EMD-PCA-SVM power transmission line abnormal discharge ultrahigh frequency monitoring method based on fusion of multiple parameters

Abstract

The invention discloses an EMD-PCA-SVM power transmission line abnormal discharge ultrahigh frequency monitoring method based on fusion of multiple parameters, and relates to the technical field of power system fault diagnosis, and the method comprises the steps: collecting the multi-parameter signal data of a power transmission line through an acoustic sensor, an optical sensor, an electromagnetic sensor and an ultrahigh frequency sensor; performing empirical mode decomposition on the multi-parameter signal data to obtain an intrinsic mode function, extracting characteristic parameters of the intrinsic mode function, and performing dimension reduction processing on the characteristic parameters through principal component analysis to obtain dimension reduction characteristic data; through organic combination of multi-parameter cooperative monitoring, EMD-PCA feature extraction and multi-core support vector machine classification, in a data acquisition link, a sound, light, electricity and ultrahigh frequency multi-parameter cooperative monitoring strategy is adopted, and a self-adaptive sampling mechanism is matched, so that the coverage and reliability of abnormal discharge detection are effectively improved.

Description

Technical Field

[0001] The invention relates to the technical field of power system fault diagnosis, and in particular to an ultra-high frequency monitoring method for abnormal discharge of a power transmission line based on EMD-PCA-SVM fusion of multiple parameters. Background Art

[0002] Abnormal discharge detection of transmission lines is a key link in the safe operation of power systems. UHF detection technology has been widely used in the field of abnormal discharge monitoring of transmission lines due to its high sensitivity and excellent anti-interference ability. At present, abnormal discharge detection of transmission lines mainly adopts UHF sensor monitoring method, optical monitoring method, acoustic monitoring method, current monitoring method and radio interference monitoring method. The UHF sensor monitoring method has high detection sensitivity and anti-interference ability, but its detection distance is limited and there is a directional effect; the optical monitoring method can directly detect the discharge phenomenon, but it is easily interfered by ambient light and the monitoring distance is limited; the acoustic monitoring method is not affected by electromagnetic interference, but the signal attenuation is fast and it is easily affected by environmental noise; the current monitoring method is suitable for the detection of live equipment, but the ability to distinguish the type of discharge is weak; the radio interference monitoring method has a wide coverage, but there is a problem of low positioning accuracy.

[0003] There are three main technical defects in the existing abnormal discharge detection methods for transmission lines: First, a single monitoring method is difficult to fully reflect the discharge characteristics, resulting in insufficient detection accuracy and reliability. Various monitoring methods have their own advantages and disadvantages, and a single monitoring method is difficult to meet the needs of high-precision detection. Secondly, traditional detection methods rely too much on manual experience, requiring professionals to judge the type of discharge based on experience, which makes it difficult to achieve intelligent and automated detection. Thirdly, the existing signal processing methods have limited feature extraction capabilities for nonlinear and non-stationary discharge signals, making it difficult to accurately extract and utilize the essential characteristics of the discharge signal, affecting the detection effect. These problems have seriously restricted the development and application of abnormal discharge detection technology for transmission lines. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the present invention provides an ultra-high frequency monitoring method for abnormal discharge of power transmission lines based on EMD-PCA-SVM fusion of multiple parameters, which can solve the problems mentioned in the background technology.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a UHF monitoring method for abnormal discharge of power transmission lines based on EMD-PCA-SVM fusion of multiple parameters, comprising: collecting multi-parameter signal data of the power transmission line through an acoustic sensor, an optical sensor, an electromagnetic sensor and a UHF sensor;

[0007] Performing empirical mode decomposition on the multi-parameter signal data to obtain an intrinsic mode function, extracting characteristic parameters of the intrinsic mode function, and performing dimensionality reduction processing on the characteristic parameters through principal component analysis to obtain reduced-dimensionality characteristic data;

[0008] The dimension-reduced feature data is input into a support vector machine classification model, the support vector machine classification model is trained and a classification recognition result of the abnormal discharge signal is output.

[0009] As a preferred solution of the UHF monitoring method for abnormal discharge of a power transmission line based on EMD-PCA-SVM fusion of multiple parameters described in the present invention, wherein: an acoustic sensor, an optical sensor, an electromagnetic sensor and an UHF sensor are installed on a tower of the transmission line at a first preset interval, the sampling frequency of the acoustic sensor is set to a first preset frequency, the wavelength response of the optical sensor is set to a first preset wavelength, the operating frequency of the electromagnetic sensor is set to a second preset frequency, and the operating frequency of the UHF sensor is set to a third preset frequency;

[0010] The signal data of the power transmission line is synchronously collected by the acoustic sensor, the optical sensor, the electromagnetic sensor and the ultra-high frequency sensor. If the amplitude of the collected signal of the acoustic sensor exceeds the first preset threshold value, the amplitude of the collected signal of the optical sensor exceeds the second preset threshold value, the amplitude of the collected signal of the electromagnetic sensor exceeds the third preset threshold value, or the amplitude of the collected signal of the ultra-high frequency sensor exceeds the fourth preset threshold value, the sampling frequency of the sensor is increased to the fourth preset frequency, otherwise, the data is collected according to the fifth preset frequency;

[0011] The signal data collected by the sensor is digitally converted according to a sixth preset frequency, and the digitally converted signal data is synchronously integrated according to timestamp information to obtain multi-parameter signal data.

