Icing detection method for power transmission line and related equipment

Abstract

The invention discloses an icing detection method for a power transmission line and related equipment, relates to the technical field of electric power safety risk control, and solves the problems of low accuracy and poor timeliness of line icing detection. The method comprises the following steps: acquiring mechanical data, electrical data, environmental data and visual data of the power transmission line; constructing an input vector based on the mechanical data, the electrical data, the environmental data and the visual data; inputting the input vector into a preset multi-core KFDA model to obtain a low-dimensional output vector output by the multi-core KFDA model; and outputting the icing risk level of the power transmission line through a preset SVR model based on the low-dimensional output vector. The high-dimensional multi-source data of the power transmission line is projected to the low-dimensional feature space through the multi-core KFDA model, the nonlinear relation of the input data is captured, the most discriminative feature of the input vector is extracted and reserved, and meanwhile, the data dimension is reduced, so that the detection accuracy and efficiency of icing of the power transmission line are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power safety risk control, and particularly relates to an ice coating detection method and related equipment for a transmission line. Background Art

[0002] The operation safety of overhead transmission lines is affected by seasonal adverse weather conditions. Especially in cold winter conditions, icing phenomena such as glaze ice and rime ice are likely to occur. These ice coatings will not only increase the load and risk of the line, but also affect the stable operation of the power system. Therefore, accurately identifying and segmenting ice coating types is crucial for power grid operation and maintenance.

[0003] Traditional ice coating detection methods rely on single meteorological or image data, ignoring other non-linear coupling factors, resulting in difficulty in detecting line icing phenomena in a timely and accurate manner.

[0004] In view of this, an ice coating detection method and related equipment for a transmission line are needed. Summary of the Invention

[0005] Aiming at the problems of low accuracy and poor timeliness in line icing detection in the prior art, the present invention provides an ice coating detection method and related equipment for a transmission line, which can detect line icing phenomena in a timely and accurate manner. The specific technical solutions are as follows:

[0006] In a first aspect, an embodiment of the present application provides an ice coating detection method for a transmission line, including:

[0007] Obtaining mechanical data, electrical data, environmental data, and visual data of the transmission line; constructing an input vector based on the mechanical data, the electrical data, the environmental data, and the visual data; inputting the input vector into a preset multi-core kernel Fisher discriminant analysis (KFDA) model to obtain a low-dimensional output vector output by the multi-core KFDA model; and outputting an ice coating risk level of the transmission line through a preset support vector regression (SVR) model based on the low-dimensional output vector.

[0008] Preferably, the mechanical data includes conductor inclination and vibration frequency spectrum, the electrical data includes leakage current and harmonic distortion rate, the environmental data includes temperature and humidity gradient, and the visual data includes ice coating thickness calculated from images of the transmission line.

[0009] Preferably, the multi-core KFDA model includes a combined kernel function, and the combined kernel function includes a Gaussian kernel function and a polynomial kernel function; the expression of the Gaussian kernel function is:

[0010] KGaussian (x i , x j ) = exp(-γ||x i - x j ||);

[0011] Among them, K Gaussian represents the value of the Gaussian kernel function, which is used to indicate the similarity between the first sample x i and the second sample x j ; The exp function is the exponential function, and γ represents the scale parameter of the kernel function;

[0012] The expression of this polynomial kernel function is:

[0013]

[0014] Among them, K Poly represents the value of the polynomial kernel function, the inner product of the first sample and the second sample, and d represents the order of the polynomial;

[0015] The expression of this combined kernel function is:

[0016] K = αK Gaussian (x i , x j ) + βK Poly (x i , x j );

[0017] Among them, K represents the value of the combined kernel function, α represents the weight coefficient of the Gaussian kernel function, β represents the weight coefficient of the polynomial kernel function, and α + β = 1.

[0018] Preferably, inputting the input vector into a preset multi-kernel kernel Fisher discriminant analysis KFDA model to obtain a low-dimensional output vector output by the multi-kernel KFDA model, including: processing elements corresponding to the mechanical data, the environmental data, and the visual data in the input vector through the Gaussian kernel function; processing elements corresponding to the electrical data in the input vector through the polynomial kernel function; obtaining the low-dimensional output vector calculated by the multi-kernel KFDA model based on the processing results of the Gaussian kernel function and the polynomial kernel function.

[0019] Preferably, the SVR model uses the Gaussian kernel function, and the penalty constant of the SVR is 1.

[0020] Preferably, after determining the icing risk level of the transmission line through a preset support vector regression SVR model based on the low-dimensional output vector, the method further includes: outputting an alarm message corresponding to the icing risk level based on the icing risk level.

