A method and system for predicting the thickness of ice on a power transmission line based on

Abstract

This invention proposes a method and system for predicting icing thickness on transmission lines, relating to the field of transmission line safety monitoring technology. The method includes: acquiring multi-dimensional meteorological data and corresponding icing thickness data of a target transmission line; preprocessing the multi-dimensional meteorological data; calculating the grey correlation value between the preprocessed meteorological data and the icing thickness data; when the grey correlation value is greater than a preset threshold, the current meteorological data is taken as strongly correlated meteorological data for icing thickness; constructing a lag feature vector between the current time and the previous L historical times based on the strongly correlated meteorological data; inputting the lag feature vector as an input vector into a hybrid kernel extreme learning machine model; optimizing the hyperparameter vector of the hybrid kernel extreme learning machine model using an improved Harris Eagle optimization algorithm; training the hybrid kernel extreme learning machine model using the optimized hyperparameters; and outputting the icing thickness prediction result using the trained model.

Description

Technical Field

[0001] This invention relates to the field of power transmission line safety monitoring technology, and in particular to a method and system for predicting the icing thickness of power transmission lines based on [the relevant technology]. Background Technology

[0002] Icing on transmission lines is one of the major natural disasters faced by power systems under severe weather conditions such as cold, high humidity, and mountainous areas. Continued icing increases the mechanical load on conductors, insulators, hardware, and towers, potentially leading to accidents such as conductor galloping, flashover, tripping, line breaks, and even tower collapse. Accurate and timely prediction of icing thickness is crucial for icing early warning, inspection and dispatching, load transfer, and de-icing decisions.

[0003] Existing icing prediction methods are mainly divided into two categories: mechanism-driven models and data-driven models. Mechanism-driven models are based on physical processes such as heat balance, mass transport, and liquid water content, and are interpretable. However, the required microscopic meteorological variables are difficult to obtain in actual monitoring systems, limiting their engineering applications. Furthermore, the complex micro-topography and micro-meteorological conditions in mountainous areas reduce the generalizability of simplified physical assumptions. Data-driven models, such as support vector regression, extreme learning machine, backpropagation neural network, and LSTM, can learn nonlinear mapping relationships from historical monitoring data, but they still have some problems in icing scenarios with small samples, strong nonlinearity, and significant time-cumulative effects. On the one hand, most models only use meteorological variables at the current moment as input, ignoring the time-cumulative effects of icing growth. On the other hand, the relationship between icing thickness and meteorological variables is highly nonlinear, and the expressive power of a single kernel function is limited.

[0004] Therefore, there is an urgent need to propose a method and system for predicting icing thickness that can model time-series cumulative effects, has strong nonlinear expression capabilities, and is robust to hyperparameter optimization, so as to monitor, analyze, and predict the icing situation of power lines. Summary of the Invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art, and proposes a method and system for predicting the icing thickness of transmission lines.

[0006] In a first aspect, embodiments of the present invention provide a method for predicting the icing thickness of transmission lines, comprising:

[0007] Acquire multi-dimensional meteorological data and corresponding icing thickness data of the target transmission line, and preprocess the multi-dimensional meteorological data;

[0008] Calculate the gray correlation value between the preprocessed meteorological data and the ice thickness data. When the gray correlation value is greater than a preset threshold, the current meteorological data is regarded as meteorological data strongly correlated with ice thickness.

[0009] Based on the strongly correlated meteorological data, a lagged feature vector is constructed between the current time and the previous L historical times. The lagged feature vector is then used as an input vector to input the hybrid kernel extreme learning machine model.

[0010] An improved Harris Eagle optimization algorithm is used to optimize the hyperparameter vector of the hybrid kernel extreme learning machine model;

[0011] The hybrid kernel extreme learning machine model is trained using optimized hyperparameters. The trained model is then used to predict the icing thickness at a predetermined future time, and the prediction results are output.

[0012] Furthermore, the meteorological data includes at least temperature, relative humidity, wind speed, rainfall, and air pressure. Preprocessing of the multi-dimensional meteorological data specifically includes: initially identifying outliers based on the Laida criterion; combining real-time meteorological data trends with historical meteorological data for the same period to determine whether extreme values ​​in various types of meteorological data are true extreme values; if the extreme values ​​are true extreme values, they are retained; if the extreme values ​​are false positives, they are removed; normalization processing is applied to various types of meteorological data after removing extreme values; and an initial dataset is constructed from the standardized meteorological data.

[0013] Furthermore, after obtaining the initial dataset, the initial dataset will be divided into a training set, a test set, and a validation set according to a preset ratio. The specific division method includes: stratifying various types of meteorological data according to the order of months, and dividing various types of meteorological data into training sets and test sets according to a certain ratio; the proportion of various types of meteorological data in different months or different seasons in the training set is the same as the proportion in the test set; the proportion of various types of meteorological data corresponding to different actual icing thicknesses in the training set is the same as the proportion in the test set.

[0014] Furthermore, the grey relational degree value between the preprocessed meteorological data and the icing thickness data is calculated. The specific method includes: using the preprocessed meteorological data as a reference sequence and the meteorological data as a comparison sequence, calculating the time-by-time absolute difference between the reference sequence and the comparison sequence; substituting the preset resolution coefficient to calculate the grey relational coefficient; and taking the mean value of all time-series relational coefficients of the single variable to obtain the grey relational degree.

[0015] Furthermore, the hybrid kernel extreme learning machine model is composed of a weighted combination of Gaussian kernel functions and polynomial kernel functions, including an input layer, a kernel mapping layer, and an output layer. The input layer receives multi-dimensional meteorological feature vectors that have been filtered and constructed by grey relational analysis and lag. The kernel mapping layer uses a hybrid kernel function that is a weighted combination of Gaussian kernels and polynomial kernels, where the Gaussian kernel is used to capture local nonlinear similarity and the polynomial kernel is used to characterize the global icing trend. The input is mapped to a high-dimensional feature space and a kernel matrix is ​​output. The output layer combines the kernel matrix, regularization coefficient, and target vector to calculate the predicted value, thereby achieving short-term multi-step prediction of icing thickness.

[0016] Furthermore, the hybrid kernel function is defined as follows:

[0017]

[0018] In the formula: For Gaussian kernel function, For polynomial kernel functions, For kernel-mixed weighting coefficients;

[0019] The Gaussian kernel function is defined as:

[0020]

[0021] In the formula: The width of the Gaussian kernel;

[0022] The polynomial kernel function is defined as:

[0023]

[0024] In the formula: The scaling factor is the polynomial scaling factor. For offset items; Let be the degree of the polynomial.

[0025] Furthermore, an improved Harris Eagle optimization algorithm is used to optimize the hyperparameter vector of the hybrid kernel extreme learning machine model. Specific methods include:

[0026] Perform initialization operations: set the size of the Harris Eagle population, the maximum number of algorithm iterations, and the upper and lower bounds of the hyperparameters to be optimized;

[0027] Initial fitness calculation and optimal solution determination: Calculate the fitness value of each individual in the population, using the mean squared error on the validation set as the fitness function; based on the calculation results, select and determine the optimal solution for the current population;

[0028] Termination condition judgment: Determine whether the current iteration number has reached the preset maximum iteration number: If the maximum iteration number has been reached, directly output the optimal hyperparameters obtained by optimization and the algorithm process ends; if the maximum iteration number has not been reached, enter the optimization iteration stage of this round.

