Substation primary equipment partial discharge source joint positioning method based on ultrasonic and infrared fusion

Abstract

The application discloses a substation primary equipment partial discharge source joint positioning method based on ultrasonic and infrared fusion, which comprises the following steps: 1, obtaining infrared images, ultrasonic images and phase images of the primary equipment; 2, pre-processing the infrared images and ultrasonic signal images of the primary equipment obtained in step 1; 3, fusing the infrared images and ultrasonic signal images based on a maximum selection-based fusion rule; and 4, realizing joint positioning of the partial discharge source based on the fused infrared ultrasonic images. The joint positioning method is more time-saving and labor-saving, the method can be used for fusing images and signal images, can reduce the influence of adverse factors, improve the recognition degree of the target source, and further ensure accurate positioning of the partial discharge source, has high detection efficiency and strong defect depth detection capability. The method avoids the interference of subjective judgment factors, is more accurate than the traditional detection method, and is safer in the detection process.

Description

Technical Field

[0001] This invention belongs to the field of equipment fault detection, and in particular relates to a method for joint localization of partial discharge sources in primary equipment of substations by combining ultrasonic and infrared technologies. Background Technology

[0002] Primary equipment in substations is crucial for power generation and distribution. During manufacturing and transportation, insulation defects are unavoidable. Under the influence of a high-voltage electric field, these defects can lead to partial discharges. The accumulation of these weak currents can cause failures in the insulation system and high-voltage components, resulting in significant economic losses. Partial discharge occurs in the early stages of insulation degradation. Since partial discharge is a major cause of insulation aging in power equipment, rapid and effective detection is essential. However, precise location of partial discharges is challenging due to interference signals at the detection site. Nevertheless, accurately identifying the source of the partial discharge allows for a more precise understanding of the equipment's insulation condition and the development of appropriate maintenance strategies. Therefore, the detection and location of partial discharges are critical for ensuring the safe operation of power equipment. Partial discharge detection is used to detect insulation defects. When a discharge occurs, charged and neutral particles collide. When partial discharge occurs in an insulating structure, it is accompanied by electrical pulses, ultrasonic waves, electromagnetic radiation, light, chemical reactions, and local heating. Most existing technologies locate the partial discharge through infrared imaging, ultrasonic partial discharge location, or infrared ultrasonic location, which uses an excitation signal to raise the temperature at the defect and then uses an infrared thermal imager to acquire an infrared image. This method is easily affected by external factors and involves a series of problems such as noise reduction, feature analysis, and image optimization during image processing, making it relatively complex. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention provides a method for joint localization of partial discharge sources in primary equipment of substations using a fusion of ultrasound and infrared technologies. The fused image obtained using this method can avoid the influence of adverse factors, improve the identification of partial discharge signals, and facilitate the localization of partial discharge sources in primary equipment.

[0004] To address the problems existing in the prior art, the present invention adopts the following technical solution:

[0005] A method for joint localization of partial discharge sources in primary equipment of substations, which integrates ultrasonic and infrared technologies, includes the following steps:

[0006] Step 1: Acquire the infrared thermogram, ultrasonic signal spectrum, and ultrasonic phase map of the primary equipment;

[0007] Step 2: Preprocess the infrared temperature difference image, ultrasonic signal spectrum, and ultrasonic phase image corresponding to the primary equipment obtained in Step 1. The preprocessing includes: dividing the infrared temperature difference image into target regions using the partial discharge signal distribution law of the ultrasonic signal spectrum and ultrasonic phase image, and dividing the ultrasonic signal spectrum and ultrasonic phase image into target regions using the thermal distribution law of the infrared temperature difference image.

[0008] Step 3: The preprocessed infrared thermogram, ultrasonic signal spectrum, and ultrasonic phase map are fused using a fusion rule based on maximum selection to form an infrared ultrasonic image;

[0009] Step 4: Import the fused infrared ultrasound image into the primary equipment image captured by the camera, and match the infrared ultrasound image information with the real equipment image to locate the partial discharge source.

[0010] Furthermore, in step 1, the infrared temperature difference image is used to obtain the thermal distribution pattern, and the ultrasonic signal spectrum and the corresponding ultrasonic phase image are used to obtain the partial discharge signal distribution pattern.