[0021] Preferably, after outputting the icing risk level of the transmission line through a preset Support Vector Regression (SVR) model based on the low-dimensional output vector, the method further includes: periodically updating the parameters of the multi-kernel KFDA model; the update formula is as follows:

[0022] θ new = θ old + η▽ θ L(θ; D new )

[0023] where θ old and θ new respectively represent the model parameters before and after the update, η is the learning rate, ▽ θ represents the gradient of the loss function with respect to the model parameters, and L is the loss function calculated based on the new dataset D new .

[0024] Second, an icing detection system for a transmission line provided by an embodiment of the present application is applied to the method as described in the first aspect, and includes:

[0025] An acquisition module, configured to acquire the mechanical data, electrical data, environmental data, and visual data of the transmission line;

[0026] A construction module, configured to construct an input vector based on the mechanical data, the electrical data, the environmental data, and the visual data;

[0027] A dimensionality reduction module, configured to input the input vector into a preset multi-kernel KFDA model to obtain a low-dimensional output vector output by the multi-kernel KFDA model;

[0028] An output module, configured to output the icing risk level of the transmission line through a preset SVR model based on the low-dimensional output vector.

[0029] Third, a computing device provided by an embodiment of the present application includes: a memory for storing a program; a processor for loading the program to execute the method as described in the first aspect.

[0030] Fourth, a computer-readable storage medium provided by an embodiment of the present application includes a stored program, wherein when the program runs, it controls the device where the computer-readable storage medium is located to execute the method as described in the first aspect.

[0031] Compared with the prior art, the beneficial effects of the present invention are as follows: by constructing an input vector based on the mechanical data, electrical data, environmental data, and visual data of the transmission line, the characteristics of the transmission line can be more comprehensively characterized; then, through the multi-kernel KFDA model, the input vector constructed from the high-dimensional multi-source data of the transmission line is projected into a low-dimensional feature space to capture the non-linear relationship of the input vector data, extract and retain the most discriminative features of the input vector, and at the same time reduce the data dimension, thereby improving the detection accuracy and efficiency of the SVR model for the icing condition of the transmission line. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for use in the description of the specific embodiments or the prior art. In all the drawings, similar elements or parts are generally denoted by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0033] Figure 1 It is a schematic flowchart of an icing detection method for a transmission line provided by an embodiment of the present application;

[0034] Figure 2 It is a schematic structural diagram of an icing detection system for a transmission line provided by an embodiment of the present application;

[0035] Figure 3 It is a schematic structural diagram of a computing device provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0036] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0037] It should be understood that when used in this specification and the appended claims, the terms "comprises" and "comprising" indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations.

[0038] It should also be understood that the terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an", and "the" are intended to include the plural forms.

[0039] It should also be further understood that the term "and / or" used in the specification and appended claims of the present invention refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0040] To solve the problems of low accuracy and poor timeliness in detecting line icing in the prior art, the present invention provides an icing detection method and related equipment for a transmission line, which can detect the line icing phenomenon in a timely and accurate manner.

[0041] Please refer to Figure 1 , Figure 1 which is a schematic flowchart of an icing detection method for a transmission line provided by an embodiment of the present application. This method is applied to a computing device and specifically includes the following steps:

[0042] Step 101, the computing device acquires mechanical data, electrical data, environmental data, and visual data of the transmission line.

[0043] Among them, the computing device can be communicatively connected to sensors and image acquisition devices arranged on electrical equipment of the transmission line to acquire various data of the transmission line in real time.

[0044] Exemplarily, the computing device can acquire the tensile force data of the transmission line through a tensile force sensor

[0045] Exemplarily, the computing device can acquire an image of the transmission line through an image acquisition device, calculate the conductor inclination angle of the transmission line based on the image of the transmission line, and measure the icing thickness in the image based on an image segmentation algorithm.

[0046] In an overhead transmission line, the conductor inclination angle refers to the angle between the lowest point of the conductor and the line connecting the two suspension points when the conductor is suspended between the poles. If the conductor inclination angle is too large, it may cause the conductor tension to be too large, exceeding the bearing capacity of the conductor, thereby triggering safety accidents such as conductor breakage; at the same time, it will also cause the pole to bear greater lateral and longitudinal loads, posing higher requirements for the stability of the pole. And if the inclination angle is too small, the conductor sag may be too large, and in harsh weather conditions (such as strong winds, icing, etc.), it is easy to be too close to the ground or other objects, resulting in safety hazards such as electric discharge.

[0047] Exemplarily, the computing device can acquire the leakage current and harmonic distortion rate of the transmission line through an electrical sensor.

