Transformer substation electrical equipment fault detection method based on Zernike moment and support vector machine

Abstract

The invention discloses a transformer substation electrical equipment fault detection method based on Zernike moment and a support vector machine. The method comprises the following steps: acquiring infrared image data of transformer substation electrical equipment; performing feature extraction on the infrared image data of the electrical equipment of the transformer substation through Zernike moments to obtain feature vectors; feature vectors are classified through a support vector machine based on a binary decision tree, a priority concept is introduced, and two classes with minimum class spacing are combined to be classified firstly, so that interference to subsequent classification can be eliminated, meanwhile, the number of SVM classifiers is reduced, and the fault type is obtained. The method provided by the invention not only improves the accuracy and efficiency of fault detection, but also realizes continuous monitoring of the operation state of the substation equipment, and is helpful for timely finding hidden dangers and implementing preventive maintenance measures, thereby avoiding power grid interruption caused by sudden faults and ensuring stable operation of a power system.

Description

Technical Field

[0001] The present invention mainly relates to the technical field of power grids, and in particular to a method for detecting faults of electrical equipment in a substation based on Zernike moments and support vector machines. Background Art

[0002] With the rapid development of the power industry, improving the safety and reliability of electrical equipment has become a key area of ​​concern in recent years. A series of measures have been taken in this field to reduce the failure rate. For more than 50 years, different maintenance and repair methods have been used in the industry. The grid status monitoring methods currently proposed include post-fault repair, maintenance and preventive repair, and efficiency repair measures. In addition to the above methods, fault prevention is currently the most important method in the maintenance and repair system.

[0003] The heating parts in the substation equipment are usually caused by loose joints, oxidation and corrosion, phase asymmetry, and damage to coil insulators. All these problems can be detected using thermal imagers in the temperature measurement of equipment, such as transformers, fuses, cables, panel equipment and all electrical equipment between overhead or ground substations. Therefore, most of them use thermal imaging technology to detect faults and prevent faults to improve the reliability of the power grid. At the same time, among thermal vision and various prevention methods, the analysis and interpretation of thermal images mostly adopt traditional manual methods, and a systematic intelligent fault detection technology has not yet been proposed. Due to the wide distribution and diverse types of substation equipment in substations, the issue of digital transformation is particularly important. Summary of the invention

[0004] In view of the technical problems existing in the prior art, the present invention provides a substation electrical equipment fault detection method based on Zernike moments and support vector machines, which improves the accuracy and efficiency of fault detection.

[0005] In order to solve the above technical problems, the technical solution proposed by the present invention is:

[0006] A method for detecting faults of electrical equipment in a substation based on Zernike moments and support vector machines comprises the following steps:

[0007] Acquire infrared image data of electrical equipment in substations;

[0008] The feature vector is obtained by extracting the feature of infrared image data of electrical equipment in substation through Zernike moment.

[0009] The feature vectors are classified by using support vector machine improved based on binary decision tree, the concept of priority is introduced, and the improved multi-class classification algorithm is used to obtain the fault type.

[0010] Preferably, the feature vectors are classified by a support vector machine based on a binary decision tree improvement, the concept of priority is introduced, and the specific process of obtaining the fault type by using an improved multi-class classification algorithm is as follows:

[0011] Each time the k-type fault training samples are trained to generate a binary tree structure, the two classes with the smallest class distance are first merged into a new cluster I to obtain a new parent node, and a child SVM classifier is trained to generate a child SVM classifier; similarly, the two classes with the smallest class distance in the remaining k-2 classes of samples are merged into a new cluster II, and another child SVM classifier is trained to generate another child SVM classifier, until the root node of the binary tree is finally obtained, that is, the training is completed; the classification process of the test sample is the reverse direction of the training process, starting from the root node for judgment, and generating left and right child nodes after the SVM classifier of the corresponding node is tested. If the child node is a leaf node, the electrical fault category to which the sample belongs is directly determined, otherwise the test continues until the leaf node is found.

[0012] Preferably, the class spacing d ij The calculation formula is:

[0013]

[0014] Where: represents the class center of the fault sample, usually the average value of all sample points; X i are all samples of class i in the training set; |X i | for X i The number of samples in .

