Fault insulator identification method and system

Abstract

The invention discloses a fault insulator identification method and system. An image processing technology and machine learning are combined to establish a zero-value insulator online detection and identification model. The insulator string infrared image processing, feature vector extraction and machine learning-based recognition and diagnosis algorithm are described in detail, the recognition accuracy of the model is high, and the recognition and diagnosis of the low-zero-value insulator string are basically realized. According to the invention, multi-level depth feature extraction is carried out on the infrared image sample of the insulator, and compared with features extracted by an existing depth model, the method has stronger identification capability, and a method of combining the infrared image and the grey wolf optimization algorithm optimization support vector machine is provided to realize detection and identification of the zero-value insulator. The low-zero-value insulator can be accurately identified on the big data level, and manpower, material resources and financial resources are reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of target detection, and in particular relates to a method and system for identifying a faulty insulator. Background Art

[0002] Line insulators are mainly responsible for connecting conductors and electrical insulation, and are an important part of the entire power system. Due to certain defects in the production process of insulators and the influence of the natural environment, insulators will naturally deteriorate, and the insulation performance will continue to weaken. The probability of flashover of insulator strings will increase, which will eventually cause instability in the operation of the power grid, and the probability of power accidents will also increase, bringing adverse effects on production and life. Therefore, regular inspection and timely maintenance of insulators are crucial to maintaining the safety of the power grid.

[0003] At present, the mainstream low-zero value insulator detection methods include: spectroscopy, ultraviolet pulse method, radial temperature method, ultrasonic detection method, etc. However, after analyzing the literature, it is found that many methods have problems such as high risk and complex algorithms, and further in-depth research is needed. Due to the obvious temperature changes of low-zero value insulators in the line, many power companies are gradually using infrared imaging technology to detect low-zero value insulators. However, the current research on online detection of power equipment using image processing technology faces difficulties in image feature extraction. The existing algorithms cannot effectively solve the multi-classification problem of insulator status detection, and face the problem of time-consuming processing of massive data detection and low detection accuracy. Therefore, how to propose a low-zero value insulator detection method for big data is an urgent problem to be solved.

[0004] With the continuous improvement of neural network algorithms, artificial intelligence theory and application research represented by deep learning are increasingly being applied to fault detection and identification. At present, many data mining methods such as BP neural network, genetic algorithm, Petri network and decision tree have been successfully applied to the field of diagnosis and identification of deteriorated insulators. Support vector machines are also used in the field of fault diagnosis of power systems. However, the direct use of support vector machine models to detect insulator samples is not ideal. How to optimize the original support vector machine model and solve the problem of insulator detection in a big data environment is a direction that needs to be focused on at present. At present, grid search method, cuckoo search algorithm, particle swarm algorithm, etc. have been successfully applied to the optimization of parameters of optimization vector machines. The combination of gray wolf optimization algorithm and support vector machine is applied in many fields, such as medical signal recognition, plant species recognition, and medical image recognition. The experimental results have been improved, but it is rarely applied to electrical equipment fault diagnosis. Therefore, the application of combining gray wolf optimization algorithm with support vector machine for low zero value insulator detection and identification is still blank. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a method and system for identifying faulty insulators, which can detect and identify low zero-value insulators by processing infrared images of insulators, extracting multi-level deep features of infrared image samples of insulators, and optimizing support vector machine parameters using the gray wolf optimization algorithm. The method proposed by the present invention has high diagnostic accuracy and efficiency.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is: A method for identifying a faulty insulator, comprising the steps of: Step 1, collecting infrared images of insulators; Step 2, preprocessing the infrared image of the insulator; Step 3, extracting infrared features of insulators; Step 4, establishing an insulator identification model based on the support vector machine optimized by the grey wolf optimization algorithm; Step 5, bringing the characteristic data of step 3 into the insulator identification model to identify low zero value insulators.

[0007] Preferably, in step 2, the preprocessing method for the insulator infrared image is: grayscale enhancement processing is performed on the insulator infrared image to make the image clear and the features obvious. Since the edge of the insulator string is close to the background grayscale value, the insulator string cannot be accurately separated, and histogram equalization is used to make the original image become an image with uniform probability distribution, thereby enhancing the image contrast.