[0048] Leakage current refers to the unexpected current generated on the surface of the insulation layer or insulator due to reasons such as aging of insulation materials, surface contamination, or moisture during the normal operation of the transmission line. Line icing may cause surface contamination of the insulator, formation of ice bridges, or conduction of the surface water film during ice melting, leading to partial discharge or flashover. These phenomena will generate high-frequency harmonics, increasing the harmonic distortion rate. By monitoring the abnormal changes in the harmonic distortion rate, the icing state of the insulator or the degree of insulation deterioration can be indirectly judged.

[0049] Exemplarily, the computing device can obtain the ambient temperature and humidity of the transmission line, as well as the temperature and humidity gradient, through a temperature and humidity sensor.

[0050] Preferably, the mechanical data includes the conductor inclination angle and the vibration spectrum, the electrical data includes the leakage current and the harmonic distortion rate, the environmental data includes the temperature and humidity gradient, and the visual data includes the ice thickness calculated from the image of the transmission line.

[0051] Among them, the sampling rate of the mechanical data collected by the mechanical sensor by the computing device can be 10 Hz; the sampling rate of the electrical data detected by the electrical sensor can be 1 kHz; the sampling rate of the environmental data collected by the temperature and humidity sensor can be 1 time per minute; the sampling rate of obtaining the image of the transmission line by the image acquisition device can be 30 frames per second (fps).

[0052] After obtaining the above data, the computing device can perform preliminary processing on these data, including data cleaning, outlier detection, and compensation. Among them, data cleaning is mainly to remove obviously incorrect data points, such as values outside the physically possible range; outlier detection can use the standard deviation method, that is, data points exceeding the range of plus or minus 3 times the standard deviation of the mean are marked as outliers and removed; data compensation is to perform interpolation processing on missing or abnormal data points, such as linear interpolation and spline interpolation.

[0053] Exemplarily, the computing device can reject outliers based on the 3σ criterion, and the calculation formula is as follows:

[0054] Assume the data set X t ={x1, x2,..., x n}, first, it is necessary to calculate the mean μ and the standard deviation σ of this data set, where t is time, and x1, x2,..., x n are the data at different times respectively.

[0055] According to the 3σ principle, the probability that the data falls within the interval (μ - 3σ, μ + 3σ) is about 99.7%. Therefore, data points not within this interval are regarded as outliers, and the computing device can remove the outliers in this data set.

[0056] Exemplarily, the calculation formula for the computing device to standardize data can be as follows:

[0057] x' = (x - μ) / σ

[0058] where x' is the data after standardization processing, and μ and σ are the mean and standard deviation corresponding to the data x respectively.

[0059] After completing the data preprocessing, the computing device can execute step 102 and store the processed data in a dedicated database for subsequent analysis. The main purpose of this step is to provide reliable basic data for subsequent analysis and fault diagnosis, and ensure the real-time, integrity and accuracy of the data.

[0060] Step 102: The computing device constructs an input vector based on the mechanical data, the electrical data, the environmental data and the visual data.

[0061] Among them, the computing device can extract the data with high correlation with the phenomenon of line icing and capable of characterizing the icing degree from the mechanical data, the electrical data, the environmental data and the visual data, and construct an input vector.

[0062] Specifically, the computing device can collect a large amount of multi-source data in the icing state and non-icing state of the transmission line, and based on the correlation analysis method, find the N variables with the highest correlation with the icing thickness, and use these N variables as the elements of the input vector to construct the input vector.

[0063] Exemplarily, the computing device can screen the variables with a strong correlation with the icing thickness through methods such as mutual information method, distance correlation method, canonical correlation analysis (CCA) method or other correlation analysis methods.

[0064] Exemplarily, the input vector F = [rate of change of conductor inclination angle, harmonic distortion rate of leakage current, temperature and humidity gradient, icing thickness].

[0065] Step 103: The computing device inputs the input vector into a preset multi-core kernel Fisher discriminant analysis (KFDA) model to obtain a low-dimensional output vector output by the multi-core KFDA model.

[0066] Among them, the KFDA model is preset in the computing device and is a KFDA model that has been trained.

[0067] Preferably, the multi-core KFDA model includes a combined kernel function, which includes a Gaussian kernel function and a polynomial kernel function; the expression of the Gaussian kernel function is:

[0068] K Gaussian (x i ,x j )=exp(-γ||x i -x j ||);

[0069] Among them, K Gaussian represents the value of the Gaussian kernel function, which is used to indicate the similarity between the first sample x i and the second sample x j ; the exp function is an exponential function, and γ represents the scale parameter of the kernel function.