[0029] Escape energy calculation and optimization phase division: Escape energy E is calculated using a nonlinear escape energy scheduling formula. Escape energy E is used for dynamic switching and the global exploration and local development capabilities of the balancing algorithm. The corresponding optimization phase is entered based on the absolute value of the escape energy |E|: when |E|≥1, the exploration phase begins; when |E|<1, the development phase begins.

[0030] Boundary control: After the optimization phase is completed, boundary control is performed on the position of all individuals in the population to ensure that the hyperparameter values ​​always fall within the preset search range.

[0031] Iterative Updates and Loops: After boundary control is completed, the fitness values ​​of all individuals are recalculated, and the current optimal solution of the population is updated synchronously. After incrementing the iteration count by 1, the decision node for "whether the maximum number of iterations has been reached" is returned. The above optimization steps are repeated until the termination condition is met, at which point the optimal hyperparameters are output and the process ends.

[0032] Furthermore, upon entering the exploration phase, the individual positions of the eagle flock are updated using a golden sine strategy. This strategy incorporates the golden ratio, which expands the coverage of the global search and prevents the algorithm from getting trapped in local optima too early. Upon entering the development phase, the native development strategy of the standard Harris Eagle Optimization algorithm is used to update the individual positions. A Gaussian random walk perturbation is applied to the current global optimum, and a greedy acceptance criterion is adopted. The perturbation result is only retained when the fitness value of the solution after perturbation is lower, further helping the algorithm to escape local optima.

[0033] Secondly, embodiments of the present invention provide a transmission line icing thickness prediction system, comprising: a data acquisition and preprocessing module, an icing thickness strongly correlated meteorological data acquisition module, a hysteresis feature vector construction module, a model optimization module, and a thickness prediction module; wherein:

[0034] The data acquisition and preprocessing module is used to acquire multi-dimensional meteorological data of the target transmission line and the corresponding icing thickness data, and to preprocess the multi-dimensional meteorological data.

[0035] The ice thickness strongly correlated meteorological data acquisition module is used to calculate the gray correlation value between the preprocessed meteorological data and the ice thickness data. When the gray correlation value is greater than a preset threshold, the current meteorological data is used as the ice thickness strongly correlated meteorological data.

[0036] The lag feature vector construction module is used to construct lag feature vectors for the current time and the previous L historical times based on the strongly correlated meteorological data, and to input the lag feature vectors as input vectors into the hybrid kernel extreme learning machine model.

[0037] The model optimization module is used to optimize the hyperparameter vectors of the hybrid kernel extreme learning machine model using an improved Harris Eagle optimization algorithm.

[0038] The thickness prediction module is used to train the hybrid kernel extreme learning machine model with optimized hyperparameters, and then use the trained model to predict the icing thickness at a preset time in the future, and output the prediction results.

[0039] Thirdly, an electronic device, characterized in that it comprises:

[0040] One or more processors;

[0041] Memory, used to store one or more programs;

[0042] When the one or more programs are executed by the one or more processors, the one or more processors implement the prediction method.

[0043] This invention provides a method and system for predicting icing thickness on power transmission lines. The method includes: acquiring multi-dimensional meteorological data and corresponding icing thickness data of a target power transmission line; preprocessing the multi-dimensional meteorological data; calculating the grey correlation value between the preprocessed meteorological data and the icing thickness data; when the grey correlation value is greater than a preset threshold, using the current meteorological data as strongly correlated meteorological data for icing thickness; constructing a lag feature vector between the current time and the previous L historical times based on the strongly correlated meteorological data; and inputting the lag feature vector as an input vector into a hybrid kernel extreme learning machine model.

[0044] An improved Harris Eagle optimization algorithm is used to optimize the hyperparameter vector of the hybrid kernel extreme learning machine model;

[0045] The hybrid kernel extreme learning machine model is trained using optimized hyperparameters. The trained model is then used to predict the ice thickness at a preset future time, and the prediction results are output. This invention integrates golden sine global search, nonlinear escape energy, and Gaussian random walk perturbation to solve the premature convergence problem of standard optimization algorithms. The hybrid kernel synchronously fits the local fluctuations and global trends of ice thickness. The lag time-series features restore the physical laws of ice accumulation and growth. In small-sample ice monitoring scenarios, the prediction accuracy and generalization ability are significantly better than existing models such as LSTM, standard HHO-HKELM, and SVR. It can support ultra-short-term and short-term ice accretion early warning and ice melting scheduling decisions for power grids.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] 1. Strong ability to model time-series cumulative effects: By introducing lagged meteorological features, the model can reflect the actual physical process of ice thickness being affected by recent meteorological history, making up for the shortcomings of existing models that only use current meteorological variables.

[0048] 2. Superior nonlinear fitting capability: The hybrid kernel function combines the complementary advantages of the Gaussian kernel (local nonlinear fitting) and the polynomial kernel (global trend fitting), making it suitable for complex nonlinear changes in the ice growth and melting process, and has higher fitting accuracy compared to a single kernel function.

[0049] 3. High robustness of hyperparameter optimization: The improved Harris Eagle algorithm integrates golden sine exploration, nonlinear escape energy scheduling and Gaussian random walk perturbation, which significantly improves global search capability and convergence stability, and effectively avoids the premature convergence problem of standard HHO under small sample conditions.

[0050] 4. Strong adaptability to small sample scenarios: By reducing the input dimension through gray relational feature filtering, reducing trainable parameters through hybrid kernel extreme learning machine, and improving the optimizer to improve search efficiency, the method of this invention can still maintain high prediction accuracy and stability in engineering scenarios with only a small number of complete icing events.

[0051] 5. Supports multi-timescale prediction: By setting different prediction step size h, this method can output icing thickness prediction results at different time scales such as 30 minutes, 1 hour, 2 hours, and 3 hours, meeting different scheduling and early warning needs. Attached Figure Description

[0052] Figure 1 A flowchart illustrating an ice thickness prediction method provided in an embodiment of the present invention;

[0053] Figure 2 This is a structural diagram of the hybrid kernel extreme learning machine model provided in an embodiment of the present invention;

[0054] Figure 3 A flowchart of the improved Harris Eagle optimization algorithm provided in this embodiment of the invention;

[0055] Figure 4 This is a comparison chart of optimization results for different models provided in the embodiments of the present invention;

[0056] Figure 5 This is a comparison chart of measured and predicted icing thickness at a 30-minute prediction step size provided in an embodiment of the present invention.

[0057] Figure 6 A comparison chart of prediction errors of various models under a 30-minute prediction step size provided in this embodiment of the invention;

[0058] Figure 7 This is a schematic diagram of the structure of an ice thickness prediction device provided in an embodiment of the present invention;

[0059] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0060] To enable those skilled in the art to better understand the technical solutions of the present invention, exemplary embodiments of the present invention are described below in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0061] Where there is no conflict, the various embodiments of the present invention and the features thereof may be combined with each other.

[0062] As used herein, the term “and / or” includes any and all combinations of one or more related enumerated entries.