[0008] Preferably, in step 3, the method for extracting the infrared features of the insulator is: using image adaptive threshold segmentation based on the Ostu algorithm, an image gray level is L, and the image gray value is The average gray value of the neighborhood is , Represents the probability that the number of pixels is i, using Represents the joint probability density. When the image is segmented with (s, t) as the threshold, the corresponding target is to separate the insulator string area as a whole from the background area. and background The mean vector of is: ; ; The corresponding total mean vector is: ; The inter-class variance between the target and the background is defined as: ; Optimal segmentation threshold for The threshold corresponding to the maximum value; each time the inter-class variance is calculated, 3st+3(Ls )(Lt ) points need to be accumulated: ; The infrared image of the insulator string taken shows various angles due to the shooting angle. To solve the problem of the tilt angle of the target insulator, the Rondon transform is used to correct the tilt of the insulator. The realization principle of Radon transform is A straight line in plane space Mapped to a point in Radon space , then the Radon transform of the continuous image is: ; Where: is a Dirac function, and finally the directional gradient histogram features of the minimum circumscribed rectangle of the insulator string and the texture features of the grayscale gradient co-occurrence matrix are extracted.

[0009] Preferably, in step 4, the insulator identification model of the support vector machine optimized based on the gray wolf optimization algorithm is: in the sample space, all data are completely mixed together, and it is impossible to find a classification function that can effectively classify the two types of data, then a nonlinear support vector machine is used. The nonlinear support vector machine refers to the kernel function on the basis of the linear separable support vector machine. At this time, the convex quadratic optimization problem of the farthest distance between the two types of samples is: ; Where: is the relaxation factor, is the penalty factor, Represents a low-dimensional space sample Linearly separable samples mapped to high-dimensional space; the final classification function solved is: ; is the kernel function, represents the optimal Lagrange multiplier.

[0010] Preferably, in step 1, feature vectors of normal insulator infrared images and feature vectors of zero-value insulator infrared images are collected and divided into two parts, one part is input into the insulator recognition model as a training set for parameter optimization training of the model; and the other part is used as a test set to verify the recognition accuracy of the established insulator recognition model.

[0011] Preferably, the method of step 5 is: Support vector machine (SVM) is widely used in fault identification due to its excellent classification performance. Gray wolf optimization algorithm (GWO) is an optimization algorithm that is widely used in various optimization problems. Using the gray wolf optimization algorithm to optimize the parameters of the support vector machine will make the classification results more accurate. Initialize the various parameters of the gray wolf optimization algorithm, where the number of wolf groups is 10 and the maximum number of iterations is 200; Initialize the penalty parameter C and kernel parameter g of the support vector machine, the penalty parameter The range is [-10,10], the kernel function parameters The Grey Wolf Optimization Algorithm starts to optimize these two parameters in the range of [-10,10], aiming at the optimal solution of particle fitness.

[0012] According to the final parameter optimization results, C and g of the support vector machine are initialized, the test set is input into the insulator recognition model, and the test samples are predicted and analyzed.

[0013] A faulty insulator identification system adopts the faulty insulator identification method and system, including: an image acquisition module for acquiring infrared images of insulators; An image processing module, used for preprocessing the infrared image of the insulator; Feature extraction module, used to extract infrared features of insulators; Identification model creation module, used to establish an insulator identification model based on the support vector machine optimized by the Grey Wolf optimization algorithm; The identification module is used to bring the characteristic data of step 3 into the insulator identification model to identify low zero value insulators.

[0014] A computer device comprising: one or more processors; The processor is used to store one or more programs; When the one or more programs are executed by the one or more processors, a fault insulator identification method as described is implemented.

[0015] A computer-readable storage medium stores a computer program, and when the computer program is executed, the method for identifying a faulty insulator as described above is implemented.

[0016] The present invention can achieve the following beneficial effects: 1. Multi-level deep feature extraction is performed on insulator infrared image samples, which has stronger identification ability than the features extracted by existing deep models.

