Icing overhead line prediction method and system based on support vector machine GA-SVM

By combining support vector machines with genetic algorithms, combining singular value decomposition to extract spatiotemporal mode features, optimize SVM model parameters, the noise interference and parameter sensitivity problems in large-scale high-dimensional data processing are solved, and the prediction of ice-covered overhead lines is significantly improved. Accuracy and timeliness provide scientific decision-making support for the power system.

CN118940127APending Publication Date: 2024-11-12GUIZHOU POWER GRID CO LTD
View PDF 0 Cites 3 Cited by

Patent Information

Application Number
CN202410882039.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-03
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The prior art has serious noise interference and outliers when processing large-scale, high-dimensional meteorological and environmental data, and the support vector machine is more sensitive in parameter selection, resulting in large deviations in the prediction results of ice-covered overhead lines and it is difficult to capture complex climate patterns.

Method used

Using a method of combining support vector machine with genetic algorithm (GA-SVM), historical data is finely processed, spatiotemporal pattern features are extracted through singular value decomposition, representative feature vectors are constructed, and parameters of SVM model are optimized through genetic algorithms, so as to Improve prediction accuracy and generalization capabilities.

Benefits of technology

It significantly improves the accuracy and timeliness of ice-covered overhead line prediction, provides scientific basis and technical support for the prevention and disaster reduction of the power system, and enhances the generalization ability and prediction accuracy of the model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118940127A_ABST
    Figure CN118940127A_ABST
Patent Text Reader

Abstract

The invention discloses an icing overhead line prediction method and system based on a support vector machine GA-SVM, and relates to the field of intelligent power grid monitoring and maintenance, and the method comprises the steps: collecting the historical data information of a target line, carrying out the data processing, and constructing an ice coverage thickness time sequence; converting the ice coverage thickness time sequence data into a multivariate time sequence data matrix, and performing singular value decomposition; based on the effect of a singular value, identifying a main mode, and forming a feature space through linear combination conversion; extracting a statistical feature of each time point from the feature space to form a feature vector for an SVM (Support Vector Machine) model; and establishing a linear SVM model, introducing a genetic algorithm to optimize parameters of the SVM model, and determining optimal regularization parameters. According to the invention, through integration of historical data processing, advanced feature extraction, the optimized machine learning model and multi-dimensional prediction integration, the accuracy, timeliness and practicability of icing prediction are significantly enhanced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of smart grid monitoring and maintenance, and in particular to a method and system for predicting ice-covered overhead lines based on a support vector machine (GA-SVM). Background Art

[0002] Existing technologies still have several deficiencies when processing large-scale, high-dimensional meteorological and environmental data. On the one hand, noise interference and outliers in the original data seriously affect the training effect of the model. Simple preprocessing methods are difficult to effectively remove these interferences, resulting in large deviations in the prediction results. On the other hand, most models directly apply the original data for learning, ignoring the potential spatiotemporal correlation structure within the data, which limits the model's ability to capture complex climate patterns. In addition, support vector machines (SVMs), as a powerful classification and regression tool, show excellent performance in dealing with small samples and nonlinear problems, but are sensitive to parameter selection, and improper parameter configuration will significantly reduce its prediction performance. Although some studies have attempted to optimize SVMs through manual parameter adjustment or grid search, this usually requires a lot of computing resources and is prone to falling into local optimal solutions.

[0003] The present invention proposes a prediction method and system for ice-covered overhead lines based on a combination of support vector machines and genetic algorithms (GA-SVM). The method first performs fine processing on the historical data of the target line, including smoothing and denoising, to construct high-quality time series data, and then uses singular value decomposition technology to extract key spatiotemporal pattern features from high-dimensional data, effectively reducing the data dimension while retaining important information. Furthermore, through statistical analysis of these spatiotemporal features, more representative feature vectors are constructed, providing a more accurate input for the SVM model. On this basis, a genetic algorithm is introduced to globally optimize the key parameters of the SVM to ensure that the model can find the best hyperparameter configuration in complex data distribution, thereby significantly improving the generalization ability and prediction accuracy of the prediction model. Finally, by integrating all prediction results, a comprehensive and accurate prediction of the ice thickness in the entire monitoring area is achieved, providing a scientific basis and technical support for ice prevention and disaster reduction in the power system. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a method and system for predicting ice-covered overhead lines based on support vector machine GA-SVM to solve the problem of high-precision prediction of ice cover thickness under complex climatic conditions by integrating spatiotemporal pattern feature representation of genetic algorithm and singular value decomposition.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In the first aspect, an embodiment of the present invention provides a method for predicting ice-covered overhead lines using a support vector machine (GA-SVM), which includes collecting historical data information of the target line, performing data processing, and constructing a time series of ice cover thickness; converting the time series data of ice cover thickness into a multivariate time series data matrix, and performing singular value decomposition; based on the role of singular values, identifying the main modes, and forming a feature space through linear combination transformation; extracting the statistical features of each time point from the feature space to form a feature vector for an SVM model; establishing a linear SVM model, introducing a genetic algorithm to optimize the parameters of the SVM model, and determining the optimal regularization parameters; using the SVM model parameters optimized by the genetic algorithm, integrating the predicted values ​​of all positions and time points, and accumulating the predicted values ​​of each component to obtain the overall ice thickness predicted value of the entire monitoring area.