[0063] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used herein, the singular forms “a” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will also be understood that when the terms “comprising” and / or “made of” are used in this specification, the presence of the stated feature, integral, step, operation, element, and / or component is specified, but the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or groups thereof is not excluded. Terms such as “connected” or “linked” are not limited to physical or mechanical connections but can include electrical connections, whether direct or indirect.

[0064] Unless otherwise specified, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art. It will also be understood that terms such as those defined in commonly used dictionaries should be interpreted as having the meaning consistent with their meaning in the context of the relevant art and the invention, and will not be interpreted as having an idealized or overly formal meaning unless expressly so defined herein.

[0065] In the technical solution of this invention, the collection, storage, use, processing, transmission, provision, and disclosure of user personal information all comply with relevant laws and regulations and do not violate public order and good morals. The use of user data in this technical solution follows relevant national laws and regulations (e.g., the "Information Security Technology - Personal Information Security Specification"). For example: appropriate measures are taken for personal information access control; restrictions are imposed on the display of personal information; the purpose of using personal information does not exceed the scope of direct or reasonable association; and explicit identity targeting is eliminated when using personal information to avoid precisely locating a specific individual.

[0066] To address at least one of the technical problems existing in the aforementioned related technologies, the present invention provides a method and system for predicting the icing thickness of transmission lines based on [the relevant technology].

[0067] This implementation discloses a method for predicting the icing thickness of transmission lines based on [the relevant technology / method]. Figure 1 ,include:

[0068] S100. Obtain multi-dimensional meteorological data and corresponding icing thickness data of the target transmission line, and preprocess the multi-dimensional meteorological data; in this embodiment, the meteorological data includes at least temperature, relative humidity, wind speed, rainfall, and air pressure; the preprocessing of the multi-dimensional meteorological data specifically includes: initially identifying outliers based on the Laida criterion, combining real-time meteorological data trends with historical meteorological data for the same period, determining whether extreme values ​​in various types of meteorological data are true extreme values, retaining them if they are true extreme values, and removing them if they are false values; normalizing the various types of meteorological data after removing extreme values; constructing an initial dataset from the standardized meteorological data.

[0069] For example, we take a complete icing event (referred to as Event 1) collected by an online icing monitoring system for a 500kV transmission line in Southwest China as the application object. This monitoring system records icing thickness and meteorological data every 30 minutes, collecting a total of 120 continuous samples, covering a total duration of 60 hours. In this icing event, the minimum icing thickness was 8.63 mm, the maximum was 23.84 mm, the mean was 16.92 mm, and the standard deviation was 4.08 mm; the temperature range was -6.37℃ to 1.82℃, with a mean of -2.41℃; the relative humidity range was 82.15% to 99.86%, with a mean of 94.28%; the wind speed range was 0.31 m / s to 7.42 m / s, with a mean of 2.76 m / s; and the rainfall range was 0.00 mm / h to 2.68 mm / h, with a mean of 0.43 mm / h.

[0070] In this embodiment, after obtaining the initial dataset, it is divided into a training set, a test set, and a validation set according to a preset ratio. The specific division method includes: stratifying various types of meteorological data according to monthly order; dividing the meteorological data into training and test sets according to a certain ratio; ensuring that the proportion of meteorological data from different months or seasons in the training set is the same as its proportion in the test set; and ensuring that the proportion of meteorological data corresponding to different actual ice thicknesses in the training set is the same as its proportion in the test set. This ensures that the distribution ratio of samples from different seasons and ice thickness ranges is consistent across the three datasets, avoiding overfitting of the model due to seasonal or ice thickness distribution shifts.

[0071] Specifically, in the data preprocessing stage, all normalization parameters are calculated solely from the training set. The min-max normalization method is used to standardize each variable; the normalization formula is as follows:

[0072]

[0073] In the formula: max(X) i ) represents the maximum value in the training set; min(X) i X is the minimum value in the training set; i (k) represents the data to be normalized.

[0074] S200. Calculate the gray correlation value between the preprocessed meteorological data and the ice thickness data. When the gray correlation value is greater than a preset threshold, the current meteorological data is taken as meteorological data strongly correlated with ice thickness.

[0075] In this embodiment, the gray correlation degree value of the preprocessed meteorological data and the icing thickness data is calculated. The specific method includes: taking the preprocessed meteorological data as a reference sequence and the meteorological data as a comparison sequence, calculating the time-by-time absolute difference between the reference sequence and the comparison sequence; substituting the preset resolution coefficient to calculate the gray correlation coefficient; and taking the mean of all time-series correlation coefficients of the single variable to obtain the gray correlation degree.

[0076] Specifically, to screen out meteorological variables strongly correlated with icing thickness, the grey relational degree between each variable and icing thickness is calculated on the training set. First, the normalized reference sequence Y′(k) (icing thickness) and the comparison sequence X are calculated. i The absolute difference between ′(k) (each meteorological variable):

[0077]

[0078] Then calculate the grey relational coefficient:

[0079]

[0080] In the formula: ; ; The distinguishing factor.

[0081] Finally, the grey relational degree of each variable is calculated:

[0082]

[0083] The larger Ri is, the stronger the association between the candidate variable and ice thickness.

[0084] For example, taking event 1 as an example, the calculated grey relational degrees are as follows: temperature 0.786, relative humidity 0.735, wind speed 0.716, rainfall 0.635, and air pressure 0.454. Based on the calculation results, the four variables with higher grey relational degrees—temperature, relative humidity, wind speed, and rainfall—are selected as the input meteorological variables for the prediction model, while air pressure is removed due to its lower relational degree (0.454).

[0085] S300. Based on the strongly correlated meteorological data, construct the lag feature vectors of the current time and the previous L historical times, and input the lag feature vectors as input vectors into the Hybrid Kernel Extreme Learning Machine (HKELM) model; preferably, when constructing the lag feature vectors, the lag order L is 3, and the current time, the previous 1, and the previous 2 sampling times are fused into a total of 3 sets of meteorological data to form 4×3=12-dimensional input features; the prediction step size h includes four types of ultra-short-term / short-term prediction scenarios: h=1 (30min), h=2 (1h), h=4 (2h), and h=6 (3h).

[0086] Specifically, considering that icing growth has a typical time-cumulative effect—that is, the icing thickness at the current moment depends not only on the current meteorological conditions but also on the meteorological conditions at several past moments—this embodiment constructs a lag input vector. The lag order is set to L=3, meaning that the meteorological information from the current moment and the previous two sampling moments (a total of 1.5 hours) are combined into one input vector. Specifically, the input vector at time k is defined as:

[0087]

[0088] Each input vector has a dimension of 4×3=12. The prediction target is the icing thickness yk+h after h sampling intervals in the future. This embodiment considers four prediction step sizes: h=1 (30 minutes), h=2 (1 hour), h=4 (2 hours), and h=6 (3 hours), corresponding to the icing thickness prediction requirements for ultra-short term, short term, and slightly longer term, respectively.

[0089] The 120 samples from Event 1 were divided chronologically: the first 80 samples (samples 1-80) were used as the training set, the middle 20 samples (samples 81-100) as the validation set, and the last 20 samples (samples 101-120) as the test set. The validation set was used for fitness evaluation during the hyperparameter optimization process, while the test set was only used for the final model performance evaluation and did not participate in any parameter learning or hyperparameter selection during the entire training and optimization process to ensure the objectivity of the evaluation.

