Insulator contamination degree identification method, system and device based on contaminated solution

Abstract

The present invention relates to the field of insulator detection technology, and specifically to an insulator contamination degree identification method, system and equipment based on a contaminated solution. The method comprises collecting a number of contaminated insulators, making contaminated solutions of upper surface insulators and contaminated solutions of lower surface insulators; photographing hyperspectral images of the contaminated solutions of upper surface insulators and contaminated solutions of lower surface insulators, performing lens correction processing on the hyperspectral images of the contaminated solutions, and calculating the hyperspectral transmittance of the contaminated solutions of upper and lower surfaces; extracting characteristic wavelengths on the average hyperspectral transmittance curve of the insulators, training to obtain a contamination degree identification model of the contaminated solution of the target insulator, and inputting the transmittance value of the contaminated solution into the trained contamination degree identification model of the contaminated solution to obtain a contamination degree level identification result. The present invention can shorten the detection time of the insulator contamination degree, improve the detection efficiency of the contamination degree, and solve the problem that moisture in the air interferes with the measurement of dust density.

Description

Technical Field

[0001] The invention relates to the technical field of insulator detection, and in particular to a method, system and equipment for identifying the contamination degree of an insulator based on a contaminated solution. Background Art

[0002] Insulators are widely used in power grid construction, playing an important role in electrical insulation and mechanical fixation. However, since their working environment is outdoors, dirt accumulation is inevitable. When these dirty substances are dissolved by moisture, a conductive layer will be formed on the surface of the insulator. Under the action of normal operating voltage, the insulator is very likely to discharge, which will further lead to serious accidents such as power outages. This is the "dirty flash" phenomenon that is a headache for power grid companies. Relevant studies have shown that the "dirty flash" phenomenon is closely related to the composition and concentration of dirty substances on the surface of the insulator. If the dirtiness information of the insulator can be accurately grasped, the cleaning work can be carried out more targeted, which has a very positive and practical significance for the prevention and control of the "dirty flash" phenomenon. At present, the relevant departments of the power grid company use the following method to detect the degree of insulator contamination: first, transfer the insulator contamination to a special wiping cloth, then soak the wiping cloth in distilled water, and then use a conductivity meter to measure the conductivity of the contaminated solution, from which the salt density of the contaminated solution can be converted. Then, the contaminated solution is filtered, and the dried contaminated filter paper is weighed, and the difference between the mass of the contaminated filter paper and the clean filter paper is used to obtain the mass of the insoluble contaminants, from which the ash density is converted. Finally, the salt density and ash density are comprehensively analyzed to obtain the degree of contamination information. The current process of measuring the degree of insulator contamination is too cumbersome, inefficient, slow, time-consuming and labor-intensive. Summary of the invention

[0003] In order to solve the technical problems existing in the prior art, the present invention provides an insulator contamination degree identification method, system and equipment based on a contamination solution, which can shorten the insulator contamination degree detection time, improve the contamination degree detection efficiency, and solve the problem of interference of moisture in the air on dust density measurement.

[0004] The first object of the present invention is to provide an insulator contamination degree identification method based on a contamination solution.

[0005] The second object of the present invention is to provide an insulator contamination degree identification system based on contamination solution.

[0006] A third object of the present invention is to provide a computer device.

[0007] The first object of the present invention can be achieved by adopting the following technical solutions:

[0008] An insulator contamination degree identification method based on a contaminated solution, the method comprising:

[0009] Collect a number of contaminated insulators, and prepare the contaminated substances on the upper and lower surfaces of the contaminated insulators into upper surface insulator contaminated solutions and lower surface insulator contaminated solutions respectively;

[0010] The hyperspectral images of the contaminated solution of the upper surface of the insulator and the contaminated solution of the lower surface of the insulator are respectively taken, the hyperspectral images of the contaminated solution are subjected to lens correction processing to obtain lens correction results, and the hyperspectral transmittances of the contaminated solutions of the upper surface and the lower surface are respectively calculated according to the lens correction results;

[0011] The characteristic wavelength on the average hyperspectral transmittance curve of the insulator is extracted through a competitive adaptive reweighted sampling algorithm. The transmittance corresponding to the characteristic wavelength on the average hyperspectral transmittance curve is taken as the feature, and the actual contamination level of the insulator is taken as the label. The support vector machine algorithm is used for training to obtain the contamination level recognition model of the insulator contamination solution.