[0012] As a preferred solution of the ultra-high frequency monitoring method for abnormal discharge of power transmission lines based on EMD-PCA-SVM fusion of multiple parameters described in the present invention, the local maximum points in the multi-parameter signal data are connected to form an upper envelope curve through cubic spline interpolation, and the local minimum points are connected to form a lower envelope curve through cubic spline interpolation, the mean curve of the upper envelope curve and the lower envelope curve is calculated and subtracted from the multi-parameter signal data to obtain a candidate component, if the difference between the number of local extreme value points and the number of zero-crossing points of the candidate component is greater than the seventh preset threshold, the candidate component is used as a new signal to be processed to repeatedly perform this step operation, if the difference between the number of local extreme value points and the number of zero-crossing points is less than or equal to the seventh preset threshold, the candidate component is determined as a first intrinsic mode function, the multi-parameter signal data is subtracted from the first intrinsic mode function to obtain a residual signal, if the number of extreme value points of the residual signal is greater than the eighth preset threshold, the residual signal is used as a new signal to be processed to repeatedly perform this step operation until the intrinsic mode function set is obtained.

[0013] As a preferred solution of the ultra-high frequency monitoring method for abnormal discharge of power transmission lines based on EMD-PCA-SVM fusion of multiple parameters described in the present invention, characteristic parameters are extracted from each intrinsic mode function in the intrinsic mode function set respectively, including: calculating the instantaneous frequency of the intrinsic mode function through Hilbert transform to obtain frequency characteristics, calculating the energy distribution of the intrinsic mode function to obtain energy characteristics, extracting the amplitude and duration of the intrinsic mode function to obtain time domain characteristics, and forming a characteristic parameter matrix with the frequency characteristics, the energy characteristics and the time domain characteristics.

[0014] As a preferred scheme of the ultra-high frequency monitoring method for abnormal discharge of transmission lines based on EMD-PCA-SVM fusion of multiple parameters described in the present invention, wherein: the characteristic parameter matrix is ​​standardized to obtain a standardized characteristic matrix, the covariance matrix of the standardized characteristic matrix is ​​calculated, the covariance matrix is ​​eigenvalue decomposed to obtain eigenvalues ​​and eigenvectors, the eigenvalues ​​are arranged in descending order, and the cumulative contribution rate is calculated. If the cumulative contribution rate corresponding to a certain eigenvector is greater than the ninth preset threshold value, all eigenvectors before the eigenvector are selected to construct a dimensionality reduction matrix, and the standardized feature matrix is ​​multiplied by the dimensionality reduction matrix to obtain dimensionality reduction feature data.

[0015] As a preferred solution of the ultra-high frequency monitoring method for abnormal discharge of power transmission lines based on EMD-PCA-SVM fusion of multiple parameters described in the present invention, wherein: the dimension reduction feature data is divided into a training data set and a test data set according to a tenth preset ratio, and a support vector machine classification model is trained based on the training data set, including: selecting a linear kernel function, a polynomial kernel function and a Gaussian kernel function to respectively construct a support vector machine model, and using a cross-validation method to optimize the penalty factor and kernel function parameters of the support vector machine model to obtain an optimized support vector machine classification model;

[0016] The test data set is input into the optimized support vector machine classification model for testing, and the classification accuracy of the support vector machine classification model is calculated. If the classification accuracy is less than the eleventh preset threshold, the support vector machine classification model is retrained. If the classification accuracy is greater than or equal to the eleventh preset threshold and the variance of the test results for twelfth consecutive preset times is less than the thirteenth preset threshold, the final support vector machine classification model is determined.

[0017] As a preferred solution of the ultra-high frequency monitoring method for abnormal discharge of power transmission lines based on EMD-PCA-SVM fusion of multiple parameters described in the present invention, the dimension reduction feature data collected in real time is input into the final support vector machine classification model, and the recognition result of the abnormal discharge type is output. If the recognition result is judged as abnormal discharge for the fourteenth consecutive preset number of times, an alarm signal is triggered and the feature data of the abnormal discharge type is recorded, otherwise, the next recognition is continued.

[0018] To further solve the above technical problems, the present invention provides the following technical solutions: A system for ultra-high frequency monitoring of abnormal discharge of power transmission lines based on EMD-PCA-SVM fusion of multiple parameters, comprising: an acquisition module, for acquiring signal data through multiple types of sensors installed on transmission line towers, adjusting the sampling frequency according to the signal amplitude, performing digital conversion and time synchronization processing on the acquired signal data, and obtaining multi-parameter signal data;

[0019] An analysis module, used for performing empirical mode decomposition on the multi-parameter signal data to obtain an intrinsic mode function set, extracting characteristic parameters from the intrinsic mode function set to form a characteristic parameter matrix, and performing dimensionality reduction processing on the characteristic parameter matrix through principal component analysis to obtain reduced-dimensionality characteristic data;

[0020] The monitoring module is used to train and optimize the support vector machine classification model based on the reduced dimension feature data, use the support vector machine classification model to identify the abnormal discharge type of the reduced dimension feature data collected in real time, and trigger an alarm signal according to continuous identification results.