[0015] Preferably, after the infrared image data of the electrical equipment in the substation is acquired, the infrared image data is converted into a binary image or a grayscale image.

[0016] Preferably, the specific process of extracting features from infrared image data of electrical equipment in a substation by Zernike moments to obtain feature vectors is as follows:

[0017] Convert the infrared image to the polar coordinate system to obtain a series of radial and angular coordinates;

[0018] The Zernike moment at each pixel is calculated by the Zernike polynomial order, and the Zernike moments of all pixels are combined to form a feature vector.

[0019] Preferably, the Zernike polynomials are an orthogonal set of complex-valued polynomials, defined as formula (1):

[0020]

[0021] In the formula, |m|<n,n> 0,x 2 +y 2 ≤1, and n-|m| is an even number, Rnm is a radial polynomial.

[0022] Preferably, the radial polynomial R nm The definition of is as shown in formula (2)-(3):

[0023]

[0024] Where, the complex Zernike moment of order n and repeated m times is given by formula (4):

[0025]

[0026] When calculating the Zernike moment ZM, the center of the image is located at the center of the circle and the rest of the image is projected into the circle.

[0027] The present invention also discloses a computer program product, comprising a computer program, wherein the computer program executes the steps of the method described above when executed by a processor.

[0028] The present invention further discloses a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the method described above are executed.

[0029] The present invention also discloses a substation electrical equipment fault detection system based on Zernike moments and support vector machines, comprising an interconnected memory and a processor, wherein a computer program is stored on the memory, and when the computer program is run by the processor, the steps of the method described above are executed.

[0030] Compared with the prior art, the advantages of the present invention are:

[0031] The present invention uses a high-precision thermal imager to perform non-contact infrared scanning on key electrical equipment in the substation, especially for vulnerable parts such as fuses, and effectively distinguishes common fault types such as cable ear damage and fuse bottom breakage through its real infrared image data, thereby identifying and locating potential fault sources in infrared images; and in the fault identification stage, relying on the feature vector calculated by ZM in the early stage, the binary decision tree is combined with the vector support machine, and the priority concept is introduced, and an improved multi-class classification algorithm is used to realize fault feature recognition, and automatic classification of image data and automatic classification of different types of faults are realized. The method of the present invention not only improves the accuracy and efficiency of fault detection, but also realizes continuous monitoring of the operating status of substation equipment, which helps to timely discover hidden dangers and implement preventive maintenance measures, thereby avoiding power grid interruptions caused by sudden faults and ensuring the stable operation of the power system. Compared with traditional fault detection methods that rely on manual inspections and experience judgment, the method of the present invention has significantly improved the level of intelligence and reliability.

[0032] The substation electrical equipment fault detection method based on Zernike moment (ZM) and improved support vector machine (SVM) of the present invention integrates thermal imaging technology and support vector machine algorithm in machine learning, uses Zernike moment as image feature extraction means, performs moment analysis on thermal imaging images of substation electrical equipment and extracts statistical features, introduces the concept of priority, combines the two classes with the smallest class distance for classification first, which is beneficial to eliminate interference with subsequent classification, and reduces the number of SVM classifiers (only k-1 need to be constructed), so as to realize intelligent evaluation of equipment health status.

[0033] The method of the present invention not only improves the accuracy and efficiency of fault detection, but also realizes continuous monitoring of the operating status of substation equipment, which helps to timely discover hidden dangers and implement preventive maintenance measures, thereby avoiding power grid interruptions caused by sudden faults and ensuring the stable operation of the power system. Compared with traditional fault detection methods that rely on manual inspections and experience judgment, the method of the present invention has significantly improved the level of intelligence and reliability.

[0034] The present invention not only realizes real-time remote monitoring of the health status of substation equipment, but also provides solid data support for the formulation of predictive maintenance strategies, thereby significantly improving the stability and reliability of power grid operation and reducing the risk of power outages caused by sudden failures and operation and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 The figure is a flow chart of a fault detection method in an embodiment of the present invention.

[0036] Figure 2 It is a schematic diagram of the improved SVM structure in the present invention. DETAILED DESCRIPTION

[0037] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments.