[0070] ||x i -x j || represents the Euclidean distance between the first sample x i and the second sample x j , which measures the geometric distance between the two samples in space.

[0071] The scale parameter γ is a parameter optimized by cross-validation. It determines the width of the Gaussian kernel function. The larger the value of γ, the smoother the curve of the Gaussian kernel function, and the wider the influence range of each data point, which means that the similarity measure between data points in the high-dimensional space is relatively looser; the smaller the value of γ, the sharper the curve, and the smaller the influence range of the data points, and the similarity measure is relatively stricter.

[0072] The expression of the polynomial kernel function is:

[0073]

[0074] Among them, K Poly represents the value of the polynomial kernel function, the inner product of the first sample and the second sample, and d represents the order of the polynomial.

[0075] C is a constant term, and its function is to adjust the characteristics of the kernel function. The value of c will affect the properties of the feature space and the complexity of the model, etc. When C is 0, the kernel function only contains the information of the vector inner product; when C is greater than 0, it is equivalent to adding some bias terms in the feature space, enabling the model to better fit the data. Especially when there is a non-linear relationship in the data and it is not completely determined by the inner product, c plays an important regulatory role.

[0076] d represents the degree / order of the polynomial. It determines the dimension of the feature space after mapping and the complexity of the model. The larger d is, the higher the dimension of the feature space after mapping, the more complex the function that the model can fit, and the stronger the fitting ability for data. However, it may also increase the risk of overfitting. For example, when d = 2, the kernel function is a quadratic polynomial kernel function, which maps the original feature space to a higher-dimensional quadratic feature space.

[0077] Preferably, the computing device can process the elements corresponding to the mechanical data, the environmental data, and the visual data in the input vector through the Gaussian kernel function; process the elements corresponding to the electrical data in the input vector through the polynomial kernel function; and obtain the low-dimensional output vector calculated by the multi-kernel KFDA model based on the processing results of the Gaussian kernel function and the polynomial kernel function.

[0078] Among them, different data sources often have different data characteristics and distribution laws. For example, image data has features such as spatial structure and pixel values, and text data has features such as word frequency and semantics. Different kernel functions have different abilities to capture data characteristics. In order to make full use of the information of different data sources, the most suitable kernel function will be selected for each data source for calculation, and then these calculation results will be integrated through linear combination.

[0079] Specifically, mechanical data (such as wire tension, vibration frequency, tower stress) usually presents non-linear characteristics. Especially when ice accretion causes changes in wire loads, the relationship between mechanical parameters may present complex local non-linear patterns; temperature and humidity data has the characteristic of continuous change, and its correlation with ice accretion formation may be non-linear; the ice accretion thickness in visual data is itself a continuous physical quantity, and its growth process is affected by the comprehensive influence of environmental conditions (wind speed, humidity) and line conditions (temperature), presenting non-linear dynamic characteristics. The Gaussian kernel function controls the locally sensitive range through the hyperparameter γ, and can effectively capture the local non-linear correlation in mechanical data and the non-subtle changes in data.

[0080] Specifically, electrical data (such as leakage current, harmonic distortion rate, voltage fluctuation) often contains polynomial relationships. The polynomial kernel function explicitly models the polynomial relationships in the data through the parameter d (polynomial degree). For example, the square relationship between the leakage current and the surface conductivity of the insulator can be directly expressed by a quadratic polynomial kernel without relying on the implicit non-linear mapping of the Gaussian kernel.

[0081] Optionally, the input vector further includes the fusion features of different types of data; the computing device can explicitly capture the combined effect between features through the cross terms of the polynomial kernel function.

[0082] The expression of the combined kernel function is:

[0083] K = αK Gaussian (x i , x j ) + βK Poly (x i , x j );

[0084] Among them, K represents the value of the combined kernel function, α represents the weight coefficient of the Gaussian kernel function, β represents the weight coefficient of the polynomial kernel function, and α + β = 1.

[0085] For example, for any two samples, the first sample x i and the second sample x j , first calculate their similarities under different data sources using the corresponding kernel functions respectively. Assume x i = (x lx,i , x dq,i ), x j = (x lx,j , x dq,j ), where x lx,i and x lx,j are the mechanical data parts, and x dq,i and x dq,j are the electrical data parts.

[0086] Calculate K Gaussian (x lx,i , x lx,j ), to obtain the similarity of the mechanical data part; calculate K Poly (x dq,i , x dq,j ), to obtain the similarity of the electrical data part.

[0087] Then, through the linear combination formula K = αK Gaussian (x i , x j ) + βK Poly (x i , x j ), the final combined similarity is obtained, where α and β are weight coefficients used to control the relative importance of the mechanical data and the electrical data in the combined kernel function.