[0090] In this implementation, the hybrid kernel extreme learning machine model is composed of a weighted combination of Gaussian kernel functions and polynomial kernel functions, including an input layer, a kernel mapping layer, and an output layer. The input layer receives multi-dimensional meteorological feature vectors that have been filtered and constructed by grey relational analysis and lag. The kernel mapping layer uses a hybrid kernel function that is a weighted combination of Gaussian kernels and polynomial kernels, where the Gaussian kernel is used to capture local nonlinear similarity and the polynomial kernel is used to characterize the global icing trend. The input is mapped to a high-dimensional feature space and a kernel matrix is ​​output. The output layer combines the kernel matrix, regularization coefficient, and target vector to calculate the predicted value, thereby achieving short-term multi-step prediction of icing thickness.

[0091] The hybrid kernel extreme learning machine model structure diagram provided in this embodiment of the invention is as follows: Figure 2 As shown, the hybrid kernel extreme learning machine model consists of three parts: an input layer, a kernel mapping layer, and an output layer. The input layer receives a 12-dimensional meteorological feature vector, filtered by grey relational analysis and constructed using lag. The kernel mapping layer maps the input vector to a high-dimensional feature space through a hybrid kernel function, which is a weighted combination of a Gaussian kernel and a polynomial kernel: the Gaussian kernel captures local nonlinear similarity, and the polynomial kernel characterizes the global icing trend; the two are fused using kernel weight coefficients λ. The kernel mapping layer outputs a kernel matrix K(Z,Z). The output layer calculates the predicted value based on the kernel matrix, the regularization coefficient C, and the target vector Y. This enables short-term, multi-step prediction of icing thickness.

[0092] The prediction model used in this embodiment is a Hybrid Kernel Extreme Learning Machine (HKELM). For the training set... The output function of HKELM is:

[0093]

[0094] In the formula: h(z) represents the hidden layer feature map, and β is the output weight vector.

[0095] KELM uses kernel matrices instead of explicit hidden layer mappings. The kernel matrix is ​​defined as follows:

[0096]

[0097] The new version of z's HKELM output calculation is as follows:

[0098]

[0099]

[0100] In the formula: C is the regularization parameter; I is the identity matrix; The target vector.

[0101] The hybrid kernel function is defined as follows:

[0102]

[0103] In the formula: For Gaussian kernel function, For polynomial kernel functions, For kernel mixed weighting coefficients, .

[0104] The Gaussian kernel function is defined as:

[0105]

[0106] In the formula: is the width of the Gaussian kernel.

[0107] The polynomial kernel function is:

[0108]

[0109] In the formula: The scaling factor is the polynomial scaling factor. For offset items; Let be the degree of the polynomial.

[0110] S400. The improved Harris Eagle optimization algorithm is used to optimize the hyperparameter vector of the hybrid kernel extreme learning machine model;

[0111] In this embodiment, the improved Harris Eagle Optimization (IHHO) algorithm is used to optimize the hyperparameter vector of the hybrid kernel extreme learning machine model. The specific method includes:

[0112] Perform initialization operations: set the size of the Harris Eagle population, the maximum number of algorithm iterations, and the upper and lower bounds of the hyperparameters to be optimized;

[0113] Initial fitness calculation and optimal solution determination: Calculate the fitness value of each individual in the population, using the mean squared error on the validation set as the fitness function; based on the calculation results, select and determine the optimal solution for the current population;

[0114] Termination condition judgment: Determine whether the current iteration number has reached the preset maximum iteration number: If the maximum iteration number has been reached, directly output the optimal hyperparameters obtained by optimization and the algorithm process ends; if the maximum iteration number has not been reached, enter the optimization iteration stage of this round.

[0115] Escape Energy Calculation and Optimization Phase Division: The escape energy E is calculated using a nonlinear escape energy scheduling formula. E is used for dynamic switching and balancing the global exploration and local exploitation capabilities of the algorithm. The algorithm enters the corresponding optimization phase based on the absolute value of the escape energy |E|: when |E|≥1, it enters the exploration phase; when |E|<1, it enters the development phase. Upon entering the exploration phase, the individual positions of the eagle flock are updated using a golden sine strategy. This strategy incorporates the golden ratio, expanding the coverage of the global search and preventing the algorithm from prematurely falling into local optima. Upon entering the development phase, the native development strategy of the standard Harris Eagle Optimization algorithm is used to update the individual positions. A Gaussian random walk perturbation is applied to the current global optimum, and a greedy acceptance criterion is used. The perturbation result is only retained if the fitness value of the solution after the perturbation is lower, further helping the algorithm escape local optima.

[0116] Boundary control: After the optimization phase is completed, boundary control is performed on the position of all individuals in the population to ensure that the hyperparameter values ​​always fall within the preset search range.

[0117] Iterative Updates and Loops: After boundary control is completed, the fitness values ​​of all individuals are recalculated, and the current optimal solution of the population is updated synchronously. After incrementing the iteration count by 1, the decision node for "whether the maximum number of iterations has been reached" is returned. The above optimization steps are repeated until the termination condition is met, at which point the optimal hyperparameters are output and the process ends.

[0118] Specifically, Figure 3 A flowchart of the improved Harris Eagle optimization algorithm provided in an embodiment of the present invention. Figure 3As shown, the algorithm first performs initialization, setting the Harris Eagle population size, maximum number of iterations, and upper and lower bounds for each hyperparameter. After initialization, the algorithm calculates the fitness value of each individual, using the mean squared error on the validation set as the fitness function, and determines the optimal solution in the current population. Then, the algorithm checks if the current iteration count has reached the maximum number of iterations. If it has, it directly outputs the optimal hyperparameters and terminates the process; otherwise, it continues with subsequent optimization steps. The algorithm first calculates the escape energy using a nonlinear escape energy scheduling formula. This escape energy is used for the global exploration and local development capabilities of the dynamic balancing algorithm. Next, the algorithm determines which stage to enter based on the absolute value of the escape energy: when the absolute value of the escape energy is greater than or equal to one, the algorithm enters the exploration stage, using a golden sine strategy to update the eagle flock position. This strategy incorporates the golden ratio to expand the global search coverage and prevent the algorithm from getting trapped in local optima too early; when the absolute value of the escape energy is less than one, the algorithm enters the development stage, using the development strategy of the standard Harris Eagle optimization algorithm to update the eagle flock position. After the development phase, the algorithm applies a Gaussian random walk perturbation to the current optimal solution, accepting the perturbation only if the fitness value of the solution is lower after the perturbation, to help the algorithm escape local optima. Subsequently, the algorithm performs boundary control on the positions of all individuals, ensuring that the hyperparameter values ​​always remain within the preset search range. After boundary control, the algorithm recalculates the fitness value of each individual and updates the current optimal solution, incrementing the iteration count by one, and then returns to check if the maximum number of iterations has been reached. This process is repeated until the termination condition is met. Through this process, the algorithm can adaptively search for the optimal hyperparameter combination of the hybrid kernel extreme learning machine.