[0008] As a preferred solution of the ice-covered overhead line prediction method using the support vector machine GA-SVM of the present invention, the specific steps of collecting historical data information of the target line, performing data processing, and constructing an ice cover thickness time series are as follows:

[0009] It connects to the power system monitoring platform through an automated data interface, collects weather station data, historical ice cover records, geographic coordinate information, and time series data from multiple latitudes, uses Python's pandas library to process the received data, sorts the time series data by date and time, and finally stores it in a structured format.

[0010] As a preferred solution of the ice-covered overhead line prediction method based on support vector machine GA-SVM described in the present invention, wherein: the ice cover thickness time series data is converted into a multivariate time series data matrix, and the specific steps are as follows:

[0011] Assume the original time series data is {X t}, where t = 1, 2, ..., N, where N represents N data points of ice thickness collected from the power system monitoring platform,

[0012] For each time point i, starting from this point, L data points are taken continuously to form a delay vector X1 = [X1, X2, ..., X L ];

[0013] Repeat the above process for each i value until all N-L+1 delay vectors are constructed. All delay vectors are stacked vertically. The formula is as follows:

[0014] X i =[X1,X2,...,X n-L+1 ] T ;

[0015] Among them, Xi is the vector of the i-th time window, L represents the window length, N is the total number of time points, and T represents the matrix transpose;

[0016] Perform singular value decomposition on the data matrix X, the formula is as follows:

[0017] X=UΣV * ;

[0018] Among them, U is an m×m left singular orthogonal matrix, V is an n×n right singular orthogonal matrix, Σ is an m×n diagonal matrix, V * is the conjugate transpose of V.

[0019] As a preferred solution of the ice-covered overhead line prediction method based on support vector machine GA-SVM described in the present invention, wherein: based on the role of singular values, the main modes are identified, and the feature space is formed through linear combination transformation. The specific steps are as follows:

[0020] In order to identify the main mode from the singular values, a relative action threshold δ is first determined, and the model is further simplified to retain only the singular values ​​that contribute significantly to the total variance. The formula is as follows:

[0021]

[0022] Among them, r is the number of singular values ​​that need to be retained, δ is the threshold of the cumulative contribution rate, σ i is the i-th singular value;

[0023] Set a specific value for the target contribution rate. r When r is exceeded for the first time, the corresponding r is the number of singular values ​​required, as follows:

[0024]

[0025] Among them, CEV r represents the cumulative explained variance ratio accumulated to the first r singular values, σ i is the i-th singular value;

[0026] Based on the above determined r, the most important r modes are screened out, and the left singular vector matrix after screening is defined as U r And the corresponding singular value diagonal matrix is ​​Σ r , the formula is as follows:

[0027] U r =[u1,u2,...,u r ];

[0028] ∑ r =diag(σ1,σ2,...,σ r );

[0029] The identified modes are transformed through nonlinear functions to weight and enhance the predictive ability in subsequent analysis. The formula is as follows:

[0030]

[0031] Among them, W i represents the adaptive weight of each mode, p i is the correlation coefficient between the ith mode and the predefined physical index, and f is a nonlinear function;

[0032] Design an orthogonal transformation matrix that represents statistical characteristics, the formula is as follows:

[0033] W=diag(w1,w2,...,w r );

[0034] Generate the feature space matrix, the formula is as follows:

[0035]

[0036] in, Each element in the diagonal matrix is ​​the square root of the corresponding singular value, and W represents the enhanced prediction ability through adaptive weights.

[0037] As a preferred solution of the ice-covered overhead line prediction method based on support vector machine GA-SVM described in the present invention, wherein: the statistical features of each time point are extracted from the feature space to generate the SVM model feature vector, and the specific steps are as follows:

[0038] Y osc Perform missing value processing, outlier detection and processing, data standardization and feature selection, and divide the processed data set into training set, validation set and test set;

[0039] For feature space Y osc Each time point i in and the corresponding feature vector y i , calculate its mean u i and variance σ i 2 , the formula is as follows:

[0040]

[0041] Among them, y ij Represents the jth data point in the feature vector corresponding to the i-th time point;

[0042] Select a sliding window w of appropriate size, and at each time point i, use the data points in the window to find the slope m through the linear fitting model y=mx+b iand intercept b i , and the model can be obtained by the least squares method, the formula is as follows:

[0043]

[0044] Among them, m i Represents the local trend slope of the time series at window i. and are the average values ​​of the time points and corresponding feature values ​​in the window, and j represents the Traverse to Take data points of half the window length before and after time point i;

[0045] For each time point i, construct a three-dimensional feature vector: F i =(u i , σ i 2 , m i ), where each pair (u i , σ i 2 ) corresponds to a statistical description of a time point.