[0088] The contribution of the kernel function can be dynamically adjusted through the weight coefficients to avoid the limitations of a single kernel function.

[0089] It can be understood that the multi-kernel KFDA model can be trained by the computing device or by a device with data processing capabilities outside the computing device. Taking the training of the multi-kernel KFDA model by the computing device as an example, the specific training process is as follows:

[0090] 1. The computing device inputs a labeled training set {(x i, y i )}。

[0091] where x i is the input vector of multi-source features, and y i is the icing risk level.

[0092] 2. Solve the generalized eigenvalue problem.

[0093] The goal of Fisher discriminant analysis (FDA) is to find a projection direction such that data of different classes are as separated as possible after projection, and data of the same class are as clustered as possible after projection. While KFDA maps the data to a high-dimensional feature space through a kernel function and performs FDA operations in this high-dimensional space.

[0094] In the multi-kernel KFDA model, it is necessary to define the within-class scatter matrix S w and the between-class scatter matrix S b in the representation of the kernel space. Let N be the total number of samples, C be the number of classes, N i be the number of samples in the i-th class, φ(x) is the function that maps the sample x to the high-dimensional feature space, K is the kernel matrix, and its element K ij = k(x i , x j ).

[0095] The within-class scatter matrix S w and the between-class scatter matrix S b in the representation of the kernel space can be constructed through the kernel matrix K. The goal of KFDA is to find the projection vector a such that the ratio J of the between-class scatter to the within-class scatter after projection is maximized, that is, to solve the following optimization problem:

[0096] J(a) = a T S b a / a T S w a

[0097] To solve the above optimization problem, it is necessary to take the derivatives of the numerator and denominator respectively and set the derivatives to zero, which is finally transformed into a generalized eigenvalue problem:

[0098] S b a = λS w a

[0099] where λ is the generalized eigenvalue and a is the corresponding generalized eigenvector.

[0100] 3. Extract the generalized eigenvectors corresponding to the N largest generalized eigenvalues and construct the projection matrix W.

[0101] The generalized eigenvectors a obtained by solving the generalized eigenvalue problem correspond to projection directions in kernel space. These projection directions allow high-dimensional kernel space data to be projected into a lower-dimensional space, thereby achieving dimensionality reduction and classification. The generalized eigenvectors corresponding to the largest generalized eigenvalues are typically chosen as projection directions because these directions maximize the separation of different data categories, improving classification accuracy.

[0102] Preferably, the dimension of the projection matrix W is 3, that is, the projection matrix W is constructed by extracting the generalized eigenvectors corresponding to the three largest generalized eigenvalues.

[0103] Thus, the computing device completes the training of the KFDA model. After that, when the input vector is input into the multi-core KFDA model, the low-dimensional output vector output by the multiple KFDA models after projecting the input vector into the low-dimensional space can be obtained.

[0104] Preferably, after completing the training of the KFDA model, the computing device may first verify the performance of the multi-core KFDA model before putting it into use. Exemplarily, the computing device may use 5-fold cross validation to evaluate the performance of the KFDA model.

[0105] First, the computing device can randomly divide the original data set into 5 non-overlapping subsets (called "folds"), and the size of each subset is roughly the same; if the data categories are unevenly distributed, stratified 5-fold cross-validation is used to ensure that the proportion of each type of sample in each fold is consistent.

[0106] Then, each of the five folds is used as a validation set in turn, and the remaining four folds are used as training sets. The model is trained on the training set, validated on the validation set, and the corresponding performance indicators are calculated. Specifically, the computing device can calculate the mean squared error (MSE) and the coefficient of determination R of the model. 2 , the calculation formula of MSE is:

[0107]

[0108] Among them, N represents the number of samples, i is the index variable used to traverse each sample; y i represents the actual value or true value of the i-th sample, represents the predicted value for the i-th sample. MSE directly measures the accuracy of the model's predictions by calculating the average of the squared differences between the predicted and true values. The smaller the MSE value, the closer the model predictions are to the true values, and the smaller the error. Furthermore, MSE can be used as the loss function for the multi-core KFDA model, allowing model parameters to be adjusted by minimizing the MSE.

[0109] R 2 The calculation formula is:

[0110]

[0111] Among them, represents the average value of the actual values of all samples, is the sum of squared errors (SSE), representing the total deviation between the predicted value and the true value; is the total sum of squares (SST), representing the total deviation between the true value and the average value. R 2 By comparing the model prediction error (SSE) with the data's own fluctuation (SST), the fitting ability of the model to the data is measured.