[0017] 2. Adaptive median filtering is used to denoise the infrared image, which well preserves the infrared image characteristics of the insulator. 3. Use grayscale transformation to enhance the contrast between the target and the background in the infrared image and improve the image quality.

[0018] 4. Histogram equalization is used to increase the contrast between the insulator string and the background, making it easier to extract the insulator string later.

[0019] 5. Adopt image adaptive threshold segmentation based on Ostu algorithm to separate the insulator string area from the background area to reduce the influence of the background.

[0020] 6. A method combining infrared images and support vector machines optimized by the Gray Wolf optimization algorithm is proposed to realize the detection and identification of zero-value insulators. It can accurately identify low zero-value insulators at the big data level and reduce manpower, material and financial resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The present invention will be further described below in conjunction with the accompanying drawings and embodiments: Figure 1 Provide a flow chart of the identification method of the present invention; Figure 2 This is the collected infrared image of the insulator; Figure 3 Grayscale enhancement and histogram equalization of insulator infrared images; Figure 4 This is the Ostu segmentation effect diagram; Figure 5 This is the process diagram of insulator string tilt correction; Figure 6 It is the minimum circumscribed rectangle process diagram of the insulator string after tilt correction and segmentation; Figure 7 It is the gradient feature extraction flow chart; Figure 8 is the gradient feature vector map; Fig. 9 Graph showing the optimization process for the parameters of the gray wolf optimization algorithm. DETAILED DESCRIPTION

[0022] The preferred solution is Figures 1 to 9 As shown, a method for identifying a faulty insulator comprises the following steps: Step 1, collecting infrared images of insulators; The feature vectors of normal insulator infrared images and zero-value insulator infrared images are collected and divided into two parts. One part is used as a training set to input into the insulator recognition model for parameter optimization training of the model; the other part is used as a test set to verify the recognition accuracy of the established insulator recognition model.

[0023] Step 2, preprocessing the infrared image of the insulator; The preprocessing method for the insulator infrared image is: grayscale enhancement processing is performed on the insulator infrared image to make the image clear and the features obvious. Since the edge of the insulator string is close to the background grayscale value, the insulator string cannot be accurately separated. Therefore, histogram equalization is used to make the original image become an image with uniform probability distribution, thereby enhancing the image contrast.

[0024] Step 3, extracting infrared features of insulators; The method for extracting infrared features of insulators is: using image adaptive threshold segmentation based on Ostu algorithm, an image gray level is L, and the image gray value is The average gray value of the neighborhood is , Represents the probability that the number of pixels is i, using Represents the joint probability density. When the image is segmented with (s, t) as the threshold, the corresponding target is to separate the insulator string area as a whole from the background area. and background The mean vector of is: ; ; The corresponding total mean vector is: ; The inter-class variance between the target and the background is defined as: ; Optimal segmentation threshold for The threshold corresponding to the maximum value; each time the inter-class variance is calculated, 3st+3(Ls )(Lt ) points need to be accumulated: ; The infrared image of the insulator string taken shows various angles due to the shooting angle. To solve the problem of the tilt angle of the target insulator, the Rondon transform is used to correct the tilt of the insulator. The realization principle of Radon transform is A straight line in plane space Mapped to a point in Radon space , then the Radon transform of the continuous image is: ; Where: is a Dirac function, and finally the directional gradient histogram features of the minimum circumscribed rectangle of the insulator string and the texture features of the grayscale gradient co-occurrence matrix are extracted.

[0025] Step 4, establishing an insulator identification model based on the support vector machine optimized by the grey wolf optimization algorithm; The insulator identification model of the support vector machine optimized by the gray wolf optimization algorithm is as follows: In the sample space, all data are completely mixed together, and it is impossible to find a classification function that can effectively classify the two types of data. In this case, a nonlinear support vector machine is used. The nonlinear support vector machine is based on the linear separable support vector machine and uses the kernel function. At this time, the convex quadratic optimization problem of the maximum distance between the two types of samples is: ; Where: is the relaxation factor, is the penalty factor, Represents a low-dimensional space sample Linearly separable samples mapped to high-dimensional space; the final classification function solved is: ; is the kernel function, represents the optimal Lagrange multiplier.

