Elman neural network overhead ground wire detection method based on adaptive filtering

Abstract

The invention discloses an Elman neural network overhead ground wire detection method based on adaptive filtering, and belongs to the technical field of overhead ground wire detection, and the method comprises the steps: collecting the residual magnetism information of the surface of an overhead ground wire through a sensor array; de-noising is carried out on the detected original signal based on an adaptive filtering algorithm of wavelet transform; performing circumferential interpolation on the filtered magnetic flux leakage data by adopting a cubic spline interpolation method, and converting the filtered magnetic flux leakage data into a grayscale image for processing; an Elman neural network is adopted to identify defects, texture features of defect images are extracted to serve as input of the neural network, and a back propagation training algorithm function is set to be trainbfg to improve network convergence. By the adoption of the Elman neural network overhead ground wire detection method based on adaptive filtering, the problems existing in an existing overhead ground wire defect detection method are solved, noise signals in original signals can be effectively restrained, the defect recognition success rate can be effectively increased, and the operation safety of the overhead ground wire is improved.

Description

Technical Field

[0001] The invention relates to the technical field of overhead ground wire detection, and in particular to an overhead ground wire detection method based on an Elman neural network with adaptive filtering. Background Art

[0002] As an important part of the transmission line, the overhead ground wire plays a key role in ensuring the safe and stable operation of the power system. Overhead ground wires are exposed to the natural environment for a long time, and are affected by various adverse climatic conditions, mechanical stress, electrochemical corrosion and other factors, and are prone to different types of defects such as wear, broken strands, corrosion, fatigue cracks, etc. If these defects are not detected and repaired in time, they will seriously threaten the reliability of the transmission line and may even cause line failures, resulting in large-scale power outages and huge losses to the society and economy.

[0003] As the operation time of power transmission lines continues to increase, the accumulation and evolution of defects make detection more difficult. Especially for power transmission lines in remote areas and complex terrains, the application of manual inspections and helicopter inspections is limited, resulting in the inability to detect hidden dangers in some areas in a timely manner.

[0004] In addition, with the continuous development of the power system, the operating environment of overhead ground wires has become increasingly complex, and traditional detection methods often fail to meet the requirements of high efficiency, accuracy, and full coverage. How to use modern technical means to improve the detection efficiency and accuracy of overhead ground wires has become a problem that the power industry needs to solve urgently.

[0005] Traditional overhead ground wire detection methods mainly include manual inspections, inspections based on helicopter-mounted detection equipment, and some conventional non-destructive testing technologies. The manual inspection efficiency of these non-destructive testing technologies is low, and the detection results rely heavily on the experience and professional skills of the inspectors, making it difficult to ensure accurate identification and comprehensive detection of minor defects. Although helicopter inspections can cover a larger detection range, they are costly and are restricted by weather conditions and flight airspace. Conventional non-destructive testing technologies, such as ultrasonic testing and eddy current testing, often require direct contact with the ground wire when applied to overhead ground wire detection. The operation is complicated, the detection speed is slow, and the defect detection effect is poor for some complex shapes and working conditions. Summary of the invention

[0006] The purpose of the present invention is to provide an Elman neural network overhead ground wire detection method based on adaptive filtering, which can solve the problems of low detection efficiency, low accuracy, high detection requirements, insufficient reliability, radiation risk, etc. in the existing overhead ground wire defect detection methods, and can effectively suppress the noise signal in the original signal and improve the defect recognition success rate, thereby improving the safety of overhead ground wire operation.