[0012] The transmittance value of the contaminated solution is input into the trained insulator contamination solution contamination degree recognition model to obtain the contamination degree level recognition result.

[0013] The second object of the present invention can be achieved by adopting the following technical solutions:

[0014] An insulator contamination degree identification system based on a contamination solution, the system comprising:

[0015] A preparation module is used to collect a number of contaminated insulators, and to prepare the contaminated substances on the upper and lower surfaces of the contaminated insulators into upper surface insulator contaminated solutions and lower surface insulator contaminated solutions respectively;

[0016] A calculation module is used to respectively take hyperspectral images of the contaminated solution of the upper surface insulator and the contaminated solution of the lower surface insulator, perform lens correction processing on the hyperspectral images of the contaminated solution to obtain lens correction results, and calculate the hyperspectral transmittance of the contaminated solution of the upper and lower surfaces according to the lens correction results;

[0017] The recognition model module is used to extract the characteristic wavelength on the average hyperspectral transmittance curve of the insulator through a competitive adaptive reweighted sampling algorithm, take the transmittance corresponding to the characteristic wavelength on the average hyperspectral transmittance curve as a feature, take the actual contamination level of the insulator as a label, and apply the support vector machine algorithm to train to obtain the contamination level recognition model of the insulator contamination solution;

[0018] The identification output module is used to input the transmittance value of the contamination solution into the trained insulator contamination solution contamination degree identification model to obtain the contamination degree level identification result.

[0019] The third object of the present invention can be achieved by adopting the following technical solutions:

[0020] A computer device comprises a processor and a memory for storing a program executable by the processor. When the processor executes the program stored in the memory, the above-mentioned insulator contamination degree identification method based on contamination solution is implemented.

[0021] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0022] The present invention provides an insulator contamination degree identification method, system and device based on a contaminated solution. Compared with the traditional insulator contamination degree detection process, the steps of filtering, drying, weighing, calculating ash density and the like are omitted, which will greatly shorten the contamination degree detection time and improve the contamination degree detection efficiency. At the same time, because only the solution is analyzed and the influence of moisture in the air on the solution is extremely small, the problem of interference of moisture in the air on ash density measurement is also solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0024] Figure 1 is a flow chart of a method for identifying the contamination degree of an insulator contamination solution in an embodiment of the present invention;

[0025] Figure 2 Schematic diagram of a hyperspectral image shooting scene of an insulator contamination solution in an embodiment of the present invention;

[0026] Figure 3 This is a high-spectrum image effect diagram of the contaminated solution on the upper and lower surfaces of the insulator in the embodiment of the present invention

[0027] Figure 4 It is a comparison effect diagram of the dirty solution image in the embodiment of the present invention;

[0028] Figure 5 It is an effect diagram of the upper surface, lower surface and overall average high spectral transmittance curve in an embodiment of the present invention. DETAILED DESCRIPTION

[0029] The technical solution of the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is obvious that the described embodiments are part of the embodiments of the present invention, not all of the embodiments, and the implementation of the present invention is not limited to this. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0030] Embodiment 1:

[0031] The method and system for identifying the degree of insulator contamination based on the contaminated solution provided by the present invention will identify the degree of insulator contamination based on the hyperspectral transmittance of the insulator contamination solution. The hyperspectral image is based on the two-dimensional image, in which each pixel also contains response information on hundreds of continuous spectral bands, and is rich in information. Existing experiments have shown that insulators with different degrees of contamination have different amplitudes and peak values ​​of the hyperspectral transmittance curves of the contaminated solutions due to their different effects on light. Therefore, the corresponding degree of contamination information can be obtained non-destructively and contactlessly through the hyperspectral transmittance of the contaminated solution.