[0021] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and is characterized in that when the processor executes the computer program, the steps of the ultra-high frequency monitoring method for abnormal discharge of a transmission line based on the EMD-PCA-SVM fusion of multiple parameters are implemented as described above.

[0022] A computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of the ultra-high frequency monitoring method for abnormal discharge of a transmission line based on EMD-PCA-SVM fusion of multiple parameters as described above are implemented.

[0023] The beneficial effects of the present invention are as follows: through the organic combination of multi-parameter collaborative monitoring, EMD-PCA feature extraction and multi-core support vector machine classification, in the data acquisition link, the acoustic, optical, electrical and ultra-high frequency multi-parameter collaborative monitoring strategy is adopted, and the adaptive sampling mechanism is used to effectively improve the coverage and reliability of abnormal discharge detection. When a single sensor is disturbed by the environment, other sensors can still maintain effective monitoring; in the feature extraction link, the adaptive decomposition of the signal is realized by the improved EMD method, and the orthogonal correction and endpoint effect processing mechanism are introduced to overcome the limitations of traditional methods in processing non-stationary signals; the PCA dimension reduction strategy based on information entropy feature optimization and multi-dimensional evaluation index is adopted to achieve data dimension reduction while maintaining key feature information, thereby improving the efficiency of subsequent processing; in the abnormal identification link, the multi-core function adaptive fusion mechanism and stratified sampling strategy are innovatively adopted to improve the adaptability of the model to different types of discharge characteristics, and the incremental learning and continuous judgment mechanism are used to reduce the system false alarm rate and realize the online optimization of the model. Through the synergistic effect of multiple technical links, the present invention realizes the effective identification and timely warning of abnormal discharge in a complex electromagnetic environment, and provides technical support for the safe operation of transmission lines. These improvements have formed a complete technical system through in-depth analysis of problems and design of innovative solutions, which has practical application value in terms of accuracy, reliability and adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0025] Figure 1 is a flow chart of the overall method in the present invention;

[0026] Figure 2 is a flowchart of the steps in the present invention;

[0027] Figure 3 It is the EMD algorithm flow chart in the present invention;

[0028] Figure 4 It is the PCA algorithm flow chart of the present invention;

[0029] Figure 5 This is a diagram of a computer device in the present invention. DETAILED DESCRIPTION

[0030] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0031] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0032] Example 1, reference Figure 1 to Figure 4 , which is an embodiment of the present invention, provides an ultra-high frequency monitoring method for abnormal discharge of transmission lines based on EMD-PCA-SVM fusion of multiple parameters.

[0033] Figure 1 The overall flow chart of a UHF monitoring method for abnormal discharge of a power transmission line based on EMD-PCA-SVM fusion of multiple parameters is shown, including: S1. collecting multi-parameter signal data of the power transmission line through an acoustic sensor, an optical sensor, an electromagnetic sensor and a UHF sensor;

[0034] S2. Performing empirical mode decomposition on multi-parameter signal data to obtain intrinsic mode functions, extracting characteristic parameters of the intrinsic mode functions, and performing dimensionality reduction processing on the characteristic parameters through principal component analysis to obtain reduced-dimensionality feature data;

[0035] S3. Input the dimension-reduced feature data into the support vector machine classification model, train the support vector machine classification model and output the classification and recognition results of the abnormal discharge signal.

[0036] Further, S11. Installing an acoustic sensor, an optical sensor, an electromagnetic sensor and an ultra-high frequency sensor on a tower of a transmission line at a first preset interval, setting the sampling frequency of the acoustic sensor to a first preset frequency, setting the wavelength response of the optical sensor to a first preset wavelength, setting the operating frequency of the electromagnetic sensor to a second preset frequency, and setting the operating frequency of the ultra-high frequency sensor to a third preset frequency;

[0037] S12. Synchronously collect signal data of the power transmission line through an acoustic sensor, an optical sensor, an electromagnetic sensor and an ultra-high frequency sensor. If the amplitude of the collected signal of the acoustic sensor exceeds a first preset threshold, or the amplitude of the collected signal of the optical sensor exceeds a second preset threshold, or the amplitude of the collected signal of the electromagnetic sensor exceeds a third preset threshold, or the amplitude of the collected signal of the ultra-high frequency sensor exceeds a fourth preset threshold, then increase the sampling frequency of the sensor to a fourth preset frequency, otherwise collect data according to the fifth preset frequency;

[0038] S13. Digitally convert the signal data collected by the sensor according to the sixth preset frequency, and synchronously integrate the digitally converted signal data according to the timestamp information to obtain multi-parameter signal data.

[0039] It should be noted that the constructed multi-parameter collaborative monitoring system mainly includes three aspects: First, by rationally arranging acoustic sensors, optical sensors, electromagnetic sensors and UHF sensors, a multi-dimensional monitoring network is formed, and each sensor is responsible for collecting signal characteristics of different physical quantities. Secondly, an adaptive sampling mechanism based on preset thresholds is designed. When any sensor detects an abnormal signal, the system automatically adjusts the sampling frequency to ensure the complete collection of key information. Finally, a timestamp synchronization mechanism is used to synchronously integrate multi-source heterogeneous data to establish a unified multi-parameter signal data set, which provides a basis for subsequent feature extraction.