[0011] Furthermore, step 2, which involves using the partial discharge signal distribution patterns of the ultrasonic signal spectrum and ultrasonic phase map to divide the target region of the infrared thermogram, includes:

[0012] Step 2.1: Based on the distribution pattern of partial discharge signals, use the threshold segmentation method to map the target area distribution of partial discharge signals within the threshold to the infrared image of the primary equipment.

[0013] Step 2.2: Delineate the target area with significantly increased temperature, set a temperature threshold, enhance the temperature difference distribution pattern of the target image using matched filtering, and use curvelet transform to segment the target image to separate the pattern, thereby obtaining the temperature difference distribution pattern map of the target image.

[0014] Furthermore, the division of target regions in the ultrasound signal spectrum and ultrasound phase map using the thermal distribution pattern of infrared thermograms includes: using the original infrared thermogram, setting a partial discharge threshold, and obtaining the ultrasound signal spectrum and corresponding partial discharge phase map of the infrared thermogram region with significant amplitude changes within the threshold.

[0015] Furthermore, step 2.2, which involves enhancing the temperature difference distribution pattern of the target image using matched filtering and segmenting the target image using curvelet transform to separate the pattern, includes:

[0016] a. Perform curve transformation on the response of the enhanced vascular image;

[0017] b. Preserve the low-frequency components and remove all other coefficients;

[0018] c. Apply the inverse fast Fourier transform with unequal spacing;

[0019] d. The temperature threshold is the average of the image's pixel values;

[0020] e. Apply length filtering and remove misclassified pixels.

[0021] Further, step 3 specifically includes: based on the temperature distribution pattern of the target image, using the fuzzy ISODATA algorithm based on amplitude-phase extremum to extract data from the segmented ultrasonic signal spectrum and phase map according to the temperature distribution pattern, and generating an ultrasonic infrared image that fuses the temperature and ultrasonic signal.

[0022] Furthermore, the fuzzy ISODATA algorithm based on amplitude-phase extremum scaling includes:

[0023] 3.1) Selecting the initial cluster center Z i (0), the infrared pulse distribution of the target image is used as the control group and divided into c groups. The ultrasound signal data and the corresponding phase information are divided into c groups. The mean of each ultrasound data sample is used as the first cluster center. Then, a mean square error is added and subtracted in each feature direction to obtain a total of c(2n+1) cluster centers, where n is the dimension of the sample.

[0024] 3.2) Calculate the membership matrix U(0). The matrix elements are calculated as follows:

[0025]

[0026] In the formula, d ij It is the initial cluster center Z of the i-th class from the j-th sample. i The distance (0) is specified in the following rule to avoid the denominator being 0: if d ij =0, then μ ij =1,μ pj (0) = 0 (p ≠ i). Therefore, d ij The larger μ is ij (0) The smaller;

[0027] 3.3) Find new cluster centers Z for each type of cluster. i (L), where L is the number of iterations:

[0028]

[0029] In the formula, the parameter m≥2 is a constant that controls the degree of fuzziness of the clustering results;

[0030] 3.4) Calculate the new membership matrix U(L+1). The matrix elements are calculated as follows:

[0031]

[0032] In the formula, d ij When the Lth iteration is completed, the j-th sample moves to the initial cluster center Z of the i-th class. i The distance (L). To avoid the denominator being 0, it is specifically stipulated that: if d ij =0, then μ ij (L+1)=1, μ pj (L+1)=0(p≠i). Therefore, d ij The larger μ is ij The smaller (L+1) is;

[0033] 3.5) Return to step 3.3) and repeat until convergence, the convergence condition being max{μ ij (L+1)-μ ij (L)}≤ε, where ε is a specified parameter;

[0034] 3.6) Construct the criterion matrix:

[0035]

[0036] In the formula, ||X j -Z i || 2 Let J represent the Euclidean distance between the j-th sample and the initial cluster center of the i-th class; J represents the sum of squared distances between all samples to be clustered and their respective class centers.

[0037] Infrared images, ultrasonic signal spectra, and ultrasonic phase diagrams of the primary equipment were acquired using experimental equipment to obtain the thermal distribution and partial discharge signal distribution patterns of the primary equipment.