[0032] The insulators or insulating sheets commonly used at present have irregular shapes, which cause most of the contaminants to hide in their grooves. Therefore, the direct hyperspectral imaging of the insulators cannot fully reflect their actual contamination status. At the same time, for glass insulators, the uncontaminated areas on them may reflect light, which will bring uncertainty to the subsequent contamination degree identification. Although the insulating sheets have regular shapes and are simpler to analyze, the contamination degree identification method formed cannot be fully applied to insulators with irregular shapes and more complex contamination conditions, and has great limitations. Whether for insulators or insulating sheets, when you want to obtain experimental samples through artificial contamination, you need to spend a lot of time to let the contaminants attach in the expected way. Due to the uncertainty of the contamination process, the expected attachment effect is also difficult to guarantee. In contrast, identifying the insulator contamination degree based on the high spectral transmittance of the insulator contamination solution has the following advantages: the preparation of the contamination solution sample is convenient, the requirements for the required equipment are not high, and the operation steps are clear, with fewer uncertain factors; all the contamination substances in the insulator are transferred to the contamination solution, so the contamination solution contains all the contamination information of the insulator. Therefore, the contamination degree result obtained by analyzing the contamination solution can fully match the contamination degree of the insulator; the preparation of the contamination solution has no special restrictions on the shape and material of the insulator, so the contamination degree identification method based on the contamination solution is widely applicable.

[0033] like Figure 1 As shown, the method for identifying the degree of insulator contamination based on the contamination solution of the present invention comprises the following steps:

[0034] S1. Collect a number of contaminated insulators, and prepare the contaminated substances on the upper and lower surfaces of the contaminated insulators into upper surface insulator contaminated solutions and lower surface insulator contaminated solutions respectively;

[0035] S2, respectively taking hyperspectral images of the contaminated solution of the upper surface insulator and the contaminated solution of the lower surface insulator, performing lens correction processing on the hyperspectral images of the contaminated solution to obtain lens correction results, and calculating the hyperspectral transmittance of the contaminated solution of the upper and lower surfaces according to the lens correction results;

[0036] S3. The characteristic wavelength on the average hyperspectral transmittance curve of the insulator is extracted through a competitive adaptive reweighted sampling algorithm. The transmittance corresponding to the characteristic wavelength on the average hyperspectral transmittance curve is used as a feature, and the actual contamination level of the insulator is used as a label. The support vector machine algorithm is used for training to obtain the insulator contamination solution contamination recognition model.

[0037] S4. Input the transmittance value of the contaminated solution into the trained insulator contamination solution contamination level recognition model to obtain the contamination level recognition result.

[0038] Specifically:

[0039] S1. Collect a certain number of contaminated insulators, and make the contaminated substances on the upper and lower surfaces of the contaminated insulators into upper surface insulator contaminated solutions and lower surface insulator contaminated solutions respectively.

[0040] In this embodiment, the preparation process of the insulator contamination solution includes:

[0041] (1) Use a special wiping cloth to transfer all the dirt on the upper and lower surfaces of the insulator, and mark them for later use.

[0042] (2) Pour 200ml and 300ml of distilled water into two clean beakers respectively and set aside.

[0043] (3) Soak the special wiping cloth with dirt on the upper surface of the insulator in 200 ml of distilled water, and soak the special wiping cloth with dirt on the lower surface of the insulator in 300 ml of distilled water. If there are too many dirt on the wiping cloth, or multiple wiping cloths are used for one surface, you can consider increasing the amount of distilled water, and pay attention to special marking. The more distilled water, the more dilute the dirt solution will be, and the higher the hyperspectral transmittance will be. Therefore, in order to control the variables, the appropriate volume of distilled water is specified. If the actual volume of distilled water used is Vml, which is not equal to the specified value, it is necessary to convert the average transmittance curve obtained later. For the upper surface, the conversion method is curve value × (V / 200); for the lower surface, the conversion method is curve value × (V / 300). Use a clean glass rod to scrape all the dirt on the special wiping cloth into the distilled water as much as possible, and the preparation of the insulator dirt solution is completed.

[0044] S2. respectively capture hyperspectral images of the contaminated solution on the upper surface of the insulator and the contaminated solution on the lower surface of the insulator, perform lens correction processing on the hyperspectral images of the contaminated solution to obtain lens correction results, and respectively calculate the hyperspectral transmittance of the contaminated solution on the upper and lower surfaces.

