect data; Use the classification effect data to update the parameter adjustment strategy, optimize the parameter selection process, and obtain the recognition result.

[0017] The steps to obtain frequency distribution data are: Monitor partial discharge signals through sensors and capture frequency points in real time and energy value , get the original signal data set; Based on the original signal data set, the energy level of each frequency point is calculated to form frequency distribution data and energy data. The calculation formula is: ; in, is the frequency point, is the corresponding energy value, is the total number of data points, for energy levels; Based on the frequency distribution data and the energy data, the frequency distribution data and the energy data are integrated and reconstructed to obtain a comprehensive data report.

[0018] Specifically, the partial discharge signal is monitored by sensors to capture the frequency points in real time. and energy value , get the original signal data set, which records the frequency and energy value of the electrical signal at each time point in real time through the monitoring equipment. This step is the key initial data collection process to ensure the integrity and reliability of the basic data for subsequent analysis. This data set contains all the original electrical signal data output from the signal capture device, providing the necessary basic information for subsequent data processing and analysis; formula The benefit is that by increasing the square of the frequency and the square root of the energy, the sensitivity to the high energy value area in the signal is enhanced, so that the energy distribution characteristics of the electrical signal can be more accurately reflected; Calculation process: Given a signal data set, where Hz, Joules, total data points, the formula is calculated as follows: ; The results show that the total energy level in a given monitoring period is 35350 Joules, which indicates that the overall energy of the electrical signal during this period is high, and there may be high-energy discharge events, which require further analysis to determine the specific cause and location; Based on the energy level data, the frequency distribution data and energy data are integrated and reconstructed to generate a comprehensive data report. This process involves matching and fusing the initially calculated energy data with the frequency data. Through data integration software processing, the frequency and the corresponding energy value are combined to form a comprehensive data report containing the frequency-energy correspondence. This report will be directly used for subsequent data analysis and fault diagnosis, as well as providing engineers with a scientific basis for decision-making support. The report records in detail the energy level corresponding to each frequency point to facilitate analysts to conduct detailed data interpretation and analysis.

[0019] The steps to obtain the main frequency band data are: Analyze frequency distribution data, identify the energy peak at each frequency point, and obtain a detailed peak data set; Based on the detailed peak data set, the frequency points whose energy exceeds the preset threshold are selected, and the preliminary main frequency band data are formed through the frequency points; Based on the preliminary main frequency band data, the energy and frequency characteristics of adjacent frequency points are analyzed to establish the main frequency band and obtain the main frequency band data.

[0020] Specifically, in the process of in-depth analysis of frequency distribution data, identifying the energy peak of each frequency point is a key step. This process is carried out by scanning each frequency point and detecting whether its energy exceeds the environmental noise baseline. In order to accurately identify the peak, threshold screening is used. The threshold is based on the environmental baseline energy plus a preset difference setting. Each frequency point exceeding this threshold will be marked and recorded. In addition, in order to smooth the frequency data, the sliding window technology is used to process the data to reduce the peak misjudgment caused by short-term noise and ensure that only the real energy peak is recorded. These identified and recorded peak data are finally aggregated into a detailed peak data set.

[0021] The peak data set obtained is used to further analyze and screen out the main frequency points that exceed the set threshold. The core of this screening process is to compare the energy of each frequency point with the preset threshold. By calculating the energy of each point, it is confirmed which frequency points have energies significantly higher than the threshold. These selected frequency points constitute the preliminary main frequency band data. The screening method involves iterative analysis of each point in the data set to ensure that each selected frequency point meets the requirement that the energy is higher than the threshold. The collection of these points provides the basic data for determining the frequency band.

[0022] Based on the preliminary main frequency band data obtained through screening, an evaluation is conducted to determine the final main frequency band. This evaluation includes a comprehensive analysis of the energy relationship and frequency position between the selected frequency points to ensure that these frequency points appear as continuous and significant areas in the frequency distribution diagram. The frequency band boundaries are determined by comparing the energy and frequency characteristics of adjacent frequency points. During the evaluation process, special attention is paid to those frequency areas where the energy and position show high concentration and significance, ensuring that the final main frequency band has high cohesion and significance, thereby providing a scientific basis for generating accurate main frequency band data.

[0023] The steps to obtain the adjustment parameter data are: Extract frequency characteristics from the main frequency band data, match the filter type according to the frequency characteristics, and obtain a list of candidate filter types; Based on the candidate filter type list, by comparing the performance parameters of each type of filter, an adaptive filter and transformation parameters are selected to obtain a selected filter and parameter setting; According to the selected filter and parameter settings, the filter configuration is adjusted and the transformation parameters are optimized to generate adjustment parameter data.