[0046] As a preferred solution of the ice-covered overhead line prediction method based on support vector machine GA-SVM described in the present invention, wherein: a linear SVM model is constructed based on the SVM model feature vector, a genetic algorithm is introduced to optimize the parameters of the SVM model, and the optimal regularization parameters are determined. The specific steps are as follows:

[0047] Using the libsvm library, based on the feature vector Y osc The features F extracted from i , build a linear SVM model, and determine the regularization parameter C. The optimization formula of C is as follows:

[0048]

[0049] Under the constraints:

[0050]

[0051] Among them, w is the weight vector, b is the bias term, l is the number of training samples, and y i is the label of the i-th sample, X i is the feature vector F of the i-th sample, ξ i is the slack variable;

[0052] Introduce genetic algorithm to construct the initial population P t , each individual C k represents a candidate value of the regularization parameter k = 1, 2, ..., N, where Ck ∈[C min , C max ];

[0053] The performance index of the model based on cross-validation is defined as the fitness function f(C k ), using the tournament selection method, from P t Randomly select T individuals from the pool for comparison, and select the individual with the highest fitness to enter the next generation;

[0054] Randomly select multiple individuals from the next generation population, use the single-point crossover strategy, randomly select a crossover point, exchange the C value after the selected point, and generate new individuals;

[0055] The C value of each new individual is mutated to C'=C+δ, where δ is a random perturbation factor;

[0056] Calculate the fitness of the new individuals generated by the evaluation mutation f(C k ), and then evaluate the performance of the new individuals, and form a new generation of population P based on all individuals generated by selection, crossover and mutation operations t+1 , enter the next round of iteration, iterate until the maximum number of iterations C is reached max or the best fitness improvement in recent generations is less than a pre-set threshold;

[0057] The optimal regularization parameter C optimized based on genetic algorithm * , input into the SVM model, retrain the model, the formula is as follows:

[0058]

[0059] where w * and b * Especially the optimized weight vector and bias, ||W|| 2 Represents the square of the Euclidean norm of the weight vector w.

[0060] As a preferred solution of the ice-covered overhead line prediction method based on support vector machine GA-SVM described in the present invention, wherein: the SVM model parameters optimized by genetic algorithm are integrated with the prediction values ​​of all positions and time points, and the overall ice thickness prediction value of the entire monitoring area is obtained by accumulating the prediction values ​​of each component. The specific steps are as follows:

[0061] Collect the meteorological condition data of each monitoring point at the current time t, including the key factors affecting the ice thickness, such as temperature, humidity, and wind speed, and combine these data into a new matrix X t , and for the input data X t Perform standardization to eliminate dimensional differences between different features;

[0062] Using the determined parameters in the optimized SVM model and The ice thickness at the next time step t+1 of each monitoring point is predicted using the following formula:

[0063]

[0064] in represents the predicted ice thickness at the i-th monitoring point at time t+1, X i (t) is the normalized eigenvector of point i at time t, and sign represents the sign function;

[0065] A weighted average strategy is introduced to integrate the prediction results. Based on historical data and expert knowledge, a weight α is assigned to each monitoring point to ensure that α i >0 and Σα i =1, Among them I i represents the historical ice coverage impact index of the i-th point;

[0066] The predicted values ​​of all monitoring points are integrated using the weighted average method to form the overall regional predicted value Y tl t+1 , the formula is as follows:

[0067]

[0068] Where N is the total number of monitoring points and α is the weight of the i-th point.

[0069] In the second aspect, the present invention provides an ice-covered overhead line prediction system based on a support vector machine GA-SVM, including a historical data preprocessing module, a time series data conversion and feature extraction module, a spatiotemporal pattern feature construction module, an SVM model optimization and establishment module, and a prediction integration and evaluation module; the historical data preprocessing module is used to collect historical ice thickness data of the line, improve data quality through smoothing and denoising technology, and construct a clear time series to provide an accurate basis for subsequent analysis; the time series data conversion and feature extraction module is used to convert the processed time series data into a multivariate time series data matrix, and perform singular value decomposition, identify key modal features, form a spatiotemporal pattern feature representation through linear combination, further extract statistical features of each time point, and prepare efficient input feature vectors for the SVM model; the spatiotemporal pattern feature construction module is used to convert the processed time series data into a multivariate time series data matrix, perform singular value decomposition, identify key modal features, form a spatiotemporal pattern feature representation through linear combination, further extract statistical features of each time point, and prepare efficient input feature vectors for the SVM model; The building module is used to identify and extract the main spatiotemporal patterns reflecting the dynamic changes of ice cover based on singular value analysis, construct a highly generalized feature representation, and enhance the model's learning ability for complex ice cover patterns; the SVM model optimization and building module is used to use the support vector machine theory and combine the genetic algorithm to globally optimize the SVM model parameters to ensure that the model can perform efficient parameter tuning for specific ice cover prediction tasks while maintaining high generalization ability, forming an optimized GA-SVM classifier; the prediction integration and evaluation module is used to use the optimized GA-SVM model to predict the ice cover thickness of all monitoring points in future time steps. By integrating all predicted values ​​and using accumulation or other statistical methods, the overall ice thickness prediction of the entire monitoring area is comprehensively obtained, providing a scientific basis for line maintenance and decision-making.