[0045] The process of obtaining the high spectral transmittance of the polluted solution is as follows:

[0046] Set up the beaker stand, place the mobile light source for the hyperspectrometer 2 to 3 cm below the beaker stand, and place the beaker on the beaker stand stably, so that the light source is directly penetrating the solution in the cup. Place the hyperspectrometer in a position where the lens is facing the beaker mouth and the beaker mouth occupies about 1 / 2 to 2 / 3 of the lens field of view. The schematic diagram of the hyperspectral image shooting scene of the insulator contaminated solution is as follows: Figure 2 shown.

[0047] The hyperspectral image of the beakers containing 200 ml and 300 ml of distilled water was triggered to capture hyperspectral images. After lens correction, these images were used as standard correction images of the contaminated solutions on the upper and lower surfaces of the insulator. The standard correction image is a small area in the hyperspectral image of pure water after lens correction, and its value is the DN value. The hyperspectral images of the contaminated solutions on the upper and lower surfaces captured by the hyperspectral imager were obtained.

[0048] like Figure 3 As shown in the figure, the effect diagram of the hyperspectral image of the contaminated solution on the upper and lower surfaces of the insulator. Generally, the upper surface of the insulator is less polluted than the lower surface, so in most cases the color of the contaminated solution on the lower surface is darker than that on the upper surface. However, since the contaminants on the lower surface are dissolved in more distilled water and sometimes the contaminants on the lower surface of the insulator are less polluted, there are exceptions. At the same time, the container of the contaminated solution on the upper surface is a small beaker, and the container of the contaminated solution on the lower surface is a large beaker, which is determined by the volume of the solution. In fact, the difference in the hyperspectral image of the contaminated solution has nothing to do with the upper and lower surfaces of the insulator, but is determined by the amount of contaminants. This is also the basis for identifying the corresponding contamination level through the hyperspectral image of the contaminated solution. The image captured by the hyperspectral instrument after the lens cover is triggered is called the background image, which reflects the inherent interference information caused by factors such as the dark current inside the hyperspectral instrument.

[0049] Due to the inherent influence of the hyperspectrometer camera lens and internal optical components, the amount of light entering each point of the lens is inconsistent, with more light entering in the middle and less light entering on both sides. Lens correction can be understood as making the amount of light entering each pixel in the image consistent. The specific calculation formula for the lens correction result is:

[0050] Y=(X-X0)*α+X0

[0051] Among them, X is the original hyperspectral data of the contaminated solution, X0 is the background image data, α is the brightness stretch coefficient, which is obtained through the standard lens correction file provided by the hyperspectral instrument company, and Y is the lens correction result.

[0052] The hyperspectral transmittance of the contaminated solution on the upper and lower surfaces is calculated respectively. The hyperspectral transmittance is used to identify the contamination level. The specific calculation formula for the hyperspectral transmittance of the contaminated solution is:

[0053] Z=(Y-X0) / (Z0-X0)

[0054] Among them, Z0 is the upper surface standard correction image or the lower surface standard correction image, and Z is the transmittance calculation result of the dirty solution, which is equivalent to assuming that the transmittance of distilled water is 1, and comparing the dirty solution with it to obtain the corresponding transmittance.

[0055] like Figure 4 As shown, the original hyperspectral image of the contaminated solution, the hyperspectral image after lens correction, and the hyperspectral image after transmittance calculation. At the same time, in order to clearly present the changes in the image data, the average transmittance curve of the local area of ​​the contaminated solution is also displayed.

[0056] Extract as Figure 3 The white dotted lines show the average hyperspectral transmittance curves of the circular areas circled by the beaker stand in the hyperspectral images of the upper and lower surface contaminated solutions, and the two are averaged again to obtain the overall average hyperspectral transmittance curve. All the above curves of a certain insulator are shown in Figure 5 shown.

[0057] The overall average hyperspectral transmittance curve data is used as features, and the actual pollution level of the corresponding insulator is used as a label to constitute the sample of the subsequent pollution level recognition model.

[0058] S3. The characteristic wavelength on the average hyperspectral transmittance curve of the insulator is extracted through a competitive adaptive reweighted sampling algorithm. The transmittance corresponding to the characteristic wavelength on the average hyperspectral transmittance curve is used as a feature. The multiple different real contamination levels corresponding to the characteristic wavelength on the average hyperspectral transmittance curve are used as labels. The support vector machine algorithm is used for training to obtain a contamination degree recognition model for insulator contamination solution.