[0040] In addition, the following technical problems are solved and corresponding beneficial effects are achieved: In the traditional single sensor monitoring scheme, different types of abnormal discharges will show significant characteristics in different physical quantities. It is difficult for a single sensor to fully capture the discharge characteristics, which is easy to cause missed detection and misjudgment. The present invention realizes the synchronous acquisition of multiple physical quantities such as sound, light, and electricity through a multi-sensor collaborative monitoring strategy, significantly improving the coverage and reliability of abnormal discharge detection. Especially in complex environments, when some sensors are interfered with, other sensors can still maintain effective monitoring, which improves the anti-interference ability of the system. In addition, the adaptive sampling mechanism solves the problems of data redundancy and information loss under the traditional fixed sampling method, while ensuring the integrity of abnormal signal acquisition, reducing the burden of data storage and transmission. The multi-source data synchronization integration mechanism overcomes the technical difficulties of heterogeneous data from different sensors, laying the foundation for subsequent fusion analysis. Through practical application verification, this step not only improves the accuracy of abnormal discharge detection, but also optimizes the efficiency of system resource utilization, with obvious technical advantages and practical value. These improvements are not achieved by the simple superposition of existing technologies, but by in-depth analysis of the physical characteristics and monitoring requirements of abnormal discharges, and innovatively designing a multi-parameter collaborative monitoring scheme.

[0041] Further, S21. connect the local maximum points in the multi-parameter signal data into an upper envelope curve through cubic spline interpolation, connect the local minimum points into a lower envelope curve through cubic spline interpolation, calculate the mean curve of the upper envelope curve and the lower envelope curve and subtract them from the multi-parameter signal data to obtain candidate components. If the difference between the number of local extreme points and the number of zero-crossing points of the candidate component is greater than the seventh preset threshold, then repeat this step with the candidate component as a new signal to be processed. If the difference between the number of local extreme points and the number of zero-crossing points is less than or equal to the seventh preset threshold, then determine the candidate component as the first intrinsic mode function, subtract the first intrinsic mode function from the multi-parameter signal data to obtain the remaining signal. If the number of extreme points of the remaining signal is greater than the eighth preset threshold, then repeat this step with the remaining signal as a new signal to be processed until a set of intrinsic mode functions is obtained.

[0042] Furthermore, S22. characteristic parameters are extracted from each intrinsic mode function in the eigenmode function set respectively, including: calculating the instantaneous frequency of the intrinsic mode function by Hilbert transform to obtain frequency characteristics, calculating the energy distribution of the intrinsic mode function to obtain energy characteristics, extracting the amplitude and duration of the intrinsic mode function to obtain time domain characteristics, and combining the frequency characteristics, energy characteristics and time domain characteristics into a characteristic parameter matrix.

[0043] Further, S23. The feature parameter matrix is ​​standardized to obtain a standardized feature matrix, the covariance matrix of the standardized feature matrix is ​​calculated, the covariance matrix is ​​eigenvalue decomposed to obtain eigenvalues ​​and eigenvectors, the eigenvalues ​​are arranged in descending order, and the cumulative contribution rate is calculated. If the cumulative contribution rate corresponding to a certain eigenvector is greater than the ninth preset threshold, all eigenvectors before the eigenvector are selected to construct a dimensionality reduction matrix, and the standardized feature matrix is ​​multiplied by the dimensionality reduction matrix to obtain reduced dimensionality feature data.

[0044] It should be noted that the construction of the EMD-PCA feature extraction framework contains three key links: first, the empirical mode decomposition method is used to adaptively decompose the multi-parameter signal, and the complex signal is decomposed into a series of intrinsic mode functions through an iterative screening process. Each intrinsic mode function represents a different frequency component of the signal. The mirror continuation method is used to deal with the endpoint effect, and the orthogonal correction is introduced to ensure the independence of the intrinsic mode functions. Secondly, the frequency characteristics, energy characteristics and time domain characteristics are extracted from the intrinsic mode functions, and a multidimensional feature parameter matrix is ​​constructed. The feature weights are calculated based on the information entropy principle, and the feature correlation analysis is combined to perform feature optimization to achieve the optimal combination of features. Finally, the principal component analysis method is used to reduce the dimension of the feature parameter matrix, and the eigenvector is obtained through eigenvalue decomposition. Combined with multidimensional evaluation indicators such as cumulative contribution rate, eigenvector variance ratio and cross-validation, the most representative feature combination is screened to achieve efficient data dimensionality reduction.