[0070] In a third aspect, an embodiment of the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the method for predicting ice-covered overhead lines based on support vector machine GA-SVM as described in the first aspect of the present invention is implemented.

[0071] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the method for predicting ice-covered overhead lines based on a support vector machine GA-SVM as described in the first aspect of the present invention is implemented.

[0072] The beneficial effects of the present invention are as follows: the present invention significantly enhances the accuracy, timeliness and practicality of ice cover prediction by integrating historical data processing, advanced feature extraction, optimized machine learning models and multi-dimensional prediction integration, and provides an accurate and efficient decision support tool for power grid ice cover management. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0074] Figure 1 This is a flow chart of the ice-covered overhead line prediction method using the support vector machine GA-SVM in Example 1. DETAILED DESCRIPTION

[0075] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0076] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0077] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0078] Example 1, reference Figure 1 This is the first embodiment of the present invention, which provides a method for predicting ice-covered overhead lines using a support vector machine GA-SVM, comprising the following steps:

[0079] S1. Collect historical data information of the target line, process the data, and construct a time series of ice cover thickness.

[0080] Furthermore, the automated data interface is connected to the power system monitoring platform to collect weather station data, historical ice records, geographic coordinate information and time series data from multiple latitudes. In the design of the automated data interface, the stable and compatible HTTP protocol is used, and TLS encryption is used for data transmission.

[0081] Use Python's pandas library to process the received data and sort the time series data by date and time.

[0082] S2. Convert the ice cover thickness time series data into a multivariate time series data matrix and perform singular value decomposition.

[0083] Furthermore, let the original time series data be {X t}where t = 1, 2, ..., N, where N represents N data points of ice thickness collected from the power system monitoring platform;

[0084] The optimal length L (which determines the model's ability to capture trends and periodicity in time series) is determined through domain knowledge and cross-validation. For each time point i, starting from that point, L data points are obtained to form a delay vector X1 = [X1, X1, X1, ..., X k ];

[0085] Repeat the above process for each i value until all N-L+1 delay vectors are constructed, and all delay vectors are stacked vertically to form X i =[X1,X2,...,X n-L+1 ] T , where X i is the vector of the i-th time window, L represents the window length, N is the total number of time points, T represents the matrix transpose, and its dimension is (L, N-L+1);

[0086] Perform singular value decomposition on the data matrix X and obtain X = U∑V * , where U is an m×m left singular orthogonal matrix, V is an n×n right singular orthogonal matrix, Σ is an m×n diagonal matrix, V * is the conjugate transpose of V, and the elements of Σ are singular values ​​σ i , where the singular values ​​σ i (shows the variance of the data on different principal components, the larger σ i means that the data change in this direction is more significant) is X·X T The non-negative eigenvalues ​​λ i The square root of

[0087] S3. Based on the role of singular values, the main modes are identified and transformed through linear combinations to form feature space.

[0088] Furthermore, in order to identify the main mode from the singular values, a relative action threshold δ is first determined, and the model is further simplified to retain only those singular values ​​that contribute significantly to the total variance, where δ can be determined by the cumulative contribution rate, where the cumulative contribution rate is given by the ratio of the first r singular values ​​to the sum of all singular values, as shown in the following formula:

[0089]

[0090] Where, for each singular value σ i 2 , will satisfy σ i2 >δσ max The condition σ i be retained, where σ max Represents the largest singular value in the matrix X;

[0091] The specific value of the target contribution rate γ is used to determine the number of singular values. r When r is exceeded for the first time, the corresponding r is the number of singular values ​​required, and the specific formula is as follows:

[0092]

[0093] Among them, CEV r represents the cumulative explained variance ratio accumulated to the first r singular values, σ i is the i-th singular value;

[0094] Based on the above determined r, the most important r modes are screened out (the main direction of data change, each singular value corresponds to a mode of the data matrix), and the left singular vector matrix after screening is defined as U r And the corresponding singular value diagonal matrix is ​​Σ r , (the original high-dimensional data is effectively compressed into a low-dimensional space, and this low-dimensional space maintains the structure and information of the original data as much as possible), where:

[0095] U r =[u1,u2,...,u r ];

[0096] ∑ r =diag(σ1,σ2,...,σ r );

[0097] The identified modes are transformed and weighted through nonlinear functions to reduce the matrix latitude and enhance the model's ability to express complex relationships, thereby improving prediction accuracy. The formula is as follows:

[0098]

[0099] where w i represents the adaptive weight of each mode, p i is the correlation coefficient between the ith mode and the predefined physical index, f(p i ) is a nonlinear function, μ is the expected correlation value, and the μ value can be obtained by performing correlation analysis on each mode and the predefined physical index. σ controls the width of the curve. The influence of different σ values ​​on the prediction performance is evaluated through cross-validation to determine the σ value. When p i When it is close to μ, the weight is close to the maximum value of 1. i Deviating from μ, the weight gradually decreases;

[0100] Design an orthogonal transformation matrix that represents statistical characteristics, the formula is as follows:

[0101] W=diag(w1,w2,...,w r );

[0102] Generate the feature space matrix, the formula is as follows:

[0103]

[0104] in, Each element in the diagonal matrix is ​​the square root of the corresponding singular value, aiming to recover the amplitude information of the original data, while W enhances the prediction ability through adaptive weights.

[0105] S4. Extract the statistical features of each time point from the feature space and generate the SVM model feature vector.

[0106] Furthermore, Y osc Perform missing value processing, outlier detection and processing, data standardization and feature selection, and divide the processed data set into training set, validation set and test set;

[0107] For feature space Y osc Each time point i in and the corresponding feature vector y i , calculate its mean u i and variance σ i 2 , the formula is as follows:

[0108]

[0109] Among them, y ij Represents the jth data point in the feature vector corresponding to the i-th time point;

[0110] Select a sliding window w of appropriate size, and at each time point i, use the data points in the window (i.e., the time point and the total w data points before and after it) to find the slope m through the linear fitting model y=mx+b i and intercept b i , and the model can be obtained by the least squares method:

[0111]

[0112] Where m i Represents the local trend slope of the time series at window i. and are the average values ​​of the time points and corresponding feature values ​​in the window, and j represents the Traverse to It means taking data points of half the window length before and after time point i;

[0113] For each time point i, construct a three-dimensional feature vector: F i =(u i , σ i 2 , m i ), where each pair (u i , σ i 2 ) corresponds to a statistical description of a time point

[0114] S5. Construct a linear SVM model based on the SVM model feature vector, introduce a genetic algorithm to optimize the parameters of the SVM model, and determine the optimal regularization parameters.

[0115] Furthermore, using the libsvm library, based on the feature vector Y osc The features F extracted from i , build a linear SVM model, and determine the regularization parameter C (control the balance between model complexity and generalization ability, a larger C may lead to overfitting, and a smaller C may lead to underfitting). The optimization formula of C is as follows,

[0116]

[0117] Under the constraints:

[0118]

[0119] Among them, w is the weight vector, b is the bias term, l is the number of training samples, and y i is the label of the i-th sample, X i is the feature vector F of the i-th sample, ξ i is a slack variable used to handle the situation where it is not completely linearly separable;

[0120] Introduce genetic algorithm to construct the initial population P t , each individual C k represents a candidate value of the regularization parameter k = 1, 2, ..., N, where C k ∈[C min , C max ];

[0121] The performance index of the model based on cross-validation is defined as the fitness function f(C k ), using the tournament selection method, from P t Randomly select T individuals from the pool for comparison, and select the individual with the highest fitness to enter the next generation;

[0122] Randomly select multiple individuals from the next generation population, use the single-point crossover strategy, randomly select a crossover point, exchange the C value after the selected point, and generate new individuals;

[0123] Mutate the C value of each new individual to C'=C+δ, where δ is a small random disturbance factor;

[0124] Calculate the fitness of the new individuals generated by the evaluation mutation f(C k ), and then evaluate the performance of the new individuals, and form a new generation of population P based on all individuals generated by selection, crossover and mutation operations t+1 , enter the next round of iteration, iterate until the maximum number of iterations C is reached max or the best fitness improvement in recent generations is less than a pre-set threshold;

[0125] The optimal regularization parameter C optimized based on genetic algorithm * , input into the SVM model, retrain the model, the formula is as follows:

[0126]

[0127] where w * and b * are the optimized weight vector and bias, ∥w∥ 2 Represents the square of the Euclidean norm of the weight w.

[0128] S6. Use the SVM model parameters optimized by genetic algorithm to integrate the predicted values ​​of all locations and time points. By accumulating the predicted values ​​of each component, the overall ice thickness prediction value of the entire monitoring area is obtained.