[0059] Specifically, a plurality of different real pollution degree levels include: a - very light, b - light, c - medium, d - heavy, and e - very heavy.

[0060] The characteristic wavelength on the average high-spectral transmittance curve of the insulator is extracted through the competitive adaptive reweighted sampling algorithm (CARS algorithm) to remove some irrelevant information and improve the stability and efficiency of the recognition model. The specific steps include:

[0061] 1) A set percentage (preferably 80%) of input samples is extracted through Monte Carlo sampling, and a partial least squares (PLS) regression model is established based on the input sample spectrum matrix. Assuming that the sample spectrum matrix is ​​X(n×p), where n is the number of samples and p is the number of variables, and the sample contamination matrix is ​​Y(n×1), then its partial least squares regression model is

[0062] Y=Xb+e

[0063] The regression coefficient b is a p-dimensional coefficient vector, and e is the n-dimensional residual. The absolute value of the regression coefficient of the i-th element in b | b i |(1≤i≤p) represents the contribution of the i-th variable to the entire partial least squares regression model. Specifically, the larger the value, the more important the corresponding wavelength variable is in predicting the pollution level of the model.

[0064] 2) The number of wavelengths is reduced by exponentially decreasing function (EDF), and the preservation rate of wavelength variables is obtained after the jth sampling. The preservation rate of wavelength variables is:

[0065] r j =ae -kj

[0066] Where a and k are constants, e is the residual error in n dimensions, and assuming that N samplings are performed in total, the CARS algorithm stipulates that in the first sampling, all wavelength variables are involved in the modeling, r1 = 1; in the Nth sampling, only two variables are left to participate in the modeling, then r N =2 / p, from which we can deduce

[0067]

[0068]

[0069] Preservation rate × number of original wavelengths is the number of wavelengths preserved and retained. This step is to force a reduction in the number of wavelengths. The selection of characteristic wavelengths is to reduce unimportant wavelength variables. This is the first reduction.

[0070] 3) Based on the Adaptive Reweighted Sampling (ARS) method, competitive wavelength variables are selected. The selection is based on the size of the weight value, which is:

[0071]

[0072] Where i = 1, 2, 3, ..., p; | b i | is the absolute value of the regression coefficient of the i-th element in b, and the weight value is w i The larger the value, the more competitive it is, and the more likely it is to be selected. The wavelength variables that best reflect the characteristics of the curve of the category to which they belong are extracted, and the wavelength variables and weight values ​​with lower weight values ​​are removed.

[0073] 4) N ​​wavelength variable subsets are obtained by N samplings, and the root mean square error of the interactive verification of the wavelength variable subsets is calculated to obtain the optimal characteristic wavelength variable combination. The subset with the smallest root mean square error of the interactive verification is the optimal characteristic wavelength variable combination. The root mean square error of the interactive verification is:

[0074]

[0075] Where y kcv is the predicted label value of the kth sample, y k is the input true label value of the kth sample. After that, for the overall average hyperspectral reflectance curve of the new contaminated insulator, it is only necessary to retain the reflectance value corresponding to the optimal characteristic wavelength as the input of the subsequent contamination level recognition model.

[0076] The transmittance corresponding to the characteristic wavelength of the contaminated solution is used as a feature, and the corresponding real contamination level (the contamination level consists of 5 levels: a-very light, b-light, c-medium, d-heavy, and e-very heavy) is used as a label. 70% of the samples are used as a training set, and 30% of the samples are used as a test set. The support vector machine algorithm (SVM algorithm) is used to train the insulator contamination solution contamination level recognition model. The SVM algorithm is to find a hyperplane in a high-dimensional space to linearly separate different categories of input samples with the maximum spacing. The SVM algorithm has guaranteed prediction accuracy and has the advantage of being lightweight.

[0077] The specific steps of using the support vector machine algorithm to train the insulator contamination solution contamination degree identification model include:

[0078] (1) Format conversion: convert the training set and test set sample data into the input data format required by the toolkit LIBSVM. <label>1:<value 1> 2:<value 2> …",in <label>It is the input sample label required for training the recognition model, and it is also the classification result output when applying the recognition model. Here it is the actual pollution level of the insulator. a~e are represented by {1,0,0,0,0}, {0,1,0,0,0}, {0,0,1,0,0}, {0,0,0,1,0}, {0,0,0,0,1} respectively; 1: <value1>The number before the colon, such as 1, represents the feature number, and the number after the colon is the specific reflectivity data.