[0007] To achieve the above object, the present invention provides an overhead ground wire detection method based on an Elman neural network with adaptive filtering, comprising the following steps:

[0008] S1. Use a sensor array to collect residual magnetic information of the overhead ground wire, and divide the collected residual magnetic information data into equal intervals to produce a sample data set;

[0009] S2, the sample data set is subjected to wavelet transform using Mallet algorithm and decomposed into high-frequency detail signals and low-frequency approximate signals;

[0010] S3, performing adaptive filtering on each segment of data processed by S2, and obtaining noise-reduced data after screening;

[0011] S4, using a cubic spline interpolation method to perform circumferential interpolation on the filtered denoised data, and converting the filtered denoised data into a grayscale image;

[0012] S5, thresholding the grayscale image of S4, automatically obtaining the threshold, and obtaining a grayscale image of overhead ground wire defects with a resolution of 300×400;

[0013] S6, extracting texture features of the defect grayscale image to obtain a quantitative recognition feature vector;

[0014] S7. Use the Elman neural network to identify defects, use the test set to verify the effectiveness of the Elman neural network, and realize the accurate detection function of overhead ground wire defects.

[0015] Preferably, the sensor array of S1 is composed of 18 high-precision giant magnetoresistance GMR sensors evenly arranged around the overhead ground wire.

[0016] Preferably, the calculation formula of the Mallat algorithm of S2 is:

[0017]

[0018]

[0019] Among them, c j+1,k is the low-frequency approximate output of the signal, d j+1,k is the high-frequency detail output of the signal, j is the signal level, k is the wavelet translation multiple, h 0 is the low-pass filter coefficient, h 1 is the high-pass filter coefficient, m is the wavelet magnification, c j,m is the input signal at j signal level.

[0020] Preferably, the calculation formula of the adaptive filtering algorithm of S3 is:

[0021] e(j)=d(j)-WT (j)X(j);

[0022] Where e is the difference between the output signal and the expected signal, d is the expected signal, j is the signal level, and W T (j)X(j) is the output signal, W is the weighted signal, and X is the input signal.

[0023] Preferably, the S4 uses grayscale normalization of MFL data for preprocessing, and the calculation formula is:

[0024]

[0025] Among them, x and y are the values ​​before and after normalization respectively; V max , V min are the maximum and minimum values ​​of the original grayscale image respectively.

[0026] Preferably, in S5, threshold processing is performed on the data to automatically obtain a threshold, the maximum value of the defect position data is calculated, the centered maximum value is taken, and 150 pixels are intercepted in the axial direction; the circumferential resolution is set to 400 to obtain a 300×400 defect grayscale image.

[0027] Preferably, the S6 extracts texture features and morphological features of the defect image as feature vectors for quantitative identification.

[0028] Preferably, the Elman neural network set in S7 is divided into four layers: input layer, hidden layer, receiving layer and output layer.

[0029] Preferably, the number of hidden layer nodes of the Elman neural network set in S7 is set to 22.

[0030] Preferably, training samples are randomly extracted from the samples, and the remaining samples are used as test samples, and the function of the back propagation training algorithm is set to trainbfg.

[0031] Therefore, the present invention adopts the above-mentioned Elman neural network overhead ground wire detection method based on adaptive filtering, which can effectively extract useful signal features from a complex noise environment and improve the signal-to-noise ratio of the signal, thereby providing a more reliable data basis for subsequent defect identification; Elman neural network, as a neural network with dynamic feedback characteristics, has unique advantages in processing time series data and complex nonlinear system modeling, and can accurately learn and classify various defect characteristics of overhead ground wires, thereby realizing intelligent and efficient detection of overhead ground wire defects, making up for the shortcomings of traditional detection methods and meeting the high-precision, high-reliability and real-time requirements of modern power systems for overhead ground wire status monitoring and maintenance.

[0032] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a schematic diagram of a data acquisition device of an embodiment of an overhead ground wire detection method based on an Elman neural network with adaptive filtering according to the present invention;

[0034] Figure 2 It is a flow chart of an overhead ground wire defect detection method according to an embodiment of an overhead ground wire detection method based on an Elman neural network with adaptive filtering according to the present invention;

[0035] Figure 3 This is a detection result diagram of Elman neural networks with different numbers of hidden layers designed in an embodiment of an Elman neural network overhead ground wire detection method based on adaptive filtering of the present invention.

[0036] Reference numerals

[0037] 1. Overhead ground wire; 2. Sensor array; 3. Permanent magnet. DETAILED DESCRIPTION

[0038] The technical solution of the present invention is further described below through the accompanying drawings and embodiments.