[0129] Furthermore, the meteorological conditions (temperature, humidity, wind speed), geographical location, altitude, conductor characteristics, line direction and layout and other key factors affecting ice thickness at each monitoring point at the current time t are collected, and these data are combined into a new matrix X t , and for the input data X t Perform standardization to eliminate dimensional differences between different features;

[0130] Using the determined parameters in the optimized SVM model and The ice thickness at the next time step t+1 of each monitoring point is predicted using the following formula:

[0131]

[0132] in represents the predicted ice thickness at the i-th monitoring point at time t+1, X i (t)is the normalized eigenvector of point i at time t, and sign represents the sign function

[0133] A weighted average strategy is introduced to integrate the prediction results. Based on historical data and expert knowledge, a weight α is assigned to each monitoring point to ensure that α i >0 and Σα i =1, Among them I i represents the historical ice coverage impact index of the i-th point;

[0134] The predicted values ​​of all monitoring points are integrated using the weighted average method to form the overall regional prediction value. The formula is as follows:

[0135]

[0136] Where N is the total number of monitoring points, and α is the weight of the i-th point;

[0137] Compare the predicted results with the actual values ​​subsequently observed, evaluate the accuracy of the prediction interval and whether the uncertainty estimate of the model is appropriate, and continuously iterate and optimize the model and prediction method.

[0138] The present embodiment also provides an ice-covered overhead line prediction system of a support vector machine GA-SVM, including: a historical data preprocessing module, a time series data conversion and feature extraction module, a spatiotemporal pattern feature construction module, an SVM model optimization and establishment module, and a prediction integration and evaluation module; the historical data preprocessing module is used to collect historical ice thickness data of the line, improve data quality through smoothing and denoising technology, and construct a clear time series to provide an accurate basis for subsequent analysis; the time series data conversion and feature extraction module is used to convert the processed time series data into a multivariate time series data matrix, and perform singular value decomposition, identify key modal features, form a spatiotemporal pattern feature representation through linear combination, further extract statistical features of each time point, and prepare efficient input feature vectors for the SVM model; the spatiotemporal pattern feature construction The module is used to identify and extract the main spatiotemporal patterns reflecting the dynamic changes of ice cover based on singular value analysis, construct a highly generalized feature representation, and enhance the model's learning ability for complex ice cover patterns; the SVM model optimization and establishment module is used to use the support vector machine theory and the genetic algorithm to globally optimize the SVM model parameters to ensure that the model can perform efficient parameter tuning for specific ice cover prediction tasks while maintaining high generalization ability, thereby forming an optimized GA-SVM classifier; the prediction integration and evaluation module is used to use the optimized GA-SVM model to predict the ice cover thickness of all monitoring points in future time steps. By integrating all predicted values ​​and using accumulation or other statistical methods, the overall ice thickness prediction of the entire monitoring area is comprehensively obtained, providing a scientific basis for line maintenance and decision-making.

[0139] This embodiment also provides a computer device, which is suitable for the case of the ice-covered overhead line prediction method using a support vector machine GA-SVM, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the ice-covered overhead line prediction method using a support vector machine GA-SVM as proposed in the above embodiment.

[0140] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0141] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for predicting ice-covered overhead lines by implementing a support vector machine GA-SVM as proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (Static Random Access Memory, referred to as SRAM), an electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), an erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), a programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), a read-only memory (Read-Only Memory, referred to as ROM), a magnetic storage device, a flash memory, a disk or an optical disk.

[0142] In summary, the present invention realizes the monitoring and understanding of user behavior through the analysis of user file operation behavior, can accurately capture the user's operation habits and behavior patterns, provide basic data for the subsequent generation of encryption strategies, and also help the system to detect and prevent abnormal behaviors; predict the user's future operations according to the user's behavior habits, so as to formulate encryption strategies more intelligently; protect user privacy more accurately, reduce the risk of data leakage, and improve the data security and confidentiality of the system; the present invention better adapts to the user's operation habits and needs, improves the intelligence level of the encryption algorithm and user experience, and enhances the flexibility and adaptability of the system.

[0143] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for predicting ice-covered overhead lines based on a support vector machine GA-SVM, characterized in that: include: Collect historical data information of the target line, process the data, and construct a time series of ice cover thickness; The ice cover thickness time series data are converted into a multivariate time series data matrix and subjected to singular value decomposition; Based on the role of singular values, the main modes are identified and transformed through linear combination to form a feature space; Extract the statistical features of each time point from the feature space and generate the SVM model feature vector; A linear SVM model is constructed based on the SVM model feature vector, and a genetic algorithm is introduced to optimize the parameters of the SVM model to determine the optimal regularization parameters. The SVM model parameters optimized by genetic algorithm are used to integrate the predicted values ​​of all locations and time points. The predicted value of the overall ice thickness in the entire monitoring area is obtained by accumulating the predicted values ​​of each component.