[0079] (2) Feature scaling: perform feature scaling on the training set data and the test set data with the same ratio to improve prediction accuracy. The target scaling range is generally [-1, +1] or [0, 1]. The same ratio means that if the range of the training set data is scaled from [-10, +10] to [-1, +1], and the original range of the test set data is [-12, +7], then the range must be scaled to [-1.2, +0.7]. The significance of feature scaling itself is to make the model pay equal attention to all features and improve the model convergence speed. Since the reflectivity data of the training set and the test set in this paper are already between 0 and 1, this step is omitted.

[0080] (3) Kernel function selection: set a training set consisting of samples x i and the corresponding label y i (y∈{1,-1} l ) constitutes a data pair (x i ,y i ), select the RBF kernel function, the RBF kernel function is:

[0081] K(x i ,x j ) = exp(-γ||x i -x j || 2 ),

[0082] Where the parameter γ>0.

[0083] LIBSVM needs to solve the optimization problem

[0084]

[0085] The condition is

[0086] y i (w T φ(x i )+b)≥1-ξ i

[0087] ξ i ≥0

[0088] Where C (C>0) is the penalty coefficient, that is, the tolerance for error. The larger C is, the better the trained model can recognize existing data, but the performance on new data may not be satisfactory, that is, it is easy to overfit; and the smaller C is, the trained model may not have reliable recognition results on existing data, that is, it is easy to underfit. Training sample vector x i Through the function φ, it is mapped to a higher-dimensional space, and K(x i ,x j )≡φ(x i ) T φ(x j ) is called the kernel function. The use of kernel functions simplifies complex calculations in high-dimensional space. The RBF kernel function is selected when the number of features is not much larger than the number of samples.

[0089] (4) Parameter optimization: The cross-validation parameter selection tool grid.py is used to find the best penalty coefficient C and parameter γ on the training set. The cross-validation process of v folds is to first divide the training set into v subsets, and then use each subset as a test set. The subset will be predicted by the classifier trained on the corresponding remaining v-1 subsets. Thus, each sample in the entire training set will be predicted once, and the cross-validation accuracy is the ratio of correctly classified samples. The grid.py tool uses a grid search method to find the best C and parameter γ through cross-validation, that is, it traverses and applies multiple pairs of pre-set C and γ parameters for cross-validation, and obtains the C and γ parameter pair with the highest cross-validation accuracy as the optimal parameter selection. The application of grid.py is implemented by the command line "…\python…\grid.py filename", where the first "…" represents the file path of python.exe to be applied; the second "…" is the path of the grid.py tool to be called; filename is the training set file name converted to the input data format required by LIBSVM.

[0090] (5) Model training: on the training set, the cross-validation parameter selection tool grid.py is used to select the optimal penalty coefficient C and parameter γ to train the pollution recognition model. The first step of the training process is to read the training set label y and feature data x through the svm_read_problem(path) function in the LIBSVM toolkit. The function parameter "path" is the path of the training set data file; the second step is to package the read label y and feature data x through problem = svm_problem(y,x) to form a description of the recognition problem problem to be solved; the third step is to set the values ​​of C and γ parameters through parameter = svm_parameter('-g*-c*'). The g in the svm_parameter function parameter corresponds to γ, c corresponds to C; the fourth step is to train the target pollution degree recognition model model according to the identification problem problem and the setting parameter parameter through model = svm_train (problem, parameter); the last step is to save the trained pollution degree recognition model through the svm_save_model ('model', model) function. The first parameter of this function is the name of the model, and the second parameter is the model trained by svm_train. After that, the svm_load_model ('model') function can be called to load and apply the pollution degree recognition model. After that, facing the new polluted insulator that needs to be detected, the transmittance value corresponding to the characteristic band of its overall average hyperspectral reflectance curve can be input into the trained pollution degree recognition model to obtain the pollution degree level recognition result.

[0091] S4. Input the transmittance value of the contaminated solution into the trained insulator contamination solution contamination level recognition model to obtain the contamination level recognition result.