[0112] Finally, the average value and standard deviation of the 5 validation results are calculated and summarized as the performance evaluation of this model.

[0113] If the performance of the final evaluation is not good, the computing device can optimize and adjust the kernel function parameters or projection dimension of this model, and then perform a validation evaluation again.

[0114] When the computing device passes through the trained KFDA model, the computing device can directly project the input vector, and the projection calculation formula is:

[0115] z = W T φ(x new )

[0116] Among them, z is the low-dimensional output vector obtained after projection, W is the projection matrix of this KFDA model, φ(x) is the function that maps the sample x to the high-dimensional feature space, and x new is the input vector.

[0117] It can be understood that directly inputting the high-dimensional features of multi-source data into the SVR model will cause problems such as high computational complexity and high overfitting risk. Through the kernel function mapping and projection of the multi-kernel KFDA model in this application embodiment, the key features of this multi-source data are extracted, the data discrimination information is retained, it is more conducive to the subsequent calculation of the SVR model, and the computational efficiency and accuracy of the SVR model can be improved.

[0118] Step 104, the computing device outputs the icing risk level of this transmission line based on this low-dimensional output vector through a preset support vector regression (SVR) model.

[0119] After the input vector undergoes kernel function mapping and projection, a low-dimensional output vector is obtained; then the computing device can pass this low-dimensional output vector to the SVR model to determine and output the icing risk level.

[0120] Preferably, the SVR model adopts a Gaussian kernel function, and the penalty constant of the SVR is 1.

[0121] Among them, the regression function expression of the SVR model is:

[0122]

[0123] Among them, α i and are Lagrange multipliers, b is the bias term, N is the number of samples, K is the kernel function, is the predicted icing level; α i is greater than or equal to 0, is less than the penalty constant, and the non-zero α i and corresponding sample points x i are support vectors, and x new is the low-dimensional output vector input to the SVR model.

[0124] It can be understood that the Gaussian kernel function is similar to the Gaussian kernel function of the above-mentioned multi-kernel KFDA model, and will not be elaborated here.

[0125] Through the close combination of feature dimensionality reduction and regression prediction, an accurate mapping from multi-source data to risk levels is achieved. Specifically, the computing device extracts the low-dimensional features of multi-source data through the multi-kernel KFDA model, reduces the data complexity, and provides optimized input for the subsequent SVR model; then, the SVR model uses the low-dimensional features to achieve efficient and accurate icing risk prediction.

[0126] Preferably, after determining the icing risk level of the transmission line through a preset support vector regression (SVR) model based on the low-dimensional output vector, the method further includes: outputting alarm information corresponding to the icing risk level based on the icing risk level.

[0127] Then, the computing device can determine the alarm information to be output based on the calculated icing risk level.

[0128] Exemplarily, when the icing risk level is greater than or equal to 9, the computing device can determine that the current warning level is urgent and output alarm information requiring immediate de-icing to a display device perceivable by maintenance personnel or management personnel; when the icing risk level is less than 9 and greater than or equal to 7, the computing device can determine that the current warning level is high and output alarm information requiring scheduled maintenance to the display device; when the icing risk level is less than 7 and greater than 5, the computing device can determine that the current warning level is medium and output alarm information requiring enhanced detection to the display device; when the icing risk level is less than or equal to 5, the computing device can determine that the current warning level is low and output alarm information only requiring routine inspection to be maintained to the display device.

[0129] Preferably, after outputting the icing risk level of the transmission line through a preset Support Vector Regression (SVR) model based on the low-dimensional output vector, the method further includes: periodically updating the parameters of the multi-kernel KFDA model; the update formula is as follows:

[0130] θ new = θ old + η▽ θ L(θ; D new )

[0131] where θ old and θ new respectively represent the model parameters before and after the update, η is the learning rate, ▽ θ represents the gradient of the loss function with respect to the model parameters, and L is the loss function calculated based on the new dataset D new . Specifically, the model parameters may include the scale parameter, polynomial order, weight coefficient in the kernel function, and the dimension number of the projection matrix.

[0132] In the embodiments of the present application, by constructing an input vector based on the mechanical data, electrical data, environmental data, and visual data of the transmission line, the characteristics of the transmission line can be more comprehensively characterized; then, through the multi-kernel KFDA model, the input vector constructed from the high-dimensional multi-source data of the transmission line is projected into a low-dimensional feature space to capture the non-linear relationship of the input vector data, extract and retain the most discriminative features of the input vector, and at the same time reduce the data dimension, thereby improving the detection accuracy and efficiency of the SVR model for the icing condition of the transmission line.

[0133] The method provided by the embodiments of the present application has been described above. Next, the system provided by the embodiments of the present application will be described.[[ID=3!]]