2. The ice-covered overhead line prediction method using the support vector machine GA-SVM according to claim 1, characterized in that: The specific steps of collecting historical data information of the target line, processing the data, and constructing the ice cover thickness time series are as follows: It connects to the power system monitoring platform through an automated data interface, collects weather station data, historical ice cover records, geographic coordinate information, and time series data from multiple latitudes, uses Python's pandas library to process the received data, sorts the time series data by date and time, and finally stores it in a structured format.

3. The ice-covered overhead line prediction method using support vector machine GA-SVM according to claim 2, characterized in that: The specific steps of converting the ice cover thickness time series data into a multivariate time series data matrix are as follows: Assume the original time series data is {X t }, where t = 1, 2, ..., N, where N represents N data points of ice thickness collected from the power system monitoring platform, For each time point i, starting from this point, L data points are taken continuously to form a delay vector X1 = [X1, X1, X1, ..., X k ]; Repeat the above process for each i value until all N-L+1 delay vectors are constructed. All delay vectors are stacked vertically. The formula is as follows: X i =[X1,X2,...,X n-L+1 ] T ; Among them, X i is the vector of the i-th time window, L represents the window length, N is the total number of time points, and T represents the matrix transpose; Perform singular value decomposition on the data matrix X, the formula is as follows: X=UΣV * ; Among them, U is an m×m left singular orthogonal matrix, V is an n×n right singular orthogonal matrix, Σ is an m×n diagonal matrix, V * is the conjugate transpose of V.

4. The ice-covered overhead line prediction method using support vector machine GA-SVM as claimed in claim 3, characterized in that: Based on the role of singular values, the main modes are identified and transformed into feature space through linear combination. The specific steps are as follows: In order to identify the main mode from the singular values, a relative action threshold δ is first determined, and the model is further simplified to retain only the singular values ​​that contribute significantly to the total variance. The formula is as follows: Among them, r is the number of singular values ​​that need to be retained, δ is the threshold of the cumulative contribution rate, σ i is the i-th singular value; Set a specific value for the target contribution rate. r When r is exceeded for the first time, the corresponding r is the number of singular values ​​required, as follows: Among them, CEV r represents the cumulative explained variance ratio accumulated to the first r singular values, σ i is the i-th singular value; Based on the above determined r, the most important r modes are screened out, and the left singular vector matrix after screening is defined as U r And the corresponding singular value diagonal matrix is ​​Σ r , the formula is as follows: U r =[u1,u2,...,u r ]; S r =diag(σ1,σ2,...,σ r ); The identified modes are transformed through nonlinear functions to weight and enhance the predictive ability in subsequent analysis. The formula is as follows: Among them, W i represents the adaptive weight of each mode, p i is the correlation coefficient between the ith mode and the predefined physical index, and f is a nonlinear function; Design an orthogonal transformation matrix that represents statistical characteristics, the formula is as follows: In=diag(in1,in2,...,in r ); Generate the feature space matrix, the formula is as follows: in, Each element in the diagonal matrix is ​​the square root of the corresponding singular value, and W represents the enhanced prediction ability through adaptive weights.

5. The ice-covered overhead line prediction method using support vector machine GA-SVM according to claim 4, characterized in that: The statistical features of each time point are extracted from the feature space to generate the SVM model feature vector. The specific steps are as follows: Y osc Perform missing value processing, outlier detection and processing, data standardization and feature selection, and divide the processed data set into training set, validation set and test set; For feature space Y osc Each time point i in and the corresponding feature vector y i , calculate its mean u i and variance σ i 2 , the formula is as follows: Among them, y ij Represents the jth data point in the feature vector corresponding to the i-th time point; Select a sliding window w of appropriate size, and at each time point i, use the data points in the window to find the slope m through the linear fitting model y=mx+b i and intercept b i , and the model can be obtained by the least squares method, the formula is as follows: Among them, m i Represents the local trend slope of the time series at window i. and are the average values ​​of the time points and corresponding feature values ​​in the window, and j represents the Traverse to Take data points of half the window length before and after time point i; For each time point i, construct a three-dimensional feature vector: F i =(u i , σ i 2 , m i ), where each pair (u i , σ i 2 ) corresponds to a statistical description of a time point.