[0038] Furthermore, step 2 includes the following steps:

[0039] Step 2.1: Based on the partial discharge signal condition, preprocess the infrared image of the primary equipment using the threshold segmentation method;

[0040] Step 2.2: Delineate the target area with significantly increased temperature, enhance the temperature difference distribution pattern of the target image using matched filtering, set a temperature threshold, and segment the target image using curvelet transform. The steps are as follows:

[0041] 2.21) Based on step 2.1, the partial discharge source distribution is obtained based on the ultrasonic signal spectrum and phase map. Then, the phase T(n) of the infrared image is analyzed using Fast Fourier Transform (DFT). Let (p, q) be the sampling points of the network. The Fast Fourier Transform of the infrared image T(K) is expressed as:

[0042]

[0043]

[0044] In the formula, W N Let W be the transformation kernel function of the N-point DFT. N =l -j2π / N .

[0045] 2.22) Perform region segmentation on the infrared image of the primary equipment and select the region with a significant increase in temperature, i.e., the region where the partial discharge source is located;

[0046] 2.23) A temperature threshold is set, and a matched filter is used to enhance the temperature distribution pattern of the target image, which can be approximated by a Gaussian-shaped curve. Based on this, a two-dimensional matched filter kernel is designed to convolve with the enhanced image to amplify the pattern. A prototype matched filter kernel is represented as:

[0047] f(x,y)=-exp(-x 2 / (2σ 2 for |y|≤L / 2

[0048] In the formula, σ is the standard deviation of the noise, L is the length of a temperature texture assumed to have a fixed direction, and the negative sign in the Gaussian function indicates the variability in the temperature distribution. Here, it is assumed that the image orientation is aligned along the y-axis. Because the image can face any angle, the kernel needs to be rotated for all possible angles. Therefore, this kernel is in 12 different directions, each 15°, preserving the maximum response of each pixel.

[0049] 2.24) The target image is segmented using curvelet transformation, and the texture extraction algorithm is as follows:

[0050] a. Perform curve transformation on the response of the enhanced vascular image;

[0051] b. Preserve the low-frequency components and remove all other coefficients;

[0052] c. Apply inverse transform;

[0053] d. The threshold is the average of the image's pixel values;

[0054] e. Apply length filtering and remove misclassified pixels.

[0055] Step 2.3: Based on the original infrared thermogram, segment the ultrasonic signal spectrum and corresponding phase map with significant amplitude changes, and adapt them to the target region. The steps are as follows:

[0056] 2.31) Data was acquired using full-matrix acquisition mode of full-focus imaging technology to obtain the ultrasonic partial discharge signal spectrum and phase diagram of the primary equipment.

[0057] 2.32) Based on the original infrared thermoelectric image, a partial discharge threshold is set to obtain an ultrasonic signal spectrum with significant amplitude changes and a corresponding partial discharge phase map, which is adapted to the infrared target image.

[0058] Step 3: Based on the temperature distribution texture map of the target image, the fuzzy ISODATA algorithm based on amplitude-phase extremum is used to extract the data of the segmented ultrasonic signal spectrum and phase map to generate an ultrasonic infrared image.

[0059] Furthermore, step 3 includes the following steps:

[0060] 3.1) Selecting the initial cluster center Z i (0), the infrared pulse distribution of the target image is used as the control group and divided into group c. The ultrasound signal data is also divided into group c. The mean of each ultrasound data sample is used as the first cluster center. Then, a mean square error is added and subtracted in each feature direction to obtain a total of c(2n+1) cluster centers, where n is the dimension of the sample.

[0061] 3.2) Calculate the membership matrix U(0). The matrix elements are calculated as follows:

[0062]

[0063] In the formula, d ij It is the initial cluster center Z of the i-th class from the j-th sample. i The distance of (0). To avoid the denominator being 0, it is specifically stipulated that: if d ij =0, then μ ij =1,μ pj (0) = 0 (p ≠ i). Therefore, d ij The larger μ is ij (0) The smaller.

[0064] 3.3) Find new cluster centers Z for each type of cluster. i (L), where L is the number of iterations.