[0092] After that, when facing new contaminated insulators that need to be tested, repeat the above operations to obtain the transmittance value corresponding to the characteristic wavelength of the overall average hyperspectral transmittance curve, which can be input into the trained insulator contamination solution contamination level identification model to obtain the contamination level identification result, without having to go through tedious and time-consuming operations such as filtering, drying, weighing, and calculation.

[0093] Embodiment 2:

[0094] This embodiment provides an insulator contamination degree identification system based on contamination solution, which includes a preparation module, a shooting and calculation module, a recognition model module and a recognition output module. The specific functions of each module are as follows:

[0095] A preparation module is used to collect a number of contaminated insulators, and to prepare the contaminated substances on the upper and lower surfaces of the contaminated insulators into upper surface insulator contaminated solutions and lower surface insulator contaminated solutions respectively;

[0096] A calculation module is used to respectively take hyperspectral images of the contaminated solution of the upper surface insulator and the contaminated solution of the lower surface insulator, perform lens correction processing on the hyperspectral images of the contaminated solution to obtain lens correction results, and calculate the hyperspectral transmittance of the contaminated solution of the upper and lower surfaces according to the lens correction results;

[0097] The identification model module is used to extract the characteristic wavelength on the average hyperspectral transmittance curve of the insulator through a competitive adaptive reweighted sampling algorithm, take the transmittance corresponding to the characteristic wavelength on the average hyperspectral transmittance curve as a feature, take the actual contamination level of the insulator as a label, and apply the support vector machine algorithm to train to obtain the insulator contamination solution contamination level identification model.

[0098] The identification output module is used to input the transmittance value of the contamination solution into the trained insulator contamination solution contamination degree identification model to obtain the contamination degree level identification result.

[0099] Embodiment 3:

[0100] This embodiment provides a computer device, which may be a server, a computer, etc., and includes a processor, a memory, an input device, a display, and a network interface connected through a system bus. The processor is used to provide computing and control capabilities. The memory includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. When the processor executes the computer program stored in the memory, the insulator contamination degree identification method based on the contaminated solution of the above-mentioned embodiment 1 is implemented as follows:

[0101] Collect a number of contaminated insulators, and prepare the contaminated substances on the upper and lower surfaces of the contaminated insulators into upper surface insulator contaminated solutions and lower surface insulator contaminated solutions respectively;

[0102] The hyperspectral images of the contaminated solution of the upper surface of the insulator and the contaminated solution of the lower surface of the insulator are respectively taken, the hyperspectral images of the contaminated solution are subjected to lens correction processing to obtain lens correction results, and the hyperspectral transmittances of the contaminated solutions of the upper surface and the lower surface are respectively calculated according to the lens correction results;

[0103] The characteristic wavelength on the average hyperspectral transmittance curve of the insulator is extracted by a competitive adaptive reweighted sampling algorithm. The transmittance corresponding to the characteristic wavelength on the average hyperspectral transmittance curve is used as a feature, and the actual contamination level of the insulator is used as a label. The contamination level recognition model of the insulator contamination solution is obtained by training with the support vector machine algorithm.

[0104] The transmittance value of the contaminated solution is input into the trained insulator contamination solution contamination degree recognition model to obtain the contamination degree level recognition result.

[0105] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention. < / label> < / label>

Claims

1. The method for identifying the degree of insulator contamination based on contamination solution is characterized in that: The following steps are involved: Collect a number of contaminated insulators, and prepare the contaminated substances on the upper and lower surfaces of the contaminated insulators into upper surface insulator contaminated solutions and lower surface insulator contaminated solutions respectively; The hyperspectral images of the contaminated solution of the upper surface insulator and the contaminated solution of the lower surface insulator are respectively taken, and the hyperspectral images of the contaminated solution are subjected to lens correction processing to obtain lens correction results. The hyperspectral transmittances of the contaminated solutions on the upper and lower surfaces are respectively calculated according to the lens correction results. The lens correction result calculation formula is: Y = (X - X0) * α + X0 Where X is the original hyperspectral data of the polluted solution, X0 is the background image data, α is the brightness stretch coefficient, and Y is the lens correction result; The high spectral transmittance calculation formula is: Z = (Y - X0) / (Z0 - X0) Wherein, Z0 is the upper surface standard calibration image or the lower surface standard calibration image, X0 is the background image data, and Z is the transmittance calculation result of the dirty solution; The characteristic wavelength on the average hyperspectral transmittance curve of the insulator is extracted by a competitive adaptive reweighted sampling algorithm. The transmittance corresponding to the characteristic wavelength on the average hyperspectral transmittance curve is used as a feature, and the actual contamination level of the insulator is used as a label. The contamination level recognition model of the insulator contamination solution is obtained by training with the support vector machine algorithm. The transmittance value of the contaminated solution is input into the trained insulator contamination solution contamination degree recognition model to obtain the contamination degree level recognition result.