[0039] Unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0040] Embodiment 1

[0041] like Figure 1 As shown, the data acquisition device is composed of an overhead ground wire 1, a sensor array 2 and a permanent magnet 3. The sensor array 2 is composed of 18 high-precision giant magnetoresistance (GMR) sensors evenly arranged around the wire rope. The magnetizing device is composed of a plurality of permanent magnets 3.

[0042] The acquisition process is as follows: first, the wire rope is magnetized using a magnetization device; then, the residual magnetic information on the surface of the wire rope is collected using the sensor array 2; the sensor array 2 moves at a constant speed along the axial direction of the wire rope, and the encoder rotates synchronously; the controller collects data from 18 channels in sequence according to the pulse signal emitted by the encoder.

[0043] Based on the above design idea, the present invention proposes an overhead ground wire detection method based on Elman neural network with adaptive filtering. 190 samples are randomly extracted from 255 samples as training samples, and the remaining 65 samples are used as test samples. The function of the back propagation training algorithm is set to trainbfg, as shown in Figure 2 As shown, the following steps are included:

[0044] S1. Use a sensor array to collect residual magnetic information of the overhead ground wire, and divide the collected residual magnetic information data into equal intervals to produce a sample data set.

[0045] S2. The sample data set is transformed by wavelet using Mallet algorithm, and the wavelet decomposition related parameters are set: the Daubechies V (db5) wavelet is used to orthogonally decompose the signal of each channel, and the decomposition layer number k = 8; the high-frequency detail signal and low-frequency approximate signal of each scale are obtained, and the low-frequency signal is cleared.

[0046] The calculation formula of the Mallat algorithm correlation coefficient is:

[0047]

[0048]

[0049] Among them, c j+1,k is the low-frequency approximate output of the signal, d j+1,k is the high-frequency detail output of the signal, j is the signal level, k is the wavelet translation multiple, h 0 is the low-pass filter coefficient, h 1 is the high-pass filter coefficient, m is the wavelet magnification, c j,m is the input signal at j signal level.

[0050] S3, perform adaptive filtering on each segment of data processed by S2, and perform adaptive filtering on high-frequency signals of various scales. The optimal weight coefficient of the filter is obtained through the minimum mean square error algorithm (LMS); when the signal of each scale meets the filtering requirements, the adaptive filtering ends, and the noise reduction data can be obtained after the filtered wavelet reconstruction. The calculation formula of the correlation coefficient of the adaptive filtering algorithm is:

[0051] e(j)=d(j)-W T (j)X(j);

[0052] Where e is the difference between the output signal and the expected signal, d is the expected signal, j is the signal level, and W T (j)X(j) is the output signal, W is the weighted signal, and X is the input signal.

[0053] S4. Use the cubic spline interpolation method to perform circumferential interpolation on the filtered denoised data, and increase the circumferential resolution from 18 to 400, and convert the filtered denoised data into a grayscale image for processing; the calculation formula of the correlation coefficient is:

[0054]

[0055] Among them, x and y are the values ​​before and after normalization respectively; V max , V min are the maximum and minimum values ​​of the original grayscale image respectively.

[0056] S5. Threshold the grayscale image of S4, automatically obtain the threshold, and use the threshold to determine whether the defect is here. Then calculate the maximum value of the defect position data, take the maximum value in the middle, and intercept 150 pixels in the axial direction. Set the circumferential resolution to 400 to obtain a 300×400 defect grayscale image.

[0057] S6. Quantifying the texture content of a region is an important method for describing an image region. The simplest way to describe texture is to use the statistical moments of the image grayscale histogram.

[0058] Extract the texture features of the defect image and obtain the quantitative identification feature vector; One of the simplest ways to describe texture is to use the statistical moment of the grayscale histogram of an image or region.

[0059] The main method to describe the shape of the histogram distribution is through the distance from the center to the center of the nth order, including the mean, standard deviation, and smoothness. The morphological features of the object are also one of the important features of image recognition, including area, verticality, and elongation.