[0119] To verify the method disclosed in this embodiment, Figure 4 This is a comparison chart of optimization results for different models provided in embodiments of the present invention. Figure 4 As shown, the three benchmark functions are: F1 Sphere function, used to test the algorithm's development accuracy; F2 Rastrigin function, used to test the algorithm's exploration ability in multimodal problems; and F3 Shekel function, used to test the algorithm's ability to avoid local optima. The algorithms compared include Particle Swarm Optimization (PSO), Gray Wolf Optimization (GSO), the standard Harris Eagle Optimization (HAO), and the improved Harris Eagle Optimization (HAO) proposed in this invention. The parameter settings for each algorithm are kept consistent, with a population size of 30 and a maximum number of iterations of 300. Figure 4 In the diagram, the horizontal axis represents the number of iterations, and the vertical axis represents the fitness value. A logarithmic coordinate system is used to facilitate observation of differences during the convergence process.

[0120] from Figure 4It can be seen that, on the F1 Sphere function, the particle swarm optimization algorithm has the slowest convergence speed but the highest final fitness value. The gray wolf optimization algorithm and the standard Harris Eagle optimization algorithm both show significant improvement, but the improved Harris Eagle optimization algorithm proposed in this invention converges the fastest and has the lowest final fitness value. On the F2 Rastrigin function, both particle swarm optimization and gray wolf optimization exhibit premature convergence to varying degrees. The standard Harris Eagle optimization algorithm shows some improvement but still suffers from local fluctuations, while the algorithm of this invention maintains a stable downward trend throughout the iteration process and obtains the lowest fitness value. On the F3 Shekel function, all algorithms converge, but the algorithm of this invention outperforms the other three comparative algorithms in terms of convergence accuracy and convergence stability. These results demonstrate that the improved Harris Eagle optimization algorithm proposed in this invention significantly outperforms particle swarm optimization, gray wolf optimization, and the standard Harris Eagle optimization algorithm in terms of convergence speed, convergence accuracy, and ability to avoid local optima, verifying the effectiveness of the introduced golden sine exploration strategy, nonlinear escape energy scheduling mechanism, and Gaussian random walk perturbation strategy.

[0121] In this patent, fixing and To reduce the search dimensionality, while C, , λ is then optimized using IHHO. Therefore, the hyperparameter vector is defined as:

[0122]

[0123] To improve the hyperparameter optimization effect, this embodiment uses the improved Harris Eagle Optimization Algorithm (IHHO) to automatically optimize the above four hyperparameters. The IHHO algorithm introduces three improvement strategies based on the standard HHO algorithm.

[0124] The first improvement is the introduction of a golden sine wave strategy during the exploration phase to enhance global search coverage. The position update formula during the exploration phase is:

[0125]

[0126] In the formula: r1 and r2 are random numbers; ; It is the golden ratio; and This represents the upper and lower bounds of the search space.

[0127] Boundary control is performed after each position update:

[0128]

[0129] The second improvement is the adoption of a nonlinear escape energy scheduling mechanism, leveraging the exploration and development capabilities of dynamic equilibrium algorithms. The formula for calculating the escape energy E is:

[0130]

[0131] In the formula: t is the current iteration number; Tmax is the maximum iteration number; rand∈(0,1) is a random number; α and β are constants, usually α=1.3 and β=1.7.

[0132] Compared to the linear decreasing strategy of the standard HHO, the nonlinear strategy of this invention decreases more slowly in the early stage and more quickly in the later stage, thereby better maintaining the global exploration capability in the early stage and the local fine search capability in the later stage.

[0133] The third improvement involves applying a Gaussian random walk perturbation to the current optimal solution during the development phase to help the algorithm escape local optima. The perturbation formula is:

[0134]

[0135] In the formula: Xbest is the current optimal solution; η is the perturbation coefficient; N(0,1) is a standard Gaussian random number; and ⊙ represents element-wise multiplication.

[0136] In this embodiment, the population size of the IHHO algorithm is set to 30, and the maximum number of iterations is 300. The search ranges of each hyperparameter are set as follows: regularization coefficient C∈[10⁻³, 10³], Gaussian kernel width σ∈[10⁻³, 10³], polynomial kernel scaling coefficient γ∈[10⁻³, 10³], and kernel weight λ∈[0,1]. The fitness value of each candidate solution is calculated by the mean squared error (MSE) on the validation set.

[0137]

[0138] S500. The hybrid kernel extreme learning machine model is trained using optimized hyperparameters. The trained model is then used to predict the icing thickness at a preset future time, and the prediction results are output.

[0139] To verify the method disclosed in this embodiment, Figure 4 This is a comparison chart of optimization results for different models provided in embodiments of the present invention. Figure 4As shown, the three benchmark functions are: F1 Sphere function, used to test the algorithm's development accuracy; F2 Rastrigin function, used to test the algorithm's exploration ability in multimodal problems; and F3 Shekel function, used to test the algorithm's ability to avoid local optima. The algorithms compared include Particle Swarm Optimization (PSO), Gray Wolf Optimization (GSO), the standard Harris Eagle Optimization (HAO), and the improved Harris Eagle Optimization (HAO) proposed in this invention. The parameter settings for each algorithm are kept consistent, with a population size of 30 and a maximum number of iterations of 300. Figure 4 In the diagram, the horizontal axis represents the number of iterations, and the vertical axis represents the fitness value. A logarithmic coordinate system is used to facilitate observation of differences during the convergence process.

[0140] from Figure 4 It can be seen that, on the F1 Sphere function, the particle swarm optimization algorithm has the slowest convergence speed but the highest final fitness value. The gray wolf optimization algorithm and the standard Harris Eagle optimization algorithm both show significant improvement, but the improved Harris Eagle optimization algorithm proposed in this invention converges the fastest and has the lowest final fitness value. On the F2 Rastrigin function, both particle swarm optimization and gray wolf optimization exhibit premature convergence to varying degrees. The standard Harris Eagle optimization algorithm shows some improvement but still suffers from local fluctuations, while the algorithm of this invention maintains a stable downward trend throughout the iteration process and obtains the lowest fitness value. On the F3 Shekel function, all algorithms converge, but the algorithm of this invention outperforms the other three comparative algorithms in terms of convergence accuracy and convergence stability. These results demonstrate that the improved Harris Eagle optimization algorithm proposed in this invention significantly outperforms particle swarm optimization, gray wolf optimization, and the standard Harris Eagle optimization algorithm in terms of convergence speed, convergence accuracy, and ability to avoid local optima, verifying the effectiveness of the introduced golden sine exploration strategy, nonlinear escape energy scheduling mechanism, and Gaussian random walk perturbation strategy.

[0141] In this patent, fixing and To reduce the search dimensionality, while C, , λ is then optimized using IHHO. Therefore, the hyperparameter vector is defined as:

[0142]

[0143] To improve the hyperparameter optimization effect, this embodiment uses the improved Harris Eagle Optimization Algorithm (IHHO) to automatically optimize the above four hyperparameters. The IHHO algorithm introduces three improvement strategies based on the standard HHO algorithm.

[0144] The first improvement is the introduction of a golden sine wave strategy during the exploration phase to enhance global search coverage. The position update formula during the exploration phase is:

[0145]

[0146] In the formula: r1 and r2 are random numbers; ; It is the golden ratio; and This represents the upper and lower bounds of the search space.