2. The method for identifying the degree of insulator contamination based on contamination solution according to claim 1 is characterized in that: The method of extracting characteristic wavelengths on the average high spectral transmittance curve of the insulator by a competitive adaptive reweighted sampling algorithm comprises the following steps: A set percentage of input samples are extracted through Monte Carlo sampling, and a partial least squares regression model is established based on the input sample spectral matrix; The number of wavelengths is forced to be reduced through the decay exponential function, and the preservation rate of the wavelength variable is obtained after the jth sampling; Competitive wavelength variables are screened out based on the adaptive reweighted sampling method; N wavelength variable subsets are obtained by N samplings, and the root mean square error of the interactive verification of the wavelength variable subsets is calculated to obtain the optimal characteristic wavelength variable combination.

3. The method for identifying the degree of insulator contamination based on contamination solution according to claim 2 is characterized in that: The partial least squares regression model is: Y = Xb + e Among them, the regression coefficient b is a p-dimensional coefficient vector, e is the n-dimensional residual, Y is the sample pollution matrix, and X is the sample spectrum matrix.

4. The method for identifying the degree of insulator contamination based on contamination solution according to claim 3 is characterized in that: The method of applying support vector machine algorithm to train and obtain the insulator contamination solution contamination degree identification model comprises the following steps: Convert the training set sample data and the test set sample data into the input data format required by the toolkit LIBSVM; Perform feature scaling on the training set data and the test set data with the same proportion; Set a training set, which contains samples X i With the corresponding label Composition data pair , select the RBF kernel function, the RBF kernel function is: ， The parameter γ > 0; Use the cross-validation parameter selection tool to find the best penalty coefficient C and parameter γ on the training set; The pollution degree recognition model is trained on the training set by using the optimal penalty coefficient C and parameter γ.

5. The insulator contamination degree identification system based on contamination solution is characterized by: The system comprises: A preparation module is used to collect a number of contaminated insulators, and to prepare the contaminated substances on the upper and lower surfaces of the contaminated insulators into upper surface insulator contaminated solutions and lower surface insulator contaminated solutions respectively; The calculation module is used to respectively take the hyperspectral images of the contaminated solution of the upper surface insulator and the contaminated solution of the lower surface insulator, perform lens correction processing on the hyperspectral images of the contaminated solution to obtain lens correction results, and calculate the hyperspectral transmittance of the contaminated solution on the upper and lower surfaces according to the lens correction results; the lens correction result calculation formula is: Y = (X - X0) * α + X0 Where X is the original hyperspectral data of the polluted solution, X0 is the background image data, α is the brightness stretch coefficient, and Y is the lens correction result; The high spectral transmittance calculation formula is: Z = (Y - X0) / (Z0 - X0) Wherein, Z0 is the upper surface standard calibration image or the lower surface standard calibration image, X0 is the background image data, and Z is the transmittance calculation result of the dirty solution; The recognition model module is used to extract the characteristic wavelength on the average hyperspectral transmittance curve of the insulator through a competitive adaptive reweighted sampling algorithm, take the transmittance corresponding to the characteristic wavelength on the average hyperspectral transmittance curve as a feature, take the actual contamination level of the insulator as a label, and apply the support vector machine algorithm to train to obtain the contamination level recognition model of the insulator contamination solution; The identification output module is used to input the transmittance value of the contamination solution into the trained insulator contamination solution contamination degree identification model to obtain the contamination degree level identification result.

6. A computer device comprising a processor and a memory for storing a program executable by the processor, characterized in that: When the processor executes the program stored in the memory, the insulator contamination degree identification method based on the contamination solution according to any one of claims 1 to 4 is implemented.

Images