[0134] Please refer to Figure 2 , Figure 2 which is a schematic structural diagram of an icing detection system for a transmission line provided by the embodiments of the present application. As Figure 2 shown, the system 200 includes:

[0135] An acquisition module 201, configured to acquire the mechanical data, electrical data, environmental data, and visual data of the transmission line;

[0136] A construction module 202, configured to construct an input vector based on the mechanical data, the electrical data, the environmental data, and the visual data;

[0137] A dimensionality reduction module 203, configured to input the input vector into a preset multi-kernel KFDA model to obtain a low-dimensional output vector output by the multi-kernel KFDA model;

[0138] An output module 204, configured to output the icing risk level of the transmission line based on the low-dimensional output vector through a preset SVR model.

[0139] Preferably, the mechanical data includes conductor inclination and vibration spectrum, the electrical data includes leakage current and harmonic distortion rate, the environmental data includes temperature and humidity gradient, and the visual data includes the icing thickness calculated from the image of the transmission line.

[0140] Preferably, the multi-kernel KFDA model includes a combined kernel function, and the combined kernel function includes a Gaussian kernel function and a polynomial kernel function; the expression of the Gaussian kernel function is:

[0141] K Gaussian (x i ,x j )=exp(-γ||x i -x j ||);

[0142] Wherein, K Gaussian represents the value of the Gaussian kernel function, and is used to indicate the similarity between the first sample x i and the second sample x j ; the exp function is an exponential function, and γ represents the scale parameter of the kernel function;

[0143] The expression of the polynomial kernel function is:

[0144]

[0145] Wherein, K Poly represents the value of the polynomial kernel function, the inner product of the first sample and the second sample, and d represents the order of the polynomial;

[0146] The expression of the combined kernel function is:

[0147] K=αK Gaussian (x i ,x j )+βK Poly (x i ,x j );

[0148] Wherein, K represents the value of the combined kernel function, α represents the weight coefficient of the Gaussian kernel function, β represents the weight coefficient of the polynomial kernel function, and α + β = 1.

[0149] Preferably, the dimensionality reduction module 203 is specifically configured to process the elements corresponding to the mechanical data, the environmental data, and the visual data in the input vector through the Gaussian kernel function; process the elements corresponding to the electrical data in the input vector through the polynomial kernel function; and obtain the low-dimensional output vector calculated by the multi-kernel KFDA model based on the processing results of the Gaussian kernel function and the polynomial kernel function.

[0150] Preferably, the SVR model uses a Gaussian kernel function, and the penalty constant of the SVR is 1.

[0151] Preferably, the system 200 further includes: an alarm module 205, configured to output alarm information corresponding to the icing risk level based on the icing risk level.

[0152] Preferably, the system 200 further includes: an update module 206, configured to periodically update the parameters of the multi-kernel KFDA model; the update formula includes:

[0153] θ new = θ old + η▽ θ L(θ; D new )

[0154] where θ old and θ new respectively represent the model parameters before and after the update, η is the learning rate, ▽ θ represents the gradient of the loss function with respect to the model parameters, and L is the loss function calculated based on the new data set D new .

[0155] The icing detection system for a transmission line provided by the embodiments of the present application can be understood by referring to the corresponding content in the foregoing method embodiment section, and will not be repeated here.

[0156] As Figure 3 shown, Figure 3 FIG. is a possible schematic logical structure diagram of a computing device provided by an embodiment of the present application. The computing device 300 includes: a processor 301, a communication interface 302, a memory 303, and a bus 304. The processor 301, the communication interface 302, and the memory 303 are interconnected through the bus 304. In the embodiments of the present application, the processor 301 is used to control and manage the actions of the computing device 300. For example, the processor 301 is used to execute Figure 1 the steps in the embodiment and / or other processes for the technologies described herein. The communication interface 302 is used to support the computing device 300 to communicate. The memory 303 is used to store the program code and data of the computing device 300.

[0157] Among them, the processor 301 may be a central processing unit, a general-purpose processor, a digital signal processor, an application-specific integrated circuit, a field-programmable gate array, or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute various exemplary logic blocks, modules, and circuits described in connection with the disclosure of this application. The processor may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a digital signal processor and a microprocessor, and so on. The bus 304 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For the sake of convenience of representation, Figure 3 only a thick line is used to represent it in Figure 3 , but it does not mean that there is only one bus or one type of bus.

[0158] In another embodiment of the present application, a computer-readable storage medium is further provided. The computer-readable storage medium includes instructions that, when run on a computer, cause the computer to execute the above Figure 1 method described in the embodiment.