[0147] Boundary control is performed after each position update:

[0148]

[0149] The second improvement is the adoption of a nonlinear escape energy scheduling mechanism, leveraging the exploration and development capabilities of dynamic equilibrium algorithms. The formula for calculating the escape energy E is:

[0150]

[0151] In the formula: t is the current iteration number; Tmax is the maximum iteration number; rand∈(0,1) is a random number; α and β are constants, usually α=1.3 and β=1.7.

[0152] Compared to the linear decreasing strategy of the standard HHO, the nonlinear strategy of this invention decreases more slowly in the early stage and more quickly in the later stage, thereby better maintaining the global exploration capability in the early stage and the local fine search capability in the later stage.

[0153] The third improvement involves applying a Gaussian random walk perturbation to the current optimal solution during the development phase to help the algorithm escape local optima. The perturbation formula is:

[0154]

[0155] In the formula: Xbest is the current optimal solution; η is the perturbation coefficient; N(0,1) is a standard Gaussian random number; and ⊙ represents element-wise multiplication.

[0156] In this embodiment, the population size of the IHHO algorithm is set to 30, and the maximum number of iterations is 300. The search ranges of each hyperparameter are set as follows: regularization coefficient C∈[10⁻³, 10³], Gaussian kernel width σ∈[10⁻³, 10³], polynomial kernel scaling coefficient γ∈[10⁻³, 10³], and kernel weight λ∈[0,1]. The fitness value of each candidate solution is calculated by the mean squared error (MSE) on the validation set.

[0157]

[0158] After the S500 step optimization, the optimal hyperparameter combination is obtained as follows: C=42.67, σ=0.38, γ=1.24, λ=0.61. Using these optimal hyperparameters, the HKELM model is retrained on the merged training and validation sets (100 samples in total), and then predictions are made on the test set (20 samples) to obtain the predicted icing thickness values ​​at four prediction steps.

[0159] Figure 5 This is a comparison chart of measured and predicted icing thickness at a 30-minute prediction step size, provided in an embodiment of the present invention. Figure 5 As shown in the figure, the horizontal axis represents the test set sample points, and the vertical axis represents the ice thickness in millimeters. The figure displays the measured ice thickness curve and the prediction curves of four models: persistent prediction, Long Short-Term Memory (LSTM) network, standard Harris Eagle optimized hybrid kernel extreme learning machine, and the improved Harris Eagle optimized hybrid kernel extreme learning machine proposed in this invention. The figure shows that the persistent prediction curve exhibits a significant time lag effect during periods of rapid ice thickness change, lagging behind the measured curve. The LTM network can capture the overall trend of ice thickness change, but it produces large deviations at local fluctuations, reflecting its instability under small sample conditions. The standard Harris Eagle optimized hybrid kernel extreme learning machine shows improvement compared to the other two, with a significantly improved fit between the predicted and measured curves. The improved Harris Eagle optimized hybrid kernel extreme learning machine proposed in this invention performs best among all the comparison models, with its prediction curve being closest to the measured curve throughout the entire time period, especially maintaining good tracking ability during rapid increases in ice thickness and peak regions.

[0160] Figure 6 This is a comparison chart of prediction errors for various models under a 30-minute prediction step size provided in an embodiment of the present invention. Figure 6As shown in the figure, the horizontal axis represents the test set sample points, and the vertical axis represents the absolute error, in millimeters. The figure displays the absolute prediction error curves for four models: persistent prediction, Long Short-Term Memory (LSTM) network, standard Harris Eagle optimized hybrid kernel extreme learning machine, and the improved Harris Eagle optimized hybrid kernel extreme learning machine proposed in this invention. It can be seen from the figure that the persistent prediction error curve has the largest fluctuation range and exhibits high peak errors at multiple sample points. The LTM network error curve shows overall improvement, but still exhibits significant fluctuations and several high peaks. The standard Harris Eagle optimized hybrid kernel extreme learning machine error curve further decreases, with a significantly narrowed error fluctuation range. The error curve of the method proposed in this invention is the lowest overall and has the smallest fluctuation among all the comparative models. The absolute error at each sample point remains within a small range, and no significant peak error appears. This result fully demonstrates that the method proposed in this invention has higher prediction accuracy and better prediction stability. The lower peak error has important engineering significance for transmission line icing early warning and de-icing decision-making, because large local errors may lead to misjudgments of the de-icing timing.

[0161] To verify the effectiveness of this invention, under the same data partitioning and prediction tasks, several comparative models were selected for performance comparison, including: Persistent Prediction (Persist), Support Vector Regression (SVR), Backpropagation (BP) Neural Network, Long Short-Term Memory (LSTM) Network, Single-Core Extreme Learning Machine (KELM), Unoptimized Hybrid Core Extreme Learning Machine (HKELM), and Hybrid Core Extreme Learning Machine optimized with the standard Harris Eagle (HHO-HKELM). All comparative models were trained and evaluated on the same training, validation, and test sets.

[0162] The predictive performance is evaluated using three metrics: Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE). The calculation formulas are as follows:

[0163] Mean absolute error:

[0164]

[0165] Root mean square error:

[0166]

[0167] Mean absolute percentage error:

[0168]

[0169] Test results show that the IHHO-HKELM method of this invention achieves optimal prediction accuracy across all prediction step sizes. For 30-minute predictions (h=1), the RMSE of this method is 0.534 mm, MAE is 0.484 mm, MAPE is 2.87%, and the maximum absolute error is 0.86 mm. In comparison, the RMSE for persistent predictions is 0.724 mm, LSTM is 1.052 mm, and the standard HHO-HKELM is 0.585 mm. For 1-hour predictions (h=2), the RMSE of this method is 0.628 mm, MAE is 0.568 mm, MAPE is 3.34%, and the maximum absolute error is 1.03 mm. For 2-hour predictions (h=4), the RMSE is 0.815 mm; for 3-hour predictions (h=6), the RMSE is 1.019 mm.

[0170] Compared to the standard HHO-HKELM, the method of this invention reduces the average RMSE by 7.64% across the four prediction steps. Compared to persistent prediction, the RMSE is reduced by 27.50%; compared to LSTM, the RMSE is reduced by 45.33%. These results demonstrate that the method of this invention has significant accuracy advantages and better robustness in ultra-short-term and short-term icing thickness prediction tasks.

[0171] To verify the contributions of each improved component, ablation experiments were conducted. Grey relational analysis (GRA), lag features, hybrid kernel, and the IHHO optimizer were removed sequentially, and the impact of each component on the final performance was observed. The results show that the RMSE of the complete IHHO-HKELM model is 0.534 mm, while the RMSE after removing the hybrid kernel (i.e., using a single Gaussian kernel) is 0.654 mm, the RMSE after removing lag features is 0.662 mm, and the RMSE after removing grey relational analysis is 0.692 mm. This indicates that each component contributes positively to the final performance, with the hybrid kernel and lag features having the most significant impact.

[0172] To verify the generalization ability of the method of this invention, additional validation was performed on two separate icing events (Event 2 with 50 samples and Event 3 with 57 samples). On these two events, only the HHO-HKELM and IHHO-HKELM methods were compared. The results show that in the 30-minute prediction of Event 2, the RMSE of the method of this invention was 0.385 mm, better than HHO-HKELM's 0.421 mm; in the 1-hour prediction of Event 2, the RMSE was 0.486 mm, better than 0.533 mm; and in both prediction steps of Event 3, the method of this invention also maintained lower errors. The average RMSE reduction of the four additional validation cases was 8.59%, indicating that the method of this invention has good stability and generalization ability on different icing events.