[0065]

[0066] In the formula, the parameter m≥2 is a constant that controls the degree of fuzziness of the clustering results.

[0067] 3.4) Calculate the new membership matrix U(L+1). The matrix elements are calculated as follows:

[0068]

[0069] In the formula, d ij When the Lth iteration is completed, the j-th sample moves to the initial cluster center Z of the i-th class. iThe distance (L). To avoid the denominator being 0, it is specifically stipulated that: if d ij =0, then μ ij (L+1)=1, μ pj (L+1)=0(p≠i). Therefore, d ij The larger μ is ij The smaller (L+1) is.

[0070] 3.5) Return to step 3.3) and repeat until convergence. The convergence condition is max{|μ ij (L+1)-μ ij (L)|}≤ε, where ε is a specified parameter.

[0071] 3.6) Construct the criterion matrix:

[0072]

[0073] In the formula, ||X j -Z i || 2 Let J represent the Euclidean distance between the j-th sample and the initial cluster center of the i-th class; J represents the sum of squared distances between all samples to be clustered and their respective class centers.

[0074] Step 4: Import the fused infrared ultrasound image into the primary equipment image captured by the camera to achieve joint localization of the partial discharge source.

[0075] The advantages and beneficial effects of this invention are:

[0076] This invention is a method for joint localization of partial discharge sources in substation primary equipment, combining ultrasonic and infrared technologies. It optimizes infrared thermography and combines it with ultrasonic partial discharge signal detection, achieving a fusion of ultrasonic and infrared identification techniques. This improves the identification of partial discharge sources, is simple to operate, highly practical, and its principles are easy to understand. The detection method is more time-saving and labor-saving. By fusing image and signal spectra, it reduces the influence of adverse factors, improves the identification of target sources, and ensures accurate localization of partial discharge sources. It boasts high detection efficiency and strong defect depth detection capabilities. It avoids interference from subjective judgment factors, is more accurate than traditional detection methods, and the detection process is safer. Attached Figure Description

[0077] The present invention will now be described in detail with reference to the accompanying drawings and embodiments:

[0078] Figure 1 This is a schematic diagram of the process of the present invention.

[0079] Figure 2 for Figure 1 Mid-infrared image preprocessing flowchart;

[0080] Figure 3 This is a flowchart of infrared ultrasound fusion.

[0081] Figure 4 Infrared image of the partial discharge source. Figure 4 (a) is the original infrared image. Figure 4 (b) is an infrared thermoelectric image of the vicinity of the partial discharge source based on ultrasonic signal localization. Figure 4 Part of (a);

[0082] Figure 5 This is a phase diagram of ultrasound partial discharge. Figure 5 (a) is the original partial discharge image. Figure 5 (b) is an ultrasonic partial discharge image near the partial discharge source located based on infrared thermography. Figure 5 Part of (a);

[0083] Figure 6 This is a network diagram of infrared temperature difference distribution.

[0084] Figure 7 This is a schematic diagram of infrared ultrasound fusion.

[0085] Figure 8 Schematic diagram of infrared ultrasound fusion image and primary equipment image fusion. Detailed Implementation

[0086] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0087] like Figure 1 As shown, this invention is a method for joint localization of partial discharge sources in primary equipment of a substation using a fusion of ultrasonic and infrared technologies. The method includes the following steps:

[0088] Step 1: Acquire the infrared thermogram, ultrasonic signal spectrum, and ultrasonic phase map of the primary equipment;

[0089] Step 2: Preprocess the infrared temperature difference image, ultrasonic signal spectrum, and ultrasonic phase image corresponding to the primary equipment obtained in Step 1. The preprocessing includes: dividing the infrared temperature difference image into target regions using the partial discharge signal distribution law of the ultrasonic signal spectrum and ultrasonic phase image, and dividing the ultrasonic signal spectrum and ultrasonic phase image into target regions using the thermal distribution law of the infrared temperature difference image.

[0090] Step 3: The preprocessed infrared thermogram, ultrasonic signal spectrum, and ultrasonic phase map are fused using a fusion rule based on maximum selection to form an infrared ultrasonic image;

[0091] Step 4: Import the fused infrared ultrasound image into the primary equipment image captured by the camera, and match the infrared ultrasound image information with the real equipment image to locate the partial discharge source.