[0159] Those of ordinary skill in the art can realize that the units of each example described in connection with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the components of each example have been generally described according to their functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0160] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described systems, devices, and units can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein.

[0161] In several embodiments provided by the embodiments of the present application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of the devices or units can be in electrical, mechanical, or other forms.

[0162] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0163] In addition, in each embodiment of the present invention, the functional units can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0164] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The foregoing storage medium includes: USB flash drives, read-only memories (ROMs), random access memories (RAMs), mobile hard disks, magnetic disks, or optical discs, and other various media that can store program codes.

[0165] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than limiting them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered by the scope of the claims and the description of the present invention.

Claims

1. A method for detecting icing on a transmission line, characterized in that, Including: Obtain the mechanical data, electrical data, environmental data, and visual data of the transmission line; Based on the mechanical data, the electrical data, the environmental data, and the visual data, construct an input vector; Input the input vector into a preset multi-kernel kernel Fisher discriminant analysis KFDA model to obtain a low-dimensional output vector output by the multi-kernel KFDA model; Based on the low-dimensional output vector, output the icing risk level of the transmission line through a preset support vector regression SVR model.

2. The method according to claim 1, wherein The mechanical data includes conductor inclination and vibration spectrum, the electrical data includes leakage current and harmonic distortion rate, the environmental data includes temperature and humidity gradient, and the visual data includes the icing thickness calculated from the image of the transmission line.

3. The method according to claim 1 or 2, characterized in that, The multi-kernel KFDA model includes a combined kernel function, and the combined kernel function includes a Gaussian kernel function and a polynomial kernel function; The expression of the Gaussian kernel function is: K Gaussian (x i ,x j ) = exp(-γ||x i -x j ||); Among them, K Gaussian represents the value of the Gaussian kernel function, which is used to indicate the similarity between the first sample x i and the second sample x j ; the exp function is the exponential function, and γ represents the scale parameter of the kernel function; The expression of the polynomial kernel function is: Among them, K Poly represents the value of the polynomial kernel function, the inner product of the first sample and the second sample, d represents the order of the polynomial, and C is a constant term; The expression of the combined kernel function is: K = αK Gaussian (x i , x j ) + βK Poly (x i , x j ); Where K represents the value of the combined kernel function, α represents the weight coefficient of the Gaussian kernel function, β represents the weight coefficient of the polynomial kernel function, and α + β = 1.

4. The method according to claim 3, wherein The step of inputting the input vector into a preset multi-kernel kernel Fisher discriminant analysis KFDA model to obtain a low-dimensional output vector output by the multi-kernel KFDA model includes: Process the elements in the input vector corresponding to the mechanical data, the environmental data, and the visual data through the Gaussian kernel function; Process the elements in the input vector corresponding to the electrical data through the polynomial kernel function; Obtain the low-dimensional output vector calculated by the multi-kernel KFDA model based on the processing results of the Gaussian kernel function and the polynomial kernel function.

5. The method according to claim 1 or 2, characterized in that, The SVR model uses a Gaussian kernel function, and the penalty constant of the SVR model is 1.

6. The method according to claim 1 or 2, characterized in that, After outputting the icing risk level of the transmission line through a preset support vector regression SVR model based on the low-dimensional output vector, the method further includes: Based on the icing risk level, output an alarm message corresponding to the icing risk level.

7. The method according to claim 1 or 2, characterized in that, After outputting the icing risk level of the transmission line through a preset support vector regression SVR model based on the low-dimensional output vector, the method further includes: Periodically update the parameters of the multi-kernel KFDA model; the update formula includes: Among them, θ old and θ new represent the model parameters before and after update respectively, η is the learning rate, and ▽ θ represents the gradient of the loss function with respect to the model parameters, and L is the loss function calculated based on the new dataset D new .

8. An icing detection system for a transmission line, characterized in that, Applied to the method according to any one of claims 1-7, including: An acquisition module for acquiring the mechanical data, electrical data, environmental data, and visual data of the transmission line; A construction module for constructing an input vector based on the mechanical data, the electrical data, the environmental data, and the visual data; A dimensionality reduction module for inputting the input vector into a preset multi-kernel kernel Fisher discriminant analysis KFDA model to obtain a low-dimensional output vector output by the multi-kernel KFDA model; An output module for outputting the icing risk level of the transmission line through a preset support vector regression SVR model based on the low-dimensional output vector.

9. A computing device, characterized in that, Including: A memory for storing programs; A processor for loading the program to execute the method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein when the program runs, it controls the device where the computer-readable storage medium is located to execute the method according to any one of claims 1-7.

Images