[0038] like Figure 1 As shown, the substation electrical equipment fault detection method based on Zernike moment and support vector machine provided by the embodiment of the present invention includes the steps of:

[0039] S1, obtaining infrared image data around the fault point of the substation electrical equipment through thermal imaging, and setting the pixel coordinates outside the fault point to zero;

[0040] S2, converting the infrared image data around the fault point of the electrical equipment into a binary image;

[0041] S3, calculating the infrared image data of the substation electrical equipment by Zernike moment to obtain a feature vector;

[0042] The Zernike moment is a polynomial based on the polar coordinate system, which is used to fit and describe the local shape in a two-dimensional image. In image processing, the image is usually converted to a polar coordinate system, and then the Zernike polynomial is used to represent the local features of the image. The Zernike moment can effectively capture the edges, corners and other shape features in the image. The specific steps include:

[0043] Convert the image to polar coordinates to obtain a series of radial and angular coordinates;

[0044] Select an appropriate order of the Zernike polynomial (usually a lower order can effectively describe the shape characteristics of the image), calculate the Zernike moment at each pixel (this involves integrating the image in the radial and angular directions), and combine the Zernike moments of all pixels to form a feature vector for subsequent image analysis and processing.

[0045] The Zernike polynomials are an orthogonal set of complex-valued polynomials, defined as in formula (1):

[0046]

[0047] In the formula, |m|<n,n> 0,x 2 +y 2 ≤1, and n-|m| is an even number, the radial polynomial R nm The definition of is as shown in formula (2)-(3):

[0048]

[0049] Where, the complex Zernike moment of order n and repeated m times is given by formula (4):

[0050]

[0051] When calculating the Zernike moment ZM, the center of the image is placed at the center of the circle and the rest of the image is projected onto the circle. The portion of the image that is outside the circle is not considered in the calculation.

[0052] In order to select ZM as the fault feature, three categories are defined for creating feature vectors according to the matrix order (n), specifically:

[0053] The first category, n = 1, 2, ... 6, 7, considers all moments of ZM as eigenvector elements. In this category, the eigenvector contains 19 elements.

[0054] In the second category, n = 4, 5 is chosen. All moments of each order contained in this category are summed up to obtain an eigenvector of size 6.

[0055] In the third category, consider n = 8, 9, 10, 11. The feature vector of this category has 22 elements. These elements are shown in Table 1:

[0056] Table 1. Eigenvector elements based on ZM

[0057]

[0058] S4. Classify the image based on the feature vector obtained in the previous stage, and perform intelligent classification through the SVM improved based on the binary decision tree.

[0059] The support vector kernel is used to map data from the input space to a high-dimensional feature space, making the problem easy to linearize. SVM can always find the global minimum because it usually tries to minimize the bound of the structural risk rather than the empirical risk. The empirical risk is defined as the measured average error rate of the training set, as shown in formula (5):

[0060]

[0061] In the formula, 1 is the number of observations, y i is the class label, x i is a sample vector. The structural risk is defined as the structure derived from the inner classes of functions in the nested subsets, i.e., the subset of functions for which the bound on the actual risk is minimized.

[0062] SVM achieves this goal by minimizing formula (6):

[0063]

[0064] In the formula, α i Positive Lagrange multiplier [10,11].

[0065] In order to improve the defect detection speed and recognition rate, the binary decision tree is combined with the vector support machine, and the priority concept is introduced. The improved multi-class classification algorithm is used to realize fault feature recognition. Figure 2 Shown is a schematic diagram of the improved multi-class SVM structure of the present invention.

[0066] Each time the k-class fault training samples are trained to generate a binary tree structure, the first step is to merge the two classes with the smallest class distance into a new cluster I to obtain a new parent node, and at the same time train to generate a child SVM classifier; similarly, the two classes with the smallest class distance in the remaining k-2 classes are merged into a new cluster II, and another child SVM classifier is trained to generate, until the root node of the binary tree is finally obtained, that is, the training is completed. The classification process of the test sample is the reverse direction of the training process, starting from the root node, and generating left and right child nodes after the SVM classifier of the corresponding node is tested. If the child node is a leaf node, the electrical fault category to which the sample belongs is directly determined, otherwise the test continues until the leaf node is found.