[0045] The following technical problems are solved and corresponding beneficial effects are achieved: Traditional signal processing methods such as Fourier transform and wavelet transform have limitations when processing nonlinear and non-stationary discharge signals, and it is difficult to accurately extract the essential characteristics of the signal. The present invention realizes adaptive decomposition of the signal through an improved EMD method, without presetting basis functions, and can better adapt to the non-stationary characteristics of the discharge signal. In the EMD decomposition process, the mirror extension method is used to process the endpoint effect, avoiding distortion at the signal boundary; the orthogonalization correction mechanism is introduced to solve the mode aliasing problem and improve the reliability of the decomposition result. In the feature extraction link, the problem of insufficient expression ability of a single feature is overcome, and comprehensive feature information is obtained through a multidimensional feature extraction strategy, and a feature optimization mechanism is established based on information entropy and feature correlation analysis to achieve the optimal combination of features. In the feature dimensionality reduction process, a scientific feature selection standard is established through the comprehensive application of multidimensional evaluation indicators such as cumulative contribution rate, feature vector variance ratio and cross-validation, ensuring that the features after dimensionality reduction have good representativeness and distinguishability. This systematic feature extraction scheme not only improves the accuracy and reliability of feature extraction, but also improves the computational efficiency through dimensionality reduction optimization, providing high-quality feature data support for subsequent abnormal discharge identification. Experimental verification shows that this solution has obvious advantages over traditional methods in terms of feature extraction accuracy, anti-interference ability and computational efficiency.

[0046] Further, S31. The dimension reduction feature data is divided into a training data set and a test data set according to a tenth preset ratio, and a support vector machine classification model is trained based on the training data set, including: selecting a linear kernel function, a polynomial kernel function and a Gaussian kernel function to construct a support vector machine model, and using a cross-validation method to optimize the penalty factor and kernel function parameters of the support vector machine model to obtain an optimized support vector machine classification model;

[0047] S32. Input the test data set into the optimized support vector machine classification model for testing, calculate the classification accuracy of the support vector machine classification model, if the classification accuracy is less than the eleventh preset threshold, retrain the support vector machine classification model, if the classification accuracy is greater than or equal to the eleventh preset threshold and the variance of the test results for the twelfth consecutive preset number of times is less than the thirteenth preset threshold, determine the final support vector machine classification model.

[0048] Furthermore, the reduced-dimensional feature data collected in real time is input into the final support vector machine classification model to output the recognition result of the abnormal discharge type. If the recognition result is judged as abnormal discharge for the fourteenth consecutive preset number of times, an alarm signal is triggered and the feature data of the abnormal discharge type is recorded. Otherwise, the next recognition is continued.

[0049] It should be noted that, at the model construction level, this step first uses stratified sampling to divide the data set, and ensures the consistency of the training set and the test set according to the distribution characteristics of the abnormal discharge type, overcoming the problem of uneven sample distribution caused by random division. In the model design, a multi-kernel function parallel construction strategy is adopted, and the support vector machine model is constructed using linear kernel functions, polynomial kernel functions and Gaussian kernel functions. An adaptive weight mechanism is introduced to dynamically adjust the contribution ratio of each kernel function according to the distribution characteristics of different types of discharge characteristics. This design breaks through the limitations of the traditional single kernel function and can adapt to the nonlinear distribution characteristics of different types of discharge characteristics. The penalty factor and kernel function parameters are optimized by the cross-validation method, and an adaptive adjustment mechanism for the model parameters is established to improve the generalization ability of the classifier. Especially when dealing with small samples and unbalanced samples, this method shows good robustness.

[0050] At the model evaluation level, the present invention designs a multi-dimensional evaluation system. First, a basic evaluation is performed through classification accuracy and variance analysis, requiring the variance of continuous test results to meet stability requirements. At the same time, ROC curve analysis and confusion matrix evaluation are introduced to comprehensively evaluate the model's sensitivity, specificity, and recognition ability of various types of discharges. This multi-dimensional evaluation mechanism effectively solves the problem of single model performance evaluation in traditional methods and ensures the reliability of classification results.

[0051] At the real-time monitoring level, the present invention establishes an abnormal discharge identification mechanism based on continuous judgment and introduces an incremental learning strategy. By setting a threshold for the number of continuous judgments, false alarms caused by single misjudgments are avoided, and the anti-interference ability of the system is improved. For feature data identified as abnormal discharge, the system not only records and stores them, but also incorporates them into the incremental learning data set. When the incremental data set accumulates to a preset scale, the model update mechanism is triggered, and the model is optimized online using new data, thereby achieving continuous self-improvement of the system. This dynamic optimization mechanism enables the model to adapt to changes in the operating status of the equipment and maintain a high recognition accuracy.

[0052] In summary, the present invention has achieved three technical innovations: first, it adopts a multi-parameter fusion monitoring strategy to improve the comprehensiveness and reliability of detection; second, it introduces a machine learning method to achieve intelligent identification of discharge types; third, it adopts the EMD-PCA algorithm to improve the accuracy of feature extraction. These technical innovations of the present invention effectively solve the problems of low detection accuracy, insufficient automation and limited feature extraction capabilities in the prior art.

[0053] Embodiment 2 is an embodiment of the present invention. A more specific solution is provided to address the defects of the conventional ultra-high frequency monitoring method for abnormal discharge of power transmission lines:

[0054] Reference Figure 2 to Figure 4 , using a variety of sensor monitoring equipment such as sound, light, electricity and UHF to collect transmission line data in all directions. Secondly, perform empirical mode decomposition (EMD) on each collected UHF signal to obtain several intrinsic mode functions (IMFs), and extract multiple features such as instantaneous frequency, energy distribution and time of each IMF. Then, principal component analysis (PCA) is used to reduce the dimension of the feature space and delete unimportant features. Finally, the remaining valid features are input into the SVM model for training and testing, and the final discharge signal classification and recognition results are output. The detailed description of each algorithm and model is explained in the subsequent explanation.