[0024] Specifically, in the process of evaluating the data of the main frequency band, the energy values ​​of each frequency point in the frequency band are first identified, and the frequency points showing high energy peaks are focused on. These frequency points are the key basis for filter selection. According to the specific values ​​of these key frequency points, a candidate list containing multiple filter types is constructed. Each filter is accompanied by a detailed description of its frequency response and processing capabilities in the list, so as to ensure that the selected filter can accurately match the main energy distribution in the frequency band. This step is performed through detailed analysis and data matching of the frequency points; Based on the selected filter candidate list, in-depth parameter comparison is carried out, including the bandwidth, damping characteristics and frequency response of each filter. These parameters are measured and evaluated, and the filter type that best matches the main frequency band data is selected. At the same time, suitable transformation parameters are determined to optimize the performance of the filter and ensure that it can achieve good signal processing effects in specific applications. This selection process is achieved through a one-by-one comparison of the performance of each filter and a detailed analysis of the actual application requirements; After determining the most suitable filter and transformation parameters, perform detailed configuration adjustments of the filter, including setting the filter cutoff frequency, gain and phase to adapt to the signal characteristics of the main frequency band. These adjustments are completed through real-time performance testing and parameter optimization to ensure that the filter can achieve the designed performance indicators during operation. After the adjustment is completed, detailed adjustment parameter data is generated. These data are directly generated based on the measurement results and performance feedback during the filter adjustment process and will be used for further signal processing applications.

[0025] The steps to obtain new spectrum data are: Applying the adjustment parameter data to the original electrical signal, adjusting the parameter configuration of the filter so that the parameter configuration matches the adjustment parameter data, and obtaining an adjusted signal output; recalculating the frequency distribution and energy level based on the adjusted signal output to obtain corrected frequency and energy data; Extract information from the corrected frequency and energy data to construct new spectrum data.

[0026] Specifically, the adjusted parameter data is applied to the original electrical signal. First, the frequency response and gain setting of the filter are adjusted to ensure that the filter parameters are adapted to the new data, which includes adjusting the gain level and selecting a suitable frequency bandwidth. Through these operation steps, the overall quality and processing efficiency of the signal are improved. The stability of the signal is closely monitored during the adjustment process to avoid signal distortion caused by excessive parameter adjustment. Such adjustment enables the filter to effectively process the adjusted signal. Recalculate the frequency distribution and energy level of the adjusted signal. This process involves a detailed frequency component analysis of the signal, by specifically measuring the energy peaks of each major frequency point and analyzing the overall spectrum distribution. Extract key information from the adjusted frequency and energy data and construct new spectrum data. This operation involves screening and analyzing key frequency points and corresponding energy levels to construct a new spectrum diagram. This spectrum diagram shows in detail the frequency distribution and energy distribution of the signal after adjusting the parameters, providing data support for further signal processing and application.

[0027] The steps to obtain detailed feature data are: Analyze new spectrum data, extract each frequency point and corresponding energy value, and generate a preliminary feature extraction report; Based on the preliminary feature extraction report, the overall change index between each frequency point is calculated using the following formula: ; in, Indicates the overall change index of the signal, is the energy value of the ith frequency point, is the average value of energy at all frequency points, is the frequency point, is the total number of frequency points; Based on the overall change index of the signal, the signal changes are analyzed to generate detailed feature data.

[0028] Specifically, the energy value of each frequency point is analyzed, and the specific frequency and corresponding energy are recorded to facilitate subsequent feature identification. These frequency points are compared and analyzed to determine the mutual influence and change pattern between the frequency points. By identifying the peaks and abnormal points in the frequency, the main characteristics of the signal can be accurately depicted, providing reliable basic data for further data processing and analysis. In addition, based on the distribution of frequency point energy, the overall stability and volatility of the signal are evaluated. These detailed analyses help understand the problems that may arise during signal transmission, thereby providing guidance for correcting and optimizing the signal processing process.

[0029] The benefit of the formula is that it provides a sensitive and comprehensive measure of signal changes by weighting the energy deviation of frequency points, and can identify those frequency regions with significant changes, which helps further signal analysis and feature extraction.