[0067] In the feature space, each class has k-1 distance values ​​from other classes. The distance d between class i and class j is calculated according to the formula shown. ij (i,j=1,2,3…k, and i≠j):

[0068]

[0069] Where: represents the class center of the fault sample, usually the average value of all sample points; X i are all samples of class i in the training set; |X i | for X i The number of samples in .

[0070] Considering the accuracy of fault classification and separation, the fault type of the identified equipment is verified and the average accuracy of the method is defined as:

[0071]

[0072] Where m is the number of experimental runs performed each time the database is randomly divided into sets, is the number of accurate classifications in the ith run, N t is the total number of test images per run.

[0073] In another embodiment, to obtain higher accuracy, the feature vector calculation based on ZM can be performed on the grayscale image, and the feature vector set calculated from the sub-image of the grayscale image fault point can be used as a new feature vector.

[0074] The present invention uses a high-precision thermal imager to perform non-contact infrared scanning on key electrical equipment in the substation, especially for vulnerable parts such as fuses, and effectively distinguishes common fault types such as cable ear damage and fuse bottom breakage through its real infrared image data, thereby identifying and locating potential fault sources in infrared images; and in the fault identification stage, relying on the feature vector calculated by ZM in the early stage, the binary decision tree is combined with the vector support machine, and the priority concept is introduced, and an improved multi-class classification algorithm is used to realize fault feature recognition, and automatic classification of image data and automatic classification of different types of faults are realized. The method of the present invention not only improves the accuracy and efficiency of fault detection, but also realizes continuous monitoring of the operating status of substation equipment, which helps to timely discover hidden dangers and implement preventive maintenance measures, thereby avoiding power grid interruptions caused by sudden faults and ensuring the stable operation of the power system. Compared with traditional fault detection methods that rely on manual inspections and experience judgment, the method of the present invention has significantly improved the level of intelligence and reliability.

[0075] The substation electrical equipment fault detection method based on Zernike moment (ZM) and improved support vector machine (SVM) of the present invention integrates thermal imaging technology and support vector machine algorithm in machine learning, uses Zernike moment as image feature extraction means, performs moment analysis on thermal imaging images of substation electrical equipment and extracts statistical features, introduces the concept of priority, combines the two classes with the smallest class distance for classification first, which is beneficial to eliminate interference with subsequent classification, and reduces the number of SVM classifiers (only k-1 need to be constructed), so as to realize intelligent evaluation of equipment health status.

[0076] The method of the present invention not only improves the accuracy and efficiency of fault detection, but also realizes continuous monitoring of the operating status of substation equipment, which helps to timely discover hidden dangers and implement preventive maintenance measures, thereby avoiding power grid interruptions caused by sudden faults and ensuring the stable operation of the power system. Compared with traditional fault detection methods that rely on manual inspections and experience judgment, the method of the present invention has significantly improved the level of intelligence and reliability.

[0077] The present invention not only realizes real-time remote monitoring of the health status of substation equipment, but also provides solid data support for the formulation of predictive maintenance strategies, thereby significantly improving the stability and reliability of power grid operation and reducing the risk of power outages caused by sudden failures and operation and maintenance costs.

[0078] Specific verification situation: The CPU used is i7-14900k, the GPU is RTX4090, and the python version is 3.8. The batch size used for model training is 8, and the number of training rounds is 500. Two common fault types of substations are selected. It is necessary to distinguish whether the two faults are fuse cable connector faults or fuse base faults. 80% are used for training and 20% are used for testing. The results are shown in Table 2:

[0079] Table 2

[0080]

[0081] It can be seen that compared with linear, quadratic or polynomial methods, the method of the present invention has significant advantages in feature extraction and classification strategies, and can achieve higher accuracy classification in image recognition of different categories.

[0082] The embodiment of the present invention further discloses a computer program product, including a computer program, which executes the steps of the above method when executed by a processor. The embodiment of the present invention further discloses a computer-readable storage medium, on which a computer program is stored, which executes the steps of the above method when executed by a processor. The embodiment of the present invention further discloses a substation electrical equipment fault detection system based on Zernike moments and support vector machines, including a memory and a processor connected to each other, the memory storing a computer program, and the computer program executing the steps of the above method when executed by the processor.