[0055] The implementation process of the UHF monitoring method for abnormal discharge of transmission lines based on EMD-RFE-SVM fusion of multiple parameters is as follows: Figure 2 As shown, the steps are as follows:

[0056] Step 1: Data collection and labeling. Install UHF sensors on transmission lines and towers to collect UHF signal data under normal operation and abnormal discharge conditions. Label the collected signals to distinguish normal signals, partial discharge signals and other interference signals, providing a basis for subsequent machine learning model training.

[0057] Step 2: Signal processing and feature extraction. Perform EMD decomposition on each collected UHF signal to obtain several IMFs. Extract features from each IMF, such as:

[0058] Instantaneous frequency: The instantaneous frequency distribution of each IMF is obtained through Hilbert transform.

[0059] Energy distribution: Calculate the energy of each IMF to reflect the energy distribution of different frequency components in the signal.

[0060] Time characteristics: Extract time domain characteristics such as amplitude and duration of IMF.

[0061] Other features: spectral characteristics of optical signals, changes in magnetic field intensity of magnetic signals, etc.

[0062] Step 3: Feature selection: PCA is used to reduce the dimension of the feature space, remove unimportant features, and retain the features that can best distinguish normal and abnormal discharge signals.

[0063] Step 4: Data partitioning and model training. Use the feature vectors obtained from step 3 to construct training and test sets. Use the training data to train the SVM model and select appropriate kernel functions (linear kernel, Gaussian kernel, etc.) and hyperparameters (C value, gamma value).

[0064] Step 5: Model evaluation. Combine cross-validation (such as k-fold cross-validation) to evaluate the performance of the model and adjust hyperparameters to optimize the accuracy of the model.

[0065] Step 6: Determine the value of the loss function. This patent sets the mean square error (MSE) as the loss function. If the value of the loss function is lower than the set threshold, the accuracy requirement is met, the iteration ends, and the next step is entered. Otherwise, jump back to step 4 and repeat the iteration until the condition is met.

[0066] Step 7: Discharge signal classification and identification: Use the trained SVM model to classify various signals and determine whether they are abnormal discharge signals.

[0067] The EMD algorithm process is as follows Figure 3 As shown, the steps are as follows:

[0068] Step 1: Select local extreme points in the signal and generate upper and lower envelope curves through interpolation method.

[0069] Step 2: Calculate the average of the upper and lower envelope curves and subtract them from the original signal to obtain the new signal component.

[0070] Step 3: Repeat the iteration until the residual signal meets the definition of IMF and the first IMF is obtained.

[0071] Step 4: Subtract the first IMF from the original signal.

[0072] Step 5: Determine whether the remaining signal is monotonic or no longer has obvious extreme points. If so, proceed to the next step; otherwise, jump back to step 2 until the conditions are met.

[0073] Step 6: Output the IMFs collection.

[0074] The PCA algorithm process is as follows Figure 4 As shown, the steps are as follows:

[0075] Step 1: Data standardization. Standardize the data so that each feature has a mean of 0 and a variance of 1. Standardization ensures that each feature contributes equally to PCA.

[0076] Step 2: Calculate the covariance matrix. Calculate the covariance matrix using the standardized data. The covariance matrix represents the correlation between different features.

[0077] Step 3: Calculate eigenvalues ​​and eigenvectors. Perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and eigenvectors. The eigenvectors represent the directions of the new feature space, while the eigenvalues ​​represent the data variance in these directions.

[0078] Step 4: Select principal components. Select the first n principal components based on the size of the eigenvalues, where n is the expected number of features after dimensionality reduction. Principal components with larger eigenvalues ​​have larger data variances and retain more information.

[0079] Step 5: Transform the data. Project the original data onto the selected principal components to obtain the reduced-dimensional data. This is achieved by performing matrix multiplication of the original data with the selected eigenvectors (principal components).

[0080] Step 6: Analyze and use the reduced data. Use the reduced data for further analysis or modeling to check whether the reduced data has achieved the desired effect.

[0081] Embodiment 3 is an embodiment of the present invention, which provides an ultra-high frequency monitoring system for abnormal discharge of power transmission lines based on EMD-PCA-SVM fusion of multiple parameters, including: an acquisition module, which is used to collect signal data through multiple types of sensors installed on transmission line towers, adjust the sampling frequency according to the signal amplitude, and perform digital conversion and time synchronization processing on the collected signal data to obtain multi-parameter signal data;

[0082] An analysis module is used to perform empirical mode decomposition on multi-parameter signal data to obtain an intrinsic mode function set, extract characteristic parameters from the intrinsic mode function set to form a characteristic parameter matrix, and perform dimension reduction processing on the characteristic parameter matrix through principal component analysis to obtain reduced-dimensional characteristic data;

[0083] The monitoring module is used to train and optimize the support vector machine classification model based on the reduced dimension feature data, use the support vector machine classification model to identify the abnormal discharge type of the reduced dimension feature data collected in real time, and trigger an alarm signal based on continuous identification results.