[0060] S7. Use Elman neural network to identify defects. Based on the basic structure of BP network, a receiving layer is added in the hidden layer as a one-step delay operator to achieve the purpose of memory.

[0061] The system has the ability to adapt to time-varying characteristics, enhances the global stability of the network, and has stronger computing power than feedforward neural networks.

[0062] Compared with BP neural network, Elman neural network has higher classification accuracy, faster approach speed and better dynamic characteristics, and is more suitable for solving pattern classification problems. A 16×22×7×7 Elman neural network is designed, and the 16 extracted defect feature vectors are used as the input of the neural network.

[0063] Among them, 22 is the number of hidden layers, and the output of the receiving layer and the output layer are both 7. The output of the neurons connected to the hidden layer and the input vector is not only the input of the output layer, but also connects other neurons in the hidden layer to the input of the hidden layer. The hidden layer uses the tansig transfer function to verify the effectiveness of the Elman neural network using the test set, and realizes the accurate detection function of the overhead ground wire defects.

[0064] The Elman network detection results are as follows Figure 3 As shown, Figure 3 (a) Figure 3 (b) Figure 3 (c) and Figure 3 (d) in the figure represents the number of hidden layers as 13, 15, 18 and 22. It can be seen from the figure that the detection accuracy of overhead ground wire defects is not exactly the same under different numbers of hidden layers. When the number of hidden layer nodes is set to 22, the recognition rate is the highest. When the recognition error is 1.17%, the recognition success rate is 98.46%, and the maximum recognition error is less than 2.5%. It can be seen that the overhead ground wire defect detection technology based on adaptive filtering and Elman neural network can effectively realize overhead ground wire fault diagnosis with high stability.

[0065] The present invention has the following advantages and positive effects:

[0066] 1. The present invention introduces adaptive filtering technology, the core principle of which is to accurately and dynamically filter out the noise interference components in the detection signal by real-time monitoring of signal characteristics and according to specific algorithm logic. This dynamic adjustment mechanism can quickly and accurately identify and eliminate various types of noise according to the real-time changes of the signal, thereby substantially improving the clarity and accuracy of the signal. At the same time, by organically combining with the Elman neural network, its powerful nonlinear mapping ability is fully utilized. Based on its unique structural design and operation mode, the neural network can deeply analyze the complex pattern information hidden in the signal, and can achieve high-precision and accurate identification of various defect types such as minor damage and corrosion in overhead ground wires. This highly accurate recognition capability fundamentally reduces the probability of missed detection and false detection, effectively and effectively guarantees the reliability and stability of the power transmission process, and provides solid technical support for the smooth operation of the power system.

[0067] 2. The present invention can monitor the overhead ground wire in real time, quickly acquire and process the detection data, and timely feedback the operating status of the overhead ground wire. Once a defect is found, an alarm can be quickly issued, so that maintenance personnel can take measures to repair it as soon as possible, effectively avoiding further deterioration of the defect, reducing power outages caused by overhead ground wire failures, ensuring the continuity of power supply, and reducing economic losses.

[0068] 3. The adaptive filtering algorithm can automatically and accurately adjust the filtering parameters according to different operating environment conditions and various different detection conditions. This adaptive adjustment capability ensures that under various complex and changeable working conditions, the system can stably and accurately extract truly valuable and useful signals from the original detection signals, providing a solid data foundation for subsequent defect identification and analysis. Similarly, the good self-learning and adaptive capabilities of the Elman neural network play a key role in the long-term operation process. With the passage of time and the change of the operating state of the overhead ground wire, the defect characteristics of the overhead ground wire will change dynamically due to the continuous influence of environmental factors and the effects of various factors such as its own aging. The Elman neural network can continuously optimize its own model parameters based on the newly acquired data samples through its internal learning mechanism, so as to always maintain a high degree of adaptability to the changes in the defect characteristics of the overhead ground wire. This feature not only significantly improves the intelligence level of the entire detection system, enabling it to better cope with complex and changeable actual operating conditions, but also greatly enhances the long-term reliability of the detection system. From the perspective of long-term operation, the frequency of manual intervention and maintenance costs are effectively reduced.