[0030] Calculation process: There is a specific data set that contains energy values ​​at five frequency points: ; The corresponding frequencies are: ; First, calculate the average value of the energy: ; Then, calculate the deviation for each frequency point, multiply the square of the deviation by the frequency, and finally sum and take the square root: ; The overall change index of the signal obtained is about 1.825, which means that the energy deviation of the frequency points in the signal is generally medium. A value greater than a certain threshold, such as greater than 2, may indicate that the signal has undergone significant changes and requires further analysis or processing.

[0031] After obtaining the specific data of each frequency change, the signal is analyzed to identify the key change areas in the signal, which usually indicate possible signal interference or source changes; the energy changes of each frequency point and the relationship between these changes and the overall signal change indicators are analyzed; the identified characteristic change points will be used to adjust and set filters or other signal processing equipment to ensure that the equipment settings can minimize unnecessary signal loss and interference; in addition, through these analyses, more effective signal optimization strategies can be formulated to improve response speed and processing accuracy, providing support for the final application or research.

[0032] The steps to obtain classification effect data are: Analyze the detailed feature data, use the machine learning model to classify the signal, and obtain the classification effect data; Based on the classification effect data, the classification accuracy is calculated using the following formula: ; in, represents the overall classification accuracy, Indicates whether the i-th data point is correctly classified. is the weight of each data point, is the total number of data points; Use classification accuracy to analyze and evaluate classification results and generate classification results data reports.

[0033] Specifically, we analyze the detailed feature data, extract key features by analyzing the spectrum of the original signal data and combining the time-frequency characteristics, and use the decision tree in the deep learning algorithm for classification. The decision tree learns the influence of each feature on the classification result by layering, and gradually refines the final category judgment. In this process, the decision of each node is based on the calculation of feature weights and thresholds, ensuring the logic and scientificity of the classification decision. The classification model is cross-validated during the training stage, and the model parameters are optimized to improve the prediction accuracy of unknown data. formula The benefit is that by introducing the weight coefficient , taking into account the importance of each data point, making the calculation of classification accuracy more precise and personalized; each The value of is obtained by analyzing historical data based on the influence of the data point in the classification decision. It is obtained by comparing the actual classification result with the expected result, 1 indicates correct classification, and 0 indicates incorrect classification; Calculation process: There are 100 data points in the data set, and the weight of each point is Taking the average value as 0.5, 70 data points are classified correctly, and the formula becomes: ; The results show that the accuracy of the classification model is 35% on the current parameters and data set. If the accuracy is lower than expected, it is necessary to adjust the model parameters or optimize the data preprocessing steps to improve the model performance; Using the calculated classification accuracy, we analyze and evaluate the details of the overall classification effect and generate a comprehensive classification effect data report. Through further analysis of the classification results, we can identify the reasons why the model performs poorly on certain types of signals, such as insufficient feature extraction or insufficient training samples for certain categories of data. We can then adjust the feature extraction algorithm, increase sample data, optimize the algorithm's classification decision tree structure, and improve the flexibility and accuracy of weight adjustment, to ensure that the model can better generalize and adapt to new data environments in future applications.

[0034] The steps to obtain the recognition results are: Analyze the classification effect data report, analyze the efficiency and accuracy of the current parameter settings, and obtain a parameter evaluation report; Based on the parameter evaluation report, update the parameter adjustment strategy, reconfigure and optimize the algorithm parameters, and obtain the optimized parameter configuration; Apply the optimized parameter configuration, execute the recognition process, and obtain the recognition result data.

[0035] Specifically, by analyzing the classification effect data, we determined the efficiency and accuracy of the current parameter settings in different recognition tasks, evaluated the sensitivity of each parameter and its impact on the results, and compared the model performance under different parameter configurations. We found that low accuracy is often caused by specific parameter missettings. By adjusting these parameters, we improved the model's ability to recognize edge cases, optimized the generalization and robustness of the entire model, and formed a comprehensive parameter evaluation report. After receiving the parameter evaluation report, we analyzed and discussed the deficiencies pointed out in the report, and decided to make minor adjustments to specific parameters to adapt to data characteristics of different complexity. During the adjustment process, we paid special attention to the performance of the model on different data sets, and compared the model performance of the original parameters with the newly adjusted parameters. Through verification tests, we ensured that the adjustment of each parameter could improve the prediction accuracy. By applying the new optimized parameter configuration and through actual application tests, relevant data of the recognition results were collected, which were used to further verify the effectiveness of the parameter adjustment.