[0173] In summary, this embodiment fully demonstrates that the proposed method for short-term multi-step prediction of transmission line icing thickness based on an improved Harris Eagle optimized hybrid kernel extreme learning machine is superior to existing technologies in terms of time-series cumulative effect modeling, nonlinear mapping capability, hyperparameter optimization robustness, and small-sample generalization capability. It can provide accurate and reliable decision support for transmission line icing early warning and ice melting scheduling.

[0174] This embodiment provides a method and system for predicting icing thickness on transmission lines. The method includes: acquiring multi-dimensional meteorological data and corresponding icing thickness data of a target transmission line; preprocessing the multi-dimensional meteorological data; calculating the grey correlation value between the preprocessed meteorological data and the icing thickness data; when the grey correlation value is greater than a preset threshold, the current meteorological data is taken as strongly correlated meteorological data for icing thickness; constructing a lag feature vector between the current time and the previous L historical times based on the strongly correlated meteorological data; inputting the lag feature vector as an input vector into a hybrid kernel extreme learning machine model; optimizing the hyperparameter vector of the hybrid kernel extreme learning machine model using an improved Harris Eagle optimization algorithm; training the hybrid kernel extreme learning machine model using the optimized hyperparameters; predicting the icing thickness at a preset future time using the trained model; and outputting the prediction result. This invention integrates golden sine global search, nonlinear escape energy, and Gaussian random walk perturbation to solve the premature convergence problem of standard optimization algorithms; it uses hybrid kernels to synchronously fit the local fluctuations and global trends of ice thickness; and it uses lag time-series features to restore the physical laws of ice accumulation and growth. In small-sample ice monitoring scenarios, its prediction accuracy and generalization ability are significantly better than existing models such as LSTM, standard HHO-HKELM, and SVR, and it can support ultra-short-term and short-term ice accretion early warning and ice melting scheduling decisions for power grids.

[0175] Based on the same inventive concept, embodiments of the present invention also provide a transmission line icing thickness prediction system, such as... Figure 7It includes: a data acquisition and preprocessing module, a meteorological data acquisition module strongly correlated with icing thickness, a hysteresis feature vector construction module, a model optimization module, and a thickness prediction module; among which:

[0176] The data acquisition and preprocessing module is used to acquire multi-dimensional meteorological data of the target transmission line and the corresponding icing thickness data, and to preprocess the multi-dimensional meteorological data.

[0177] The ice thickness strongly correlated meteorological data acquisition module is used to calculate the gray correlation value between the preprocessed meteorological data and the ice thickness data. When the gray correlation value is greater than a preset threshold, the current meteorological data is used as the ice thickness strongly correlated meteorological data.

[0178] The lag feature vector construction module is used to construct lag feature vectors for the current time and the previous L historical times based on the strongly correlated meteorological data, and to input the lag feature vectors as input vectors into the hybrid kernel extreme learning machine model.

[0179] The model optimization module is used to optimize the hyperparameter vectors of the hybrid kernel extreme learning machine model using an improved Harris Eagle optimization algorithm.

[0180] The thickness prediction module is used to train the hybrid kernel extreme learning machine model using optimized hyperparameters, and then use the trained model to predict the icing thickness at a preset time in the future, and output the prediction results.

[0181] The specific working methods of the data acquisition and preprocessing module, the meteorological data acquisition module with strong correlation to icing thickness, the hysteresis feature vector construction module, the model optimization module, and the thickness prediction module have been described in detail in the above-mentioned method for predicting the icing thickness of transmission lines, and will not be repeated here in this embodiment.

[0182] Based on the same inventive concept, embodiments of the present invention also provide an electronic device. Figure 8 This is a structural block diagram of an electronic device provided in an embodiment of the present invention. Figure 8 As shown, an embodiment of the present invention provides an electronic device including: one or more processors 101, a memory 102, and one or more I / O interfaces 103. The memory 102 stores one or more programs, which, when executed by the one or more processors, cause the one or more processors to implement any of the prediction methods described in the above embodiments; the one or more I / O interfaces 103 are connected between the processors and the memory, configured to enable information interaction between the processors and the memory.

[0183] The processor 101 is a device with data processing capabilities, including but not limited to a central processing unit (CPU); the memory 102 is a device with data storage capabilities, including but not limited to random access memory (RAM, more specifically SDRAM, DDR, etc.), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), and flash memory (FLASH); the I / O interface (read / write interface) 103 is connected between the processor 101 and the memory 102, and can realize information interaction between the processor 101 and the memory 102, including but not limited to a data bus (Bus).

[0184] In some embodiments, the processor 101, memory 102, and I / O interface 103 are interconnected via bus 104, and thus connected to other components of the computing device.

[0185] In some embodiments, the one or more processors 101 include a field-programmable gate array.

[0186] This invention also provides a computer-readable medium. The computer-readable medium stores a computer program, which, when executed by a processor, implements the steps of any of the prediction methods described in the above embodiments. The computer-readable storage medium may be volatile or non-volatile.

[0187] This invention also provides a computer program product, including computer-readable code, or a non-volatile computer-readable storage medium carrying computer-readable code, wherein when the computer-readable code is run in the processor of an electronic device, the processor in the electronic device executes the above-described prediction method.

[0188] Those skilled in the art will understand that all or some of the steps, systems, and apparatuses disclosed above, and their functional modules / units, can be implemented as software, firmware, hardware, or suitable combinations thereof. In hardware implementations, the division between functional modules / units mentioned above does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed collaboratively by several physical components. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit (ASIC). Such software can be distributed on a computer-readable storage medium, which may include computer storage media (or non-transitory media) and communication media (or transient media).

[0189] As is known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable program instructions, data structures, program modules, or other data. Computer storage media includes, but is not limited to, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), static random access memory (SRAM), flash memory or other memory technologies, portable compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, it is known to those skilled in the art that communication media typically contain computer-readable program instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0190] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0191] The computer program instructions used to perform the operations of this invention may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing state information from the computer-readable program instructions. This electronic circuitry can execute the computer-readable program instructions to implement various aspects of the invention.

[0192] The computer program product described herein can be implemented specifically through hardware, software, or a combination thereof. In one alternative embodiment, the computer program product is specifically embodied in a computer storage medium; in another alternative embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0193] Various aspects of the present invention are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.

[0194] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processor of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner; thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0195] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.

[0196] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction, which contains one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than those shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0197] Example embodiments have been disclosed herein, and while specific terminology has been used, it is for illustrative purposes only and should be construed as such, and is not intended to be limiting. In some instances, it will be apparent to those skilled in the art that features, characteristics, and / or elements described in conjunction with particular embodiments may be used alone, or in combination with features, characteristics, and / or elements described in conjunction with other embodiments, unless otherwise expressly indicated. Therefore, those skilled in the art will understand that various changes in form and detail may be made without departing from the scope of the invention as set forth in the appended claims.