[0092] Step 1 includes the following steps:

[0093] Infrared images, ultrasonic signal spectra, and ultrasonic phase diagrams of the primary equipment were acquired using experimental equipment to obtain the thermal distribution and partial discharge signal distribution patterns of the primary equipment.

[0094] Step 2, which utilizes the partial discharge signal distribution patterns of the ultrasonic signal spectrum and ultrasonic phase map to divide the target region in the infrared thermogram, includes the following steps:

[0095] Step 2.1: Based on the distribution pattern of partial discharge signals, use the threshold segmentation method to match the region within the partial discharge signal threshold with the infrared image of the primary equipment;

[0096] Step 2.2: Delineate the target area with significantly increased temperature, set a temperature threshold, enhance the temperature difference distribution pattern of the target image using matched filtering, and segment the target image using curvelet transform.

[0097] Specifically, the distribution pattern of partial discharge sources is obtained from the ultrasonic signal spectrum and phase map. The phase T(n) of the infrared image is analyzed using Fast Fourier Transform (DFT). Let (p, q) be the sampling points of the network. The Fast Fourier Transform of the infrared image T(K) is expressed as:

[0098]

[0099]

[0100] In the formula, W N Let W be the transformation kernel function of the N-point DFT. N =l -j2π / N The infrared image of the primary equipment is segmented to identify areas with significantly elevated temperatures, which are the regions where partial discharge sources are located.

[0101] By setting a temperature threshold and employing a matched filter to enhance the temperature distribution patterns in the target image, a pattern that can be approximated by a Gaussian-shaped curve can be used. Based on this, a two-dimensional matched filter kernel is designed to convolve with the enhanced image to amplify the patterns. A prototype matched filter kernel is represented as:

[0102] f(x,y)=-exp(-x 2 / (2σ 2 for |y|≤L / 2

[0103] In the formula, σ is the standard deviation of the noise, L is the length of a temperature texture assumed to have a fixed direction, and the negative sign in the Gaussian function indicates the variability of the temperature distribution. Here, it is assumed that the image orientation is aligned along the y-axis. Because the image can face any angle, the kernel needs to be rotated for all possible angles. Therefore, this kernel is in 12 different directions, each 15°, preserving the maximum response of each pixel;

[0104] The target image is segmented using curvelet transform, and the texture extraction process is as follows:

[0105] a. Perform curve transformation on the response of the enhanced vascular image;

[0106] b. Preserve the low-frequency components and remove all other coefficients;

[0107] c. Apply the inverse fast Fourier transform with unequal spacing;

[0108] d. The threshold is the average of the image's pixel values;

[0109] e. Apply length filtering and remove misclassified pixels.

[0110] The target region is divided using the thermal distribution patterns of infrared thermograms and ultrasonic phase maps: Based on the original infrared thermogram, the ultrasonic signal spectrum and corresponding phase map with significant amplitude changes are segmented to fit the infrared target region. Specifically:

[0111] Data is acquired using full-matrix acquisition mode of full-focus imaging technology. After obtaining the ultrasonic partial discharge signal spectrum and partial discharge signal phase map of the primary equipment, based on the original infrared thermoelectric image, a partial discharge threshold is set to obtain the ultrasonic signal spectrum and corresponding partial discharge phase map with significant amplitude changes, which is adapted to the infrared target image.

[0112] Using the temperature distribution texture map of the target image, a fuzzy ISODATA algorithm based on amplitude-phase extremum is employed to extract the data from the segmented ultrasound signal spectrum and phase map, generating an ultrasound infrared image. The extraction of the segmented ultrasound signal spectrum and phase map data using the fuzzy ISODATA algorithm based on amplitude-phase extremum includes the following steps:

[0113] 3.1) Selecting the initial cluster center Z i (0), the infrared pulse distribution of the target image is used as the control group and divided into group c. The ultrasound signal data is also divided into group c. The mean of each ultrasound data sample is used as the first cluster center. Then, a mean square error is added and subtracted in each feature direction to obtain a total of c(2n+1) cluster centers, where n is the dimension of the sample.