[0026] Step 5, bringing the characteristic data of step 3 into the insulator identification model to identify low zero value insulators.

[0027] Support vector machine (SVM) is widely used in fault identification due to its excellent classification performance. Gray wolf optimization algorithm (GWO) is an optimization algorithm that is widely used in various optimization problems. Using the gray wolf optimization algorithm to optimize the parameters of the support vector machine will make the classification results more accurate. Initialize the various parameters of the gray wolf optimization algorithm, where the number of wolf groups is 10 and the maximum number of iterations is 200; Initialize the penalty parameter C and kernel parameter g of the support vector machine, the penalty parameter The range is [-10,10], the kernel function parameters The Grey Wolf Optimization Algorithm starts to optimize these two parameters in the range of [-10,10], aiming at the optimal solution of particle fitness.

[0028] According to the final parameter optimization results, C and g of the support vector machine are initialized, the test set is input into the insulator recognition model, and the test samples are predicted and analyzed.

[0029] Using the same training and test sample data, the parameter optimization results of the support vector machine under different optimization algorithms are shown in Table 1. It can be seen from the table that the final optimization results of the penalty parameter C and the kernel parameter g by different optimization algorithms are very different. However, from the final insulator diagnosis results, it can be seen that the support vector machine optimized by the Grey Wolf optimization algorithm has the highest recognition and diagnosis accuracy.

[0030] Table 1 Comparison of SVM operation results

[0031] From the results of optimizing support vector machines using three algorithms, it can be seen that GridSearch is time-consuming, has low recognition accuracy, and has shortcomings such as high complexity and large amount of calculation when searching for the best solution. Particle swarm optimization (PSO) takes less time, but it also has obvious disadvantages. For functions with multiple local extreme points, it is easy to fall into local extreme points. This can be seen from the final optimization results of the two optimization algorithms. The particle swarm optimization algorithm falls into the local optimum. Although the final recognition accuracy is not low, there are more problems with misclassification results. The recognition accuracy of the gray wolf optimization algorithm can reach 95.246%, the optimization time is short, and the parameter search efficiency is high. The gray wolf optimization algorithm makes full use of prior knowledge and avoids the risk of falling into the local optimum. Therefore, compared with the particle swarm search algorithm and the grid search algorithm, the gray wolf optimization can efficiently identify and detect low zero value insulators.

[0032] A faulty insulator identification system adopts the faulty insulator identification method and system, including: an image acquisition module for acquiring infrared images of insulators; An image processing module, used for preprocessing the infrared image of the insulator; Feature extraction module, used to extract infrared features of insulators; Identification model creation module, used to establish an insulator identification model based on the support vector machine optimized by the Grey Wolf optimization algorithm; The identification module is used to bring the characteristic data of step 3 into the insulator identification model to identify low zero value insulators.

[0033] A computer device comprising: one or more processors; The processor is used to store one or more programs; When the one or more programs are executed by the one or more processors, a fault insulator identification method as described is implemented.

[0034] A computer-readable storage medium stores a computer program, and when the computer program is executed, the method for identifying a faulty insulator as described above is implemented.

[0035] The above embodiments are only preferred technical solutions of the present invention and should not be regarded as limiting the present invention. The protection scope of the present invention shall be the technical solutions recorded in the claims, including equivalent replacement solutions of the technical features in the technical solutions recorded in the claims. That is, equivalent replacement improvements within this scope are also within the protection scope of the present invention.

Claims

1. A method for identifying a faulty insulator, characterized in that The following steps are involved: Step 1, collecting infrared images of insulators; Step 2, preprocessing the infrared image of the insulator; Step 3, extracting infrared features of insulators; Step 4, establishing an insulator identification model based on the support vector machine optimized by the grey wolf optimization algorithm; Step 5, bringing the characteristic data of step 3 into the insulator identification model to identify low zero value insulators.