[0084] Example 4, reference Figure 5 , is an embodiment of the present invention, which is different from the previous embodiment in that: if the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0085] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0086] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0087] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0088] Example 5 is an embodiment of the present invention, which provides an ultra-high frequency monitoring method for abnormal discharge of transmission lines based on EMD-PCA-SVM fusion of multiple parameters. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0089] In order to verify the effectiveness of the present invention, the present invention has designed a comprehensive comparative experiment. The experiment was carried out in a 330kV substation, and 10 groups of different types of abnormal discharge samples (including corona discharge, suspended discharge, surface discharge and hollow discharge) were selected, and each type of sample contained 100 groups of test data. The control group adopted the traditional single ultra-high frequency detection method and the feature extraction method based on wavelet transform, combined with the conventional SVM classifier for identification. The experimental group adopted the multi-parameter collaborative detection method of the present invention, combined with the improved EMD-PCA feature extraction and multi-kernel function adaptive SVM classification method. The experimental environment temperature is 25±2℃, the relative humidity is 65±5%, the background noise is -75dB, the sampling frequency is set to 100MHz, and the data acquisition time is 10ms.

[0090] In the feature extraction stage, the orthogonality index of EMD decomposition, the influence of endpoint effect and the feature dimension reduction ratio were calculated respectively. In the classification and recognition stage, the model performance was evaluated through 5-fold cross validation, and indicators such as accuracy, recall rate, F1 score and ROC curve area were calculated. At the same time, the anti-interference ability of the system was evaluated by testing under different electromagnetic interference intensities (20dB, 30dB, 40dB signal-to-noise ratio). In order to verify the adaptive ability of the system, an online monitoring experiment lasting 72 hours was also carried out, and the system performance indicators were recorded every 8 hours to evaluate the stability of the model and the self-optimization effect.

[0091] Table 1 Experimental index comparison analysis table

[0092]

[0093]

[0094] The following conclusions can be drawn from the analysis of experimental data: the detection sensitivity of the present invention is improved by 7dB compared with the traditional method, and the performance degradation in a strong interference environment is also smaller; the signal acquisition completeness rate is increased by 12.3% and 21.4% under no interference and strong interference conditions, respectively, indicating the effectiveness of the multi-parameter collaborative monitoring strategy; the feature extraction time is reduced by 32%, and the feature dimension is reduced from 256 to 96, which significantly improves the calculation efficiency; the improved EMD method reduces the impact of the endpoint effect to 5.3%, and the orthogonality index is increased to 0.95, which is better than the traditional method; in terms of classification performance, the accuracy of the present invention reaches 93.8%, the false alarm rate is reduced to 4.2%, and the performance degradation is small in a strong interference environment, and the classification accuracy remains at 89.2%; the anti-interference ability of the system is improved by 13dB, the model convergence time is shortened by 28.9%, and the system response delay is reduced by 33.3%. These data fully prove the comprehensive advantages of the present invention in detection sensitivity, anti-interference ability, calculation efficiency and recognition accuracy. Especially in a strong interference environment, the present invention exhibits better robustness, and the attenuation amplitude of various performance indicators is significantly smaller than that of traditional methods, which is of great significance for practical engineering applications.

[0095] It is important to note that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A UHF monitoring method for abnormal discharge of power transmission lines based on EMD-PCA-SVM fusion of multiple parameters, characterized in that: include: Collect multi-parameter signal data of power transmission lines through acoustic sensors, optical sensors, electromagnetic sensors and UHF sensors; Performing empirical mode decomposition on the multi-parameter signal data to obtain an intrinsic mode function, extracting characteristic parameters of the intrinsic mode function, and performing dimensionality reduction processing on the characteristic parameters through principal component analysis to obtain reduced-dimensionality characteristic data; The dimension-reduced feature data is input into a support vector machine classification model, the support vector machine classification model is trained and a classification recognition result of the abnormal discharge signal is output.

2. The UHF monitoring method for abnormal discharge of power transmission lines based on EMD-PCA-SVM fusion of multiple parameters as claimed in claim 1 is characterized by: Installing an acoustic sensor, an optical sensor, an electromagnetic sensor and an ultra-high frequency sensor on a tower of a power transmission line at a first preset interval, setting the sampling frequency of the acoustic sensor to a first preset frequency, setting the wavelength response of the optical sensor to a first preset wavelength, setting the operating frequency of the electromagnetic sensor to a second preset frequency, and setting the operating frequency of the ultra-high frequency sensor to a third preset frequency; The signal data of the power transmission line is synchronously collected by the acoustic sensor, the optical sensor, the electromagnetic sensor and the ultra-high frequency sensor. If the amplitude of the collected signal of the acoustic sensor exceeds the first preset threshold value, the amplitude of the collected signal of the optical sensor exceeds the second preset threshold value, the amplitude of the collected signal of the electromagnetic sensor exceeds the third preset threshold value, or the amplitude of the collected signal of the ultra-high frequency sensor exceeds the fourth preset threshold value, the sampling frequency of the sensor is increased to the fourth preset frequency, otherwise, the data is collected according to the fifth preset frequency; The signal data collected by the sensor is digitally converted according to a sixth preset frequency, and the digitally converted signal data is synchronously integrated according to timestamp information to obtain multi-parameter signal data.