[0036] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A high-frequency partial discharge signal intelligent identification method, characterized in that: The following steps are involved: Monitor partial discharge signals, calculate the signal frequency and energy in real time, and obtain frequency distribution data and energy data; determine the main frequency band according to the peak value of the frequency distribution data, and generate the main frequency band data; Based on the main frequency band data, determining matching parameter settings, selecting filter types and transformation parameters, adjusting the filters and transformation parameters, and generating adjustment parameter data; applying the adjustment parameter data to the signal, recalculating the frequency distribution and energy level of the adjusted signal, and obtaining new spectrum data; According to the new spectrum data, performing detail feature extraction, calculating signal changes, and generating detail feature data; performing signal classification according to the detail feature data, analyzing classification accuracy, and obtaining classification effect data; The classification effect data is used to update the parameter adjustment strategy, optimize the parameter selection process, and obtain the recognition result.

2. The high-frequency partial discharge signal intelligent identification method according to claim 1 is characterized in that: The steps for obtaining the frequency distribution data are: Monitor partial discharge signals through sensors and capture frequency points in real time and energy value , get the original signal data set; Based on the original signal data set, the energy level of each frequency point is calculated to form frequency distribution data and energy data. The calculation formula is: ； in, is the frequency point, is the energy value, is the total number of data points, for energy levels; Based on the frequency distribution data and energy data, the frequency distribution data and energy data are integrated and reconstructed to obtain a comprehensive data report.

3. The high-frequency partial discharge signal intelligent identification method according to claim 1 is characterized in that: The steps for obtaining the main frequency band data are: Analyzing the frequency distribution data, identifying the energy peak of each frequency point, and obtaining a detailed peak data set; Based on the detailed peak data set, frequency points whose energy exceeds a preset threshold are selected, and preliminary main frequency band data are formed through the frequency points; Based on the preliminary main frequency band data, the energy and frequency characteristics of adjacent frequency points are analyzed to establish the main frequency band and obtain the main frequency band data.

4. The high-frequency partial discharge signal intelligent identification method according to claim 1, characterized in that: The steps for obtaining the adjustment parameter data are as follows: Extracting frequency characteristics from the main frequency band data, matching filter types according to the frequency characteristics, and obtaining a list of candidate filter types; Based on the candidate filter type list, by comparing the performance parameters of each type of filter, selecting an adaptive filter and transformation parameters to obtain a selected filter and parameter setting; According to the selected filter and parameter settings, the configuration of the filter is adjusted and the transformation parameters are optimized to generate adjustment parameter data.

5. The high-frequency partial discharge signal intelligent identification method according to claim 1, characterized in that: The steps for acquiring the new spectrum data are: Applying the adjustment parameter data to the original electrical signal, adjusting the parameter configuration of the filter so that the parameter configuration matches the adjustment parameter data, and obtaining an adjusted signal output; recalculating frequency distribution and energy level based on the adjusted signal output to obtain revised frequency and energy data; Information is extracted from the modified frequency and energy data to construct new spectrum data.

6. The high-frequency partial discharge signal intelligent identification method according to claim 1, characterized in that: The steps for obtaining the detailed feature data are as follows: Analyze the new spectrum data, extract each frequency point and corresponding energy value, and generate a preliminary feature extraction report; Based on the preliminary feature extraction report, the overall change index between each frequency point is calculated, and the calculation formula is: ； in, Indicates the overall change index of the signal, is the energy value of the ith frequency point, is the average value of energy at all frequency points, is the frequency point, is the total number of frequency points; Based on the overall change index of the signal, the change of the signal is analyzed to generate detailed feature data.

7. The high-frequency partial discharge signal intelligent identification method according to claim 1, characterized in that: The steps for obtaining the classification effect data are as follows: Analyze the detailed feature data, use a machine learning model to perform signal classification, and obtain classification effect data; Based on the classification effect data, the classification accuracy is calculated using the following formula: ； in, represents the overall classification accuracy, Indicates whether the i-th data point is correctly classified. is the weight of each data point, is the total number of data points; The classification accuracy is used to analyze and evaluate the classification effect and generate a classification effect data report.

8. The high-frequency partial discharge signal intelligent identification method according to claim 1, characterized in that: The steps for obtaining the recognition result are: Analyze the classification effect data report, analyze the efficiency and accuracy of the current parameter settings, and obtain a parameter evaluation report; Based on the parameter evaluation report, update the parameter adjustment strategy, reconfigure and optimize the parameters of the algorithm, and obtain an optimized parameter configuration; Apply the optimized parameter configuration, execute the recognition process, and obtain recognition result data.

Images