2. A method for identifying a faulty insulator according to claim 1, characterized in that: In step 2, the preprocessing method for the insulator infrared image is: grayscale enhancement processing is performed on the insulator infrared image to make the image clear and the features obvious. Since the edge of the insulator string is close to the background grayscale value, the insulator string cannot be accurately separated. Therefore, histogram equalization is used to make the original image into an image with uniform probability distribution, thereby enhancing the image contrast.

3. A method for identifying a faulty insulator according to claim 1, characterized in that: In step 3, the method for extracting the infrared features of the insulator is: using image adaptive threshold segmentation based on the Ostu algorithm, an image gray level is L, and the image gray value is The average gray value of the neighborhood is , Represents the probability that the number of pixels is i, using Represents the joint probability density. When the image is segmented with (s, t) as the threshold, the insulator string area is separated from the background area. and background The mean vector of is: ； ； The corresponding total mean vector is: ； The inter-class variance between the target and the background is defined as: ； Optimal segmentation threshold for The threshold corresponding to the maximum value; each time the inter-class variance is calculated, 3st+3(Ls )(Lt ) points need to be accumulated: ； The infrared image of the insulator string taken shows various angles due to the shooting angle. To solve the problem of the tilt angle of the target insulator, the Rondon transform is used to correct the tilt of the insulator. The realization principle of Radon transform is A straight line in plane space Mapped to a point in Radon space , then the Radon transform of the continuous image is: ； Where: is a Dirac function, and finally the directional gradient histogram features of the minimum circumscribed rectangle of the insulator string and the texture features of the grayscale gradient co-occurrence matrix are extracted.

4. A method for identifying a faulty insulator according to claim 1, characterized in that: In step 4, the insulator identification model of the support vector machine optimized by the gray wolf optimization algorithm is: in the sample space, all data are completely mixed together, and it is impossible to find a classification function that can effectively classify the two types of data. In this case, a nonlinear support vector machine is used. The nonlinear support vector machine refers to the kernel function on the basis of the linear separable support vector machine. At this time, the convex quadratic optimization problem of the farthest distance between the two types of samples is: ； Where: is the relaxation factor, is the penalty factor, Represents the sample vector For sample Dimensions; For the corresponding sample The type tag, is the normal vector, represents the classification threshold, Represents a low-dimensional space sample Linearly separable samples mapped to high-dimensional space; the final classification function solved is: ； is the kernel function, represents the optimal Lagrange multiplier.

5. A method for identifying a faulty insulator according to claim 1, characterized in that: In step 1, the feature vectors of normal insulator infrared images and zero-value insulator infrared images are collected and divided into two parts. One part is input into the insulator recognition model as a training set for parameter optimization training of the model; the other part is used as a test set to verify the recognition accuracy of the established insulator recognition model.

6. A method for identifying a faulty insulator according to claim 1, characterized in that: The method for step 5 is: Using the gray wolf optimization algorithm to optimize the parameters of the support vector machine will make the classification results more accurate; Initialize the various parameters of the gray wolf optimization algorithm, where the number of wolves is 10 and the maximum number of iterations is 200; Initialize the penalty parameter C and kernel parameter g of the support vector machine, the penalty parameter The range is [-10,10], the kernel function parameters The range of [-10,10], the Grey Wolf Optimization Algorithm begins to optimize these two parameters, aiming at the optimal solution of particle fitness; According to the final parameter optimization results, C and g of the support vector machine are initialized, the test set is input into the insulator recognition model, and the test samples are predicted and analyzed.

7. A fault insulator identification system, characterized in that: A faulty insulator identification method and system according to any one of claims 1 to 6 is adopted, comprising: an image acquisition module for acquiring infrared images of insulators; An image processing module is used for preprocessing the infrared image of the insulator; Feature extraction module, used to extract infrared features of insulators; Identification model creation module, used to establish an insulator identification model based on the support vector machine optimized by the Grey Wolf optimization algorithm; The identification module is used to bring the characteristic data of step 3 into the insulator identification model to identify low zero value insulators.

8. A computer device, characterized in that: include: one or more processors; The processor is used to store one or more programs; When the one or more programs are executed by the one or more processors, a fault insulator identification method as described in any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed, a fault insulator identification method according to any one of claims 1-6 is implemented.

Images