3. The UHF monitoring method for abnormal discharge of power transmission lines based on EMD-PCA-SVM fusion of multiple parameters as claimed in claim 2 is characterized by: The local maximum points in the multi-parameter signal data are connected to form an upper envelope curve through cubic spline interpolation, and the local minimum points are connected to form a lower envelope curve through cubic spline interpolation. The mean curve of the upper envelope curve and the lower envelope curve is calculated and subtracted from the multi-parameter signal data to obtain a candidate component. If the difference between the number of local extreme points and the number of zero-crossing points of the candidate component is greater than a seventh preset threshold, the candidate component is used as a new signal to be processed and this step is repeatedly performed. If the difference between the number of local extreme points and the number of zero-crossing points is less than or equal to the seventh preset threshold, the candidate component is determined as a first intrinsic mode function, and the multi-parameter signal data is subtracted from the first intrinsic mode function to obtain a residual signal. If the number of extreme points of the residual signal is greater than an eighth preset threshold, this step is repeatedly performed on the residual signal as a new signal to be processed until the intrinsic mode function set is obtained.

4. The UHF monitoring method for abnormal discharge of power transmission lines based on EMD-PCA-SVM fusion of multiple parameters as claimed in claim 3 is characterized by: Extracting characteristic parameters from each intrinsic mode function in the eigenmode function set respectively, including: calculating the instantaneous frequency of the eigenmode function by Hilbert transform to obtain frequency characteristics, calculating the energy distribution of the eigenmode function to obtain energy characteristics, extracting the amplitude and duration of the eigenmode function to obtain time domain characteristics, and combining the frequency characteristics, the energy characteristics and the time domain characteristics into a characteristic parameter matrix.

5. The UHF monitoring method for abnormal discharge of power transmission lines based on EMD-PCA-SVM fusion of multiple parameters as claimed in claim 4 is characterized by: The feature parameter matrix is ​​standardized to obtain a standardized feature matrix, the covariance matrix of the standardized feature matrix is ​​calculated, the covariance matrix is ​​eigenvalue decomposed to obtain eigenvalues ​​and eigenvectors, the eigenvalues ​​are arranged in descending order, and the cumulative contribution rate is calculated. If the cumulative contribution rate corresponding to a certain eigenvector is greater than a ninth preset threshold, all eigenvectors before the eigenvector are selected to construct a dimensionality reduction matrix, and the standardized feature matrix is ​​multiplied by the dimensionality reduction matrix to obtain dimensionality reduction feature data.

6. The UHF monitoring method for abnormal discharge of power transmission lines based on EMD-PCA-SVM fusion of multiple parameters as claimed in claim 5 is characterized by: The dimension reduction feature data is divided into a training data set and a test data set according to a tenth preset ratio, and a support vector machine classification model is trained based on the training data set, including: selecting a linear kernel function, a polynomial kernel function, and a Gaussian kernel function to respectively construct a support vector machine model, and using a cross-validation method to optimize the penalty factor and kernel function parameters of the support vector machine model to obtain an optimized support vector machine classification model; The test data set is input into the optimized support vector machine classification model for testing, and the classification accuracy of the support vector machine classification model is calculated. If the classification accuracy is less than the eleventh preset threshold, the support vector machine classification model is retrained. If the classification accuracy is greater than or equal to the eleventh preset threshold and the variance of the test results for twelfth consecutive preset times is less than the thirteenth preset threshold, the final support vector machine classification model is determined.

7. The UHF monitoring method for abnormal discharge of power transmission lines based on EMD-PCA-SVM fusion of multiple parameters as claimed in claim 6 is characterized by: The reduced-dimensional feature data collected in real time is input into the final support vector machine classification model, and the recognition result of the abnormal discharge type is output. If the recognition result is determined to be abnormal discharge for the fourteenth consecutive preset number of times, an alarm signal is triggered and the feature data of the abnormal discharge type is recorded, otherwise, the next recognition is continued.

8. A system using the ultra-high frequency monitoring method for abnormal discharge of power transmission lines based on EMD-PCA-SVM fusion of multiple parameters as described in any one of claims 1 to 7, characterized in that: include: The acquisition module is used to collect signal data through multiple types of sensors installed on the transmission line tower, adjust the sampling frequency according to the signal amplitude, perform digital conversion and time synchronization processing on the collected signal data, and obtain multi-parameter signal data; An analysis module, used for performing empirical mode decomposition on the multi-parameter signal data to obtain an intrinsic mode function set, extracting characteristic parameters from the intrinsic mode function set to form a characteristic parameter matrix, and performing dimensionality reduction processing on the characteristic parameter matrix through principal component analysis to obtain reduced-dimensionality characteristic data; The monitoring module is used to train and optimize the support vector machine classification model based on the reduced dimension feature data, use the support vector machine classification model to identify the abnormal discharge type of the reduced dimension feature data collected in real time, and trigger an alarm signal according to continuous identification results.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the ultra-high frequency monitoring method for abnormal discharge of power transmission lines based on EMD-PCA-SVM fusion of multiple parameters as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the ultra-high frequency monitoring method for abnormal discharge of a transmission line based on EMD-PCA-SVM fusion of multiple parameters as described in any one of claims 1 to 7 are implemented.

Images