[0114] 3.2) Calculate the membership matrix U(0). The matrix elements are calculated as follows:

[0115]

[0116] In the formula, d ij It is the initial cluster center Z of the i-th class from the j-th sample. i The distance of (0). To avoid the denominator being 0, it is specifically stipulated that: if d ij =0, then μ ij =1,μ pj (0) = 0 (p ≠ i). Therefore, d ij The larger μ is ij (0) The smaller.

[0117] 3.3) Find new cluster centers Z for each type of cluster. i (L), where L is the number of iterations.

[0118]

[0119] In the formula, the parameter m≥2 is a constant that controls the degree of fuzziness of the clustering results.

[0120] 3.4) Calculate the new membership matrix U(L+1). The matrix elements are calculated as follows:

[0121]

[0122] In the formula, d ij When the Lth iteration is completed, the j-th sample moves to the initial cluster center Z of the i-th class. i The distance (L). To avoid the denominator being 0, it is specifically stipulated that: if d ij =0, then μ ij (L+1)=1, μ pj (L+1)=0(p≠i). Therefore, dij The larger μ is ij The smaller (L+1) is.

[0123] 3.5) Return to step 3.3) and repeat until convergence. The convergence condition is... Where ε is a specified parameter.

[0124] 3.6) Construct the criterion matrix:

[0125]

[0126] In the formula, ||X j -Z i || 2 Let J represent the Euclidean distance between the j-th sample and the initial cluster center of the i-th class; J represents the sum of squared distances between all samples to be clustered and their respective class centers.

[0127] Finally, the fused infrared ultrasound image is imported into the primary equipment image captured by the camera to achieve joint localization of the partial discharge source.

[0128] Figure 4 (a) to (b) represent the infrared target region acquisition. Figure 4 (a) is the original infrared image. Figure 4 (b) is an infrared thermoelectric image of the vicinity of the partial discharge source based on ultrasonic signal localization. Figure 4 Part of (a), corresponding to step 2.2, Figure 5 Images (a) to (b) show the acquisition of the ultrasound partial radiation target area. Figure 5 (a) is the original partial discharge image. Figure 4 (b) is an ultrasonic partial discharge image near the partial discharge source located based on infrared thermography. Figure 5 Part of (a), corresponding to step 2.2, Figure 6 Temperature texture map after curve enhancement segmentation Figure 7 This is an infrared ultrasound diagram combining temperature pulses and ultrasonic partial discharge signals. Areas with significant changes in temperature pulses and partial discharge amplitude are clearly visible. Figure 8 It is a partial discharge source localization map obtained by fusing infrared ultrasound fusion image with primary equipment image, thereby realizing the localization of partial discharge source.

Claims

1. A method for joint localization of partial discharge sources in primary equipment of a substation, combining ultrasonic and infrared technologies, characterized in that... The method includes the following steps: Step 1: Acquire the infrared thermogram, ultrasonic signal spectrum, and ultrasonic phase map of the primary equipment; Step 2: Preprocess the infrared temperature difference image, ultrasonic signal spectrum, and ultrasonic phase image corresponding to the primary equipment obtained in Step 1. The preprocessing includes: dividing the infrared temperature difference image into target regions using the partial discharge signal distribution law of the ultrasonic signal spectrum and ultrasonic phase image, and dividing the ultrasonic signal spectrum and ultrasonic phase image into target regions using the thermal distribution law of the infrared temperature difference image. Step 3: The preprocessed infrared thermogram, ultrasonic signal spectrum, and ultrasonic phase map are fused using a fusion rule based on maximum selection to form an infrared ultrasonic image; Step 4: Import the fused infrared ultrasound image into the primary equipment image captured by the camera, and match the infrared ultrasound image information with the real equipment image to locate the partial discharge source.

2. The method for joint localization of partial discharge sources in primary equipment of a substation using ultrasonic and infrared fusion as described in claim 1, characterized in that, In step 1, the infrared temperature difference image is used to obtain the thermal distribution pattern, and the ultrasonic signal spectrum and the corresponding ultrasonic phase image are used to obtain the partial discharge signal distribution pattern.