[0069] Therefore, the present invention adopts the above-mentioned Elman neural network overhead ground wire detection method based on adaptive filtering, which solves the problems existing in the existing overhead ground wire defect detection method, can effectively suppress the noise signal in the original signal and improve the defect identification success rate, thereby improving the safety of overhead ground wire operation.

[0070] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. An overhead ground wire detection method based on Elman neural network with adaptive filtering, characterized in that: The following steps are involved: S1. Use a sensor array to collect residual magnetic information of the overhead ground wire, and divide the collected residual magnetic information data into equal intervals to produce a sample data set; S2, the sample data set is subjected to wavelet transform using Mallet algorithm and decomposed into high-frequency detail signals and low-frequency approximate signals; S3, performing adaptive filtering on each segment of data processed by S2, and obtaining noise-reduced data after screening; S4, using a cubic spline interpolation method to perform circumferential interpolation on the filtered denoised data, and converting the filtered denoised data into a grayscale image; S5, thresholding the grayscale image of S4, automatically obtaining the threshold, and obtaining a grayscale image of overhead ground wire defects with a resolution of 300×400; S6, extracting texture features of the defect grayscale image to obtain a quantitative recognition feature vector; S7. Use the Elman neural network to identify defects, use the test set to verify the effectiveness of the Elman neural network, and realize the accurate detection function of overhead ground wire defects.

2. The method for detecting overhead ground wires based on an Elman neural network using adaptive filtering according to claim 1, characterized in that: The sensor array of S1 consists of 18 high-precision giant magnetoresistance (GMR) sensors evenly arranged around the overhead ground wire.

3. The method for detecting overhead ground wires based on an Elman neural network using adaptive filtering according to claim 1, characterized in that: The calculation formula of the Mallat algorithm of S2 is: Among them, c j+1,k is the low-frequency approximate output of the signal, d j+1,k is the high-frequency detail output of the signal, j is the signal level, k is the wavelet translation multiple, h0 is the low-pass filter coefficient, h1 is the high-pass filter coefficient, m is the wavelet magnification multiple, c j,m is the input signal at j signal level.

4. The method for detecting overhead ground wires based on an Elman neural network using adaptive filtering according to claim 1, characterized in that: The calculation formula of the adaptive filtering algorithm of S3 is: e(j)=d(j)-W T (j)X(j); Where e is the difference between the output signal and the expected signal, d is the expected signal, j is the signal level, and W T (j)X(j) is the output signal, W is the weighted signal, and X is the input signal.

5. The method for detecting overhead ground wires based on an Elman neural network using adaptive filtering according to claim 1, characterized in that: The S4 uses grayscale normalization of MFL data for preprocessing, and the calculation formula is: Among them, x and y are the values ​​before and after normalization respectively; V max , V min are the maximum and minimum values ​​of the original grayscale image respectively.

6. The method for detecting overhead ground wires based on an Elman neural network using adaptive filtering according to claim 1, characterized in that: In the S5, the data is thresholded to automatically obtain the threshold, the maximum value of the defect position data is calculated, the centered maximum value is taken, and 150 pixels are intercepted in the axial direction; the circumferential resolution is set to 400 to obtain a 300×400 defect grayscale image.

7. The method for detecting overhead ground wires based on an Elman neural network using adaptive filtering according to claim 1, characterized in that: The S6 extracts the texture features and morphological features of the defect image as feature vectors for quantitative identification.

8. The method for detecting overhead ground wires based on an Elman neural network using adaptive filtering according to claim 1, characterized in that: The Elman neural network set in S7 is divided into four layers: input layer, hidden layer, receiving layer and output layer.

9. The method for detecting overhead ground wires based on an Elman neural network using adaptive filtering according to claim 8, characterized in that: The number of hidden layer nodes of the Elman neural network set in S7 is set to 22.

10. The method for detecting overhead ground wires based on an Elman neural network using adaptive filtering according to claim 1, characterized in that: Training samples are randomly extracted from the samples, and the remaining samples are used as test samples. The function of the back propagation training algorithm is set to trainbfg.

Images