[0083] The product, medium and system of the present invention correspond to the above method and also have the advantages described in the above method.

[0084] The present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by hardware related to computer program instructions. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned method embodiment when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. Computer-readable storage media include: any entity or device that can carry computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. The memory is used to store computer programs and / or modules, and the processor implements various functions by running or executing computer programs and / or modules stored in the memory, and calling data stored in the memory. The memory may include high-speed random access memory and may also include non-volatile memory, such as a hard disk, an internal memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0085] The above are only preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.

Claims

1. A method for detecting faults of electrical equipment in a substation based on Zernike moments and support vector machines, characterized in that: Includes steps: Acquire infrared image data of electrical equipment in substations; The feature vector is obtained by extracting the feature of infrared image data of electrical equipment in substation through Zernike moment. The feature vectors are classified by using support vector machine improved based on binary decision tree, the concept of priority is introduced, and the fault type is obtained by using improved multi-class classification algorithm.

2. The method for detecting faults of electrical equipment in a substation based on Zernike moments and support vector machines according to claim 1, characterized in that: The specific process of classifying feature vectors by using support vector machines improved based on binary decision trees, introducing the concept of priority, and using improved multi-class classification algorithms to obtain fault types is as follows: Each time the k-type fault training samples are trained to generate a binary tree structure, the two classes with the smallest class distance are first merged into a new cluster I to obtain a new parent node, and a child SVM classifier is trained to generate a child SVM classifier; similarly, the two classes with the smallest class distance in the remaining k-2 classes of samples are merged into a new cluster II, and another child SVM classifier is trained to generate another child SVM classifier, until the root node of the binary tree is finally obtained, that is, the training is completed; the classification process of the test sample is the reverse direction of the training process, starting from the root node for judgment, and generating left and right child nodes after the SVM classifier of the corresponding node is tested. If the child node is a leaf node, the electrical fault category to which the sample belongs is directly determined, otherwise the test continues until the leaf node is found.

3. The method for detecting faults of electrical equipment in a substation based on Zernike moments and support vector machines according to claim 2, characterized in that: The class spacing d ij The calculation formula is: Where: represents the class center of the fault sample, usually the average value of all sample points; X i are all samples of class i in the training set; |X i | for X i The number of samples in .

4. The method for detecting faults of electrical equipment in a substation based on Zernike moments and support vector machines according to claim 1, 2 or 3, characterized in that: After acquiring the infrared image data of the electrical equipment in the substation, the infrared image data is converted into a binary image or a grayscale image.

5. The method for detecting faults of electrical equipment in a substation based on Zernike moments and support vector machines according to claim 1, 2 or 3, characterized in that: The specific process of extracting features from infrared image data of substation electrical equipment through Zernike moment and obtaining feature vectors is as follows: Convert the infrared image to the polar coordinate system to obtain a series of radial and angular coordinates; The Zernike moment at each pixel is calculated by the Zernike polynomial order, and the Zernike moments of all pixels are combined to form a feature vector.

6. The method for detecting faults of electrical equipment in a substation based on Zernike moments and support vector machines according to claim 5, characterized in that: The Zernike polynomials are an orthogonal set of complex-valued polynomials, defined as in formula (1): In the formula, |m|<n,n> 0,x 2 +y 2 ≤1, and n-|m| is an even number, R nm is a radial polynomial.

7. The method for detecting faults of electrical equipment in a substation based on Zernike moments and support vector machines according to claim 6, characterized in that: Radial Polynomial R nm The definition of is as shown in formula (2)-(3): Where, the complex Zernike moment of order n and repeated m times is given by formula (4): When calculating the Zernike moment ZM, the center of the image is located at the center of the circle and the rest of the image is projected into the circle.

8. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are performed.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the computer program performs the steps of the method according to any one of claims 1 to 7.

10. A substation electrical equipment fault detection system based on Zernike moments and support vector machines, comprising a memory and a processor connected to each other, wherein a computer program is stored in the memory, characterized in that: When the computer program is executed by a processor, the computer program performs the steps of the method according to any one of claims 1 to 7.

Images