3. The method for joint localization of partial discharge sources in primary equipment of a substation using a fusion of ultrasound and infrared as described in claim 2, characterized in that, Step 2, which involves using the partial discharge signal distribution patterns of the ultrasonic signal spectrum and ultrasonic phase map to divide the target region in the infrared thermogram, includes: Step 2.1: Based on the distribution pattern of partial discharge signals, use the threshold segmentation method to map the target area of ​​partial discharge signal distribution within the threshold range to the infrared image of the primary equipment, thereby achieving region segmentation of the infrared image; Step 2.2: Delineate the target area with significantly increased temperature, set a temperature threshold, enhance the temperature difference distribution pattern of the target image using matched filtering, and use curvelet transform to segment the target image to separate the pattern, thereby obtaining the temperature difference distribution pattern map of the target image.

4. The method for joint localization of partial discharge sources in primary equipment of a substation using ultrasonic and infrared fusion as described in claim 1, characterized in that, The method of dividing the target region of ultrasound signal spectrum and ultrasound phase map by using the thermal distribution law of infrared thermogram includes: using the original infrared thermogram, setting a partial discharge threshold, and obtaining the ultrasound signal spectrum and corresponding partial discharge phase map of the infrared thermogram region with significant amplitude changes within the threshold.

5. The method for joint localization of partial discharge sources in primary equipment of a substation using ultrasonic and infrared fusion as described in claim 3, characterized in that, Step 2.2, which involves enhancing the temperature difference distribution pattern of the target image using matched filtering and segmenting the target image using curvelet transform to separate the pattern, includes: a. Perform curve transformation on the response of the enhanced vascular image; b. Preserve the low-frequency components and remove all other coefficients; c. Apply the inverse fast Fourier transform with unequal spacing; d. The temperature threshold is the average of the image's pixel values; e. Apply length filtering and remove misclassified pixels.

6. The method for joint localization of partial discharge sources in primary equipment of substations using ultrasound and infrared fusion as described in claim 3, characterized in that, Step 3 specifically includes: based on the temperature distribution pattern of the target image, using the fuzzy ISODATA algorithm based on amplitude-phase extremum to extract data from the segmented ultrasonic signal spectrum and phase map according to the temperature distribution pattern, and generating an ultrasonic infrared image that fuses the temperature and ultrasonic signal.

7. The method for joint localization of partial discharge sources in primary equipment of substations using ultrasound and infrared fusion as described in claim 6, characterized in that, The fuzzy ISODATA algorithm based on amplitude-phase extremum scaling includes: 3.1) Selecting the initial cluster center Z i (0), the infrared pulse distribution of the target image is used as the control group and divided into c groups. The ultrasound signal data and the corresponding phase information are divided into c groups. The mean of each ultrasound data sample is used as the first cluster center. Then, a mean square error is added and subtracted in each feature direction to obtain a total of c(2n+1) cluster centers, where n is the dimension of the sample. 3.2) Calculate the membership matrix U(0). The matrix elements are calculated as follows: In the formula, d ij It is the initial cluster center Z of the i-th class from the j-th sample. i The distance (0) is specified in the following rule to avoid the denominator being 0: if d ij =0, then μ ij =1,μ pj (0) = 0 (p ≠ i). Therefore, d ij The larger μ is ij (0) The smaller; 3.3) Find new cluster centers Z for each type of cluster. i (L), where L is the number of iterations: In the formula, the parameter m≥2 is a constant that controls the degree of fuzziness of the clustering results; 3.4) Calculate the new membership matrix U(L+1). The matrix elements are calculated as follows: In the formula, d ij When the Lth iteration is completed, the j-th sample moves to the initial cluster center Z of the i-th class. i The distance (L). To avoid the denominator being 0, it is specifically stipulated that: if d ij =0, then μ ij (L+1)=1, μ pj (L+1)=0(p≠i). Therefore, d ij The larger μ is ij The smaller (L+1) is; 3.5) Return to step 3.3) and repeat until convergence, the convergence condition being max{|μ ij (L+1)-μ ij (L)|}≤ε, where ε is a specified parameter; 3.6) Construct the criterion matrix: In the formula, ||X j -Z i || 2 Let J represent the Euclidean distance between the j-th sample and the initial cluster center of the i-th class; J represents the sum of squared distances between all samples to be clustered and their respective class centers.

Images