Claims

1. A method for predicting icing thickness on transmission lines based on [previous method name], characterized in that, include: Acquire multi-dimensional meteorological data and corresponding icing thickness data of the target transmission line, and preprocess the multi-dimensional meteorological data; Calculate the gray correlation value between the preprocessed meteorological data and the ice thickness data. When the gray correlation value is greater than a preset threshold, the current meteorological data is regarded as meteorological data strongly correlated with ice thickness. Based on the strongly correlated meteorological data, a lagged feature vector is constructed between the current time and the previous L historical times. The lagged feature vector is then used as an input vector to input the hybrid kernel extreme learning machine model. An improved Harris Eagle optimization algorithm is used to optimize the hyperparameter vector of the hybrid kernel extreme learning machine model; The hybrid kernel extreme learning machine model is trained using optimized hyperparameters. The trained model is then used to predict the icing thickness at a predetermined future time, and the prediction results are output.

2. The prediction method according to claim 1, characterized in that, The meteorological data includes at least temperature, relative humidity, wind speed, rainfall, and air pressure. Preprocessing of the multi-dimensional meteorological data specifically includes: initially identifying outliers based on the Laida criterion; combining real-time meteorological data trends with historical meteorological data for the same period to determine whether extreme values ​​in various types of meteorological data are true extreme values; if extreme values ​​are true extreme values, they are retained; if extreme values ​​are false positives, they are removed; normalization processing is applied to various types of meteorological data after removing extreme values; and an initial dataset is constructed from the standardized meteorological data.

3. The prediction method according to claim 2, characterized in that, After obtaining the initial dataset, the initial dataset will be divided into a training set, a test set, and a validation set according to a preset ratio. The specific division method includes: stratifying various types of meteorological data according to the order of months, and dividing various types of meteorological data into training sets and test sets according to a certain ratio; the proportion of various types of meteorological data in the training set is the same as the proportion in the test set for different months or different seasons; the proportion of various types of meteorological data corresponding to different actual icing thicknesses in the training set is the same as the proportion in the test set.

4. The prediction method according to claim 1, characterized in that, The grey relational degree between preprocessed meteorological data and icing thickness data is calculated using the following methods: using preprocessed meteorological data as a reference sequence and meteorological data as a comparison sequence, calculating the time-by-time absolute difference between the reference sequence and the comparison sequence; substituting the preset resolution coefficients to calculate the grey relational coefficients; and taking the mean of all time-series relational coefficients of the single variable to obtain the grey relational degree.

5. The prediction method according to claim 1, characterized in that, The hybrid kernel extreme learning machine model is composed of a weighted combination of Gaussian kernel functions and polynomial kernel functions, including an input layer, a kernel mapping layer, and an output layer. The input layer receives multi-dimensional meteorological feature vectors that have been filtered and constructed by grey relational analysis and lag. The kernel mapping layer uses a hybrid kernel function that is a weighted combination of Gaussian kernels and polynomial kernels, where the Gaussian kernel is used to capture local nonlinear similarity and the polynomial kernel is used to characterize the global icing trend. The input is mapped to a high-dimensional feature space and a kernel matrix is ​​output. The output layer combines the kernel matrix, regularization coefficient, and target vector to calculate the predicted value, thereby achieving short-term multi-step prediction of icing thickness.

6. The prediction method according to claim 5, characterized in that, The hybrid kernel function is defined as follows: In the formula: For Gaussian kernel function, For polynomial kernel functions, For kernel-mixed weighting coefficients; The Gaussian kernel function is defined as: In the formula: The width of the Gaussian kernel; The polynomial kernel function is defined as: In the formula: The scaling factor is the polynomial scaling factor. For offset items; Let be the degree of the polynomial.

7. The prediction method according to claim 1, characterized in that, An improved Harris Eagle optimization algorithm is used to optimize the hyperparameter vector of a hybrid kernel extreme learning machine model. Specific methods include: Perform initialization operations: set the size of the Harris Eagle population, the maximum number of algorithm iterations, and the upper and lower bounds of the hyperparameters to be optimized; Initial fitness calculation and optimal solution determination: Calculate the fitness value of each individual in the population, using the mean squared error on the validation set as the fitness function; based on the calculation results, select and determine the optimal solution for the current population; Termination condition judgment: Determine whether the current iteration number has reached the preset maximum iteration number: If the maximum iteration number has been reached, directly output the optimal hyperparameters obtained by optimization and the algorithm process ends; if the maximum iteration number has not been reached, enter the optimization iteration stage of this round. Escape energy calculation and optimization phase division: Escape energy E is calculated using a nonlinear escape energy scheduling formula. Escape energy E is used for dynamic switching, global exploration, and local development capabilities of the balancing algorithm. The corresponding optimization phase is entered based on the absolute value of escape energy |E|: when |E|≥1, the exploration phase is entered; when |E|<1, the development phase is entered. Boundary control: After the optimization phase is completed, boundary control is performed on the position of all individuals in the population to ensure that the hyperparameter values ​​always fall within the preset search range. Iterative Updates and Loops: After boundary control is completed, the fitness values ​​of all individuals are recalculated, and the current optimal solution of the population is updated synchronously. After incrementing the iteration count by 1, the decision node for "whether the maximum number of iterations has been reached" is returned. The above optimization steps are repeated until the termination condition is met, at which point the optimal hyperparameters are output and the process ends.

8. The prediction method according to claim 7, characterized in that, Once in the exploration phase, the individual positions of the eagle flock are updated using the golden sine strategy. This strategy incorporates the golden ratio, which expands the coverage of the global search and prevents the algorithm from getting trapped in local optima too early. Once in the development phase, the native development strategy of the standard Harris Eagle Optimization algorithm is used to update the individual positions. A Gaussian random walk perturbation is applied to the current global optimum, and a greedy acceptance criterion is adopted. The perturbation result is only retained if the fitness value of the solution after the perturbation is lower, which further helps the algorithm escape local optima.

9. A transmission line icing thickness prediction system based on [the aforementioned method], employing the prediction method described in any one of claims 1-8, characterized in that, include: The module includes a data acquisition and preprocessing module, a meteorological data acquisition module strongly correlated with icing thickness, a hysteresis feature vector construction module, a model optimization module, and a thickness prediction module; among which: The data acquisition and preprocessing module is used to acquire multi-dimensional meteorological data of the target transmission line and the corresponding icing thickness data, and to preprocess the multi-dimensional meteorological data. The ice thickness strongly correlated meteorological data acquisition module is used to calculate the gray correlation value between the preprocessed meteorological data and the ice thickness data. When the gray correlation value is greater than a preset threshold, the current meteorological data is used as the ice thickness strongly correlated meteorological data. The lag feature vector construction module is used to construct lag feature vectors for the current time and the previous L historical times based on the strongly correlated meteorological data, and to input the lag feature vectors as input vectors into the hybrid kernel extreme learning machine model. The model optimization module is used to optimize the hyperparameter vectors of the hybrid kernel extreme learning machine model using an improved Harris Eagle optimization algorithm. The thickness prediction module is used to train the hybrid kernel extreme learning machine model with optimized hyperparameters, and then use the trained model to predict the icing thickness at a preset time in the future, and output the prediction results.

10. An electronic device, characterized in that, include: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the prediction method as described in any one of claims 1 to 8.

Images