6. The ice-covered overhead line prediction method using support vector machine GA-SVM according to claim 5, characterized in that: A linear SVM model is constructed based on the SVM model feature vector, and a genetic algorithm is introduced to optimize the parameters of the SVM model to determine the optimal regularization parameters. The specific steps are as follows: Using the libsvm library, based on the feature vector Y osc The features F extracted from i , build a linear SVM model, and determine the regularization parameter C. The optimization formula of C is as follows: Under the constraints: y i (w x i +b)≥1-ξ i ,x i ≥0, Among them, w is the weight vector, b is the bias term, l is the number of training samples, and y i is the label of the i-th sample, X i is the feature vector F of the i-th sample, ξ i is the slack variable; Introduce genetic algorithm to construct the initial population P t , each individual C k represents a candidate value of the regularization parameter k = 1, 2, ..., N, where C k ∈[C min , C max ]; The performance index of the model based on cross-validation is defined as the fitness function f(C k ), using the tournament selection method, from P t Randomly select T individuals from the pool for comparison, and select the individual with the highest fitness to enter the next generation; Randomly select multiple individuals from the next generation population, use the single-point crossover strategy, randomly select a crossover point, exchange the C value after the selected point, and generate new individuals; The C value of each new individual is mutated to C'=C+δ, where δ is a random perturbation factor; Calculate the fitness of the new individuals generated by the evaluation mutation f(C k ), and then evaluate the performance of the new individuals, and form a new generation of population P based on all individuals generated by selection, crossover and mutation operations t+1 , enter the next round of iteration, iterate until the maximum number of iterations C is reached max or the best fitness improvement in recent generations is less than a pre-set threshold; The optimal regularization parameter C optimized based on genetic algorithm * , input into the SVM model, retrain the model, the formula is as follows: where w * and b * Especially the optimized weight vector and bias, ||W|| 2 Represents the square of the Euclidean norm of the weight vector w, which is to maximize the decision boundary and solve w under the same constraints. * and b * So that the model can achieve the best classification on the given training data set.

7. The ice-covered overhead line prediction method using support vector machine GA-SVM according to claim 6, characterized in that: The SVM model parameters optimized by the genetic algorithm are integrated with the predicted values ​​of all locations and time points, and the predicted value of the overall ice thickness of the entire monitoring area is obtained by accumulating the predicted values ​​of each component. The specific steps are as follows: Collect the meteorological condition data of each monitoring point at the current time t, including the key factors affecting the ice thickness, such as temperature, humidity, and wind speed, and combine these data into a new matrix X t , and for the input data X t Perform standardization to eliminate dimensional differences between different features; Using the determined parameters in the optimized SVM model and The ice thickness at the next time step t+1 of each monitoring point is predicted using the following formula: in represents the predicted ice thickness at the i-th monitoring point at time t+1, X i (t) is the normalized eigenvector of point i at time t, and sign represents the sign function; A weighted average strategy is introduced to integrate the prediction results. Based on historical data and expert knowledge, a weight α is assigned to each monitoring point to ensure that α i >0 and Σα i =1, Among them I i represents the historical ice coverage impact index of the i-th point; The predicted values ​​of all monitoring points are integrated using the weighted average method to form the overall regional predicted value Y tl t+1 , the formula is as follows: Where N is the total number of monitoring points and α is the weight of the i-th point.

8. A support vector machine GA-SVM ice-covered overhead line prediction system, based on the support vector machine GA-SVM ice-covered overhead line prediction method according to any one of claims 1 to 7, characterized in that: include: Historical data preprocessing module, time series data conversion and feature extraction module, spatiotemporal pattern feature construction module, SVM model optimization and establishment module, and prediction integration and evaluation module; The historical data preprocessing module is used to collect historical ice thickness data of the line, improve data quality through smoothing and denoising technology, build a clear time series, and provide an accurate basis for subsequent analysis; The time series data conversion and feature extraction module is used to convert the processed time series data into a multivariate time series data matrix, perform singular value decomposition, identify key modal features, form a spatiotemporal pattern feature representation through linear combination, further extract the statistical features of each time point, and prepare efficient input feature vectors for the SVM model; The spatiotemporal pattern feature construction module is used to identify and extract the main spatiotemporal patterns reflecting the dynamic changes of ice cover based on singular value analysis, construct a highly generalized feature representation, and enhance the model's ability to learn complex ice cover patterns; The SVM model optimization and establishment module is used to use the support vector machine theory and the genetic algorithm to globally optimize the SVM model parameters, ensuring that the model can perform efficient parameter tuning for specific ice cover prediction tasks while maintaining high generalization capabilities, thus forming an optimized GA-SVM classifier; The prediction integration and evaluation module is used to use the optimized GA-SVM model to predict the ice thickness of all monitoring points in future time steps. By integrating all predicted values ​​and using accumulation or other statistical methods, the overall ice thickness prediction of the entire monitoring area is comprehensively obtained, providing a scientific basis for line maintenance and decision-making.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the ice-covered overhead line prediction method using the support vector machine GA-SVM described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the ice-covered overhead line prediction method using a support vector machine GA-SVM according to any one of claims 1 to 7 are implemented.

Citation Information

Cited By

  • GNSS coordinate time sequence prediction method for optimizing SVM based on genetic algorithm

    CN119939258A

  • Perioperative period hypothermia prediction system based on artificial intelligence

    CN119993506A

  • Method, device and equipment for predicting sedimentation of dust remover, storage medium and program product

    CN120910459A